Azabicyclic vinyl sulfones for residue-specific dual protein labelling† †Electronic supplementary information (ESI) available: Detailed methods and additional characterisation. See DOI: 10.1039/c9sc00125e

We have developed [2.2.1]azabicyclic vinyl sulfone reagents that simultaneously enable cysteine-selective protein modification and introduce a handle for further bioorthogonal ligation.


Introduction
For certain biological applications, the installation of two or more distinct synthetic modications into a protein of interest is desirable. 1 To introduce two distinct modications onto a protein two options are possible: (1) site-selectively modify two different amino-acid residues within the protein's sequence 2,3 or (2) use a scaffold that simultaneously displays two different modications and one handle for site-selective protein modication. 4,5 The latter, which uses multifunctional scaffolds, is perhaps the most straightforward because it does not require orthogonal chemoselective protein reactions. 1 Senter and coworkers developed a double-drug carrier with a hydrophilic tail that reacts with the protein through a cysteine-reactive handle (maleimide). 5 In another example, Gonçalves and co-workers reported dichlorotetrazine as a trivalent platform for conjugation to cysteine and showed dual labelling of albumin with a macrocyclic chelator for simultaneous nuclear imaging and a uorescent probe for imaging. 6 In addition, Chudasama, Caddick and co-workers have explored dibromopyridazinedione reagents that are able to site-specically redbridge disulde bonds and-at the same time-leave two click-reactive handles available for subsequent modications with uorophores, PEG derivatives or drugs through strain-promoted or coppermediated azide-alkyne cycloaddition reactions. 4,7,8 Watersoluble allyl sulfones were also shown by Weil and co-workers to rebridge disuldes whilst enabling dual protein functionalisation. 9 In another example, Swarts and co-workers developed a bicyclo[6.1.0]nonyne-based cyclooctyne reagent for the modi-cation of azide-labelled biomolecules with functional handles that allow both photo-crosslinking and subsequent detection/ enrichment of binders. 10 The latter example is the only one in which the ability to perform dual protein labelling was shown in cells through bioorthogonal labelling, with the others being used to prepare dual-labelled conjugates in the test tube. Thus, the development of simple and robust methodologies that enable the installation of two specic modications at a single site on a protein specially when one handle can be used for bioorthogonal ligation in cells is an area of prominent interest.
Herein, we present strained [2.2.1]azabicyclic vinyl sulfones as efficient and versatile reagents for residue-specic dual protein labelling. Our design includes a highly reactive and selective alkene handle for bioorthogonal inverse electron demand Diels-Alder (iEDDA) ligation and one easily accessible functionalisation site through N-substitution reactions on the azanorbornadiene scaffold.
Vinyl sulfones are useful Michael acceptors and dipolarophiles in cycloaddition reactions for synthesis and medicinal chemistry applications, 11 and as handles with multi-purpose functions in protein modication. 12 Their rst use for protein modication dates from 1988 in which simple methyl and ethyl vinyl sulfones were used to alkylate thiol, 3-amino or imidazole side-chain amino acids. 13 Since that time, there have been a number of reports that describe the development of allyl 9 and vinyl 14,15 sulfones for disulde rebridging 16 and cysteine 17 conjugation methodologies, respectively. Under slightly basic conditions, the sulydryl side-chains of the cysteine residues are typically more nucleophilic than the amino groups of lysine and imidazole, and the hydroxyl groups of serine and threonine and thus, selective modication can be achieved. 18 Strained systems offer an unique opportunity for protein labelling because the signicant energy stored in the system is liberated either in part or fully in a click-like bioconjugation process. 19 Among them, [2.2.1]bicyclic systems, such as norbornenes, exhibit both high reactivity and high stability, and are synthetically accessible. (Hetero)norbornenes and (hetero) norbornadienes can be easily obtained in one step through Diels-Alder reactions between a cyclic diene (cyclopentadiene/ furan/pyrrole derivatives) and the appropriate dienophile (electron-decient alkene or alkyne). Of particular interest are the 7-oxanorbornadiene derivatives developed by Finn and coworkers, which were previously used for site-selective modication of cysteine residues in proteins by conjugate addition (Fig. 1a). 20,21 Aer cysteine ligation on the protein, oxanorbornadiene undergoes a controllable retro-Diels-Alder (rDA) fragmentation that allowed their use as drug delivery systems. 22 Strained [2.2.1]bicyclic systems, especially oxanorbornadiene and norbornene systems, 19 have also been widely used in the eld of bioorthogonal chemistry as handles, for example, for fast reactions with tetrazines in iEDDA. 23 On the contrary, the use of [2.2.1]azabicyclic systems for bioconjugation is rare. A single example showed the use of a 7-azanorbornene as hub molecule for the incorporation of a uorophore for imaging and biotin as an affinity probe for protein recognition. 24 This system lacks a handle that would allow for site-selective covalent protein modication.
In this work, we envisioned N-substituted azanorbornadienes of type A, which combine a vinyl sulfone functionality, an additional reactive double bond and a functionalisable bridging nitrogen as an attractive reagent for cysteine-specic and dual protein labelling (Fig. 1b). The reactive double bond of the resulting azanorbornene upon cysteine modication offers an attractive handle for further bioorthogonal modication at the same residue through iEDDA. 23,25 Finally, the bridging nitrogen can be used to incorporate relevant fragments for protein studies/functions, such as uorophores, affinity probes or drugs. Thus, with the designed chemical probe we can selectively target cysteine within a protein sequence and functionalise the molecule through simple N-derivatisations of the azanorbornadiene and further modify it in situ through fast iEDDA reactions with tetrazine reagents.

Development of [2.2.1]bicyclic vinyl sulfones for cysteine modication
Based on a reported procedure, 26,27 we rst carried out the synthesis of racemic azanorbornadiene 1 through a Diels-Alder reaction between inexpensive N-tert-butyloxycarbonyl (Boc)pyrrole and ethynyl p-tolyl sulfone. We then tested the reactivity of azanorbornadiene 1 towards cysteine by using N-Boc-Cys-OMe as a model system. When a stoichiometric amount of bicyclic system 1 was reacted with the cysteine model in a buffered mixture of NaP i (100 mM, pH 7.3) and N,N-dimethylformamide (DMF) at room temperature for 10 min, complete conversion to the corresponding trans thioether adduct was detected (Scheme 1a). This experimental data conrmed the high reactivity of this particular vinyl sulfone embedded in the strained 7-azanorbornadiene skeleton towards thiol-Michael addition. The exo-facial preference of the nucleophilic attack is well documented in other [2.2.1]azabicyclic systems. [27][28][29][30] Furthermore, to attest the chemoselectivity of our strained [2.2.1]azabicyclic reagent we performed a competition experiment between N-Boc-Cys and N-Boc-Lys. Notably, under our experimental conditions [buffered mixture of NaP i (50 mM), pH 7.3, dimethylsulfoxide (DMSO), room temperature, 10 min] only the Cys adduct was obtained and no modication of the 3amino group of lysine was detected (Scheme 1b and Fig. S6 †). Together our data show that [2.2.1]azabicyclic vinyl sulfones can react rapidly under aqueous conditions in a chemo-and stereoselective manner with cysteine residues.

Stability of cysteine-[2.2.1]azabicyclic vinyl sulfone towards rDA
Having shown the chemoselectivity of 1 towards cysteine residues, we then studied the stability of the formed thioether because we were concerned by some rDA reports of cleavage with similar bicyclic systems. 22,[31][32][33][34] Thus, fragmentation of azanorbornene 2 into N-Boc-pyrrole and thio-vinyl sulfone 9 was studied by 1 H NMR spectroscopy at 37 C in CDCl 3 ( Table 1, entry 1; see Fig. S8 and S9 † for NMR spectroscopic monitoring). Compound 2 has a half-life of z16 h, which is nearly half of the longest half-life reported for the corresponding oxanorbornadiene derivative. 22 As reported, the rDA fragmentation has a tuneable rate depending upon the substituents of the norbornene so we decided to modulate the stability of our reagent. To prove the concept, we introduced two bridgehead methyl groups on the bicyclic system (vinyl sulfone 5, ESI † for synthetic details), which signicantly decreased the half-life (<5 min) of the corresponding cysteine adduct (Table 1, entry 2). In fact, in this case, the corresponding bicyclic adduct was not observed and afforded directly the rDA products. Next, instead of modifying the diene structure of the pyrrole ring, we decided to change the protection of the amino group from Boc in 1 to an acetyl group in 7 (ESI † for synthetic details), which doubled the half-life of the cysteine adduct 8 (37 h; Table 1, entry 3). The stabilising effect was even more predominant when we changed to more polar DMSO-d 6 , which resulted in a half-life of z59 h, which is nearly 4 times higher than the initial conditions ( Table   1, entry 4). Moreover, when a mixture of DMF and buffered aqueous solution was chosen to test the potential rDA, aer 59 h at 37 C only 37% of adduct 8 had decomposed into the rDA products.
Quantum mechanical calculations on model compounds (Fig. S12 in the ESI †) reproduced the experimentally observed slight slowing down of the reaction when replacing the Boc by the Ac group, and when using a more polar solvent (DMSO vs. CHCl 3 ), reecting the high nonpolar character of the rDA transition structures. However, and unlike in the case of Finn's oxanorbornadiene thiol and amine adducts, which have different substitution patterns and exhibit much more asynchronous transition structures as studied by Houk, 35 the difference in reactivity of the bridgehead-methylated and nonmethylated analogues was not so apparent computationally, and only using nearly complete models could reproduce the experimental trend. While the rDA can take place under physiologically relevant conditions, our data shows that this reaction is very slow (>59 h) for conjugates formed through modication of cysteine with [2.2.1]azabicyclic vinyl sulfone reagents, which is promising for their potential utility to build conjugates for in vivo studies.

[2.2.1]azabicyclic vinyl sulfones for residue-specic protein labelling
We decided to evaluate whether our ndings on small molecules could be translated into cysteine-tagged proteins. We started by adding D-biotin to the bridging nitrogen of the strained [2.2.1]azabicyclic vinyl sulfone scaffold for potential use in imaging or enrichment experiments. The bridging nitrogen is ideal as a point of attachment because it is not expected to inuence the reactivity of the vinyl sulfone towards thiol-Michael addition. Briey, acidic Boc-deprotection of 1 followed by acylation with D-biotin acid chloride 10 afforded biotinylated azanorbornadiene 11 in 75% overall yield (see Scheme S1 in the ESI †). An identical synthetic route may be used to incorporate other motifs routinely used for protein modication including PEG, drugs or uorophores (e.g. dansyl moiety, see Scheme S2, † compound 12). As a model protein for labelling, we decided to produce a mutant of a multivalent protein-ubiquitin (Ub-K63C)that has been engineered to feature a surface-exposed cysteine residue at position 63 (for further details see the ESI †). 36 By using an equimolar amount of biotin-functionalized [2.2.1]azabicyclic vinyl sulfone 11 in NaP i buffer (20 mM, pH 7.0) and DMF as co-solvent (10%) at room temperature for 30 min, Ub-K63C was fully converted into a single and homogenous product ( Fig. 2a and b) as analysed by LC-MS. The crude reaction was then subjected to excess Ellman's reagent and aer 1 h at 37 C, no change was noted, which conrms that all cysteine was consumed during the Michael reaction with vinyl sulfone 11 (see Fig. S18 †), and further conrmed the chemoselectivity of our method. Importantly, we showed that the Ub-K63C-biotin conjugate is relatively stable in plasma media. Aer 24 hours incubation at 37 C with plasma, only modest degradation of the product (<20%) was observed as a result of retro-Diels-Alder reaction (see Fig. S19 †). Moreover, the original thioether linkage remained unaltered and did not exchange with other biological thiols present in plasma media. Our data show a cysteine-specic protein modication with a [2.2.1]azabicyclic vinyl sulfone that bears a synthetic modication inserted at the bridging nitrogen of the reagent that leaves the remaining alkene unreacted, which may be further used for iEDDA ligation in a test tube or in cells.

Imaging C2Am-11-labelled apoptotic cells
Encouraged by our results, we envisioned to apply our biotinfunctionalised reagent for cysteine-selective protein modication and further labelling with a streptavidin probe for imaging. For these studies we used the C2A domain of Synaptotagmin-I (C2Am) that binds to phosphatidylserine (PS), which is localized on the external leaet of the plasma membrane during apoptosis. In previous studies a C2Am mutant that has a single cysteine at the position 95 for modication with a radionuclide, a uorescent probe, or a magnetic resonance-detectable tag was used for imaging apoptotic tissues in vivo. 37 Thus, we reacted this protein mutant with compound 11 by using similar conditions to those described above (5 equiv., 20 mM of NaP i buffer/10% DMF at pH 7.0, room temperature, 30 min reaction) to give a single homogenous product ( Fig. 3a and b) determined by LC-MS. The efficiency of the biotinylation reaction was also examined by Western blot. Non-modied C2Am (control) and biotinylated protein C2Am-11 were resolved by SDS-PAGE. Biotinylation was conrmed by using a streptavidin uorescent probe (Streptavidin-Alexa555) aer transfer to a polyvinylidene diuoride membrane (Fig. 3c). We then tested if this C2Am derivative retains its inherent functionality of binding to the PS phospholipid on the surface of apoptotic cells. Therefore, C2Am-11 was used for imaging apoptotic cells by pre-targeting followed by labelling with Streptavidin-Alexa555. 38 Apoptosis in HeLa cells was induced by treatment with actinomycin D (Act D, 1 mM, 12 h). Following treatment with Act D, cells were washed and incubated with C2Am-11 (6 mM, 20 min) at 37 C. Aer pre-incubation, apoptotic cells were further washed and incubated with the counterpart of biotin, Streptavidin-Alexa555 (2.5 mL of 1 mg mL À1 , 20 min). Blocking studies were performed by incubating apoptotic cells with a large excess of non-modied C2Am and then with biotinylated surrogate C2Am-11 and Streptavidin-Alexa555. Fluorescent microscopy images showed staining of only apoptotic cells by C2Am-11 + Strep-Alexa555, whereas healthy control cells remained unstained (Fig. 3d). Moreover, in the blocking experiments, in which cells were rst incubated with the non-labelled C2Am, a signicant decrease of the mean uorescence intensities (MFI) was observed (Fig. 3e). These results attest the ability of modied C2Am-11 to bind to PS to enable specic labelling of apoptotic cells.

Reactivity of cysteine-[2.2.1]azabicyclic reagents towards tetrazines
Aer having studied the reactivity, stability and utility of the cysteine attachment point of [2.2.1]azabicyclic vinyl sulfones, we explored the possibility of a second reactive handle for site-specic modication with the addition of a proper counterpart. In our case, the subsequent modication would take advantage of the remaining double bond present in the bicyclic system by its reaction with tetrazines through iEDDA. Based on a precedent with reactions between azanorbornene derivatives and tetrazines, 24 we evaluated the reactivity of thioether bicyclic system 2 with different tetrazines (13a-13c) by monitoring the decrease of tetrazine absorbance over time. 23 We found that the dienophile present in 2 can quickly and easily undergoes iEDDA reactions with tetrazines. 23 Considering the properties of the diene to perform an iEDDA most electron-decient tetrazines 13a and 13b resulted in a faster reaction, whereas more electron-rich tetrazine 13c reacted an order of magnitude slower ( Table 2). As demonstrated by quantum mechanical calculations (Fig. 4 and S13 †), and contrary to what is commonly assumed, the superior reactivity of N-substituted azanorbornenes compared to their non-bicyclic analogues does not arise from any special strain at the double bond, as it is the case for other highly strained p-systems such as trans-cyclooctene or cyclooctyne. This can be deduced from the very similar double bond lengths and angles in non-bicyclic and bicyclic analogues. Moreover, such apparent activation of the alkene group strongly depends on the type of substitution at the bridging atom. While N-alkyl azanorbornenes or norbornenes show nearly the same (or even slower) calculated iEDDA reaction rates with tetrazine 13a than their non-bicyclic vemembered analogues (1-alkyl-2,5-dihydro-1H-pyrroles and cyclopentene, respectively), calculated activation barriers for Ncarbonyl (carbamate or amide)-substituted azanorbornenes are 5-6 kcal mol À1 smaller than those of their non-bicyclic counterparts, resulting in a 3-4 orders of magnitude acceleration. The source of this acceleration is the distortion of the otherwise planar 1-carbonyl-2,5-dihydro-1H-pyrrole ring into a puckered conformation closely resembling that at the transition state, similarly to what described for norbornenes vs. (Z)-but-2-ene or cyclohexene. 39 Fig. 4 shows that the iEDDA reaction causes a puckering of the planar ve-membered 2,5-dihydro-1H-pyrrole ring in model system A1 c of z13 at the transition state, while for bicyclic analogue A1 b the structural changes at the transition state are minimal. This reects the pre-distorted nature of azanorbornenes and thus their superior ability to achieve transition state geometries through lower activation barriers, which ultimately translates into reaction rate acceleration.
These data highlight the suitability of strained [2.2.1]azabicyclic vinyl sulfone reagents to site-specically modify cysteine residues whilst simultaneously leaving a reactive alkene free for further iEDDA labelling. With the chemoselectivity towards cysteine residues evaluated (one attachment point studied with an affinity tag, and the viability of chemically modifying the second reactive handle in our [2.2.1]azabicyclic vinyl sulfones with tetrazines appraised), we then needed to see whether this last achievement was also reproducible in a physiological environment. This would allow us to demonstrate the dual-labelling character present in this vinyl sulfone.
We planned to pre-target apoptotic cells with our previously synthesized bioconjugate with the affinity tag (C2Am-11) and then use uorogenic and commercially available 6-methyl- Table 2 Kinetic studies for the iEDDA reaction between azanorbornene 2 and tetrazines 13a-13c

Entry
Tetrazine 13c 0.143 AE 0.005 Fig. 4 Kinetic studies for the iEDDA reaction between azanorbornene 2 and tetrazines 13a-13c. Lowest-energy conformations calculated for N-Moc-protected (Moc ¼ methyl carbamate) cyclic 2,5-dihydro-1H-pyrrole (A1 c ) and bicyclic azanorbornene (A1 b ) models and their iEDDA reactions with tetrazine 13a calculated with PCM DMF /M06-2X/ 6-31+G(d,p). Five-membered ring puckering is represented by the dihedral angle f and relative reaction rate constants at 37 C (310 K, k 310 rel ) are calculated from activation free energies (DG ‡ ). A significant distortion of the dihydropyrrole ring is required to achieve the transition state geometry in A1 c , while in A1 b such distortion is already developed up in the dienophile, translating into a reaction rate acceleration of several orders of magnitude. tetrazine-sulfo-Cy3 (Tz-Cy3) to functionalise the remaining double bond (Fig. 5a). The protocol for targeting cells with C2Am-11 was similar to the one mentioned above. Aer incubation with the bioconjugate, cells were washed and further incubated with uorogenic tetrazine for 90 min. Tetrazine labelling of pre-targeted C2Am-11 allowed ready visualization of cells rendered apoptotic by treatment with Act D (Fig. 5a). Similarly, cells previously blocked with "native" C2Am, and then labelled with C2Am-11 + Tz-Cy3 following the same protocol, showed a signicant decrease of uorescence (Fig. 5c). These experiments conrm that the specic affinity of C2Am-11 for apoptotic cells is retained and the double bound is accessible by a tetrazine uorophore in a cellular environment (Fig. 5d). Reactivity of the double bond towards the uorescent tetrazine was studied by SDS-PAGE (Fig. 5b). Aer 90 min a uorescent band that corresponded to C2Am-11-Tz-Cy3 modied protein was observed (Fig. 5c and d). We further tested the versatility of this system for dual modication of proteins. Reaction with precursor 12 (dansylglycine-azanorbornadiene) enabled installation of a uorescent tag on ubiquitin that could be further modied by reaction with a tetrazine compound (accessed by MS and SDS-PAGE in-gel uorescence; Fig. S20 and S21 †). Overall, the azanorbornadiene handle can be used as a platform for double modication of the same protein without changing its native properties.

Conclusions
In this work, we explored a [2.2.1]azabicyclic vinyl sulfone for cysteine-specic dual protein labelling. We have demonstrated the attachment of synthetic modications through N-substitutions of a pyrrole ring and showed the utility of having a second reactive handle to perform bioorthogonal ligation in cells through a site-specic reaction between a bicyclic alkene and a tetrazine. Furthermore, we have proven the stability of the anchorage in plasma, and demonstrated that the modied proteins retain their initial function. The simplicity and synthetic accessibility of these azanorbornadienes makes them ideal reagents for cysteine-specic dual protein modication. This strategy may be used for example for the attachment of two different drugs to a targeting protein/antibody, the incorporation of different tags for target identication or even a cytotoxic molecule and a uorophore for following cancer treatment.

Conflicts of interest
There are no conicts to declare.