Probing the Mechanism of Photoaffinity Labeling by Dialkyldiazirines through Bioorthogonal Capture of Diazoalkanes

Abstract

Dialkyldiazirines have emerged as reagents of choice for biological photoaffinity labeling studies. The mechanism of crosslinking has dramatic consequences for biological applications where instantaneous labeling is desirable, as carbene insertions display different chemoselectivity and are much faster than competing mechanisms involving diazo or ylide intermediates. Here, deuterium labeling and diazo compound trapping experiments are employed to demonstrate that both carbene and diazo mechanisms operate in the reactions of a dialkyldiazirine motif that is commonly utilized for biological applications. For the fraction of intermolecular labeling that does involve a carbene mechanism, direct insertion is not necessarily involved, as products derived from a carbonyl ylide are also observed. We demonstrate that a strained cycloalkyne can intercept diazo compound intermediates and serve as a bioorthogonal probe for studying the contribution of the diazonium mechanism of photoaffinity labeling on a model protein under aqueous conditions.

Graphical Abstract

Photoaffinity labeling has become a ubiquitous tool in chemical biology and drug discovery.¹ Aromatic diazirines, arylazides, and benzophenone derivatives have long served in photoaffinity labeling applications,^1–3 and recent probes based on nitrile imines⁴ and photoredox catalysis⁵ offer new possibilities. A consideration for the design of any photoaffinity label (PAL) is the minimization of size and hydrophobicity in order to ameliorate any detrimental impact of the PAL on the function, solubility, permeability, or subcellular localization of the biological molecules being studied.⁶ Accordingly, dialkyldiazirines have emerged as favored probes for photoaffinity labeling in cellular experiments^7–9 with applications that include probing small molecule–protein interactions,^6,7 protein–protein interactions,¹⁰ nucleic acid–protein interactions,¹¹ metabolic oligosaccharide engineering,¹² lipid–protein interactions,^13,14 and the study of biological membranes.¹⁵

While their physiochemical properties are advantageous for chemical biology, there are important photochemical differences between dialkyldiazirines and more traditional aromatic diazirines. Analogs of α-trifluoromethyl-α-phenyldiazirine are well established biological probes that upon irradiation give high intermolecular labeling yields via formal carbene addition even with aliphatic CH bonds,¹⁶ and because α-trifluoromethyl-α-phenylcarbene lacks α-CH bonds, it cannot participate in intramolecular α-C–H insertion processes. α-Trifluoromethyl-α-phenyldiazirine also leads to ∼35% of α-trifluoromethyl-α-phenyldiazomethane.¹⁶ While this side product undergoes photochemistry with shorter wavelength UV (∼300 nm),¹⁷ it does not absorb at the wavelengths typically used for diazirine activation (>350 nm) and displays high stability in the dark under neutral or acidic conditions.¹⁶ Photochemical reactions of dialkyldiazirines are relatively challenging to study because of the susceptibility of both alkylcarbenes and their photoexcited nitrogenous precursors to undergo intramolecular rearrangements with α-hydrogens.¹⁸ However, several dialkylcarbenes have been characterized and both singlet and triplet ground states have been observed.¹⁹

When diazirines are applied in photoaffinity labeling, it is frequently assumed that carbene insertion reactions are responsible for the bimolecular capture of biological targets, but further consideration is required with dialkyldiazirine precursors (1, Figure 1A). Like aromatic diazirines, aliphatic diazirines partially rearrange to diazo compounds upon irradiation.¹⁸ However, aliphatic diazo compounds (2) are readily protonated at neutral pH to give diazonium compounds (4) which are reactive alkylating agents.^20,21 While both carbene (3) and diazonium mechanisms can lead to intermolecular labeling, the chemoselectivity of alkyldiazonium labeling is likely to differ from that of a carbene mechanism. Labeling by alkyldiazonium 4 is known to favor esterification reactions,^22–25 whereas the reactions of free carbenes are less selective and include aliphatic C–H insertion as a possibility.²⁶ Moreover, crosslinking reactions are extremely rapid for the carbene pathway,²⁶ whereas the alkylation reactions of alkyldiazonium ions are much slower; for example, the reaction of a methanediazonium ion with water has a relatively long half-life of ∼0.4 s in water/THF solution.^27,28 Thus, in biological target identification experiments, alkyldiazonium ions are likely to display a much larger labeling radius than more reactive carbene intermediates. Additionally, when alkyldiazonium ions do modify proteins they are likely to modify carboxylic acid residues,^22–25 producing esters which may be subsequently hydrolyzed by esterases^22–24 in live cell experiments and potentially in the proteomic workflow. Understanding the relative contributions of carbene and diazonium intermediates is therefore critical to the interpretation of biological data from photoaffinity experiments.

Figure 1.

Potential mechanisms for photoaffinity labeling with dialkyldiazirine probes. (A) Both carbene and diazonium intermediates can potentially lead to crosslinking. (B) For diazirines with pendant carbonyls, carbene mechanisms do not necessarily involve direct insertion, but may instead involve formation of ylide and oxocarbenium intermediates.

In seminal studies, Kirmse, Platz and coworkers studied the photolyses of spirobicyclic diazirines including norbornane-derived 7, which upon irradiation in MeOH gives a mixture of products that could derive from either carbene 8 or diazo compound 9 (Figure 2A).¹⁸ By studying the changes in product distribution in the presence of dipolarophile traps for diazocompounds and by analyzing levels of deuterium incorporation for photolyses in MeOD, it was possible to deduce that diazirine 7 produces both carbene 8 and diazo compound 9 in significant amounts. These studies with bicyclic diazirines not only demonstrated the importance of diazonium ions to intermolecular ether formation but also revealed a structural dependence on the contribution of carbenes to bimolecular chemistry.

Figure 2.

(A) Studies of spirobicyclic diazirines by Kirmse and Platz. (B) Structures of “linear” alkyl diazirines more commonly used in biological photoaffinity labeling.

While “linear” dialkyldiazirines have emerged as reagents of choice for photoaffinity labeling (e.g., 10¹² and 11,⁷ Figure 2B), studies on the mechanism of intermolecular labeling are lacking for the substrate types that are most often used for biological experiments. Another mechanistic consideration for such systems is that carbonyl groups are generally used to attach the probe to a molecule of interest, which opens a pathway for carbenes to react via carbonyl ylides (5) and subsequently formed oxocarbenium ions (6) (Figure 1). Recent LC-MS/MS studies on diazirine photo crosslinking of proteins have uncovered a preference for labeling carboxylic acid residues, suggesting an important role of diazo intermediates.^25,29,30 Here, we present direct evidence that both carbene and diazonium mechanisms operate in the reactions of a dialkyldiazirine motif commonly utilized for photoaffinity labeling. We demonstrate that a strained cycloalkyne can intercept diazocompound intermediates²⁰ and serve as a bioorthogonal probe for studying the contribution of the diazonium mechanism of photoaffinity labeling on a model protein under aqueous conditions.

Our study began by observing the products formed when methanol solutions of diazirine 12 (60 mM) were irradiated at 350 nm in the air to produce three alkenes 13–15, lactone 16, and the methanol adduct 17³¹ (Figure 3A). Based on corrected GC yields there was a 64% yield of the intramolecular products 13–16 and a 32% yield of the intermolecular product 17. Alkene products 13–15 may arise via carbene 18 or via diazo 21/diazonium 22 (Figure 3B,C). For carbene 18, the alkenes can form either by 1,2 hydride (1,2 H–) shift or via the intermediacy of ylide 19 and oxocarbenium 20. Among these mechanisms, only the carbene concerted 1,2 H-shift does not involve protonation. Thus, for irradiations carried out in methanol-d, the D/H ratio reports on the contribution of the carbene 1,2 H-shift mechanism for the formation of alkenes 13–15,¹⁸ as alkenes 13-d, 14-d, and 15-d arise from protonation/elimination of diazo 21 or ylide 19. As shown in Figure 3C, the level of deuterium incorporation for 13, 14, and 15 was 48%, 31%, and 50%, respectively. Hence, the carbene concerted 1,2 H-shift mechanism is responsible for at least 50–69% of the alkene products 13–15.

Figure 3.

(A) Products from photolysis of diazirine 12. (B) Mechanistic pathways of intramolecular product formation for photolyses in MeOD leading to deuterated and non-deuterated alkenes 13–15. (C) D/H ratio serves as a reporter of the carbene 1,2 H-migration pathway. (D) The ratio of alkene:ether products from amide 12a is similar to that from ester 12.

Additionally, the diazirinyl amide 12a was prepared and irradiated in MeOH (350 nm, 2 h) and by NMR analysis was observed to give alkene (13a–15a) and ether (17a) products in 61% yield and ratios that were similar to what was observed with diazirinyl ester 12. Lactone 16 was not observed, nor was the cyclic butyrolactam, 5-methyl-1-propylpyrrolidin-2-one.³²

Dipolarophiles such as fumaronitrile and diethyl fumarate participate in cycloaddition reactions with diazo compounds formed photochemically from diazirines.³³ These dipolarophiles have been used as mechanistic probes in the reactions of spirocyclic diazirines,^18,34 but are also potent electrophiles toward biological nucleophiles. We therefore sought to identify a bioorthogonal dipolarophile that could intercept intermediate diazo compounds under biologically relevant conditions. Cycloalkynes including the cyclooctyne BCN participate in a number of bioorthogonal reactions³⁵ including 3 + 2 cycloadditions with diazo compounds.²⁰ As shown in Figure 4A, the photolysis of diazirine 12 in MeOH in the presence of BCN (1.1 equiv) produces cycloadduct 23 as a mixture of diastereomers in 22% isolated yield, thus demonstrating the ability of BCN to intercept diazo intermediate 21. As depicted in Figure 4B,D, the inclusion of dipolarophile in photolyses of 12 (60 mM) in MeOD also has influence on D-incorporation for alkene products 13–15. Interception of diazo 21 by BCN is expected to lead to a decrease in the yield and the D/H ratio for alkenes 13–15. In the absence of BCN, alkene products were formed in combined 52% yield and with 31–50% D incorporation. However, the photolysis of 60 mM 12 in the presence of 70 mM BCN gave alkenes 13–15 with a yield decrease to 33% and with only 6–9% D incorporation.³⁶ This decrease in alkene yield and D-incorporation is consistent with the capture by BCN of diazo 21, which is partly responsible for the formation of alkenes 13-d, 14-d, and 15-d. Similar decreases in yield and D-incorporation were noted for the photolysis in the presence of fumaronitrile.¹⁸

Figure 4.

(A) Preparative reaction of diazirine 12 with the cycloalkyne BCN to produce cycloadduct 23 is proposed to occur by [3 + 2] cycloaddition with intermediate diazo compound 21, which (B) influence formation of intramolecular products by decreasing the yield and level of D-incorporation for alkenes 13–15, and (C) on intermolecular chemistry by decreasing the yield of product 17. (D) Effect of product formation by dipolarophiles BCN and fumaronitrile.

As depicted in Fig 4C,D, the inclusion of dipolarophiles in photolyses of 12 also decreases the yield of bimolecular product 17. The yield of ether 17 without added dipolarophile is 29%, whereas irradiation of 12 (60 mM) with either BCN (70 mM) or fumaronitrile (200 mM) gives 17 in 14% yield. Thus, intercepting diazo intermediate 21 leads to a major reduction in bimolecular product formation. These results illustrate the importance of diazo compound protonation/alkylation for intermolecular product formation.

Persistently formed in the photolyses is lactone 16, which is proposed to arise via protonation of ylide 19 and subsequent hydrolysis of oxocarbenium 20 by adventitious water. Intermediates 19 and 20 may arise from either the carbene pathway (via cyclization of 18) or the diazo pathway (via cyclization of either 21 to 19, or 22 to 20 (Figure 3B)). Consistent with all of these mechanisms, 16 shows >99% D-incorporation when photolyses are carried out in MeOD. The formation of ylide 19 from carbene 18 provides a possible explanation for the incomplete ability of dipolarophiles to suppress D-incorporation in alkenes 13–15, as would be expected if the carbene 1,2 H-shift was the sole pathway (Figure 2).³⁷ The formation of 19 from 18 also explains the incomplete suppression of bimolecular adduct 17 (Figure 4B,C). Thus, while carbene 18 may lead to the formation of an intermolecular product 17, this is most likely preceded by an intramolecular reaction to produce ylide 19 based on earlier conclusions that intramolecular H-shifts are generally too rapid for intermolecular reactions to compete.^18,38–40 Also consistent with the formation of ylide 19 from carbene 18 is the incomplete ability of dipolarophiles to suppress the formation of lactone 16, suggesting that 16 arises from a combination of the carbene and diazo pathways.

We next sought to use BCN to probe the mechanism of protein photoaffinity labeling. Rhodamines are known to non-selectively bind to bovine serum albumin (BSA), and Jewett had previously demonstrated fluorescent labeling of BSA with a rhodamine–diazirine conjugate.⁴¹ Accordingly, we prepared a fluorescent diazirine, TAMRA-DAz, and the water-soluble BCN-PEG₁₂ (Figure 5A). BSA was reduced by DTT and alkylated with iodacetamide before usage to minimize cysteine labeling by the BCN reagent.^42,43 As shown in Figure 5B, irradiating a solution of BSA (22 μM) with TAMRA-DAz (24 μM) in PBS/tris buffer for 15 min at 365 nm led to fluorescent labeling as judged by SDS-PAGE (lane 3), whereas controls without light gave no labeling (lanes 2, 9, 10). The inclusion of BCN-PEG₁₂ (2–10 mM) in photolysis experiments with BSA led to decreased labeling by TAMRA-DAz, with up to a 69% decrease at the highest BCN-PEG₁₂ concentration (lanes 4–7). However, no suppression of fluorescence was observed upon irradiation in the presence of 24, an analog of BCN-PEG₁₂ that lacks the reactive alkyne (lane 8). As further evidence that the diazonium mechanism is partly responsible for protein labeling, an additional control was carried out in the presence of sodium acetate, which would be expected to decrease labeling by competing for the diazonium intermediate.^22–25 As shown in Figure S1, inclusion of NaOAc (1 M) during the irradiation of TAMRA-DAz with BSA led to a 45% decrease in labeling.

Figure 5.

(A) Experiment involving BSA, TAMRA-DAz, BCN-PEG₁₂, and 24. (B) Lane assignments in SDS-PAGE. (C) Relative fluorescence with 0 to 10 mM BCN-PEG₁₂.

These protein labeling results are consistent with the results in methanol that show that intermolecular labeling can be suppressed by capturing diazoalkane intermediates, and support the hypothesis that the diazonium alkylation pathway is a major pathway in photoaffinity labeling. The inability to completely quench labeling by addition of BCN-PEG₁₂ does suggest a role for the a carbene intermediates in diazirine photoaffinity labeling. However, the mechanism by which the carbenes studied here label proteins is unclear and may not involve commonly assumed carbene insertion reactions, but could instead involve the intermediacy of ylide and oxocarbenium intermediates. These mechanistic details have consequences for chemical biology applications, as alkylations of diazonium or oxocarbenium ions are much slower than carbene insertion reactions which may in turn negatively impact the radius of labeling in protein identification experiments. Additionally, reactions of diazo compounds with proteins are known to selectively react with carboxylic acid residues to give esters,^22–25 which may not be stable in live cell applications due to native esterases and/or the proteomic workflow employed.^22–24

The long-recognized problem of detangling the chemistry of carbenes from the reactions of their precursors remains incredibly relevant today.^{18,39,44–48} The field of chemical biology will benefit from future studies to develop photoaffinity tags with desirable physiochemical properties that can label biological targets via a “real” carbene insertion mechanism.