Selective targeting of cancer cells is a crucial requirement for the improvement of anti-tumor therapies to avoid toxicity in non-neoplastic cells [102]. It is becoming evident that certain altered glycans aiding tumor onset and progression can be used as selective targets for improved diagnostics and therapeutic strategies. Here, we present a novel bispecific lectibody that targets the glycosphingolipid Gb3 on Burkitt´s lymphoma-derived cell lines and engages T cells for a specific and powerful anti-tumor response. The format of a bispecific lectibody is inspired by the therapeutic class of BiTEs, which have proven efficient in redirecting immune cells, primarily T cells, towards target cells, thereby inducing anti-tumor activity. For example, the CD19 × CD3 canonical BiTE blinatumomab has achieved impressive efficacy in treating B cell malignancies [103]. Therefore, our study aimed to show that a bispecific lectibody with the lectin Stx1B as tumor-targeting domain and scFv OKT3 as a T cell engager could redirect T lymphocytes´ cytotoxicity towards Gb3-expressing lymphoma-derived cells.
The potential of rationally engineered lectins to produce lectibodies has been already investigated in several studies in the past few years [104]. Through the genetic fusion of a lectin and an antibody´s crystallizable fragment (Fc) of immunoglobulin G (IgG), the resulting lectibody molecules have shown potential as antiviral proteins. The recognition of carbohydrates on the envelope of viruses by the lectin domain, coupled to the antibody effector functions of the Fc fragment ˗ such as complement-dependent cytotoxicity (CDC), antibody-dependent cell-mediated cytotoxicity (ADCC) or antibody-dependent cell-mediated phagocytosis (ADCP) ˗ have led to successful neutralization of viral entry into host cells and clearance of cells infected by viruses [105]. Our new Stx1B-scFv OKT3 lectibody resembles the previously described lectibodies as it combines a lectin with an antibody fragment, nevertheless, the architecture is quite different. In lieu of the Fc fragment, we employed the scFv OKT3 as a T cell engager and selectively conjugated it to the Gb3-binding Stx1B lectin instead of genetically fusing it. This approach allowed us to select the attachment points for maximum functionality of both components and to join them with a predefined spacing using linker molecules. The lectin Stx1B was selected to successfully recognize Gb3-expressing tumors as previously described [18, 44–47]. Due to the invariant property of CD3 chains in the TCR [106], CD3 was here chosen as a T cell surface target for the lectibody. The monoclonal antibody OKT3 recognizes a region of the antigenic CD3ε on human T cells and induces an immune response by T cell activation and proliferation [107]. Full-length antibodies can be produced in E. coli [108–110]. However, the balanced expression of the heavy- and light chains constituting the large antibody molecule, which requires several correctly formed disulfide bonds for its functional folding, seriously challenges the prokaryotic host [111]. Moreover, antibodies produced in E. coli are not glycosylated, which can affect their stability [112]. For these reasons, E. coli has been used to produce antibody fragments such as Fabs and scFvs rather than full-length antibodies. In this study, we employed the OKT3 single chain variable fragment because it readily binds to the CD3 antigen, is stable and efficiently produced in E. coli [86, 113]. These properties render scFv OKT3 an excellent candidate as the T cell engager component of the lectibody. To crosslink Gb3-expressing tumor cells with T cells via CD3 of the TCR, the lectin Stx1B and the scFv OKT3 must be physically linked. The linkage can be affected at the gene level or by chemical ligation on the protein level. In general, genetic fusions open a narrow window for conjugating proteins either on the N-terminus or on the C-terminus. The approach leaves no room for site-selective conjugations, offers only limited options for linkers and poses limitations when the biological activity of one (or both) of the fusion partners depends on the free terminus [114]. Nevertheless, gene fusions such as of human α-L-iduronidase on the N-terminus of RTB lectin [115] and the Fc region of human IgG1 to the C-terminus of Avaren lectin [105] were previously demonstrated as functional drug delivery systems. Fucose binding rBC2LCN lectin fused to pseudomonas exotoxin A proved effective in colorectal cancer treatment [116]. Linkage at the protein level is often realized by chemical ligations involving lysine- and cysteine residues in proteins. Yet, more than one lysine or cysteine in the protein sequence will spoil the regioselectivity [117]. The conjugation reactions are inefficient if the lysine- or cysteine residues are buried in the protein structure [118]. A few chemically conjugated lectin-drug delivery vehicles were tested as potential biotherapeutics. For instance, wheat germ agglutinin (WGA)-conjugated liposomes loaded with amoxicillin showed potent antimicrobial activity [119]. To overcome the above-mentioned limitations of genetic fusions and bioconjugation at canonical amino acids such as lysine and cysteine, we embarked on a bioorthogonal conjugation strategy for Stx1B and the scFv OKT3. Bioorthogonal conjugation exploits unique reactive groups that are installed at predefined positions in the protein conjugation partners. To site-selectively install the unique reactivity, ncAAs carrying a corresponding reactive side chain can be introduced into the target protein(s) at an in-frame amber stop codon [117]. Here we used the pyrrolysyl-tRNA synthetase from Methanosarcina mazei and its cognate amber suppressor MmtRNACUA for this purpose. The MmPylRS/MmtRNACUA pair is orthogonal in E. coli, which means that the MmPylRS does not charge any of the E. coli tRNAs, nor is the MmtRNACUA charged with a canonical amino acid by any of the host aminoacyl-tRNA synthetases. Wild-type MmPylRS accepts a palette of (pyrol)lysine derivatives such as AzK [87]. We selected the sites for incorporation of AzK into Stx1B and scFv OKT3 such that the reactive azido-group would be surface exposed and as distant as possible from the glycan- or antigen-binding sites, respectively. These conditions were excellently fulfilled by residues K9 and E129 of Stx1B and scFv OKT3 (Fig. 1).
The direct bioorthogonal conjugation of two proteins requires that they are functionalized with compatible reactive groups. Unless an orthogonal pair accepts both corresponding ncAAs, individual orthogonal pairs are necessary [87]. Orthogonal pairs for the incorporation of azido-ncAAs [87, 120], cyclooctynyllysine derivatives [121], ncAAs with trans-cyclooctene- and bicyclooctyne- [122] as well as tetrazine- [123] side chain moieties were previously devised. While the azido-ncAAS such as AzK and para-azido-L-phenylalanine can be purchased, the ncAAs with reactive handles for SPAAC or IEDDA are either extremely expensive or commercially unavailable, which severely limits their use in a biotechnology lab. Being confronted with these issues, we decided to take an alternative path. Instead of directly linking Stx1B and scFv OKT3 via compatible reactive groups for SPAAC or IEDDA, we embarked on a two-step strategy (Fig. 3a): We first functionalized both proteins with an azido-group by the site-specific incorporatin of AzK as outlined above. AzK was efficiently incorporated into scFv OKT3 and Stx1B as reflected by titers corresponding to 70 % and 93 % of the corrspondingwild-type proteins. The azide groups allowed us to perform a SPAAC reaction with bi-functional linker molecules that carried each an azide-reactive DBCO group on the one end and a trans-cyclooctyne- or a tetrazine-moiety for IEDDA on the other. The resulting methyltetrazine functionalized Stx1B K9Tz and trans-cyclooctene functionalized scFv OKT3 E129TCO spontaneously assembled into the Stx1B-scFv OKT3 lectibody in the second conjuagtion step. Since the conjugate would be used for cell assays and to avoid copper-induced toxic effects in cell lines [124], we did not consider copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC) for the first-step conjugations [125]. Besides employing the commercially available SPAAC-reactive AzK that can be efficiently incorporated into target proteins, the two-step conjugation strategy offered another benefit. Selecting the linker lengths allowed flexibility in the orientation of Stx1B and scFv OKT3 in the Stx1B-scFv OKT3 conjugate such that the bispecific properties were retained. This flexibility is particularly crucial for the conjugation of large molecules such as proteins. Our prediction of the lectibody structure suggested that this would be the case. Indeed, the flow cytometry analysis confirmed that the Stx1B and scFv OKT3 modules preserved their affinity and recognition towards Gb3 and CD3 antigens, respectively, upon conjugation. Moreover, we observed the selective killing of Gb3-expressing cancer cells, which indicates that the T cells were brought in close contact with the cancer cells.
While the structural prediction of the lectibody indicated that five scFv OKT3 molecules can be conjugated to a Stx1B pentamer, our SEC results showed that only three scFv OKT3 molecules were conjugated (Table S3). This result adds another hint to the fact that bioorthogonal conjugation efficiencies rarely derive a 100% conjugate yield. Varying conjugation efficiencies were reported previously even with small molecules such as fluorescein and with DNA oligomers. For instance, Synakewicz et al. reported 47% yield for conjugating proteins with azide-modified DNA oligomers by SPAAC [76], and Maggi et al. observed a remarkable 62 % onjugation of trastuzumab-tetrazine with TCO-fluorescein by the IEDDA reaction [126]. These findings illustrate that conjugation efficiencies vary not only with the conjugation partners, i.e. protein with small molecule or DNA oligomer, but also with the conjugation method used. In this study, we were able to successfully conjugate three large ~ 30 kDa scFv OKT3 molecules to one ~ 45 kDa Stx1B pentamer via linkers by two successive SPAAC and IEDDA click reactions. Although studies by Li et al. [127] observed reduced cytotoxicity with the linker-based drug conjugate, our linker-based lectibody mediated crosslinking of T cells to tumor cells and induced high specific killing. As demonstrated, the Stx1B-scFv OKT3 promotes up to 93% of Burkitt´s lymphoma-derived tumor cells elimination within 48 hours of treatment. Strikingly, low nanomolar concentrations (7 nM and 35.6 nM) of the lectibody induced a potent T cells response against Gb3+ Ramos cells and promoted CD8+ T cells activation. On the one hand, linkers aid in the flexible orientation of the conjugated proteins and, on the other hand, they cause non-specific interactions leading to low yields during conjugation due to their hydrophobicity as reported elsewhere [128, 129]. It is key to understand what role the hydrophopathy of the linkers plays in the conjugation reaction to tackle the incomplete decoration of the Stx1B pentamer with the scFv OKT3 modules. For instance, Rahim et al. showed that 90% of the TCO groups without a linker attached on the surface of monoclonal antibodies were masked by their hydrophobic interactions with the antibodies. They had overcome this problem by adding hydrophilic PEG to the linker between the reactive groups [128]. In the similar way, to improve the conjugation efficiency of Stx1B with scFv OKT3 in the future, water-soluble methyltetrazine-PEG-DBCO linkers of short varying lengths might be used to reduce non-specific interactions.
Noticeably, the Stx1B-scFv OKT3 lectibody did not affect the viability of Gb3˗ Namalwa cells, excluding the presence of undesired off-target cytotoxicity. The observed poor unspecific binding to off-targets by linker-based drug conjugates are in line with Li et al. [127]. To this matter, the excellent specificity of BiTEs has made this platform an exquisite system for treating cancer. Due to their bispecific configuration, the induction of a strong T cell response is accomplished by physically bridging cells, and the linkage of T cells to tumor cells is crucial to the BiTE´s cytotoxic mechanism. Single-sided binding of BiTE molecules is not sufficient in driving the activation of T cells [130], nor to induce cytokines expression – including interferon gamma, tumor necrosis factor alpha, IL-6, and IL-10 – that are upregulated upon T cells stimulation. Lack of dual binding, as proven for Gb3˗ Namalwa cells, demonstrated the ability of Stx1B-scFv OKT3 to circumvent undesired T cell activation. This finding suggests the strict dependence of this lectibody on simultaneous T cell-tumor cell engagement, only in the presence of the GSL Gb3.
According to SEC analysis, the purified lectibody consisted of a conjugate of 140 kDa in size. This lectibody is roughly three times bigger in size than other recombinant bispecific antibodies able to redirect T lymphocytes to tumor cells. Generally, most bispecific antibody platforms – including BiTEs, tandem scFv molecules (taFv), diabodies (Db), or single chain diabodies (scDb) [131–133] – consist of small molecules with molecular masses of 50–60 kDa. However, due to their small size, these formats often suffer from poor pharmacokinetic (PK) and pharmacodynamic (PD) properties in vivo. For example, they are rapidly cleared from circulation and their half-life is less than 30 minutes, which makes it difficult to target them to the site-of-action for long duration times [134–136]. This trait hampers their therapeutic applications, as they require multiple doses and repeated injections or infusions of treatment. Several antibody-based platforms have been engineered to improve their PK properties, and most attempts have been directed so far to increase their molecular sizes. The addition of a third scFv to the polypeptide chain of bispecific antibodies led to the design of single chain triplebody formats (sctbs) with a mass of about 90 kDa, exhibiting one domain for the recruitment of effector cells and two specific binding sites for antigens on tumor cells [137]. Such design intended to promote a bivalent binding of cancer cells, thus targeting low-density tumor antigens, and a prolonged half-life of the format, whose molecular size was above the kidney exclusion limit. Several sctb prototypes were tested and published, and potent lysis of CD19-positive lymphoma cells and acute myeloid leukemia cells was reported [138–140]. The trispecific arrangement of the recombinant antibody showed favorable features in cell culture as well as in mice, redirecting the cytotoxicity of CD16+ effector cells towards CD19+, CD123+, CD33+ tumor cells and doubling the serum half-life as a consequence of the increased mass. Other common strategies comprise chemical coupling of polyethylene glycol (PEG) chains [141], multimerization design [135, 142], or fusion with long-circulating serum proteins such as albumin [143, 144]. Due to its larger size, the Stx1B-scFv OKT3 lectibody presented in this study has a great potential to avoid clearance from serum within few hours. To this end, future in vivo studies are required to establish the clearance rate and renal excretion of Stx1B-scFv OKT3, assessing the lectibody´s circulation times and protection from catabolism. On the other end, potential off-target cytotoxicity should be evaluated as a consequence of the lectibody persistence and recirculation in plasma. This includes the induction of T cell-mediated lysis in cells and tissues which present only a mild expression of this GSL, and the overactivation of the immune system after a stable binding of the scFv OKT3 to CD3 receptors on the surface of CD4+ and CD8+ T cells.
The remarkable performance of the Stx1B-scFv OKT3 lectibody in enabling T cell-mediated lysis of tumor cells is accompanied by activation of the cytotoxic CD8+ T lymphocytes. The treatment elicited a notable upregulation of crucial markers at the surface of T cells – namely CD69, CD71, and CD25 – at 24 and 48 hours post-treatment. The appearance of such activation markers led to the identification of different stages in the T cells activation process, which was exclusively driven by the co-presence of lectibody and Gb3+ target cells. CD69, CD71 and CD25 were significantly increased and reached a peak in surface expression that overlaps with the onset of cytotoxic activity, precisely at 24 and 48 hours. This increment in expression was elicited solely when Gb3+ Ramos cells and Stx1B-scFv OKT3 were added to the culture, indicating the simultaneous engagement of effector and target cells and the target-specific killing induced by our treatment. These findings are in line with the therapeutic mechanism of action reported for BiTEs. Accordingly, a unique feature of BiTE-induced T cell activation is the lack of dependence upon T cell co-stimulation. Several studies of BiTEs with various specificity have demonstrated that these molecules induce significant cytotoxicity by T cells in the absence of co-stimulation, such as through anti-CD28 antibodies and IL-2 [130, 145].
Bifunctional molecules represent a promising anti-cancer arsenal, targeting a variety of tumor-associated antigens on both solid and hematologic tumors [63]. In a combinatorial approach, our lectibody might improve the outcome of existing therapies for the eradication of those tumor cells which present dramatic alterations of glycosylation at the surface. When targeting cancer-associated glycans, glycan-binding proteins (GBPs), such as lectins and anti-glycan antibodies, can be used to discriminate between tumor and normal cells. As tumor-targeting ligands, lectins can be used to increase the selectivity and efficacy of anti-cancer treatments and enhance their concentration at the tumor sites. Interestingly, in addition to Stx1B, several other bacterial lectins have found application in tumor detection or treatment. For example, the cholera toxin (Ctx) from Vibrio cholerae, belonging to the AB5 family of microbial toxins, has been proven effective in a number of studies for tumor targeting and imaging [38]. The affinity for the ganglioside GM1, which is highly expressed in the blood-brain barrier, resulted in the investigation of its B-subunit (CtxB) for new anti-glioma chemotherapy strategies, where CtxB nanoparticles loaded with paclitaxel induced apoptosis of intracranial glioma cells in vivo and extended survival in mice models [146]. At the same time, this lectin was reported to be highly efficient in sensitive retrograde neuronal tracing [147–150]. More recently, we demonstrated the efficiency of the engineered FS-Janus lectin consisting of two carbohydrate-binding domains in detecting and targeting pathological hypersialylation on non-small cell lung cancer via its multivalent architecture, and with remarkable nanomolar avidity [151, 152]. The chimeric, bispecific lectin was reported to crosslink glyco-decorated giant unilamellar vesicles and lung epithelial tumor cells, leading to the intracellular uptake of liposomal content and unraveling its potential in lectin-mediated drug delivery. Moreover, Meléndez et al. developed a panel of lectin-based chimeric antigen receptors (CARs) T cells, which demonstrated high therapeutical potential towards a variety of hematological malignancies and solid tumors expressing Gb3. In their studies, the Gb3-binding lectins StxB from Shigella dysenteriae, LecA from Pseudomonas aeruginosa, and the engineered Mitsuba from Mytilus galloprovincialis were employed to recognize the TACA Gb3 and fused to a second-generation CAR, achieving excellent target-specific cytotoxicity against Burkitt’s lymphoma-derived cell lines as well as colorectal and triple-negative breast-cancer [153]. Overall, these studies together with our current study support the potential of lectins as tools in many therapeutical applications where the glycome plays a crucial role in the development and sustainment of pathological conditions.
The described observations of dual-binding dependence, target-specific killing and absence of co-stimulation suggest a model for our lectibody-mediated cytotoxicity, where multiple lectibody-dependent binding events occur between T cells and tumor cells in culture, promoting clustering of T cell receptors and activation of cytotoxic signaling. Indeed, the presence of several scFvs OKT3 in the lectibody allows this bispecific T cell engager to be presented in a polyvalent form. This might induce the formation of an immunological synapse and release of cytolytic granules resulting in tumor cell lysis, as observed for BiTEs. There is much further knowledge to gain regarding the mechanism of action of the lectibody. Foremost, the investigation of its efficacy and potency in animal models must be carried out to further characterize the potential of this bispecific format in cancer therapy, as well as a better understanding of how the lectibody is processed by the body and of the side effects that might appear during therapy. In addition, examining Stx1B-scFv OKT3 in combination with other therapies will be necessary. For instance, it will be interesting to see how the lectibody performs in association with immune checkpoint inhibitors, which promote a greater T cell activation. A combinatorial strategy of this type could increase Stx1B-scFv OKT3 efficacy by enhancing and maintaining T cell activity for tumor eradication [154]. As with most new immunotherapies, it would be attractive to examine if this lectibody is clinically useful against solid malignancies, and whether its performance can overcome the challenges imposed by the tumor microenvironment. Nevertheless, it is important to point out that the GSL Gb3 is present on the cell membrane of a variety of non-transformed cells, which could lead to the onset of off-target cytotoxicity upon administration of the lectibody. A key factor that must be taken into account is the difference in Gb3 expression levels between pathological and physiological states, highlighted by the impact they have on StxB participation to receptor recognition [155–158]. It has been proven that the binding of the toxin and its targeting to a specific intracellular transport pathway within Gb3+ cells is determined by the heterogeneity in Gb3 isoforms and the abundance of Gb3 in the lipid rafts of the plasma membrane [159]. According to these studies, Melendez et al. correlated the cytotoxicity of Gb3-targeting CAR T cells towards distinct Gb3+ and Gb3⁻ cells lines with Gb3 abundance and isoform variations via mass spectrometry analysis [153]. The analysis revealed a strong dependency between Gb3 presence and structural variability and specific killing by the lectin-CAR T cells. Ultimately, it becomes evident the necessity for a screening of cytological specimens to determine the Gb3 status of patients prior to making such a therapeutic choice.
Finally, the lectibody concept described here possesses versatile features, since it can be adapted to other TACAs by exchanging Stx1B with another lectin. Recent advances in the use of lectins in research and medicine suggest they are potential tools for many applications, such as drug delivery and selective targeting of pathological conditions with a focus on glycosylation changes [49, 50]. Lectins engineering, as proven in this study, may offer the possibility to target glycan epitopes on tumor cells and boost the efficacy of current tumor therapies.