Rhein Inhibits AlkB Repair Enzymes and Sensitizes Cells to Methylated DNA Damage*

The AlkB repair enzymes, including Escherichia coli AlkB and two human homologues, ALKBH2 and ALKBH3, are iron(II)- and 2-oxoglutarate-dependent dioxygenases that efficiently repair N1-methyladenine and N3-methylcytosine methylated DNA damages. The development of small molecule inhibitors of these enzymes has seen less success. Here we have characterized a previously discovered natural product rhein and tested its ability to inhibit AlkB repair enzymes in vitro and to sensitize cells to methyl methane sulfonate that mainly produces N1-methyladenine and N3-methylcytosine lesions. Our investigation of the mechanism of rhein inhibition reveals that rhein binds to AlkB repair enzymes in vitro and promotes thermal stability in vivo. In addition, we have determined a new structural complex of rhein bound to AlkB, which shows that rhein binds to a different part of the active site in AlkB than it binds to in fat mass and obesity-associated protein (FTO). With the support of these observations, we put forth the hypothesis that AlkB repair enzymes would be effective pharmacological targets for cancer treatment.

The nucleic acids in living cells are subject to modification by both endogenous and environmental agents (1). Direct-acting chemicals constantly damage nucleic acids and generate various methyl lesions with mutagenic and/or cytotoxic consequences (2,3). O 6 -Methylguanine (O 6 mG) 2 and N 3 -methyladenine lesions have the highest potential for methylating damage by an SN1 agent such as N-methyl-NЈ-nitro-N-nitrosoguanidine (MNNG), which block replication and are thought to be toxic (4,5). For the most part, the SN2 agent such as methyl methane sulfonate (MMS) produces N 1 -methyladenine (m 1 A) and N 3 -methylcytosine (m 3 C) lesions in single-stranded DNA (ssDNA). Accumulation of these adducts can lead to cell death (6,7). Organisms have evolved several mechanisms to efficiently remove various methyl lesions, including suicidal methyltransferases, DNA glycosylases, and the AlkB family dioxygenases (see Fig. 1A) (8,9).
To date, AlkB repair appears to be the major natural defense mechanism with the power to restore the canonical base structure in vivo. Escherichia coli AlkB and its human homologues, ALKBH2 and ALKBH3, utilize iron(II) and 2-oxoglutarate (2OG) to achieve oxidative demethylation of m 1 A and m 3 C (see Fig. 1B) (10 -12). The lack of AlkB repair results in increased sensitivity to MMS, elevated level of mutations, and reduced cell proliferation (13)(14)(15)(16). Furthermore, the accumulation of m 1 A and m 3 C lesions could also occur on RNA. Research suggests that the oxidative demethylation in mRNA and tRNA acts as a part of AlkB or ALKBH3 repair to protect cells against MMS (17,18). The scope of substrates for AlkB repair has been largely extended to all simple N-alkyl lesions at the Watson-Crick base-pairing interface on the four bases (19), thus indicating the importance of oxidative demethylation for cell survival. In addition, human enzymes have been broadly linked to cancer. The housekeeping enzyme in mammalian cells, ALKBH2, plays a crucial role in pediatric brain tumors during chemotherapy treatment (20). Essential for prostate cancer progression, ALKBH3 presents a potential target for effective therapy in prostate cancer (21,22).
Structural characterizations of AlkB repair enzymes have provided insights into the understanding of demethylation mechanism and substrate recognition (23)(24)(25)(26). Similar to members of the 2OG-dioxygenase family, 2OG could inhibit AlkB at high concentrations (27). Recently, the method dynamic combinatorial chemistry, which is linked to mass spectrometric analyses, was used to identify AlkB inhibitors; these inhibitors showed improved potency and selectivity (28). A set of inhibitors of ALKBH3 were obtained from a random screening, and some of them significantly suppress tumor formation in a mice xenograft model (29). These data demonstrate that AlkB repair enzymes are amenable to potent inhibition by small molecules. What these studies lack, however, is 2-fold: one, a profile of the target engagement of inhibitor and two, a full elucidation of the inhibitor mode of action, which should provide a better understanding of the biological consequences of the replication-blocking m 1 A and m 3 C lesions (30).
As we reported in previous studies, the natural product rhein inhibits FTO demethylation of N 6 -methyladenine in vitro and elevates the level of N 6 -methyladenine within mRNA in HeLa cells (31,32). In this paper, we provide significant new data that describe the in vitro and in vivo effects of rhein as an inhibitor of AlkB repair enzymes.

Experimental Procedures
Chemicals, DNA Oligonucleotides, Antibodies, E. coli Strains, and Cell Culture-Rhein, MMS, MNNG, and temozolomide (TMZ) in cell biology grade were purchased from Sigma-Aldrich. Bromaminic acid (BA) and JIB-04 were purchased from Tokyo Chemical Industry Co. and Selleck Chemicals, respectively. The DNA oligonucleotide were synthesized on an Expedite DNA synthesizer (PerSeptive Biosystems). Antibodies anti-m 3 C, anti-ALKBH2, anti-ALKBH3, anti-H3K9me3, and anti-␤-actin were produced by Active Motif (61180), Sigma (SAB3500534), Millipore (09882), ABCAM (ab8898), and CWBIO (0096a), respectively. The rabbit polyclonal anti-AlkB antibody was generated by Shanghai Immune Biotech using the purified AlkB protein as the antigen. The E. coli AB1157 and HK82 strains were kind gifts from Dr. H. Krokan's lab (Norwegian University of Science and Technology). While this paper was in revision, the HK82 strain that has been widely used was reported to contain additional mutations affecting MMS sensitivity (33). U87 cells were purchased from the Cell Bank of Type Culture Collection of Chinese Academy of Sciences and cultured in minimum essential medium (41500034; Gibco) supplemented with 10% FBS.
Restriction Endonuclease Digestion Assay-The AlkB, ALKBH2, and ALKBH3 proteins were expressed and purified as described (34 -36). We cloned the expression plasmids of the C-terminal domain of Ada (C-Ada) and AlkA by incorporating the two genes into pET28a vector, and then the expression and purification of the two recombinant proteins followed known procedures (37,38). The DNA oligonucleotides used in the enzymatic assays were 5Ј-TAGACATTGCCATTCTCGATA-GG(m 1 A)TCCGGTCAAACCTAGACGAATTCCA-3Ј or 5Ј-ATTGCCATTCTCGATAGG(m 1 A)TCCGGTCAAACCTA-GACGAA-3Ј for AlkB or ALKBH3 repair, 5Ј-TGGAATTCG-TCTAGGTTTGACCGGATCCTATCGAGAATGGCAATG-TCTA-3Ј or 5Ј-TTCGTCTAGGTTTGACCGGATCCTATC-GAGAATGGCAAT-3Ј as the complementary DNA sequence for the duplex substrates of AlkB or ALKBH2 repair, 5Ј-GCC-ATTCTCGATAGGCGCA(O 6 mG)CTGAGCTCGCGTCCG-GTCA-3Ј complementary to 5Ј-TGACCGGACGCGAGCTC-AGCTGCGCCTATCGAGAATGGC-3Ј for Ada repair, and 5Ј-CGATAGCATCCTGCCTTCTCTCCAT-3Ј complementary to 5Ј-ATGGAGAGAAGGAAGGATGCTATCG-3Ј for AlkA repair, respectively. All reactions were run in 100-l scale. The conditions were 50 mM Tris-HCl (pH 7.5), 1 M DNA, 1 M enzyme, 0.28 m⌴ (NH 4 ) 2 Fe(SO 4 ) 2 , 2 mM L-ascorbic acid, and 50 M 2OG in AlkB repair, 50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 1 mM DTT, 1 M DNA, and 1 M C-Ada in Ada repair, and 50 mM Tris-HCl (pH 8.5), 1 mM EDTA, 1 mM DTT, 0.1 mg/ml bovine serum albumin, 5 M AlkA, 1 M DNA, and ionic strength adjusted to 100 mM with sodium chloride in AlkA glycosylase assay, respectively. The oxidative demethylation and C-Ada repair assays were run at room temperature for 1 h and quenched by heating at 50°C for 10 min, and then a DpnII or PvuII digestion was performed at 37°C. The AlkA assay was continually executed at 37°C for 24 h and heated at 70°C for 30 min in the presence of 0.2 M NaOH. All reactions were detected on 20% nonreducing PAGE. The gel was stained in the presence of GelRed (41003; Biotium) at room temperature for 0.5 h and then was photographed by UV light in a gel imaging system.
HPLC-based Quantification Assay-The reactions were run under the same conditions as described in the DpnII digestion assay in which a 15-mer ssDNA (5Ј-ATTGTCA(m 1 A) CAGCAGA-3Ј) was assayed at 5 M. After the reaction was terminated, ssDNA was digested by nuclease P1 (N8630; Sigma) and alkaline phosphatase (P4252; Sigma), and then the nucleosides were analyzed by HPLC using an Agilent Eclipse XDB-C18 column (4.6 mm ϫ 250 mm). The program was run with the following settings: mobile phase A (50 mM CH 3 COONH 4 , pH 5.0) and B (acetonitrile) at a flow rate of 1 ml/min. The IC 50 values were then calculated from the variation as a percentage of demethylation at different inhibitor concentrations using nonlinear regression, with a normalized dose-response fit on GraphPad Prim 5.0 using the equation, where Bottom and Top are the minimal and maximal inhibition percentages, respectively. All reactions were repeated in triplicate.
Plate Killing Assay-Overnight culture of E. coli was diluted to A 600 0.005 or 0.008 (transformed with pET28a). After serial dilution by 10-fold, each volume of 10 l of bacteria was spotted onto LB solid medium in the presence of rhein or MMS at varying concentrations, and isopropyl ␤-D-1-thiogalactopyranoside at 50 M was supplemented to induce AlkB overexpression. After 10 h, static cultures at 37°C, the plates were photographed with the gel imaging system (Tanon 2500R).
E. coli Colony-forming Unit Count-E. coli Gold, AB1157, or HK82 was cultured overnight and diluted 1:100 to fresh LB medium and grown to logarithmic phase. Then the bacteria were separated equally and grown for 5 h with MMS or rhein or both. The bacteria were diluted serially and cultured overnight on LB agar. The E. coli colonies were counted relative to the DMSO control group. All tests were performed in triplicate.
Western Blot and Dot Blot Analysis-Overnight E. coli culture was diluted 1:100 to fresh LB and incubated to the logarithmic phase. Compounds were added, and cultures were grown at 37°C with constant shaking at 230 rpm for 5 h. The cells were collected and divided into two aliquots. One aliquot was used to isolate genomic DNA by means of a PureLink genomic DNA mini kit (Invitrogen). 2-fold serial dilutions were loaded onto the nylon membrane and cross-linked by exposure to UV, and then m 3 C antibody was probed to the genomic DNA. The other aliquot was sonicated, and the supernatant was quantified by BCA assay. The equal 25-g cell extracts were separated by 15% SDS-PAGE. Both AlkB and m 3 C were visualized by SuperSignal West Femto kit (Thermo).
Cellular Thermal Shift Assay (CETSA)-After exposure to 500 M MMS with (for intact cell assay) or without rhein (for cell lysate assay), E. coli cells were collected and washed several times to avoid any excess compound residue. Then the cells were sonicated and centrifuged at 12,000 rpm for 20 min at 4°C to fractionize the supernatant. For the cell lysate assay, rhein was added to the supernatant and incubated at 20°C for 20 min. After denaturing at various temperatures for 5 min on Mastercycler pro PCR instrument (Eppendorf), samples were centrifuged to eradicate the precipitate. The supernatants were analyzed by Western blot, and the density of AlkB protein bands was plotted using GraphPad Prism 5.0 TM . All performances were repeated in triplicate (39,40).
Isothermal Titration Calorimetry-Microcal iTC200 (GE Healthcare) was operated to measure the binding affinity at 25°C (41). The titration buffer contained 50 mM Tris (pH 8.0), 150 mM NaCl, and 10 mM ␤-mercaptoethanol. Each sample was thoroughly degassed before titration. The first titration into the sample cell was a 0.5-l injection, followed by 24 successive 1.5-l injections at 180-s intervals. The Experimental data were analyzed with the Microcal ORIGIN V7.0 software (Microcal Software, MA).
Differential Scanning Fluorimetry-The experiments were performed using a RT-PCR detection system (ABI 7500 Fast) (42,43). Each well consisted of a 30-l solution containing 50 mM HEPES (pH 7.5), 50 M MnCl 2 , and 1.25 M protein, 1ϫ SYPRO orange (Invitrogen), and tested compound. The unfolded protein was monitored by SYPRO orange using the wavelength 492 nm for excitation and 610 nm for emission, respectively. Fluorescence intensities were obtained every 0.4°C in the range of 25°C to 95°C, which was heated at 1% ramp rate. The data were processed using Protein Thermal Shift TM software (Applied Biosystems). The T m was calculated by fitting the Boltzmann equation to the sigmoidal curve. All conditions were tested in triplicate.
Kinetics Analysis-To obtain the initial rate of inhibition, 20 nM AlkB was incubated with 2OG (5, 10, and 20 M) and rhein at 0, 2, 4, and 8 M, respectively. Another analysis was assayed in the presence of 50 nM AlkB and 20 M 2OG with 15-mer ssDNA (2.5, 5, and 10 M) and rhein at 0, 3, 6, and 12 M, respectively. The consumption of m 1 A-containing ssDNA was adjusted to less than 20%. After digestion by nuclease P1 and alkaline phosphatase, the nucleosides were separated by HPLC. The kinetics parameters were determined with the Michaelis-Menten equation fits and Lineweaver-Burk plot in GraphPad Prism 5.0.
Cell Viability Assay-100 l U87 cells (8 ϫ 10 3 ) were seeded on a 96-well plate and cultured overnight. For ALKBH2/3 knockdown, cells were seeded as 60% confluent of each well. After additional culture with compounds for 24 -48 h, cells in each well were mixed with 10 l of MTT (5 mg/ml; Amresco). The precipitate was resuspended in 150 l of DMSO, and the absorbance was detected at the wavelength of 570 nm. All assays were performed in triplicate.
RNA Interfering of ALKBH2 and ALKBH3-The assays were carried out using Lipofectamine RNAiMAX transfection re- agent (Invitrogen). The target mRNA sequence of ALKBH2 is 5Ј-CACGGGAGCUUACUAAUGA-3Ј, and that of ALKBH3 is 5Ј-UGCCCAAAGAAUACCACUCUA-3Ј. Cells were collected for Western blot analysis after 24 h of culture.
Crystal Structure of AlkB-Rhein Complex-The purified AlkB was incubated with 20-fold excess rhein in 50 mM HEPES (pH 7.5), 500 M MnCl 2 at room temperature for 20 min. The crystallization was set up as 1:1 hanging drops at 20°C. Crystals were grown in the reservoir solution containing 200 mM ammonium sulfate, 100 mM HEPES (pH 7.5), and 25% (w/v) polyethylene glycol 3,350. Crystals were flash cooled in liquid N 2 with a well solution diluted to 20% (v/v) glycerol. All x-ray data were collected at the Shanghai Synchrotron Radiation Facility Beamline 17U and processed in HKL2000 (44). The structures were solved by molecular replacement using the AlkB-2OG complex (PDB code 3I3Q) as a search model in the CCP4 suite (45). The model building was manually performed in COOT (46), and iterative refinements were carried out in REFMAC5 (47).

Results
Rhein Inhibits AlkB in Vitro-AlkB and FTO share the homologous catalytic domain for oxidative demethylation. Therefore, we wondered whether the FTO inhibitor rhein could inhibit AlkB in vitro (Fig. 1B). The restriction endonuclease digestion assay was run to evaluate the inhibition of AlkB by rhein (Fig. 1C). We observed full activity of AlkB repair of m 1 A on ssDNA or dsDNA in the control experiment, whereas AlkB was inhibited in the presence of rhein in a concentration-dependent manner. The IC 50 was quantitatively determined to be 12.7 M in the HPLC-based detection when assayed at 50 M 2OG (Fig. 1D) (48). BA is a moderate inhibitor of FTO (Fig. 1B) (31,49). AlkB repair remained intact, however, even in the presence of 100-fold excess BA (Fig. 1E); thus BA was employed as an inactive control. Taken together, these results show that rhein effectively inhibits E. coli AlkB in vitro.
Rhein Does Not Inhibit Ada or AlkA-In addition to AlkB, Ada and AlkA were also induced in E. coli in response to MMS. The C-Ada specifically repairs O 6 mG by transferring the methyl group to an active Cys (50). AlkA glycosylase recognizes N 3 -methyladenine and mismatch pairs (51). We performed selectivity analysis to rule out the potential phenotypic consequences of inhibiting C-Ada or AlkA repair by rhein in E. coli. As expected, rhein does not inhibit the demethylation activity proceeded by either C-Ada methyltransferase or AlkA glycosylase in vitro (Fig. 1, F and G). These results clearly show that rhein selectively inhibits AlkB rather than Ada or AlkA of E. coli.
Rhein Sensitizes E. coli to MMS-The alkB mutant E. coli strain becomes more sensitive to MMS compared with the wild-type or AlkB-overexpressed strain. This is due to the lack of AlkB repair, which accumulates methyl lesions on nucleic acids (17). We wondered whether the AlkB inhibitor rhein would similarly result in the impaired viability of E. coli when exposed to MMS threats. The growth of E. coli Gold was minimally impaired in the presence of either 50 -75 M MMS or 100 g/ml rhein ( Fig. 2A). Interestingly, the combination of 50 M MMS and 100 g/ml rhein reduces bacterial survival on agar medium. Moreover, the reduction in bacterial survival is reinforced with elevated MMS (75 M). In addition, we performed a measurement of cfu assay to quantitatively determine the sensitization of E. coli to MMS by rhein. As shown in Fig. 2B, rhein significantly makes E. coli sensitive to MMS threats when grown in liquid medium. In the control experiment, BA failed to sensitize the growth of E. coli in the presence of MMS (Fig.  2C). These data show that rhein sensitizes E. coli to MMS and ultimately impairs bacterial growth.
Rhein Promotes the Accumulation of DNA Methyl Lesions-To test whether rhein sensitization of E. coli to MMS is a result of the inhibitory effects of AlkB repair, we sought to quantify the cellular m 3 C lesions on genomic DNA (52). Performance of Western blot analysis confirms that MMS indeed induces AlkB expression in the time course and that rhein had no effect on AlkB abundance (Fig. 2D). The content of genomic m 3 C was quantified in dot blot analysis. As expected, rhein left the cellular level of m 3 C unimpaired but did promote the accumulation of m 3 C lesions in a dose-dependent manner during continuous MMS exposure (Fig. 2D), which should be a result of the inhibition of AlkB by rhein in E. coli.
In Vivo Target Engagement of Rhein-To probe the cellular target of rhein, we explored the impact on the potency of rhein when AlkB is overexpressed or mutated. The growth of E. coli Gold complemented with pET28a empty vector was minimally impaired in the presence of either 75 M MMS or 100 g/ml rhein but significantly inhibited in the presence of both 75 M MMS and 100 g/ml rhein (Fig. 3A, left panel). However, the AlkB-overexpressed E. coli strain became more resistant to the combined treatment (Fig. 3A, right panel). Obviously, the overexpressed AlkB could be able to repair methyl lesions even in the presence of the inhibitor rhein. Rhein also sensitizes wildtype AB1157 to MMS (Fig. 3B, left panel). The E. coli HK82, which is a typical alkB mutant strain derived from wild-type AB1157, became more sensitive to MMS because of deficient AlkB repair. In addition, rhein did not further increase the toxicity of MMS toward HK82 strain (Fig. 3B, middle panel). However, the HK82 strain, when transformed with a low copy pJB658 for AlkB complement, regained resistance to 750 M MMS. As expected, rhein obviously reduced the viability of AlkB-complement HK82 during MMS exposure (Fig. 3B, right  panel). These results show that the inhibitory activity of rhein is dependent on the abundance and function of the cellular AlkB enzyme.
Next, we compared the cellular sensitivity of E. coli to SN2 agent versus SN1 agent in the presence of rhein. MMS is an SN2-alkylating agent that induces the formation of m 1 A and m 3 C lesions repaired specifically by AlkB (53). In contrast, SN1 agents, MNNG for example, induce predominantly O 6 mG lesions that are repaired by O 6 -methyltransferases (54). MNNG at 5 M displayed minimal cytotoxic effects on E. coli growth. As expected, rhein could not enhance bacterial sensitivity to MNNG (Fig. 3C). In addition, rhein could not sensitize E. coli growth to other DNA-damaging agents such as H 2 O 2 (Fig. 3D).
These studies would also provide strong evidence that rhein is specifically targeting AlkB in vivo.
To further probe the in vivo target engagement of rhein, we performed a CETSA. AlkB in E. coli lysate shows a distinct shift in the melting curve (Fig. 3E), indicating that rhein directly binds to and stabilizes AlkB in the cell lysate. Next, the CETSA experiment in intact cells was also performed. E. coli were exposed to 500 M MMS in the presence of 100 g/ml rhein. The quantization of soluble AlkB reveals an obvious thermal shift in rhein-treated cells compared with that in the control experiment (Fig. 3F). The observations in CETSAs clearly show that rhein stabilizes AlkB in E. coli, perhaps through a direct binding.
Mechanistic Study of Rhein Inhibition of AlkB-We next investigated the mechanism of the inhibition of AlkB by rhein. Rhein binds tightly to the AlkB-Mn 2ϩ complex (K d ϭ 0.29 Ϯ 0.03 M) as observed in the isothermal titration calorimetry assay (Fig. 4A), which is an enthalpically driven process (⌬H ϭ Ϫ5.8 kcal mol Ϫ1 , ⌬S ϭ 10.4 cal/mol/deg). As expected, no binding response was detected when compound BA was titrated into the AlkB-Mn 2ϩ complex (Fig. 4B), which explains why BA is an inactive inhibitor of AlkB (Fig. 1E). In addition, the performance of a fluorescence-based thermal shift (differential scanning fluorimetry) assay showed that rhein significantly stabilizes AlkB as a reflection of the increasing melting temperature (T m ) of AlkB over 8°C in the presence of 20-fold excess molar of rhein (Fig. 4C). Furthermore, we have performed an enzyme kinetics analysis. The initial rates linear fits by HPLC quantification were of good quality (Fig. 4D, left panel). Rhein acts as a 2OG competitive inhibitor (K i ϭ 1.3 M; Fig. 4D, right panel) but a noncompetitive inhibitor toward substrate DNA (K i ϭ 2.1 M; Fig. 4E). Taken together, we conclude that rhein competes on 2OG binding to AlkB and acts as a competitive inhibitor of AlkB.
Structural Insights into the Inhibitor Binding-We have determined the x-ray crystal structure of AlkB-rhein complex at a resolution of 1.5 Å ( Table 1). The final R work and R free were 11.7 and 15.0%, respectively. The AlkB structure is not changed upon binding by rhein or 2OG, because the RMSD observed in the superimposition of our AlkB-rhein complex, and the known AlkB/2OG (PDB code 3I3Q) is only 0.28 Å (Fig. 5A). The F o Ϫ F c OMIT density contoured to 3.0 confirm that rhein is indeed bound (Fig. 5B). The binding site of rhein is partially overlapping with the 2OG binding site. This explains the com-petitive property of rhein inhibition of AlkB. The side chain of the carboxyl acid in rhein is positioned to form a salt bridge or hydrogen bonding with the side chains of Arg 204 and Tyr 122 . Rhein is positioned to chelate Mn 2ϩ in a bidentate manner by means of the carbonyl and hydroxyl groups. In addition to these major interactions, which are similar to 2OG binding to AlkB, rhein utilizes more interactions for AlkB binding, which take their origin from hydrogen-bonding between the other carbonyl in rhein and the side chains of Trp 178 and Ser 145 in AlkB.
Rhein is a competitive inhibitor of both AlkB and FTO (31); however, the mode through which rhein binds to AlkB is different from the mode through which the inhibitor binds to FTO (Fig. 5C). Rhein mainly occupies the 2OG-binding site to bind to AlkB (Fig. 5B). In contrast, rhein fully occupies the methylated DNA binding site in FTO (49). The structural superimposition of FTO and the complex of rhein bound to AlkB clearly shows that rhein would not bind to the 2OG pocket of FTO because of steric clashes to Tyr 295 and Met 297 (Fig. 5D, left panel). On the other hand, a second site for rhein binding to AlkB appears likely in the structural alignment of AlkB to the FTO-rhein complex, which is similar to that observed in FTOrhein complex (Fig. 5D, right panel). However, the functional relevance of this putative binding site of rhein in AlkB needs to be investigated.
Rhein Inhibition of ALKBH2 and ALKBH3 Sensitizes U87 Cells to MMS-The AlkB human homologues ALKBH2 and ALKBH3 are crucial to maintaining the integrity of the genome (55,56). In glioblastoma multiforme, knockdown of ALKBH2 reduces alkylation resistance to MMS (57). Silence of the ALKBH3 sensitizes the prostate cancer cells to MMS threats (15). We wondered whether the inhibitor of ALKBH repair would similarly reduce the methylation resistance of mammalian cells. We confirmed the inhibitory activity of rhein toward ALKBH2 and ALKBH3 demethylation by using the DpnII digestion assay (Fig. 6A). Then we performed HPLC-based assay to quantitatively measure the inhibitory activity of rhein. The IC 50 is 9.1 and 5.3 M for the inhibition of ALKBH2 repair of m 1 A in dsDNA and ALKBH3 repair of m 1 A in ssDNA, respectively, assayed at 50 M 2OG (Fig. 6B). Rhein promotes the thermal stability of ALKBH2 and ALKBH3 by enhancing T m values in a concentration-dependent manner (Fig. 6C), which indicates that rhein also binds to ALKBH2 and ALKBH3 enzymes. Next, we evaluated whether rhein sensitizes U87, a glioblastoma multiforme cell line, to SN2-alkylating agents. As demonstrated in the MTT assay, the proliferation of U87 cells was minimally affected by either 80 M rhein or 500 M MMS after 48 h (Fig. 6D). Strikingly, rhein significantly sensitized U87 cells to MMS in a dose-dependent manner. The combined treatment of 80 M rhein with 250 or 500 M MMS caused a severe struggle for the U87 cells to survive, indicating that rhein and MMS inhibit U87 cell proliferation in a synergistic way.
Target Selectivity of Rhein in U87 Cells-To investigate whether the rhein-dependent sensitization of U87 cells to the MMS threat requires the expression and oxidative demethylation of cellular ALKBH2 and ALKBH3, we silenced both ALKBH2 and ALKBH3 by RNAi (Fig. 6E). After 24 h of exposure to MMS, rhein showed enhanced cytotoxicity to U87 cells in the siRNA control group. Of note, knockdown of ALKBH2 and ALKBH3 removed the synergistic effect of MMS and rhein. In addition, rhein and the SN1-alkylating agents such as MNNG and TMZ, a common chemotherapy drug, failed to produce a synergistic effect on the proliferation of U87 cells (Fig. 6F). These data indicate the rhein-dependent sensitization of human cells to MMS alkylation damage, which is a result of the inhibition of ALKBH2-and ALKBH3 repair of m 1 A and m 3 C lesions by rhein.
To further address target selectivity in vivo, we tested whether rhein inhibits other demethylation dioxygenases. Rhein is inactive for the inhibition of prolyl-4-hydroxylase in the in vitro screens, indicating that rhein is not a broad spectrum inhibitor of the 2OG-dependent hydroxylases (58). Rhein could moderately inhibit JMJD2A and JMJD2E in vitro, two iron(II) and 2OG-dependent histone demethylases. We monitored the abundance of trimethylated H3K9me3 that is the downstream target of JMJD2A and JMJD2E, when U87 cell is treated with rhein or MMS. Rhein or MMS or the combination of rhein and MMS did not alter the cellular abundance of H3K9me3, indicating that rhein fails to inhibit JMJD2A or JMJD2E at the tested concentration (Fig. 6G, top panel). In the control experiment, we observed that JIB-04, a known selective inhibitor of JMJD family enzymes (59), increased the abundance of H3K9me3 in a concentration-dependent manner (Fig. 6G, bottom panel). These data indicate that rhein-dependent sensitization of human cells to MMS alkylation damage is due to ALKBH2 and ALKBH3 inhibition rather than perturbation of other dioxygenase-dependent cellular pathways.

Discussion
DNA-methylating agents constantly damage nucleic acids in cells, frequently with mutagenic and/or cytotoxic consequences. The accumulation of m 1 A and m 3 C lesions is generally very cytotoxic and even leads to cell death. In E. coli, AlkB repair of m 1 A and m 3 C appears to be the major natural defense mechanism in protecting bacteria survival on methylation threats. To our knowledge, several inhibitors of AlkB repair enzymes have been developed, but the cellular tractability by small molecules has not been explored (27,28,60). The validation of biological targets and profiling of cellular pathway remains a major challenge for those inhibitors.
We have carried out a number of assays to investigate the mechanism by which rhein inhibition of E. coli AlkB sensitizes bacterial cells to MMS methylation. Rhein efficiently inhibits AlkB without inhibiting the glycosylase AlkA or the methyltransferase Ada in vitro (Fig. 1). Rhein promotes E. coli death during continuous MMS exposure, which could be a result of rhein inhibition of AlkB repair of the major methyl lesion, such as m 3 C (Fig. 2). Indeed, we detected an increased amount of m 3 C lesion in genomic DNA, which could be responsible for bacterial death. Rhein binds to AlkB enzyme and promotes its stability in vitro. Structural features observed within the AlkB-rhein complex have further demonstrated the inhibition mechanism of rhein, which is consistent with the enzyme kinetics result that rhein is a competitive inhibitor of AlkB (Figs. 4 and 5). Taken together, we conclude that rhein inhibition of AlkB repair sensitizes E. coli to MMS-induced methylation damage by increasing the accumulation of cellular m 3 C lesions. The cellular target engagement of rhein has been adequately addressed (Fig. 3). Either overexpression of AlkB or deletion of AlkB demethylation makes E. coli resistant to rhein during MMS exposure, suggesting that the inhibitory activity of rhein is dependent on the abundance and the function of the cellular AlkB enzyme. Furthermore, rhein fails to sensitize E. coli to MNNG, a SN1-alkylating agent, which indicates that rhein specifically inhibits AlkB-catalyzed repair in E. coli. Rhein increases the thermal stability of AlkB in vivo, thus revealing its ability for binding to the AlkB enzyme. All these data have pointed toward a direct inhibition of AlkB by rhein in vivo.
Human oxidative repair enzymes have been broadly linked to cancer. DNA damage created by alkylation agents would lead to apoptosis and is one of the major mechanisms of cytotoxic anticancer drugs such as TMZ (61,62). We have demonstrated that rhein inhibits ALKBH2 and ALKBH3 in vitro and synergistically suppresses the proliferation of U87 cells during exposure to MMS (Fig. 6). Strikingly, this synergistic effect vanished in the absence of ALKBH2 and ALKBH3 enzymes. In addition, rhein enhances the sensitivity of U87 cells to an SN2 agent but not an SN1 agent, which also provides strong evidence that rhein is specifically targeting ALKBH-catalyzed DNA repair. Together, rhein could efficiently target the ALKBH-catalyzed repair, thus acting to sensitize mammalian cells to MMS threats.
In summary, we have characterized that the natural product rhein inhibits AlkB repair enzymes (AlkB, ALKBH2, and FIGURE 5. Structural insights into the mode of rhein binding to AlkB. A, structure alignment of the AlkB-rhein (PDB code 4RFR) and 2OG-bound AlkB (PDB code 3I3Q) performed in PyMOL with RMSD ϭ 0.28 Å. The AlkB-rhein structure is colored cyan, AlkB/2OG is magenta, and the oxygen atom is red, respectively. Mn 2ϩ is shown as a sphere and colored orange. Rhein and 2OG are shown as sticks. B, an m͉F o ͉ Ϫ D͉F c ͉ map was calculated within the PHENIX program suite after omission of rhein from the complex model and subsequent simulated annealing. The map density is contoured to 3.0 . The coordination of Mn 2ϩ by ligands and hydrogen bonding are denoted by dotted dark lines. The map is shown in blue. The superimposition of rhein and 2OG is presented. C, structural superimposition of AlkB-rhein and FTO-rhein complexes performed in PyMOL. The FTO/rhein (PDB code 4IE7) is colored orange. Rhein is shown as sticks. D, zoomed-in view token from C to show the pocket for rhein binding to AlkB and FTO, respectively. Rhein could not bind to FTO similarly to AlkB because of the steric clashes by Tyr 295 and Met 297 (left panel). A likely binding pocket is observed in AlkB for rhein binding similarly in FTO (right panel). ALKBH3) in vitro and decreases cellular resistance to MMS. All data of the mechanism investigations point to rhein as the direct inhibitor of AlkB repair enzymes. Our proof of principle study supports the findings that ALKBH2 and ALKBH3 enzymes would be effective pharmacological targets to overcome tumor resistance to methylating anticancer drugs.
Author Contributions-C.-G. Y. designed the project and wrote the paper with Q. L., and Q. L. performed most of the experiments. Y. H. provided technical assistance in protein purification and biochemical experiments. X. L. and H. C. assisted in the isothermal titration calorimetry experiment. J. G. processed the x-ray data and solved the AlkB-rhein structure. All authors reviewed the results and approved the final version of the manuscript.