Weight loss weightloss weight loss programs weight loss foods weight loss tips
Cancer chemotherapy targeting frequent loss of heterozygosity events is an attractive concept, since tumor cells may lack enzymatic activities present in normal constitutional cells. To find exploitable targets, we map prevalent genetic polymorphisms to protein structures and identify 45 nsSNVs (non-synonymous small nucleotide variations) near the catalytic sites of 17 enzymes frequently lost in cancer. For proof of concept, we select the gastrointestinal drug metabolic enzyme NAT2 at 8p22, which is frequently lost in colorectal cancers and has a common variant with 10-fold reduced activity. Small molecule screening results in a cytotoxic kinase inhibitor that impairs growth of cells with slow NAT2 and decreases the growth of tumors with slow NAT2 by half as compared to those with wild-type NAT2. Most of the patient-derived CRC cells expressing slow NAT2 also show sensitivity to 6-(4-aminophenyl)-N-(3,4,5-trimethoxyphenyl)pyrazin-2-amine (APA) treatment. These findings indicate that the therapeutic index of anti-cancer drugs can be altered by bystander mutations affecting drug metabolic genes.
Recent targeted anti-cancer therapies exploit acquired genetic differences between cancer and normal cells, such as activation by mutation of a specific oncogene, inactivation by mutation of a tumor suppressor gene, or perturbation of pathways involved in the maintenance of genome integrity to achieve preferential killing of cancer cells1. Drugs targeting protein tyrosine kinases are mainstays of clinical cancer care2,3,4, and strategies addressing loss-of-function of tumor suppressor genes such as TP535,6 and RB7 are in development. Collateral lethality, exploiting vulnerabilities left in cancer cells after a passenger gene in the vicinity of a tumor suppressor is lost, has emerged as a new therapeutic avenue. The frequent 1p36 deletion in glioblastoma, entailing loss of ENO1, sensitizes tumor cells to ENO2 inhibition8. Similarly, complete losses of POLR2A, MTAP, and ME39,10,11,12, and partial losses of PSMC213, SF3B114, and MAGOH15 have been identified as potential therapeutic targets in human cancers. Targeting of the loss of heterozygosity (LOH) events occurring as cancer genomes inactivate tumor suppressor genes has been achieved by allele-specific inhibition, where variants of essential genes such as the 70-kDa subunit of replication protein A (RPA70) near TP53 are silenced using antisense oligonucleotides16. However, allele-specific LOH therapy targeting proteins tractable to inhibition by systemically administered agents has previously not been demonstrated. We here demonstrate that a recurring loss of heterozygosity event affecting a drug metabolic activity (NAT2) can increase the sensitivity to a low molecular weight cytotoxic compound.
Identification of N-acetyltransferase 2 (NAT2) as a target for allele-specific inhibition
To identify target proteins, we mapped variants observed in the 1092 individuals in the 1000 Genomes project to functional domains and crystal structures to identify those that could alter protein structure in catalytic or substrate binding sites (Fig. 1a). The variants were first mapped to transcripts, localizing 482,280 small nucleotide variants (SNVs) to protein coding regions. Of these, 78% were likely to be present in only one population17. To enrich for prevalent targets of higher utility in LOH-directed therapies, 23,532 non-synonymous SNVs in functional protein domains with allele frequency ≥0.5% were selected. Of these, 1367 SNVs (~5.8%) mapping to 566 crystal structures had both the SNV and active or substrate binding sites defined in the structure. After visual inspection, 56 SNVs in 45 intracellular proteins resulted in amino-acid substitutions near catalytic residues or substrate binding pockets, including 36 SNVs in 26 genes with >5% heterozygosity in all 1000 Genomes populations (Supplementary Data 1). Finally, retaining only the genes with >15% LOH frequency in common human cancers18 and with a gene expression profile matching the tissues of interest, yielded 17 potential target enzymes for LOH-directed cancer therapies (Supplementary Table 1). From the putative target genes, we selected polymorphisms in NAT2, AKR7A2, and SULT1A1 for validation as these genes were known to be involved in drug metabolism and are non-essential for cell survival. Other genes with common variants were considered but not prioritized because of likely limited chemical space of substrates (HSD17B), small reduction in catalytic activity by the variant (GSTP1), unknown importance (HAAO), known substrate redundancy with related enzymes (GSTP1, SULTs), or too few LOH events in CRC (ABP1 and AKR7A). To confirm the expected prevalence of candidate SNVs and frequency of LOH events, we genotyped the selected SNVs in 74 patients with chromosomally unstable colorectal cancers (CRCs) and could detect all in heterozygous states (Supplementary Table 2). Next, tumors from heterozygous individuals were assessed for somatic LOH events. The highest frequency of allelic loss was observed for NAT2, where ~7% heterozygous for the rs1799930 polymorphism retained only one allele in their tumors because of LOH (Fig. 1b; Supplementary Table 2). We observed a similar likelihood for CRCs to lose either allele (Fig. 1b), in an LOH event detected in stage II as well as stage III and IV CRCs (Supplementary Fig. 1), in line with NAT2 being a bystander gene on 8p which is lost early in CRC development19. After genotyping, based on the frequency of the LOH events NAT2 was prioritized as a proof of concept gene over AKR7A2 and SULT1A1. The NAT2 gene encodes one of the two human N-acetyltransferases involved in xenobiotic metabolism of arylamines and hydrazine compounds. NAT2 is highly polymorphic with 108 allelic variants identified in human populations [http://nat.mbg.duth.gr/]. The SNV rs1799930 defines the NAT2*6 group of variant alleles (R197Q) with ≥10-fold reduced activity compared with wild-type NAT2, considered to encode a rapid acetylator phenotype20,21 (Fig. 1c). Whereas NAT1 is expressed in essentially all tissues, NAT2 is confined to the liver and gastrointestinal tract (GI)22,23. From the putative target proteins identified, we selected NAT2 because of its (a) known role in drug metabolism, (b) non-redundant and well defined substrate specificity, (c) GI restricted expression, (d) location on a chromosome arm frequently lost during transition from colorectal adenoma to carcinoma24, (e) 10-fold activity difference between the gene products encoded by the wild-type and a common variant allele, and (f) clinically relevant substrates as ~10 commonly used drugs are subject to NAT2 metabolism, among them the cytotoxic drug amonafide25, where rapid acetylators have ~4.5-fold higher plasma ratio of N-acetyl-amonafide to amonafide than slow acetylators leading to increased systemic toxicity26.
Discovery of compounds selectively toxic to slow NAT2 cells
We reasoned that anti-cancer drugs rendered less toxic by NAT2 metabolism are also likely to exist and that CRC cells having lost a rapid NAT2 allele through LOH could be sensitized to treatment with a cytotoxic NAT2 substrate relative to other constitutional cells retaining the rapid allele (Fig. 1d). Therefore, cell systems for small molecule library screening were engineered in human CRC RKO and DLD-1 cells by transfection with NAT2 expression vectors encoding slow NAT2*6A (rs1799930) or rapid NAT2*13A (wt) alleles (Figs. 2a, b). Both cell lines are homozygous for the slow NAT2*6A allele and have low endogenous NAT2 expression. The acetylation velocities of the NAT2 substrates amonafide and procainamide were ≥8-fold higher in rapid NAT2 clones when compared with slow NAT2 in RKO (p < 0.0001) and DLD-1 (p < 0.05) clones (Figs. 2c, d). To identify agents with selective toxicity toward CRC cells only expressing slow NAT2, we searched a chemical library for arylamine compounds with demonstrated cytostatic or cytotoxic effects, resulting in 176 candidate substances from a total of 189,018 (Fig. 2e). Next, dose-response experiments were performed to determine cytotoxicity of these potential NAT2 substrate compounds in RKO cells (Supplementary Fig. 2 and Supplementary Table 3). Ten substances showed cytotoxicity at 10 µM and one (6-(4-aminophenyl)-N-(3,4,5-trimethoxyphenyl)pyrazin-2-amine; APA) selectively killed slow NAT2 cells (Fig. 2e). A ~3-fold difference in growth inhibition of the rapid NAT2 clones (EC50 0.97 µM for RKO and 2.94 µM for DLD-1 cells) versus NAT2-deficient cells (EC50 0.33 µM for RKO and 1.21 µM for DLD-1) was detected following APA treatment (Figs. 2f, g). Silencing of endogenous NAT2 in HCT116 CRC cells harboring rapid and slow NAT2 alleles sensitized them to APA, but not to 5-fluorouracil (Supplementary Fig. 3). To translate the observed difference in cytotoxicity to NAT2 efficiency, cell-based enzyme kinetics were determined. The rapid NAT2 RKO cells had increased affinity of APA acetylation, and the specificity of rapid versus slow NAT2 (Vmax Km−1) was 40-fold at similar levels of NAT2 protein (Figs. 2a and 3a). Recombinant assays with human NAT1 and NAT2 further proved APA acetylation to be specifically mediated by NAT2 (Vmax Km−1 > 20-fold) (Fig. 3b). Together, APA is a cytotoxic agent which is directly inactivated by NAT2.
Mediators of APA toxicity in slow NAT2 cells
Structural similarities between APA and kinase inhibitors27 prompted investigation of protein kinases as mediators of APA toxicity. We reasoned that N-acetyl-APA (NAPA), the less toxic NAT2 metabolite of APA, should not effectively inhibit the target kinases. When the binding affinity of APA and NAPA to 468 protein kinases was determined, APA had ≥10% binding affinity to 235 human kinases, while NAPA showed reduced binding affinity to 89 of these (Supplementary Fig. 4A). To enrich for kinase interactions that could explain the preferential killing of slow NAT2 cells, we compared the binding affinity of each kinase with APA and NAPA, selected 44 targets with preferential binding to APA and excluded kinases non-essential for cell survival and those not expressed in CRC cells (Supplementary Fig. 4A, Supplementary Table 4). The remaining kinases DYRK1A, AURKA, and CDK7, were subject to siRNA knockdown in RKO and DLD-1 cells (Supplementary Fig. 5A, B) and knockdown of AURKA (Aurora kinase A) resulted in growth inhibition of both (Figs. 3c, d). The growth inhibition of RKO and DLD-1 following siRNA-mediated AURKA knockdown (Figs. 3c, d) reflected the relative efficiency of protein downregulation (Supplementary Fig. 5C, D). While APA bound to AURKA (Kd = 13.5 µM), NAPA did not (Supplementary Fig. 4B). Thermal shift assay demonstrated a thermal unfolding profile of AURKA in presence of APA similar to the AURKA inhibitor Tripolin A28, albeit weaker than that of S145129 (Supplementary Fig. 4C). AURKA regulates cell cycle progression through association with other centrosome and spindle-associated proteins. Its activation requires auto-phosphorylation at Thr288 and ensures spindle pole integrity, centrosome duplication as well as adequate chromosomal alignment and segregation30, while its inhibition leads to multipolar spindles with acentrosomal poles28,31. Upon APA treatment, DLD-1 cells lacking NAT2 had >40% reduced phosphorylation of AURKA at Thr288 (pAURKA), while rapid NAT2 cells did not differ from DMSO-treated control (Figs. 3e, f and Supplementary Fig. 6A). S1451 reduced the fraction of pAURKA levels by >90% regardless of NAT2 expression (Fig. 3f and Supplementary Fig. 6A). Total AURKA protein levels were decreased after APA or S1451 treatment (Fig. 3g)32. Treatment with APA increased the levels of the mitotic marker histone H3 pSer10 (Fig. 3g). Further, both APA and nocodazole arrested NAT2-deficient cells in mitosis, as demonstrated by high levels of histone H3 pSer10, whereas only nocodazole arrested NAT2-proficient cells (Supplementary Fig. 6B). The number of cells with mitotic spindles after APA treatment increased by ~15% in DLD-1 cells lacking NAT2, but not in rapid NAT2 cells where <5% cells had mitotic spindles (Supplementary Figs. 6B, C). While the full compendium of mechanisms for APA cytotoxicity remains to be characterized, these results show that AURKA is a target of APA but not NAPA in vitro, and that the cytotoxicity resulting from kinase inhibition may be mediated by AURKA inhibition.
Evaluation of the in vivo anti-tumor activity of APA
Systemic administration of APA would likely result in limited tumor exposure, as the compound was prone to rapid oxidative metabolism in liver microsomes (Supplementary Table 5). We therefore used sterically-stabilized liposomes to increase tumor targeting via the enhanced perfusion and retention (EPR) effect, protecting normal cells from the toxic substrate and enriching the drug in the tumor33. To determine the anti-tumor efficacy of APA, athymic mice xenografted on each flank with rapid and slow NAT2 tumors received free APA, liposomal APA, or control liposomes. Slow NAT2 tumors treated by liposomal or free APA were 50% smaller compared with those treated with control liposomes (p < 0.05, Supplementary Fig. 7A). This difference was not observed with rapid NAT2 tumors (Supplementary Fig. 7B). Further, a modest, but measurable >30% decrease in the ratio of slow to rapid NAT2 tumor volumes was observed for animals treated with liposomal APA (p < 0.05, Fig. 4a), meaning that rapid NAT2 tumors on average grew faster than the corresponding slow NAT2 tumor on each animal. In comparison, the same ratio was unchanged in mice treated with control liposomes and free drug (Fig. 4b; Supplementary Fig. 7C). The growth inhibitory effect of liposomal but not free APA supports the idea that encapsulation minimizes APA metabolism en route to the tumor by mouse Nat enzymes.