Dichloro Butenediamides as Irreversible Site‐Selective Protein Conjugation Reagent

Abstract We describe maleic‐acid derivatives as robust cysteine‐selective reagents for protein labelling with comparable kinetics and superior stability relative to maleimides. Diamide and amido‐ester derivatives proved to be efficient protein‐labelling species with a common mechanism in which a spontaneous cyclization occurs upon addition to cysteine. Introduction of chlorine atoms in their structures triggers ring hydrolysis or further conjugation with adjacent residues, which results in conjugates that are completely resistant to retro‐Michael reactions in the presence of biological thiols and human plasma. By controlling the microenvironment of the reactive site, we can control selectivity towards the hydrolytic pathway, forming homogeneous conjugates. The method is applicable to several scaffolds and enables conjugation of different payloads. The synthetic accessibility of these reagents and the mild conditions required for fast and complete conjugation together with the superior stability of the conjugates make this strategy an important alternative to maleimides in bioconjugation.


Introduction
Over the last decade protein site-selective conjugation has become at hriving research field. [1] Ap lethora of methods have been reported for the construction of protein conjugates with aw ide array of applications in chemical biology and medicine. [2] Cysteine [3] (Cys) and lysine [4] (Lys) remain the main target residues in protein modification although recently,methodologies that target alternative residues,such as aspartic/glutamic acid, [5] tryptophan, [6] methionine, [7] serine, [8] or tyrosine, [9] have been described. Cys-targeting methods are particularly ubiquitous,a sthey yield structurally homogeneous conjugates in proteins that contain native [10] or genetically engineered free Cys. [11] Multiple strategies have been reported for Cys-selective conjugation, usually based on alkylation [12] /arylation [13] reagents or Michael acceptors. [14] Michael acceptors,p articularly maleimides,r emain the most commonly used reagent for the construction of conjugates for biological applications [15] because of their associated fast reaction kinetics.I nf act, an umber of Food and Drug Administration-approved conjugates,s uch as antibody-drug conjugates (ADCs) Brentuximab vedotin [15] and Tr astuzumab emtansine [16] or PEGylated conjugate Certolizumab pegol, [17] contain at hio-succinimide adduct derived from maleimide conjugation. [18] However,i tis well known that thio-succinimide adducts can undergo fast and uncontrolled disruptive cleavage by thiol-exchange in plasma, which ultimately compromises the safety and efficacy of the conjugate. [19] Considerable efforts have been reported on how to increase the stability of maleimide-based constructs. [20] Them ost common approach consists of hydrolyzing the thio-succinimide linkage to form as table,l inear thio-succinamic acid species. [21] Different strategies to control the hydrolysis kinetics have been described. Forinstance,the addition of electron-withdrawing groups in the payload, [22] tuning the amino-acid environment near the inserted Cys, [21a] or mild ultrasonication [23] can accelerate thio-succinimide hydrolysis.A lternatively,m aleimides with different leaving groups attached to the vinylic bond or similar structures have been reported to form conjugates with higher stability based on maleimide or maleamic acid linkages. [24] There are also acyclicr eagents,s uch as benzoyl acrylic reagents [11,25] or fumarate derivatives. [26] They produce conjugates with linear linkages,w hich can present enhanced stability in plasma relative to maleimides. [11,25] Despite these advances,m ost of these strategies fail to combine fast conjugation kinetics along with the formation of stable conjugates,w hich is required for biological applications,a nd where maleimides remain the first choice.T he challenge remains to find ag eneral, simple alternative for Cys bioconjugation with comparable kinetics and synthetic accessibility to maleimides,but complete linkage stability.
We hereby report on diamide and amido ester derivatives from maleic acid as new,e fficient reagents for Cys-selective protein conjugation. These reagents undergo fast Michael addition reactions with free Cys-containing proteins in stoichiometric amounts under physiological conditions.C onjugation occurs through an unprecedented mechanism in which, upon Cys addition, spontaneous cyclization is promoted by attack of an amide moiety to the distal carbonyl group to form as uccinimide link (Scheme 1). By expanding this approach to 2,3-dichloro butenediamides,w ec an selectively target free Cys residues that form chloro-maleimide intermediate linkages,which quickly undergo hydrolysis.The final conjugates are fully resistant to thiol-exchange and cleavage in the presence of biological thiols and human plasma. Through this approach an array of free Cys-containing proteins could be irreversibly tagged, such as ubiquitin (Ub), human serum albumin (HSA), or different antibody fragments,w ith different tags including azide or alkyne functionalities.T he favorable kinetics,s tability,a nd stoichiometry of the conjugation makes aflexible platform to access complex constructs,such as ADCs.

Results and Discussion
With the goal of finding alternative approaches to access homogeneous conjugates with fully stable linkages in free Cys-containing proteins,weset out to investigate maleic acid derivatives with labile functionalities,w hich we envisioned could undergo fast Michael addition with thiols under physiological conditions and, upon conjugation, would hydrolyze to form succinamic acid structures or equally stable analogues.
Ther eactivity of b-mercaptoethanol (BME) was evaluated with different maleic acid derivatives (1-10,F igure 1, Conditions A,F igures S3-S6) to assess their reactivity as Michael acceptors.H aving determined the reactivity of 1-10 in small molecule experiments,w en ow exposed these reagents to the presence of af ree Cys-containing protein ( Figure 1,C onditions B,F igures S25-S31). As am odel protein we chose Ub,which was engineered with areactive Cys at position 63 (K63C). Surprisingly,the reactivity of the different species varied depending on the setup.Some species showed high reactivity in small molecule reactions but no protein conjugation reactivity,w hereas others proved to be very efficient protein conjugation reagents but were unreactive in as mall molecule setup.I ns uccessful conjugations,C ysselectivity was confirmed after digestion by MS/MS studies ( Figure S82, S83) and by using Ellmansreagent as achemical control ( Figure S32).
Mass analysis of the protein conjugates formed revealed au nique conjugate with am ass of 8754 Da (protein mass + 188 Da) for all the examples.I nterestingly,t he mass of the newly formed conjugate did not correspond to the addition of  the full reagent. Instead, we observed the formation of athiosuccinimide linkage.T he origin of such al inkage is through intramolecular attack of the amide to the distal carbonyl group (Scheme 2a), which can happen prior to or after Cys conjugation. Similar cyclizations are known although never triggered by aqueous solution or thiol conjugation. [27] Furthermore,the cyclization step seems to play akey role in the viability of the conjugation, because only compounds with at least one secondary amide,capable of driving such ac yclization, underwent conjugation. To gain better mechanistic insight, compounds 2,a nd 5-10 were stirred for 2hat 25 8 8C in DMSO/D 2 O( NaP i pH 8, 50 mm)t od etermine if the cyclization was induced as aresult of the pH of the solvent or the thiol addition. Amido esters 2 and 8 cyclized spontaneously in the presence of D 2 Op H8 to form am aleimide species (Figures S1,S2). However,t his was not the general behavior for amido esters;c ompound 6 together with diamides 5, 7, 9,a nd 10 was completely stable under these conditions.T his aqueous stability supports the idea that thiol addition in the protein environment is responsible for triggering the cyclization. By using diamide 10 and Ub as am odel, the reaction showed great tolerance for pH [5- When ar eagent with two different secondary amide groups,such as 9,isconjugated, aregioselectivity issue arises. Thepresence of two secondary amide moieties allows for two possible cyclizations, each driven by one of the amides (Scheme 2b); each pathway leads to adifferent conjugate: 9a or 9b.T he preferred pathway could be controlled by the pH at which the reaction was done.U nder slightly acidic conditions,pH6-7, the aryl amide group-directed cyclization was favored, whereas under basic conditions,p H8-9, the opposite selectivity was observed ( Figures S38, S39).
Another appealing aspect of this conjugation reaction is the controlled release of small molecules in ap articular microenvironment of the protein sequence.Upon cyclization, an amine (diamides) or alcohol (amidoesters) molecule is released, which could open new approaches for monitoring reactions (release of fluorophores), Cys-targeted decaging of payloads or site-directed bifunctionalization of proteins.W e are currently investigating this directed release in our laboratories and will report it in due course.
Thet hio-succinimide linkages obtained are known to hydrolyze over time under physiological conditions to give alinear and stable N-substituted thio-succinamic acid linkage, but the kinetics of hydrolysis is extremely slow and often yields mixtures of conjugates.H arsh hydrolyzing conditions, such as high pH and temperature,could ultimately lead to the complete hydrolysis to succinamic acid linkages,b ut such conditions are not compatible or desirable with proteins because they can result in unfolding and loss of functional properties.T oaddress this problem, we focused on accelerating the hydrolysis of the linkages created by the conjugation of maleic acid diamides.
Aqueous instability of dichloro maleimides [24c] suggests that hydrolysis kinetics of maleimides can be exponentially increased by the introduction of halogens into the ring.T his led us to think that, by introducing chlorine substituents into the alkene scaffold of maleic acid diamides,the stability of the compounds and their reactivity would not be affected, although now instead of obtaining conjugates with succinimide linkages,c hloro-maleimide linkages would be generated, which would quickly hydrolyze to stable linear chloro maleamic acid species.B yf ollowing reported procedures [28] we synthesized symmetrical dichloro butenediamides (11-14, Figure 2a)a nd, to our delight, their reactivity still included ac yclization step which yielded ac yclic chloro-maleimide linkage with enhanced hydrolysis kinetics.A ddition of 11 (3 equiv) to asolution of Ub resulted in complete conjugation (  Figures S67-S70)],w hich further supports the hypothesis of rapid hydrolysis of the linkage.T he mechanism associated with this conjugation is similar to that of dibromo maleimides [23] in that the linkages obtained are similar, but an additional cyclization step and much more favorable hydrolysis kinetics make this approach advantageous to obtain homogeneous,stable maleamic acid linkages.
Theg enerality of the method was demonstrated by conjugation of 7 and 14 to alternative free Cys-containing proteins.U nder similar conditions,H SA, the N-terminal domain of phage repressor R434 and ac amelian nanobody targeted at amyloid precursors (HET nanobody) [29] were effectively conjugated (Figures S40-S42) to 7.Inall examples aM ichael addition/cyclization mechanism to yield as uccinimide was observed, which supports the generality of the mechanism in different protein sequence microenvironments. Analysis of the circular dichroism (CD) profiles of the conjugates relative to native proteins confirmed that overall Scheme 2. Regioselectivity of the cyclization step with compound 9.

Angewandte Chemie
Research Articles the tertiary structure of the proteins was maintained throughout conjugation (Figures S85, S86).
An array of dichloro butenediamide analogues with different functionalities were synthesized (compounds [11][12][13][14]. Alyphatic 11 and aromatic 12 amide species could be accessed as well as amides with an incorporated alkyne 13 or azide 14 functionality.T hese new functionalities were incorporated into the protein structure to enable further functionalization through copper(I)-catalyzed alkyne-azide cycloadditon. Compounds 12-14 showed similar reactivity to 11 upon reaction with Ub to form stable,l inear chloromaleamic acid linkages (Figures S44-S46). However,w hen compound 14 was tested with different proteins,u pon conjugation-cyclization, different reactivity was observed for some of these proteins.I nU b, C2Am [30] or peptide ASCATNt he expected sequential Michael addition-cyclization-hydrolysis occurs to give afinal conjugate with achloromaleamic acid linkage.W hen the scope is expanded to proteins like HSA, H3K4C,o rt he full-length IgG antibody Gemtuzumab,a na lternative pathway consisting of as ubse-quent Michael addition of ap roximal nucleophilic residue occurs,i nstead of the hydrolysis step (Figure 2b). This alternative pathway forms cyclic maleimide linkages with two residues on the protein. Both linkages cannot undergo retro-Michael addition and form stable conjugates,which was demonstrated by analyzing the stability of Ub-14;t he conjugate remained stable in buffer, and no degradation was observed after 48 hat378 8Cinthe presence of glutathione (1 mm)ori nh uman plasma (Figures S73, S74).
Theenhanced plasma stability displayed by the conjugates formed from the reaction of Cys with reagents 11-14 has potential for the construction of stable ADCs.H owever, because two alternative linkages can be formed, this could constitute an issue regarding homogeneity of the final construct. Next, we analyzed the ratio between cyclic and linear conjugates using LCMS.T he conjugate derived from Gemtuzumab V205C showed am ixture of species,[ ratio 2:1 cyclic vs.l inear ( Figure S53)],b ut all others formed selectively one of the two potential linkages.C onjugation of HER2-targeted IgG Tr astuzumab V205C with 14 led to asimilar result (ratio 4:1, Figure S54).
To explain this result at the atomic level, we performed molecular dynamics (MD) simulations on the Fabfragment of Tr astuzumab conjugated to compound 14 (Figures 2b and  S87) and examined the distance from all lysine residues to the reactive carbon (CÀCl). According to these calculations, lysine at position 207 had the smallest average distance and was presumably the residue that directed the second nucleophilic attack to form the cyclic derivative.T he simulations indicated also that the incorporation of this cyclic derivative into Fabd oes not significantly disrupt its structure (Figure 2b). Thei nvolvement of K207 in the formation of the cyclic linkage was further confirmed by MS/MS studies on Tr astuzamab V205C-14 ( Figure S84). By mutating lysine 207 into an alanine,t he new Tr astuzumab V205C + K207A was conjugated to 14 and this time complete selectivity to the linear linkage was obtained ( Figure S55). This result supported the idea that nearby lysine residues were the driving factors for the formation of the cyclic linkage,allowing us to predict the final linkage formed and if mixtures were to arise, avoid the formation of the cyclic species by introducing cysteines in the appropriate place or mutating the problematic lysine residues.W ef ound, however, that mutation of lysine 207 in Tr astuzumab V205 + K207A reduces the nucleophilicity of cysteine 205 and only approx. 50 %conjugation is achieved under the previously described conditions (Figure S55).
To find suitable positions to genetically encode areactive Cys that allow complete conversion, yet selective formation of the stable linear linkage,w eproduced Fc fragments of an IgG with Cys insertions at different positions (239iC,2 68C, 274C,2 89C and 442C). Under identical reaction conditions with 14,t he Fc fragments with 274C and 289C reacted completely and selectively formed the linear linkage (Figures S56-S60). These simple maleic acid derivatives enable installation of modifications on antibodies at specific sites to form homogeneous and stable antibody conjugates.
To compare the kinetics of conjugation between maleic acids and maleimides,a ne quimolar mixture of 11 and its analogous maleimide reagent were conjugated to Fc fragment 289iC.Under the same conditions, 11 and maleimide showed comparable conjugation kinetics as a1:1 mixture of Fc-11 and Fc-maleimide constructs was obtained ( Figure S81). However,when we compared the stability of the constructs in the presence of GSH (1 mm)a fter 66 ha t3 78 8C, the constructs (trastuzumab-11,Fc274iC-11,Fc289iC-11)conjugated to 11 showed higher stability than the corresponding species conjugated with maleimide ( Figures S75-S80). While the conjugates formed with reagent 11 remain intact in the presence of at hiol, constructs built using maleimides form am ixture of hydrolyzed maleimide and unconjugated antibody.T hese data show ak ey advantage of our reagents to build stable conjugates and illustrate how enhanced hydrolysis kinetics of the cyclic linkages allows the formation of fully stable species.

Conclusion
We describe the protein conjugation reactivity of aseries of maleic acid derivatives and identify amido esters and diamides as viable conjugation reagents to produce succinimide linkages through an ew stepwise mechanism. The strategy was optimized to use dichloro butenediamides,which promote af urther hydrolysis step to form completely stable linkages to proteins.T his latter species shows comparable conjugation kinetics relative to maleimides yet benefits from superior stability.T he general mechanism described composes aMichael addition followed by spontaneous cyclization to form cyclic linkages similar to those obtained by conjugation to maleimides.I ntroduction of Cl atoms into the cyclic linkages promotes subsequent hydrolysis,l eading to fully stable linkages.Nearby nucleophilic residues can promote an alternative mechanism where as econd nucleophilic attack leads to the formation of acyclic linkage,but this pathway can be controlled and depleted if undesired conjugate mixtures are obtained. Thea pproach allows for the creation of conjugates with azide/alkyne moieties and opens up possibilities to create tailored conjugation for an umber of relevant proteins,i ncluding nanobodies or IgGs relevant for the formation of ADCs.