Enamine N-Oxides: Synthesis and Application to Hypoxia-Responsive Prodrugs and Imaging Agents.

Tumor hypoxia induces the large-scale adaptive reprogramming of cancer cells, promoting their transformation into highly invasive and metastatic species that lead to highly negative prognoses for cancer patients. We describe the synthesis and application of a hypoxia-responsive trigger derived from previously inaccessible enamine N-oxide structures. Hypoxia-dependent reduction of this motif by hemeproteins results in the concomitant activation of a caged molecule and a latent electrophile. We exploit the former in a hypoxia-activated prodrug application using a caged staurosporine molecule as a proof-of-principle. We demonstrate the latter in in vivo tumor labeling applications with enamine-N-oxide-modified near-infrared probes. Hypoxia-activated prodrug development has long been complicated by the heterogeneity of tumor hypoxia in patients. The dual drug release and imaging modalities of the highly versatile enamine N-oxide motif present an attractive opportunity for theranostic development that can address the need not only for new therapeutics but paired methods for patient stratification.

Introduction

Tumor hypoxia describes a state of oxygen deficiency in tumor tissue arising from the inadequate and irregular vascularization of rapidly proliferating cancer cells.[1−4] Diffusion and perfusion limitations in these regions can lead to both persistent and fleeting levels of hypoxia that feature oxygen tensions less than 2% and, in the most severe cases of radiobiological hypoxia, below 0.1%.[5,6] Malnourishment and insufficient oxygenation of hypoxic tissue lead to large-scale adaptive reprogramming of cancer cells, transforming them into highly invasive and metastatic species with vastly altered metabolism and enhanced potential for proliferation and survival.[7−9] The onset of hypoxia often leads to highly negative prognoses for patients for whom few treatments exist. Radiotherapy is ineffective against hypoxic tissues given the essential role that oxygen plays as a radiosensitizer; chemotherapies, which target actively proliferating cells, are ineffective against cells in hypoxia-induced quiescence; and surgical options are often curtailed by the enhanced metastatic spread of cancers exhibiting hypoxia.[8,10−13] Hypoxia continues to present a challenge for the treatment of cancer.

Over the past six decades, dozens of hypoxia-activated prodrugs (HAPs) have undergone development, and 11 have entered the clinic.[14−18] Clinical success, however, has been elusive.[19] Most recently, HAPs tirapazamine[20] and TH-302[21] each failed to meet their primary endpoints in phase 3 clinical trials,[22,23] sparking a reassessment of the field’s approach to prodrug development.[16,19] In particular, these studies highlighted the imperative of patient stratification.[15,19] The presence, extent, and severity of hypoxia vary greatly among patients, yet no clinical factors, such as size or stage, or genomic markers sufficiently predictive of hypoxia have yet been identified.[16,19] Further establishing the critical importance of such prognostic factors, retrospective studies from the tirapazamine trial revealed that while efficacy was not established in the general population, among the subgroup of patients in whom tumor hypoxia was detected by [18F]-misonidazole (MISO)-based positron emission tomography (PET) imaging, a significant reduction in locoregional failure was observed in the treatment versus control cohorts.[24] Noninvasive hypoxia-responsive imaging agents are critical to the selection of patients likely to benefit from HAPs and essential for further prodrug development.

In this manuscript, we present a small, modular hypoxia-responsive trigger that is suitable for both imaging and prodrug applications, offering a complementary set of diagnostic and therapeutic tools that operate through parallel but divergent mechanisms. The presence of hypoxia is a necessary but insufficient determinant of patient response to HAP treatment. For instance, ample tumor expression of the requisite bioreductive enzymes responsible for prodrug activation is also obligatory. Hypoxia imaging agents, appropriately selected, have the potential to report not only on the existence of hypoxia but on additional factors relevant to the efficacy of the HAP itself.[25]

Clinically relevant hypoxia-activated prodrugs, with a notable exception (vide infra), have been designed around three key chemical motifs: nitroarenes,[21,26−28] quinones,[29−31] and pyrazine di-N-oxides.[20,32] Mechanistically, each engages in a futile redox cycle under normoxic conditions, undergoing continual 1e– reduction and oxidation by flavin-dependent oxidoreductases and molecular oxygen, respectively (Figure a).[15] Disruption of this balance in hypoxia results in a steady-state buildup of the reduced species, which decomposes into an active drug after one to three further 1e– reductions. Oxygen-independent 2e– reduction by flavoproteins acts to short-circuit this process and is responsible for the off-target toxicity of nitroarenes and quinones as well as the metabolic inactivation of pyrazine di-N-oxides.[5]

Figure 1

Mechanism and design of hypoxia-activated prodrugs. (a) Hypoxia-activated prodrugs commonly exploit a futile redox cycle to achieve selectivity. Oxygen continually reverses the reduction of the prodrug by 1e– reductases. (b) In contrast, two sequential 2e– reductions by hemeproteins convert aliphatic N-oxide prodrug AQ4N into the cytotoxic agent AQ4. (c) Design of new hypoxia-activated prodrugs termed enamine N-oxides. Enamine N-oxides can release small molecules upon 2e– bioreduction selectively under hypoxic conditions. The resulting unsaturated iminium ion can readily react with biological nucleophiles. (d) Retro-Cope elimination between alkynes and dialkylhydroxylamine provide access to novel enamine N-oxides under mild reaction conditions. LG = leaving group.

In the area of noninvasive hypoxia imaging, the nitroimidazole functional group has almost exclusively emerged as the principal hypoxia-responsive agent incorporated into PET radiopharmaceuticals.[33−35] [18F]-MISO,[36] the most well-studied of these compounds, has demonstrated good correlation with immunohistochemistry and patient prognosis in several cancers.[24,33,37,38] As with its cognate prodrugs, its hypoxia-induced activation mechanism involves four successive 1e– reductions, which ultimately produces an electrophile that covalently labels the hypoxic tissue. Notably, [18F]-MISO faces an important limitation that roughly 80% of activated probe fails to traverse the full sequence of reductions,[39] contributing to low tumor-to-background ratios and variable reproducibility.[40,41]

Challenges in translating the preclinical success of redox cycling prodrugs to the clinic have led us to explore an alternative mode of hypoxia activation based on the unique chemistry of amine N-oxides, as epitomized by the anthracenedione antineoplastic agent AQ4N (1, Figure b). AQ4N is a hypoxia-activated topoisomerase II inhibitor featuring two aliphatic amine N-oxides, which undergo two sets of 2e– reductions by hemeproteins in a single irreversible and oxygen-inhibited step to achieve hypoxia selectivity.[42−44] Importantly, hemeproteins responsible for AQ4N activation, such as iNOS[45] and CYP2S1,[46] are upregulated under hypoxic conditions by the hypoxia-inducible transcription factor HIF1 as are other AQ4N-reducing cytochrome P450s CYP2W1 and CYP3A4, which are generally upregulated in cancers.[15,47] Furthermore, the irreversibility of the AQ4N reduction is better suited to address perfusion-limited acute hypoxia than reversible redox cycling agents.[25,48]

AQ4N reduction to the diamine AQ4 (2) results in a cytotoxin that is protonated at physiological pH to afford a dicationic DNA intercalator.[43] Electrostatic interactions enhance the DNA-binding affinity of the anthracene core and enable it to persist until cell cycle resumption. Here, the N-oxide plays an intimate role in both hypoxia sensing and drug function. In contrast, a general hypoxia-responsive trigger would require the sensor and effector roles of the N-oxides to be fully divorced.

Prior applications of the N-oxide in Fe(II) sensing,[49] bioorthogonal reactions,[50] and the photoacoustic detection of hypoxia[51,52] have seen the integration of the N-oxide motif onto probes with redox-induced fluorescence/photoacoustic emission turn-on mechanisms. In this work, we design a new mode of N-oxide reduction-induced signal to output conversion suitable for prodrug and imaging applications.

Results and Discussion

Design of a Hypoxia-Responsive Chemical Motif with Drug Release and Labeling Properties

In designing an N-oxide-based hypoxia trigger, we posited that major structural changes immediately surrounding the N-oxide would adversely impact its ability to coordinate the heme in hemeproteins and impede its ability to act as a bona fide competitor of molecular oxygen. Instead, the N-oxide to N-lone pair transition could be relayed to a distal position through conjugation. Introduction of α,β-unsaturation on the amine N-oxide provided our candidate structure: the enamine N-oxide (Figure c).

A signal output mechanism was designed into the structure by embedding a leaving group at the allylic position. Enamine N-oxide reduction produces an enamine from which β-elimination would generate two functionally relevant species: (1) a leaving group and (2) an electrophilic α,β-unsaturated iminium ion. If a prodrug is desired, the allylic leaving group (LG) could be a caged drug, if a probe, a fluorophore, and if nothing, an inert halogen or chalcogen. The function of the electrophilic component would likewise be defined by the payload appended at the allylic position (R3). Affinity tags such as biotin or an alkyne, probes such as a fluorophore, or PET tracers such as an [18F] fluorine atom could be installed to suit the application. Both labeling and release potential are captured in the minimalist design of the enamine N-oxide.

Synthesis of Enamine N-Oxides

Initially, the synthesis of enamine N-oxides proved prohibitive, as reports of these structures were scarce, and their substrate scopes were limited. The chemical motif could be accessed by elimination of β-halogenated amine N-oxides[53,54] or by retro-Cope elimination of highly activated alkynes;[55] however, the former was impractical for use in prodrug applications of complex structure, and few instances of the latter existed, each of which involved the hydroamination of either strongly π-deficient Michael acceptors or an ynol ether. These products were unsuitable for accessing enamine N-oxides with the requisite γ-leaving groups.

Intermolecular retro-Cope elimination of unactivated alkynes is complicated by the propensity of enamine N-oxides to undergo Cope elimination[56,57] and [1,2]-Meissenheimer rearrangement[54,58,59] at mildly elevated temperatures. While activation has previously been accomplished exclusively through carbonyl and sulfonyl groups having strong mesomeric influence,[55] we wondered whether inductive effects would suffice (Figure d). Using p-fluorophenyl propargyl ether (3) as a model substrate, an initial temperature screen from room temperature to 80 °C in chloroform (CHCl3) demonstrated that the desired enamine N-oxide can be obtained, but there is a significant trade-off between conversion and degradation with a steep dropoff in yield within a tight ±10 °C window. Still, the maximum yield was 50% (Supporting Information, Table S1).

A solvent screen indicated that the hydroamination rates were fastest, but the products were the most prone to degradation in low polarity aprotic solvents (CH2Cl2, CHCl3, CCl4, DCE, PhMe), while reaction conversions were lower in polar protic ones (MeOH, EtOH, iPrOH, nBuOH) where fewer degradation products were observed (Tables S2–S4). Given the centrality of the N-oxide oxygen atom in both Cope (Supporting Information, Figure S1) and Meissenheimer degradation processes, we explored the role of solvent pKa and discovered that 2,2,2-trifluoroethanol (TFE) mitigates degradation better than the less acidic alcohols isopropanol and n-butanol likely through increased enamine N-oxide stabilization. Lower pKa solvents can, however, adversely affect the reaction rate presumably by inhibition of the hydroxylamine reagent through protonation as was observed for 1,1,1,3,3,3-hexafluoroisopropanol (HFIP).

Ultimately, we found that the best balance between reaction rate and product stability could be achieved by employing a low polarity solvent supplemented with a minimal quantity of a strong hydrogen-bond-donating solvent additive (entries 10 and 11, Table ). The hydroamination of alkyne 3 with N,N-diethylhydroxylamine in 20% TFE/CHCl3 (v/v) at 60 °C for 18 h provided the corresponding enamine N-oxides in 96% yield (entry 14, Tables and S5).

Table 1

Reaction Optimization for the Hydroamination Reaction between Alkynes and N,N-Dialkylhydroxylaminesa

					yield (%)b
entry	solvent	temp. (°C)	time (h)	conv. (%)	4	5 + 5′
1	CHCl3	50	6	65	5	50
2	CHCl3	60	6	84	23	48
3	CHCl3	70	6	100	49	31
4	CH2Cl2	60	6	100	46	43
5	DCE	60	6	89	45	20
6	iPrOH	60	6	59	3	47
7	TFE	60	6	49	2	47
8	iPrOH	60	12	77	26	51
9	TFE	60	12	81	4	76
10	50% TFE/CHCl3	60	6	55	<1	52
11	20% TFE/CHCl3	60	6	62	<1	61
12c	20% TFE/CHCl3	60	6	33	<1	28
13d	20% TFE/CHCl3	60	6	35	<1	26
14	20% TFE/CHCl3	60	18	95	<1	96

Conditions: alkyne 3 (0.2 mmol, 1 equiv, 0.2 M), N,N-diethylhydroxylamine (1 mmol, 5.0 equiv, 1 M).

Yields were determined by NMR using benzotrifluoride as an internal standard.

N,N-Diethylhydroxylamine (2 equiv) was used.

Concentration of 0.1 M. Temp. = temperature; conv. = conversion.

Substrate Scope of the Hydroamination Reaction

These reaction conditions provided access to enamine N-oxide products of propargylic ethers (5–9), alcohols (10, 15), esters (11), carbamates (12), carbonates (13), halides (16, 17), and acetals (18). Importantly, reactions were completed faster with greater regioselectivities in favor of the desired anti-Markovnikov product when more strongly electron-withdrawing substituents were present at the propargylic position (Figure ). We found that propargylic branching enhances the regioselectivity of the reaction and enables the hydroamination of secondary propargylic halide substrates (16, 17). Primary propargylic halides were not compatible due to competing SN2 displacement. Substrates featuring propargylic nitrogen, sulfur, or carbon substitutents (19–22) were insufficiently activated for the reaction, and their regioselectivities moderately or strongly favored the undesired Markovnikov product.

Figure 2

Hydroamination reaction between alkynes and N,N-dialkylhydroxylamines. (a) Alkyne substrate scope. (b) Hydroxylamine substrate scope. The major regioisomer of the product is depicted, and yields are reported as the average isolated yield from two experiments. Regioisomeric ratios (r.r.) represent the ratio of major to minor products as determined by 1H NMR analysis. When no r.r. is presented, only the depicted regioisomer is observed. Reactions were monitored by thin layer chromatography for disappearance of limiting reagent, and the reaction times are provided. alkyne (1 equiv), N,N-diethylhydroxylamine (5.0 equiv). N,N-dialkylhydroxylamine (1 equiv), alkyne (2.0 equiv).

Nonetheless, these are the first instances of branched enamine N-oxide products that have been reported. To access linear allylic nitrogen, sulfur, or primary halogen-substituted enamine N-oxides, one can instead access them via allylic phosphate 14. The diethyl phosphate superbly activates the alkyne for hydroamination while tempering its susceptibility to propargylic substitution. The product is suitable for further functionalization by nucleophilic displacement.

We also evaluated the substrate scope of the hydroxylamine component. Hydroamination of propargyl carbonates is agnostic to changes in the sterics of unbranched N,N-dialkylhydroxylamines (13, 23, 24) and likewise tolerates cyclic hydroxylamines such as N-hydroxymorpholine (31); however, the reaction with sterically encumbered substrates with α-branching (25, 26) poses a significant but not insurmountable challenge for the reaction. Finally, the compatibility of the reaction with a free hydroxyl (27), nitrile (28), azide (29), and amide (30) boded well both for the implementation of this reaction on complex substrates and for the fine-tuning of hypoxia selectivity in downstream applications (vide infra).

Validation of the Release and Labeling Functions of the Enamine N-Oxide

With enamine N-oxides readily accessible, we turned to their function, evaluating whether their reduction could induce the expulsion of a leaving group and formation of an electrophile. Treatment of enamine N-oxide 6 with either B2(OH)4 or ferrous sulfate as reducing agents in HEPES buffer at pH 7.4 indeed triggered the release of a p-cresol leaving group positioned at the γ-position (Figure S2a). In contrast, the corresponding aliphatic amine N-oxide S4 was unable to do the same when subjected to identical reducing conditions, confirming the essentiality of the enamine N-oxide unsaturation for drug release applications (Figure S2b). Additionally, reaction of B2(OH)4 with enamine N-oxide S6 in the presence of excess benzyl mercaptan generated Michael adduct S7 in 94% isolated yield, confirming that the reduction of enamine N-oxides produces an α,β-unsaturated electrophile with which thiols can react. Thiols do not react directly with the enamine N-oxides, as no reaction was observed for enamine N-oxide S6 in the presence of benzyl mercaptan alone (Figures S3 and S4).

Hypoxia-Dependent Reduction of Enamine N-Oxides by Hemeproteins in Microsomal Assays

We then set out to assess the hemeprotein-dependent reactivity of enamine N-oxides under hypoxic conditions.

Using probes featuring the release of a 2-nitroaniline chromophore, six enamine N-oxide probes 32a–f were assembled and each treated with human liver microsomes under anaerobic conditions.

A wide range of reactivities was observed among probes in this panel with a 35-fold difference in reactivity between the fastest and slowest substrates (Figures a and S5). Reactivities were similar between the three smallest substrates and decreased markedly with increasing steric congestion of the N-oxide substituents. These data are consistent with N-oxide reduction occurring in enzyme active sites and suggest that the overall rate and oxygen-dependent selectivity of these hypoxia triggers might be amenable to tuning through N-oxide structural manipulation.

Figure 3

Enamine N-oxides are bioreduced in an oxygen-dependent manner in vitro and in cells. (a) A panel of chromogenic enamine N-oxide probes that release 2-nitroaniline upon reduction are depicted. These probes were incubated with human liver microsomes under anaerobic conditions (0% pO2), and their initial rates of reduction were measured and are reported as relative rates of reduction (krel) normalized to probe 32a. (b) The time-dependent reduction of enamine N-oxide probe 32c by human liver microsomes under hypoxic (blue line) and normoxic (red line) conditions is reported. The data represent the concentration of 2-nitroaniline released based on its absorbance measurement at λ = 412 nm. There is a 21-fold enhancement in the initial rate of reduction under hypoxic conditions. (c) Oxygen, NADPH depletion, microsomal heat inactivation, as well as a panel of CYP450 inhibitors were evaluated for the ability to inhibit the reduction of enamine N-oxide probe 32c in our A412 microsomal assay using human liver microsomes. Standard conditions: 0% pO2, human liver microsomes (0.2 mg/mL), NADPH (1 mM), no inhibitors, 100 mM phosphate buffer, rt, 1 h. The x-axis label reflects deviation from standard conditions. Inhibitors were applied at 200 μM. (d) Synthesis of enamine-N-oxide-caged staurosporine 37 and nonreducible control compound 36. (e) Dose response curves of prodrug 37, staurosporine, and the nonreducible alkyne derivative 36 under both normoxic and hypoxic conditions in A431 cells. The pO2 of each condition is denoted in parentheses. (f) Comparison of the HCR values of compound 37 and AQ4N in H460 and A431 cells. TFA = trifluoroacetic acid; HCR = hypoxic-to-normoxic cytotoxicity ratio; NADPH = nicotinamide adenine dinucleotide phosphate; H.I. = heat-inactivated; DPI = diphenyleneiodonium chloride; TAO = troleandomycin; DDC = diethyldithiocarbamate.

We next took the most reactive of the chromogenic probes, probe 32c, and evaluated its oxygen-dependent reactivity. Incubation of this probe with human liver microsomes under both anaerobic (0% pO2) and aerobic (21% pO2) conditions in this assay revealed the significant impact hypoxia has on N-oxide reduction. A 21-fold higher initial rate of reduction was observed under the anaerobic conditions (Figure b).

We further found that when N-oxide probe 32c was incubated with human liver microsomes in the absence of NADPH or in the presence of an irreversible CYP450 reductase inhibitor diphenyleneiodonium chloride (DPI),[60]N-oxide reduction was nearly abolished. Heat-inactivated microsomes were similarly unable to activate the N-oxide probe appreciably. In these microsomal assays, enamine N-oxide reduction appears to be mediated by properly folded hemeproteins in an NADPH-dependent manner. Furthermore, a screen of common metal cations in the presence of probe 6 revealed that Fe2+ is unique among the evaluated cations in its ability to reduce enamine N-oxides (Figure S6). Incomplete ablation of reductase activity in heat-inactivated microsomes may partially result from the reduction of N-oxides by a free heme or ferrous ion. We also found that addition of different isoform-specific CYP450 inhibitors troleandomycin (TAO), diethyldithiocarbamate (DDC), quinidine, furafylline, and sulfaphenazole as well as a pan-CYP450 inhibitor ketoconazole individually had a more muted effect on N-oxide reduction than with a cocktail of these inhibitors.[61] This is suggestive of the polypharmacology of enamine N-oxides and is in line with what is observed for AQ4N.[62]

In Vitro Demonstration of Hypoxia-Dependent Drug Release in Prodrug Applications

Having confirmed that the enamine N-oxide can be reduced selectively under hypoxic conditions in vitro, we set out to evaluate the suitability of this motif for hypoxia-selective drug release in cells. As a proof-of-principle, proapoptotic pan-kinase inhibitor staurosporine (33)[63] was first modified with a propargyl carbamate on its γ-lactam and subsequently derivatized to N,N-diethyl enamine N-oxide 37 (Figure d). The staurosporine γ-lactam recognizes a similar set of residues as the adenine of ATP in the ATP-binding site of kinases,[64] and we determined that functionalization of this position would sufficiently decrease its activity. We evaluated the activity of staurosporine in multiple cancer cell lines including those of the skin, pancreas, lung, brain, and cervix and proceeded with further prodrug studies using the A431 epidermoid carcinoma and H460 lung carcinoma cell lines, which displayed the greatest intrinsic sensitivity to the parent drug with IC50s of 208 and 571 nM, respectively, in cell viability assays.

In these cell lines, we first confirmed that staurosporine itself does not display hypoxia-dependent activity and that our negative control, propargyl carbamate 36, which cannot be uncaged, exhibits hypoxia-independent activity and is >100-fold less potent than the parent drug. Importantly, the hypoxic-to-normoxic cytotoxicity ratios (HCRs), reported as the ratio between the cell viability IC50 values of cells treated with prodrug for 48 h under normoxic (20% pO2) and hypoxic (0.1% pO2) conditions, were 4.00 and 3.20 for A431 and H460 cells, respectively (Figures e and S7). We also observed that under hypoxic conditions, the IC50 of prodrug 37-treated cells better recapitulates that of staurosporine-treated cells when A431 rather than H460 cells are employed, potentially indicating a greater extent of prodrug activation in the former. Strikingly, AQ4N displays significant hypoxia selectivity in H460 cells with an HCR of 9.61 while displaying minimal selectivity in A431 cells (Figures f and S8).[65] While being inferior to AQ4N in the former, compound 37 compares favorably against AQ4N in the latter. These cell-line-dependent differences in hypoxia selectivity between AQ4N and prodrug 37 are not wholly unexpected given the differences in the mechanisms of action of the uncaged drugs. This finding raises the potential for the complementary use not only of AQ4N and compound 37 against different types of cancer but also of different enamine N-oxide prodrugs featuring distinct payloads.

To ensure that prodrug activation could be the source of the observed hypoxia selectivities, we monitored the amount of N,N-diethyl enamine-N-oxide-caged and -uncaged staurosporine present in the organic soluble extract of the cell and tissue culture media using HPLC analysis. Prodrug release was both oxygen- and cell-dependent with a 4-fold increase in the initial rate of staurosporine release under hypoxic conditions (Figure S9). A dose response assay of caspase 3/7 activity and a Western blot against apoptosis markers pro/p17-caspase 3 and PARP1 also recapitulated the findings of our cell viability assay (Figure S10).

Importantly, we emphasize that staurosporine merely serves as a proof-of-principle demonstrating that hypoxia-selective drug release can be accomplished using enamine N-oxides. The strength of the method resides in the ease with which drugs can be caged by this motif and the ease with which structural variations can be made for further prodrug optimization.

In Vitro and In Vivo Demonstration of Hypoxia-Dependent Cellular Labeling for Imaging Applications

We next sought to demonstrate the diagnostic potential of enamine N-oxides. Based on our previous observation that N-oxide reduction reveals an electrophilic α,β-unsaturated iminium ion, we anticipated that these reactive species could be used to covalently modify local proteins via their nucleophilic side chains. Accordingly, alkyne-functionalized probe compounds were designed for in-gel fluorescence visualization of labeled proteins after copper-catalyzed azide–alkyne cycloaddition (CuAAC)[66] with tetramethylrhodamine (TAMRA)-azide (Figure a).[67] We first synthesized a small panel of these probes (38–40, Figure b) that contained various leaving groups such as hydroxide, fluoride, and chloride ions. Satisfyingly, when A431 cells were treated with these probes at 10 μM for 48 h, a significant increase in protein labeling was observed in hypoxia (0.1% pO2) over the normoxia (20% pO2) background using each of the probes. Probe structure was consequential in the labeling, as N-oxide 39, containing the fluorine leaving group, had the highest hypoxic-to-normoxic labeling ratio at 10:1 (Figure c). Additionally, when probe 39 was used in a panel of cancer cell lines (Figure e), labeling was observed among all six cancer types tested with hypoxic-to-normoxic labeling ratios that ranged from 7.6:1 in U251 glioblastoma cells to 30:1 in HeLa cells indicating the broad applicability of the hypoxia-dependent labeling mechanism across cell types.

Figure 4

Hypoxia-specific bioreduction of enamine N-oxides leads to intracellular protein labeling in cells and in vivo. (a) Workflow for visualizing hypoxia-dependent cellular or tumor tissue slice labeling by alkyne-containing enamine N-oxide probes. Under hypoxic conditions, these probes are reduced and covalently modify proteins with an alkyne handle. Cell lysates or tumor tissue slices from probe-treated samples are labeled with a TAMRA-azide fluorophore via copper-catalyzed azide–alkyne cycloaddition (CuAAC) click chemistry. Lysates are visualized by in-gel fluorescence, and tissue slices are visualized by fluorescence microscopy. (b) Structures of alkyne-containing enamine N-oxide imaging probes and pimonidazole. (c) A431 epidermoid carcinoma cells were treated with probes 38–40 for 48 h and visualized by in-gel fluorescence after CuAAC with a TAMRA-azide fluorophore. (d) Enamine N-oxide probe 39 is activated at ∼1% pO2 over 48 h and shows oxygen-dependent labeling in cell culture using a BxPC-3 pancreatic cancer cell line. (e) The labeling profile of probe 39 shows hypoxia selectivity in a series of cancer cell lines. (f) Tumor tissue slices were obtained from BxPC-3 xenografts in mice that were intraperitoneally inoculated with enamine N-oxide 39 and pimonidazole. Localization of compound 39 (red) shows strong correlation with the staining patterns from immunofluorescent labeling of known hypoxia markers GLUT1, HIF1α, CAIX, and pimonidazole (green). Scale bar represents 100 μm. H/N = hypoxic-to-normoxic ratio; Nec = region of necrosis; TAMRA = tetramethylrhodamine; DAPI = 4′,6-diamidino-2-phenylindole.

In like manner, we looked at cellular labeling at a series of oxygen concentrations in one of the cell lines. Hypoxia can range from 2% pO2 to complete anoxia, and a diagnostic agent with a dynamic range suitable for detecting both mild and severe hypoxia can be impactful. BxPC-3 pancreatic adenocarcinoma cells treated for 48 h with probe 39 displayed a clear oxygen-dose-dependent effect on cell labeling (Figures d and S11). No significant differences in labeling were detected between hypoxic and normoxic conditions until 1% pO2 was reached. At this level of hypoxia, the enamine N-oxide probe reached a hypoxia-to-normoxia labeling ratio of 4.5:1, which is nearly half the maximal ratio of 12:1 obtained at the radiobiological hypoxia level of 0.1% pO2. In contrast, labeling by the 2-nitroimidazole-based tumor hypoxia marker pimonidazole (41) around oxygen tensions of 1% pO2 is minimal, and its half-maximal labeling is not reached until oxygen levels are further depressed (Figure S12).

We then decided to test the capacity of this technology to label regions of tumor hypoxia in an in vivo setting. We chose to examine a human tumor xenograft model in mice derived from the BxPC-3 pancreatic cancer cell line based on the superiority of its hypoxia response among the panel of cell lines evaluated in our cellular labeling studies described above as well as the reported propensity for extensive regions of hypoxia to develop in pancreatic tumors.[68,69] Mice bearing xenografts were intraperitoneally injected with a cocktail containing both probe 39 and pimonidazole, and then, their tumor tissue was analyzed by immunofluorescence. We visualized enamine N-oxide probe localization with TAMRA-azide using CuAAC and examined its colocalization with the well-studied hypoxia markers glucose transporter 1 (GLUT1), carbonic anhydrase IX (CAIX), and HIF1α, along with pimonidazole (Figures f, S15, and S16).[35] We found the probe to colocalize well with both pimonidazole and GLUT1 surrounding an area of necrosis within the tumor (Figures f and S13). Tumoral necrosis is expected under hypoxic conditions as oxygen deprivation reaches levels that induce cell death. HIF1α and CAIX expression did not overlap but surrounded the area of probe localization. This is a common staining pattern in hypoxia as HIF1α, CAIX, and pimonidazole are induced by different levels of hypoxia. Furthermore, CAIX is a secreted protein that can also be induced by factors other than HIF1α.[70,71] We subsequently examined the staining pattern of the probe relative to the perfusion marker Hoechst 33342 (Figure S13) and endothelial marker CD31 (Figure S14). We observed probe and hypoxia marker localization away from these areas of vascularization, supporting our hypothesis that we are labeling an area of low oxygen content.[4]

In Vivo Imaging of Tumors in Live Mice with a Near-Infrared Hypoxia-Activatable Imaging Agent

To demonstrate the feasibility of using enamine N-oxides in hypoxia imaging applications in vivo, we performed near-infrared (NIR) fluorescence imaging of human tumor xenograft models in mice (Figure ). Encouraged by the effective tumor labeling of fluorinated probe 39, we designed enamine N-oxide probe 42 (Figure a), which is conjugated to a Si-rhodamine fluorophore.[72] BxPC-3-xenografted mice were intraperitoneally injected with NIR probe 42 (20 mg/kg) and imaged at 6, 24, and 30 h postinjection (Figures b and S18). Distinct fluorescent signals associated with the subcutaneously implanted BxPC-3 tumors were observed at the first time point of 6 h together with strong signals in the kidneys. The nontumor associated signals subsided over the course of the next 24 h consistent with renal clearance of the probe. Clear tumor localization with low background fluorescence was observed at both the 24 and 30 h time points. No signal was observed in the corresponding section of the opposing flank as expected. Further analysis by immunofluorescent staining of tumor tissue showed significant colocalization of probe 42 with pimonidazole confirming the presence of hypoxia in each of the labeled tumors (Figure S17). Importantly, the NIR probe was installed on the imaging agent at a nonprivileged position on the molecule, which points to the feasibility of integrating other imaging modalities such as [18F]-PET tracers or MRI contrast agents. Overall, these in vivo data suggest that enamine-N-oxide-derived probes could prove useful as diagnostic imaging agents for the noninvasive detection of hypoxic tumors.

Figure 5

Hypoxia-responsive bioreduction of enamine N-oxides enables near-infrared (NIR) fluorescence imaging of tumors in vivo. (a) Structure of alkyne-containing enamine N-oxide NIR probe 42. (b) Near-infrared imaging of a BxPC-3 xenograft mouse model showed preferential accumulation of enamine N-oxide probe 42 in the tumor. hpi = hours postinjection.

Conclusions

We have described a general method for the synthesis of enamine N-oxides. This chemical motif undergoes hypoxia-selective and hemeprotein-dependent reduction to induce the concomitant activation of a small molecule and a latent electrophile, each of which can be leveraged for hypoxia-responsive prodrug and imaging applications, respectively. In studies demonstrating the use of this motif in prodrugs, we found that enamine-N-oxide-caged cytotoxin staurosporine displays hypoxic-to-normoxic cytotoxicity ratios that compare favorably with and are complementary to those of AQ4N, a well-investigated aliphatic amine N-oxide hypoxia-activated prodrug from which the current structure was inspired. Importantly, we also confirmed the dual function of the enamine N-oxide, demonstrating both in cells and in in vivo tumor xenograft mouse models that this motif can be used to selectively label hypoxic tumor tissue. Highlighting the modularity and generality of this hypoxia trigger, the enamine N-oxide was conjugated to a near-infrared probe and used to image hypoxic tumors in mice.

Tumor hypoxia is highly correlated with low survival rates and negative prognoses for cancer patients with advanced solid tumors, and therapeutic agents targeting tumor hypoxia are urgently needed. Given the significant heterogeneity in hypoxia that can develop between tumors of the same type and across patient populations, the clinical exploitation of hypoxia will require codevelopment of therapeutics with companion diagnostic agents. We envision that the enamine N-oxide scaffold described herein could serve as a starting point for the development of hypoxia-selective theranostic agents that are useful for both the identification and treatment of cancer patients with advanced solid tumors.