Your mother was right, washing matters: an alkyne-analog of ibuprofen reveals unwanted reactivity of aromatic compounds with proteins during copper-catalyzed click chemistry

Abstract

Bioorthogonal chemistry, in particular the copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC), has enabled the robust identification of covalent protein targets of probes and drugs. Ibuprofen is commonly used pain and fever reducer and is sold as an enantiomeric racemate. Interestingly, the stereoisomers can be enzymatically converted through an ibuprofen-CoA thioester intermediate, which might non-specifically react with protein nucleophiles. Here, we use an alkyne-analog of ibuprofen to make two discoveries. First, we find that ibuprofen likely does not result in notable chemical labeling of proteins. However, we secondly find that aromatic compounds can react with proteins during the CuAAC reaction unless they are appropriately washed out of the mixture. This second discovery of false positive labeling has important technical implications for the application of this approach.

Graphical Abstract

Ibuprofen is one of the most common non-steroidal anti-inflammatory drugs (NSAIDs) for the treatment of mild pain and fever reduction in adults and children.¹ Ibuprofen is sold as a racemate, but only the S enantiomer is a biologically active anti-inflammatory drug. Interestingly, the inactive R enantiomer can be converted into the S enantiomer in the body by the consecutive action of two enzymes: the Acetyl CoA synthetase and the α-methylacyl-CoA racemase (Figure 1a).² The first enzyme generates an ibuprofen-CoA thioester that can then be epimerized by the racemase thereby interconverting the R and S ibuprofen stereoisomers. Finally, the thioester is hydrolyzed to release the ibuprofen acid. Because the reactive ibuprofen thioester leaves the active site of one enzyme before entering the next, we hypothesized that it might result in the covalent modification of proteins through reaction of the protein nucleophiles with the ibuprofen thioester. This hypothesis is supported by the fact that other CoA thioesters, like acetyl-CoA, have been shown to non-enzymatically modify lysine residues when they build up to sufficient concentrations.³ Additionally, simple thioester probes have been developed to directly monitor this type of chemical modification and show significant protein labeling at μM concentrations.⁴ While the exact concentration of ibuprofen in tissues is difficult to measure, μM concentrations in plasma are achieved after the recommended dosage of 400 mg in adults⁵ and some patients on chronic ibuprofen regiments are receiving significantly higher doses with correspondingly higher plasma concentrations.⁶

Figure 1. Ibuprofen metabolism and probe synthesis.

a) Inactive R-ibuprofen can be enzymatically racemized to the active S-enantiomer through a CoA-thioester. b) Structure of our alkyne-analog of ibuprofen “IbuAlk.” c) Synthesis of IbuAlk.

Here, we tested this hypothesis through the synthesis and evaluation of an alkyne analog of ibuprofen (IbuAlk, Figure 1b). We reasoned that any covalent protein-labeling could be readily visualized/identified using a subsequent copper(I)-catalyzed azide-alkyne clycoaddition (CuAAC) reaction.⁷ This approach is essentially identical to our development and the subsequent application of an alkyne analog of the activated phenolic ester of aspirin,^8–10 another NSAID. Incubation of IbuAlk with cell- and liver-lysates initially resulted in the robust labeling of multiple proteins, supporting our hypothesis. However, subsequent experiments will live cells resulted in no protein modification. Following up these seemingly contradictory results, we found that ibuprofen likely “sticks” to proteins during our lysate protocol, resulting in background covalent modification of proteins during the CuAAC reaction. However, we also found this false-positive signal can be alleviated through more rigorous protein precipitation and washing before CuAAC. Our results suggest that ibuprofen does not result in notable protein modification during its metabolism and that care should be taken to avoid background reaction of aromatic compounds with proteins during CuAAC-based ligand discovery experiments.

To generate our ibuprofen probe, we chose to introduce an alkyne due to its small size and robust bioorthogonal labeling through the CuAAC reaction. To preserve the reactivity of the carboxylic acid toward the 2-arylpropionyl-coenzyme A epimerase, we chose to place the alkyne group at the end of the isobutyl chain found on the other end of ibuprofen (Figure 1b), which we also reasoned would be easily accessible for CuAAC reactions. To generate IbuAlk, we took the total synthesis route shown in Figure 2. First, we performed a Heck arylation of a methyl-subsituted allylic alcohol to give aldehyde derivative 1 with high yield.¹¹ Next, we reduced the aldehyde to alcohol 2, as we were concerned that the aldehyde would not tolerate the conditions encountered along the synthetic route. After optimization, the Friedel-Craft acylation of the compound 2, allowed us to introduce the acyl group and also, to protect the free hydroxyl with the acetyl to give 3. Next, the cyanohydrine 4 was generated by treatment with titanium chloride and trimethyl silane cyanide. The first attempt of the cyanide hydrolysis under acid conditions led to the E2 elimination of the acetyl group, to form the correspondent isobutyl alkene. However, under alkaline conditions, we observed no elimination of the acetyl group, and the cyanide was fully hydrolyzed into the acid derivative 5 in excellent yield. Next, hydrogenation in presence of acetic and sulfuric acid yielded the hydroxy-ibuprofen compound 6. The carboxylic acid was protected as the corresponding methylester 7, and we then used pyridinium chlorochromate to reintroduce the aldehyde. Finally, we utilized the Bestmann-Ohira reagent to convert aldehyde 8 into alkyne derivative and hydrolyzed the methyl ester to give IbuAlk in 10% overall yield from 10 steps.

Figure 2. Characterization of IbuAlk protein labeling.

a) IbuAlk displays dose-dependent labeling of rat liver lysates. Non-denatured lysates were treated with the indicated concentrations of IbuAlk before CuAAC and in-gel fluorescence. b) Time dependence of IbuAlk labeling. Rat liver lysates were treated with IbuAlk (1mM) for the indicated lengths of time before CuAAC and in-gel fluorescence. c) Ibuprofen pre-treatment does not strongly compete IbuAlk labeling. Rat liver lysates were treated with the indicated concentrations of ibuprofen for 30 min before addition of IbuAlk (1 mM) for an additional 30 min. Labeling was visualized by CuAAC and in-gel fluorescence. d) IbuAlk does not labeling living cells. Huh-7 and HEK293 cells were treated with IbuAlk (1 mM) or DMSO vehicle for 18 h before lysis, CuAAC, and in-gel fluorescence. e) IbuAlk labels cell and liver lysate. Non-denaturing cell or rate liver lysate was incubated with IbuAlk (1 mM) or DMSO vehicle for 30 min before CuAAC and in-gel fluorescence. f) Protein denaturation does not prevent IbuAlk labeling. Rat liver lysates were either left non-denatured or subjected to protein unfolding conditions (4% SDS and 100 °C) before IbuAlk treatment, CuAAC, and in-gel fluorescence.

Previous studies have shown high Co-A epimerase activity in rat liver,¹² for this reason, we prepared non-denatured lysates from fresh rat livers and incubated them with a range of IbuAlk concentrations for 2 h. We then performed CuAAC with tetramethylrhodamine azide (TAMRA-azide) and analyzed any protein labeling by in-gel fluorescence (Figure 2a). Supporting our initial hypothesis, we observed dose-dependent labeling of multiple proteins by IbuAlk. We then performed an analogous timecourse experiment using 1 mM IbuAlk and observed the strongest labeling at 30 min of incubation (Figure 2b). With this timing in mind, we next attempted to compete IbuAlk labeling by pretreatment of lysates with increasing concentrations of ibuprofen for 30 min. While we did see some amount of competition (Figure 2c), it was never more than ~50%, making us question the nature of the labeling.

To further explore these inconclusive results, we next treated live Huh7 (human hepatocytes) or HEK293 (human embryonic kidney) with IbuAlk (1 mM), followed by lysis, CuAAC with TAMRA-azide, and in-gel fluorescence (Figure 2d). Treatment of neither cell-line resulted in labeling above background, which we rationalized might be explained by a lack of CoA synthetase and racemase in these particular cell-lines or the low membrane permeability of the IbuAlk. Therefore, we next subjected non-denatured lysates from rat liver, Huh7, or HEK293 cells to IbuAlk (1 mM) before CuAAC and in-gel fluorescence and observed notable labeling potentially consistent with low cell-permeability of IbuAlk (Figure 2e). However, given the high availability of ibuprofen to multiple tissues in the body, we felt that these results instead indicate some sort of background “false-positive” labeling in cell lysates. To begin to test this possibility, we examined the enzyme-dependence of IbuAlk labeling. Specifically, we generated non-denatured liver lysates and either treated it directly the IbuAlk or subjected it to denaturation (4% SDS with boiling) for 20 or 30 min prior to IbuAlk treatment. We observed very little loss of IbuAlk labeling by in-gel fluorescence upon protein denaturation (Figure 2f), demonstrating that CoA synthetase activity is not required for IbuAlk signal.

With this data in hand, we choose to finally investigate the nature of IbuAlk background labelling in lysates and potentially develop a protocol to avoid this false positive result. During the CuAAC protocol, we precipitate and wash proteins using cold methanol (−80 °C) to remove any unreacted IbuAlk remaining free in solution. Scrutiny of the IbuAlk structure, suggested to us that the probe may co-precipitate with the proteins and the side reaction might largely originate from the aromatic ring during the subsequent CuAAC reaction. To test this possibility, we compared IbuAlk to phenyl-1-propyne and cyclohexyl-1-propyne in liver lysates using the same MeOH precipitation protocol. Supporting our hypothesis concerning the importance of the aromatic ring for background labeling, we observed robust signal from both IbuAlk and phenyl-1-propyne but notably reduced signal form cyclohexyl-1-propyne (Figure 3a). We then examined whether alternative precipitation conditions and washing might reduce IbuAlk background lysate labeling. A common alternative to cold methanol is a mixture of chloroform and methanol. Accordingly, we treated non-denatured rat liver lysate with IbuAlk and subjected it to the two different precipitation protocols, as well as additional washes using chloroform/methanol, and found that extensive washing did indeed reduce protein labeling (Figure 3b). Finally, we confirmed that this washing approach might be generalizable by testing whether we could reduce the protein labeling of phenyl-1-propyne. Gratifyingly, we found that aggressive washing with chloroform and methanol did indeed notably reduce the labeling signal compared to simple methanol washing (Figure 3c), suggesting that this approach will work for other probes.

Figure 3. IbuAlk labeling likely results from insufficient washing and aromatic reactivity during CuAAC.

a) IbuAlk and phenyl-1-propyne result in notable protein labeling while cyclohexyl-1-propyne does not. Non-denatured rat liver lysates were treated with the indicated small molecules (1 mM) for 30 min before CuAAC and in-gel fluorescence. b) Thorough washing removes most of the IbuAlk labeling. Non-denatured rat liver lysates were treated with IbuAlk (1 mM) for 30 min before being subjected to different precipitation and/or washing conditions prior to CuAAC and in-gel fluorescence. c) The thorough washing protocol also reduces the background labeling of phenyl-1-propyne. Lysates were treated as in b with phenyl-1-propyne in place of IbuAlk.

In summary, our results strongly suggest that while ibuprofen is activated as a CoA thioester for racemization, it likely does not build-up to concentrations necessary for notable protein labeling (Figure 4a). This is supported by several results. First, we could not label proteins in living cells despite the robust modification of many proteins when lysates from the same cells are treated (Figures 2d&e). Second, we could not compete IbuAlk labeling with ibuprofen (Figure 2c). Finally, denaturation of the lysates did not prevent labeling, effectively ruling out the need for an enzyme like acetyl-CoA synthase for protein labeling (Figure 2f). Notably, Whitby and co-workers also observed increased protein labeling by a tienilic acid probe in liver lysates compared to in vivo experiments. While the in vivo labeling is due to liver metabolism, the liver lysate experiments were not subjected to precipitation or washing, raising the possibility that some of the observed difference might be due to the same background reaction.¹³

Figure 4. Our model that explains the results.

a) The on-target labeling pathway would involve the generation of an IbuAlk thioester that then reacts with nucleophilic residues (e.g., lysine). CuAAC with TAMRA-azide would then detect the resulting labeling. Our data suggests that this pathway does not result in detectable levels of protein labeling. b) The off-target labeling pathway involves the co-precipitation of IbuAlk with proteins. Curing subsequent CuAAC, the aromatic ring of IbuAlk reacts in an unknown way with proteins in parallel to the normal azide-alkyne cycloaddition reaction.

Interestingly, while performing our investigation we observed what turned out to be unexpected background labelling caused by what appears to be co-precipitation of the probe and proteins and by a CuAAC side reaction (Figure 4b). We believe that the co-precipitation probably results from hydrophobic interactions between the hydrocarbon-rich ibuprofen and newly exposed cores of proteins. Follow up experiments demonstrated that this background labeling is likely due to the aromatic ring in IbuAlk, as phenyl-1-propyne but not cyclohexyl-1-propyne also gave significant protein labeling in cell lysates (Figure 3a). We do not think that this difference results from a difference in the co-precipitation of these two compounds but rather from the chemistry that the aromatic ring can undergo during the CuAAC reaction. We do not know exactly what these reactions are or where they occur on proteins (e.g., specific side-chains, backbones, etc.). However, proteomic analysis using appropriate software packages like MSFragger^14,15 should be able to better characterize the types and locations of modifications resulting from compounds like IbuAlk, as has been recently demonstrated for a library of electrophiles.¹⁶ We then reasoned that washing the cell pellet may remove a significant fraction of the co-precipitated IbuAlk, and this proved to be correct in the case of both IbuAlk (Figure 3b) and phenyl-1-propyne (Figure 3c). While these results are somewhat disappointing, we hope that this cautionary tale will be of use to the bioorthogonal probe community and suggest using aggressive washing conditions to remove most of the unreacted probe material before the CuAAC reaction.