Targeting RING domains of Mdm2–MdmX E3 complex activates apoptotic arm of the p53 pathway in leukemia/lymphoma cells

Reactivation of tumor-suppressor p53 for targeted cancer therapy is an attractive strategy for cancers bearing wild-type (WT) p53. Targeting the Mdm2–p53 interface or MdmX ((MDM4), mouse double minute 4)–p53 interface or both has been a focus in the field. However, targeting the E3 ligase activity of Mdm2–MdmX really interesting new gene (RING)–RING interaction as a novel anticancer strategy has never been explored. In this report, we describe the identification and characterization of small molecule inhibitors targeting Mdm2–MdmX RING–RING interaction as a new class of E3 ligase inhibitors. With a fluorescence resonance energy transfer-based E3 activity assay in high-throughput screening of a chemical library, we identified inhibitors (designated as MMRis (Mdm2–MdmX RING domain inhibitors)) that specifically inhibit Mdm2–MdmX E3 ligase activity toward Mdm2 and p53 substrates. MMRi6 and its analog MMRi64 are capable of disrupting Mdm2–MdmX interactions in vitro and activating p53 in cells. In leukemia cells, MMRi64 potently induces downregulation of Mdm2 and MdmX. In contrast to Nutlin3a, MMRi64 only induces the expression of pro-apoptotic gene PUMA (p53 upregulated modulator of apoptosis) with minimal induction of growth-arresting gene p21. Consequently, MMRi64 selectively induces the apoptotic arm of the p53 pathway in leukemia/lymphoma cells. Owing to the distinct mechanisms of action of MMRi64 and Nutlin3a, their combination synergistically induces p53 and apoptosis. Taken together, this study reveals that Mdm2–MdmX has a critical role in apoptotic response of the p53 pathway and MMRi64 may serve as a new pharmacological tool for p53 studies and a platform for cancer drug development.

Activation of tumor-suppressor p53 as a targeted non-genotoxic cancer therapy has been pursued for many years, 1,2 because p53 possesses potent tumor-suppressing activity in vivo. [3][4][5] p53 can inhibit cancer cell growth by cell cycle arrest or terminate their proliferation by inducing apoptosis and senescence. 6 The p53-based therapy is particularly attractive for cancer types including retinoblastoma, neuroblastoma and leukemia/lymphoma in which p53 is rarely mutated 7 and p53-dependent apoptotic pathway is a predominant endpoint. [8][9][10] Except for cancer-selected p53 mutations, the p53 activity is mainly inhibited by p53-binding proteins Mdm2 and MdmX ((MDM4), mouse double minute 4) in normal and cancer cells. 11,12 Prior focus of p53 reactivation strategy has been on targeting the Mdm2-p53 and/or MdmX-p53 interface. This has led to the discovery of a list of potent Mdm2-p53 inhibitors 13 with several compounds of this class being advanced to phase I clinical trials in hematological neoplasia and solid tumors. 2 However, the therapeutic effects of these Mdm2-p53 inhibitors can be attenuated by MdmX overexpression. [14][15][16] Although peptide inhibitors with dual functions of inhibiting both Mdm2-p53 and MdmX-p53 interactions will overcome this problem and enhance p53-dependent cancer killing; 17,18 these inhibitors will not inhibit Mdm2 E3 ligase activity toward non-p53 targets such as retinoblastoma protein (RB), p21 and DAXX (death domain-associated protein), [19][20][21] which to a different extent contributes to the p53-dependent biological effects.
Recent genetic studies indicated that really interesting new gene (RING) domains of Mdm2 and MdmX are required for in vivo inhibition of p53 activity during development. [22][23][24] MdmX was reported to stimulate Mdm2-mediated p53 multiple monoubiquitination using glutathione S-transferase (GST) fusion Hdm2 proteins. 25,26 Using non-GST Hdm2 proteins in in vitro biochemical assays, we found that MdmX-Mdm2 RING-RING interaction is essential for p53 polyubiquitination and proteasome-dependent degradation. 26 These findings established that Mdm2-MdmX complex is the key regulator of p53 activity and Mdm2-MdmX RING-RING interaction is a critical but an unexplored interface for drug targeting. 27 Identification of E3 ligase inhibitors for cancer therapy presents a huge opportunity but with great challenges. 28 In this report, we describe successful identification and characterization of small molecule inhibitors for the E3 ligase activity of Mdm2-MdmX E3 complex. Among seven specific MMRis (Mdm2-MdmX RING domain inhibitors), MMRi64 was followed up in detail in this report. MMRi64 has several unique features that distinguish it from Mdm2-p53 inhibitor Nutlin3a. MMRi64 disrupts Mdm2-MdmX interaction in vitro and inhibits the E3 ligase activity of Mdm2-MdmX without affecting the E3 ligase activity of Mdm2 RING domain homodimers. MMRi64 induces p53 accumulation without induction of Mdm2 and p21 in lymphoma cells, which is distinct from the effects of Nutlin3a. Finally, MMRi64 induces PUMA (p53 upregulated modulator of apoptosis) but strongly downregulates MdmX and Mdm2, consequently activating the apoptotic arm of the p53 pathway in leukemia/lymphoma cells without the induction of growth arrest.

Results
High-throughput screening of small molecule inhibitors for the E3 ligase activity of Mdm2-MdmX E3 complex. We previously reported that Mdm2-MdmX RING-RING interaction is required for p53 polyubiquitination. 26 This interaction also stimulates Mdm2 autoubiquitination and MdmX ubiquitination (Figure 1a and Wang et al. 26 ). To establish a biochemical assay for screening small molecule inhibitors of Mdm2-MdmX RING-RING interaction, we took advantage of an in vitro assay for MdmX-stimulated Mdm2 autoubiquitination as a readout of the interaction effect. To facilitate its application in high-throughput screening (HTS), we adapted our in vitro ubiquitination assay to a fluorescence resonance energy transfer (FRET)-based quantification system described previously. 29 This system uses homogeneous time-resolved fluorescence (HTRF TM ) to quantify ubiquitin chain reactions. In this system, the fluorescence signals are generated by FRET from two fluorophore-labeled components in proximity, one is ubiquitin and the other is ubiquitinated substrates. In our case, as illustrated in Figure 1b, FRET signals were generated between anti-HA-XL665 that binds to HA-Mdm2 and HA-ubiquitin and ubiquitin cryptate. The total FRET signal from the reaction collectively reflects ubiquitin chains formed on Mdm2 and MdmX. Compounds that disrupt the Mdm2-MdmX interaction will result in reduced E3 ligase activity of Mdm2-MdmX complex consequently reducing the amounts of ubiquitinated Mdm2 and ubiquitinated MdmX and the FRET signals. In the absence of MdmX, FRET signals generated by ubiquitin cryptate and HA-Mdm2 were very low, which was defined as baseline. Under our optimized conditions, addition of MdmX produced~8-fold increase in FRET signals in an MdmX concentration-dependent manner ( Figure 1c) and reaction time-dependent manner (Figure 1d). After adaption of this assay in HT format, we performed an initial screen of~650 samples. The Z′-factor of this HTS assay was determined to be 0.52 (Figure 1e), indicating a suitable and reliable HTS screen assay (Figure 1e). 30 This validated HTS assay was then used to screen a diversity library (DIVERSetTM, ChemBridge). Out of 55 230 compounds, we identified a number of positive hits at different inhibition cutoffs as summarized in Figure 1f. The results indicated that our HTS was robust, considering the library size we used and hit rates obtained, 31 as it identified 119 hits at 90% inhibition cutoff and 371 hits at 70% inhibition cutoff out of~50 000 compounds (Figure 1f). We followed up all the 371 hits for validation using our bench-top biochemical assay.
Identification of lead compounds that specifically target Mdm2-MdmX E3 ligase activity. To evaluate the hits identified by HTS for specific inhibition of Mdm2-MdmX E3 activity, we examined their ability to inhibit Mdm2 autoubiquitination and p53 polyubiquitination by Mdm2-MdmX using our in vitro biochemical assays. We used NEDD4-1 autoubiquitination as a control for nonspecific inhibitors of E3 ligase activity in replicate experiments. In these assays, the ubiquitinated products were monitored by western blotting (WB) instead of FRET assay. After evaluating the availablẽ 350 hits, the hits fall into three categories: (1) 301 hits that fail to inhibit any of the E3 ligase activities (HTS false positive); (2) 42 pan-inhibitors of the ubiquitination system that inhibit all three reactions; (3) seven specific inhibitors of Mdm2-MdmX RING-RING E3 complex that inhibit both MdmX-stimulated Mdm2 autoubiquitination and Mdm2-MdmX-mediated p53 polyubiquitination, but not NEDD4-1 autoubiquitination (designated MMRi, which stands for Mdm2-MdmX RING domain inhibitors). As summarized in Figure 2, among the seven MMRis, three MMRis strongly inhibited p53 ubiquitination by Mdm2-MdmX. We used Nutlin3a as a negative control. In contrast to MMRis, Nutlin3a as an Mdm2-p53-binding inhibitor had no effect on Mdm2-MdmX-mediated p53 polyubiquitination at the same concentration ( Figure 2c, Nutlin3a), indicating that MMRis inhibit Mdm2-MdmX via distinct mechanisms than Nutlin3a. Intriguingly, MMRi1, MMRi4 and MMRi6 all partially inhibited Mdm2 autoubiquitination to a similar extent ( Figure 2a). However, MMRi6 significantly inhibited p53 ubiquitination, whereas MMRi1 and MMRi4 only partially inhibited p53 ubiquitination by Mdm2-MdmX in vitro. The exact mechanisms underlying these differences are unclear at the moment. However, we speculate that these differences may be attributed to the differential effects of the individual MMRis on the ubiquitin transfer preference of Mdm2-MdmX to its adjacent peptides. Of note, this study also identified a group of compounds that inhibits both Mdm2-MdmX and NEDD4-1 E3 ligases designated as MMNi (Mdm2-MdmX and NNEDD4-1 inhibitors) as represented in Figure 2.
Characterization of MMRi6 and its analogs as disruptors of Mdm2-MdmX RING-RING interaction. MMRi6 was further followed up because several MMRi6 analogs were available and MMRi6 appeared to be the most potent inducer of p53 stabilization in cell culture among the available MMRis (data not shown). Other MMRis of different chemical classes including MMNi will be followed up in separate studies. We obtained 13 commercially available analogs of MMRi6 for further evaluation (structures of MMRi6 and MMRi64 are shown Figure 3d). We confirmed that MMRi6 and its MMRi64 potently induces p53-dependent and p53independent apoptosis in lymphoma cells. We further tested the antitumor effect of MMRi64 in leukemia/lymphoma cells. We focused on the induction of apoptosis by this compound because p53-dependent apoptosis is a critical mechanism for preventing lymphomagenesis 33,34 and determines the outcome of lymphoma treatment. 35 In NALM6 cells, MMRi64 at 1 μM showed a time-dependent induction of PUMA, a critical pro-apoptotic downstream gene product of p53. 36,37 Interestingly, p21, the growth-arresting effector target gene of p53, was transiently induced then downregulated to a level lower than basal p21 expression at 24 h of the treatment. Accompanied with the activation of p53's proapoptotic arm of the p53 response, cleavage of poly ADP ribose polymerase (PARP) by activated caspase 3 was evident at 8 h and further increased at 24 h after treatment. These data indicated that MMRi64 triggered activation of the intrinsic apoptosis pathway. We performed a similar experiment with Nutlin3a and our results showed that Nutlin3a induced stronger p53 accumulation and PUMA induction than MMRi64 at the same concentration of MMRi64 (Figure 5a, middle panel). However, in contrast to MMRi64, Nutlin3a also exhibited stronger expression of growth-arresting effector p21 (Figure 5a, right panel). Accompanied with these molecular events, cleavage of PARP and activation of caspase 3 were barely detectable in Nutlin3a-treated cells for 24 h, although Nutlin3a induced a similar level of p53 accumulation at this time point (Figure 5a right panel). Together, these results indicated that p53 activation by Nutlin3a mainly results in cell growth arrest, whereas p53 activation by MMRi64 mainly causes apoptosis in NALM6 cells. To determine whether MMRi64-induced apoptosis in lymphoma cells is p53 dependent, we performed experiments with Emu-myc mouse lymphoma cells of different p53 status. As expected, MMRi6 induced p53 accumulation in wt-p53 cells at a concentration as low as 0.1 μM (Figure 5b, upper panel). PARP cleavage was detected at 24 h by~0.5 μM MMRi6 treatment in wt-p53 Emu-myc lymphoma cells but not in p53-null Emu-myc lymphoma cells (Figure 5b, lower panel). Therefore, the MMRi6-induced apoptosis in lymphoma cells contains a p53dependent component. The ability of MMRi64 to induce apoptosis was further analyzed by flow cytometry. Our results showed that MMRi64 at 0.5 and 1 μM for 48 h induced 7.3% and 20% sub-G1 population, respectively. In contrast, Nutlin3a at 0.5, 1 and 2 μM for 48 h only induced 0.4%, 0.8% and 3.0% sub-G1 populations, respectively ( Figure 5c). Together, these results indicate that MMRi64 preferentially induces apoptosis in NALM6 cells.
To further confirm the p53 dependence of MMRi64-induced growth inhibition, we used mouse Emu-myc lymphoma cells of wt-p53 and p53-null background in growth inhibition experiments. A 72-h cell proliferation assay for MMRi6 showed IC50s of~0.5 μM and~3 μM in wt-p53 and p53-null Emu-myc lymphoma cells, respectively, indicating that p53 contributes to a~6-fold difference in MMRi6 sensitivity in this set of mouse lymphoma cells (Figure 6a, data from two doses were shown). Furthermore, we used HCT116 and HCT116-p53 − / − cells to test the contribution of p53 to MMRi64-induced anti-growth effect. As shown in Figure 6b, at equimolar concentrations of Nutlin3a, p53 contributes to a maximal~35% more growth inhibition than that of HCT116-p53 − / − cells, whereas p53 contributes to a maximal~10% more growth inhibition in MMRi64 treatment. Therefore, MMRi64 inhibits cell growth through both p53-dependent and p53-independent mechanisms.
To understand whether low concentrations of MMRi64 and Nutlin3a will synergistically inhibit cell growth by apoptosis, we performed combination experiments with the two compounds in NALM6 cells. First, we looked at apoptotic PARP cleavage during single and combination treatment. Our results showed that 1 μM

Discussion
Targeting Mdm2-p53 interaction for p53-based cancer therapy has been pursued for many years. Several promising compounds with therapeutic activity in preclinical systems have been reported and have recently advanced into early clinical trials. 2 Nutlin3a was the first potent specific inhibitor of Mdm2-p53 interaction. 1 It has served as a prototype for chemical optimization and has fostered the discovery of compounds with better drug properties and efficacy. [38][39][40] However, one prominent concern is that MdmX overexpression confers resistance to Nutlin3a. [14][15][16] This problem is likely to hinder the use of other Mdm2-p53 targeting compounds.
Together with the radioresistant lymphoma phenotype of nondegradable MdmX mutant mice, 41 these findings point MdmX as another valid drug target for p53-based cancer therapy. 13 Using an elegant mouse model, Even's group demonstrated that MdmX is a better drug target than Mdm2 in lymphoma. 42 To overcome MdmX-mediated resistance, dual inhibitors that target both Mdm2-p53 and MdmX-p53 interfaces were also explored. Results from a dual inhibitor peptides and compounds demonstrated better p53-dependent cytotoxic effects in breast and colon cancer cells. 17,43 Differing from the focus of the field, we turned our interest in Mdm2-MdmX RING-RING interaction based on the biochemical findings from our lab and genetic evidences from other's studies. 23,24,26 This report is the first attempt to identify specific inhibitors of Mdm2-MdmX RING domain interaction and assess the effectiveness of targeting this newly established interface. To our surprise, MMRi64 not only distuptedMdm2-MdmX interaction, but also induced a significant MdmX downregulation in leukemia cells. Mdm2 upregulation usually follows p53 activation because of a feedback regulatory loop. 44,45 However, Mdm2 was not induced in  (Figure 4b). The mechanisms underlying the cell type difference in Mdm2 induction are not known at present. We speculate that this may be due to the differential effect of MdmX on Mdm2 protein stability, that is, MdmX has a crucial role to stabilize Mdm2 proteins in leukemia cells but is not so critical in colon cancer cells. This hypothesis needs further experimentation. In our opinion, this cell type-specific effect of MMRi64 on Mdm2 and MdmX makes it a very unique compound for leukemia/ lymphoma treatment.
We tested MMRi64 mainly in leukemia/lymphoma cells because the p53 pathway is critical for apoptosis induction, 46 lymphoma development 10,33,34 and restoration of p53 in mice leads to regression of autochthonous lymphomas. 4 Importantly, p53 mutation rate is relatively low (4.2% of diffuse large B-cell lymphoma, TCGA data) and about 80-90% of lymphoma patients have a wt-p53 status. 47 Therefore, p53based therapy would benefit a large group of lymphoma/ leukemia patients. The p53-based therapy relies on druginduced apoptosis. 4 In sharp contrast to Nutlin3a, MMRi64 preferentially induces apoptosis in leukemia/lymphoma cells, probably due to PUMA induction and simultaneous shutdown of pro-growth-arrest p21 (Figures 4 and 5a), and downregulation of Mdm2 and MdmX. Whether the selective effect of p53 downstream gene induction by MMRi64 is due to its effect on MdmX or Mdm2 needs to be tested in future experiments.
The cellular effect of Nutlin3a is very moderate at concentrations similar or higher than that of MMRi64: 1-3% increase in G1 or G2 populations and 4-7% reduction in S-phase cells (Figure 5b), suggesting that there are modifiers of the p53 downstream effect in NALM6 cells. The mechanisms to why MMRi64 preferentially induces apoptosis compared with Nutlin3a are presently unknown. Downregualtion of Mdm2 and/or MdmX by MMRi64 may be an explanation and several lines of evidence support this notion. The downregulation of Mdm2 by MI-219 appears to be associated with p53-dependent apoptosis in follicular lymphoma. 48 Lozano's group demonstrated that p53 restoration in Mdm2overexpressing tumors inhibits proliferation but does not induce apoptosis, 49 suggesting that high levels of Mdm2 has anti-death activity. Moreover, siRNA knockdown experiments revealed that Mdm2 is actually required for p53-dependent induction of p21 to cause growth arrest. 50 These findings are consistent with the reduced levels of Mdm2 and weak induction of p21 by MMRi64 treatment. Shutdown of p21 induction by p53 in MMRi64-treated cells is expected to favor apoptosis induction in MMRi64-treated cells, as p21 serves as an inhibitory effector for p53-dependent apoptosis shown in colon cancer cells. 51 In addition, Mdm2 was reported to have p53-independent lymphomagenic activity, a role revealed by Mdm2 splice isoforms that do not bind to p53 but promote Emu-myc lymphomagenesis in a manner comparable with fulllength Mdm2. 52 At present, we cannot exclude the possibility that MMRi64 kills leukemia/lymphoma cells independently of p53 or PUMA, which requires more mechanistic studies in the future. The mechanisms underlying downregulation of MdmX and Mdm2 in the presence of MMRi64 in cells are not clear either. However, we speculate that the binding of MMRi64 to MdmX disrupts Mdm2-MdmX interaction, which may make MdmX a target of protein quality control system leading to its ubiquitin-dependent degradation. This is highly likely as compound-bound MdmX may adopt a conformation of ill-folded protein, which will be recognized by protein quality control E3 ligase CHIP. Once MdmX is degraded, it will affect Mdm2 expression as MdmX is required for binding p53 to  53 In addition, MdmX can stabilize Mdm2 protein by inhibiting its autoubiquitination. 54 Of note, MEL23 and MEL24 were identified by a cell-based Mdm2-luciferase fusion reporter screening effort as inhibitors of Mdm2-MdmX E3 ligase. 55 However, their mechanisms of action were not characterized and their potential in apoptosis induction was not reported. As MEL23/MEL24 strongly stabilize Mdm2 and MdmX, 55 our MMRi64 has unique property for better targeting oncogenic Mdm2-MdmX complex for robust cancer cell killing.
MMRi64 belongs to a chemical class of quinolinol family in which inhibitors of botulinum neurotoxins and two anticancer derivatives were recently reported. [56][57][58] Owing to the unique effects of MMRi64 on the critical components of the p53 regulatory loop and downstream effectors, MMRi64 may be used as a pharmaceutical tool to dissect the molecular regulation of p53-dependent transactivation program. Therefore, this study identified a new class of chemicals that may be useful in basic research on p53-dependent biology as well as development of new p53/Mdm2/MdmX-based cancer therapeutics. In summary, MMRi represents a new class of p53-activating agents with promising anticancer activities in cell culture. Future studies will be needed to address the biophysical property of drug-target interaction, p53/Mdm2/ MdmX-dependency and the issues of off-targets and genotoxicity in lead optimization process.

Materials and Methods
Plasmids, protein purification and chemical reagents. FLAG-MdmX and HA-Mdm2 (human) constructs for insect cell expression and protein purification were described previously. 26 HA-ubiquitin construct was generated by inserting HA-tag to the N-terminus of ubiquitin in pET28a vector. The mammalian expression plasmid pcDNA3.1-HdmX was a gift from Dr. Gokul . HA-Mdm2 RING domain was generated by site-directed mutagenesis to loop out aa28-299 using pFAST-bac-HA-Mdm2 as an template and the recombinant baculovirus was prepared and protein was expressed in insect cells as described previously. 26 HCT-8 was used in our recent studies 59  FRET-based in vitro ubiquitination assay. FRET-based in vitro ubiquitination assay was adapted from the protocol of HTRF TM described previously. 29 HTRF is a homogeneous method, which combines standard FRET technology with    For the FRET measurement in Perkin Elmer Envision 2103 Multilabel Reader, there is a 100 μs time delay between the excitation (320 nm) and measurement at two different wavelengths (615 and 665 nm), then calculating the ratio for each well individually (ratio = 665 nm/615 nm x10 4 ). The 10 4 multiplying factor is introduced for convenient data processing.
In order to test the compound specificity, we used two sets of ubiquitination reaction. One was MdmX-stimulated Mdm2 autoubiquitination, the other was NEDD4-1 autoubiquitination. The two reactions shared the same constitutes in the premixture: 40 mM Tris-HCl (pH 7.5), 5 mM MgCl 2 , 2 mM DTT, 5 mM ATP, 20 nM E1 and 350 nM E2 UbcH5c and 10 μM of ubiquitin. The former reaction included 100 nM HA-tagged Mdm2 and 200 nM MdmX, and latter reaction included 200 nM HA-tagged NEDD4-1. After adding compound to a final concentration of 10 μM, the reaction was started by incubation at 30°C in a water bath for 1 h. Then the reaction was stopped by adding SDS sample buffer, followed by SDS-PAGE and WB analysis for HA or NEDD4-1. In vitro assays for p53 ubiquitination by Mdm2-MdmX were performed as described previously. 26 Briefly, reactions were carried out as described above for Mdm2 autoubiquitination except for addition of 100 nM p53. Compounds in dimethyl sulfoxide (DMSO) or DMSO of final concentrations of 10 μM were added in the reaction before starting the reaction at 30°C for 1 h followed by WB of p53 with DO-1 antibody.
Proliferation and apoptosis assays. Growth inhibition assay and apoptosis assay was described previously. 59 RING domain interaction between Mdm2 and MdmX by in vitro pulldown assay. HA-tagged Mdm2 RING domain (500 nM), Flag-tagged MdmX (250 nM) and testing compound (10 μM) were mixed together in 50 μl NP40 buffer (0.5% NP40, 150 mM NaCl, 20 mM Tris-HCl, pH 8.0). After 30-min incubation, the protein mixture was diluted with 450 μl NP40 buffer-0.5% BSA. In all, 10 μl of anti-FLAG antibody conjugated M2 beads (Sigma: A2220) preincubated with NP40 buffer containing 0.5% BSA for 30 min at RT were added into the mixture to pulldown the Flag-MdmX by rotating at RT for 2 h. After washing five times by the NP40 buffer-0.5% BSA, the M2 beads were eluted with 45 μl 0.2 mg/ml 3xFlag peptides (in 20 mM Tris-HCl, pH 7.5, 10 mM NaCl, 1 mM DTT) to release FLAG-MdmX and its interacting proteins. The FLAG-MdmX-bound HA-Mdm2 RING was detected by SDS-PAGE followed by WB analysis for HA-Mdm2 by using ant-HA antibody.
Docking analysis. Docking studies were carried out in order to investigate the preferential binding mode geometry of the compounds. The steps were performed as instructed by DOCK6 software package (dock.compbio.ecsf.edu) including all default parameter set. 3-D grids were defined for the interaction with the compounds. Docking results were examined using the Chimera program (http://www.cgl.ucsf.edu/chimera/) to observe the interaction precisely at atomic level. Chimera was also used to manipulate the Mdm2-MdmX RING (30) structures with the utilities for deleting solvents and adding charges. 32