Conformationally Strained trans-Cyclooctene with Improved Stability and Excellent Reactivity in Tetrazine Ligation

Abstract

Computation has guided the design of conformationally-strained dioxolane-fused trans-cyclooctene (d-TCO) derivatives that display excellent reactivity in the tetrazine ligation. A water soluble derivative of 3,6-dipyridyl-s-tetrazine reacts with d-TCO with a second order rate k₂ 366,000 (+/− 15,000) M⁻¹s⁻¹ at 25 °C in pure water. Furthermore, d-TCO derivatives can be prepared easily, are accessed through diastereoselective synthesis, and are typically crystalline bench-stable solids that are stable in aqueous solution, blood serum, or in the presence of thiols in buffered solution. GFP with a genetically encoded tetrazine-containing amino acid was site-specifically labelled in vivo by a d-TCO derivative. The fastest bioorthogonal reaction reported to date [k₂ 3,300,000 (+/− 40,000) M⁻¹s⁻¹ in H₂O at 25 °C] is described herein with a cyclopropane-fused trans-cyclooctene. d-TCO derivatives display rates within an order of magnitude of these fastest trans-cyclooctene reagents, and also display enhanced stability and aqueous solubility.

Introduction

The use of strain energy to enable molecular reactivity is a venerable concept in the field of organic synthesis, and has in recent years become an increasingly important tool for enabling bioorthogonal reactions—those transformations that retain selectivity in a biological context.^1-12 In 2004, Bertozzi and co-workers first described the use of strain to accelerate the rate of the bioorthogonal cycloaddition between alkynes and azides¹³— a transformation that otherwise requires transition metal catalysis.^14-16 Subsequently, strained molecules have been employed as reactive cycloaddition partners across a range of applications where efficient bioorthogonal labeling is required.^{1, 8, 9, 17-21}

In 2008, our group developed a photochemical flow-method for producing functionalized trans-cyclooctenes (TCO’s),²² and subsequently demonstrated that tetrazines combine with trans-cyclooctenes with unusually fast rates (k₂ >10³ M⁻¹s⁻¹) in bioorthogonal reactions.²³ Contemporaneous with our work on TCO, bioorthogonal reactions of tetrazines with norbornenes²⁴ and the Reppe Anhydride²⁵ were described. More recently, cyclopropenes,^26-31 cyclooctynes^32-35 and simple terminal alkenes³⁶ have been used as dienophiles in bioorthogonal reactions with tetrazines. Each of these dienophiles offers unique advantages. Still, there remains a high level of interest in trans-cyclooctene dienophiles due to their high reactivity, and as such trans-cyclooctenes have been used in a range of applications including cellular imaging and nuclear medicine.^37-42

While simple trans-cyclooctene derivatives combine with s-tetrazines with very fast rates, several situations have arisen where even faster rates are necessary. Reaction rate is always a premium consideration for applications to nuclear medicine due to the low concentrations and short half-lives intrinsic to radiochemistry.^{40, 43-51} Fast bioorthogonal reactivity is especially important for pretargeted imaging and radioimmunotherapy where the bioconjugation event takes place in vivo.^43-48 The development of more highly reactive TCO probes has also proven necessary for applications to intracellular labeling,^{35, 52-57} as tetrazines that display the best long term in vivo stability also tend to be less reactive in Diels-Alder reactions. For example, it was possible to genetically encode an electron-rich tetrazine-containing amino acid into proteins such as GFP (2, Scheme 1c); however subsequent labelling by 5-hydroxy-trans-cyclooctene was limited by sluggish reactivity.⁵⁸

Scheme 1.

(a) Conformation analysis of ground states show that the half- chair conformation of TCO is higher in energy by 5.6–5.9 kcal/mol relative to the lowest energy crown conformation. (b) As predicted computationally, the conformationally constrained s-TCO (1b) combines with 3,6-diphenyl-s-tetrazine in with a rate that is 160 times faster than trans-cyclooctene itself. (c) For GFP with a genetically encoded tetrazine-containing amino acid, s-TCO enables rapid fluorogenic labelling inside living bacteria.

Computation has emerged as an important tool for understanding and predicting the reactivity of bioorthogonal reaction partners.^{29, 34, 59-63} In parallel with studies by van Delft on the strained cycloalkyne, bicyclo[6.1.0]non-4-yn-9-ylmethanol,¹⁸ we described how the reactivity of TCO could be increased through the introduction of conformational strain. (Scheme 1a).⁶² Ab initio calculations predicted the lowest energy ‘crown’ conformation (1a) of trans-cyclooctene to be 5.6–5.9 kcal/mol lower in energy than the ‘half-chair’ conformation (1b). Recognizing that trans-cyclooctenes with cis-ring fusions would be forced to adopt strained conformations similar to 1b (Scheme 1b), ^{60, 62} computation was used to design a strained trans-cyclooctene (termed “s-TCO”) with a cis-fused cyclopropane ring that enforces a highly strained half-chair conformation for the eight-membered ring. Transition state calculations with 3,6-diphenyl-s-tetrazine predicted that s-TCO would react much faster than the crown conformation of TCO (1a) (ΔΔG^‡‡ = 3.34 kcal/mol at 25 °C). Experimentally, s-TCO combined with 3,6-diphenyl-s-tetrazine in MeOH at 25 °C with a rate of 3100 M⁻¹s⁻¹— 160 times faster than trans-cyclooctene (Scheme 1b). The experimentally measured ΔΔG^‡ (3.0 kcal/mol) correlated closely with this predicted value. With more reactive tetrazines under aqueous conditions, rates can exceed 10⁶ M⁻¹s⁻¹, making s-TCO the most reactive dienophile reported to date for bioorthogonal applications.^{44, 62, 64} With GFP derivative 2, fluorogenic labelling by s-TCO takes place rapidly inside living bacteria (Scheme 1c).⁵⁸

While the reactivity of s-TCO in tetrazine ligation is desirable, there are some limitations that also emerge from its high reactivity. As we have discussed previously, s-TCO was found to isomerize in the presence of high thiol concentrations (30 mM), presumably via a free radical mediated pathway.⁶² Furthermore, many s-TCO derivatives tend to be only moderately stable to prolonged storage, and must be kept in cold solution to avoid polymerization and isomerization.⁶⁵ Additionally, s-TCO is hydrophobic, and its synthesis is not stereoselective, necessitating chromatographic separation of syn- and anti- product diastereomers.

For the reasons outlined above, we considered the design of new conformationally strained trans-cyclooctene derivatives that would maintain high reactivity toward s-tetrazines but display enhanced stability and be more easily prepared. To this end, we considered cis-dioxolane-fused trans-cyclooctenes (d-TCO) derivatives of structure 3 (Schemes 2 and 3). It was expected that the cis-fused ring would enforce a strained half-chair conformation of the cyclooctene, imparting high reactivity toward tetrazines. However, it was also reasoned that the alkene of d-TCO would be less electron rich (and hence more stable) than that of s-TCO as a result of the inductively electron withdrawing oxygens. The hope for d-TCO was that a modest compromise in reactivity toward tetrazines would be compensated by enhanced stability. We also projected that the dioxolane functionality would impart improved aqueous solubility for d-TCO.

Scheme 2.

Transition state calculations (level) for 3,6-diphenyl-s-tetrazine with (a) trans-cyclooctene and (b) dioxolane-fused trans-cyclooctene 3a. The cis-ring fusion of 3a confines the 8-membered ring to a strained half-chair conformation, resulting in a significantly lower transition state barrier for the cycloaddition.

Scheme 3.

Synthesis of d-TCO derivatives

Our studies began with a computational comparison of the reactions of 3,6-diphenyl-s-tetrazine with the crown conformation of trans-cyclooctene (1a) and that of d-TCO derivative 3a, which is locked in the half-chair conformation. Transition state calculations in the gas phase between 1a and 3,6-diphenyl-s-tetrazine at the M06L/6(311)+G(d,p) level were described previously, and shown to proceed with a barrier of ΔE^‡ = 13.31 kcal/mol and ΔG^‡ = 16.09 kcal/mol (Scheme 2a). By comparison, the computed reaction of 3,6-diphenyl-s-tetrazine and d-TCO 3a proceeds with a significantly lower barrier (ΔE^‡ = 10.82 kcal/mol and ΔG^‡ = 13.27 kcal/mol, Scheme 2b). These calculations predict that the reactions of d-TCO would be considerably faster than that of the parent TCO 1.

Encouraged by this computational prediction, we prepared derivatives of 3a and 3b as shown in Scheme 3. The diol 4 can be prepared in multigram quantity from 1,5-cyclooctadiene by Upjohn dihydroxylation.⁶⁶ Diol 4 combined with glycolaldehyde dimer at room temperature to give 5 as a 12:1 mixture in favor of the syn-diastereomer. When dioxolane formation was conducted at 80 °C, both syn- and anti- diasteromers of 5 were produced in a 1:1 ratio. This kinetic preference for the syn-diastereomer is known for cyclic acetals,^67-69 and has been ascribed to the relief of 1,3 allylic strain⁷⁰ in the cyclization of the oxonium intermediate. Photoisomerization using the closed-loop flow reactor^{22, 71} developed in our labs gave d-TCO 3a in 59% yield. Likewise, acetal 6 was prepared from adipic semialdehyde methyl ester, and photoisomerized to give d-TCO 3b in 36% yield, predominantly as the syn-diastereomer (>50:1). The brief and diastereoselective nature of the syntheses of d-TCO derivatives greatly facilitates material throughput, as illustrated by a gram-scale preparation of d-TCO 3a. A conjugatable derivative of 3a could be prepared by treatment with p-nitrophenylchloroformate to give the activated carbonate 7 in 56% yield, and can be readily conjugated to amines under standard conditions using Hunig’s base, as demonstrated by the synthesis of biotin-mini-PEG derivative 8 (Scheme 3).

Rate constants for the reaction of d-TCO derivatives with tetrazine derivatives were measured under several conditions (Scheme 4). Diphenyl-s-tetrazine with d-TCO 3b in MeOH (25 °C) reacts with a rate of 520 (+/− 3) M⁻¹s⁻¹. The same tetrazine was found to react with the parent trans-cyclooctene with a rate of 19.1 (+/− 1) M⁻¹s⁻¹.⁶² The observed 27-fold rate enhancement (ΔΔG^‡ = 2.0 kcal/mol) for d-TCO is in good alignment with the computationally predicted value (ΔΔG^‡ = 2.8 kcal/mol; (ΔΔE^‡ = 2.5 kcal/mol) described above. To ensure that the rate acceleration for d-TCO was due to the ring fusion, and not due to the electronic influence of the hydroxyl groups, we compared the rate of diphenyl-s-tetrazine with diol 9, which was prepared by photoisomerizing diol 4. The observed rate constant k₂ = 41 (+/− 8.1 × 10⁻²) M⁻¹s⁻¹ was significantly slower than the rate of the d-TCO 3b. This observation provides strong evidence that the conformational strain of the 8-membered ring is indeed responsible for the observed rate acceleration.

Scheme 4.

Comparison of second order rate constants

Under aqueous conditions, reactions are significantly faster due to acceleration by the hydrophobic effect, and as such a stopped-flow technique was used to measure rate constants. Previously, a rate of k₂ 5235 (+/− 258) M⁻¹s⁻¹ was measured for the equatorial diastereomer of 5-hydroxy-trans-cyclooctene (11) with tetrazine 10 at 25 °C in 45:55 H₂O:MeOH.³⁵ Under similar conditions, d-TCO 3a reacts with tetrazine 10 with a rate k₂ 167,000 (+/− 7000) M⁻¹s⁻¹. As expected, the fastest rates are measured under purely aqueous conditions. Improved aqueous solubility of the d-TCO derivatives made kinetic measurements in aqueous conditions possible. At 25 °C in water, the syn-diastereomer of d-TCO (syn-3a) combines with the freely water soluble derivative 12 with a rate k₂ 366,000 (+/− 15,000) M⁻¹s⁻¹. The anti-diastereomer (anti-3a) combines with a rate k₂ 318,000 (+/− 2900) M⁻¹s⁻¹. In a recent study, Robillard reported that the axial diastereomer of 5-hydroxy-trans-cyclooctene is more reactive than the equatorial diastereomer.⁴⁴ Under similar conditions, d-TCO derivatives 3a and 3b combine with 12 with a rate that is significantly faster than either diastereomer of 5-hydroxy-trans-cyclooctene: the axial diastereomer 13 reacts with k₂ 80,200 (+/− 200) M⁻¹s⁻¹ , and the equatorial diastereomer 11 reacts with k₂ 22,600 (+/− 40) M⁻¹s⁻¹. As expected, faster rates were observed with s-TCO derivatives. Water soluble s-TCO 14 reacts with 12 at a rate of k₂ 3,300,000 (+/− 40,000) M⁻¹s⁻¹, which to the best of our knowledge, makes this the fastest conjugation rate measurement to date.

While d-TCO derivatives are extremely reactive toward tetrazines, they also display excellent stability. While s-TCO derivatives can be handled as neat materials, for storage we find it is necessary to keep most⁷² s-TCO compounds in solution at freezer temperatures. By contrast, the stability of d-TCO derivatives are greatly improved. Both 3a and 3b are crystalline solids that can be stored uneventfully on the bench and are soluble (logP = 0.94 for syn-3a) and stable in aqueous solutions. As a safeguard for long-term storage, these compounds are kept as solids in the freezer (−20 °C), where they are stable for at least 14 months. In phosphate buffered D₂O (pD 7.4), no isomerization or decomposition of syn-3a (20 mM) was noted after 14 days. In concentrated solution (0.5 M in CD₃OD), syn-3a and anti-3a isomerizes very slowly at room temperature, with 5% isomerization after three days. For an 5 mM solution of syn-3a in human serum at room temperature, no isomerization or decomposition was observed after 24 hours, and the compound was >97% trans after 4 days. In the presence of rabbit reticulocyte lysate (10%) in buffered D₂O (pD 7.4), compound syn-3a isomerized slowly with 12%, 22% and 41% isomerization after 10, 24 and 96 hours, respectively. The isomerization of syn-3a in the presence of reticulocyte lysate is possibly a thiol-radical promoted process, as both glutathione and mercaptoethanol were found to promote d-TCO isomerization. Thus, mercaptoethanol (30 mM) promoted isomerization of syn-3a slowly at pH 6.8 (44% isomerization after 48 hours), and more rapidly at pH 7.4 (43% isomerization after 5 hours). Previously, we noted that s-TCO (Scheme 1b) is susceptible toward isomerization in the presence of thiols,⁶² presumably by a radical mediated process. Robillard has also noted the higher propensity of s-TCO to isomerize in vivo.⁷³ While d-TCO appears more robust than s-TCO toward thiol-promoted isomerization, both of these strained TCOs are less resilient than conformationally unstrained TCOs such as 9.^{74, 75} Full details of isomerization studies are provided in the Supplementary Information.

In a prior study, it was shown that the tetrazine derived amino acid 15 could be site-specifically encoded into proteins such as GFP (2) by an evolved Methanococcus jannaschii (Mj) tyrosyl-tRNA synthetase/tRNA_(CUA) pair (Scheme 5).⁵⁸ When genetically encoded into GFP, this amino acid could be tagged in vivo by reactive s-TCO derivatives. Given the advantages of d-TCO in terms of synthesis and stability, we investigated the ability of this dienophile to label protein 2. As described previously, the tetrazine of 2 quenches the fluorescence of the attached GFP. After tetrazine ligation the fluorescence is restored, thus providing a convenient handle for monitoring the reaction. With d-TCO syn-3a and 3b, in vitro labelling takes place rapidly with a rate constant of k₂ 95 (+/− 0.3) M⁻¹s⁻¹ and k₂ 99 (+/− 0.6) M⁻¹s⁻¹ to give conjugates 16a and 16b, respectively (Scheme 5a). As expected, labelling by the d-TCOs was much faster than labelling by 5-hydroxy-trans-cyclooctene (13), and within an order of magnitude of the more reactive s-TCO (Scheme 1b). In vivo labelling of 2 with d-TCO syn-3a was confirmed by mass spectroscopy. Thus, E. coli containing expressed 2 were washed and incubated with syn-3a for 2 h in PBS buffer at room temperature. After washing the cells to remove excess syn-3a, purification was carried out using His-tagged protein affinity chromatography. Q-TOF MS of the purified protein confirmed the molecular mass of 16a, thereby successfully demonstrating site-specific labelling in vivo (Scheme 5b).

Scheme 5.

Characterization of tetrazine d-TCO reaction on GFP (a) in vitro labelling rates of d-TCOs with tetrazine unnatural amino acid derived GFP 2 (b) Q-TOF mass spectrum from the in vivo labelling of syn-3a and 2. Near quantitative conversion of 2 with syn-3a demonstrating specific conversion to 16a (expected 28130 Da; observed 28130 ± 1 Da). Each sample did show a small peak at −131 ±1 Da indicating minor amounts of peptidase-based removal of N-terminal methionines and +22 sodium adducts.

Conclusions

Computation was used to design of conformationally strained dioxolane-fused trans-cyclooctene (d-TCO) derivatives that also display excellent reactivity in the tetrazine ligation. The cis-dioxolane ring introduces conformational strain and improved aqueous solubility to the cyclooctene derivatives. A water-soluble derivative of 3,6-dipyridyl-s-tetrazine reacts with d-TCO with a second order rate k₂ 366,000 (+/− 15,000) M⁻¹s⁻¹ at 25 °C in pure water. d-TCO derivatives can be prepared easily through diastereoselective synthesis, and are typically crystalline bench-stable solids that are stable in aqueous solution, blood serum, or in the presence thiols in buffered solution. GFP with a genetically encoded tetrazine-containing amino acid was site specifically labelled in vivo by a d-TCO derivative. This combination of reactivity and stability make d-TCOs attractive for bioorthogonal applications where fast rates and stability are of importance.