Amber
Barendrecht‡
a,
Heleen H. C.
Peeters‡
a,
Diana
Torres-García‡
a,
M. Thierry
Shema
a,
Alexi J. C.
Sarris
a,
Shimrit
David
a,
Göktuğ
Aba
ab,
Camille M.
Le Gall
c,
Martin
Wilkovitsch
d,
Martijn
Verdoes
cef,
Hannes
Mikula
d,
Mark A.
Travis
g and
Sander I.
van Kasteren
*a
aDivision of Chemical Biology and Immunology, Leiden Institute of Chemistry, Universiteit Leiden, Gorlaeus Laboratory, Einsteinweg 55, Leiden, The Netherlands. E-mail: s.i.van.kasteren@chem.leidenuniv.nl
bLeiden University Medical Center, Albinusdreef 2, Leiden, The Netherlands
cDepartment of Medical BioSciences, Radboud University Medical Center, Nijmegen, The Netherlands
dInstitute of Applied Synthetic Chemistry, TU Wien, Getreidemarkt 9/163, Vienna, Austria
eInstitute for Chemical Immunology, Nijmegen, The Netherlands
fCurrent address: Department of Immunology, Leiden University Medical Center, Leiden, The Netherlands
gManchester Collaborative Centre for Inflammation Research, University of Manchester, Manchester, UK
First published on 4th June 2025
Immunocytokines have emerged as a promising modality in cancer therapy, capitalizing on the precision of antibodies to deliver cytokines selectively to tumours. Yet, the toxicity of the cytokine portion of these antibody-cytokine constructs remains a major dose-limiting issue. We present a new approach to control cytokine function without affecting binding of the targeting moiety. By modifying the cytokine with trans-cyclooctene carbamates at the lysine positions, we can reduce the binding to the receptor of various highly pro-inflammatory cytokines. Then, using a click-2-release (C2R)-approach, we can reactivate the cytokine activity by reacting it with a variety of tetrazines, through a Diels–Alder-pyridazine-elimination cascade. Finally, we show that the caged cytokines can be conjugated via a sortase motif to an unmodified targeting nanobody resulting in a targetable caged immunocytokine construct.
Extensive effort has therefore gone into detoxifying immunocytokines to reduce this off-site on-target side-effect profile, including targeted mutagenesis,9–11 local activation of the cytokine through dimerization,12 and increasing steric hindrance to lower cytokine affinity for its receptor.13,14 However, all these current methods are cytokine-specific and require extensive modification of the cytokine, potentially leading to reduced efficacy of the cytokine in the tumour, or even the cytokine being recognised as foreign and being cleared. We envisaged a method whereby the cytokine could be detoxified whilst accumulating in a tumour, and only locally activated, would be a powerful approach to detoxify immunocytokines.
We hypothesized that a click-2-release (C2R) strategy would be a great approach here.15 This method, uses the modification of a key heteroatom with a trans-cyclooctene group, first reported by Robillard and co-workers16 and then by Chen and co-workers,17 followed by inducing release with a tetrazine reagent (Fig. 1).18 This method has been used extensively to locally activate, and/or deliver prodrugs19 through the chemical protection, or ‘caging’ of key heteroatoms, such as amines and alcohols,20 followed by controlled deprotection using an inverse electron-demand Diels–Alder (IEDDA) reaction with a tetrazine.21 The reason is that the reaction is highly in vivo compatible, having been shown in mice,22 and now even pursued in human clinical trials.23 This method is also employed for the selective activation of chemotherapeutic agents, such as doxorubicin,16 and monomethyl auristatin19 at tumour sites. The IEDDA reaction enables a rapid and bioorthogonal cycloaddition between trans-cyclooctene (TCO) and tetrazine moieties, ensuring that therapeutic agents remain inactive until a tetrazine-based trigger is administered.24,25 It is currently even being pursued for the subcellular release of prodrugs26 and antigens.27,28 We envisaged that this approach would be highly beneficial for detoxifying immunocytokines as the introduction of one,29 or in the case of cytokines, multiple TCOs could lead to the blocking of the interaction of the cytokines with their receptors until the point in time of tetrazine addition. We postulated that an amber codon suppression approach29 in which a single key lysine is modified would not be of use due to the large contact site between the cytokine and its receptor, as multiple lysine residues are involved in cytokine activity in most cases.
Here we describe our first efforts in making caged immunocytokines that can be deprotected after antigen target binding, and their in vitro evaluation (Fig. 2). Unlike the only reported caged cytokine where the cytokine is not targeted, but has to be deprotected with an antibody tetrazine conjugate,30 we here describe a method in which we first cage the cytokine portion of an immunocytokine and conjugate it to an uncaged targeting moiety (a nanobody in this case) using a sortase-based approach. Such an approach yielded a caged immunocytokine in which the targeting was not affected by the caging, and only the function of the cytokine could be controlled. We also found that by lowering the temperature of the caging reaction, IL-1β, IL-2, TNF-α, and IFN-γ could be caged in a manner that allowed recovery of function after uncaging. Interleukin 2 (IL-2) and TNF-α are the most extensively investigated cytokines for immunocytokine production, due to their high toxicity and high potential clinical benefit in treating tumours.31 Immunocytokines based on these proteins have been used in the treatment of e.g. metastatic melanoma.32,33 Interferon gamma (IFN-γ), is critical for antitumour responses, and directly suppresses tumourigenesis, shown by heightened tumour susceptibility in IFN-γ-deficient models.34,35 IL-1β-based immunocytokines are a particular interest, as they not only have potentially toxic on-target off-tumour effects, but also have been reported to play controversial roles in cancer, abetting both development36–41 and tumour regression.42,43
IL-1β is highly amenable to a caging strategy as it has 6 surface-exposed lysine residues (K88, K92, K93,K94) that are involved in its receptor binding (Fig. S1, ESI†).44,45 IL-1β is produced as an inactive 31 kDa pro-protein and undergoes proteolytic cleavage to generate the active 17 kDa form.46 IL-1β interacts with two receptors: IL-1 receptor I (IL1RI), which triggers a pro-inflammatory signalling cascade via NF-κB activation, and IL-1 receptor II (IL1RII), which serves as a decoy receptor and inhibits downstream signalling.43,47,48 IL-1β is known to be involved in the progression of multiple myeloma49 as it causes the production of IL-6 which causes the development of diseased plasma cells.50 Contradictory to these findings, it has been found that mice with SA1 sarcoma benefit from injections with external IL-1β, causing the regression of the tumour.51 Its capacity to drive inflammation and its association with secondary cytokine release have made its systemic administration particularly challenging.45 The use of IL-1β in immunocytokine formats has been explored, including monoclonal antibodies F8 (specific to the alternatively-spliced extra-domain A domain of fibronectin, a marker of tumor angiogenesis) designed to selectively deliver IL-1β to tumour stromal components.52 However, despite the potential for improved targeting, these constructs have exhibited unacceptable systemic toxicity, ultimately preventing their clinical development. We therefore started by determining whether the activity of IL-1β could be controlled through caging with a TCO-modality in such a manner that it could be reactivated upon treatment with tetrazine.
This initial optimisation was performed using murine IL-1β as a model cytokine due to its well-characterized structure and high lysine content near the receptor-binding interface.54 This made this cytokine particularly suitable for lysine-targeted modification strategies. Initial caging experiments were performed with different concentrations of NHS-TCO in 20 mM HEPES buffer (pH 8) and caging efficiency was assessed by an IL-1β-specific ELISA (Fig. S2A–C, ESI†).54 Increasing NHS-TCO concentrations indeed reduced ELISA signal (Fig. S2A, ESI†), however, incubation beyond 1 hour analysis resulted in precipitation of the protein (as measured by SDS-PAGE, Fig. S2B, ESI†). Optimal conditions for caging (not taking decaging into account) were found to be incubation 1 hour with between 4.0–8.0 mM NHS-TCO at 37 °C (Fig. S2C, ESI†).
To determine whether TCO-modification also affected IL-1β-receptor binding, HEK-Blue IL-1β and RAW-Blue assays were performed.55,56 HEK-Blue cells detect cytokine activity via NF-κB-mediated SEAP expression, quantified using QUANTI-Blue.56–58 RAW-Blue cells are murine macrophages that are also engineered to express SEAP upon NF-κB activation.55,59 Colorimetric evaluation of SEAP activity therefore allowed the quantitative assessment of IL-1β-receptor activation by the TCO-modified IL-1β construct.58 Both the HEK-Blue IL-1β-assay (Fig. 3A–C) and the RAW-Blue assay (Fig. 3D–F) showed that increasing the NHS-TCO concentration during the modification reaction, resulted in a product that was less capable of activating these cells (Fig. 3A and D). The highest concentration of NHS-TCO (8 mM) reduced the activity to 10% of the original value. Under 18-hour control conditions, IL-1β activity increased non-significantly, but still unexpectedly (Fig. 3E), whereas extended incubation (24 hours; Fig. 3B) results in decreased activity. These observations suggest that prolonged single-protein incubation at 37 °C may induce protein folding changes that alter cytokine activity, potentially contributing to variability in experimental outcomes. Lowering the caging temperature did not affect caging efficiency (Fig. 3C and F).
To correlate IL-1β activity loss to lysine modification levels, ESI LC-MS was performed on caged samples (Fig. S3A–G, ESI†). For sufficient signal, protein concentration was increased to 11 μM. Murine IL-1β contains 15 surface-exposed lysines, of which 14 are accessible for modification.54 LC-MS analysis showed that caging with 8.0 mM NHS-TCO resulted in an average of 7–8 modified lysines (Fig. S3B, ESI†), while 1.6 mM NHS-TCO reduced this to 4 (Fig. S3A, ESI†). Increasing NHS-TCO to 16 mM did not raise the average modification level but decreased the proportion of minimally modified species (Fig. S3B and C, ESI†).
It was next determined whether any of the above-caged proteins could also be reactivated upon reaction with tetrazines. And here the project hit a snag: initial decaging experiments (Fig. S4A–C, ESI†) showed that IL-1β could not be restored to its functional form for any of the above caging levels. This may be due to incomplete removal of TCO groups or irreversible inactivation due to protein unfolding during caging. We postulated that, given that IL-1β has 14 accessible lysines44,60 extensive modification at elevated temperatures and high NHS-TCO concentrations could lead to the disruption of the protein fold, leading to an irreversible loss of activity. In an attempt to address this, we assessed whether modifications at lower temperatures could prevent this permanent deactivation during the TCO reaction. IL-1β was therefore modified with 3.5 mM, 2.1 mM or with 0.7 mM NHS-TCO at 10 °C for 24 h (Fig. 4B). HEK-Blue IL-1β confirmed that caging with this lower NHS-TCO concentration at the lower temperature/longer reaction time resulted in similar reduction of IL-1β (Fig. 4B). Using these caging conditions, decaging was tested in the HEK-Blue IL-1β assay using tetrazines (2–6, Fig. 4A and C). Tetrazines 2–4 yielded significant recovery of IL-1β activity in this assay, with 3,6-dimethyl-tetrazine 2 showing the highest IL-β activity recovery of approximately 60%.
Following promising caging and decaging results with IL-1β, the strategy was extended to a sortaggable IL-1β variant expressed from a pET28a(+) vector, which showed comparable behaviour to wild-type IL-1β. The approach was also applied to therapeutically relevant cytokines, TNF-α, IL-2, and IFN-γ. IL-2 (K35, K43 and K64)61 and IFN-γ (K108, K125, K128 and K130)62 contain lysine residues critical for receptor binding. For TNF-α this is more complex, as no lysine residues are directly involved in receptor binding or trimerization. However, certain lysine residues are critical for proper TNF-α (K112) folding, and their modification may impair the structural integrity of the cytokine, thereby indirectly reducing its affinity for TNF receptors (TNFRs).63 Initial caging experiments at 37 °C again resulted in proteins that could not be efficiently reactivated by tetrazine treatment (Fig. S5A–I, ESI†). However, like for IL-1β, the cold caging strategy with low-temperature incubations with reduced NHS-TCO concentrations resulted in efficient inactivation for all cytokines (Fig. 5A–C). Decaging with tetrazine 2 of TNF-α, IL-2 and IFN-γ, yielded some promising results: IL-2 function was restored to 90% at a concentration of 75 μM 2 (Fig. 5D). IFN-γ function was restored to 60% at a concentration of 25 μM 2 (Fig. 5E), increasing demethylsufide (DMT, 2) concentration did not result in more sufficient decaging (data not shown). Decaging of TNF-α (Fig. S6A, ESI†) restored only 20% compared to caged cytokine activity. Further optimisation of TNF-α caging did not improve either the caging, nor the decaging activity. A novel tetrazine, 2,2′-(1,2,4,5-tetrazine-3,6-diyl)bis(pyridin-3-ol) ((2PyrH)2Tz),64 outperformed other tetrazines for IL-1β and IL-2 decaging (Fig. 5F and G), but failed to restore TNF-α activity (Fig. S6B, ESI†). Moreover, higher concentrations of (2PyrH)2Tz induced cytotoxicity (data not shown). Decaging of IFN-γ using (2PyrH)2Tz was inconclusive due to high background signal in the Quanti-Luc assay. However, a test with Tz4, that showed promising decaging for IL-1β, restored activity of IFN-γ to approximately 75% of the original activity (Fig. 5H). It may be valuable to explore alternative tetrazine designs similar to those described by Fan et al. (2016).65 In this work, the authors introduced a series of unsymmetrical tetrazines, optimized with an electron-withdrawing group (EWG) at the 3-position to accelerate the initial cycloaddition with TCO, and a small, non-EWG group at the 6-position to facilitate the subsequent elimination (release) step. Their best-performing derivative achieved >90% decaging efficiency in living cells within 4 minutes, highlighting the potential of such structural designs to further enhance bioorthogonal decaging performance.
Building on the successful control of cytokine activity via click-2-release chemistry, the next objective was to generate an immunocytokine by conjugating the caged cytokine to a tumour-targeting moiety. While immunocytokines are typically produced via genetic fusion, this strategy is incompatible with non-selective lysine modification, as it would indiscriminately modify both the cytokine and targeting domain. Although this would result in the desired loss of cytokine activity, it would also result in undesired loss of antigen binding capacity. The targeting-reagent therefore had to be coupled to the cytokine only after it had been modified with TCO-carbamates. To achieve this, it was decided to use a sortase A-based approach in which a small single-chain fragment of a camelid antibody was genetically modified with an N-terminal tag and the cytokine with a C-terminal tag (LPETG). Sortase A from Staphylococcus aureus catalyses the ligation of a LPETG-motif and a poly-glycine tail in peptide–protein66 and protein–protein67,68 systems. This strategy was employed to ligate a TCO-modified cytokine carrying a C-terminal LPETGG-tag to a nanobody equipped with an N-terminal glycine tag. Reported ligation yields (40–85%), would be sufficient to produce caged immunocytokines for further evaluation.
In order to obtain the immunocytokine in which the cytokine was inactive, a construct consisting of a nanobody targeting group is used, linked by a sortase reaction to the murine IL-1β-gene construct. The choice was made to use a nanobody instead of an antibody as a targeting group. Until now immunocytokines were based on intact antibodies (IgGs)67,69,70 or single chain variable fragments (scFv).71–77 However nanobodies were selected over full-length antibodies or scFvs due to their smaller size, thermal and pH stability, lack of glycosylation and hydrophobic domains, and efficient bacterial expression.78,79 IL-1β was modified with the sortase motif at the C-terminus for two reasons: first, N-terminal TCO-modification would preclude sortase ligation; second, N-terminal modification has been reported to reduce IL-1β activity.80 The IL-1β gene (Gene ID: 16176) was engineered with a C-terminal GGGGS spacer, followed by the LPETGG sortase recognition motif and a 6His-tag for purification (Fig. 6).81–84
To obtain IL-1β-LPETGG-6His the sequence was expressed from pET28a(+)-vector using E. coli ArcticExpress (DE3) RP system, which can express proteins with rare codons.85,86 Expression was induced with 0.5 mM IPTG under T7 promoter control, and carried out at 10 °C for 3 days (Fig. S7A, ESI†).87 Lysis was performed with lysozyme and sonication, followed by purification from the soluble fraction via nickel affinity chromatography (Fig. S7b, ESI†). Sortase A was expressed from the pET28aSrtAΔ59-expression vector as previously described.87 The anti-CD11c nanobody or VHH-CD11c modified with a pelB signal sequence was expressed from pET22b-vector in E. coli BL21 (DE3) pLysS as reported.88
Next, the sortase-mediated coupling reaction of the nanobody to the uncaged cytokine was optimised using unmodified IL-1β-LPETGG. The initial ligation reactions were performed in 50 mM Tris/150 mM NaCl/10% glycerol pH 7.5 for 1 hour at 37 °C using equal equivalents of both IL-1β-LPETGG (19.7 kDa) and VHH-CD11c-SIINFEKL (15 kDa 4 μM each) and 0.75 equivalents of sortase A (3 μM). SDS-PAGE (Fig. S8A, ESI†) confirmed formation of the expected 35 kDa conjugate (red arrow). The gel showed additional bands at molecular weights lower then IL-1β-LPETGG or sortase A, even in the absence of the nanobody. Further research into these proteins was not performed. The assumption was made that these bands originated from hydrolysis of the bond formed between sortase A and IL-1β-LPETGG. The result is cleavage at the LPETGG site, generating a smaller fragment of IL-1β. Optimisation experiments indicated that higher sortase concentrations or prolonged reaction times (≥2 h) reduced product yield and shorter reaction times (1–15 min) improved efficiency (Fig. S8B–D, ESI†). This was most likely due to the fact that sortase A can also hydrolyse the formed bonds between the nanobody and the IL-1β-LPETGG.89 The most effective ligation conditions, minimizing side product formation due to sortase A–mediated hydrolysis, were 4 μM cytokine, 8 μM nanobody, and 3 μM sortase A with a 15-minute reaction time. Although the reaction was not complete, the yield was sufficient for subsequent experiments (Fig. 7).
These conditions were then used to ligate the caged cytokine to the anti-CD11c nanobody. IL-1β-LPETGG (5.0 μM) was caged using 3.5 mM NHS-TCO and the caged product was coupled to VHH-CD11c-SIINFEKL by sortase A using the optimal conditions found in Fig. S8B–D (ESI†). The red arrow indicated the 35 kDa product formed between caged IL-1β-LPETGG and VHH-CD11c-SIINFEKL, which was only formed when both substrates and sortase A were present (Fig. S9A, ESI†). The green arrow indicated an intermediate (38 kDa) formed between sortase A and caged IL-1β-LPETGG in the absence of VHH-CD11c-SIINFEKL (Fig. S9A, ESI†). Finally, the dark purple arrow indicated the presence of uncoupled caged IL-1β-LPETGG (Fig. S9A, ESI†). Product formation was assessed with IL-1β-specific western blot (Fig. S9B, ESI†) and, for the TCO-modified products, by reaction with BODIPY-TMR-tetrazine (Fig. S9A, ESI†). Tetrazine reacts with trans-cyclooctene (TCO), resulting in decaging of most substrates; however, fluorophores are highly sensitive, and while release is initiated, a sufficient amount of the TCO–tetrazine intermediate remains stabilized to be detected via gel analysis. Additionally, the steric bulk of the fluorophore can hinder release.
To optimise the deprotection step, crude mixtures of IL-1β-VHH-CD11c-SIINFEKL were tested. These were made by first caging IL-1β-LPETGG with 3.5 mM, 2.1 mM, or 0.7 mM NHS-TCO, followed by ligation to the nanobody. The resulting mixtures were added to HEK-Blue IL-1β cells to measure activity. Caged IL-1β binding to its IL-1R receptor compared to uncaged IL-1β was 2% with 3.5 mM NHS-TCO, 5% with 2.1 mM NHS-TCO, and 16% with 0.7 mM NHS-TCO (Fig. 7B, Fig. S10A and B, ESI†). Treatment with (2PyrH)2Tz successfully restored full activity for the 2.1 and 0.7 mM samples, but activity recovery for the 3.5 mM sample remained below 20%. Purification of the crude mixtures via size exclusion chromatography (SEC) yielded limited success, and needs to be further optimised (Fig. S11A–D, ESI†).
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5cb00113g |
‡ Contributed equally. |
This journal is © The Royal Society of Chemistry 2025 |