Benzyl Ammonium Carbamates Undergo Two‐Step Linker Cleavage and Improve the Properties of Antibody Conjugates

Abstract

Targeted payload delivery strategies, such as antibody‐drug conjugates (ADCs), have emerged as important therapeutics. Although considerable efforts have been made in the areas of antibody engineering and labeling methodology, improving the overall physicochemical properties of the linker/payload combination remains an important challenge. Here we report an approach to create an intrinsically hydrophilic linker domain. We find that benzyl α‐ammonium carbamates (BACs) undergo tandem 1,6–1,2‐elimination to release secondary amines. Using a fluorogenic hemicyanine as a model payload component, we show that a zwitterionic BAC linker improves labeling efficiency and reduces antibody aggregation when compared to a commonly used para‐amino benzyl (PAB) linker as well as a cationic BAC. Cellular and in vivo fluorescence imaging studies demonstrate that the model payload is specifically released in antigen‐expressing cells and tumors. The therapeutic potential of the BAC linker strategy was assessed using an MMAE payload, a potent microtubule‐disrupting agent frequently used for ADC applications. The BAC‐MMAE combination enhances labeling efficiency and cellular toxicity when compared to the routinely used PAB‐Val‐Cit ADC analogue. Broadly, this strategy provides a general approach to mask payload hydrophobicity and improve the properties of targeted agents.

Towards the goal of identifying an intrinsically polar linker domain, we report that benzyl α‐ammonium carbamates (BAC) undergo a tandem 1,6–1,2 elimination sequence to release secondary amines. A zwitterionic BAC linker variant is found to improve the labeling and reduce antibody aggregation when compared to a conventional and a cationic BAC linker. Finally, fluorescence imaging studies demonstrate that the model payload is released in antigen‐expressing cells and tumors with high specificity.

Targeted drug delivery strategies combine a targeting agent and a payload molecule through a cleavable linker.[ ¹ , ² ] Extensive efforts to develop antibody‐drug conjugates (ADCs) have validated the significant therapeutic potential of this approach.[ ³ , ⁴ , ⁵ ] The success of an ADC depends on the careful selection of appropriate payloads, optimization of antibody characteristics, and the use of linkers that cleave only after target engagement.[ ⁶ , ⁷ ] A key challenge lies in managing the hydrophobic nature of most small molecule payloads and linkers, as their hydrophobicity can compromise the efficiency of protein labeling, lead to physical instability (i.e. protein aggregation), and contribute to off‐target toxicities.[ ⁸ , ⁹ , ¹⁰ ] An appealing strategy is to engineer the linker domain to improve the overall physical properties of the ADC.

Existing linkers often rely on the self‐immolative para‐amino benzyl (PAB) spacer unit, which was first introduced for biological applications by Katzenellenbogen in 1981. ^[11] Based on an efficient and general 1,6‐elimination (Figure 1A), this approach has been widely used to create cleavable linkers that respond to various stimuli. ^[12] Nevertheless, the PAB spacer introduces a hydrophobic element into the linker domain, and several studies have found that the PAB linker contributes to ADC off‐target uptake.[ ⁸ , ¹³ , ¹⁴ ]

Figure 1.

(A) Payload release with conventional para‐aminobenzyl spacer and (B) with zwitterionic benzyl α‐ammonium carbamate spacers described here.

The key role of the chemical components of mAb (monoclonal antibody)‐conjugate targeting has been investigated extensively in imaging efforts.[ ¹⁵ , ¹⁶ , ¹⁷ ] These studies typically sought to develop mAb and other conjugates for diagnostic and interoperative purposes by attaching variably substituted imaging probes to a range of targeting units.[ ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ ] A critical insight to emerge from these studies is that charged, but net neutral (i.e. zwitterionic) functional groups can effectively shield the hydrophobicity of the payload and improve in vivo targeting performance. We hypothesize that suitable chemistry could introduce otherwise hydrophobic payloads onto targeting mAbs without compromising the properties of the parent antibody. While several efforts have looked at pegylation strategies,[ ⁹ , ²⁴ , ²⁵ , ²⁶ , ²⁷ ] no approaches that incorporate zwitterionic linker domains have been developed.

Herein, we detail a cleavable zwitterionic linker strategy utilizing the tandem 1,6–1,2‐elimination of benzyl α‐ammonium carbamates (BACs). This design was inspired by prior examples that applied the individual reaction components. The 1,6‐elimination of quaternary amines has been applied to prepare ADCs with tertiary‐ and heteroaryl‐amine payloads.[ ²⁸ , ²⁹ ] The triggered use of 1,2‐elimination reactions – part of the hemiaminal to carbonyl hydrolysis pathway – has been applied in several contexts and neutral methylene alkoxy carbamates have been used as ADC linkers.[ ³⁰ , ³¹ , ³² ] In this work, we fuse these two fundamental chemistries such that upon trigger cleavage, 1,6‐elimination unveils a nitrogen lone pair that undergoes spontaneous 1,2‐elimination to release the payload (Figure 1B).

A feature of the BAC strategy is that the cleavage site is independent from the conjugation chemistry. This enables the use of conventional protease‐cleavable linkers, but also less common terminating‐cleavable groups. To demonstrate the latter, a para azido aryl group and a para nitro aryl group were employed. Aryl azides can be reduced by biologically compatible phosphines to provide a non‐enzymatic test of the cleavage chemistry. Nitroarenes are known to be reduced to corresponding aniline by nitroreductases, which are overexpressed in hypoxic tumors.[ ³³ , ³⁴ ] To demonstrate the former, we used a valine‐alanine (Val‐Ala) dipeptide, which can be cleaved by ubiquitous lysosomal cathepsins and has been extensively applied for ADC development.[ ³⁵ , ³⁶ ] Finally, an acetamido group was used as a non‐cleavable control.

To model the payload component, we chose a far‐red hemicyanine dye HcyNMe (see Figure 2A and Scheme S1 for molecular structure). The corresponding primary aniline has been used extensively as a fluorogenic probe, as its emission is significantly reduced by attaching a carbamate to the nitrogen.[ ³⁷ , ³⁸ , ³⁹ ] For synthetic reasons described below, we use an N‐methyl derivative, which exhibits similar optical properties (Φ_F=0.09) to the parent probe (Figure S1).

Figure 2.

(A) Synthesis of N₃‐Hcy, NTR‐Hcy, CatB‐Hcy, NC‐Hcy. (B) Fluorescence emission spectra and (C) time course study of 10 μM N₃‐Hcy in the absence (black) or presence (red) of TCEP (100 μM). n=3. (D) Releasd HcyNMe from the probes (10 μM) in the presence of different physiological species or under different pHs after 2 h: TCEP (100 μM), NTR (nitroreductase, 20 U/mL), CatB (cathepsin B, 15 U/mL), elastase (1 U/mL), β‐Glu (β‐glucuronidase, 50 U/mL), GGT ($\gamma$ ‐glutamyltranspeptidase, 100 U/mL), peroxidase (100 U/mL), pH 4.0 PBS (100 mM), pH 8.0 PBS (100 mM), GSH (glutathione, 10 mM), cysteine (100 μM), Na₂S (100 μM), H₂O₂ (100 μM). n=3. (E) Structures of control probes CTRL‐1 and CTRL‐2. (F) Generation and purification of Pan‐probe conjugates. (G) Representative FPLC chromatograms of Pan‐probe conjugates. (H) Summary of antibody conjugation efficiency – % area of polydispersed fraction (A_poly). DOL, and % yield (n=3). (I) Release of HcyNME from 100 μg/ml Pan‐probe conjugates after treatment with TCEP (100 μM), NTR (20 U/mL), CatB (15 U/mL), elastase (1.0 U/mL), β‐Glu (β‐glucuronidase, 50 U/mL), or PBS (10 mM, pH 7.2) for 12 h. Data are expressed as the mean±SD of three separate measurements.

The syntheses of the probes are described in Figure 2A and Scheme S1–S3. Briefly, HcyNMe was converted to the corresponding chloromethyl carbamate, which undergoes a displacement reaction with benzyl piperidine nucleophiles to create the key BAC component. Critically, we found this reaction proceeds readily only in the case of secondary anilines and amines, and not with any primary examples we have tested to date. However, tertiary carbamates using payloads with secondary amines, such as MMAE, are much more common than secondary carbamates, likely due to stability issues.[ ²⁹ , ⁴⁰ ] A negatively charged sulfonate functionalized amine linker, which we have used in previous studies, ^[41] was further attached to the spacer and the tert‐butyl protecting group was cleaved to afford the four carboxylic acid derivatives, namely N₃‐Hcy, NTR‐Hcy, CatB‐Hcy and NC‐Hcy.

To evaluate the cleavage chemistry and stability of the linker, the spectroscopic responses of the four molecules were first examined in vitro. We first tested the chemically activated azide variant. Exposure of 10 μM N₃‐Hcy to 100 μM tris(2‐carboxyethyl)phosphine (TCEP) led to a 45‐fold fluorescence enhancement at 715 nm (Figure 2B), indicating a TCEP‐triggered tandem linker cleavage. A time course study (Figure 2C) showed that the payload was quantitatively released within about 30 min. The reaction of NTR‐Hcy with nitroreductase (85 %) and CatB‐Hcy with cathepsin B (17 %) also provided significant release of free HcyNMe (Figure 2D, S2, and S3). The detection limits, kinetic parameters K _m (Michaelis constant) and V _max (maximum of initial reaction rate) for probe NTR‐Hcy and CatB‐Hcy were measured with values comparable to reported probes for nitroreductase[ ⁴² , ⁴³ ] and cathepsin B[ ⁴⁴ , ⁴⁵ ] (Figure S2, S3). Probe CatB‐Hcy with the Val‐Ala linker also responds to neutrophile elastase, which has been reported in previous contexts.[ ⁴⁶ , ⁴⁷ , ⁴⁸ ] A range of other common physiological species do not induce significant release of free HcyNMe (Figure 2D), verifying the trigger specificity of the BAC linker. The release of HcyNMe fluorophore was further verified by HPLC analysis, and the tertiary amine intermediate has not been detected to date suggesting that the 1,2‐elimination step proceeds efficiently (Figure S4). Overall, these results demonstrate that the zwitterionic BAC linker is likely to be stable in biological settings and undergoes triggered payload release.

The hydrophilicity of these molecules was then assessed. For comparison, two additional control compounds were synthesized with a Val‐Ala linker or a cationic BAC linker, namely CTRL‐1 and CTRL‐2 (Figure 2E, Scheme S4). Their lipophilicity was then examined by measuring the probes partitioning between n‐octanol and PBS (1X, pH 7.2) to determine the Log P_oct values. These values were found to decrease along the series CTRL‐1 (1.15)>CTRL‐2 (0.47)>CatB‐Hcy (−0.17)>NTR‐Hcy (−0.46)>N₃‐Hcy (−0.51)>NC‐Hcy (−1.01) (Table S1), indicating that the zwitterionic BAC linkers increase the overall hydrophilicity of the linker‐payload combination.

Motivated by these promising proof‐of‐concept results, we then set out to assess the utility of the linker for protein conjugation. Probes with carboxylic acids were converted to the corresponding NHS‐esters. Standard lysine labeling with panitumumab (Pan), a clinical monoclonal antibody that targets epidermal growth factor receptor (EGFR), was then carried out in PBS (pH 7.2) to prepare Pan‐N₃‐Hcy, Pan‐NTR‐Hcy, Pan‐CatB‐Hcy, Pan‐NC‐Hcy and control conjugates Pan‐CTRL‐1 and Pan‐CTRL‐2 (Figure 2F, Scheme S3, S4). The conjugation reactions were initially purified using a desalting PD‐10 column and further purified/assessed by size‐exclusion fast protein liquid chromatography (FPLC). FPLC chromatograms (Figure 2G) showed that the main products of the conjugation were monodispersed Pan‐probe conjugates that eluted at ~18 min. The lower intensity peaks eluted at ~15 min and 27 min represent polydispersed antibody aggregates and free probes, respectively.[ ⁴⁹ , ⁵⁰ ] The conjugation efficiency was quantitively analyzed as summarized in Figure 2H. Using 10 equiv. of the corresponding NHS esters produced Pan‐probe conjugates with degree of labeling (DOL 2.9–3.2, Figure S5) and high monomer mass recoveries (62 %–67 %). Under the same conditions, CTRL‐1 formed the polydispersed species to a substantial degree, which resulted in a poor conversion (27 %) for Pan‐CTRL‐1. Similarly, CTRL‐2 provided only low DOL (1.2) conjugates. Dynamic light scattering (DLS) was used to characterize the z‐average size (hydrodynamic diameter d_H ) of the FPLC fractions (Figure S6, S7).[ ⁵¹ , ⁵² ] As anticipated, the main monodispersed peaks all had z‐average sizes of ~13 nm, which were identical to that of the unlabeled antibody. However, the polydispersed fractions vary as a function of the small molecule component, with CTRL‐1 and CTRL‐2 providing dramatically larger polydispersed fractions (z‐average size=100 nm and 483 nm, respectively) than those obtained with the zwitterionic BAC linkers (z‐average size ~20 nm). Overall, these results support the notion that the zwitterionic BAC molecules improve the efficiency of mAb labeling and the physical properties of the resulting conjugates.

The cleavability and susceptibility of the Pan‐probe conjugates were then evaluated. 95 %, 47 %, and 30 % of HcyNMe was released from Pan‐N₃‐Hcy, Pan‐NTR‐Hcy and Pan‐CatB‐Hcy after incubation with TCEP, nitroreductase and cathepsin B for 12 h, respectively (Figure 2I). As seen previously, the Pan‐CatB‐Hcy conjugate was also cleaved by elastase.[ ⁴⁷ , ⁴⁸ ] In addition, all four conjugates showed ~10 % release of HcyNMe payload after 72 h in human plasma independent of the trigger chemistry (Figure S8), indicating their good plasma stability.

Next, we investigated the properties of Pan‐NTR‐Hcy and Pan‐CatB‐Hcy in a cellular context. EGFR‐positive JIMT‐1 cells and EGFR‐negative MCF‐7 cell lines were used to examine the relationship between targeting and payload release. All of the Pan‐probe conjugates displayed negligible cytotoxicity (Figure S9). Confocal microscopy images were recorded longitudinally under same conditions and flow cytometry was used for quantitative analysis. After incubation with 10 μg of Pan‐NTR‐Hcy in hypoxic conditions (1 % O₂) for 24 h, the fluorescence intensity from JIMT‐1 cells was 3.5 times higher than that from MCF‐7 cells (Figure 3A, 3B, S10, S11), indicating the targeting ability of the conjugate was maintained. Moreover, the signal from Pan‐NTR‐Hcy ‐treated hypoxic JIMT‐1 cells was 1.7‐fold higher than that from Pan‐NTR‐Hcy‐treated normoxic JIMT‐1 cells, which could be ascribed to the overexpression of nitroreductase in hypoxic environments. A similar analysis comparing Pan‐CatB‐Hcy and Pan‐NC‐Hcy‐treated cells further supports a role for receptor‐mediated uptake and cathepsin‐mediated cleavage (Figure S12, S13). Colocalization analysis showed that most of the fluorescent signals originated from mitochondria (Figure S14), in line with the hydrophobic, cationic features of HcyNMe dye.[ ⁵³ , ⁵⁴ ] Taken together, these results support the notion that antibody conjugates constructed with zwitterionic BAC linkers undergo both receptor‐mediated targeting and selective cleavage.

Figure 3.

(A) Confocal fluorescence images of JIMT‐1 cells (a, b) and MCF‐7 cells (c, d) incubated with 10 μg Pan‐NTR‐Hcy under hypoxic (1 % O₂) or normoxic conditions for 24 h. λ _ex=633 nm, λ _em=650–750 nm. The second and fourth rows are the corresponding differential interference contrast (DIC) images. Scale bars=20 μm. (B) Corresponding normalized flow cytometry histograms and (C) quantification of fluorescence intensities in Alexa Fluor 700‐A channel (n=3). λ _ex=633 nm, λ _em=685–735 nm. Data points are displayed as the mean±SD. Statistical analysis performed by Student's t‐test (**p≤0.01).

Next, the in vivo performance of Pan‐NTR‐Hcy, Pan‐CatB‐Hcy and Pan‐NC‐Hcy was investigated by intravenously injecting 250 μg of the Pan‐probe conjugates into female athymic nude mice implanted with JIMT‐1 tumors in the axilla mammary fat pad. The mice were imaged at 4 h, 24 h, 48 h and 72 h post‐injection using a commercial in vivo imaging system (IVIS). After 24 h, a strong fluorescence signal was observed from tumors injected with Pan‐NTR‐Hcy (Figure 4A, 4B, S15), and the signal persisted over 72 h, with maximum tumor to background ratios (TBRs) reaching 2.7 (±0.36) at about 48 h (Figure S16A). Tumors injected with Pan‐CatB‐Hcy displayed tumor signals that increased until 72 h post‐injection, with a TBR reaching 2.5 (±0.50). In contrast, the Pan‐NC‐Hcy group exhibited weak fluorescence signals at all time points, with TBRs between 1.4–1.7. Due to the excitation and emission wavelengths, we observed some autofluorescence from the feed, as indicated by the control mouse only injected with PBS (Figure S15, S17). This signal can be effectively removed using spectral unmixing methods (Figure S18). Overall, these results are consistent with the cellular data, suggesting that conjugates with BAC linkers enable the effective release of payloads within the region of interest. Notably, previous mouse studies using hydrophobic heptamethine cyanine‐mAb conjugates exhibit significant non‐specific liver signal, especially within the initial 48 h.[ ¹⁶ , ⁵⁵ ] In contrast, mice injected with conjugates in this study did not show detectable liver signals at any time point (Figure 4C, S16B), with liver‐to‐background ratios (LBRs) ranging from 0.8 to 1.1. We attribute this targeting performance to the incorporation of the zwitterionic spacer that reduces the overall hydrophobicity of the conjugates. The mice were sacrificed at 72 h post‐injection, and fluorescence imaging of the excised organs corroborated the high tumor specificity of the Pan‐probe conjugates (Figure 4D, 4E, S19). Overall, these imaging results suggest that zwitterionic BAC linkers provide mAb conjugates that selectively target and release payloads in solid tumors.

Figure 4.

(A) Representative in vivo brightfield and fluorescence images of the JIMT‐1 tumor‐bearing mice after intravenous injection of Pan‐probe conjugates (250 μg dose) and imaging at 4, 24, 48, and 72 h post‐injection. Tumors are highlighted in dotted red circles. (B) Quantification of tumor signal (total radiant efficiency normalized to tumor size, n=5). Statistical analysis performed by Student's t‐test (*p≤0.05, **p≤0.01, ***p≤0.001). (C) Quantification of LBR (n=5). (D) Representative overlaid brightfield and fluorescence images of excised organs. (E) Biodistribution of Pan‐probe conjugates as determined by quantifying the fluorescence of excised organs (n=5). Tu: tumor, Li: liver, Lu: lung, Ki: kidneys, Pa: pancreas, Mu: muscle, Bl: bladder, In: small intestines, Sp: spleen. λ _ex=660–690 nm, λ _em=710–730 nm. Data points are displayed as the mean±SD.

Finally, we tested the suitability of the BAC strategy for constructing ADCs with therapeutic payloads. To do this, two ADCs were prepared by conjugating the potent microtubule‐disrupting agent MMAE to panitumumab through a zwitterionic Val‐Ala linker or a widely used PAB‐Val‐Cit linker, namely Pan‐BAC‐VA‐MMAE and Pan‐PAB‐VC‐MMAE (Figure 5A, see Figure S5 for full structures).[ ⁵⁶ , ⁵⁷ ] The DOLs of Pan‐BAC‐VA‐MMAE and Pan‐PAB‐VC‐MMAE are 3.3 and 2.0, respectively, as measured by HPLC‐Q‐TOF‐MS analysis (Figure 5B, 5C), indicating that the BAC linker exhibits higher labeling efficiency. FPLC analysis showed that both ADCs formed neglectable aggregates (Figure S20). We also carried out an in vitro cleavable assay and found that cathepsin B treatment leads to meaningful release of MMAE payload from Pan‐BAC‐VA‐MMAE and Pan‐PAB‐VC‐MMAE (Figure S21). Cytotoxicity studies confirmed the high potency of free MMAE in JIMT‐1 (IC₅₀=0.031 nM) and MCF‐7 cells (IC₅₀=0.062 nM, Figure S22). Notably, Pan‐BAC‐VA‐MMAE (IC₅₀=0.18 nM with respect to payload concentration) exhibited greater potency than Pan‐PAB‐VC‐MMAE (IC₅₀=1.27 nM) in EGFR‐positive JIMT‐1 cell lines (Figure 5D) and demonstrated lower off‐target toxicity in EGFR‐negative MCF‐7 cell lines (Figure 5E). These results suggest that the BAC linkers can improve mAb labeling efficiency, improve potency, and reduce cellular off‐target uptake.

Figure 5.

(A) Structures of Pan‐BAC‐VA‐MMAE and Pan‐PAB‐VC‐MMAE. Representative HPLC Q‐TOF‐MS analysis of (B) Pan‐BAC‐VA‐MMAE and (C) Pan‐PAB‐VC‐MMAE. Cytotoxicity of Pan‐BAC‐VA‐MMAE and Pan‐PAB‐VC‐MMAE in (D) JIMT‐1 cells and (E) MCF‐7 cells. The results are the mean±SD of three separate measurements.

In summary, a new cleavable ADC linker strategy has been developed utilizing the tandem 1,6–1,2‐elimination of benzyl α‐ammonium carbamates. We demonstrate that this linker enhances yield/labeling density while mitigating the formation of aggregated conjugates when compared to PAB strategies. Importantly, unlike the PAB approach, this cleavage chemistry can be applied not only to conventional proteolytic cleavage but also to terminal hypoxia responsive nitroaryl triggering groups. While we have implemented this strategy without site‐specific conjugation, it holds promise for integration with various other conjugation methods. Moreover, this approach may prove beneficial in peptide, protein‐fragment, or particle‐based targeting strategies, where the characteristics of small molecule labeling chemistry can significantly influence in vivo properties.[ ⁵⁸ , ⁵⁹ , ⁶⁰ ] Ultimately, the BAC‐based linker may facilitate the development of particularly well‐tolerated drug delivery constructs, and efforts towards this goal are currently underway.

Conflict of Interests

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Li X., Patel N. L., Kalen J., Schnermann M. J., Angew. Chem. Int. Ed. 2025, 64, e202417651. 10.1002/anie.202417651

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.