Recombinant yeast screen for new inhibitors of human acetyl-CoA carboxylase 2 identifies potential drugs to treat obesity

Abstract

Acetyl-CoA carboxylase (ACC) is a key enzyme of fatty acid metabolism with multiple isozymes often expressed in different eukaryotic cellular compartments. ACC-made malonyl-CoA serves as a precursor for fatty acids; it also regulates fatty acid oxidation and feeding behavior in animals. ACC provides an important target for new drugs to treat human diseases. We have developed an inexpensive nonradioactive high-throughput screening system to identify new ACC inhibitors. The screen uses yeast gene-replacement strains depending for growth on cloned human ACC1 and ACC2. In “proof of concept” experiments, growth of such strains was inhibited by compounds known to target human ACCs. The screen is sensitive and robust. Medium-size chemical libraries yielded new specific inhibitors of human ACC2. The target of the best of these inhibitors was confirmed with in vitro enzymatic assays. This compound is a new drug chemotype inhibiting human ACC2 with 2.8 μM IC₅₀ and having no effect on human ACC1 at 100 μM.

Acetyl-CoA carboxylase (ACC) catalyzes the addition of CO₂ to acetyl-CoA to make malonyl-CoA, the form in which two carbon units are added to growing chains of fatty acids. Many eukaryotes have two multidomain ACC isozymes, localized in different cellular compartments, the products of genes duplicated independently in different lineages. Plant cells make all their fatty acids de novo. On the contrary, animals derive most of their fatty acids from food and protozoan parasites have fatty acid scavenging capability. Despite these external sources of fatty acids, animals and at least some protozoan parasites retain the ability to synthesize malonyl-CoA and to make fatty acids de novo. In yeast, which can take up long-chain fatty acids from the medium, the ACC1 gene remains essential as well. The essential nature and central role in cellular metabolism makes ACC a potentially valuable target for new drugs (1).

In wheat, ACC1 has a leader sequence that directs it to the plastid, where it is involved in fatty acid synthesis. The cytosolic enzyme (ACC2) makes malonyl-CoA for very long-chain fatty acids, flavonoids, and signaling compounds. The plastid form of the enzyme in wheat and other grasses is sensitive to three classes of highly effective herbicides: aryloxyphenoxypropionates, cyclohexanediones, and pinoxaden (2–5). We have shown that the parasite Toxoplasma gondii has two ACCs as well, one located in the apicoplast, where it is involved in de novo fatty acid synthesis for lipids and the lipoic acid cofactor of pyruvate dehydrogenase. We showed that the apicoplast isozyme is sensitive to aryloxyphenoxypropionates such as clodinafop and haloxyfop (6–8). These compounds are strong enough inhibitors to kill parasites in human fibroblasts grown in culture and are not toxic to human cells, but they are not strong enough to be useful as drugs. Nevertheless, our results support the validity of Toxoplasma ACC as a potential drug target.

Early experiments from the Wakil laboratory showed that there are two isozymes of ACC in mammals (9). The two isozymes of human ACC are similar in amino acid sequence over most of their length (∼2,400 amino acids). An N-terminal extension on ACC2 directs this form of the enzyme to mitochondria (10 and 11). There, ACC2-catalyzed synthesis of malonyl-CoA leads to suppression of fatty acid transport into mitochondria by a system involving carnitine palmitoyl transferase 1 (CPT1): malonyl-CoA inhibits CPT1. ACC2 is expressed mainly in muscle. Deletion of the ACC2 gene in mice leads to continuous fatty acid oxidation and affects insulin sensitivity, validating ACC2 as a potential target for drugs to treat obesity (12–15). ACC1, on the other hand, is an essential enzyme responsible for fatty acid synthesis in lipogenic tissues (liver and adipocytes). Deletion of the ACC1 gene in mice is embryo-lethal and has a pronounced effect on liver and adipose tissue lipid metabolism (16–18). Furthermore, lipogenesis is up-regulated in many tumors, increasing demand for ACC-made malonyl-CoA (19). A role of malonyl-CoA in hypothalamic sensing of energy, metabolite balance, and control of feeding behavior has been suggested (20). The level of malonyl-CoA is also controlled directly by malonyl-CoA decarboxylase (21). To develop new drugs for obesity or cancer one needs compounds that inhibit ACC and do nothing else. Because human ACC1 and ACC2 produce two separate pools of malonyl-CoA with dramatically different functions, isozyme-specific inhibitors are highly desirable. The current arsenal of small-molecule inhibitors of mammalian ACC includes several classes of compounds with different chemical cores and submicromolar IC₅₀ and, in some cases, a modest isozyme specificity (22–27). No drugs targeting human ACC have yet been developed, based on these compounds or others.

Previously, we showed that growth of yeast gene-replacement strains, in which the yeast ACC1 gene is replaced with genes expressing foreign ACCs, with ACC inhibitors correctly reflects the inhibitor specificity and the enzyme sensitivity. These observations present a convenient method for testing ACC inhibitors by monitoring yeast growth rather than by measuring enzymatic activity. We also showed that a comparative analysis of gene-replacement yeast strains containing various ACCs and their chimeras can be used to determine the specificity and binding site of ACC inhibitors (4 and 5).

In this article, we describe yeast gene-replacement strains suitable for high-throughput screening and the identification of unique inhibitors of eukaryotic ACCs, including both of the human ACC isozymes. This technology can be used for the discovery and characterization of compounds with unique central cores and binding specificity.

Results

Yeast ACC Gene-Replacement Strains.

Full-length cDNAs encoding human ACC1 and ACC2 were assembled from large cDNA fragments (see Materials and Methods) in the pRS423 vector carrying the GAL10 promoter and the 5′-UTR and 3'-UTR from the yeast ACC1 gene (28). A chimeric gene consisting of wheat and human ACC coding sequences was prepared by replacing a large fragment of the wheat cytosolic ACC coding region with the corresponding coding sequence of human ACC2 in a construct described previously (7 and 28). An overview of the domain structure of the ACCs found in eukaryotes and the structure of various synthetic ACC genes used in this study is shown in Fig. 1. The ACC coding sequences included in these constructs are described in Table S1. The wheat cytosolic ACC, the wheat cytosolic/plastid chimera, the T. gondii apicoplast ACC1, other similar chimeric genes encoding wheat, and Toxoplasma ACCs were described previously, along with the effect of known herbicides on the growth of yeast gene-replacement strains carrying such chimeric genes (4–7, 28). All of the constructs carrying ACC genes complemented the yeast ACC1 null mutation as previously described (28). The chimeric genes all use a GAL10 promoter to drive ACC expression, so growth of the haploid gene-replacement strains obtained by tetrad dissection is galactose-dependent.

Fig. 1.

An overview of the domain structure of the multidomain ACC and the structure of various chimeric genes complementing the ACC1 null mutation in yeast. The apicoplast targeting signal of Toxoplasma ACC was deleted in the construct shown (7). Two additional human ACC2 genes were constructed in which the N-terminal mitochondrial targeting signal was deleted (29- and 148-amino acid deletion, respectively). Only the coding parts of the synthetic genes in yeast shuttle vectors pRS423 carrying a GAL10 promoter and the 5′-UTR and 3′-UTR from the yeast ACC1 gene (28) are shown. Numbers indicate amino acid positions in native ACCs. Positions of mutations and other features described in the text and Table S1 are also shown. Names of the corresponding yeast gene-replacement strains (Table S1) are shown on the left, underlined.

Two human ACC1 gene variants, with and without a small alternatively-spliced exon in the middle of the coding sequence (Fig. 1), complement the yeast ACC1 null mutation. The full-length ACC2 gene with the N-terminal signal domain and the extra exon shown in Fig. 1 also complements the yeast null mutation. A variant without the extra exon does not complement the yeast null mutation. The initial human ACC2-complemented yeast strains grew poorly on plates but several mutants with improved growth properties were isolated. Our results indicate that these mutations are in the yeast genome. One of these mutants (strain ACC2m1, Table S1) was selected for future experiments. The strain is cold sensitive—it grows at 30 °C but not at 25 °C. We also isolated secondary mutants that grow at both temperatures. A chimeric ACC, composed of wheat cytosolic ACC and human ACC2 (Fig. 1) also complemented the yeast null mutation (Fig. 1). The resulting gene-replacement strain grew poorly, but again we were able to isolate mutants with improved growth properties (Table S1). These mutations are located either on the yeast chromosome (strain C50ACC2m5) or on the plasmid used for complementation (C50ACC2m3), shown by complementation with a rescued plasmid. The latter mutation (F1,498L, Fig. 1) was identified by sequencing the plasmid rescued from the mutant.

The relative growth rates of the gene-replacement strains showed that although active human ACC2 is made in yeast, its level is much lower than that of human ACC1. The low level of activity of the full-length human ACC2 in yeast could be due to misreading of the N-terminal mitochondrial targeting signal by the yeast transport machinery. Therefore, two variants of the human ACC2 with N-terminal deletions (29 and 148 amino acids, Table S1) were tested: both complemented the yeast null mutation, and for both of them the gene-replacement strains grow significantly faster then either strain ACC2m1 or C50ACC2m5. Even with the N-terminal deletions, the enzymatic activity of human ACC2 in protein extracts prepared from gene-replacement strains was very low. In contrast, the activity of human ACC1 could easily be detected in such extracts. We also found, using the yeast complementation test, that adding a His-tag to the C terminus or N terminus of full-length human ACCs does not abolish their activity.

All of these yeast gene-replacement strains, except ACC1/fl1 (T. gondii) and C100, grow significantly slower than wild-type (WT) haploid yeast strains, with doubling times 1.5–7 times higher than for the WT yeast (Table S1). The slow growth phenotype, presumably caused by a limiting level of ACC activity, is desirable in screening because it makes the strains more sensitive to inhibitors and allows more reliable identification of weak- and medium-strength inhibitors (7).

Growth Inhibition of Yeast ACC Gene-Replacement Strains by Known ACC Inhibitors.

A protocol was developed for high-throughput screening of chemical libraries using the yeast gene-replacement strains. This experiment relies on growing 150 μL yeast cultures in a 96-well plate with shaking and aeration under high humidity conditions for a period of up to 96 h. Yeast growth was assessed by measuring culture turbidity at 580–620 nanometers, using a plate reader. Based on a series of experiments using our collection of yeast gene-replacement strains and several inhibitors with known potency against various ACCs (e.g., herbicides affecting wheat plastid ACC or inhibitors of human ACCs) we determined the optimal culture volume, initial cell density, culture conditions, times for yeast growth determination for different strains depending on the doubling time, and 100 μM as the most informative compound concentration. Growth of yeast tester strains measured for an individual well was within 20% of a multiwell average. Growth greater than 80% of a control without inhibitor is considered “no inhibition” and growth less than 20% in the presence of an inhibitor is considered “strong inhibition.” In the high-throughput protocol, all compounds inhibiting yeast growth by more than 50% were rescreened for confirmation. Wild-type yeast is used in a counterscreen to eliminate compounds with general yeast toxicity.

The validation of our screening system came from studies using known human ACC inhibitors and yeast gene-replacement strains with either one of the full-length human ACCs: strain ACC2m1 and strain ACC1. Fig. S1 illustrates growth of these strains at two concentrations of CP-640186, an inhibitor developed by Pfizer (26): the ACC1 strain is only partially affected at 100 μM, whereas the ACC2 strain is completely inhibited. Growth of the ACC2m1 strain in the presence of 20 μM CP-640186 shows that the inhibitory effect is most pronounced when the yeast culture, in the absence of an inhibitor, reaches density corresponding to OD 1. None of the other yeast gene-replacement strains (Table S1) were inhibited by CP-640186, indicating strong specificity of this compound for human ACCs. Furthermore, our results show that although the carboxy transferase (CT) domain is targeted by CP-640186, another domain of human ACC2 is required for full inhibition: the strain C50ACC2m5, in which the N-terminal half of ACC is from wheat, was inhibited less strongly than the strain ACC2m1 with full-length human ACC2. The latter result confirms the suggestion from structural work which indicated involvement of domains other than the CT domain in effective binding of CP-640186 (29) and shows that the yeast system can be used for predicting the target domain of a new inhibitor. Inhibition of strain ACC2m1 by CP-640186 requires a relatively high compound concentration despite its IC₅₀ measured in vitro in the 0.1 μM range (Table S2). CP-640186 is known to be metabolically unstable (22) and most likely is converted to an inactive derivative in yeast. Despite this apparent instability, the screening system consistently identifies CP-640186 as a strong ACC2 inhibitor.

We also tested four other inhibitors of human ACCs described in (24). The yeast growth experiments (Fig. S2) reflect properly the specificity and potency of these compounds towards human ACC1 and ACC2, determined using enzymatic assays and recombinant proteins (Table S2). Compounds 9a, 9b, 9n, and 12 inhibit growth of the ACC2 strain completely (at 10 μM), whereas the ACC1 strain is inhibited strongly only by compound 12 and weakly by compound 9b. Lack of strong inhibition of strain C50ACC2m5 indicates that the CT domain of ACC2 is not sufficient to confer full sensitivity to these compounds. Weak inhibition of strain C50P50 by compounds 9b and 9n suggests that the CT domain of the wheat plastid ACC (P50 in C50P50) is targeted by these compounds as well. Growth of WT yeast and strains C100 (full-length wheat cytosolic ACC) and ACC1/fl1 (full-length T. gondii apicoplast ACC1) is not affected by these compounds.

In earlier experiments we showed that yeast gene-replacement strains carrying wheat ACCs and their various chimeras respond correctly to known inhibitors (herbicides) of the wheat plastid ACC isozyme (3–5). We tested the effect of six aryloxyphenoxypropionates (clodinafop, haloxyfop, quizalofop, diclofop, fenoxaprop, and fenoxaprop-P), three cyclohexanediones (sethoxydim, tralkoxydim, and tepraloxydim) and pinoxaden on the growth of strain C50P50 (Table S1) containing the CT domain from the wheat plastid ACC carrying the herbicide-binding site (Fig. 1). In all cases, these inhibitors exerted the expected strong growth inhibitory effect on yeast growth with IC₅₀ values similar to IC₅₀ values measured in ACC enzymatic assays using wheat germ extract as a source of active enzyme consisting predominantly of the plastid ACC activity. This finding further demonstrates the sensitivity and robustness of the yeast screening system for many compounds representing different chemical families.

Human ACC1 and ACC2 for in Vitro Enzymatic Assays.

In view of the low expression levels of human ACCs in yeast, we used instead the baculovirus/insect Sf9 cell expression system to produce both proteins for biochemical studies. Successful application of this technology for expression of human ACCs has been reported previously (30). The expression constructs were assembled by moving the gene cassette encoding full-length human ACC1 or truncated human ACC2 (148 amino acid N-terminal deletion) from the yeast construct into the baculovirus vector. Active human ACC1 (with His₆-tag fused at the C terminus) and ACC2 (with His₆-tag fused at the N terminus) are produced in these cells at a level of 2 mg per L of cell culture (determined after affinity purification and gel filtration on Sephadex G100).

Identification of New ACC Inhibitors.

Three screening libraries were developed by ChemDiv. “ACC1 Structure-Activity Relationship Library”—comprising 3,068 compounds selected based on ChemDiv’s internally developed “molecular descriptors similarity” approach using information about known ACC inhibitors as a starting point and applying drug-likeness criteria and bioisosteric morphing procedures. “Diverse Screening Library”—comprising 30,000 compounds selected from more than 1,000,000 compounds representing more than 12,000 different templates. This library included nature-like compounds, annotated biologically active privileged structures, and nonnatural peptidomimetic structures with drug-like characteristics, selected based on diversity criteria. Third, the “ACC2 Analog Screening Library”—was comprised of 828 analogs of potential ACC2 inhibitors identified during the screening of the first two libraries.

Screening of the ACC1 Structure-Activity Relationship Library produced three hits (0.1% rate) with potential selective inhibition of ACC2. The Diverse Screening Library yielded seven hits (0.02% rate) with a few of them being either nonselective regarding ACC2, or “toxic” as indicated by growth inhibition of WT yeast. Finally, the ACC2 Analog Screening Library produced 28 ACC2 selective hits (3.5% rate). The hits belong to five chemotypes: thieno[2,3-d]pyrimidine (11 hits), 5H-pyrrolo[1,2-a]quinoxalin-4-one (one hit), oxazolo[4,5-b]pyridines (3 hits), 2-piperidin-1-yl-1H-benzoimidazole (17 hits), and 1H-pyrazolo[3,4-b]pyridine (6 hits). Potency and specificity of these inhibitors towards strain ACC2m1, ACC1, WT yeast, and ACC2 enzymatic activity are shown in Fig. 2.

Fig. 2.

Specificity of the best hits. Growth inhibition of yeast strain ACC2m1 versus strains WT and ACC1 (A and B), and correlation between growth inhibition of strain ACC2m1 versus inhibition of ACC2 enzymatic activity (C) by the best hits found by screening chemical libraries described in the text. Arrow points to data points for compound CD-017-0191. Chemotypes: green, thieno[2,3-d]pyrimidine; purple, 5H-pyrrolo[1,2-a]quinoxalin-4-one; orange, oxazolo[4,5-b]pyridines; blue, 2-piperidin-1-yl-1H-benzoimidazole; red, 1H-pyrazolo[3,4-b]pyridine. Inhibition data are shown in Table S3.

Compared to the whole screening set, the octanol-water partition coefficient (cLogP) distribution of the ACC2 hits (Fig. S3A) as well as their molecular weights were shifted towards higher values. Both the cLogP values and molecular weights (two of the four Lipinsky’s rule of five criteria (31) were under 5 and 500, respectively, for the majority of the hits. The other two parameters, the number of hydrogen bond acceptors (< 10) and number of hydrogen bond donors (< 5), were also satisfied for the majority of the hits. Distribution of polar surface area, which correlates well with good intestinal absorption (32), for the majority of the hits was shifted toward smaller values (Fig. S3B). Thus, the hits found possess physicochemical properties that conform to Lipinski’s rule of five for compounds with good drug-likeness characteristics.

Compound CD-017-0191 (Fig. 3) from the first library inhibits human ACC2 activity in vitro with IC₅₀ = 2.8 μM, very similar to the IC₅₀ = 3.8 μM observed in the yeast growth experiment (Fig. 4, Table S2). For comparison, the Abbott reference inhibitors 9a, 9b, 9n, and 12 tested in parallel with CD-017-0191 showed in vitro IC₅₀ = 0.01–0.05 μM. CD-017-0191 did not inhibit growth of WT yeast or the ACC1 strain even at 100 μM. ACC1 enzymatic activity was also not inhibited at 100 μM. Compound CD-017-0191 is not toxic to human cells in culture at concentrations below 10 μM. Two additional compounds representing the same chemotype were found in the third library, both less potent than CD-017-0191 (Fig. 2, Table S3).

Fig. 3.

Structure of CD-017-0191, 2-[[[5-methyl-2-[4-(1-methylethoxy)phenyl]-4-oxazolyl]methyl]thio]-oxazolo[4,5-b]pyridine), and CD-017-0213, 4-(2-methyl-5,6,7,8-tetrahydrobenzo[4,5]thieno[2,3-d]pyrimidin-4-yl)-1-ethoxycarbonylpiperazine.

Fig. 4.

(A). Growth inhibition of yeast gene-replacement strains ACC2m1 and ACC1 (Fig. 1, Table S1) and WT yeast by compound CD-017-0191. (B). Inhibition of ACC1 and ACC2 enzymatic activity in vitro by compound CD-017-0191. Squares, growth of strain ACC2m1 and enzymatic activity of ACC2; circles, growth of strain ACC1 and enzymatic activity of ACC1; triangles, growth of strain WT.

Specificity of compound CD-017-0191 was further tested using yeast gene-replacement strains with ACCs from different species (Fig. 5). As expected, growth of strain ACC2Δ148, whose ACC differs from ACC2m1 by deletion of 148 amino acids at the N terminus, is strongly inhibited by CD-017-0191. Growth of the ACC1 strain and WT yeast is not inhibited, consistent with the screening results and IC₅₀ measurements shown in Fig. 4. Growth of other strains is not inhibited or is only 20% inhibited (for wheat ACCs) at 100 μM CD-017-0191, indicating strong specificity of this compound for human ACC2. Lack of any inhibition of strains C50ACC2m3 and C50ACC2m5 indicates that the CT domain of human ACC2 is either not the target of CD-017-0191 or it does not contain the entire binding site of the inhibitor.

Fig. 5.

Growth inhibition of a collection of yeast gene-replacement strains by CD-017-0191. ACC composition in different yeast strains is shown in Fig. 1. Inhibitor concentration: lighter gray bar, 10 μM; darker gray bar, 100 μM. Culture density in the absence of inhibitor = 100%.

Compound CD-017-0213 (Fig. 3), which was initially scored as a strong inhibitor of strain ACC2m1, showed 20 μM IC₅₀ for both ACC2 and ACC1 in vitro (Table S2). Follow-up tests showed that growth of the yeast strains used in the experiment illustrated in Fig. 5 is inhibited by 50%–80% by 100 μM CD-017-0213. These results suggest that this compound is a nonspecific ACC inhibitor.

Compounds, other than CD-017-0191 and CD-017-0213 l, identified in the screens as inhibitors of the ACC2m1 strain, showed 10%–90% inhibition of ACC2 activity in vitro at 100 μM (Fig. 2C). No or partial inhibition of ACC2 activity by some of these compounds may be due to their conversion to active forms during the yeast assay but not under conditions of the in vitro assay. Compound conversion in the medium or in vivo by yeast enzymes may be beneficial as it could increase the effective number of compounds tested in the screen, for example, if inactive esters are hydrolyzed to active free acids, they may help identify new chemotypes for further analysis by combinatorial chemistry chemical modification or identification of the active converted product followed by its de novo synthesis. Potentially important lead compounds may be identified this way which otherwise could not be found using in vitro ACC enzymatic assays.

Tests of three aryloxyphenoxypropionates and their esters showed that both forms of these compounds are equally effective inhibitors of yeast strain C50P50 carrying a chimeric wheat ACC sensitive to these compounds. Aryloxyphenoxypropionates are effective only as free acids (33). This result shows that yeast can hydrolyze these esters to the active acids.

Compound conversion during the yeast assay could prevent unstable compounds from being detected by the yeast screening system. The 100 μM compound concentration used in the screen is high enough to ensure that the impact of degradation on screening results is restricted to compounds converted to inactive derivatives with decay half-times of hours. CP-640186 was reproducibly scored as a strong inhibitor of human ACC despite its known metabolic instability (above). In addition to CP-640186, we tested four other classes of ACC inhibitors: compounds 9a, 9b, 9n, and 12 targeting human ACCs as well as aryloxyphenoxypropionates and their esters (9 compounds), cyclohexanediones (3 compounds) and pinoxaden targeting wheat ACC. All of the compounds with low IC₅₀ values in the in vitro ACC assay showed strong inhibition of appropriate yeast gene-replacement strains in the proposed HTP screening measurements.

No structural chemical similarity has been found between CD-017-0191 and CD-017-0213 and 158 known ACC inhibitors available from the Integrity database (Prous Science SAU, Barcelona, Spain). There are no references to CD-017-0191 and its analogs (16 compounds) in SciFinder database. There is only one reference to CD-017-0213 and its analogs (10 compounds) in the SciFinder database, but it is not related to ACC inhibition. Only very low structural chemical similarity has been found between our molecules and compounds compiled in the SciFinder database (Fig. S4). CD-017-0191 and CD-017-0213 therefore represent unique chemotypes with moderate potency and very low structural relatedness to known ACC inhibitors as well as known drugs in general.

Discussion

An inexpensive, nonradioactive, high-throughput screening system, using yeast gene-replacement strains, suitable for the identification of new strong inhibitors of ACC has been developed. The strategy is to screen libraries using the yeast strains and to verify targets of any positive hits using an in vitro ACC enzymatic assay. Validation of the screening system with known potent inhibitors of human ACCs demonstrated its robust sensitivity allowing for high-throughput identification of new inhibitors.

Extensive studies of five different classes of strong ACC inhibitors, three classes of herbicides targeting plant multidomain ACC, and two classes of compounds targeting human ACCs, showed that growth inhibition of the yeast gene-replacement strains reflects faithfully the sensitivity of the foreign ACC to these inhibitors. Furthermore, these results have shown that the screening principle works for a variety of compounds with different chemical structures and properties. Even in the case of a metabolically unstable inhibitor of human ACC2, an appropriate yeast gene-replacement strain responded properly to the inhibitor although at concentrations significantly higher than expected from the compound’s IC₅₀ measured in enzymatic assay in vitro.

Yeast gene-replacement strains with ACCs from several species and their chimeras have been established to test specificity and mode of action of any new ACC inhibitor. These strains can serve as screening strains for inhibitors/drug candidates with an alternative specificity. They can also be used to isolate inhibitor-resistant ACCs to identify amino acid residues lining the binding pocket of newly discovered inhibitors and determine their mode of action.

Yeast gene-replacement strains can serve as a source of different foreign ACCs and their chimeras for enzymatic studies on new inhibitors. Transferring gene cassettes identified in the yeast system to the baculovirus system is simple and efficient. The baculovirus sf9 insect cell expression system is superior and capable of producing large quantities of highly active multidomain ACCs for biochemical and structural studies.

Screening of designer chemical libraries identified a series of compounds that are potential specific inhibitors of human ACC2. These chemicals will serve, upon further analysis, as a first set of lead compounds for the development of drug candidates targeting human ACC2 to treat obesity. Specificity and potency of one of these compounds was confirmed in in vitro ACC enzymatic tests. Its IC₅₀ for human ACC2 is 2.8 μM. This compound does not inhibit human ACC1 at all (in vitro or in vivo), up to 100 μM. It does not affect ACCs from several other sources either. Using a gene-replacement strain with a chimeric ACC, we have shown that this compound targets either a domain other than the CT domain or one that includes additional structural elements. Another compound identified in the screen was shown to be a nonspecific ACC inhibitor with in vitro IC₅₀ of 20 μM for both human ACCs.

Our experiments confirm the feasibility of a high-throughput screen of large chemical libraries using the yeast gene-replacement-based technology. The system could be useful for identifying new drugs for the treatment of parasite diseases, since parasite ACCs such as those of Plasmodium and Leishmania differ in sequence from both human ACCs.

The extensive studies of human ACCs, including the weight-loss experiments with the ACC2 knockout mouse from the Wakil laboratory, validated the enzyme as a potential target for drugs to treat obesity. Specific small-molecule inhibitors of the enzyme, such as the compound identified in this study, provide useful leads for future drug development efforts.

Materials and Methods

A full description of the materials and methods used in this work is provided in SI Text. This information includes the sources of chemical inhibitors, the assembly of the clones shown in Fig. 1, the complementation tests used to characterize each cloned gene, the preparation of acetyl-CoA carboxylase from recombinant yeast strains as well as from insect cells, the conditions used for assay of the enzymes, and the conditions used for assay of toxicity of the inhibitors toward human cells in culture.