Identification of Aldo-Keto Reductase AKR1B10 as a Selective Target for Modification and Inhibition by Prostaglandin A₁: Implications for Antitumoral Activity

Cyclopentenone prostaglandins (cyPG) are endogenous reactive eicosanoids which display varied biological actions including inhibition of proinflammatory gene expression and modulation of cell proliferation and redox status (1–3). cyPG arise from the spontaneous dehydration of various PG or from nonenzymatic peroxidation of arachidonic acid. These eicosanoids possess an α,β-unsaturated carbonyl group in the cyclopentene ring which confers them a high reactivity toward nucleophiles, such as thiol groups, and can lead to the formation of covalent adducts by Michael addition. This property is essential for cyPG biological actions (4), including PPAR activation (5) and inhibition of NF-κB and AP-1 transcription factors (1, 6). A- and J-series cyPG were initially observed to share several protein targets and biological actions (7–9). However, recent studies have shown that cyPG with different structures may target selective subsets of cellular proteins (10), and even different residues within the same protein. Ras proteins and histone deacetylases constitute examples of this intra- and intermolecular selectivity (9, 11). Identification of the factors involved in the selectivity of protein modification by different cyPG, as well as of the selectively modified targets, could unveil potential opportunities for drug discovery.

Using proteomic approaches we have identified proteins of the aldo-keto reductase (AKR) family as targets for selective modification by PGA₁ compared with 15-deoxy-Δ^12,14-prostaglandin J₂ (15d-PGJ₂). AKR enzymes are involved in the metabolism of several ketones and aldehydes and play important roles in pathophysiology (12). Among them, AKR1B10 exerts key roles in tumor biology. AKR1B10, first identified in human hepatocarcinoma (13), is overexpressed in colorectal, breast and lung cancer, for which it is considered a diagnostic marker (14, 15), and may play a pathogenic role in hepatocellular carcinoma (16) and in tobacco-related carcinogenesis (14, 17). Proposed AKR1B10-mediated tumorigenic mechanisms include retinoic acid depletion and cancer cell dedifferentiation (18, 19) as well as chemoresistance due to metabolism of carbonyl group–bearing anticancer drugs (20). Therefore, AKR1B10 constitutes a primary target for the development of inhibitors with anticancer potential.

Here we have confirmed the direct modification of AKR1B10 by PGA₁-B and show that this cyPG inhibits AKR activity. Moreover, PGA₁ treatment potentiated the effects of the carbonyl group–containing drug doxorubicin. These results provide a novel approach for the inhibition of AKR1B10 which could be exploited in the design of strategies to overcome cancer chemoresistance.

Prostanoids and their biotinylated analogues [15d-PGJ₂-biotin (15d-PGJ₂-B) and prostaglandin A₁-biotin (PGA₁-B)] were from Cayman Chemical. Horseradish peroxidase (HRP)–streptavidin and enhanced chemiluminescence (ECL) reagents were from GE Biosciences. Neutravidin-agarose was from Pierce. Anti-aldose reductase (sc-33219) and anti-GAPDH (sc-32233) were from Santa Cruz Biotechnology and anti-AKR1B10 from Origene.

NIH-3T3, H1299, and COS-7 cells were from ATCC. A549 and Calu-3 cells were gifts of Dr. F. Rodríguez-Pascual (21, 22) and Dr. A. Silva, respectively. All cells were used within 6 months after resuscitation. NIH-3T3 and COS-7 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% FBS, 2 mmol/L glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin. A549 cells were grown in RPMI1640 with the above supplements plus 50 μg/mL gentamycin. For treatments, cells at a confluence of 80% to 90% (COS-7) or 60% to 70% (A549) were incubated in serum-free medium, unless otherwise stated. Cell fractionation into S100 and P100 fractions was done as described (23).

S100 fractions from control or PGA₁-B–treated NIH-3T3 cells were purified on neutravidin beads and analyzed by liquid chromatography/tandem mass spectrometry (LC/MS-MS; ref. 23). Total cell lysates (100 μg of total protein) from control, and 15d-PGJ₂-B- or PGA₁-B–treated NIH-3T3 fibroblasts were analyzed by 2D-electrophoresis as described (24). Gels run in duplicate were either transferred to Immobilon-P membranes to specifically detect proteins which had incorporated the biotinylated PG with HRP–streptavidin and ECL or stained with colloidal Coomassie blue. Gel images were aligned and spots of interest were excised and proteolyzed with trypsin as detailed previously (23, 25). Digests (0.5 μL) were spotted onto a MALDI target in 1 μL of matrix (2,5-dihydroxybenzoic acid). Matrix-assisted laser desorption/ionization–time-of-flight mass spectrometry (MALDI-TOF MS) was carried out on a calibrated Ultraflex time-of-flight mass spectrometer in reflectron mode (Bruker Daltonik). All mass spectra were internally calibrated with trypsin autolysis peaks at m/z 842.51 and m/z 2,211.10. Peptide masses were searched against the IPI Human database using MASCOT version 2.0.02 (Matrix Sciences). Identifications were accepted when a minimum of 6 peptide masses matched a particular protein (mass error of ± 50 ppm allowing 1 missed cleavage), sequence coverage was more than 30%, MOWSE scores were higher than the threshold value (P = 0.05), and the predicted protein mass agreed with the gel-based mass.

The 1.65 Å resolution crystal structure of NADP⁺-bound human aldose reductase (AKR1B1) was used as a model (PDB: 1ADS code). Addition of missing hydrogen atoms and computation of the protonation state of ionizable groups at pH 6.8 were carried out by using the H++ Web server (26), which relies on AMBER (27) force-field parameters and finite difference solutions to the Poisson–Boltzmann equation. The charge distribution for PGA₁-B, in which B was replaced by a methyl group, was obtained by fitting the quantum mechanically calculated (RHF 6-31G*\\3-21G*) molecular electrostatic potential to a point charge model, as implemented in Gaussian 03 (Gaussian, Inc.).

The Lamarckian genetic algorithm implemented in AutoDock 3.0.5 (28) was used to generate automated poses of PGA₁ docked within the binding site of AKR employing a dielectric constant of 2. Consistent bonded and nonbonded AMBER “gaff” parameters (parm03) were assigned to PGA₁ atoms and a molecular dynamics (MD) simulation of the selected complex in the presence of explicit water (a truncated octahedron containing 4 Na⁺ ions to achieve electroneutrality) was run for 5 nanoseconds by using periodic boundary conditions. Initially, positional restraints of 5 kcal/mol on the protein α-carbons and harmonic restraints on the distances between the keto-group of PGA₁ and His111 N^ϵ and between the electrophilic carbon of PGA₁ and the sulfur atom of Cys299 were applied. After 1 nanosecond, all the restraints were removed and the solvated complex was allowed to evolve freely.

The plasmid pCMV6-XL5 containing human AKR1B10 ORF was from Origene. The insert was cloned into the pCEFL-KZ-AU5 plasmid to yield AU5-AKR1B10 wild type (wt). The C299S, H111A, and Y49F mutants were generated by site-directed mutagenesis by using oligonucleotides, 5′-CTGGAGGGCCTCTAACGTGTTGCAATCCTCTCATTTGG-3′, 5′-GGACGTCTATCTTATTGCCTGGCCACAGGGATTCAAGTCTGG-3′, and 5′-CTGTGCCTATGTCTTTCAGAATGAACATGAAGTGGG-3′, and their complementary reverse oligonucleotides, respectively. All constructions were sequenced. COS-7 cells in 6-well dishes were transfected with Lipofectamine 2000, using 2 μg of DNA per well.

Cells were lysed in 50 mmol/L sodium phosphate buffer (pH 6.8), and S100 fractions were used for AKR activity measurements. Assay mixtures contained 50 μg of protein, 10 mmol/L d,l-glyceraldehyde, and 200 μmol/L NADPH in 100 mmol/L sodium phosphate buffer (pH 6.8). After 20-minute incubation at room temperature NADPH consumption was monitored by measuring the absorbance at 340 nm in an Ultrospec 4300 pro spectrophotometer (GE Biosciences). Enzymatic activity was expressed in nmol/min/mg of protein.

Intracellular doxorubicin was detected by confocal fluorescence microscopy on a Leica SP5 confocal microscope, using excitation and emission wavelengths of 470 and 590 nm (29). Doxorubicin fluorescence was quantitated by flow cytometry on an EPICS Coulter analyzer, using the above excitation and emission settings.

For wound-healing assays cells were grown to confluence and the monolayer was scratched with a 200 μL pipette tip. The width of the scratch was measured at 3 points in each well, and at various time points after treatment with the indicated agents. Relative migration was calculated as the percentage of distance migrated compared with the control well. Anchorage-independent growth was assessed as described (30). Cells were cultured for 3 weeks with 3 weekly medium replacements. Colonies were visualized by ethidium bromide staining on an UV transilluminator.

Cells treated with the various agents were detached with trypsin/EDTA and fixed with 70% ethanol in PBS. Immediately before flow cytometry analysis, cells were centrifuged and resuspended in 0.5% NP-40, 25 μg/mL propidium iodide and 25 μg/mL RNase A, and incubated in this solution for 15 minutes at room temperature. Apoptosis was detected by binding of Annexin V and propidium iodide staining, as described (31).

Experiments were carried out at least 3 times. Results are presented as average values ± SEM. Average values were compared by Student's t test for unpaired observations.

We have previously reported that biotinylated analogues of the cyPG 15d-PGJ₂ and PGA₁, selectively modify specific subsets of cellular proteins in NIH-3T3 fibroblasts (10). To identify selective targets of each cyPG we incubated cells with concentrations of 15d-PGJ₂-B or PGA₁-B resulting in a similar extent of total protein modification (10), and subsequently carried out subcellular fractionation. We detected a prominently labeled band at 37 kDa, appearing selectively in S100 fractions from PGA₁-B–treated fibroblasts (Fig. 1A). To identify this selective PGA₁-B target we employed a proteomic strategy following the enrichment of biotin-containing proteins by avidin chromatography, as schematized in Figure 1B and detailed in Supplementary Figure S1. Consistent with our previous work, initial proteomic analysis of this band identified glyceraldehyde-3-phosphate dehydrogenase (GAPDH; ref. 23). However, immunoprecipitation of GAPDH did not recover the PGA₁-B–selective target, which remained in the nonretained fraction (Supplementary Fig. S1), indicating that the presence of the abundant protein GAPDH was hampering identification of the comigrating PGA₁-B target. Consequently, combining GAPDH immunodepletion, avidin chromatography, and LC/MS-MS analysis (Fig. 1B; and Supplementary Fig. S1) allowed the identification of aldose reductase (AKR1B1) as a putative selective target for PGA₁-B (Supplementary Fig. S1). To directly evidence covalent addition of PGA₁-B to AKR enzymes we conducted 2D-electrophoresis. Samples from control, 15d-PGJ₂-B, and PGA₁-B–treated cells were resolved in duplicate gels used for transfer and detection of AKR proteins or biotin-positive spots, or for Coomassie blue staining and protein identification (Fig. 1C). Two spots were detected by the anti-AKR antibody; spot 1 showed pI and apparent molecular weight consistent with AKR1B1; MALDI-TOF MS analysis of spot 2 identified aldose reductase-related protein-2 (AKR1B8), an AKR family member expressed in murine fibroblasts. Both spots matched biotin positive spots only in the sample from PGA₁-B–treated cells (Fig. 1C, right), confirming the selectivity of their modification by PGA₁-B. Moreover, the presence of AKR enzymes in the avidin-bound fraction from PGA₁-B–treated cells, but not from control cells was confirmed by Western blot (Fig. 1D).

In our proposed binding mode for PGA₁ in the AKR1B1 active site, the keto-group is accepting hydrogen bonds from both the N^ϵ-protonated His111 and the hydroxyl of Tyr49 (Fig. 2A). These interactions, which remained stable during the unrestrained MD simulations, facilitate the orientation of the cyclopentenone ring in such a way that the electrophilic carbon is at a suitable distance from the sulfur atom of Cys299 (i.e., ≤4.0 Å) thus making it possible for the Michael reaction to proceed and yield a covalent adduct. Furthermore, in the case of the biotinylated analogue, PGA₁-B, the amide bond between PGA₁ and the spacer (Fig. 4) can establish an additional hydrogen bonding interaction with the side-chain carboxamide of Gln50 whereas the biotin moiety would be found facing the solvent. Consistent with the observed selectivity of the modification, the interactions predicted by this model could not be established in the case of the J series cyPG, 15d-PGJ₂ (data not shown).

The AKR family consists of more than 100 members, and significant homology exists among many of them (Fig. 2B). Moreover, the residues predicted by the molecular model to be involved in the interaction of PGA₁ with the enzyme are highly conserved among several members of this family. Given the high degree of sequence identity between AKR1B1 and AKR1B10, especially within the active site region (Fig. 2B), the mode of binding of PGA₁ and PGA₁-B to these 2 enzymes is expected to be identical. Given its significance in tumor biology, we used AKR1B10 for subsequent studies addressing the interactions of PGA₁ with AKR enzymes and their functional implications.

PGA₁-B clearly formed a stable adduct with AKR1B10 in cells transiently transfected with an AU5-AKR1B10 construct, as assessed by SDS-PAGE and biotin detection (Fig. 3A). Remarkably, mutation of Cys299 to Ser virtually abolished the incorporation of PGA₁-B into AKR1B10, indicating that this residue is critical for PGA₁-B binding. Interestingly, mutation of either Tyr49 or His111 reduced the incorporation of PGA₁-B, suggesting that the presence of these residues favors the interaction of the cyPG with the active site of the enzyme as predicted by the model.

We next addressed the functional consequences of AKR1B10 modification by PGA₁. Endogenous AKR activity was negligible in COS-7 cells (see Fig. 3C). Transfection of AU5-AKR1B10 produced measurable AKR activity, which was inhibited by treatment with either PGA₁ or PGA₁-B by more than 60% (Fig. 3B), without altering AU5-AKR1B10 protein levels. In contrast, 15d-PGJ₂-B at a concentration eliciting a degree of total protein modification similar to that achieved with PGA₁-B, induced only marginal AKR1B10 labeling, and a weak inhibition of AKR activity (15%) was observed in 15d-PGJ₂–treated cells (data not shown). Remarkably, PGA₁-elicited inhibition was abolished in cells transfected with AKR1B10 Cys299Ser, confirming the importance of Cys299 for both PGA₁ binding and inhibitory effect (Fig. 3C). Consistent with previous observations, the AKR1B10 Cys299Ser mutant displayed lower activity than the wt variant, reportedly due to reduced affinity for the d,l-glyceraldehyde substrate (32). Mutation of Tyr 49 or His111 abolished activity (data not shown), confirming their importance in catalysis (33).

Given the potential biomedical interest in AKR inhibitors, we characterized PGA₁ effect in more detail. PGA₁ inhibition of AKR1B10 activity in COS-7 cells was time- and concentration-dependent (IC₅₀, 43 μmol/L; Supplementary Fig. S2). Several PGA₁-related compounds including PGA₂, 8-iso-PGA₁, and 8-iso-PGA₂ inhibited AKR1B10 activity (Fig. 4). Interestingly, the isoprostanes 8-iso-PGA₁ and 8-iso-PGA₂, isomers of the respective PG, were significantly less effective than their PG counterparts, suggesting that the trans orientation of the PG side chain with respect to the prostane ring is more favorable for interaction with the enzyme. Remarkably, 2-cyclopentenone did not inhibit AKR1B10, suggesting that the chains of the PG are important for docking at the active site. Of the 2 reported inhibitors of AKR enzymes, fenofibrate (34) and AD-5467 (35), only the latter showed a potent inhibitory effect. AKR and actin levels were not affected under the conditions of the assay (Fig. 4).

In search for a biologically relevant model to assess PGA₁ effects, we explored the levels of AKR1B10 in A549, H1299, and Calu-3 lung cancer cells, of which only A549 cells showed detectable levels (Supplementary Fig. S3). PGA₁ reduced cell viability and induced cell-cycle alterations in all cells studied to variable extents (Supplementary Fig. S3). Consistent with a role of AKR1B10 in detoxification and cell survival, A549 cells showed reduced susceptibility to PGA₁ effects. Treatment of A549 cells with PGA₁-B elicited the selective retention of endogenous AKR proteins, and specifically of AKR1B10, on avidin–agarose beads (Fig. 5A), indicating that PGA₁-B modifies AKR enzymes in this cell type. Moreover, PGA₁ and PGA₁-B effectively inhibited AKR activity without altering AKR protein levels (Fig. 5B). According to previous evidence, reduced glutathione (GSH) forms stable adducts with PGA₁-type PG potentially reducing interactions with proteins (36). We previously showed that GSH depletion enhances PGA₁-B binding to cellular proteins (10). Consistent with this, treatment of A549 cells with buthionine sulfoximine (BSO), a GSH synthesis inhibitor, improved PGA₁-B incorporation into proteins and increased retention of biotin-tagged proteins onto neutravidin beads (Supplementary Fig. S4). Moreover, GSH depletion specifically increased AKR retention on avidin after PGA₁-B treatment (Fig. 5C). Furthermore, BSO showed a tendency to reduce AKR activity per se and significantly potentiated PGA₁-elicited AKR inhibition (Fig. 5D). Taken together, these results indicate that maintenance of GSH levels or of redox status is important for AKR function. In addition, GSH depletion facilitates AKR modification and inhibition by PGA₁.

PGA₁ inhibited A549 cell migration in a wound healing assay (Fig. 6A) and reduced anchorage-independent clonogenic growth (Fig. 6B), thus suggesting an inhibition of tumorigenic potential. AKR1B10 is involved in the metabolism of a variety of anticancer compounds, including doxorubicin and daunorubicin, potentially leading to chemoresistance (32, 37). Therefore, we explored whether PGA₁ could improve doxorubicin efficiency. The main mechanism for doxorubicin action at clinically relevant concentrations is the stabilization of a cleaved ternary topoisomerase II–doxorubicin–DNA complex, resulting in a blockade in the G₂ phase of the cell cycle (38–40). Doxorubicin treatment induced a concentration-dependent increase in the proportion of A549 cells present in the G₂–M phases of the cell cycle (Fig. 6C, left), an effect that was clearly potentiated by PGA₁ (Fig. 6C, right). This effect was associated with an improvement of doxorubicin availability in cells. PGA₁ clearly potentiated intracellular accumulation of doxorubicin at several concentrations of the drug, as assessed by flow cytometry (Fig. 6D). Confocal fluorescent microscopy confirmed PGA₁-elicited potentiation of doxorubicin retention, mainly at the nuclear compartment (Supplementary Fig. S5). The AKR inhibitor AD-5467 mimicked the effects of PGA₁ both on cell-cycle distribution and doxorubicin accumulation (data not shown). Taken together, these results suggest that PGA₁ treatment through inhibition of AKR activity in A549 cells, may increase the accumulation of the AKR substrate doxorubicin and potentiate its biological effects, helping to counteract multidrug chemoresistance.

Proteomic studies are shedding light into the signaling and targets of endogenous electrophiles, unveiling novel mechanisms of action and providing opportunities for drug discovery (23). Our results identify AKR1B10, an enzyme involved in tumor development and cancer chemoresistance, as a selective target for modification and inhibition by A-class cyPG. Moreover, we showed that PGA₁ inhibits AKR activity in lung cancer cells, reduces their migration and anchorage-independent growth, reflecting reduced invasiveness, and potentiates the effect of the antitumoral agent doxorubicin, a known AKR1B10 substrate. Therefore, the PGA₁–AKR1B10 interaction could be explored for the development of novel strategies against tumor chemoresistance.

Insight into the structural basis for the preferential interaction of PGA₁ with AKR enzymes was obtained by molecular docking and confirmed by biochemical characterization of site-specific mutants of AKR1B10. AKR1B10, like other AKR enzymes, is a monomeric protein with an (α/β)₈ barrel structure containing the active site. Residues His111, Tyr49, K79, and Asp44 line the active site, play a key role in catalysis, and are highly conserved in the AKR family (12). Cys299 is located close to the active site and modulates the interaction of substrates and inhibitors with the enzyme. Molecular modeling studies postulated the formation of a Michael adduct between PGA₁ and AKR involving Cys299, with Tyr49 and His111 playing a facilitating role. Interestingly, mutagenesis studies confirmed the requirement for Cys299 for covalent addition of PGA₁-B to AKR1B10, whereas Tyr49 and His111 were required for optimal binding. Cys299 was also required for PGA₁-mediated inhibition of AKR1B10, confirming the functional importance of the interaction. Importantly, both PGA₁ and PGA₁-B displayed similar inhibitory effects, indicating that, in agreement with the modeling results, the biotin moiety does not interfere with the interaction. Indeed, we have confirmed modification and inhibition of recombinant AKR1B1 by both compounds (B.D-D., unpublished data).

Given the important implications of AKR1B10 in tumor development and cancer chemoresistance, the search for inhibitors of this enzyme is a thriving area of research. Several compounds have been found to display AKR1B10 inhibitory properties in vitro, including some nonsteroidal anti-inflammatory drugs (41), butein (42), and curcumin derivatives (43). Our results suggest that PGA-type cyPG may constitute a novel class of AKR inhibitors. PGA₁ binding to AKR1B10 shows unique characteristics among several recently identified inhibitors for this enzyme (41, 43), for which Gln114, Val301, and Gln303 or Ser304 were found to be the critical interacting residues. It would be interesting to explore whether modifications of the PGA₁ structure could be designed to increase the specificity or potency of the inhibition. Concentrations of PGA₁ used in this work should be considered pharmacologic because they are at least 3 orders of magnitude higher than those measured in biological systems (44).

A549 cells constitute a widely used cellular model of lung carcinoma and express several AKR isoforms, of which AKR1B10 is the most abundant (45). As AKR1B10 is involved in detoxification of various carbonyls, it could also confer increased resistance to PGA₁ or to reactive species generated during PGA₁ treatment. The finding that blocking GSH synthesis improves PGA₁-elicited modification and inhibition of AKR enzymes suggests that GSH depletion could constitute a strategy to improve the antitumoral action of PGA₁-type cyPG. We have also observed that AKR inhibition in A549 cells is associated with an improvement of the effects of doxorubicin, a widely used drug in the therapy of lung (46) and other cancers (47). Doxorubicin clinical efficacy, however, is limited by the development of cancer cell resistance due to induction of drug detoxification mechanisms, including conversion by AKR enzymes to the less effective metabolite doxorubicinol (37). It should be noted that although most studies on doxorubicin action have been conducted with micromolar concentrations, the effects herein reported occur at nanomolar doxorubicin concentrations. This is closer to the levels measured in the plasma of patients receiving the drug (25–250 nmol/L) and that expected to occur in vivo within the tumor microenvironment (48). The fact that the AKR inhibitor AD-5467 mimics PGA₁-elicited potentiation of doxorubicin actions suggests that inhibition of AKR may be involved in PGA₁ effects. Nevertheless, because PGA₁ may bind to multiple cellular targets (23), it could exert other cellular actions contributing to the improvement of doxorubicin effectiveness, including induction of reactive oxygen species, modulation of GST activity or GSH content, and interference with drug efflux. These possibilities will be the subject of further studies.

In summary, our study identifies AKR enzymes as prominent PGA₁-selective targets. Moreover, PGA₁ binds covalently and inhibits AKR1B10 establishing specific interactions with the enzyme and potentiates the effects of the antitumoral drug doxorubicin. These results unveil novel possibilities to overcome cancer chemoresistance.

D. Pérez-Sala and B. Díez-Dacal are coinventors of a related patent applied for by CSIC. The other authors disclosed no potential conflicts of interest.

This work was supported by grants SAF2009-11642 from MiCInn and RETICS RD07/0064/0007 from ISCIII to D. Pérez-Sala and SAF2006-12713-C02-02 (CICYT) and S-BIO/0214/2006 (CAM) to F. Gago. J. Gayarre was the recipient of fellowships from CSIC (I3P), EMBO, and FEBS (short term fellowships). C. Coderch was supported by a MiCInn FPU (AP2007-01225) fellowship. Part of this work was undertaken at UCLH/UCL who received a proportion of funding from the Department of Health's NIHR Biomedical Research Centres funding scheme. Feedback from COST Action CM1001 is appreciated.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.