SC134-deruxtecan, a fucosyl-GM1 targeting ADC for small cell lung cancer therapy

Abstract

Background

Small cell lung cancer (SCLC) remains a difficult disease to treat with poor long-term survival rates. New therapies offer modest overall survival benefit beyond that of chemotherapy alone, necessitating the development of improved therapies. Fucosyl-GM1 (FucGM1) is a glycolipid highly expressed on SCLC cells, but virtually absent in normal tissues, suggesting strong potential for targeted therapy. We have developed SC134-deruxtecan, an antibody drug conjugate (ADC) targeting FucGM1 in SCLC, and characterized its preclinical activity.

Methods

SC134 binding specificity and affinity were tested through enzyme-linked immunosorbent assay (ELISA), surface plasmon resonance (SPR), flow cytometry against several SCLC cell lines, and immunohistochemistry (IHC) of clinical samples and animal tissues. In silico modelling supplemented the FucGM1 binding specificity. Internalization kinetics and colocalization of SC134 with the lysosomes were investigated through imaging flow cytometry. Direct cytotoxicity as well as bystander killing and antibody dependent cell cytotoxicity (ADCC) by SC134-deruxtecan were determined using in vitro cell cytotoxicity assays. SC134-deruxtecan’s efficacy was evaluated in vivo using a DMS79 xenograft model.

Results

SC134 specifically targets FucGM1, without GM1 cross-reactivity, with nanomolar affinity. In silico modelling of the SC134 FucGM1 binding site revealed a relatively narrow binding pocket, occupied by the terminal three glycans with multiple Fucose-engaging interactions. Robust FucGM1 expression in frozen SCLC patient tissues was evident, whilst tissue cross-reactivity analysis indicated non-human primates as well as mice as suitable tox models. FucGM1 binding by SC134 led to effective internalization, with a 6.9-h half-life, lysosomal colocalization, culminating in sub-nanomolar drug delivery efficiency, across a range of payloads. Covalent deruxtecan conjugation of SC134 with a DAR 8 and a cleavable linker showed effective (nanomolar) in vitro killing of SCLC cell lines such as DMS79 and DMS153, with concentration-dependent bystander killing of FucGM1-negative AGS cells. SCLC cell killing was further augmented through ADCC. Potent in vivo DMS79 xenograft killing was seen at 3mg/kg SC134-deruxtecan, which was well tolerated.

Conclusion

The tumour-specific nature of FucGM1, combined with the potent SCLC killing by SC134-deruxtecan underscore the development potential of SC134 for use as an ADC therapy against SCLC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06940-2.

Background

Lung cancer is the second most common cancer and is the leading cause of cancer death in both men and women [1]. Small cell lung cancer (SCLC), constituting up to 15% of all lung cancers, is an aggressive disease with poor prognosis. At the limited stage (LS-SCLC), surgery followed by chemo/radiotherapy, with or without immunotherapy, can be effective. The majority of patients, however, present at the disseminated (extensive) stage (ES-SCLC), where etoposide plus platinum-based chemotherapy with or without radiotherapy (standard of care, SOC) initially shows good impact, but patients soon progress. In recent years, the addition of immunotherapy (PD-L1 inhibitors) in the first line (1L) setting has shown further modest improvements (2–3 months overall survival (OS) improvement), but patients still progress whilst on maintenance immunotherapy [2–4]. A number of second line (2L) options are available such as Topotecan, Lubirnectin, or checkpoint inhibitors, but the 5-year survival rates remain extremely low, underscoring the need for additional treatments [5].

Glycosphingolipids are abundant in the outer leaflet of the plasma membrane, where they impact on membrane protein function, through microdomain formation or lateral association [6, 7]. Dramatic changes in membrane glycosphingolipid composition occur during normal development and malignant transformation [8, 9]. More recently, functional genomics screening revealed an essential role for membrane sphingolipids in enabling tumour growth in immunocompetent mice, highlighting their importance in immune evasion [10]. Fucosyl GM1 (FucGM1) is a ganglioside originally discovered as being highly expressed by SCLC samples [11, 12]. In these early reports, FucGM1 expression was shown to be specifically associated with both primary and metastasised samples from SCLC patients with extremely limited expression on normal tissues [13]. FucGM1 is a monosialoganglioside composed of a ceramide lipid tail which inserts into the membrane and a surface exposed oligosaccharide headgroup comprising Fucα1-2Galβ1-3GalNAcβ1-4(Neu5Acα2-3)Galβ1-4Glc (Fig. 1A) [14]. GM1, the precursor to FucGM1, is expressed in the nervous system where it plays a role in the maturation and maintenance of neuronal tissues [15]. The α(1,2)-fucosyltransferase genes, Fut1 and Fut2, are responsible for fucosylation of FucGM1 in a species and tissue-specific manner [16, 17]. Analysis of a panel of SCLC cell lines, however revealed frequent inactivating Fut2 mutations that did not impact on the FucGM1 surface expression, indicating that Fut1 is likely the key driver for FucGM1 expression in SCLC. Additionally, a role for Fut1 in maintaining pluripotency during embryonic development has also recently been established [18].

Fig. 1.

SC134 exhibits avid FucGM1 binding, facilitated by a narrow binding pocket A FucGM1 cartoon with sphingosine and hydrophobic lipid tail, the importance of the Fucose residue in binding is highlighted; B ELISA binding of FucGM1 (ceramide and BSA conjugate) to SC134, absence of GM1 cross reactivity; C Biacore analysis of real-time SC134 binding to FucGM1-BSA compared to GM1a-BSA); D In silico modelling of FucGM1 glycolipid binding to SC134

FucGM1 has been pursued as an immunotherapy target in a range of settings. Dickler et al. trialled vaccination using FuGM1-linked KLH in the adjuvant setting, with most patients mounting a FucGM1-specific serological response [19]. More recently, BMS-986012, optimised for effector function and showing encouraging preclinical efficacy, has progressed to PhII/III in combination with checkpoint inhibition [20, 21], supporting the notion that FucGM1 is an effective target for SCLC therapy and paving the way for other FucGM1-targeting modalities to be evaluated. We recently described a FucGM1-targeting T cell engaging bispecific, SC134-TCB, showing potent ex vivo and in vivo SCLC tumour cell killing [22]. Additionally, antibody drug conjugates (ADC) continue to show great promise in solid tumour settings. Here we present results on the FucGM1-targeting SC134-ADC, comprising the humanised SC134 antibody conjugated to the topoisomerase 1 inhibitor (TOP1i), deruxtecan, showing encouraging preclinical in vitro and in vivo efficacy. Importantly, we provide a structural model for selective FucGM1 binding by SC134, as well as FucGM1 target distribution in SCLC patient tissues.

Methods

Materials, cells, and antibodies

DMS79, DMS153, DMS53, H526, and AGS cells were obtained from American Type Culture Collection (ATCC) and maintained at low passage number. DMS79 and H526 cells were cultured in RPMI 1640 (Sigma), supplemented with 10% FBS (Sigma). AGS cells were cultured in Kaighn’s Modification of Ham’s F-12 Medium (F12-K) (ATCC), supplemented with 10% FBS. DMS153 and DMS53 cells were cultured in Waymouth’s medium (Gibco), supplemented with 10% FBS. Cell lines were tested monthly for the presence of mycoplasma and were maintained at low passage number for experiments.

Generation of SC134

Hybridoma 134 was created via BALB/c mice immunizations with a complex immunisation regimen [22]. The hybridoma-derived antibody, a murine (m) IgG1 with κ light chain, was subsequently chimerized (huIgG1), followed by humanisation using Composite Human Antibody™ Technology (Abzena) to generate SC134.

Target binding analysis via ELISA

ELISA plates were coated with 200 ng/well FucGM1 or GM1 ceramide (both Matreya LLC) or FucGM1-BSA (Elicityl) followed by blocking with 2% BSA-PBS. Primary antibody dilutions were detected using anti–human IgG (γ-chain specific)–biotin antibody (Sigma), followed by streptavidin–HRP conjugate.

Indirect immunofluorescence and flow cytometry

SC134 cell binding was analysed by flow cytometry. Briefly, 1 × 10⁵ SCLC cells were incubated with serially diluted SC134 for 1 h at 4 °C. Following washing, bound SC134 was detected using FITC-labelled anti-human IgG Fc (Sigma). Stained samples were washed and resuspended in IC Fixation buffer (Invitrogen) followed by analysis on a MACSQuant 16 flow cytometer, equipped with MACSQuant software version 2.13.3. EC50 values were determined via three parameter non-linear regression. The number of antibody binding sites was determined using QIFIKIT (Agilent), according to the manufacturer’s protocol. Specific antibody binding capacity (SABC) was deduced by interpolation onto the calibration curve using Graphpad Prism v10.2.3.

Kinetic internalization

Imaging flow cytometry was carried out to determine antibody internalization. Briefly, 100,000 DMS79 cells were incubated with Alexa Fluor 647 (AF647) labelled SC134 on ice for 1 h. Cells were washed with ice cold FACS medium, before being span down at 300 g for 5 min at 4 °C. Cells were cultured in their respective culture medium for 0, 1, 3, 6, 24, and 48 h. For the final 10 min of incubation Live/Dead yellow (Thermo Fisher) viability dye was added at a 1:50 dilution before being washed with ice cold FACS medium, span down at 300 g for 5 min, at 4 °C, and resuspended in IC fixation buffer. The samples were processed on an ImageStreamX MkII imaging flow cytometer and data was interpreted using IDEAS software, version 6.2. The software was used to give an internalization score for viable DMS79 cells which is determined as the ratio of internal fluorescence over total fluorescence, on a log scale. Positive internalization scores represent a population with a majority of internal fluorescence, negative scores represent a population with a majority of external (membrane-bound) fluorescence, and a score of 0 represents a population with equal levels of internal and external fluorescence. The half-life for internalization was calculated using the formulae ln(2)/K, where K is the rate constant.

Lysosomal colocalization

Imaging flow cytometry was used to determine the degree of colocalization between the lysosomal marker LAMP1 and internalized SC134-AF647. Briefly, 4 × 10⁶ DMS79 cells were incubated with Alexa Fluor 647 labelled SC134 on ice for 1 h. Following incubation, cells were washed with ice cold FACS medium. 500,000 labelled cells were added per well in a 24-well plate in complete medium and incubated for 0, 1, 3, 5.5, 24, and 48 h at 37 °C, 5% CO₂. For the final 10 min of incubation Live/Dead yellow (Thermo Fisher) viability dye was added at a 1:50 dilution. Following incubation, cells were washed with ice cold FACS and fixed using IC Fixation buffer for 20 min. Following fixation, cells were permeabilised using permeabilization buffer (Invitrogen) and labelled with anti-LAMP1-AF488 (Cell Signalling Technology). Following an incubation of 30 min, cells were washed and resuspended in IC Fixation buffer. The samples were processed on an ImageStreamX MkII imaging flow cytometer and data was interpreted using IDEAS software, version 6.2. The bright detail similarity feature was used to determine the degree of colocalization between AF647 and AF488 fluorescence, and Live/Dead yellow dye was used to exclude dead cells from the analysis.

Kinetic binding analysis Biacore T200

The kinetics parameters for SC134 binding to immobilised FucGM1-BSA, used as a surrogate for the FucGM1 glycolipid, was determined by Surface Plasmon resonance (SPR, Biacoare™ T200, Cytiva). Covalent immobilisation of FucGM1-BSA (immobilised density, 178.6RU) or GM1a-BSA (immobilised density, 132.9RU, both from Elicityl) was achieved using a CM5 chip, with an HSA-coated flow cell used as a reference cell. For kinetics analysis, SC134 binding to FucGM1-BSA or GM1a-BSA was monitored by flowing SC134 across the coated surfaces at increasing concentration (1.56–50 nmol/L) using the multi-cycle kinetics (MCK) option and regeneration with 10 mM Glycine pH 1.5. Data processing and analysis were performed using Biacore T200 evaluation software V 2.0.1. All sensorgrams from the analysis were double referenced.

In silico structural modelling (Phyre2.2 and HADDOCK)

Homology modelling (Phyre2.2) using PSI-blast was used to create a theoretical SC134 model [23]. The top model from the list (PDB 2GHW, crystal structure of SARS spike protein receptor binding domain in complex with a neutralizing antibody, 80R), was selected based on the percentage confidence and coverage. The coordinates for the 3D structural homology model were saved as a PDB, after deletion of hetero atoms and water molecules, followed by model validation using MolProbidity. The coordinate file for FucGM1 was obtained from RCSB Protein Data Bank (PDB 6HMY: Crystal structure of Cholera toxin classical B-pentamer in complex with FucGM1). PDB files for SC134 and FucGM1 were uploaded to the HADDOCK (HADDOCK 2.4) submission interface for molecular docking. Inputs for molecules 1 and 2 were SC134 model and Glycan (FucGM1) respectively. For SC134, all the CDRs were selected as active residues and for FucGM1 all residues except terminal beta-D-Glucopyranose (BGC1) were selected for the docking run. Docking parameters were left on default. From the output data, clusters were generated and within each cluster are 4 models. HADDOCK score, cluster size, z-score and other scoring are generated for each cluster. The top cluster usually considered most reliable were downloaded and analysed in PYMOL to assess SC134 and FucGM1 interactions. Final figures were created using PYMOL and Biorender.

Drug conjugation

Conjugation was carried out by Sterling Pharma Solutions. SC134 and Rituximab were fully reduced to generate eight free thiols for alkylation with Deruxtecan. Deruxtecan was linked via Cathepsin B-cleavable glycine-glycine-phenylalanine-glycine linker. Once manufactured, the ADC was purified by Sephadex G25 exchange into PBS, and residual toxin linker removed via contact with activated carbon. Average Drug to antibody ratio (DAR) was determined through reversed-phase high-performance liquid chromatography (RP-HPLC) (Supplemental Fig. 2).

Immunohistochemistry

Immunohistochemistry was carried out as previously described [22] using SC134 (10μg/ml), and Human IgG1 isotype (#31154, Thermofisher Scientific), combined with the human-on-human kit (HOH-3000, Vector Laboratories) according to the manufacturer’s protocol. Where the murine (m) 134 (10 μg/ml) was used, mouse IgG1 isotype (Abcam-ab18443) was included combined with horseradish peroxidase (HRP) labelled goat anti-mouse polymer (MHRP520L, Biocare Medical). Additional markers comprised: Thyroid transcription factor 1 (TTF-1) (Abcam-ab133737) and Delta like ligand-3 (DLL3) (Cell Signalling-71804T) ab with rabbit IgG Isotype (Abcam-ab172730) in combination with goat anti-rabbit polymer and CD56 (Agilent- M730429-2). Frozen tissue sections were air-dried and rehydrated with cold PBS followed by blocking of endogenous peroxidase activity and protein block. Primary antibodies were applied using antibody diluent mixture, followed by the relevant horseradish peroxidase (HRP) labelled polymer and DAB. Hematoxylin-modified Mayer’s solution was used for counterstaining. Slides were dehydrated, mounted, and scanned on NanoZoomer SQ. QuPath V.3 was used for scoring and H-scores were assigned ranging from 0 to 300, considering staining intensity and percentage of positive cells. Details for the human patient tissues and animal tissues are described in Supplemental Table 1 and 2, respectively.

In vitro cytotoxicity

A water-soluble tetrazolium salt assay (WST-8) (Abcam) or Cell Titer Glo assay (Promega) was used to determine cytotoxicity, following treatment with ADC, in vitro. Briefly, 2 × 10³ or 5 × 10³ target cells were cultured for 72 h with titrated ADC or titrated primary antibody plus ADC anti-Fc secondary antibody (Moradec, final concentration 0.2 ug/ml). After incubation Wst-8 reagent was added and cells were cultured for a further 3 h to allow colour to develop. Absorbance was then read at 450 nmol/L using a Tecan Infinate M Plex plate reader and cell survival was calculated compared to a medium only control well. If the Cell Titer Glo assay was used for calculation of cell survival, after 72 h incubation, Cell Titer Glo reagent was added 1:1 and placed on an orbital shaker for 2 min. The plate was further incubated for 10 min. Luminescence was then detected using a Tecan Infinite M Plex plate reader. Cell survival was calculated using the formula: A450/luminescence test wells ÷ A450/luminescence cell only control wells × 100. Linear regression was used to calculate EC₅₀ values (Graphpad Prism v10.2.3). For the bystander killing assay, 1 × 10³ target negative AGS cells were co-cultured with increasing ratios of DMS79 cells, in the presence of SC134-deruxtecan, or Rituximab-Deruxtecan. After 6 days of culture, DMS79 cells were stained with SC134-AF647 and the AGS cell count/ml determined.

In vitro antibody dependent cell cytotoxicity (ADCC)

Target DMS79 (20,000 cells) were co-cultured with or without effector PBMC at a ratio of 20:1 effector:target. PBMC were isolated from Leukopaks (BioIVT) through density gradient centrifugation. Ethics approval was obtained from the IRB (tracking number 20190318). All donors signed informed consent. The cells were cultured in the presence of SC134-deruxtecan or Rituximab-Deruxtecan, serially diluted to 100, 10, 1, 0.1, 0.01, and 0 nmol/L. After 12 h incubation, 50 µl of supernatant was taken for measurement of cell cytotoxicity via lactate dehydrogenase (LDH) release, using CytoTox 96 Non-radioactive cytotoxicity assay, according to the manufacturer’s recommendation (Promega). The percentage cytotoxicity of test wells with DMS79 and PBMC co-culture was determined with the formula: (experimental LDH release − effector spontaneous LDH release − target spontaneous LDH release)/(target maximum LDH release − target spontaneous LDH release) × 100. Percentage cytotoxicity of wells with DMS79 only were calculated using the formula: experimental LDH release/target maximum LDH release × 100.

In vivo anti-tumour model

In vivo work was carried out by Crown Biosciences under home office project licence number PP1144128. Animal welfare for the study complied with the UK Animals Scientific Procedures Act 1986 (ASPA). BALB/c nude mice (Charles River, n = 10/group, total number of 30 mice) were inoculated with 2 × 10⁶ DMS79 cells. Upon tumour growth to an average size of 100 mm³, mice were randomised into three groups and dosed intravenously with 3 mg/kg SC134-Deruxtecan, 3 mg/kg Rituximab-Deruxtecan, or vehicle (PBS, 10 ml/kg). Mice were dosed twice, two weeks apart, following the day of randomization. Body weight and tumour growth were measured biweekly until termination on day 39. A Two-way ANOVA Model with Geisser-Greenhouse correction (Graphpad Prism v10.2.3) was applied to tumour growth data to determine significant differences between groups. Last observation carried forward (LOCF) was used for the control group on day 37, as a result of loss of animals due to reaching the tumour volume endpoint.

Statistical analyses

All statistical analyses were calculated using Graphpad Prism, version 10.2.3. Significant P values were taken as P ≤ 0.05.

Results

SC134 exhibits avid FucGM1 binding, facilitated by a narrow binding pocket

We previously reported the creation of a FucGM1-targeting monoclonal antibody (mAb) obtained via a complex Balb/c immunisation regimen [22]. SC134, the humanised lead mAb, retained the original FucGM1 binding characteristics, binding FucGM1 glycolipid with a half maximal effective concentration (EC₅₀) of 0.08 nmol/L on ELISA (Fig. 1B). Additionally, SC134 also bound FucGM1-BSA (comprising the FucGM1 hexasaccharide headgroup grafted onto the BSA carrier) albeit with lower efficiency (EC₅₀, 0.2 nmol/L). Critically, no meaningful cross-reactivity towards GM1 was detected, in spite of using an amplified (biotin-streptavidin) detection setup, suggesting the terminal α1-2 Fucose is crucial for binding specificity. This was further corroborated by SPR analysis, where SC134 exhibited a concentration-dependent response to FucGM1-BSA, but critically, showed no cross-reactivity towards GM1a-BSA (Fig. 1C), confirming the ELISA results.

In order to further understand the structural nature of the SC134-FucGM1 binding specificity, and the role of Fucose, we performed in silico modelling of SC134 combined with docking of the FucGM1 hexasaccharide. SC134 engaged FucGM1 (supplemental Fig. 1A) through a narrow binding pocket, with the terminal three glycans (Fucose (Fuc5), Galactose (Gal4) and N-acetyl galactosamine (NGA3) driving the interaction (Fig. 1D and Supplemental Fig. 1B). FucGM1 binding to SC134 was largely mediated by residues in CDRH2, CDRH3, CDRL1 and CDRL3. Fuc5, residing in the centre of the binding pocket, made the majority of the interactions with SC134 through a combination of electrostatic interactions (PRO100, VAL101 and ARG102 in CDRH3) and hydrogen bonds with ASP151 and HIS153 (CDRL1). Additionally, GAL4 sitting on the rim of the binding pocket, interacted with CDRH3 GLU99 and formed H-bond with main chain GLY210. NGA3 provided support to both GAL4 and FUC5 by bridging between CDRH2 HIS53 and CDRL3 HIS211 with a bond distance of 3.1 Angstrom (Å) on either side, consequently forming H-bonds (Supplemental Fig. 1B). In contrast, galactose (GAL2) and sialic acid (SIA6), both being more solvent exposed, were further removed from the binding pocket, and BGC1 due to its membrane proximity was furthest removed. The model, whilst requiring experimental validation, indicates a crucial role for Fucose interactions stabilising the FucGM1 binding to SC134 thereby providing a potential explanation for the lack of GM1 cross-reactivity by SC134.

FucGM1 exhibits a superb differential tumour versus normal tissue distribution and is expressed in preclinical toxicology models

High-level target expression in tumour tissues combined with restricted normal tissue distribution are critical attributes for ADC development. Due to the lipid nature of FucGM1, its expression can only be evaluated on fresh or frozen tissues. Historic analyses have demonstrated FucGM1 expression in 58–90% of SCLC tumours, with a much smaller fraction of NSCLC tissues (squamous as well as adenocarcinoma] being positive [11–13, 24]. We previously reported more than 80% positive FucGM1 expression in frozen SCLC patient-derived xenograft tissues (PDX) [22]. Comparison of this FucGM1 expression profile with that of DLL3, an alternative clinically validated SCLC target, indicated increased FucGM1 expression compared to DLL3 in eleven out of thirteen tissues (Fig. 2A). We expanded on these PDX tissues by analysing commercially sourced frozen SCLC patient tissues (Supplemental Table 1). Robust FucGM1 target expression was evident in all five examined SCLC tissues, with H-scores up to 102 (Fig. 2B). DLL3 expression was again lower than FucGM1. Thyroid Transcription Factor 1 (TTF-1), a marker for primary lung tumours was widely expressed in all patient tissues, with the strongest expression coinciding with the stronger FucGM1 expression. Across a frozen NSCLC array comprising adenocarcinoma as well as squamous carcinoma tissues, FucGM1 expression was evident in five out of twenty squamous NSCLC, with four tissues displaying a H-score between 5 and 150. SC134 displayed an excellent normal tissue distribution with previously reported low-level expression in only three normal human tissues: skin, pituitary and thymus [22]. The absence of meaningful GM1 cross-reactivity was evident from the fact that neither frozen neuronal tissues such as cerebellum, cerebrum [22] nor full-face peripheral nerve exhibited staining for FucGM1 (Supplemental Fig. 3B).

Fig. 2.

FucGM1 exhibits a superb differential tumour versus normal tissue distribution and is expressed in preclinical toxicology models A Comparative H-scores for DLL3 expression by PDX tissues (Supplemental Fig. 3A) compared to FucGM1 from [22]; B Comparative H-scores and staining for FucGM1, DLL3 and TTF-1 in patient SCLC samples (patient details in Supplemental Table 1); C H-score distribution of FucGM1 in squamous NSCLC tissues (staining in Supplemental Fig. 3C); D FucGM1 positive mouse tissues: (A) breast, (B) colon, (C) skeletal muscle, (D) pancreas, (E) skin, (F) small intestine (G) stomach (H) testes, (I) thymus, (J) Pituitary (C57BL/6); E FucGM1 positive rat tissues (A) mesothelium, (B) stomach (foveolar cells), (C) stomach (crypt cells); F FucGM1 positive cynomolgous monkey tissues: (A) thymus, (B) skin and (C) small intestine

The FucGM1 glycolipid is conserved across all animal species, however species-specific normal tissue distribution is evident, largely driven by the tissue-specific expression of the biosynthetic enzymes [16, 25]. Normal mouse tissues, in contrast to humans, exhibited a wider FucGM1 expression, comprising ten FucGM1-positive tissues, three of which overlapped with human (skin, thymus and pituitary) (Fig. 2D). Strong FucGM1 expression was observed in stomach, with more moderate to weak expression in breast, colon, skeletal muscle, pancreas, skin, small intestine, testes, thymus, and pituitary. Rat showed a more restricted distribution with two positive tissues: mesothelium and stomach: foveolar and crypt cells, neither of which were seen in humans (Fig. 2E). Frozen cynomolgus monkey tissues mimicked the restricted FucGM1 distribution seen in humans, with thymus and skin exhibiting weak FucGM1 staining (Fig. 2F), but unlike the human tissues, cynomolgus small intestine was also weakly positive for FucGM1.

SC134 targets SCLC cell lines with nanomolar efficiency and displays efficient internalization

SC134 binding of SCLC cell surface occurred with nanomolar efficiency (Fig. 3A). DMS79, a high-FucGM1 expressing cell line, showed high binding efficiency (EC₅₀ 43 nmol/L), followed by DMS153 and DMS53 (EC₅₀ 171 nmol/L and 231 nmol/L, respectively), and somewhat lower efficiency on the low-expressing H526 (EC₅₀ 7.7 µmol/L). The binding histograms revealed a relatively broad distribution of SC134 binding on DMS53 and H526 (Fig. 3A). Quantification of the antibody binding capacity (ABC) (at saturation), using the murine version of SC134, revealed DMS79 to have approximately 10⁶ ABC units and DMS53 of the order of 10⁵ ABC. DMS153 consistently showed ABC in the same order of magnitude as DMS79. H526 is estimated to have a lower ABC based on the binding data in Fig. 3A.

Fig. 3.

SC134 targets SCLC cell lines with nanomolar efficiency and displays efficient internalization A Flow cytometry analysis of SC134 cell binding to SCLC cell lines. Histograms show binding of 50 nM SC134 (DSM79, DMS53, H526) or 500 nM SC134 (DMS153) (black) against control (gray); B SC134 internalization on DMS79, anti-transferrin receptor antibody was included as a positive control; i. representative cellular fluorescence of AF647-SC134 or AF647-anti-transferrin receptor over time, out of approximately 500–1500 gated events. Image number corresponds to the recorded event acquisition number. Brightness is increased for transferrin receptor labelling compared to SC134 to help visualization of antibody location. This does not affect internalization values. ii. internalization score of each antibody over time. iii. Percentage of AF647-positive cells with a positive internalization score, over time. C Co-localization of SC134-AF647 (red) and lysosomal LAMP1 (green) following incubation of DMS79 cells in the presence of SC134-AF647 for 0 h (i) and 24 h (ii)

A prerequisite for ADC functionality is cellular internalization and release of the cytotoxic drug via lysosomal proteases. Consequently, the SC134 internalization kinetics and lysosomal colocalization using DMS79 cells were investigated using AF647-conjugated SC134 and imaging flow cytometry (Fig. 3B, C). Over the 48 h incubation period, progressive internalization of SC134 was observed (Fig. 3Bi), with a calculated half-live of 6.9 h (Fig. 3Bii). The percentage of cells with internal fluorescence gradually increased over time to a maximum of 55% at 48 h (Fig. 3Biii). AF647-anti-transferrin receptor antibody showed faster internalization (internalization half-live of 3.4 h), with a positive internalization score from 1 h onwards, in line with its literature-reported fast internalization and lower surface expression levels [26, 27]. SC134 showed increasing colocalization with the lysosome (based on lysosomal associated membrane protein 1 (LAMP1) colocalization) over time (Fig. 3C, supplemental Fig. 4), with visible colocalization from 1 h (Supplemental Fig. 4), and reaching a maximum at 24 h, suggesting SC134 is successfully delivered to the lysosome following internalization.

Indirect ADC screens reveal potent target-dependent killing of SCLC tumour cells by SC134 across a range of DNA-targeting drugs

The effective internalization of SC134 prompted its evaluation for cytotoxic drug delivery. Intracellular drug delivery was evaluated, initially, using Fab-based drug-conjugated anti-human Fc constructs (Moradec LLC). This allowed for the screening of drugs with a range DNA-targeting actions such as the duocarmycin (DMDM), pyrolobenzodiazepin (PBD), anthracylin (PNU) and TOP1i, exatecan mesylate (DX8951). Over the 72 h incubation period, impact on cell survival was evident with PBD being the most efficient at killing, with an EC₅₀ of 16 pmol/L, followed by DMDM (EC₅₀ 49 pmol/L), DX8951 (EC₅₀ 0.15 nmol/L) and PNU (EC₅₀ 0.42 nmol/L) (Supplemental Fig. 5A–D). The window for target-dependent impact of PNU was much reduced compared to the other drugs (Supplemental Fig. 5D). The most complete cell reduction was seen with the TOP1i, DX8951 (Supplemental Fig. 5C), reducing cell survival to 25%, informing the decision to create SC134-deruxtecan through covalent conjugation.

SC134-deruxtecan shows potent in vitro SCLC cell killing with bystander and effector cell-mediated impact

SC134-deruxtecan, as well as a non-targeting control (Rituximab-deruxtecan) with peptidyl cleavable linker and an average DAR4 or DAR8 were created using interchain cysteine conjugation. SC134 binding to DMS79 was unaffected by conjugation to deruxtecan (Supplemental Fig. 5E). SC134-deruxtecan DAR8 showed somewhat improved killing efficiency (EC₅₀ 1.3 nmol/L) compared to DAR4 (EC₅₀ 5.6 nmol/L, Supplemental Fig. 5F), therefore this ADC was taken forward for further in vitro and in vivo studies. SC134-deruxtecan showed potent killing activity against the high FucGM1 expressing cell line DMS79 and DMS153 with nanomolar efficiency (EC₅₀ 1.7 and 14 nmol/L, respectively, Fig. 4A). Killing was also seen against the moderate expressing cell line DMS53 with EC₅₀ 47.8 nmol/L, but little activity was seen against the low expressing H526 (Fig. 4A). The target-dependent nature of the cell killing was evident from the lack of impact by the non-targeting Rituximab-deruxtecan as well as the negligible impact on the target-negative AGS (Supplemental Fig. 5G).

Fig. 4.

SC134-deruxtecan shows potent in vitro SCLC cell killing, with bystander and effector cell mediated impact A Cell survival assay, following 72 h culture with SC134 and non-targeting rituximab control ADC using DMS79, DMS153, DMS53 and H526. B Bystander assay showing AGS cell number following 6-day co-culture with increasing DMS79 ratios and ADC concentrations for 6 days. AGS cell number was determined through flow cytometry and negative gating with SC134-AF647. C ADCC activity measured as percentage cytotoxicity following co-culture of DMS79 target cells with PBMC in the presence of SC134-deruxtecan or rituximab-deruxtecan control. DMS79 alone with SC134-deruxtecan was used as a negative control for ADCC

Bystander killing was examined through co-culture of target-positive DMS79 and target-negative AGS cells with SC134-deruxtecan or non-targeting Rituximab-deruxtecan (Fig. 4B). AGS cell numbers decreased at 10 nmol/L SC134-deruxtecan, the efficiency of which correlating with increasing DMS79 cells numbers. Target-dependency of the bystander killing was confirmed by the absence of killing by Rituximab-deruxtecan. Further to direct and bystander killing, SC134-deruxtecan mediated target cell death via antibody dependent cell cytotoxicity (ADCC), through its intact Fc region, with nanomolar efficiency (EC₅₀ 0.7 nmol/L) and up to 40% cell death (Fig. 4C). Cell killing was only evident in the presence of PBMC, indicating this is ADCC-mediated and not due to direct killing by the ADC.

Excellent in vivo tumour control against DMS79 xenograft by SC134-deruxtecan

To assess in vivo efficacy, SC134-deruxtecan and the non-targeting Rituximab-deruxtecan control ADC were tested against DMS79 in BALB/c Nude mice. Mice were randomised at day 10 with an average tumour volume of 102 mm³ and treated with 3 mg/kg ADC, once every two weeks for four weeks (Fig. 5, supplemental Fig. 6). SC134-deruxtecan showed significant (p < 0.0001) target-dependent tumour control/eradication, with 80% of the mice being tumour-free at the end of study. The non-targeting control Rituximab-deruxtecan had a marginal impact on tumour growth. SC134-deruxtecan treatment was well tolerated and had no impact on mouse bodyweight (Fig. 5B).

Fig. 5.

Excellent in vivo tumour control against DMS79 xenograft by SC134-deruxtecan A Anti-tumour efficacy of SC134-deruxtecan against DMS79 xenograft. Mice (n = 10/group) were randomised on day 10 when tumour size was ~ 102 mm³. SC134-deruxtecan and rituximab-deruxtecan were dosed at 3 mg/kg on day 11 and 25. Symbols represent mean tumour volume + SEM. Mixed effects model with Geisser-Greenhouse correction used for statistical tests. ****p ≤ 0.0001. B Relative mean bodyweight over time, as a percentage of bodyweight at start of dosing (day 11), following dosage of SC134-deruxtecan and non-targeting Rituximab-deruxtecan control. Symbols represent mean bodyweight + SEM

Discussion

SC134 exhibited excellent FucGM1 binding specificity, without GM1 cross-reactivity. The latter attribute is critical as GM1 is abundantly expressed in neuronal tissues [28]. This suggests that the terminal α1-2 linked Fucose is essential for binding by SC134 and our in silico modelling unveils key interactions between FUC5 and residues within CDRH3 and CDRL1 in SC134, that require further experimental verification. These interactions are absent for GM1 (not containing the Fucose) and could provide a rational why SC134 did not cross-react with GM1, as seen on ELISA and SPR.

The analysis of FucGM1 distribution in SCLC patient samples is challenging due to the need for frozen tissues, combined with the fact that SCLC is rarely resected. Notwithstanding, IHC analysis on frozen PDX tissues [22], as well as five SCLC patient samples indicated robust expression of FucGM1, with occasional focal distribution, matching the literature-reported expression levels, as well as those from our model cell line DMS79, previously reported to have an H-score of between 97 and 147 [22]. A subset of patient tumours had lower H-scores, however, we anticipate SC134-deruxtecan to still maintain success against these tumours, as evidenced from in vitro activity against moderate expressing cell lines such as DMS53. Limited in vitro effect was demonstrated against the low expressing cell line, H526, however, this could potentially be increased by extracellular cleavage of the SC134-deruxtecan in the tumour microenvironment, as has been shown for Enhertu [29]. Conversely, the lack of impact on ultra-low expressing tissues may be advantageous in reducing potential off-tumour toxicity. On comparison with an alternative SCLC target, Delta-like ligand 3 (DLL3), FucGM1 expression levels were usually higher, in spite of methodological differences, as DLL3 expression via IHC is predominantly analysed on FFPE specimen. Additionally, IHC does not distinguish between cell surface-associated DLL3, compared to the abundant intracellular pool [30, 31]. In the PDX cohort, the majority of cases expressed both targets, in concordance with DLL3 being associated with tumours of neuroendocrine origin [32], but some only expressed FucGM1, suggesting FucGM1 may not fully overlap with DLL3 expression. FucGM1 expression was also evident in a subset (~ 20%) of squamous NSCLC, as has been reported in the literature [33]. A limitation of our analysis is the small number of patient samples, a direct result of the fact that SCLC is rarely resected and the fact that frozen tissues are required for FucGM1 IHC analysis. PDX material has been used to address this limitation [22].

In humans FucGM1 is virtually absent from normal tissues, with only very low-level expression in skin, pituitary and thymus [22]. Non-human primates match this restricted distribution with only three FucGM1 tissues, two of which overlapped with humans (skin and thymus). In contrast, rodents exhibited wider (mice) and differential (rats) FucGM1 distribution, in line with species-specific glycosyl transferase expression [16, 25]. In mice, however, three out of ten positive tissues overlapped with humans, suggesting it could still be a relevant preclinical rodent toxicity model.

Clinical validation of FucGM1 as a target for SCLC therapy stems from BMS-986012, a FucGM1 targeting antibody with improved effector function and currently in clinical development in combination with chemotherapy and checkpoint inhibitors, where it shows a modest improvement in OS and was well tolerated [21]. We previously validated the use of SC134 as a T cell redirecting bispecific for SCLC therapy [22]. This paper examined the merits of FucGM1 as a target for ADC development. SC134 exhibited effective internalization as demonstrated by antibody imaging and delivery of cytotoxic payloads to FucGM1-positive SCLC cells. Internalization kinetics demonstrated an internalization half-life of 6.9 h, which is in line with Trastuzumab internalization on cells expressing similar levels of target [27]. High target expressing cells show increased intracellular drug exposure combined with extracellular drug release (which could contribute to bystander killing), and are modelled to have faster degradation rates, compared to lower expressing cells, despite a slower internalization half-life [27]. A half-life of 6.9 h could be considered a slow rate of internalization in comparison to for instance the transferrin receptor, however, studies have shown that slower internalization rates allows for better tumour penetration and in vivo efficacy [34]. This could explain the excellent tumour control showed by SC134-deruxtecan in vivo. Other gangliosides, such as GM1 and GD3, have shown recycling to the membrane following antibody binding and internalization [35, 36]. The current study, however, demonstrated colocalization of SC134 with LAMP1, a lysosomal marker, suggesting that SC134 is delivered to the lysosomes for degradation, a prerequisite for ADC functionality.

Following internalization, SC134 delivered its payload intracellularly, either directly conjugated or indirectly through the use of drug-conjugated secondary constructs, thereby causing target cell death. This was further augmented by ADCC via the unmodified Fc region of SC134. Indirect conjugates based on DNA-targeting activity all exhibited impact on DMS79. The TOP1i DX8951, exatecan mesylate, exhibited nanomolar efficiency, consequently deruxtecan was selected for covalent conjugation using the cleavable peptidyl linker: glycine-glycine-phenylalanine-glycine, as used by Trastuzumab-deruxtecan (T-DXd) [37]. Furthermore, this linker reduces hydrophobicity, allowing for an increased DAR of 8 [38]. SC134-deruxtecan was highly efficacious against DMS79 cells, with nanomolar EC₅₀. Efficacy correlated with FucGM1 expression levels across a range of SCLC cell lines without target-independent impact. DMS79 was selected as the model for in vivo testing, as IHC analysis demonstrated it to be relevant to the clinical tissues as well as the PDX models [22]. Potent target-mediated DMS79 killing was also evident in vivo after two weekly 3 mg/kg dose, with no apparent toxicities, in spite of a number of FucGM1 positive mouse tissues. In vivo, solid tumours can often show an “edge effect” where antibodies bind at the tumour periphery or in the immediate vicinity of blood vessels and show reduced penetration due to target binding. The use of a less potent drug (e.g. deruxtecan) allows for higher doses to be used, to saturate binding sites at the tumour edge and allow improved penetration. Interestingly, in spite of FucGM1 expression in several normal mouse tissues, notably the gastrointestinal tract tissues such as stomach, small intestine and colon, SC134-deruxtecan was well tolerated with no impact on bodyweight, suggesting that the target may not be accessible from the circulation (e.g. the lumen of the stomach) or that FucGM1 expression levels and/or internalisation kinetics in other normal tissues is suboptimal for cytotoxicity. Testicular toxicity was avoided by using only female mice.

SC134-deruxtecan exhibited ADCC capability. Maintaining a fully functional Fc region may allow for improved efficacy via effector function contributions, as well as having a beneficial impact on ADC half-live [39], there is however, some evidence that this could also contribute to ADC-associated interstitial lung disease (ILD) [40, 41]. Recent work suggests that one way by which ILD could occur is through Fc-mediated ADC uptake by alveolar macrophages, independent of target expression, however the precise mechanisms have not been fully explored. Indeed, ILD is observed across a range of targets, drugs and linker designs, suggesting it is a multi-factorial process. To diminish the risk of ILD via macrophage uptake, SC134-deruxtecan could be Fc-ablated prior to moving into the clinic.

The current ADC landscape in SCLC comprises several targets, notably: DLL3, B7H3, SEZ6 and Trop2, the majority of which using TOP1i-based drugs [42–50], in spite of the earlier clinical failure of the DLL3-targeting Rovalpituzumab tesirine (Rova-T) [51]. FucGM1 has several advantages compared to the aforementioned targets including a very high tumour specificity with very few normal human tissues expressing FucGM1, conservation of the target across preclinical tox models and critically, high-level tumour expression and efficient internalization. Of note, M3554 targeting the glycosphingolipid GD2 is a TOP1i ADC in development for glioblastoma and sarcoma, further validating glycolipids as promising targets for targeted drug delivery [52].

Conclusion

In summary, SC134-deruxtecan with its high specificity for FucGM1, a clinically relevant SCLC target, and its proficient in vitro and in vivo killing of SCLC, is a compelling candidate for clinical advancement. The combination of its direct target-mediated impact as well as its potential for bystander and immune effector cell mediated cytotoxicity likely all contributed to its potent in vivo tumour eradication, underscoring its potential for therapeutic benefit.

Abbreviations

1L

First line

2L

Second line

ABC

Antibody binding capacity

ADC

Antibody drug conjugate

ADCC

Antibody dependent cell cytotoxicity

AF

AlexaFluor

B7H3

B7 Homolog 3

BGC1

Beta-D-Glucopyranose

BSA

Bovine serum albumin

CDR

Complementarity-determining region

DAR

Drug-antibody ratio

DLL3

Delta-like canonical Notch ligand 3

DMDM

Duocarmycin

DNA

Deoxyribonucleic acid

DX8951

Exatecan mesylate

EC₅₀

Half maximal effective concentration

ES-SCLC

Extensive stage small cell lung cancer

Fc region

Fragment crystallizable region

FBS

Fetal bovine serum

FFPE

Formalin fixed paraffin embedded

Fuc5

Fucose 5

FucGM1

Fucosyl GM1

GAL2

Galactose 2

GAL4

Galactose 4

H-bond

Hydrogen bond

HRP

Horseradish peroxidase

HSA

Human serum albumin

IHC

Immunohistochemistry

ILD

Interstitial lung disease

LDH

Lactate dehydrogenase

LS-SCLC

Limited stage small cell lung cancer

mAb

Monoclonal antibody

MCK

Multi-cycle kinetics

MOA

Mechinism of action

NGA3

N-acetyl galactosamine 3

NK

Natural killer

NSCLC

Non-small cell lung cancer

OS

Overall survival

PBD

Pyrolobenzodiazepin

PBMC

Peripheral blood mononuclear cells

PDL1

Programmed death ligand 1

PDX

Patient-derived xenograft

PNU

Anthracylin

Rova-T

Rovalpituzumab tesirine

RP-HPLC

Reversed-phase high-performance liquid chromatography

SCLC

Small cell lung cancer

SIA6

Sialic acid 6

SOC

Standard of care

SPR

Surface plasmon resonance

TTF-1

Thyroid transcription factor-1

SEZ6

Seizure protein 6

SG

Sacituzumab govitecan

TCB

T cell bispecific

T-DXd

Trastuzumab deruxtecan

TOP1i

Topoisomerase 1 inhibitor

Trop-2

Tumor-associated calcium signal transducer 2

Wst8

Water-soluble tetrazolium salt assay

Author contributions

Bryony Heath: Formal analysis, investigation, methodology, writing–original draft, Bubacarr G. Kaira: Formal analysis, investigation, methodology. Dhruma Thakker: Formal analysis, methodology. Omar J. Mohammed: Formal analysis, methodology. Ruhul Choudhury: Formal analysis, methodology. Foram Dave: Formal analysis, methodology. Poonam Vaghela: Formal analysis, methodology. Elena Dubinina: Formal analysis, methodology. Tina Parsons: Resources, validation. Lindy Durrant: Conceptualization, supervision, funding acquisition, validation, writing–review. Mireille Vankemmelbeke: Conceptualization, supervision, validation, writing–original draft, writing–review and editing.

Funding

Funding was provided by Scancell Holdings PLC.

Data availability

The data that support the findings of this study are available upon reasonable request to the corresponding author.

Declarations

Ethics approval and consent to participate

Animal studies complied with the UK Animals Scientific Procedures Act 1986 (ASPA) and were carried out in concordance with Crown Bioscience’s UK Guidelines and Standard Operating Procedures. Patient-derived tumor material was sourced from Crown Bioscience International and was acquired after approval by the IntegReview Ethical Review Board. Patient SCLC tissues were acquired through AccioBiobank, BioMedica CRO and Indivdumed GmbH, according to local institutional review board (IRB) approval and with Informed Consent (IC). Human leukocytes were obtained by BioIVT under IRB approved protocols. All donors signed IC forms.

Consent for publication

Not applicable.

Competing interests

LD has ownership interest in a patent. LD is a director and shareholder in Scancell Ltd. All other authors are employees of Scancell Ltd.