2-Alkyl Furans Undergo Radiolytic Oxidative Protein Crosslinking

Abstract

Induced covalent bond formation, or crosslinking, is a powerful strategy for interrogating and modulating biological systems. Current strategies rely largely on photochemical processes, which generally restrict their use to cellular or other optically accessible settings. Broadening the chemistry underlying the crosslinking-initiating event could extend these approaches to more complex biological systems. Ionizing radiation - a cornerstone of cancer therapy – is an attractive but unexplored trigger for covalent chemistry. Here, we detail the systematic comparison of a series of rationally designed radiolytic crosslinking candidates. These studies reveal that 2-alkyl furans efficiently label a representative protein in a radiation dose-dependent manner. Mechanistic analyses suggest that radiolysis induces hydroxyl radical–mediated oxidation of the furan, generating an electrophilic dicarbonyl intermediate that reacts with lysine residues to form stable adducts. We apply this strategy to enforce the covalent capture of an otherwise reversible small-molecule–protein interactions in cells. Together, these findings establish ionizing radiation as an effective trigger for covalent chemistry, expanding the scope of radiolytic chemistry to include bond formation. This work provides a foundation for future applications leveraging spatially targeted radiation for target identification or the development of next-generation therapeutic strategies.

Graphical Abstract

Introduction

Triggered bond formation, or crosslinking, is a powerful strategy to modulate and monitor biological function. The most common of these methods use UV light to directly form reactive intermediates that modify proximate biomolecules. Diazirine, azide and benzophenone cross-linking reagents have become essential tools for diverse applications spanning ligand target identification and materials science.^{1, 2} Related to direct crosslinking are proximity-labeling methods, in which biomolecules (e.g. small molecules, proteins) of interest are tagged with a photocatalyst that activates a latent reactive species to tag – rather than directly crosslink – nearby partners.³ Mechanistically, these approaches involve either direct energy transfer from the excited photocatalyst to the tagging group or energy/electron transfer to generate reactive oxygen species (ROS).^4–8 The latter can involve creating a reactive functional group capable of protein modification or producing oxidized residues for identification (Figure 1)^9–11. Recent efforts have extended the wavelength of activation into the tissue-penetrative far-red and near-infrared range, and these approaches are finding broad use in increasingly complex settings.^12–14 However, strategies that enable direct crosslinking of small molecules and biomolecules using tissue-tolerant wavelengths are largely unexplored.

Figure 1:

Key background of photolytic labeling chemistries and radiolytic strategy described here.

This methodological gap led us to ask the question: could ionizing radiation be used as an initiator of covalent bond formation? Pioneering studies focused on defining the consequences of ionizing radiation in cancer therapy have described the ability of targeted radiation to trigger functional group transformations, typically via hydroxyl radical–driven pathways.^{15, 16} Recent efforts have applied radiolytic transformations to develop novel uncaging chemistries that use targeted radiation to mediate the local release of therapeutic small molecules.^17–19 In addition, organic and inorganic afterglow particles that emit visible and near-infrared light in response to targeted irradiation are being developed.²⁰ These studies illustrate the potential for ionizing radiation to initiate controlled chemistries in biological systems. However, the use of radiolytic chemistry to trigger covalent bond formation had never been systematically investigated prior to these efforts.

These studies report the discovery and application of a radiolytic crosslinking reaction. Synthesis and screening of a series of rationally designed crosslinking candidates led to the identification of 2-alkyl furans as latent electrophiles activated by ionizing radiation. Mechanistic studies indicate that radiolytic activation of furans is a hydroxyl radical–mediated process that leads to radiation dose–dependent formation of a persistent electrophilic species capable of modifying protein lysine residues. A functionalized furan probe enables radiation-dependent biotinylation of live cells. Finally, this reaction is applied to covalently enforce an otherwise reversible small-molecule–protein interaction in complex proteomes. These studies expand the scope of radiolytic chemistry to bond formation, setting the stage for new applications of ionizing radiation.

Results and Discussion

Synthesis and evaluation of a panel of candidate X-ray activatable latent electrophiles

To identify functional groups capable of undergoing covalent bond formation upon exposure to ionizing radiation, we synthesized a panel of rationally designed biotin- and desthiobiotin-labeled probes. As an initial test for covalent reactivity, we incubated each probe with bovine serum albumin (BSA) and irradiated each solution with X-ray doses of 10–30 Gy using a biological irradiator (320 keV). Crosslinking of each molecule to BSA was assessed by streptavidin blot-based detection and quantified by using densitometric analysis of detectable bands. To provide a uniform benchmark for these studies, we labeled BSA with a known electrophile (desthiobiotin N-hydroxysuccinimide ester; DTB-NHS) at a degree of labeling (DOL) of 1 (Figure S1). The crosslinking efficiency (CE) of the nineteen candidates was quantified relative to this control (Figure 2).

Figure 2.

Initial screen for radiolytically activated crosslinking. (A) Schematic of the BSA crosslinking assay and quantification of chemiluminescence signal following 30 Gy X-ray irradiation, normalized to degree-of-labeling 1 (DOL-1) BSA–biotin, alongside the chemical structures of the biotinylated probe library. All probes were tested at 50 μM in pH 7.0 PBS containing 0.5 mg/mL BSA (7.53 μM) and irradiated at 0, 10, or 30 Gy. (B) Representative streptavidin/anti-biotin–HRP blots for probes exhibiting detectable radiation-dependent crosslinking, with corresponding Ponceau-stained membranes shown to confirm equal protein loading. (C) Radiation dose–response analysis of the three furan analogues displaying the highest levels of radiation-triggered crosslinking activity. Signals are reported relative to DOL-1 BSA–biotin. Compounds annotated with an asterisk are modified with D-biotin; all others are modified with D-desthiobiotin.

Furans undergo ring opening to form reactive dicarbonyls in response to oxidative triggers such as N-bromosuccinimide (NBS) and photosensitization.^21–23 Evaluation of furan-containing probes 1a–1c following irradiation at 30 Gy revealed that all three underwent radiation-dependent crosslinking with BSA (Figure 2A–C). Notably, the 2-alkyl furan analogue 1a exhibited crosslinking levels qualitatively indistinguishable from the DTB-NHS-labeled standard (CE = 1.05). The 3-furan 1b (CE = 0.25) and 2-phenyl furan 1c (CE = 0.03) reagents also modified BSA at this radiation dose, albeit to a lesser extent. Encouraged by these findings, two related series of probes were analyzed: thiophenes 2a-2b, in which the ring oxygen is replaced with sulfur, as well as substituted furans 3a–3e, in which the electronic properties of the heterocycle are modulated by halogen and electron-withdrawing substituents. In contrast to the strong labeling exhibited by unsubstituted furans, no evidence of radiation-responsive crosslinking was observed with any of these compounds. Mass spectrometric analysis further revealed that, unlike 1a–1c (see below), 2a–2b and 3a–3e showed no evidence of radiolytic reactivity, even at X-ray doses as high as 50 Gy.

A series of potentially reactive electron-rich heterocycles and aryl derivatives were also examined. Imidazoles exhibit both radioprotective and radiosensitizing effects in radiotherapy through hydroxyl radical-mediated mechanisms.^24–28 Nitroimidazole 4a was designed based on the drug nimorazole, which exhibits radiosensitizing effects in hypoxic cancers.²⁹ Thioimidazole 4b was designed with the hypothesis of forming a imidazole-centered radical, possibly formed by bond cleavage adjacent to the sulfur atom. However, neither 4b nor the nitroimidazole 4a exhibited detectable BSA reactivity. Studies have shown dimethoxy aryls can be hydroxylated upon exposure to ionizing radiation, an approach applied in uncaging strategies.³⁰ To test this chemistry, we synthesized 5a, which we hypothesized could access aryl-aryl crosslinking pathways, and 5b, potentially capable of forming an electrophilic quinone-methide intermediate. However, neither reagent exhibited measurable BSA crosslinking.

Finally, we examined several commonly used crosslinking motifs used with photolytic and enzymatic processes. These included phenol and aniline derivatives, which find broad utility in photosensitization-based and enzymatic proximity labeling strategies.^31–33 We observed weak but detectable radiation-dependent crosslinking by both phenol 6a (CE = 0.02) and aniline 6b (CE = 0.01) following irradiation at 30 Gy (Figure 2B). Finally, we explored a set of azides and diazirines (7a-c), inspired by their known conversion into reactive intermediates upon exposure to photoirradiation or in certain cases, ionizing radiation.^34–36 Measurable radiation-dependent crosslinking was not observed for any of these compounds. Together, these screening results indicate that 2-alkyl furans represent a uniquely radiation-responsive functional group, prompting further investigation into the mechanism and applications of this chemistry.

Products of furan crosslinking reaction

The conversion of furans to biological electrophiles is generally understood to involve ring-opening to form a dicarbonyl intermediate capable of reacting with amine nucleophiles.^{9, 37} This transformation has been previously shown to be triggered by photosensitization and chemical oxidation (Figure 3A,B).^{23, 38–41} To examine how a similar mechanism might be triggered by ionizing radiation, we carried out mass spectrometry analyses of model reactions. Rose Bengal-driven photooxidative chemistry of furans has been characterized extensively and the product distribution resulting from reactions of unsaturated dicarbonyls with biological nucleophiles, particularly amines, has been explored in a variety of previous studies.^42–44 Therefore, we irradiated 1a with 535 nm light in the presence of Rose Bengal (RB) to use its product profile as a benchmark against the radiolysis products. The resulting product profiles were then compared to those derived from X-ray irradiation (50 Gy). Furan 1a (10 μM) was irradiated in water in the presence or absence of a model nucleophile ethylamine. Irradiating 1a alone in water with either photosensitization or ionizing radiation led to the observation of a +16 Da product, which corresponds to the dicarbonyl intermediate i (Figure 3A,B, Figure S2). We also observe the formation of a species with an m/z ion corresponding to carboxylic acid ii, which is formed to a much greater extent in the X-ray irradiation experiments. To determine what ROS species are specifically responsible for 1a oxidation to i and ii, we subjected 1a to reactions with independently generated ¹O₂ and •OH. Hydrogen peroxide treatment of 1a in the presence of sodium hypochlorite and UV light (254 nm) was used to generate ¹O₂ and •OH, respectively. Comparing the resulting product profiles to that of 1a X-ray radiolysis revealed that all three oxidations produce a species with m/z corresponding to i, while detectable amounts of m/z ion for ii were only produced by UV photolysis of H₂O₂ and X-rays (Figure S3A–B). Employing a 1:1 mixture of nominal (H₅) and deuterated (D₅) ethylamine allowed dominant and trace amine reaction products to be readily identified based on their unique isotopic signature. Prior studies have found that dihydropyrrolinones are the major amine adduct formed from furan (photo)oxidation and reaction of the resulting dicarbonyl intermediates.^45–48 Upon RB photooxidation, we observe compound iii, with mass signature that corresponds to the previously characterized dihydro-pyrrolinone products. Critically, the major amine adduct formed upon radiolysis, is also product iii, with an identical retention time and mass signature (Figure 3A). We also observe a hydroxylated variant of iii, whose structure we tentatively assign as dihydropyrrolinone, iv, which is formed only upon radiolysis and may result from hydroxyl radical- mediated oxidation of iii. Overall, these results indicate that furan radiolysis creates a reactive dicarbonyl, previously known as the major product of photosensitized oxidation, that reacts with amines to form dihydro-pyrrolinone products.

Figure 3.

Chemical characterization of the furan radiolytic crosslinking mechanism. (A) Extracted ion chromatogram (EIC) traces for m/z values corresponding to species of interest and proposed structures for X-ray or photosensitization adducts. Results show transformation of 1a (10 μM in H₂O) in the presence or absence of ethylamine after 20 min 535 nm photo-irradiation (with 0.5 mol% Rose Bengal (RB)) or 50 Gy X-ray irradiation. Amine-containing samples were heated for 2 hours at 37 °C followed by quenching with 1 v/v% formic acid. (B) Proposed structures arising from irradiation-induced transformation of 1a, with calculated m/z values. (C) Proposed mechanisms for furan activation via photosensitization by Rose Bengal and direct activation by ionizing radiation. (D) Quantification of BSA crosslinking following irradiation of 1a (50 μM, 0.5 mg/mL BSA) by 535 nm hn (5 min, 1 mol% Rose Bengal) or X-rays under the indicated conditions. Samples were prepared in PBS alone or supplemented with 1 M of the indicated quenchers. Chemiluminescence signals are normalized to irradiation of 1a in aerated PBS. (E) LC–MS analysis of radiolysis products of 1a in water at varying concentrations and irradiated at 40 Gy. Ion count ratios were calculated relative to an internal standard for “1a+O” (+16 m/z oxidation product: 324 Dal.) at 0 Gy and 40 Gy. Samples were incubated at room temperature for several hours post-irradiation, then treated with 1% formic acid prior to LC–MS analysis. Blot images and quantified chemiluminescence signal relative to DOL-1 BSA–biotin obtained from BSA treated with varying concentrations of 1a and irradiated at 30 Gy (F) and with one molar equivalent of 1a (7.5 μM) and irradiated at increasing X-ray doses (G). BSA samples were heated at 37 °C overnight.

The reduced protein crosslinking activity of furans 1b-c relative to 1a could be the result of lower radiation-responsiveness to form the oxidized species and/or diminished reactivity of the resulting electrophiles. To distinguish between these possibilities, we carried out mass spectrometry analysis following exposure to ionizing radiation in the presence of H₅/D₆ ethylamine. For 1b, only trace levels of the +16 adduct were observed upon high radiation doses (30 Gy). Conversely, 1c readily generates the +16 product at radiation doses as low as 10 Gy, but this oxidized species does not appear to efficiently form the stable pyrrolin-2-one adduct upon introduction of ethylamine (Figure S4). These results highlight the importance of both activation and electrophile reactivity in radiolytic furan crosslinking reactions.

Radiolytic furan crosslinking depends on hydroxyl radical formation

Next, we assessed the role of ROS on radiolytic crosslinking. Radiolysis of water leads to hydroxyl radical formation, which is a potent oxidant and the major mediator of radiolytic chemistry.^{49, 50} Prior studies have suggested the hydroxyl radical, generated through UV photolysis of hydrogen peroxide, leads to furan oxidation.⁴⁰ Theoretical and experimental studies have proposed mechanisms through which these •OH-furan reactions lead to the formation of the unsaturated dicarbonyls (Figure 3C). ^{23, 38, 51}By contrast, the most well-characterized mechanism of furan photosensitization occurs via a singlet oxygen-mediated (¹O₂) cycloaddition reaction.^{22, 37} To evaluate the contributions of these potential pathways to radiolytic crosslinking, we employed 1a in BSA crosslinking experiments mediated by either i) RB-mediated photosensitization or ii) X-ray irradiation. For each reaction we assessed the influence of sodium azide, which quenches both singlet oxygen and hydroxyl radical, as well as sodium formate, which quenches only hydroxyl radical.⁵² Our expectation was that sodium formate would show enhanced inhibition if radiolytic crosslinking operated through a hydroxyl radical-dependent pathway.⁵² Indeed, while sodium azide significantly reduced crosslinking in both conditions, sodium formate exclusively inhibited the X-ray-mediated reaction (Figure 3D). Given that hydrated electrons (e⁻_aq) are also a primary product of water radiolysis, we utilized potassium formate as a e⁻_aq scavenger and found that there was no impact on BSA crosslinking (Figure S5A).⁵³

We then examined the impact of potential quenching species, the role of radiation dosing, and furan concentration. We first evaluated crosslinking in the presence of glutathione (GSH), the most abundant cellular thiol, which could act as either as a scavenger of hydroxyl radical or a nucleophile following furan ring opening, and found mM GSH levels dramatically reduced reaction efficacy (Figure S5B). This finding indicates this approach may be most well-suited for crosslinking in a non-reductive environment, for example in the extracellular setting. Molecular oxygen has the potential to play a role in both the photosensitized and radiolytic crosslinking mechanisms (Figure 3D).^{45, 54, 55} To study this role we degassed protein samples prior to irradiation in the presence of 1a, leading to a 63% and 48% reduction in crosslinking within the RB and X-ray-mediated reactions, respectively. Furthermore, if crosslinking is dependent on hydroxyl radical, its activation should be saturatable (i.e. a function of the amount of hydroxyl radical produced) with previous studies finding that 1 Gy of ionizing radiation produces a fixed quantity (0.27 μM) of hydroxyl radical in water.⁵⁶ To test this, we used mass spectrometry to monitor formation of the +16 Da (+O) furan-derived species after 40 Gy irradiation (Figure 3E). Consistent with our expectation, formation of the oxidized intermediate becomes saturated at 10 μM of 1a, as would be expected if all formed hydroxyl radical is consumed. An analysis of the radiation dose dependance reveals the extent of 1a consumption in H₂O to be 0.1 μM/Gy (Figure S6). We then assessed the role of furan concentration and radiation on BSA crosslinking. The crosslinking yield of 1a exhibits a rapid linear increase from 1–10 μM but then approaches saturation, suggesting a dependence on radiation-mediated ROS generation (Figure 3F). By contrast, a linear increase in crosslinking is observed when a single concentration of 1a is subjected to escalating radiation doses (Figure 3G). Overall, these results define key parameters that inform the application of furan-based crosslinking to more complex applications. We also find that furan activation and protein crosslinking are maintained under fractionated irradiation regimens (3 × 10 Gy over 24 h or 15 min intervals), a common clinical strategy, as well as a single 30 Gy exposure in both small‑molecule LCMS and BSA crosslinking assays (Figure S7). Overall, these studies support a mechanism for furan radiolytic activation that involves a hydroxyl radical- and oxygen-mediated oxidative process, and inform efforts to apply this approach.

Radiolytic crosslinking of ligands to the surface of living cells

As an initial test of the furan radiolytic crosslinking reaction in a biological system, we evaluated its ability to immobilize ligands to the surface of living cells (Figure 4A, Figure S8). To enable these experiments two probes were synthesized consisting of an avidin-detectable desthiobiotin tethered via a PEG₁₂ linker to either a 2-furan (8a) or non-furan control (8b, Figure 4A). To evaluate cell surface crosslinking, MDA-MB-468 breast cancer cells grown under normal culture conditions were washed and incubated with solutions of 8a or 8b in PBS for 20 minutes. Following exposure to crosslinking probes, cells were irradiated, washed, stained with APC-streptavidin and analyzed by flow cytometry. Analysis of fluorescence data revealed clear radiation-dependent increases in fluorescence for 8a but not control 8b, confirming a requirement for the furan moiety in the crosslinking mechanism (Figure 4A, C). In the absence of irradiation, similar levels of fluorescence were observed in cells treated with 50 μM 8a and 8b. This suggests that, under these conditions, non-specific activation of the furan does not contribute significantly to cellular labeling. Next, we tested crosslinking across a range of 8a probe concentrations at a constant radiation dose (Figure 4B,D). Here we observed 8a to display a significant increase in cell surface crosslinking when stepping the concentration from 0 to 10 μM. At higher concentrations the signal continues to increase, but less rapidly. This observation is consistent with radiation-dependent formation of hydroxyl radicals serving as a limiting factor for furan crosslinking, in line with our observations in the model reactions described above. These results establish the ability of radiolytic crosslinking to immobilize small molecule ligands to the surface of living cells, validating its compatibility with complex biological environments.

Figure 4.

Radiation-induced furan cell surface labelling. (A) Schematic illustrating procedure for cell surface labeling experiments. (B) Flow cytometry results from desthiobiotin probes showing radiation dose series for furan-containing 8a and COOH-bearing negative control 8b (both probes dosed at 50 μM); radiation dose response observed only with 8a (B). (C) Demonstrating concentration dose response by 8a at varying concentrations in response to 0 Gy or 30 Gy X-ray dose. Probes were tested in MDA-MB-468 cells using ~20-minute probe incubation. (C) and (D) Plots representing data from experiments in B (D) and C (E) representing radiation and concentration dose response data, respectively. Data shown as average between two replicates (4,922–10,000 events acquired per sample) with error bars representing standard deviation.

Radiolytic crosslinking covalently captures transient small molecule-protein interactions

A key application of covalent crosslinking is the ability to trap or capture interactions that are otherwise transient. Previous studies have demonstrated that oxidative or photocatalytic activation of furans can crosslink peptides to their receptors, trigger interstrand DNA crosslinking, and enable RNA-protein crosslinking.^{44, 57–59}

To define parameters relevant to biological environments, we first evaluated the persistence of the activated furan intermediate under physiological conditions. Irradiation of 1a in PBS followed by incubation at 37 °C prior to protein addition resulted in sustained protein crosslinking activity for several hours (Figure S9A). Complementary small-molecule studies were performed by adding ethylamine to irradiated 1a and monitoring adduct formation by LCMS. Average ion intensities for the m/z corresponding to the pyrrolinone adduct iii decreased by ~50% after 24 hours at 37 °C (Figure S9B), providing an estimate of the functional lifetime of the reactive intermediate in water. This is distinct from the reactivity of commonly used crosslinkers such as diazirines and benzophenones, which rapidly revert or decompose following activation in aqueous solutions.

We next examined the radiolytic reactivity of furan 1a through experiments in which ethylamine was introduced after irradiation to isolate the reactivity of the radiolytically generated electrophile. In addition to the dihydropyrrolinones amine adduct iii described above, a second compound with the same m/z was also observed and is assigned as the imine v based on its lability to acidic conditions (Figure 5A and Figure S9C). At longer time points, we also observed the formation of an amide product vi, previously identified in other furan oxidative crosslinking studies.⁶⁰ These findings are consistent with prior studies of oxo-enal compounds and suggest that the radiolytically generated intermediate is susceptible to amine capture followed by subsequent conversion to amide products.^{46, 61–63}

Figure 5.

(A) Time-dependent formation of 1a-nucleophile adducts following incubation of pre-irradiated 1a (11.1 μM, 50 Gy X-rays – filled circles; 0 Gy – open circles) with amino acid surrogates; ethylamine, imidazole and methyl thioglycolate. EIC traces shown were derived from m/zs corresponding to the nucleophile adducts iii (351.2 m/z), v (351.2 m/z), vi (369.2 m/z), vii (401.3 m/z) and (439.2 m/z) viii. Data presented for 1 mM EtNH₂ alone (left), 1 mM EtNH₂ with 0.25 mM imidazole (middle) and 1 mM EtNH₂ with 0.25 mM methyl thioglycolate (right) all with 10 μM 1a in plain water. Samples were heated at 37°C for the indicated durations followed by addition of the internal standard and transfer to −20 °C storage to stop each reaction. All 0 Gy samples were heated for 48 hours. (B,C) MS/MS fragmentation for BSA peptide from trypsin digestion containing lysine 431 modified with 1a (500 μM); +305 Da (B) and +323 Da (C) corresponding to pyrrolinone and ketoamide modifications, respectively. BSA (0.5 mg/mL) was irradiated with 50 Gy X-rays along with 1a and incubated at 37°C overnight.

To investigate whether other amino acid side chains could participate in crosslinking, we performed time-course LCMS experiments with ethylamine in the presence of either imidazole or a thiol, methyl thioglycolate (Figure 5A, Figure S10A). Imidazole accelerated consumption of the acid-labile species v and increased formation of adducts iii and vi, suggesting that spatially organized nucleophilic residues within a protein may similarly enhance crosslinking. Consistent with this prediction, rapid and efficient BSA crosslinking was observed within 1 h at 37 °C, independent of protein tertiary structure or thiol availability (Figure S10B). We also detected a transient imidazole–pyrrole adduct, which disappeared at longer incubation times. Similar adducts have been identified in previous 1,4-dicarbonyl reactivity studies, although their instability is consistent with their highly electron-rich character and the absence of isolable reports of 3-imidazole-N-alkyl pyrroles.^{46, 63, 64} In the presence of thiol, we observed a mass spectral signature for a crosslinked species modified by both ethylamine and methyl thioglycolate. This product is assigned as the isomeric 3- or 4-thiopyrrole product viii in line with previous examples of related reactivity arising from 1,4-unsaturated dicarbonyl electrophiles.^{46,63, 65–68,69, 70} Together, these results suggest that crosslinking can proceed through multiple adducts and is shaped by the local environment.

To understand whether furan crosslinking prefers specific lysines or uniformly labels all surface-exposed residues, we performed tryptic digestion and LC-MS/MS analysis of BSA labeled by 1a across a nine-point probe concentration series (0–500 μM) after activation using either Rose Bengal photosensitization or X-ray irradiation (Table S1). Proteomic analysis revealed that specific lysine residues (K211 and K431) were preferentially modified even at low probe-to-protein ratios (Figure S11A–B). Two types of labelling were readily observed on these residues; 305 Da and 323 Da increments corresponding to pyrrolinone and ketoamide modifications, respectively, were each detected under both activation modes (Figure 5B–C, Figure S11A–B, Table S1). Notably, these residues are proximal (< 6 Å) to hydroxyl-containing side chains, suggesting that local chemical environments may enhance reactivity, consistent with previous observations for related electrophiles (Figure S11C).^71–75

Next, we investigated ligand-targeted radiolytic crosslinking. Targeting integrin α_vβ₃ with the modest affinity RGD ligand (K_d = 89 nM) was chosen as a model system.⁷⁶ Initial biochemical studies established that a series of RGD desthiobiotin-pegylated probes exhibit significantly reduced BSA labeling compared to the unmodified 1a, suggesting that structural changes modulate crosslinking (Figure S12).⁷⁶ Furthermore, two pegylated probes, which vary the distance between the RGD tripeptide and furan were compared for their ability to modify α_vβ₃ in an in vitro setting. Indeed, the variant with the longer RGD-furan spacer emerged as the most efficient crosslinker of α_vβ₃, outperforming the non-targeting 8a under the same conditions (Figure S12). Consistent with prior crosslinking studies, we find that β₃ is preferentially crosslinked likely reflecting RGD binding geometry in the heterodimer.^{77, 78}

Finally, we set out to ask whether we could harness these insights to enable integrin-specific radiolytic crosslinking in a cellular context. A set of biotinylated probes (9a, 9b and 9c) were synthesized, enabling cross-linking to be detected following enrichment using LC-MS/MS proteomic analysis (Figure 5A,B). Probes 9a-c were each incubated at 50 μM with U87-MG cells (0 °C, 30 – 40 minutes). To differentiate specific and non-specific labeling, in a subset of experiments cells were pre-incubated with high-affinity α_vβ₃ cycRGDfK (50 μM), which we expected to compete for radiolytic crosslinking events mediated by specific molecular recognition of the integrin α_vβ₃ active site. Following irradiation and cell lysis, biotinylated proteins were isolated through immobilization on avidin beads, washed, digested, eluted, and analyzed by quantitative tandem mass tagging (TMT) proteomics. Although all three probes demonstrated efficient crosslinking towards cellular proteins, only 9c, which places the furan distal to the RGD motif, displayed α_vβ₃-specific crosslinking through the enrichment of ITGβ₃ and competition by cycRGDfK (Figure 5C). Notably, intracellular proteins such as ITPR3, SYNE2, NUMA1 and SON also displayed measurable competition by cycRGDfK (Figure S13). Given these proteins are not known to be associated with α_vβ₃ biology, we hypothesized that the observed competition may be due to concentration-dependent off-target effects (Figure 5D,E,F). Consistent with this interpretation, at a reduced concentration of 9c (10 μM), ITGβ₃ remained the protein whose enrichment was most significantly competed by cycRGDfK (1 μM), and the only protein for whom competition was observed at both concentrations (Figure 5F). Further inspection of the data found that ITGα_v is also both enriched and competed by cycRGDfK, albeit to a lesser degree than ITGβ₃ - a property consistent with other RGD pull down studies.⁷⁸ Taken together, these experiments demonstrate that radiolytic activation of furans can induce ligand-directed irreversible crosslinking of new chemical functionalities to cell surface targets.

Conclusion

Here we establish that 2-alkyl furans act as effective radiation-responsive latent electrophiles for covalent bond formation in biological settings. A rationally designed panel of candidate X-ray activatable motifs was synthesized and evaluated in protein crosslinking assays, revealing that unsubstituted 2-alkyl furans undergo robust, radiation-dependent conjugation. Mechanistic experiments, combining mass spectrometry, radical quenching, and dose–response studies, support a pathway in which ionizing radiation generates hydroxyl radicals that oxidize the furan ring to a persistent dicarbonyl electrophile that reacts with lysine side chains. Our findings suggest that the oxidized dicarbonyl intermediate can be intercepted by amines, imidazole, and/or thiols to form potentially several distinct stable products, making crosslinking highly dependent on the local nucleophilic environment. When incorporated into biotinylated probes and RGD-containing ligands, substituted 2-alkyl furans enable radiation-triggered labeling of living cell surfaces and covalent capture of integrin α_vβ₃ in complex proteomes.

Additional efforts will expand upon and fully define aspects of the described chemistry. Crosslinking efficiency is ultimately limited by the yield of radiolytically generated hydroxyl radicals and molecular oxygen, leading to saturable behavior with respect to probe concentration. Although our data are consistent with a major pathway involving furan-to-dicarbonyl oxidation followed by lysine capture, there is likely an additional key role for adjacent amino acids in enhancing and, perhaps, participating in the cross-linking event. Given that the key electrophilic intermediates are relatively long-lived, careful optimization of probe concentration is required to limit off-target protein labeling. Finally, further studies to either enhance the reactivity of furans or identify alternate chemotypes that reduce the required radiation dose would facilitate translation into living systems.

Looking ahead, the ability to cross-link proteins, and potential other biomolecule, with tissue penetrating ionizing radiation offers unique opportunities in a range of settings. Modern stereotactic radiotherapy delivers high-dose irradiation to small tumor targets with high spatial precision.⁷⁹ When paired with radiolytic crosslinking chemistry, this sharply confined irradiation could spatially localize covalent bond formation. Low molecular weight furan-based crosslinkers could be embedded within small molecules, proteins, nucleic acids, or nanomaterials to enable spatially and temporally resolved capture of extracellular target interactions in tissues inaccessible to optical activation. This approach could provide a complementary alternative to whole body cross-linking imaging approaches that rely on tissue clearing,⁸⁰ and facilitate target identification and mechanistic dissection directly in disease-relevant complex environments.⁸⁰ Furthermore, successful implementation of such strategies could be coupled with radiotherapy or radiopharmaceutical treatments to stabilize low affinity protein–ligand complexes, broadening the utility of specific but weak binders. Such approaches may have significant utility in efforts to improve the specificity and targeting of immunotherapies. Overall, we hypothesize that new classes of radiation-activated probes and therapeutics have significant potential to define and manipulate biological systems in a spatially regulated manner.

Supplementary Material

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.6c03350. Contains synthesis procedure and characterization of key compounds and intermediate compounds, details of key experiments, uncropped version of all gel images, and supplementary figures (PDF). Table SI, with the tryptic peptide-level LC-MS/MS identifications of BSA modification, in also included as an .xslx file.

Figure 6.

Assessing targeted integrin crosslinking by RGD-containing furan probes. (A) Chemical structures of the RGD-containing furan probes along with their diagrammatic representations. (B) Illustration of the α_vβ₃ crosslinking experiment leading up to immobilization of biotinylated proteins using Streptavidin (SA)-bound magnetic beads and subsequent proteomics readout. The purified protein or harvested cells are treated with the RGD-containing probes and after ~45 minutes of incubation, irradiation proceeds, followed by EDTA treatment to disrupt non-covalent interactions. (C) Graphical representation of fold change and statistical significance of competition by cycRGDfK inhibitor observed for probes 9a-c (50 μM) in U87MG cells. (D) Waterfall plot showing fold-change values for the 37 proteins competed by cycRGDfK (with −log₂FC > 1) when 50 μM of both 9c and the cyclo-RGD is used. (E) Volcano plot from 9c-treated (10 μM) U87-MG lysates irradiated at 50Gy demonstrating effect of cycRGDfK (1 μM) competition. (F) Venn diagram showing overlap between 50 μM and 10 μM conditions, for proteins competed in 9c-treated U87MG experiments.

Abbreviations

APC

allophycocyanin

BSA

bovine serum albumin

CE

crosslinking efficiency

COOH

carboxylic acid

DOL

degree of labeling

DTB

desthiobiotin

DTB‑NHS

desthiobiotin N‑hydroxysuccinimide ester

EDTA

ethylenediaminetetraacetic acid

Gy

gray

hν

photon energy (light)

HRP

horseradish peroxidase

ITGα_v

integrin alpha‑v

ITGβ₃

integrin beta‑3

K_d

dissociation constant

LCMS

liquid chromatography–mass spectrometry

LC–MS/MS

liquid chromatography–tandem mass spectrometry

NBS

N‑bromosuccinimide

PBS

phosphate‑buffered saline

PEG

polyethylene glycol

RB

Rose Bengal

RGD

arginyl‑glycyl‑aspartic acid

ROS

reactive oxygen species

SA

streptavidin

UV

ultraviolet