Bioorthogonal retro-Cope elimination reaction of N,N-dialkylhydroxylamines and strained alkynes

Abstract

A bioorthogonal reaction between N,N-dialkylhydroxylamines and cyclooctynes is described. This reaction features a highly regioselective transformation between small, easily functionalizable reaction components with second order rate constants reaching 84 M⁻¹s⁻¹. The reaction is orthogonal to the inverse-electron demand Diels-Alder reactions between tetrazine and strained alkenes, and its components exhibit exquisite stability and chemoselectivity in cell lysate. This retro-Cope elimination reaction introduces a new member to the bioorthogonal reaction compendium outside the prolific class of cycloaddition reactions.

Graphical abstract

The growing compendium of bioorthogonal reactions has enabled the visualization, isolation, and manipulation of biomolecules in complex biological settings both in vitro and in vivo.^1-3 These reactions have been instrumental in the study of primary and secondary metabolites such as proteins,¹ sugars,^4-6 and lipids^7-9 as well as biomacromolecules¹⁰ whose modification by genetic means is neither practical nor possible. Demand continues to exist for additional bioorthogonal tools, particularly those that are more compact,^11-12 rapid,^13-15 stable,^16-17 regioselective,¹⁸ functionally diverse,^19-23 and orthogonal both to biology and to themselves.^24-25 The development of new reactions making simultaneous advances along not just one or two, but several of these axes, is an enticing yet elusive aspiration.^26-27

In search of bioorthogonal reactions, we deliberately explored transformations outside the well-studied class of cycloaddition reactions cognizant of the chemical properties that made these reactions effective. In particular, we contemplated whether other reactions might also engage with both carbons of a π-bond concertedly, circumventing the formation of a charged intermediate and the attendant need for significant polarization of the π-system that would render it susceptible to reaction by cellular nucleophiles. Modest asynchronicity in the transformation would, however, enable us to differentiate the two poles of the π-bond and deliver a regioselective transformation. The retro-Cope elimination reaction proved most serviceable (Figure 1).^28-30 In this manuscript, we describe the bioorthogonal reaction of N,N-dialkyhydroxylamines and cyclooctynes to form stable enamine N-oxide ligation products in a rapid and regioselective manner with reaction components comprising as few as three non-hydrogen atoms.

Figure 1.

Bioorthogonal retro-Cope elimination reaction between cyclooctynes and N,N-dialkylhydroxylamines.

We first evaluated the viability of the retro-Cope elimination reaction through density functional theory calculations³¹ and ascertained the activation barriers for the reaction of N,N-dimethylhydroxylamine with a variety of cyclooctynes (Figure 2A). Unmodified cyclooctyne itself proved promising with a calculated activation energy of 18.9 kcal/mol – sufficiently low for the reaction to proceed at room temperature. The absence of steric factors impinging on the incipient O··H··C2 bond in the transition state structure was equally noteworthy as it portended the importance that steric ambivalence toward propargylic substituents would have on the adaptability, mutual orthogonality, and reactivity of the cyclooctynes.

Figure 2.

Computational studies of the retro-Cope elimination reaction between cyclooctynes (COT) and N,N-dimethylhydroxylamine. Geometries were optimized at the M06-2X/6-31G(d,p) level of theory and single point energies were computed at the M06-2X/6-311G(2d,p) level of theory. (A) Computational reaction model to evaluate the reactivity of cyclooctynes. (B) Calculated transition state structure and activation energy for cyclooctyne hydroamination. (C) Additional ring strain of bicyclo[6.1.0]nonyne results in a lower activation barrier. (D) Calculated free energies of activation (ΔG^‡) as well as distortion (ΔE^‡_dist.) and interaction energies (ΔE^‡_int.) highlight the rapidity of the retro-Cope elimination reaction and the central role of hydroxylamine and alkyne distortion energies in lowering the activation barrier. R = p-NO₂Ph.

Calculations of bicyclo[6.1.0]nonyne also tempered our initial expectations that additional strain could be harnessed to the same effect as for cycloaddition reactions (Figure 2C).³² Instead, we focused on the electronic modulation^{4, 33} of cyclooctyne, which proved more profound (Figure 2D). Further distortion/interaction energy analysis indicated a sizable reduction in distortion energy^34-35 for the cyclooctynes versus their linear counterpart as expected, but unlike for cycloadditions, this reaction does not benefit from enhancement of interaction energy upon addition of electronegative substituents; the rate acceleration is driven by a decrease in the distortion energy of both components. The counterintuitive increase in the interaction energy is likely due to the reactant-ward shift of the transition state in accordance with Hammond’s postulate. Commensurate increases in the lengths of the C1··N and C2··H bonds further support this interpretation (see Supporting Information).

Kinetics experiments corroborated the reactivity trends predicted by computation. Using NMR spectroscopy to monitor reaction progress, we determined the second order rate constants for the reaction of N,N-diethylhydroxyl amine (1) with a panel of cyclooctynes 2–10 in d₃-acetonitrile at room temperature (Figure 3). Cyclooctyne (2) proved remarkably reactive, displaying a second order rate constant of 3.25×10⁻² M⁻¹s⁻¹ – an order of magnitude faster than its reaction with benzyl azide.³⁶ Further strain enhancements provided a 6.7-fold rate acceleration as predicted for bicyclo[6.1.0]nonyne 3, but far from the 100-fold increase observed for analogous azide-alkyne cycloadditions.³² Still, the second order rate constant of 2.17×10⁻¹ M⁻¹s⁻¹ compared favorably with the fastest azide-based reactions involving BARAC.¹³

Figure 3.

Kinetics studies for the hydroamination of cyclooctynes 2–10 by N,N-diethylhydroxylamine (1). Second-order rate constants were obtained using a 1:1 ratio of cyclooctyne to hydroxylamine in CD₃CN at rt. Measurements made in 50% H₂O/CD₃CN are in parentheses. The rate constant for difluorocyclooctyne 10 was derived from competition experiments with carbamate 9.

The hydroamination reaction of cyclooctynes was particularly sensitive to the inductive effects of propargylic substituents, and progressive rate enhancements were observed with increasing electronegativity (Figure 3). The most reactive of the cyclooctynol-derived substrates was carbamate 9 featuring a rate constant of 3.87 M⁻¹s⁻¹, a 120-fold improvement over that of cyclooctyne (2). Importantly, the minimalistic cyclooctyn-1-ol substructure proved versatile, being both easy to synthesize and derivatize; it was amenable to conjugation via an ester, carbamate, or a ketal linkage without incurring significant costs in size or reactivity. Indeed, elaborate functionalization of the core cyclooctyne was not only unnecessary, it at times proved deleterious. Reaction of N,N-diethylhydroxylamine with dibenzoazacyclooctyne 8 (DIBAC)³⁷ was rapid, yet still inferior in reaction rate to carbamate 9 and not appreciably superior to its more austere counterparts. Notably, the ligation product of the dibenzoazacyclooctyne 8 plus N,N-diethylhydroxylamine was prone to degradation, being uniquely unstable to purification by both standard and reverse phase flash chromatography.

We surmised from prior reports that difluorocyclooctyne 10 operates at the limits of bioorthogonality^{4, 38} and would provide a reasonable upper bound on hydroamination kinetics that can be achieved using electronically-tuned cyclooctynes in biological settings. Astoundingly, a competition experiment with cyclooctyne 9 revealed its rate constant to be 83.6 M⁻¹s⁻¹ in CD₃CN.

Notably, the retro-Cope elimination reaction is highly directed by substrate electronics and produced only a single observable regioisomer for cyclooctynes 2–7, 9, and 10. Accordingly, when a symmetrical N,N-dialkylhydroxylamine is employed, a single product is formed selectively.

To test the bioorthogonality of the hydroamination reaction, we performed in vitro protein labeling experiments. Fluorophore-conjugated hydroxylamine 13 was first assembled from 6-carboxytetramethylrhodamine and hydroxylamine 12, which was in turn synthesized by nucleophilic displacement of iodide 11 with N-methylhydroxylamine hydrochloride (Figure 4A). Separately, lysozyme was functionalized with cyclooctyne via N-hydroxysuccinimide ester 14 (Figure 4B). With both reaction components in hand, cyclooctyne-functionalized lysozyme 15 was treated with hydroxylamine 13 (0–200 μM) in PBS for 2 h and analyzed by in-gel fluorescence (Figure 4C). Labeling occurred in a concentration-dependent manner as expected, and labeling was saturated at 100 μM hydroxylamine. We also verified that the reaction occurs in a time-dependent manner (Figure 4D). Modified lysozyme 15 was treated with hydroxylamine 13 (200 μM) and quenched with N,N-diethylhydroxylamine (20 mM) at various time points. In-gel fluorescence analysis revealed signal saturation by 1 h. We further verified that the desired adducts were formed on the protein by mass spectrometry. Lysozyme 15 was incubated with hydroxylamine 13 (100 μM) in PBS, and the complete conversion of mono- and dicyclooctyne functionalized lysozymes 15 to mono- and dienamine N-oxides 16 was verified by ESI-MS (Figure 4E).

Figure 4.

Demonstration of protein labeling using the retro-Cope elimination reaction. (A) Synthetic route for fluorophore-hydroxylamine conjugate 13. (B) Lysozyme was modified using NHS-ester 14 to provide cyclooctyne-containing lysozyme 15. The modified protein lysozyme-COT 15 was labeled with fluorescent hydroxylamine 13. (C) In-gel fluorescence analysis of lysozyme-COT 15 (0.14 mg/mL) incubated with various concentrations of hydroxylamine 13 (10–200 μM) in PBS at room temperature for 2 h. (D) In-gel fluorescence analysis of lysozyme-COT 15 (0.14 mg/mL) incubated with hydroxylamine 13 (200 μM) for 1–120 min in PBS at room temperature. (E) Complete conjugation was observed via intact mass spectrometry of lysozyme-fluorophore conjugate 16 obtained by incubation of lysozyme-COT 15 (0.58 mg/mL) and hydroxylamine 13 (200 μM) in PBS at room temperature for 6 h.

We next verified the stability of both the enamine N-oxide and hydroxylamine species under a variety of biologically relevant conditions at various time points (Figure 5A, B). Hydroxylamine 13 was first incubated in PBS at room temperature, and HPLC analysis of the solution indicated that the compound was >86% intact for up to 8 h. Approximately 40% of the hydroxylamine had decomposed by the 24 h time point, however. The primary degradation products were consistent with hydrolysis of the regioisomeric nitrones likely generated by autoxidation. Consequently, we found that this degradation pathway could be abrogated by addition of cellular reductants such as ascorbic acid (5 mM) or glutathione (5 mM).³⁹ Negligible degradation was observed over 24 h. We similarly found that hydroxylamine 13 is stable in HEK293T cell lysate (1 mg/mL), experiencing no degradation above background over 24 h even when unbuffered with exogenous reductants.

Figure 5.

Demonstration of bioorthogonality. (A) Synthesis of enamine N-oxide 17. (B) The stability of hydroxylamine 13 and enamine N-oxide 17 was studied in PBS at pH 7.4 in the presence of glutathione (5 mM), cell lysate (1 mg/mL), microsomes (0.2 mg/mL), or without additives. The protective effect of sodium ascorbate (5 mM) was additionally evaluated for hydroxylamine 13. (C) In-gel fluorescence analysis of the reaction between hydroxylamine 13 (200 μM) and lysozyme-COT 15 in the presence of cell lysate (2.5 mg/mL) for 2 h showed exclusive labeling of lysozyme. (D) Live cell labeling with mutually orthogonal hydroamination and tetrazine-trans-cyclooctene reactions. HeLa cells transiently co-transfected with cell surface GFP-HaloTag and SNAP-Tag-BFP were treated with COT-chloroalkane (10 μM) and TCO-benzylguanine (10 μM) for 30 min, washed, treated with TAMRA-hydroxylamine (50 μM) and Cy5-tetrazine (1 μM) for 1 h at 37 °C, fixed, then imaged by confocal microscopy. Scale bar = 25 μm.

As with the hydroxylamine, the stability of enamine N-oxide 17 was evaluated by HPLC under biologically relevant conditions. Gratifyingly, it showed no evidence of degradation alone or in the presence of 5 mM glutathione in PBS at room temperature over the course of 24 h. Furthermore, while N-oxides do undergo reduction in a hemeprotein-dependent manner under hypoxic conditions, this process is sufficiently inhibited by aerobic conditions.⁴⁰ Incubation of enamine N-oxide 17 with human liver microsomes (0.2 mg/mL) under ambient air resulted in negligible degradation after 24 h.

Further demonstrating the bioorthogonality of the reaction, we combined TAMRA-hydroxylamine 13 and lysozyme-COT 15 in the presence and absence of HEK293T cell lysate in PBS for 2 h (Figure 5C). In-gel fluorescence shows that the lysozyme is labeled exclusively and that the degree of labeling is unperturbed by the presence of lysate. There appears to be no cross-reactivity between dialkylhydroxylamine and other proteins under these conditions.

Finally, we decided to explore the cross-compatibility of this reaction with other bioorthogonal systems to identify mutually orthogonal substrate combinations that could be used in tandem (Figures S6-S8).²⁵ We first evaluated whether tetrazines would be compatible with sterically congested cyclooctynes featuring tetrasubstitution at the propargylic position. Indeed, no product could be detected when cyclooctyne ketal 5 and tetrazine S18 were combined at 5 mM concentrations for 1 h in d₃-acetonitrile. Uncertain of whether steric constraints imposed by the fully substituted carbon⁴¹ or the electronics of the ketal were primarily responsible for inhibiting the inverse-electron demand cycloaddition, we evaluated electron-deficient cyclooctynes 9 and 10 under the same conditions to similar effect. Electronics, alone or in combination with sterics, can render the two reactions orthogonal. Similarly, we also evaluated whether our hydroxylamine reagents would be compatible with strained alkenes. 5 mM N,N-diethylhydroxylamine (1) combined with 5 mM cyclopropene S21⁴² or trans-cyclooctene S19⁴³ proved unreactive in d₃-acetonitrile. We also found that N,N-dialkylhydroxylamines do not react with aldehydes or engage in the copper-catalyzed azide-alkyne cycloaddition.⁴⁴

Dual labeling of cellular components using mutually orthogonal bioorthogonal reactions was performed on HeLa cells expressing cell surface GFP-HaloTag⁴⁵ or SNAP-Tag-BFP⁴⁶ and pre-treated with COT-chloroalkane 18 or TCO-benzylguanine 19 (Figures 5D, S9). Labeling by TAMRA-hydroxylamine 13 or Cy5-tetrazine 20 occurred if and only if cells expressed the corresponding fluorescent protein and were treated with the cognate reagent. Furthermore, intracellular labeling was accomplished in HeLa cells expressing cytosolic GFP-HaloTag using chloroalkane 18 and hydroxylamine 13 (Figure S10).

In this manuscript, we have identified a new bioorthogonal ligation reaction between N,N-dialkylhydroxylamines and cyclooctynes. The reaction features rapid kinetics with second order rate constants as high as 84 M⁻¹s⁻¹, exquisite regioselectivity, and small reaction components. The N,N-dialkylhydroxylamine reagent can be pared down to as few as three non-hydrogen atoms, and the cyclooctyne is likewise supremely effective even when unfunctionalized. Cyclooctynes can be attached conveniently at their propargylic positions without incurring costs to reactivity. We have also demonstrated that the hydroxylamine reagent and enamine N-oxide product are sufficiently stable under aqueous conditions in the presence of thiols or components of the cellular milieu found in the cell lysate, particularly on timescales that are germane to the ligation of small molecules to biomolecules. Both components, however, do have their sensitivities: hydroxylamines to air and enamine N-oxides to microsomes absent oxygen. We have identified factors that mitigate against these processes and ensure the bioorthogonality of the reaction. This retro-Cope elimination reaction adds an orthogonal member to the growing compendium of bioorthogonal reactions and demonstrates that there are yet untapped reaction modalities to be explored. Most importantly, this reaction paves the way for further expansion of our bioorthogonal arsenal’s functional repertoire.