Stimuli‐Responsive Prodrug Linkers That Simultaneously Release Cargo and Neutralize In Situ Generated (Aza)Quinone Methides

Abstract

Self‐immolative linkers that use p‐amino/hydroxy‐benzyloxycarbonyl (PABC/PHBC) spacers are essential to the mechanism of many prodrugs. However, a highly reactive (aza)quinone methide is generated as a potential toxic byproduct. To remove the methide as it forms, we synthesized a series of novel tripartite prodrugs, comprising different triggers (nitro, amide, azide, boronate) and a PABC/PHBC‐type self‐immolative spacer with an integrated nucleophile (amine). Upon reductive, hydrolytic, or oxidative‐trigger activation, the release of the cargo is facilitated via a 1,6‐elimination that generates a reactive (aza)quinone methide. With the built‐in nucleophile, the (aza)quinone methide is rapidly self‐quenched to generate tetrahydroisoquinolines (THIQs). One of the selected THIQs does not exhibit an anti‐proliferative effect on the A431 mammalian tumor cell line. The new prodrug strategy has broad scope, enabling the use of a trigger that matches the targeted stimulus, while allowing for a diverse range of drug/cargo attachment. This proof‐of‐concept study adds a new linker strategy that quenches the electrophilic (aza)quinone methide generated in many self‐immolative linker systems and could find applications in prodrug and antibody‐drug conjugate strategies, or as a linker for probes in chemical biology.

Highly reactive (aza)quinone methides are electrophiles generated from many stimuli‐responsive prodrugs and probes. A new linker has been designed to quench the electrophilic methide, forming a tetrahydroisoquinoline (THIQ) heterocycle in high yields, even when a competing nucleophile is present. Upon neutralizing the electrophile, the THIQ itself is relatively nontoxic, boding well for future prodrug linker strategies

.

1. Introduction

The past several decades have seen significant progress in enhancing pharmacodynamic and pharmacokinetic properties of biologically active compounds through prodrug approaches.^{[ 1 ]} A typical tripartite self‐immolative prodrug design consists of a biologically active drug, self‐immolative linker/spacer, and a trigger group that responds to a specific stimulus. Among several self‐immolative linkers, the para‐aminobenzyloxycarbonyl (PABC) and para‐hydroxybenzyloxycarbonyl (PHBC) self‐immolative linkers have gained widespread popularity due to their versatility.^{[ 2 ]} The activation of a tripartite prodrug occurs in two‐steps. First, the trigger is activated by chemical,^{[ 3 ]} enzymatic,^{[ 4 ]} or light‐induced means;^{[ 5 ]} and second, the linker unit undergoes spontaneous chemical breakdown, leading to the release of the active drug at the target site, along with the spacer by‐product, an azaquinone methide or quinone methide. (Aza)quinone methides, as Michael acceptors, are polar and more reactive than their parent (aza)quinone due to the presence of an electrophilic methylene group.^{[ 6 ]} The versatility of these linkers is appealing for synthetic and medicinal chemistry applications and has been demonstrated to amplify oxidative stress in cancer cells by alkylation of glutathione (GSH).^{[ 7a,b ]} However, challenges associated with nonspecific trapping of the reactive (aza)quinone methide by other nucleophiles in noncancer prodrugs and chemical biology probes is an important area that needs more attention.^{[ 7c–e ]}

During prodrug activation, it is anticipated that the (aza)quinone methide reacts with water (a weak but abundant nucleophile) to generate an amino (or hydroxyl) benzyl alcohol. However, it is well‐documented that (aza)quinone methides are electrophilic species, often described as resonance‐stabilized carbocations due to their charge‐separated aromatic structure.^{[ 8 ]} Due to their high reactivity, (aza)quinone methides are usually not observed, reacting rapidly with various nucleophiles, including DNA, antioxidants (e.g., glutathione/GSH), water, and amino acids found in enzymes and proteins (e.g., cysteine).^{[ 8} , ^{9 ]} As an example, interaction with DNA can result in the formation of alkylated DNA adducts with potential to cause genetic mutations and contribute to carcinogenesis.^{[ 8} , ^{9 ]}

To mediate the electrophilicity associated with highly reactive azaquinone and quinone methides, we developed a novel, stimuli‐responsive prodrug linker that neutralizes the in situ generated quinone methide. We hypothesized that a tripartite PABC‐based (1,6‐self‐immolative) linker strategy, comprising a stimuli‐responsive trigger, a linker (spacer group) with nucleophilic handle, and a drug/probe, could activate drug and simultaneously quench the (aza)quinone methide. After activation and drug release, the generated methide may engage in one of the two possible reactions: (1) attack of the appended nucleophile (amine) via an intramolecular reaction (6‐endo‐trig cyclization) resulting in the formation of a stable tetrahydroisoquinoline (THIQ) (Figure 1, Path A), as favored by Baldwin's rules,^{[ 10 ]} or (2) an intermolecular reaction with nucleophiles present in the biological system such as water,^{[ 11 ]} GSH, DNA, or enzymes^{[ 12 ]} (Figure 1, Path B) resulting in the formation of a benzyl adduct (e.g., benzyl alcohol).

Figure 1.

Design of tripartite prodrug to neutralize the highly reactive (aza)quinone methide while generating a THIQ. Nu, nucleophile.

Herein, we report the synthesis of novel self‐immolative linkers and demonstrate prodrug activation using nitro‐, amide‐, azide‐ and boronate‐triggers, affording the corresponding THIQs. The data show that THIQ formation is favored, even in the presence of competing nucleophiles like GSH and N‐acetyl cysteine. In vitro anti‐proliferation assays, whereby an azide prodrug was activated by hydrogen sulfide (H₂S), indicate the linker/THIQ does not have significant toxicity against the A431 tumor cell line. The isolated THIQ also had no significant anti‐proliferative effect in the tumor cell line, suggesting the in situ THIQ cyclization could provide a new prodrug strategy when alkylation of biological nucleophiles is problematic.^{[ 7} , ⁸ , ⁹ , ^{12 ]}

2. Results and Discussion

2.1. Self‐Immolative Linker Design

The design of methide‐trapping self‐immolative prodrugs presented here offers several key advantages. The trigger can be tailored to include various functional groups (e.g., amide, azide, nitro, and boronate groups), making the system responsive to specific stimuli. After self‐immolation, the reactive azaquinone methide is rapidly quenched (de‐toxified) by an intramolecular nucleophilic 6‐endo‐trig‐1,6‐addition, resulting in stable amino/hydroxy‐THIQs (Figure 1) and increasing the scope of PABC/PHBC linkers in prodrug and chemical biology applications. Upon enzymatic hydrolysis (amide), reduction (nitro, azide), or oxidation (boronate), the linker releases the model phenol or amine drug. The methide is then trapped by a secondary benzylamine, generating a THIQ (Figure 2a). In this proof‐of‐concept study, eleven model prodrugs (with self‐immolative linkers) were synthesized, all of them containing a nucleophilic benzylamine handle (Figure 2b). The nucleophile was appended at the ortho‐position of the aromatic ring (relative to the generated methide), ensuring that the methide would be quenched via an intramolecular 6‐endo‐trig cyclization.

Figure 2.

a) Scope of tripartite self‐immolative prodrugs explored in this work. b) Chemical structures of the nitro, amide, azide, boronic ester‐functionalized self‐immolative prodrugs 1–11, (THIQs) 12/13 and 7‐hydroxycoumarin (coumarin) 14.

2.2. Synthesis of Reduction‐Triggered Nitro‐Prodrugs 1–6

Nitroaromatic compounds are versatile building blocks and play a crucial role as intermediates in synthetic organic chemistry.^{[ 13 ]} Despite the broad applications, nitro‐bearing drugs have been associated with significant toxicity issues, including carcinogenicity, hepatotoxicity, mutagenicity, and bone‐morrow suppression.^{[ 14 ]} In recent years, nitroaromatics have been widely used to mask active drugs as prodrugs that can be bioactivated through enzymatic or chemical reduction^{[ 15 ]} The nitro‐functionalized self‐immolative linker has gained considerable attention in this context,^{[ 16 ]} and under hypoxic conditions is reduced to the corresponding amine by nitroreductase to release the biologically active metabolite.^{[ 17 ]} Due to interest in prodrug activation strategies, nitro functionalized prodrugs 1–6 with N‐benzylamine as a representative nucleophilic handle were synthesized over 8–9 steps (Scheme 1a). The cargo was conjugated via a carbamate, carbonate, or ether linkage that, once released, would mimic a drug with a 1° amine (1, 2), 2° amine (3), alcohol (4, 5), and phenol (6), respectively.

The diester 16 was synthesized via an Ullmann coupling reaction^{[ 18 ]} between commercially available 2‐chloro‐4‐nitrobenzoic acid 15 and diethylmalonate in the presence of 60% sodium hydride and CuBr. The geminal diester 16 was then treated with NaOH in water and methanol, promoting ester hydrolysis and decarboxylation to yield the diacid 17. The nitro‐substituted cyclic anhydride 18 was obtained by heating the diacid 17 with acetyl chloride in acetone at reflux (60 °C) for 1 hour.^{[ 19 ]} From intermediate 18, the nucleophilic trap was installed, with benzylamine selected as the amine for this proof‐of‐concept study. Selective reduction of the amide and carboxylic acid groups in 19 was accomplished using a 1.0 M borane tetrahydrofuran complex (BH₃.THF) in a pressure tube, resulting in the secondary amine nucleophilic handle of 20. In prodrugs 1–5, a potentially acid‐sensitive carbamate or carbonate was required for model drug conjugation, thus an orthogonal protection/deprotection approach for the secondary benzylamine was required. A base‐labile Fmoc‐ protecting group was employed, with prodrugs 1–5 synthesized from Fmoc‐linker 21 in two steps (via 22). The coumarin prodrug 6 (drug linked via an ether) was synthesized in three steps (via 24) using a Boc‐protecting group for the secondary benzylamine 23. The cyclization of prodrugs 1/6 and 11 to THIQs 12 and 13, respectively, is shown in Scheme 2b and 2c (discussed later).

Scheme 2.

a) Mechanism for formation of (i) THIQ 12 and release of model drug coumarin 14 with possible side reaction products (ii) 29 and (iii) 40 also indicated. b) LC‐MS spectrum of THIQ 12. (R_t = 4.45 min) (Figure S41). The yields of 12 and 14 were determined by HPLC analysis (Figure S14, Table S2) using standard curves (Figures S11–12) and LC‐MS analysis for by‐products 29 (m/z 257.16 [M+H]⁺) and 40 (m/z 546.25 [M+H]⁺) (Figure S46).

2.3. Synthesis of Enzyme‐, Reduction‐ (H₂S), and Oxidation‐Triggered Prodrugs 7–11

In this study, the scope of the stimuli‐responsive trigger group was also investigated with three additional trigger groups; amide for enzymatic activation, azide for H₂S activation, and boronate for oxidative (H₂O₂) activation. The synthesis of these prodrugs is shown in S1–S3 of the supporting information.

Amide‐triggered prodrugs 7–8 (Figure 2) were synthesized from intermediates 23/21 (Scheme 1a) over 12 and 11 steps, respectively (Scheme S1), so they could be examined for enzymatic activation using penicillin G amidase (PGA)^{[ 20 ]} at neutral (pH 7.4). The azide group, a well‐known bioorthogonal reagent,^{[ 21 ]} has been reported as a triggerable moiety for prodrug and pro‐fluorophore activation via bioorthogonal click‐to‐release mechanisms^{[ 22 ]} and reduction by hydrogen sulfide (H₂S).^{[ 23 ]} Since the reduction of the azide group by H₂S is performed at or close to physiological pH (unlike the acidic zinc‐mediated reduction of nitro groups), it was selected as a trigger for in situ azaquinone methide and THIQ synthesis from azide‐functionalized prodrugs 9 and 10 (Figure 1). From intermediate 20 (Scheme 1a), the azide prodrugs 9 and 10 were prepared over five to six additional steps (Scheme S2).

Scheme 1.

a) Synthesis of nitro‐functionalized prodrugs (1–6). Synthesis of b) THIQ 12 and c) THIQ 13 from prodrugs 1, 6 and 11, respectively (Chromatographically isolated yields of THIQs shown).

The aryl boronate self‐immolative linker represents a significant advancement in targeted drug delivery and responsive materials.^{[ 24 ]} These linkers are designed to be activated by specific stimuli, such as reactive oxygen species (ROS) including hydrogen peroxide (H₂O₂).^{[ 25 ]} Their sensitivity toward ROS makes them particularly suitable for environments with elevated ROS levels, such as inflamed tissues, cancerous tumors, or infection sites.^{[ 9d ]} The boronic ester‐functionalized prodrug 11 (Figure 1) was synthesized via a nine‐step reaction sequence from 4‐bromo‐2‐methylbenzoic acid 35 and maintained the synthesis of a cyclic anhydride^{[ 26 ]} 36a as the point to which other nucleophiles could be installed (Scheme S3). Upon activation, cyclization of the quinone methide afforded phenolic‐THIQ 13 in an isolated yield of 51% (Scheme 1c).

2.4. Synthesis and Isolation of THIQ 12 via Methide‐Trap

To prove the concept of the methide‐trap, the synthesis and isolation of THIQ 12 was performed with nitroaromatic prodrugs 1 and 6 (Scheme 1b). Reduction of the nitro‐trigger was initiated with zinc‐acetic acid (Zn/AcOH) in MeCN. The reduction proceeded at room temperature in 10 minutes. The zinc was removed by filtration and the filtrate subjected to flash silica gel column chromatography. THIQ 12 was isolated in good yield (70% and 72%), confirming that self‐immolation and trapping of the azaquinone methide occurred, even under mildly acidic (AcOH) and room temperature conditions. Further supporting evidence of self‐immolation and azaquinone methide formation was provided in enzymatic and oxidative activation carried out under neutral pH (see later discussion). Two control experiments were also conducted to rule out direct S_N2 attack by the secondary amine. Prodrug 6 was reacted under identical conditions, but in the absence of either (1) acetic acid or (2) zinc. Conversion to the THIQ 12 and release of coumarin 14 was monitored by HPLC. In both control experiments the unreacted prodrug was observed, with no detection of THIQ 12 or coumarin 14, indicating that reduction by zinc and self‐immolation must occur first before cyclization via the 6‐endo‐trig intramolecular trapping of the methide.

Evidence confirming the structure of THIQ 12 was provided by comparing the ¹H NMR spectrum to prodrugs 1 and 6, and the potential by‐product aminobenzyl alcohol adduct 29 (Figure 3, synthesis in Scheme S2). The chemical shift of the benzylic protons (Figure 3, labeled A) were indicative of cyclization and formation of THIQ 12. For prodrugs 1 and 6, the benzylic protons were relatively deshielded by the adjacent oxygen atom (1, δ = 5.25 ppm and 6, δ = 5.33 ppm) compared to THIQ 12 (δ 3.52 ppm) where the equivalent protons are in a more shielded environment (adjacent nitrogen atom). For further comparison, the aminobenzyl alcohol adduct 29 (synthesized by reduction of nitrobenzyl alcohol 20, Scheme S2) displayed a benzylic proton peak at δ 4.49 ppm (labeled A, Figure 3). A full suite of 2D‐NMR experiments (COSY, HSQC, HMBC) (Figures S1–S5) on the isolated product further validated the cyclic structure of THIQ 12, with the key correlation of the benzylic protons (A) to the protons at positions B and D (Figure 3) observed. High‐resolution mass spectrometry (HRMS) analysis provided additional support for a molecule that matched the molecular weight of THIQ 12, with a peak at m/z 239.1526 assigned as the [M+H]⁺ ion.

Figure 3.

¹H NMR comparison of prodrug 1, 6, THIQ 12, and benzyl alcohol adduct 29.

2.5. Activation of Prodrugs Monitored by HPLC

Following successful synthesis, isolation, and characterization of THIQ 12 via prodrug 1 and 6, a reverse phase HPLC‐UV assay was employed to monitor the activation (Zn/AcOH‐mediated) of the six prodrugs (1‐6) at lower concentrations (200 µM) and partial aqueous solvent. To establish quenching of the reactive azaquinone methide at the relatively low micromolar concentrations expected in biological systems, prodrugs 1 and 6 (Scheme 2) were selected as exemplars for further activation‐release‐cyclization studies conducted in the presence of GSH, a thiol‐based competing nucleophile. The inclusion of a competing nucleophile like GSH, which has very high intracellular concentrations (0.1–15 mM),^{[ 27 ]} simulates a biological environment where multiple nucleophiles are present, and in this experiment would give an acyclic GSH adduct 40 if it outcompetes the nucleophilic trap. Coumarin 14 (pKa ∼ 7.8)^{[ 28 ]} was considered a good model cargo due to its distinct retention time on HPLC and low cytotoxicity, which is advantageous for measuring the in vitro cytotoxicity of the linker/THIQ (see below), and benzyl amine (pKa = 9.34)^{[ 29 ]} was selected as a representative cargo for amine‐containing drugs.

The nitro‐substituted prodrugs 1 and 6 (200 µM) were reduced with excess Zn/AcOH in 1:1 MeCN: PBS and subjected to quantitative HPLC analysis at different timepoints (2 minutes, 1, 3, and 6 hours). Analysis of the HPLC assay confirmed in situ formation of THIQ 12 and enabled quantification for the THIQ and release of model drug (coumarin 14 from prodrug 6). Prodrug 1 (R_t = 18.72 minutes), prodrug 6 (R_t = 20.04 minutes), THIQ 12 (R_t = 4.26 minutes) and coumarin 14 (R_t = 8.55 minutes) standards were all detectable on the HPLC trace (Figure S6) and their retention times used to identify and quantify THIQ 12 and coumarin 14 released (Figures S8 and S13). Further, evidence of THIQ 12 was provided by collection of the analytes from the HPLC instrument and mass spectrometry analysis, which for R_t = 4.26 minutes and R_t = 8.55 minutes, provided an [M+H]⁺ ion matching the expected structure for THIQ 12 and coumarin 14 (released drug), respectively (Figure S6). The release of benzyl amine from prodrug 1 could not be accurately assessed due to poor retention of benzyl amine to the HPLC stationary phase used in the assays (Figure S7). Efforts to improve the retention time of benzyl amine were unsuccessful.

In the absence of competing nucleophile, THIQ 12 formation from prodrugs 1 (Figure S8, Table S1) and 6 (Figure 4a, Figure S13, Table S1) was 85% and 97% (6 hours post‐activation), respectively. The release of coumarin 14 from prodrug 6 closely mirrored the THIQ formation (96%), indicating that as coumarin 14 was released, the methide was trapped by the nucleophile (2^ο amine) on the spacer group (Scheme 2a). In the presence of GSH (1 mM, fivefold excess), coumarin 14 release and cyclization to THIQ 12 was unhindered (Figure 4a, Table S2, Figures S9 and S14). To observe any trace amounts of H₂O adduct (benzylamine 29) or GSH adduct 40, LC‐MS analysis of the reduction was performed. Activation of prodrug 1 and 6 in the presence of 1 mM GSH, and up to 6 hours of monitoring the reaction progress with LC‐MS, revealed the primary product was THIQ 12 (Scheme 2b and Figures S43–S46), supporting the HPLC data. However, trace amounts of the benzyl alcohol adduct 29 and the GSH adduct 40 were detected upon LC‐MS analysis (Scheme 2a, Figures S45 and S46).

Figure 4.

Formation of THIQ 12 and release of coumarin 14 in the absence and presence of competing nucleophiles. a) Nitro prodrug 6, b) amide prodrug 7, c) azide prodrug 10, and d) boronic ester prodrug 11 (THIQ 13 formed). Yields for THIQs 12/13 and percentage release of coumarin 14 were determined by HPLC analysis (Figures S13–S14, S17–S18, S30–S31, S35–36 and S39–S40). Data shown as the mean ± SD (n = 3).

Next, release studies analyzed by HPLC‐UV for prodrugs 2–5 enabled the assessment of the self‐immolative linker's ability to carry other drug (leaving groups) and generate THIQ 12. The formation of THIQ 12 was higher (∼80‐85%) in carbamate‐linked prodrugs (2 and 3), whereas lower levels of THIQ were observed in carbonate‐linked prodrugs (∼61‐66%) (4 and 5) (Table S1). The reduced THIQ 12 formation in the carbonate‐linked prodrugs 4/5 is likely due to competing hydrolysis of the acid‐labile carbonate bond before self‐immolation and methide formation, which further confirms our proposed route to methide‐driven THIQ formation.

During the nitro‐reduction studies, a limitation in the Zn/AcOH reduction experiments was identified; the pH of the reaction mixture was more acidic (pH 4.25–4.70) than expected for in vivo activation (pH 6–7.4), potentially reducing the nucleophilicity of GSH (pKa = 8.7)^{[ 30 ]} and the 2^ο amine nucleophilic trap (pKa = 9.3)^{[ 29 ]} on the self‐immolative linker. Therefore, three alternate triggers (amide, azide, boronate (200 µM)) on coumarin‐based prodrugs 7–11 (Figure 1) were explored (Figure 4 [for prodrugs 6, 7, 10, 11], and Figures S17, S20, S27, S30, S35, S36 for HPLC traces of 7–11). Due to the low reactivity of GSH (pKa = 8.7), and to demonstrate the preference for the intramolecular trap over other nucleophiles, an alternate thiol, N‐acetyl cysteine (NAC, pKa = 9.5)^{[ 31 ]}, was selected for experiments with amide prodrug 7 (Figure 4b) and azide prodrug 10 (Figure 4c). The amide‐prodrugs 7 and 8 (Figure S17 and S20) were activated by the enzyme (PGA) at pH 7.4, azide prodrugs 9 (Figure S27) and 10 (Figure S30) by H₂S (sodium hydrosulfide (NaSH) salt as the donor for H₂S) at pH 6.96, and boronate 11 (Figures S35–S36) by hydrogen peroxide (H₂O₂) at pH 7.6. A competing nitrogen nucleophile was selected for our final experiments with the aryl boronate trigger prodrug 11 (Figure 4d and Figures S39,S40 for HPLC traces), to compare directly with the 2^ο amine of the nucleophilic trap.

Overall, the same trend seen for the acid‐catalyzed reductive activation of nitro‐trigger prodrug 6 (Figure 4a) was observed for the amide prodrug 7 (Figure 4b) under neutral, enzyme‐activated conditions (pH 7.4). Complete consumption of the prodrug 7 occurred by 3 hours with considerable quantities of coumarin 14 and THIQ 12 (∼60%) by 1 hour (Figure 4b, Figure S17). With and without competing nucleophile (NAC, 1 mM), released coumarin 14 was within ∼10% of THIQ 12 synthesized (Figure 4b, Figure S18). In the absence of NAC, the formation of THIQ 12 from prodrug 8 was 76% and in presence of 1 mM NAC, the yield remained similar after 24 hours (Figures S20,S21, Tables S3,S4). The control experiment of prodrugs 7 and 8 (no PGA enzyme), to ensure no nonspecific activation occurred, was performed under similar reaction conditions (PBS at 37 °C for 24 hours). The results of these experiments confirmed that activation and cyclization only occur in the presence of the PGA‐enzyme‐triggering reaction (Figure S22).

The H₂S‐mediated reduction of the azide group to the corresponding amine was performed using 2 mM NaSH at neutral conditions (pH = 6.96). Compared to previous prodrugs 6–8, THIQ 12 formation from azide‐prodrugs 9 and 10 (Figure 4c, Figures S27, S30, and Table S5) was slower, with 69% and 75% detected after 24 hours, respectively. For prodrug 10, a maximum of 87% of coumarin 14 was released, indicating self‐immolation had occurred and that approx. 12% of the linker was yet to generate THIQ 12. While not substantial, the 12% difference was not observed for the other three prodrugs in Figure 4, and could be attributed to undetected polar adducts, possibly from the large excess of H₂S (from NaSH) acting as a competing nucleophile to generate a benzyl thiol. Upon addition of competing nucleophile (NAC, 1 mM), THIQ 12 generation from prodrug 10 mirrored the nucleophile‐free experiments (Figure 4c, See Figure S31, Table S6 and Figure S28, Table S6 for prodrug 9 data), suggesting NAC did not act as a competing nucleophile or may form a reversible adduct which undergoes a retro‐Michael type 1,6‐elimination, equilibrating as the THIQ. Following the HPLC analysis, LC‐MS studies were conducted and supported the formation of THIQ 12 and the release of coumarin 14, both in the absence and presence of NAC (Figures S48–S51 [for prodrugs 7 and 8]) and Figures S53,54 [for prodrug 10]).

Finally, prodrug 11 was evaluated for activation with H₂O₂ ^{[ 25 ]} alongside an experiment whereby a competing nitrogen nucleophile, tryptamine was added (Figure 4d). Oxidation of the aryl boronate produced a phenol and boric acid (by‐product), thus THIQ 13 with a phenolic substituent was generated from the quinone methide. Within 3 minutes of activation, approx. 15% of the cyclized THIQ 13 and 14% of coumarin 14 was released. After 30 minutes, approx. 56% of coumarin 14 and 63% of THIQ 13 were generated and no further increases were observed over longer timepoints (Figure 4d), indicating that the reaction had reached its maximum release (Figures S35,S36, Table S7). In the presence of 1 mM of tryptamine, the maximum release observed was 56% for coumarin 14 and 62% for THIQ 13 (Figures S39,S40, Table S8). This result suggests that tryptamine did not affect the extent of THIQ 13 formation (Figure 4d).

2.6. In Vitro Prodrug Activation and Cytotoxicity Assessment

With evidence for THIQ cyclization confirmed, we next conducted in vitro activation studies for one of the prodrugs. The azide‐functionalized prodrug 10 and the control (nonquenching) azide‐functionalized coumarin prodrug 41, lacking the benzylamine sidechain (prepared using our previous synthesis;^{[ 32 ]} structure shown in the supporting information Section 7, Scheme S4), were selected due to the ease in which the azide could be activated with a source of H₂S (NaSH salt) using the epidermoid carcinoma cell line A431 (Figure 5, Scheme S4 and Figure S55). Also, coumarin 14 is known to display antiproliferative effects in this cell line,^{[ 33 ]} thus could be used to detect any additional toxicity imposed by the prodrug linker by‐products or THIQ 12. The A431 cells were treated with the azido‐prodrug 10 at concentrations of 40–300 µM and then activated using 2 mM NaSH (H₂S donor) to trigger the self‐immolative process (Scheme S4a), with NaSH (2 mM) added to the cell culture every 30 minutes over a period of 6.5 hours, to ensure a continuous presence of the activating agent which is volatile.^{[ 34 ]} Following this activation phase, the cells were incubated for 72 hours. As a control, the A431 cells were treated with 2 mM NaSH and incubated for 72 hours. There was no significant change in cell numbers compared to untreated control (Figure 5a), suggesting NaSH itself is nontoxic. Next, the cells were treated with coumarin 14, prodrug 10, and prodrug 41 with or without NaSH. The IC₅₀ for coumarin 14 (Figure 5b) was 93 µM ± 22, and when 2 mM NaSH was added, the IC₅₀ for coumarin 14 was 77 µM ± 10, showing NaSH does not significantly augment the effects of coumarin 14 itself (Figure S55). Without NaSH, no significant toxicity was observed for prodrug 10. However following activation with NaSH, the prodrug exhibited a concentration dependent toxicity with an IC₅₀ of 112 µM ± 26 (Figure 5b). The control prodrug 41 displayed similar toxicity following activation with NaSH, with an IC₅₀ of 118 µM ± 27. This was done to investigate whether the azaquinone methide intermediate, formed after the release of coumarin 14, might contribute additional toxicity through its reaction with cellular nucleophiles or water to form 4‐amino benzyl alcohol 42 (Scheme S4b) or other adducts. However, the IC₅₀ values for each prodrug following activation were found comparable to coumarin 14 (Figure 5b), supporting toxicity primarily associated with 14 and not azaquinone methide‐related byproducts like benzyl alcohols or THIQ 12.

Figure 5.

a) Viability of A431 cells incubated with NaSH. b) Viability of cells with coumarin 14, prodrug 10, prodrug 10 + 2 mM NaSH, and control prodrug 41 + 2 mM NaSH. c) Viability of cells with THIQ 12, benzyl alcohol adducts 29 and 4‐amino benzyl alcohol 42 for 72 hours. Data shown are the mean ± SEM, n = 3. Statistical significance conducted via two‐way ANOVA followed by Tukey's multiple comparison test. *, p < 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.

To investigate the activity of THIQ 12 and the linkers independently, the cells were treated with benzyl alcohol adducts 29 and 42 and THIQ 12 (Figure 5c). The cyclized THIQ 12 (Figure 5c) was nontoxic. Even at the highest tested concentration of 300 µM, cell survival exceeded 50%. While THIQs themselves could have biological activity, the lack of toxicity in the A431 cell line implies that THIQ 12 will limit additional off‐target cytotoxicity. Benzyl alcohol adduct 29 (Scheme S4a, Figure 5c) was nontoxic at lower concentrations, but increased toxicity was observed at 300 µM. Like THIQ 12, 4‐aminobenzyl alcohol 42 (Scheme S4b, Figure 5c) was nontoxic at all concentrations. This observation underscores the importance of careful design in prodrug systems to minimize the formation of unwanted reactive species. As evidenced by the concern of others in this field, the (aza)quinone methide generated toxicity^{[ 7} , ⁹ , ^{35 ]} could become more obvious as the prodrug is activated in vivo in the vicinity of multiple cellular nucleophiles.

3. Conclusion

A library of linkers with broad scope in drug delivery and chemical biology were synthesized and evaluated for their activation and self‐quenching capability. Essential to the linker design was a nucleophilic handle near to the leaving group (e.g., drug) of the self‐immolative linker. The nucleophile could then become a trap, acting as a sponge to mop‐up the highly reactive (aza)quinone methide. Activation and release of cargo (coumarin 14) promoted the rapid attack of the nucleophile on the electrophilic (aza)quinone methide, providing two novel (THIQs) 12 and 13 in high yield. The addition of a competitive nucleophile (thiol or amino‐containing molecules) did not affect the yield of THIQ. Promisingly, proof‐of‐concept cytotoxicity assays in the A431 tumor cell line showed that THIQ 12 itself was not toxic. Therefore, we envision the methide‐trapping strategy could act as a valuable tool for medicinal chemists and chemical biologist who need to prevent alkylation of essential biological nucleophiles in situ.

Supporting Information

The data supporting the findings of this study are available in the Supporting Infomation. The Supporting Information includes supporting figures/schemes, full experimental details for characterization of all new compounds (¹H NMR, ¹³C NMR for all compounds and 2D NMR spectra for prodrug 6), analytical studies, and biological assays (including statistical analysis). The authors have cited additional references within the Supporting Information.^{[ 36} , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ^{43 ]}

Conflict of Interests

The authors declare no conflict of interest.

Supporting information

Supporting Information

Data Availability Statement

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