Recent development of ATP-competitive small molecule phosphatidylinostitol-3-kinase inhibitors as anticancer agents

Phosphatidylinostitol-3-kinase (PI3K) is the potential anticancer target in the PI3K/Akt/ mTOR pathway. Here we reviewed the ATP-competitive small molecule PI3K inhibitors in the past few years, including the pan Class I PI3K inhibitors, the isoform-specific PI3K inhibitors and/or the PI3K/mTOR dual inhibitors.

To date, a total of eight PI3Ks have been identified, which are divided into four classes (I, II, III and IV) based on their sequence homology. As the most relevant to the PI3K/Akt/mTOR pathway (Figure 1), Class I PI3Ks are always referred to as PI3Ks [21]. Generally, Class I PI3Ks are further divided into IA and IB based on their different regulatory subunits and upstream activators [22]. Class IA PI3Ks are activated by RTKs and GPCRs, which contains three isoforms (PI3Kα, PI3Kβ and PI3Kδ) with the respective p110 catalytic subunit (p110α, p110β, and p110δ) bound to the p85 regulatory subunit [21]. Class IB PI3K consists of PI3Kγ, with the p110γ bound to p101 or p84, which is mainly activated by GPCRs such as chemokine receptors [23]. PI3Kα is known to play an important role in tumor genesis, which has been detected with persistent mutations and amplification in most human cancers including breast, ovarian, colorectal, stomach and gastric cancers [17,24,25]. PI3Kβ involves in the development of thrombotic diseases by activating platelets, while PI3Kγ and δ are the therapeutic targets of inflammatory and auto-immune diseases [22]. Beside PI3Kα, the other three isoforms (β, γ and δ) are also involved in tumor genesis, especially in the case of PTEN loss or inactivation. Moreover, as PI3K mutation and PTEN inactivation have been shown to be the causes of resistance to other targeted cancer therapies [26], the PI3K may even circumvent drug resistance to current chemotherapy in combination with other anticancer drugs [27].

DEVELOPMENT OF PI3K INHIBITORS
The major PI3K inhibitors currently available are reversibly ATP-competitive. The X-ray crystal structure of PI3K [21] and those of its complexes with ATP, Wortmannin (1), LY294002 (3) [28,29] and other diverse inhibitors facilitated and accelerated the development of PI3K inhibitors.

Review
Driven by efforts in computer-based rational drug design and SAR (Structure-Activity Relationship) studies, numerous promising PI3K inhibitors have been developed and a dozen of them have entered clinical trials for treatment of cancer or other diseases (Table 1) [19]. The first approved PI3K inhibitor Idelalisib (Gilead Sciences, Inc., also known as CAL-101 and GS-1101), an orally bioavailable PI3Kδ selective inhibitor with high potency and selectivity (p110δ IC 50 = 2.5 nM), was approved by the FDA in July 2014 for the treatment of several hematological malignancies, in combination with rituximab.
However, due to the high sequence homology of the catalytic domains and the conserved ATP-binding site, the key point for the development of PI3K inhibitors is to gain sufficient isoform-selectivity (δ and/or γ vs. α and β) and cross-kinase selectivity. Although it's not an easy task, the discovery of isoform-specific PI3K inhibitors were facilitated by the elucidation of the X-ray crystal structure of PI3K isoforms and those of its complexes with diverse inhibitors [27,29,30]. Additionally, because mTOR is PI3K-related kinase that has similar ATP site with PI3K, a number of PI3K inhibitors could also exhibit inhibitory activity against mTOR (PI3K/mTOR dual inhibitors), which may be more effective by delivering a powerful two-spot inhibition of the pathway and have the advantage of being less susceptible to PI3K drug resistance and IGF) or chemokine. Subsequent activation of the lipid PI3K leads to the phosphorylation of PIP2 to PIP3, which activates AKT and PDK1. Besides direct activation by PIP3, Akt could also be activated by PDK1 and mTORC2 (Rictor-mTOR). Then mTORC1 (Raptor-mTOR) was finally activated, which regulates cell growth, glucose and lipid metabolism, autophagy as well as protein synthesis, while mTORC2 regulates cell survival and actin reorganization. Additionally, the pathway is negatively regulated by PTEN.
abrogating the compensatory effects of mTOR inhibitors [31]. In fact, most of the drug candidates in this area were PI3K/mTOR dual inhibitors (Table 1).
In this review we provide a recent view about the PI3K inhibitors including the pan PI3K inhibitors, the isoform-specific PI3K inhibitors and the PI3K/mTOR dual inhibitors, as anticancer drugs in the PI3K/Akt/ mTOR pathway. As there are many well-written reviews in this field before, the novel PI3K inhibitors we emphasized here are those developed in the past eight years. Based on their core structures, the inhibitors were divided into five series: natural product derivatives; pyrimidines and quinazolines; pyridines, quinolines and indoles; pyrazines and quinoxalines; azoles and others.
To overcome the disadvantage of toxicity due to lack of selectivity, poor solubility, and low stability [33], derivatives with good pharmacodynamics were studied and PX-866 (4), the stable, furan-ring-opened derivatives of Wortmannin was thus identified (PI3Kα/δ/γ IC 50 = 5.5/9.0/2.7 nM), which is currently being evaluated in phase I/II trials for the treatment of patients with advanced solid tumors.
Besides, many recent studies have demonstrated that a variety of natural products (or nutraceuticals) isolated from plants (e.g. fruits, vegetables, spices, nuts, legumes, herbs, etc.) also inhibit PI3K signaling, and exhibit potent anticancer activities. In 2013, Huang [35] briefly summarized that Apigenin, a family member of flavonoids, abundant in fruits (oranges, apples, cherries,grapes), vegetables (onions, parsley, broccoli, sweet green pepper, celery, barley, tomatoes)and beverages (tea, wine); Cryptotanshinone, one of the major tanshinones isolated from the roots of the plant Salviamiltiorrhiza Bunge (Danshen); Curcumin (diferuloylmethane), a polyphenol natural product of the plant Curcuma longa; Fisetin, a family member of flavonoids, occuring in fruits and vegetables( such as strawberries, apples, persimmons and onions); Indoles , natural compounds in cruciferous vegetables (such as broccoli, cauliflower, cabbage and brussels sprouts), especially indole-3-carbinol and it's in vivo dimeric product 3,3-diindolylmethane (DIM); Isoflavones, a class of flavonoid phenolic compounds, rich in soybean; Quercetin, a polyphenolic compounds, mainly from consumption of tea, onions, red grapes, and apples; Resveratrol, a natural polyphenol rich in red grapes and red wine; Tocotrienols, members of vitamin E superfamily; and many other natural products( such as caffeine , epigallocatechin gallate (EGCG, in green tea), celastrol (in traditional Chinese medicine named "Thunder of God Vine"), butein (in the stems of Rhus verniciflua, used as a food additive and as an herbal medicine in Asia), capsaicin (in chili peppers) and β-elemene (from the traditional Chinese medicinal herb Rhizoma zedoariae), etc.), have been reported to act as anticancer agents at least partly by inhibiting PI3K, Akt or mTOR activity.
In 2008, series of thieno [3, 2-d] pyrimidine derivatives were prepared and evaluated as inhibitors of PI3K p110α by Folkes et al [36]. The lead (9) was reported as a potent PI3Kα inhibitor, but with poor pharmacokinetic profile. Derivatives with substitution of 6-positions (10) and the replacement of phenol group (11) were synthesized. Indazoles (12) as the replacements of phenols serve as a hydrogen bond donor with Tyr836, while reduced glucuronidation and resulted in acceptable oral bioavailability. This resulted in the discovery of GDC-0941(compound 13, PI3Kα/β/δ/γ IC 50 = 3/3.3/3/7.5 nM, mTOR IC 50 = 0.58µM), a potent, selective, orally bioavailable inhibitor of Class I PI3K including the p110α mutant enzymes, and is currently being evaluated in human clinical trials for cancer treatment.
Liu et al [40] in Pfizer discovered 4-methylpteridinones (36) as orally active and selective PI3K/mTOR dual inhibitors (PI3Kα Ki ~ 2-82 nM, mTOR Ki ~ 0.85-3940 nM), with non-selective inhibitor 2-aminopyridopyrimidinone (35) as the lead. The 4-methylpteridinones were designed based on a small special pocket within PI3K and mTOR binding pocket to improve selectivity against other kinases, which was to be able to accommodate the methyl group of compound 37 and 38. Series of compounds (e.g. 39) with excellent selectivity for PI3K and mTOR were discovered. In addition, small changes in the C-6 aryl group will have profound effects on either PI3K or mTOR potency.
Malagu et al [42] discovered a novel series of mTOR kinase inhibitors (45,46), but the PI3K inhibitory activity was not revealed, except that of compound 47, which was 8.9µM against PI3Kα, 1000-fold less potent than that against mTOR.
In 2010, Chen et al [10] reported a series of 4-morpholinopyrrolopyrimidine derivatives as PI3K inhibitors, by the modification of compound 48, an imidazolopyrimidine derivative with good PI3Kα activity (PI3Kα IC 50 = 63 nM). Followed by modification on the N5 of the imidazole ring, the 3-hydroxyl group on the phenyl ring and the N7 position (49-50), 4-ureidobenzamide derivatives with extended amino groups (51)(52)(53) were synthesized with excellent cell potency. As the most potent compound, 54 (PI3Kα IC 50 = 0.9 nM, mTOR IC 50 = 0.6 nM) also demonstrated in vivo antitumor efficacy. The replacement of the 3-hydroxy methyl group with 4-arylurea is outstanding, which not only improved metabolic stability but also increased enzyme potency and cell potency [7] .
In 2011, Burger et al (Novartis) [49] discovered a series of 2-morpholino, 4-substituted, 6-heterocyclic morpholino pyrimidines (84-86, PI3Kα IC 50 ~ 2-7740 nM) as potent PI3K inhibitors. The lead compound 83 was a potent pan class I PI3K inhibitors (PI3Kα IC 50~5 0 nM) with poor pharmacokinetic properties due to the phenol group. Then the C6 phenol moiety was replaced by diverse heterocycles, and the aminopyridine turned to be the best choice, being equipotent to the phenol. After the C4 position was further optimized, pharmacokinetic and efficacy study conducted, compound 87 (PI3Kα IC 50 < 2 nM) was identified with efficacy and suitable in vivo pharmacokinetic properties.
Then in their continued study [50] , C4'modified, C6 pyridyl or pyrimidyl substituted 2-morpholino 4-aminoquinolyl pyrimidines (88) were synthesized and evaluated, aiming to improve potency and reduce the in vivo CL values. "Incorporation of a morpholine group at the C4 position increased the aqueous solubility while maintaining potency, selectivity, and in vivo properties". This led to the discovery of substituted 6-aminoheterocyclic 2, 4-bis morpholino pyrimidines (89), of which the highest soluble and the most potent compound was compound 90 (NVP-BKM120, PI3Kα IC 50 ~ 30nM, mTOR IC 50 ~ 4600nM) that has entered into Phase II clinical trials for the treatment of cancer.
Heffron et al (Genentech) [52] described two chemical series achieving PI3Kα selectivity versus PI3Kβ, which could be explained using homology model of PI3Kβ. In the thienopyrimidine series (96, PI3Kα Ki ~ 0.4-47 nM, PI3Kβ Ki ~ 7-1167 nM), the selectivity (e.g. 98 and 99) was derived from "a hydrogen bonding with Arg770 of PI3Kα that is not attained with the corresponding Lys777 of PI3Kβ". In the benzoxepine series (97), the selectivity (e.g. 100 and 101) was due to the "electrostatic potential differences between the two isoforms in a given region".
Using PI-103 (102) as the lead, Large et al [53] designed two series of trisubstituted pyrimidines, 3-hydroxyphenol analogues (103)(104) and bioisosteric replacements (105) , as PI3K inhibitors. The 3-phenolic motif was replaced by three surrogate types (A, B and C), to avoid the glucuronidation in vivo. The most potent inhibitor was 6-aryl substitution compound 106 (PI3Kα IC 50 =62 nM), with similar activity against PI3Kβ and δ. All three surrogate types had metabolic stabilities and inhibitory activity similar to those of parent phenols.
To discover dual pan-PI3K/mTOR inhibitors, Poulsen et al [67] generated a pharmacophore model and designed a series of novel compounds based on a purine scaffold. Three scaffolds (A-C) having a purine core substituted with a morpholine, a phenol headgroup, and a hydrophobic substituent were initially designed from three reference compounds PI-103 (102) Dihydropyrrolopyrimidine derivative (180, PI3KαIC 50 = 42 nM) was identified as a metabolically stable and potent PI3K inhibitor initially, while with poor oral bioavailability. To remove the H-bond acceptor and recover the water solubility, Kawada et al [68] designed pyridine, benzylamine and benzamide derivatives (181, PI3KαIC 50 ~ 14-220 nM), by adding amine or amide (piperazine, morpholine) as a solubilizing group, replacing pyridine with a phenyl moiety and introducing an orthosubstituent in the phenyl group. Finally, compound 182 was identified with good pharmacokinetic profiles (oral bioavailability in monkey 8 times better than that of compound 180) and PI3Ka inhibition (PI3Kα IC 50 = 42 nM).

PYRIDINES, QUINOLINES, INDOLES AND INDAZOLES
Mostly, the few PI3K inhibitors based on pyridine, quinoline and indole structures reported since 2010 have been identified to be selective against PI3Kα or mTOR, including GSK2126458 (143), the most potent PI3K inhibitor with low picomolar activity.
Hong et al [73] identified a series of [3, 5-d]-7azaindole analogs (202-203) as PI3Kα inhibitors, by varying groups on the 3, 5-positions of azaindole. In their pharmacophore-directed design, through fragmentbased approach, 7-azaindole possessing both H-donor and H-acceptor was selected as a scaffold, and the pyridyl sulfonamide pharmacophore was introduced at C5position to interact with the back pocket (DFG-motif, gate keeper and catalytic lysine). These 7-azaindole derivatives exhibited modest to good activity in cellular proliferation (PI3KαIC50=3~5200nM) and in apoptosis assays.
As their continued study, they [76] discovered Torin 2 (213, PI3K/mTOR EC 50 = 200/0.25 nM) as a potent, selective, and orally available mTOR inhibitor, utilizing a focused medicinal chemistry approach guided by cellular assays and pharmacokinetic and pharmacodynamic assays of compounds 212. The co-crystal structure revealed that the aminopyridine group formed three hydrogen bonds in the hydrophobic pocket.
Nishimura et al [77] reported the discovery of a series of substituted quinolines and quinoxalines derivatives (215-217), as potent PI3K/mTOR dual inhibitors (e.g. 203, PI3Kα Ki = 0.6 nM) with excellent pharmacokinetic properties and in vivo efficacies, using compound (214) as the lead. Initially, analogues with 6, 6-bicyclic heterocycles (quinoline, isoquinoline, quinoxaline, quinazoline, cinnoline, and naphthyridine) were designed, to replace the benzothiazole, as hinge linker binder. Then by incorporating suitable substituents at the 4-position of the quinoline or the 3-position of the quinoxaline rings, excellent cellular potencies were achieved, which indicated that the ribose pocket of the enzyme can be effectively utilized in optimizing both the potency and the physicochemical properties of PI3K inhibitors [77] .
In 2013, Li et al [78] synthesized HS-106 (219, PI3Kα IC 50 =11nM), by screening the above chemical library of imidazopyridine derivatives [74]. They found this compound "suppressed breast cancer cell proliferation and induced apoptosis by inducing apoptosis and suppressing angiogenesis", which could be a potential drug for breast cancer treatment.
In order to design and optimize 3-pyridine heterocyclic derivatives as PI3K/mTOR dual inhibitors, molecular docking and 3D-quantitative structure-activity relationship (3D-QSAR) studies based on the ligand alignment and receptor alignment were applied using the comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA) were carried out by Yang et al [79] . Highly accurate and predictive 3D-QSAR models for designing new PI3K/mTOR inhibitors were constructed (Skelton 220-222), which would be useful for predicting activity and guiding the ligand modification of PI3K/mTOR inhibitors.

Pyrazines and quinoxalines
Given that the pan PI3K inhibitors XL-765 (236) and XL-147 (237) in clinical trials was quinoxaline derivatives, few PI3K inhibitors based on pyrazine and quinoxaline core have been developed, while with limited potency mostly.

Azoles
Azoles (including diazole, triazole, thiazole, oxadiazole, etc.), five-member heteroaryls of pyrimidine isosteres, are also the proper scaffolds for PI3K inhibitors. Compounds in this series always achieved isoform selectivity, e.g. PI3Kβ selective inhibitor SAR260301 (332, PI3K α/β/δ/γ IC 50  Angelo et al [90] identified benzothiazole compound 265 (PI3Kα Ki = 53 nM, mTOR Ki > 25 μM) as an initial hit from HTS, and the crystal structure suggested that the ribose pocket might accommodate larger groups than pyrimidine. Extensive SAR studies, including the link atom (sulfur, oxygen and nitrogen) and the pyridine replacing pyrimidine (266-268), led to the sulfonamide 269 (PI3Kα Ki = 38 nM, mTOR Ki = 269 nM) for as an early lead, with high in vitro and in vivo clearance. Liu et al (Pfizer) [91] discovered tetra-substituted thiophenes as highly selective PI3K inhibitors, with compound 275 (PI3Kα Ki = 230 nM) as the lead, which was initially optimized by replacing the free carboxylic acid moiety with carboxylic amide (276, PI3Kα Ki =13 nM). As nitrogen atom could form H-bond binding with a water molecule in the ATP binding site, different amide bioisosteres of compound 276 were designed (277-281), of which compound 277 with a 1, 2, 4-triazole group stand out (PI3Kα Ki = 1.7 nM, mTOR Ki = 434 nM). C-4 phenyl moiety was then replaced by diverse aryl and heteroaryl groups to maximize the mTOR selectivity. Finally, very potent compounds 282 (PI3Kα Ki=0.35 nM, mTOR Ki=2470 nM) and 283 (PI3Kα Ki=0.6 nM, mTOR Ki = 1440 nM) with excellent selectivity over mTOR (up to 7000-fold) was discovered, which demonstrated good potency in vitro and in vivo, as well as the desired pharmacokinetic properties.
Compound 298 was identified as an initial hit (PI3Kγ IC 50 = 5 nM) through HTS by Oka et al [97] and the docking mode indicated that nitrogen atoms in the acetylaminothiazole formed hydrogen bonds to the hinge Val882 ,while the benzoic acid moiety interacted with Lys807 and Lys833. Thus, they optimized the central heterocycles as the replacement of thiazole and identified oxazole derivative 299 (PI3Kγ IC 50 = 12 nM) as the lead for further optimization. A novel series of 2-aminothiazole-oxazole derivatives (300) were synthesized and evaluated as PI3Kγ inhibitors (PI3Kγ IC 50~3 -346 nM), of which the trifluoroethyl and tert-butyl derivatives displayed good enzymatic and cellular activities.
Staben et al [102] (Genentech) had disclosed HTS derived thienobenzoxepin series (311) with aniline amide substituents as PI3K inhibitors, while the aniline amide was undesired and contributed to high clearance. To improve the clearance, stability and potency through interactions with the affinity pocket, they replaced he aniline amide with heterocyclic amide isosteres (312-320). Overall, simple branched alkyl substituted triazoles had better properties than those halo aryl substituted derivatives. The replacement of 'cis'N-methyl aniline amides led to compound 321 (PI3Kα/β/γ/δ IC 50 = 4.0/29/2.2/3.9 nM), a potent and selective PI3K inhibitor with high permeability, solubility and bioavailability.

Triazines
In 2010, Richard et al [108] (Wyeth research) identified a variety of potent triazine mTOR inhibitors containing the (R)-3-methyl morpholine moiety and a pyridylureidophenyl group, which demonstrated good selectivity (greater than 500-fold) over the related PI3Kα (344-346, mTOR IC 50 ~ 0.2-3.6 nM, PI3Kα IC 50 ~ 41-1894nM). SAR studies revealed that "the addition of basic amines at the 4-position of the ureidophenyl ring was welltolerated and offered the opportunity to develop inhibitors with improved physicochemical properties, while amide derivatives at the 4-position of the arylureidophenyl ring resulted in reduced selectivity over PI3Kα but enhanced cellular activity".

Others
To discover novel PI3Kγ inhibitors as anticancer agent, Taha et al [116] explored the pharmacophoric space of PI3Kγ via diverse inhibitors and used CATALYST-HYPOGEN to identify high quality binding model in 2014. Then QSAR model was assessed within training inhibitors (78 collected PI3Kγ inhibitors, scaffold 370-378) and two associated models were validated by screening for new PI3Kγ ligands. 19 NCI hits (379-397) exhibited good to moderate potencies against PI3Kγ (IC 50 =105-9157nM) in vitro, which suggested that "the combination of pharmacophoric exploration and QSAR analysis could be useful to find new and diverse PI3Kγ inhibitors".

CONCLUSION
Unregulated activation of the PI3K/Akt/mTOR pathway is a prominent feature of many human cancers and PI3K is activated or over-expressed in all major cancers. This makes PI3K as one of the most attractive anticancer targets, which may even circumvent drug resistance to current chemotherapies proved by preclinical and clinical evidences. The discovery of PI3K inhibitors brought a lot of promising compounds as drug candidates, a dozen of which have been advanced into preclinical or clinical trials for cancer treatment. Furthermore, the first approved PI3K inhibitor, Idelalisib (p110δ selective) has already been used for the treatment of various hematological malignancies. However, there are many issues remained to be addressed.
Currently, the key point for the further development of PI3K inhibitors is selectivity. Much effort has been made to the development of class I PI3K inhibitors that exhibit sufficient isoform-selectivity and cross-kinase selectivity, with the help of the elucidation of the X-ray crystal structures of PI3K isoforms and those of their complexes with diverse inhibitors. As each PI3K isoform has its own function and is correspondingly involved in various diseases, it was assumed that the isoformspecific PI3K inhibitors may obtain lower toxicity, better tolerability and safety, while the pan-PI3K inhibitors could offer enhanced therapeutic efficacy. Likewisely, the PI3K/mTOR dual inhibitors was considered to be more effective by delivering a powerful two-spot inhibition of the pathway and have the advantage of being less susceptible to PI3K drug resistance. "Will the isoformspecific inhibitors be more tolerable than pan-PI3K inhibitors", "Whether the dual inhibition of PI3K and mTOR is superior to inhibiting PI3K alone", "How to find the proper balance between the safety (only through kinase selectivity) and the therapeutic efficacy" are still questions remained to be addressed. And the answer will not be known until the completion of ongoing clinical trials.