IPI-549

Targeting phosphatidylinositol 3‐kinase gamma (PI3Kγ): Discovery and development of its selective inhibitors

Jingyu Zhu1 | Kan Li1 | Li Yu2 | Yun Chen1 | Yanfei Cai1 |
Jian Jin1 | Tingjun Hou3

1School of Pharmaceutical Sciences, Jiangnan University, Wuxi, Jiangsu, China
2School of Inspection and Testing Certification, Changzhou Vocational Institute of Engineering, Changzhou, Jiangsu, China
3Hangzhou Institute of Innovative Medicine, College of Pharmaceutical Sciences, Zhejiang University, Hangzhou, Zhejiang, China

Correspondence
Jian Jin, School of Pharmaceutical Sciences, Jiangnan University, Wuxi, 214122 Jiangsu, China.
Email: [email protected]

Tingjun Hou, Hangzhou Institute of Innovative Medicine, College of Pharmaceutical Sciences, Zhejiang University, Hangzhou 310058 Zhejiang, China.
Email: [email protected]

Funding information
Fundamental Research Funds of Changzhou Vocational Institute of Engineering, Grant/Award Number: 11130300117010; Fundamental Research Funds for the Central Universities, Grant/Award Number: JUSRP51703A; National Natural Science Foundation of China, Grant/Award Numbers: 21575128, 21807049

Jingyu Zhu, Kan Li, and Li Yu are equivalent authors.
Med Res Rev. 2020;1–23. wileyonlinelibrary.com/journal/med © 2020 Wiley Periodicals LLC | 1

1 | INTRODUCTION

The phosphatidylinositol 3‐kinase (PI3K) pathway is one of the most important signaling pathways in cells
because of its essential roles in various vital functions, such as cell survival, proliferation, differentiation, metabolism, and motility.1 PI3Ks are a family of lipid kinases that catalyze the phosphorylation of the inositol ring of phosphoinositide and are secondary messengers that help to transduce signals.2 PI3Ks are divided into three classes (I, II, and III) according to their substrate preference and primary structure.3 Class I PI3Ks have attracted extensive interest over the past two decades, and several isoforms have been confirmed as pro- mising drug targets for the treatment of cancer and immune diseases.4,5 Class I PI3Ks are further subdivided into Class IA and Class IB based on their regulatory proteins and signaling pathways. Class IA PI3Ks consist of three isoforms, PI3Kα, PI3Kβ, and PI3Kδ, each of which exists as a heterodimer formed by a p110 catalytic
subunit and an SH2‐domain‐containing regulatory subunit (p85). Class IA PI3Ks are activated by receptor
tyrosine kinases (RTKs) and other tyrosine kinase‐coupled receptors. Class IB has only one member, PI3Kγ, which is composed of a catalytic subunit (p110γ) and a p84/p87 or p101 regulatory subunit.4,6 PI3Kγ signaling
occurs through G protein‐coupled receptors (GPCRs) via the interaction between its regulatory subunit and the β‐subunits of GPCRs.7–9
A large body of evidence has demonstrated that PI3Ks participate in a wide variety of important cellular processes, such as cell metabolism, proliferation, survival, motility, and so forth.10–12 Biological inhibition of the
PI3K pathway has shown remarkable efficacy in treating tumors and other immune‐related diseases such as
rheumatoid arthritis and systemic lupus erythematosus.13–16 PI3K was discovered in the late 1980s, and the first
PI3K inhibitors (Wortmannin and LY294002) were described in 1994,17 followed by a wealth of PI3K inhibitors that have been identified in the past two decades. In 2014, Idelalisib (CAL‐101), a PI3Kδ‐selective inhibitor, was approved by the United States Food and Drug Administration (FDA) for the treatment of chronic lymphocytic leukemia (CLL) or non‐Hodgkin lymphoma (NHL). In 2018, Duvelisib (IPI‐145), a dual PI3Kδ/γ inhibitor, was approved by the FDA for the treatment of CLL or small lymphocytic lymphoma (SLL) after at least two prior therapies. These inhibitors and their efficacy paved the way for the development of PI3K isoform‐selective inhibitors.18
The PI3K/AKT/mTOR (mammalian target of rapamycin) pathway is one of the most important signaling pathways in tumor development. The mutation and deficiency of phosphatase and tensin homolog deleted on chromosome ten (PTEN) lead to the abnormal expression of PI3Kγ.4,19 The blockade of PI3Kγ activity has shown remarkable efficacy in treating advanced tumors such as CLL and NHL.20,21 Furthermore, PI3Kγ plays a central
role in a variety of cell functions, and its loss impairs the chemokine‐dependent migration of neutrophils and
macrophages, both in vitro and in vivo.22 Hence, targeting PI3Kγ would be a useful strategy for treating autoimmune and inflammatory diseases.16 Because of its major role in leukocyte recruitment, PI3Kγ is also a promising therapeutic target toward obesity, insulin resistance, and Type 2 diabetes.23 Unlike the ubiquitously expressed PI3Kα and PI3Kβ, PI3Kγ is preferentially expressed in the hematopoietic system, specifically in
leukocytes.16 The restricted expression pattern of PI3Kγ can alleviate the risk of undesirable side effects when
inhibiting PI3Kγ, which has motivated the development of PI3Kγ‐specific inhibitors. However, the development of PI3Kγ‐selective inhibitors is quite challenging because of the high structural and sequence homology with the other PI3K isoforms.24 A number of PI3Kγ‐selective inhibitors have been discovered; however, only a few of
them have advanced to clinical trials. In this review, we outline the unique structural biology of PI3Kγ, especially focusing on the structural features of the PI3Kγ binding site that are critical for γ‐isoform selectivity.
Furthermore, we summarize the status of the representative PI3Kγ‐selective inhibitors in clinical trials. Finally,
we discuss recent advances in the development of PI3Kγ‐selective inhibitors through experimental and com- putational approaches.

2 | STRUCTURAL BIOLOGY OF PI3KΓ

PI3Kγ is a heterodimeric enzyme that contains a p110γ catalytic subunit and a p101 or p84/p87 PIKAP regulatory
subunit. The crystal structure of the p110γ catalytic subunit (residues 144‐1102) was reported in 1999,25,26 and it consists of four domains: a Ras‐binding domain (RBD), a C2 domain, a helical domain, and a catalytic or kinase domain (Figure 1A). The RBD of p110γ (residues 220–311) is comprised of five‐stranded, mixed β‐sheets flanked by two α‐helices, a fold that is similar to that of the RBDs of the other two representative Ras effectors, rapidly
accelerated fibrosarcoma (Raf) and Ral guanine nucleotide dissociation stimulator (RalGDS; Table 1).27,28 As the
RBD is adjacent to the catalytic domain, it is postulated that Ras may activate p110γ through a fuzzy allosteric mechanism.26,29,30 The C2 domain of p110γ (residues 357–522) is an eight‐stranded antiparallel β‐sandwich
formed by two four‐stranded β‐sheets (Table 1), and it is characterized by the same fold of the Type II C2 domain in

FIGU RE 1 (A) Overall structure and the domain organization of PI3Kγ with the four domains: RBD colored in red, C2 domain colored in green, helical domain colored in orange, catalytic domain with N‐lobe colored in marine, and C‐lobe colored in magnet. (B) Interaction between ATP and PI3Kγ in the binding site, the backbones of the
residues are colored according to Figure 1A, and the hydrogen bonds are colored in light green. (C) The binding mode of ATP and the general regions within the active pocket of PI3Kγ. PI3Kγ, phosphatidylinositol 3‐kinaseγ; RBD, Ras‐binding domain [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 1 Definition of the important regions in PI3Kγ
Catalytic subunit (p110γ) 114–1102
C2 domain 357–522
Catalytic domain 726–1092
C‐terminal lobe 883–1092
P‐loop 803–811
Gatekeeper 879
DFG motif 964–966

phospholipase C‐delta‐1.31 The C2 domain is supposed to be directly involved in membrane binding. The helical domain of p110γ (residues 545–725) consists of five A/B pairs of anti‐parallel helices (Table 1). Although it has a fold similar to that of the HEAT‐repeat‐containing proteins involved in protein‐protein interactions (PPIs), the
specific function of the helical domain has not been fully elucidated. The catalytic domain of p110γ (residues
726–1092) contains a smaller N‐terminal lobe (residues 726–883) and a larger C‐terminal lobe (residues 884–1092; Table 1). The N‐ and C‐ terminal lobes of the p110γ catalytic domain are linked by a loop, which forms the deepest boundary of the adenosine triphosphate (ATP)‐binding pocket (Figure 1A). The structure of this domain, especially the ATP‐binding pocket, is quite similar in different kinases,32 which poses a great challenge to the design of isoform‐selective inhibitors. The structural details of this binding pocket are summarized below.

2.1 | ATP‐binding pocket

The X‐ray crystal structure (PDB ID: 1E8X)33 shown in Figure 1B illustrates the typical binding mode of ATP in the active site of the catalytic subunit of PI3Kγ.33 The adenine ring forms two hydrogen bonds with two residues (Glu880 and Val882) in the hinge region between the N‐ and C‐lobes of the catalytic domain (Figures 1B,C). The adenine moiety is sandwiched by several hydrophobic residues (Ile831, Ile879, Phe961, and Ile963) at the roof and base of the ATP‐binding site.33 The phosphate tail of ATP interacts with the P‐loop (residues 803–811; Figure 1B and Table 1). The ribose ring points toward the hydrophobic region II and does not form any specific interaction
with the protein (Figures 1B,C). The α‐, β‐, and γ‐phosphates of ATP form three hydrogen bonds with Lys833, Ser806, and Asn951 (Figure 1B).33 Reports of high‐resolution crystallographic structures of PI3Kγ bound with specific inhibitors, such as PI3Kγ/AS‐604850 (PDB ID: 2A4Z),14 PI3Kγ/compound 20 (PDB ID: 4PS3),34 and PI3Kγ/ thiazolopiperidine 8 (PDB ID: 4XZ4),35 provide details of affinity‐related and selectivity‐decided sites. There are four regions within the ATP‐binding pocket that have been defined based on their importance for inhibitor binding:
the hinge region, specificity pocket, affinity pocket, and hydrophobic region II (Figures 1C and 2A).36

FIGU RE 2 (A) The active site of phosphatidylinositol 3‐kinaseγ (PI3Kγ) in complex with the inhibitor, PIK‐39. The hinge region and hydrophobic region II are shown in orange, and different regions of the binding site are labeled. (B) The structural alignment of the specificity pockets in PIK‐39/PI3Kγ and apo‐PI3Kδ. The surface of p110γ is colored in gray and that of p110δ in cyan. The key residues γMet804 (δMet752) and the protuberant part
in the specificity pocket are highlighted in red. (C) Sequence alignment of the hinge regions and hydrophobic regions II for the four PI3K isoforms [Color figure can be viewed at wileyonlinelibrary.com]

2.1.1 | Hinge region

Most small molecule inhibitors of PI3K are designed to anchor the hinge region via the hydrogen bonds with the backbone amides of the hinge region by mimicking the adenine ring of ATP, which is also called the “hinge‐binder” motif. For PI3Kγ, the hinge region consists of the residues from Ile879 to Thr887 (Figure 2C and Table 1).37 Most
residues of PI3Kγ in this region are highly conserved, such as Ile879, Glu880, Val882, and Thr887. However, some adjacent residues are not conserved, such as Lys883 and Asp884 (Figure 2C). Our previous study has demon- strated that this region plays a significant role in determining the binding affinity between PI3Ks and their
inhibitors.38,39 It may provide opportunities for de novo design or modification of some specific “hinge‐binder”
moieties to improve PI3Kγ‐isoform selectivity.

2.1.2 | Specificity pocket

The specificity pocket was first reported in the binding of PIK‐39 (Figure 3A),40 which is a PI3Kδ‐selective inhibitor with mid‐nanomolar inhibitory potency against PI3Kδ. PIK‐39 exhibits more than 100 times selectivity for PI3Kδ over PI3Kβ and PI3Kγ, and no inhibition of PI3Kα at concentrations up to 100 μM (Table 2).40 The binding of PIK‐ 39 to p110γ induces an outward movement of a conserved methionine residue (Met752 in p110δ or Met804 in
p110γ) in the P‐loop to create a hydrophobic binding pocket in the active site (PDB ID: 2CHW, Figure 2A).40 However, this conformational change does not happen in the apo‐structure of p110δ (PDB ID: 2WXF, Figure 2B).36 Thus, this pocket is known as the “specificity pocket”.41 The specificity pocket has been observed in the crystal
structures of all Class I PI3K isoforms; however, the same inhibitor may form different binding patterns with this

FIGU RE 3 Two‐dimensional‐structures of representative phosphatidylinositol 3‐kinase inhibitors

pocket in different PI3K isoforms.42–45 For the binding of PIK‐39, the opening of the specificity pocket in p110γ requires a more extensive conformational change compared with that in p110δ, whose conformational movement is limited to a local change in the P‐loop.36,40 For PI3Kγ, the conformational change can be propagated through the
P‐loop to the helices of the N‐lobe.36,40 The PI3K inhibitors that can bind to the specificity pockets adopt a
“propeller” conformation, and those that cannot access the specificity pockets possess a relatively flat con- formation.46 Originally, the opening of the specificity pocket was considered one of the most important structural
determinants of PI3Kδ specificity, because it was only discovered in the case of PI3Kδ‐selective inhibitors.40

TABLE 2 Class I phosphatidylinositol 3‐kinase (PI3K) selectivity profile for the mentioned inhibitors (μM)
Inhibitors PI3Kγ PI3Kα PI3Kβ PI3Kδ Reference
PIK‐39 17.000 >20 11.000 17.000 Knight et al.40

IC87114 29.00 >100 75.00 0.500 Sadhu et al.48

PIK‐293 10 >90 >90 0.237 Williams et al.49

SW13 0.033 1.240 0.221 0.0007 Williams et al.49

SW14 0.021 8.91 0.697 0.009 Williams et al.49

CAL‐101 2.100 8.600 4.000 0.019 Lannutti et al.50 and Castillo et al.51

IPI‐145 0.027 1.602 0.085 0.003 Blair52

IPI‐549 0.016 3.200 3.500 >8.4 Evans et al.20

AS‐604850 0.250 4.500 ≥20 ≥20 Camps et al.14

AS‐605240 0.008 0.060 0.270 0.300 Camps et al.14

AS‐252424 0.033 0.935 ≥20 ≥20 Pomel et al.53

PIK‐C98 0.740 0.590 1.640 3.650 Zhu et al.54

CZC19945 0.025 1.585 2.512 ‐ Bell et al.55

CZC24832 0.025 7.943 >10 1.259 Bell et al.55

22a 0.002 0.132 0.042 0.122 Collier et al.34

However, X‐ray crystal structures of PI3Kγ‐inhibitor complexes reveal that most PI3Kγ‐selective inhibitors also adopt a propeller‐shaped conformation.20,45,47 Moreover, the residues directly involved in the opening of the specificity pockets are conserved in different isoforms. Therefore, the particular residues in the specificity pockets
or that are involved in the opening of this pocket are not the only factors determining PI3Kγ selectivity; a more complicated interaction network between inhibitors and residues within the ATP‐binding pocket may need to be taken into consideration when designing PI3Kγ‐selective inhibitors.

2.1.3 | Affinity pocket

The affinity pocket is another name of the hydrophobic Region I. It is a large hydrophobic pocket, which is controlled by a specific residue called the “gatekeeper” residue (Ile879 in PI3Kγ, Figure 1C).56,57 The affinity pocket has been meticulously described in the studies of the crystal structures of PI3Kα40,58 and the structure–activity relationship (SAR) analyses of PI3Kδ inhibitors.36,59 The affinity pocket is conserved in all PI3K isoforms. For PI3Kγ, the affinity pocket is surrounded by Lys833, Asp841, Tyr867, Ile963, and Asp964, which make favorable con- tributions to the binding of inhibitors (Figures 1C and 2A). Interestingly, although the affinity pocket is formed by conserved residues, many potent inhibitors that occupy the affinity pocket show excellent selectivity. For example,
the first PI3Kδ‐selective inhibitor, IC87114 (Figure 3B), reported by Sadhu et al.,48 selectively inhibits PI3Kδ with
an IC50 of 0.5 μM, which represents higher selectivity toward PI3Kδ of approximately 58‐fold than toward PI3Kγ and more than 100‐fold than toward PI3Kα, PI3Kβ, and a panel of protein kinases (Table 2). However, the binding between IC87114 and PI3Kδ reported by Berndt et al.36 reveals that the purine group of IC87114 does not fully
occupy the affinity pocket (PDB ID: 2X38). Subsequently, several phenol‐containing analogs of IC87114 were synthesized (PIK‐293, Figure 3C) and certain analogs, such as SW13 and SW14 (Figure 3D,E), showed high potency toward PI3Kδ.49 SW13 has the best activity profile, with IC50 values of 0.7, 1240, 221, and 33 nM toward PI3Kδ,

PI3Kα, PI3Kβ, and PI3Kγ (Table 2), respectively.49 The crystal structure of p110δ‐SW13 (PDB ID: 2WXG)36 shows that the modified group would fill the affinity pocket, leading to improvement in activity and selectivity.36,49 The high PI3Kδ selectivity of SW13 may be modulated by the hydrogen bonding networks formed by the residues
within the affinity pocket and some nonconserved residues surrounding the pocket.60 This improvement in se- lectivity due to the binding to the affinity pocket was also observed during the development of PI3Kγ‐selective inhibitors. The AS series, AS‐604850 (PDB ID: 2A4Z)14 reported by Camps and Pomel (Figure 3F,G), was the first group of PI3Kγ‐selective inhibitors with remarkable selectivity over PI3Kβ and PI3Kδ and moderate selectivity
over PI3Kα.14,53 The analysis of PI3Kγ‐inhibitor interactions shows that the rhodanine moiety in these inhibitors formed strong interactions with residues in the affinity pocket of PI3Kγ, especially Lys833, Tyr867, Ile963, and Asp964,14,53,60 which considerably improved PI3Kγ selectivity. In summary, accumulated evidence suggests
that although the affinity pocket is conserved in all PI3K isoforms, it still affects PI3Kγ‐selectivity through its conformational flexibility.

2.1.4 | Hydrophobic Region II

The hydrophobic Region II is also known as the ribose‐binding pocket, which is a region in the C‐terminal of the hinge region that consists of eight residues (Lys883‐Lys890 for PI3Kγ; Figures 1C and 2A; Table 1). Although the residues, Thr886 and Lys890, always orientate toward the ATP‐binding pocket at the inlet of the catalytic site, they are not conserved (Figure 2C). Therefore, these two residues may be vital for PI3Kγ‐selective binding.61,62 The
residues at four positions, i.e., Lys883, Asp884, Thr886, and Lys890, in the region in PI3Kγ show complete variability among the four isoforms, and Ala885 in PI3Kγ is different from the corresponding residues (Ser) in
the other three isoforms (Figure 2C). The formation of favorable interactions with the non‐conserved residues in
this region is an effective strategy for the development of isoform‐selective inhibitors. In 2014, Collier et al.34 reported a hydrophobic binding cleft adjacent to the hinge region, which is formed by Ala885 and another nonconserved residue, Glu829 (PDB ID: 4PS3), with their novel 4‐aza‐isoindolinone PI3Kγ‐specific inhibitors. The representative PI3Kγ inhibitor, compound 22, whose modified alkylated imidazole group occupied the cleft,
showed outstanding selectivity over the other three isoforms.34

2.2 | Lipid‐substrate binding site

To date, the crystallographic structure of PI3Kγ binding with its lipid substrate, phosphatidylinositol (4,5) bi-
sphosphate (PIP2), has not been reported. However, Gablli et al.63 reported a crystal structure of the wild‐type PI3Kα in complex with its lipid substrate, diC4‐PIP2 (PDB ID: 4OVV). This study also modeled a structural alignment of the ATP‐bound PI3Kγ catalytic subunit (PDB ID: 1E8X)33 and the PIP2‐bound PI3Kα (PDB ID: 4OVV).63 Their model uncovered the PIP2‐binding site within PI3K isoforms and meticulously illustrated the relationship between the two substrates. As shown in Figure 4, the diC4‐PIP2 is located in a groove surrounded by the P‐loop, the activation loop of the catalytic domain, and the iSH2 domain of the regulatory subunit (Figures 1C and 4; Table 1).63 The lipid‐substrate‐binding site is bordering the ATP‐binding site, and the 3ʹ hydroxyl group of
PIP2 is orientated toward the phosphate region of ATP (Figures 1C and 4).63 Both the activation loop and the iSH2 domain contain several key residues recognizing the 4‐ and 5‐phosphate, which make great contributions to the lipid substrate specificity.63 Interestingly, this study discovered that PI3Kα binds an additional PIP2 molecule, and
its binding site is far from the ATP‐binding site.63 Further study is required to determine whether PI3Kγ has the same additional lipid‐binding site.

FIGU RE 4 The illustration of the lipid‐substrate‐binding site and its relationship with the ATP‐binding site. The backbone of the p110γ is colored according to Figure 1A: the hinge region is colored in light green, and the
regulatory subunit is colored in yellow [Color figure can be viewed at wileyonlinelibrary.com]

2.3 | “DFG” motif

The majority of the kinase inhibitors target the ATP‐binding site with the activation loop in the active form, which is characterized by the “in” position of the conserved triad, aspartate‐phenylalanine‐glycine (DFG). Kinase in- hibitors can be classified into different types according to the two conformational states of DFG, namely DFG‐in and DFG‐out.57 Type I inhibitors bind to the ATP‐binding pocket in the active DFG‐in conformation, whereas Type II inhibitors can not only bind to the ATP‐binding pocket but also occupy an adjacent hydrophobic pocket in the inactive DFG‐out conformation.64 In PI3Kγ, the DFG motif is referred to as Asp964–Phe965–Gly966, and the
activation loop extends from Gly977 to Phe991 (Table 1). Several researchers have studied the role of the DFG motif in the regulation of isoform selectivity of PI3Kγ inhibitors.37,65 In the AZ compound series reported by
Gangadhara et al.,37 AZ2 with an extra alkyl tail (Figure 3I) showed more than 1000‐fold PI3Kγ selectivity over
PI3Kα and PI3Kβ, while AZ3 without the alkyl motif was much less potent and selective (Figure 3J). Comparison of
the crystal structures of AZ3/PI3Kγ (PDB ID: 6GY0)37 and AZ2/PI3Kγ (PDB ID: 6FTN)66 suggests that the N‐alkyl tail of AZ2 induced an obvious movement of Asp787, leading to a conformational change of the “DFG” motif.37
Ultimately, a special pocket, referred to as the “alkyl tail pocket”, opened to accommodate these alkyl tail analogs. This is the first example of PI3K inhibitors achieving their selectivity by modulating the conformation of the DFG motif. 37,60

2.4 | PI3Kγ selective inhibitors in clinical trials

Although extensive structural information is available for PI3Kγ, the development of PI3Kγ‐selective inhibitors is still quite challenging. Most residues surrounding the ATP‐binding pockets of Class I PI3K isoforms are conserved, and the side chains of most residues that can distinguish PI3Kγ from the other isoforms usually rotate away from
the pocket, which greatly hinders the development of PI3Kγ‐selective inhibitors.24 Encouragingly, some potential

PI3Kγ‐selective inhibitors have been discovered through the extensive efforts made by a larger number of med- icinal chemists. Up to now, in 2020, only one PI3Kγ‐specific inhibitor, IPI‐549, has been identified and is under clinical investigation.
It is interesting to observe that the development of isoform‐specific inhibitors of PI3Ks has been revived after almost four years, and PI3K‐selective inhibition has been considered an important strategy for the treatment of many diseases. The first important milestone was the approval of Idelalisib (CAL‐101) in 2014.50,51 As the first FDA‐approved PI3K inhibitor, it opened new avenues for the development of PI3K inhibitors and proved the targetability of PI3Ks (Figure 3K and Table 2). Although CAL‐101 prolonged survival of patients with CLL, SLL, and
NHL, pneumonitis, inflammation of lung tissue and alveoli, or even death were reported in a small number of patients.67 Therefore, there is an urgent need for the development of more efficient and safer PI3K isoform‐
specific inhibitors. The second critical milestone was the FDA approval of Duvelisib (IPI‐145) in 2018. Duvelisib is a
PI3Kδ/γ dual inhibitor (Figure 3l) that was initially developed by Intellikine and further developed by Infinity Pharmaceuticals.52 On September 24, 2018, it was approved by the FDA for the treatment of relapsed CLL and SLL in combination with Rituximab or Ofatumumab.52 On the same day, Duvelisib also received an accelerated ap- proval for the treatment of relapsed or refractory follicular lymphoma in adults who have received at least two
prior systemic therapies. Compared with Ofatumumab, Duvelisib showed improved progression‐free survival and
overall response rate in those patients for whom chemo‐immunotherapy was not suitable, which may promote the development of PI3Kγ inhibitors. Recently, in May 2019, Alpelisib (BYL‐719), a PI3Kα‐specific inhibitor, was launched onto the market.68,69 All of these achievements highlight the greater clinical significance of PI3K isoform‐
specific inhibitors compared to nonspecific inhibitors.
IPI‐549 (Figure 3M), an oral PI3Kγ‐selective inhibitor, has an IC50 value of 0.29 nM against PI3Kγ, which is considerably lower than 17 nM against PI3Kα, 82 nM against PI3Kβ, and 23 nM against PI3Kδ; moreover, it possesses a favorable drug‐like profile (Table 2).20 To date, IPI‐549, developed by Infinity Pharmaceuticals, is the
only PI3Kγ‐specific inhibitor under clinical investigation. The first clinical trial was started in December 2015 in the
USA (NCT02637531), and a dose‐escalation study was performed to evaluate the safety, tolerability, pharmaco- kinetics, and pharmacodynamics of the IPI‐549 monotherapy and IPI‐549 in combination with Nivolumab in subjects with advanced solid tumors, non‐small cell lung cancer, melanoma, and squamous cell cancer of the head
and neck.20 In January 2019, a Phase II clinical trial (NCT03795610) was conducted for advanced HPV+ and HPV−
head and neck squamous cell carcinoma. In May 2019, another Phase II clinical trial of IPI‐549 (NCT03961698) was conducted for triple‐negative breast cancer or renal cell carcinoma. In addition, in June 2019, the Phase II clinical trial was started to evaluate the efficacy and safety of IPI‐549 in combination with Nivolumab in patients
with advanced urothelial carcinoma (NCT03980041). Thus, this series of clinical trials highlights the therapeutic potential of PI3Kγ‐specific inhibition.
The identification of IPI‐549 was aided by the success of Duvelisib (IPI‐145; Figure 3l). Even though the crystal
structure of PI3Kγ bound with IPI‐145 has not been reported so far, the X‐ray crystal structures of IPI‐549 in complex with the human PI3Kγ (PDB ID: 6XRL) was recently reported by Drew et al.70 in August 2020. Starting from the N‐phenyl‐8‐chloroisoquinolinone moiety in IPI‐145 as a core scaffold to mimic the adenine ring of ATP
(Figure 5A), hinge‐binding motifs were explored by introducing at least one hydrogen bond acceptor and 0‐2
hydrogen bond donors at the R1 position (Figure 5B) to maintain hydrogen bond interactions with PI3Kγ.71 This led
to the identification of a prazolo[1,5‐α]pyrimindin‐2‐amine hinge‐binding motif with almost 10‐fold PI3Kγ se- lectivity over PI3Kδ (Figure 5B).20 Compared with the co‐crystal structures of the propeller‐shaped ligand, PIK‐39 (Figure 3A), in complex with PI3Kδ (PDB ID: 2WXF)36 and PI3Kγ (PDB ID: 2CHW),40 IPI‐549 contains a different propeller‐shaped conformation in the active site of PI3Kγ. There is a twisted orientation between the quinazoli-
none core and the aminopyrazolopyrimidine amide moiety, which lead to two hydrogen bonds between the aminopyrazolopyrimidine and Val882 of PI3Kγ; in addition, the quinazolinone moiety forms π‐stacking interaction with Trp812 of PI3Kγ.70 It was postulated that the substituent at the R2 position (Figure 5C) may access a
nonconserved region within PI3Kγ, and the alkyne substituent may form unfavorable interactions with Thr750 of

FIGU RE 5 Structural optimization of IPI‐549 based on IPI‐145 [Color figure can be viewed at wileyonlinelibrary.com]

PI3Kδ (corresponding to Lys802 of PI3Kγ),20,72 which may improve PI3Kγ potency and selectivity, which led to the identification of a methyl alkyne substituent at R2.20 Although the methyl alkyne analog had more than 40‐fold selectivity for PI3Kγ over PI3Kδ, it showed poor metabolic stability in vitro. To improve the metabolic stability, less
hydrophobic groups were introduced at the R3 position (Figure 5D),73,74 and a number of analogs with smaller heterocycles at the end of the alkyne were synthesized and evaluated. Finally, the N‐methylpyrazole analog
(IPI‐549) was found to have excellent potency and selectivity for PI3Kγ over PI3Kδ in enzymatic assays and
excellent metabolic stability in vitro (Figure 5E).

2.5 | Methods to discover PI3Kγ‐selective inhibitors

As discussed above, the discovery and development of PI3Kγ‐selective inhibitors are still quite challenging, given the high degree of structural homology shared among the highly conserved ATP‐binding pockets of the PI3K
isoforms. In this section, the experimental and molecular modeling technologies that can be used to design and discover potential PI3Kγ inhibitors are discussed.

2.6 | Experimental high‐throughput screening

The development and advancement of experimental high‐throughput screening (HTS) technologies have facilitated the identification of hits targeting PI3Kγ. The screening procedure can be divided into three components. The first component is the selection of high‐quality assays with the ability to identify potential PI3Kγ inhibitors accurately and quickly; this can be considered “Bioware” (Figure 6A). The second component is the liquid handling and robotics mechanism, called “Hardware” (Figure 6B), which is indispensable in handling large‐scale chemical libraries. The third component is the processing of existing workflows and information management, which can be
regarded as “Software” to analyze the detection results and identify potential hits (Figure 6C).75
The “Bioware” refers to the bioassays which are vital for the identification of specific PI3Kγ‐isoform inhibitors,
while both “Hardware” and “Software” are the general components in traditional HTS protocol. The bioassays in
HTS targeting PI3Kγ involve the incubation of the kinase with its substrate (usually phosphoinositide), a radi- olabeled precursor ([γ‐32P]c ATP or [γ‐33P] ATP), and the tested compound.1 Rearrangement or modification of these basic components may have a profound effect on the throughput and sensibility of assays, as observed in some HTS technologies developed for detecting phosphoinositide‐modifying enzymes, such as the scintillation proximity assay (SPA) technology, which is based on the properties of weakly emitting isotopes (such as 3H and
33P).76 Nowadays, with the development of more kinase inhibitors, many screening systems are commercially available such as LanthaScreen Eu kinase binding assay from Thermo Fisher Scientific, ADP‐Glo PI3Kγ assay from Promega, HTRF kinEASE assay provided by Cisbio, and so on.
Meanwhile, High‐content screening (HCS), another reliable screening technology platform, has been designed
to define the temporal and spatial activities of genes, proteins, and other cellular constituents in the living cells.75,77,78 As mentioned earlier, PI3Kγ responds to the activation of GPCRs, including chemokine receptors or receptors for the chemoattractants, fMLP or complement 5a (C5a), leading to the generation of
phosphatidylinositol‐3,4,5‐triphosphate and consequent phosphorylation of PKB/AKT (Figure 6D).79 This

FIGU RE 6 General workflows for high‐throughput screening and high‐content screening [Color figure can be viewed at wileyonlinelibrary.com]

GPCR‐mediated assay was used not only as a primary cellular filter, but also to assess the potency of these inhibitors in cell lines (Figure 6). Compared with classical HTS, HCS has the following technical advantages. First, high‐content assays on cells are more complicated than cell‐free assays within HTS. Specifically, HCS can detect the effects of drugs on complicated molecular events, such as signal transduction pathways and cell functions, by automatically extracting the multicolor fluorescence information derived from specific fluorescence‐based re- agents introduced into cells.80 Second, optical systems have the capacity to measure spatial and temporal dynamics
of cells after treatment with drugs. However, the more sophisticated and deeper biological information obtained
from the assays requires advanced and computerized methods. Wolff et al.81 established an automated HCS assay in a 384‐well format to measure cellular PI3K activity using a model system based on Chinese Hamster Ovary (CHO) cells that were stably transfected with human insulin receptor (hIR) and an AKT1‐enhanced GFP (EGFP)
fusion construct. In this cellular context, PI3K was activated, and AKT1‐EGFP was recruited to the plasma membrane upon stimulation with insulin like growth factor‐1 (IGF‐1). The AKT1‐EGFP redistribution assay was robust, and the IGF‐1‐induced translocation of AKT1‐EGFP could be quantified using a specific image analysis
algorithm that measured the fluorescence intensity associated with the plasma membrane. Then, the IC50 of the
tested compound towards PI3K could be estimated using does‐response curves.81 As described, HCS methods with high and specific expression of PI3Kγ or PI3Kγ signaling cellular models are powerful tools to discover PI3Kγ‐ selective inhibitors.
Medicinal chemists of Serono Pharmaceutical Research Institute identified several PI3Kγ‐selective inhibitors by HTS technologies, which were the first reported PI3Kγ inhibitors based on the rhodanine or thiazolidinedione scaffold.14 They performed a low‐throughput enzyme screening with an automated lipid‐kinase assay in vitro and a
systematic, high‐content, cell‐based screening, in combination with structure‐based design (Figure 6).76 The high‐
throughput PI3Kγ lipid kinase assay was firstly employed for the screening of the thiazolidinedione analogs combined SPA technology with the capacity of neomycin (a polycationic antibiotic) to bind phosphatidylinositol
with high affinity and specificity. Incubation of neomycin‐coated SPA beads with recombinant PI3Kγ and radio-
active ATP, allows the detection of phosphorylated radioactive lipid substrates through their specific binding to neomycin (Figure 6A).76 The assay was performed in 96‐well plates (Figure 6B) and the emission generated by the
scintillating supports were then measured for the last statistical analysis (Figure 6C). And then, a cell‐based HCS
was constructed for the second round of screening (Figure 6D). In the HCS, the cellular potency, selectivity, and function of these inhibitors were investigated on the basis of C5a‐mediated AKT phosphorylation in RAW264
mouse macrophage (Figure 6E). Finally, two ATP‐competitive PI3Kγ inhibitors, AS‐604850 (Ki = 0.18 μM) and AS‐
605240 (Ki = 0.0078 μM; Table 2), were identified with impressive and selective inhibition of PI3Kγ‐mediated signaling and chemotaxis in vitro and in vivo (Figure 3F,G).14 Later on, aiming to improve PI3Kγ selectivity, a set of 59,000 compounds was subsequently selected from the same group’s in‐house compound library collection and tested with this robust screening method. Consequently, a new series of potent and selective PI3Kγ inhibitors was
discovered and it provides valuable chemical structures for the subsequent substructure analysis and in silico screening. Ultimately, AS‐252424 (Figure 3H), with an IC50 of 33 nM toward PI3Kγ and more than 30‐fold
selectivity over PI3Kα and more than 600‐fold selectivity over PI3Kβ and PI3Kδ (Table 2) was identified.53

2.7 | Computer‐aided virtual screening

As a complementary approach to experimental HTS, virtual screening (VS) is able to screen a large compound database and has become an alternative or even better choice for both academic groups and pharmaceutical
industries for lead identification.82 The VS method can be roughly divided into two categories: ligand‐based VS
(LBVS), starting from known ligands of a target, and target/structural‐based VS (TBVS/SBVS), which starts from the three‐dimensional (3D) crystal structure of a target (Figure 7).83–85

FIGU RE 7 General workflows for structure based virtual screening (SBVS) and ligand based virtual screening (LBVS) [Color figure can be viewed at wileyonlinelibrary.com]

Some success stories demonstrate the viability of VS for the discovery of PI3Kγ inhibitors. For example, a novel PI3K inhibitor, PIK‐C98, with potent preclinical activities against multiple myeloma, was identified through docking‐based VS against PI3Kγ by Zhu et al. (Figure 3N).54,86 The detailed workflow is illustrated in Figure 8.
Based on the crystal structure of PI3Kγ (PDB ID: 3APC),87 the Chembridge database with 800,000 compounds was virtually screened by Glide and filtered on the basis of Lipinski’s Rule‐of‐Five (RO5).54,88,89 Then, the top‐ranked 1,500 compounds were subjected to structural clustering, and 148 potential candidates with the highest chemical
diversity were purchased for bioassays. Six out of the 148 tested compounds showed potential PI3K inhibitory activity in the cell‐based assays. Encouragingly, PIK‐C98 could preferentially inhibit PI3Kγ with nanomolar potency compared with LY294002 (0.74 μM vs. 3.43 μM) and had no inhibitory effect on AKT or mTOR activity (Table 2).
However, C98 showed only moderate selective inhibition toward PI3Kγ, necessitating further structural optimi-
zation of this hit compound. Furthermore, PIK‐C98 could promote cell apoptosis in multiple myeloma by speci- fically inhibiting the PI3K signaling pathway. Oral administration of PIK‐C98 fully suppressed the tumor growth in two independent human myeloma xenograft models in nude mice, but did not show any overt toxicity.54,86 Thus,
the combination of SBVS and bioassays is a powerful strategy to discover PI3Kγ inhibitors.

FIGU RE 8 The structure based virtual screening (SBVS) approach for developing PIK‐C98 [Color figure can be viewed at wileyonlinelibrary.com]

Taha et al.90 reported the discovery of a set of PI3Kγ‐selective inhibitors with nanomolar bioactivities through LBVS based on pharmacophore modeling and quantitative structure–activity relationship (QSAR) models (Figure 9). Initially, a total of 78 PI3Kγ inhibitors with specific IC50 toward PI3Kγ were used to develop and validate pharmacophore models. Then, based on approximately 100 physicochemical descriptors and the fit values com-
puted by mapping all of the collected PI3Kγ inhibitors onto the pharmacophore hypotheses (Figure 9A), QSAR models were built by using the genetic function approximation (GFA) algorithm (Figure 9B). Finally, two refined pharmacophore models with exclusion volumes were used as the searching queries to screen the 3D flexible molecular database of the National Cancer Institute (NCI) with 238,819 compounds (Figure 9C). The 224 highest ranked molecules predicted by the QSAR models were assessed by Veber’s and Lipinski’s rules89,91 to yield 10 potential hits that were tested in vitro: 19 exhibited nanomolar to low micromolar potencies.90 Thus, this combination of pharmacophore modeling and QSAR models provides an effective way to identify potent PI3Kγ inhibitors with novel scaffolds.

2.8 | Assay‐based SAR

Most hit compounds discovered by HTS or VS have relatively weak binding affinity ranging from high micromolar
to millimolar levels, low target selectivity, and unfavorable pharmacokinetic profiles; hence, they require optimi- zation towards “lead‐likeness.”92 Structure–activity relationships (SAR) play a very important role in the entire process of lead discovery such as LBVS; lead optimization; and prediction of absorption, distribution, metabolism,
excretion, and toxicology (ADMET).92,93 During the last two decades, a large number of potent and specific PI3Kγ

FIGU RE 9 The ligand based virtual screening (LBVS) approach for developing the PI3Kγ‐selective inhibitors [Color figure can be viewed at wileyonlinelibrary.com]

inhibitors have been reported, and although the majority of these inhibitors have not entered clinical trials, they provide valuable information for the study of structure‐selectivity‐activity relationships (SSAR) for PI3Kγ inhibitors.60
Assay‐based SAR is commonly recognized as traditional SAR or experimental‐based SAR. The physicochemical properties and biological activity of a compound are mainly determined by its chemical structure. It is well known
that an identified hit can contain one or multiple structural core scaffolds. The interactions between these scaffolds and the corresponding targets can be clarified by molecular modeling or cocrystallography techniques. This aids the synthesis of analogs containing these core scaffolds and the subsequent assessment of their biological activities in experimental assays.92,93 Finally, new SAR will be developed based on the new analogs with higher potency, better specificity, and lower side effects, followed by further design attempts until the discovery of the final drug candidate.
Bell et al. reported a detailed SAR optimization for the two PI3Kγ hits identified by HTS.55 Before the SAR
optimization, a kinase‐focused library comprising 16,000 compounds was screened using a high‐throughput,
chemoproteomics binding assay, and two hits with the triazolopyridine core were identified (Figure 10A).55,60
The researchers hypothesized that the exocyclic NH along with one of the nitrogen atoms in the triazolo- pyridine core could form the classical bi‐dentate hydrogen bond donor‐acceptor interaction with the hinge region of PI3Kγ. The pendant phenyl ring (R1) was identified as a suitable region for the initial modification of

FIGU RE 10 Rational drug design based on the structure–activity relationships studies with a series of triazolopyridine PI3Kγ‐selective inhibitors [Color figure can be viewed at wileyonlinelibrary.com]

the molecules; thus, an array of molecules with diversely substituted phenyl groups was synthesized. The
binding affinity was evaluated by using a chemoproteomic binding assay, and the cellular activity was assessed by a PI3Kγ‐dependent, cellular C5a‐induced AKT phosphorylation assay (pAKT) and an fMLP neutrophil mi- gration assay. However, these substituted phenyl analogs failed to enhance PI3Kγ selectivity. Hence, several heterocyclic substituents at the R2 position of the core were explored (Figure 10B), resulting in the identifi-
cation of a compound containing a methylsulfonyl‐substituted pyridine with a 30‐fold increase in potency for PI3Kγ (Figure 10B). The crystal structure of this pyridine sulfone compound in PI3Kγ (PDB ID: 4AOF)94 reveals that the exocyclic NH and N‐1 of the triazolopyridine form a bi‐dentate hydrogen bond interaction with Val882 in the hinge region of PI3Kγ. The triazolopyridine core and pendant pyridine extend into the flat
hydrophobic pocket of PI3Kγ to form multiple hydrophobic contacts. Meanwhile, a hydrogen bond was ob- served between the sulfone and Lys833 in the PI3Kγ affinity pocket.55 However, as this compound exhibited poor cellular permeability in vivo, and metabolic studies showed that the loss of the acetyl group involves a
major metabolic pathway, another array of analogs (2‐amino series) was synthesized by increasing the lipo-
philicity of the R3 substituent (Figure 10C). Among them, CZC19945 (Figure 3O) showed strong PI3Kγ‐binding
affinity and high stability. Molecular modeling studies of CZC19945 suggest that a potential hydrogen bond forms between the NH of the sulfonamide and the conserved Asp964 of the DFG unit. Meanwhile, the SAR around the triazolopyridine core was explored by adding extra groups (R4) at the 5, 7, or 8 positions to identify where substituents could be tolerated to further optimize the physical properties of the molecules
(Figure 10D). CZC24832 (Figure 3P) with an 8‐fluoro group exhibited excellent bioactivity and was almost
100‐fold selective over PI3Kβ and PI3Kδ (Table 2).55

2.9 | Computational‐aided SAR

Rational drug design based on SAR analyses heavily relies on the experience of chemists. Over the past several decades, many computational approaches have been developed to improve the accuracy of SAR models. These in
silico methods could build SAR models rapidly and efficiently and visualize the models.92 With the growing number of high‐solution PI3Kγ crystal structures and the advancement of computer technologies, computer‐aided SAR has become more important in organizing, mining, and interpreting acquired data to guide further drug discovery and optimization. By integrating and summarizing experimental data of existing PI3Kγ‐selective compounds, valuable structural information can be acquired for structural modification and de novo drug design.
Compound 22a, a potent PI3Kγ‐selective inhibitor, was discovered and reported by Collier et al.34 during an assay‐based SAR study on a series of benzothiazole compounds (Figure 3Q and Table 2). The inhibitors of this series had evolved from a reported pan‐PI3K inhibitor with a phenylthiazole core.34 To uncover the SAR of these PI3Kγ‐selective inhibitors, Zhu et al.95 reported a theoretical study by utilizing multiple in silico methods including
molecular docking, molecular dynamics (MD) simulations and binding free energy calculations (Figure 11). Initially, 40 compounds were docked into the ATP‐binding pocket of PI3Kγ using three different docking protocols: rigid
receptor docking, induced‐fit docking, and quantum mechanics‐polarized ligand docking.96,97 The binding struc-
tures predicted by induced‐fit docking were analyzed by MD simulations, MM/GBSA binding free energy calcu- lations, and free energy decomposition.98,99 The SAR analysis could identify not only the residues vital for PI3Kγ
binding, but also the residues that are important for PI3Kγ isoform selectivity.95 Recently, this group reported another improved SAR model.39 A series of potent PI3Kγ‐selective inhibitors reported by Leahy et al.100 was used
to build the 3D‐QSAR models with both traditional and advanced comparative molecular fields analysis
(CoMFA).100 The steric and electrostatic contour maps were subsequently generated to visualize the SARs (Figure 11A). Meanwhile, molecular docking, MD, and free energy calculations and decompositions were employed to detect some “hot residues” that are critical for PI3Kγ selectivity (Figure 11B).39 A group of residues within the PI3Kγ active pocket that are related to inhibitor binding were identified, including Ile831, Ile879, Ile881, Val882, and Ile963.39 Lys833 and Met953 were postulated to be critical for the specific binding of the inhibitors.39 These
researchers also designed 10 new inhibitors based on the QSAR models and the “hot residues”, and one inhibitor showed lead potential (Figure 11C).39,100 Thus, computer‐aided SAR analysis has the ability to reveal the selective binding mechanisms of PI3Kγ inhibitors, and more importantly, provide critical guidance for the rational design of novel and drug‐like PI3Kγ selective inhibitors.

FIGU RE 11 The workflow of computational‐based structure–activity relationships studies with two series of PI3Kγ selective inhibitors [Color figure can be viewed at wileyonlinelibrary.com]

3 | CONCLUSION

Genetic and pharmacological experiments have confirmed the important role of PI3Kγ in cancer pathogenesis and
immune responses during the last two decades. Structural biological studies have revealed the overall framework of the ATP‐binding pocket of PI3Kγ, which greatly enhances our understanding of the isoform‐selective mechanism from a structural perspective. Some small‐molecule PI3Kγ‐selective inhibitors have been discovered for the treatment of advanced tumor as well as inflammatory and allergic diseases. Even though there is no successful
mono‐PI3Kγ‐specific drug, the progress of IPI‐549 in clinical trials indicates that this kinase is a viable and druggable target. From the perspective of drug development, HTS is still the most widely utilized technology for the discovery of PI3Kγ‐selective hits. Combining HTS with systematic SAR‐guided optimization strategies, potent and selective PI3Kγ inhibitors with satisfactory in vivo efficacy and minimal side effects can be identified effi- ciently. Along with high‐resolution PI3Kγ crystal structures, computational methods are playing important roles in
the design and discovery of PI3Kγ‐selective inhibitors. Developing highly selective inhibitors targeting the PI3Kγ
isoform within a short period has become possible by combining experimental strategies and computational techniques.

ACKNOWLEDGMENTS
The study was supported by the National Natural Science Foundation of China (21807049, 21575128), the Fundamental Research Funds for the Central Universities (JUSRP51703A), the Fundamental Research Funds of Changzhou Vocational Institute of Engineering (11130300117010), and the Postgraduate Research & Practice Innovation Program of Jiangsu Province (KYCX19_1888).

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

ORCID
Tingjun Hou http://orcid.org/0000-0001-7227-2580

REFERENCES
1. Zhu J, Hou T, Mao X. Discovery of selective phosphatidylinositol 3‐kinase inhibitors to treat hematological malignancies. Drug Discov Today. 2015;20(8):988‐994.
2. Wymann MP, Pirola L. Structure and function of phosphoinositide 3‐kinases. Biochim Biophys Acta. 1998;1436(1‐2): 127‐150.
3. Wei M, Wang X, Song Z, et al. Targeting PI3Kdelta: emerging therapy for chronic lymphocytic leukemia and beyond.
Med Res Rev. 2015;35(4):720‐752.
4. Stark AK, Sriskantharajah S, Hessel EM, Okkenhaug K. PI3K inhibitors in inflammation, autoimmunity and cancer.
Curr Opin Pharmacol. 2015;23:82‐91.
5. Garces AE, Stocks MJ. Class 1 PI3K clinical candidates and recent inhibitor design strategies: a medicinal chemistry perspective. J Med Chem. 2019;62(10):4815‐4850.
6. Zhu J, Wang M, Cao B, Hou T, Mao X. Targeting the phosphatidylinositol 3‐kinase/AKT pathway for the treatment
of multiple myeloma. Curr Med Chem. 2014;21(27):3173‐3187.
7. Krugmann S, Hawkins PT, Pryer N, Braselmann S. Characterizing the interactions between the two subunits of the p101/p110gamma phosphoinositide 3‐kinase and their role in the activation of this enzyme by G beta gamma subunits. J Biol Chem. 1999;274(24):17152‐17158.
8. Vanhaesebroeck B, Guillermet‐Guibert J, Graupera M, Bilanges B. The emerging mechanisms of isoform‐specific PI3K signalling. Nat Rev Mol Cell Bio. 2010;11(5):329‐341.
9. Hawkins PT, Anderson KE, Davidson K, Stephens LR. Signalling through class I PI3Ks in mammalian cells. Biochem Soc Trans. 2006;34:647‐662.
10. Markman B, Dienstmann R, Tabernero J. Targeting the PI3K/Akt/mTOR pathway: beyond rapalogs. Oncotarget.
2010;1(7):530‐543.

11. Ito K, Caramori G, Adcock IM. Therapeutic potential of phosphatidylinositol 3‐kinase inhibitors in inflammatory respiratory disease. J Pharmacol Exp Ther. 2007;321(1):1‐8.
12. Ali AY, Wu X, Eissa N, et al. Distinct roles for phosphoinositide 3‐kinases gamma and delta in malignant B cell migration. Leukemia. 2018;32(9):1958‐1969.
13. Maxwell MJ, Tsantikos E, Kong AM, Vanhaesebroeck B, Tarlinton DM, Hibbs ML. Attenuation of phosphoinositide 3‐kinase delta signaling restrains autoimmune disease. J Autoimmun. 2012;38(4):381‐391.
14. Camps M, Rückle T, Ji H, et al. Blockade of PI3Kgamma suppresses joint inflammation and damage in mouse models
of rheumatoid arthritis. Nat Med. 2005;11(9):936‐943.
15. Barber DF, Bartolomé A, Hernandez C, et al. PI3Kgamma inhibition blocks glomerulonephritis and extends lifespan in a mouse model of systemic lupus. Nat Med. 2005;11(9):933‐935.
16. Cushing TD, Metz DP, Whittington DA, McGee LR. PI3Kdelta and PI3Kgamma as targets for autoimmune and
inflammatory diseases. J Med Chem. 2012;55(20):8559‐8581.
17. Knight ZA. Small molecule inhibitors of the PI3‐kinase family. Curr Top Microbiol Immunol. 2010;347:263‐278.
18. Zhu J, Ke K, Xu L, Jin J. Discovery of a novel phosphoinositide 3‐kinase gamma (PI3Kγ) inhibitor against hematologic malignancies and theoretical studies on its PI3Kγ‐specific binding mechanisms. RSC Adv. 2019;9(35):20207‐20215.
19. Fruman DA, Chiu H, Hopkins BD, Bagrodia S, Cantley LC, Abraham RT. The PI3K pathway in human disease. Cell. 2017;170(4):605‐635.
20. Evans CA, Liu T, Lescarbeau A, et al. Discovery of a selective phosphoinositide‐3‐kinase (PI3K)‐gamma inhibitor
(IPI‐549) as an immuno‐oncology clinical candidate. ACS Med Chem Lett. 2016;7(9):862‐867.
21. Bheemanaboina RRY. Isoform‐selective PI3K inhibitors for various diseases. Curr Top Med Chem. 2020;20(12): 1074‐1092.
22. Wymann MP, Björklöf K, Calvez R, et al. Phosphoinositide 3‐kinase gamma: a key modulator in inflammation and allergy. Biochem Soc Trans. 2003;31(Pt 1):275‐280.
23. Wymann MP, Solinas G. Inhibition of phosphoinositide 3‐kinase γ attenuates inflammation, obesity, and cardio- vascular risk factors. Ann N Y Acad Sci. 2013;1280:44‐47.
24. Ruckle T, Schwarz MK, Rommel C. PI3Kgamma inhibition: towards an ‘aspirin of the 21st century’? Nat Rev Drug Discov. 2006;5(11):903‐918.
25. Vanhaesebroeck B, Stephens L, Hawkins P. PI3K signalling: the path to discovery and understanding. Nat Rev Mol
Cell Biol. 2012;13(3):195‐203.
26. Walker EH, Perisic O, Ried C, Stephens L, Williams RL. Structural insights into phosphoinositide 3‐kinase catalysis and signalling. Nature. 1999;402(6759):313‐320.
27. Nassar N, Horn G, Herrmann C, Scherer A, McCormick F, Wittinghofer A. The 2.2 A crystal structure of the Ras‐ binding domain of the serine/threonine kinase c‐Raf1 in complex with Rap1A and a GTP analogue. Nature. 1995; 375(6532):554‐560.
28. Huang L, Hofer F, Martin GS, Kim S‐H. Structural basis for the interaction of Ras with RaIGDS. Nat Struct Biol. 1998; 5(6):422‐426.
29. Moodie SA, Paris M, Villafranca E, Kirshmeier P, Willumsen BM, Wolfman A. Different structural requirements within the switch II region of the Ras protein for interactions with specific downstream targets. Oncogene. 1995;
11(3):447‐454.
30. Rodriguez‐Viciana P, Warne PH, Khwaja A, et al. Role of phosphoinositide 3‐OH kinase in cell transformation and control of the actin cytoskeleton by Ras. Cell. 1997;89(3):457‐467.
31. Essen LO, Perisic O, Lynch DE, Katan M, Williams RL. A ternary metal binding site in the C2 domain of phosphoinositide‐specific phospholipase C‐delta1. Biochemistry. 1997;36(10):2753‐2762.
32. Taylor SS, Radzio‐Andzelm E, Madhusudan, Cheng X, Ten Eyck L, Narayana N. Catalytic subunit of cyclic AMP‐
dependent protein kinase: structure and dynamics of the active site cleft. Pharmacol Ther. 1999;82(2‐3): 133‐141.
33. Walker EH, Pacold ME, Perisic O, et al. Structural determinants of phosphoinositide 3‐kinase inhibition by wort- mannin, LY294002, quercetin, myricetin, and staurosporine. Mol Cell. 2000;6(4):909‐919.
34. Collier PN, Martinez‐Botella G, Cornebise M, et al. Structural basis for isoform selectivity in a class of benzothiazole inhibitors of phosphoinositide 3‐kinase gamma. J Med Chem. 2015;58(1):517‐521.
35. Collier PN, Messersmith D, Le Tiran A, et al. Discovery of highly isoform selective thiazolopiperidine inhibitors of phosphoinositide 3‐kinase γ. J Med Chem. 2015;58(14):5684‐5688.
36. Berndt A, Miller S, Williams O, et al. The p110 delta structure: mechanisms for selectivity and potency of new PI(3)K
inhibitors. Nat Chem Biol. 2010;6(2):117‐124.
37. Gangadhara G, Dahl G, Bohnacker T, et al. A class of highly selective inhibitors bind to an active state of PI3Kγ. Nat Chem Biol. 2019;15(4):348‐357.

38. Zhu J, Ke K, Xu L, Jin J. Theoretical studies on the selectivity mechanisms of PI3Kdelta inhibition with marketed idelalisib and its derivatives by 3D‐QSAR, molecular docking, and molecular dynamics simulation. J Mol Model. 2019; 25(8):242.
39. Li K, Zhu J, Xu L, Jin J. Rational design of novel phosphoinositide 3‐kinase gamma (PI3Kgamma) selective inhibitors: a computational investigation integrating 3D‐QSAR, molecular docking and molecular dynamics simulation. Chem Biodivers. 2019;16(7):e1900105.
40. Knight ZA, Gonzalez B, Feldman ME, et al. A pharmacological map of the PI3‐K family defines a role for p110alpha in insulin signaling. Cell. 2006;125(4):733‐747.
41. Zhu J, Pan P, Li Y, et al. Theoretical studies on beta and delta isoform‐specific binding mechanisms of phosphoi- nositide 3‐kinase inhibitors. Mol BioSyst. 2014;10(3):454‐466.
42. Furet P, Guagnano V, Fairhurst RA, et al. Discovery of NVP‐BYL719 a potent and selective phosphatidylinositol‐3 kinase alpha inhibitor selected for clinical evaluation. Bioorg Med Chem Lett. 2013;23(13):3741‐3748.
43. Hon WC, Berndt A, Williams RL. Regulation of lipid binding underlies the activation mechanism of class IA PI3‐ kinases. Oncogene. 2012;31(32):3655‐3666.
44. Zhang X, Vadas O, Perisic O, et al. Structure of lipid kinase p110beta/p85beta elucidates an unusual SH2‐domain‐ mediated inhibitory mechanism. Mol Cell. 2011;41(5):567‐578.
45. Certal V, Carry JC, Halley F, et al. Discovery and optimization of pyrimidone indoline amide PI3Kbeta inhibitors for the treatment of phosphatase and tensin homologue (PTEN)‐deficient cancers. J Med Chem. 2014;57(3):903‐920.
46. Elmenier FM, Lasheen DS, Abouzid KAM. Phosphatidylinositol 3 kinase (PI3K) inhibitors as new weapon to combat
cancer. Eur J Med Chem. 2019;183:111718.
47. Zheng Z, Miller MS, Jennings IG, Thompson PE. Mechanisms of PI3Kbeta‐selective inhibition revealed by reciprocal mutagenesis. ACS Chem Biol. 2013;8(4):679‐683.
48. Sadhu C, Masinovsky B, Dick K, Sowell CG, Staunton DE. Essential role of phosphoinositide 3‐kinase delta in neutrophil directional movement. J Immunol. 2003;170(5):2647‐2654.
49. Williams O, Houseman BT, Kunkel EJ, et al. Discovery of dual inhibitors of the immune cell PI3Ks p110delta and p110gamma: a prototype for new anti‐inflammatory drugs. Chem Biol. 2010;17(2):123‐134.
50. Lannutti BJ, Meadows SA, Herman SEM, et al. CAL‐101, a p110delta selective phosphatidylinositol‐3‐kinase in-
hibitor for the treatment of B‐cell malignancies, inhibits PI3K signaling and cellular viability. Blood. 2011;117(2): 591‐594.
51. Castillo JJ, Furman M, Winer ES. CAL‐101: a phosphatidylinositol‐3‐kinase p110‐delta inhibitor for the treatment of lymphoid malignancies. Expert Opin Investig Drugs. 2012;21(1):15‐22.
52. Blair HA. Duvelisib: first global approval. Drugs. 2018;78(17):1847‐1853.
53. Pomel V, Klicic J, Covini D, et al. Furan‐2‐ylmethylene thiazolidinediones as novel, potent, and selective inhibitors of phosphoinositide 3‐kinase gamma. J Med Chem. 2006;49(13):3857‐3871.
54. Zhu J, Wang M, Yu Y, et al. A novel PI3K inhibitor PIK‐C98 displays potent preclinical activity against multiple myeloma. Oncotarget. 2015;6(1):185‐195.
55. Bell K, Sunose M, Ellard K, et al. SAR studies around a series of triazolopyridines as potent and selective PI3K- gamma inhibitors. Bioorg Med Chem Lett. 2012;22(16):5257‐5263.
56. Zunder ER, Knight ZA, Houseman BT, Apsel B, Shokat KM. Discovery of drug‐resistant and drug‐sensitizing
mutations in the oncogenic PI3K isoform p110 alpha. Cancer Cell. 2008;14(2):180‐192.
57. Zuccotto F, Ardini E, Casale E, Angiolini M. Through the “gatekeeper door”: exploiting the active kinase conformation. J Med Chem. 2010;53(7):2681‐2694.
58. Huang CH, Mandelker D, Schmidt‐Kittler O, et al. The structure of a human p110alpha/p85alpha complex
elucidates the effects of oncogenic PI3Kalpha mutations. Science. 2007;318(5857):1744‐1748.
59. Sutherlin DP, Baker S, Bisconte A, et al. Potent and selective inhibitors of PI3Kdelta: obtaining isoform selectivity from the affinity pocket and tryptophan shelf. Bioorg Med Chem Lett. 2012;22(13):4296‐4302.
60. Miller M S, Thompson P, Gabelli S. Structural determinants of isoform selectivity in PI3K Inhibitors. Biomolecules.
2019;9:82.
61. Frazzetto M, Suphioglu C, Zhu J, et al. Dissecting isoform selectivity of PI3K inhibitors: the role of non‐conserved residues in the catalytic pocket. Biochem J. 2008;414(3):383‐390.
62. Zheng Z, Amran SI, Thompson PE, Jennings IG. Isoform‐selective inhibition of phosphoinositide 3‐kinase: identifi-
cation of a new region of nonconserved amino acids critical for p110alpha inhibition. Mol Pharmacol. 2011;80(4): 657‐664.
63. Miller MS, Schmidt‐Kittler O, Bolduc DM, et al. Structural basis of nSH2 regulation and lipid binding in PI3Kα.
Oncotarget. 2014;5(14):5198‐5208.
64. Schindler T, Bornmann W, Pellicena P, Miller WT, Clarkson B, Kuriyan J. Structural mechanism for STI‐571 in- hibition of abelson tyrosine kinase. Science. 2000;289(5486):1938‐1942.

65. Giuliano KA, Post PL, Hahn KM, Taylor DL. Fluorescent protein biosensors: measurement of molecular dynamics in living cells. Annu Rev Biophys Biomol Struct. 1995;24:405‐434.
66. Pemberton N, Mogemark M, Arlbrandt S, et al. Discovery of highly isoform selective orally bioavailable phos-
phoinositide 3‐kinase (PI3K)‐γ inhibitors. J Med Chem. 2018;61(12):5435‐5441.
67. Coutré SE, Barrientos JC, Brown JR, et al. Management of adverse events associated with idelalisib treatment: expert panel opinion. Leuk Lymphoma. 2015;56(10):2779‐2786.
68. Ando Y, Iwasa S, Takahashi S, et al. Phase I study of alpelisib (BYL719), an α‐specific PI3K inhibitor, in Japanese
patients with advanced solid tumors. Cancer Sci. 2019;110(3):1021‐1031.
69. Mayer IA, Abramson VG, Formisano L, et al. A phase Ib study of Alpelisib (BYL719), a PI3Kalpha‐specific inhibitor, with Letrozole in ER+/HER2‐ metastatic breast cancer. Clin Cancer Res. 2017;23(1):26‐34.
70. Drew SL, Thomas‐Tran R, Beatty JW, et al. Discovery of potent and selective PI3Kγ inhibitors. J Med Chem. 2020; 63(19):11235‐11257.
71. Winkler DG, Faia KL, DiNitto JP, et al. PI3K‐delta and PI3K‐gamma inhibition by IPI‐145 abrogates immune
responses and suppresses activity in autoimmune and inflammatory disease models. Chem Biol. 2013;20(11): 1364‐1374.
72. Zhu J, Zhang H, Yu L, et al. Computational investigation of the selectivity mechanisms of PI3K delta inhibition with
marketed idelalisib: combined molecular dynamics simulation and free energy calculation. Struct Chem. 2020. https://doi.org/10.1007/s11224-020-01643-4
73. Wei M, Zhang X, Wang X, et al. SAR study of 5‐alkynyl substituted quinazolin‐4(3H)‐ones as phosphoinositide 3‐kinase delta (PI3Kdelta) inhibitors. Eur J Med Chem. 2017;125:1156‐1171.
74. Bui M, Hao X, Shin Y, et al. Synthesis and SAR study of potent and selective PI3Kdelta inhibitors. Bioorg Med Chem Lett. 2015;25(5):1104‐1109.
75. Hertzberg RP, Pope AJ. High‐throughput screening: new technology for the 21st century. Curr Opin Chem Biol. 2000;
4(4):445‐451.
76. Camps M, Chabert C, Perrin D, Martin T, Wymann MP, Rommel C Scintillation proximity assays for aminoglycoside binding molecules. 2002.
77. Giuliano KA, DeBiasio RL, Dunlay RT, et al. High‐content screening: a new approach to easing key bottlenecks in the drug discovery process. J Biomol Screen. 1997;2(4):249‐259.
78. Giuliano KA, Haskins JR, Taylor DL. Advances in high content screening for drug discovery. Assay Drug Dev Technol. 2003;1(4):565‐577.
79. Hirsch E. Central role for G protein‐coupled phosphoinositide 3‐kinase gamma in inflammation. Science. 2000;
287(5455):1049‐1053.
80. Giuliano KA, Taylor DL. Measurement and manipulation of cytoskeletal dynamics in living cells. Curr Opin Cell Biol. 1995;7(1):4‐12.
81. Wolff M, Haasen D, Merk S, et al. Automated high content screening for phosphoinositide 3 kinase inhibition using
an AKT 1 redistribution assay. Comb Chem High Throughput Screen. 2006;9(5):339‐350.
82. Wagner JR, Churas CP, Liu S, et al. Continuous evaluation of ligand protein predictions: a weekly community challenge for drug docking. Structure. 2019;27(8):1326‐1335.
83. Oprea TI, Matter H. Integrating virtual screening in lead discovery. Curr Opin Chem Biol. 2004;8(4):349‐358.
84. Ferreira LG, Dos Santos RN, Oliva G, Andricopulo AD. Molecular docking and structure‐based drug design stra- tegies. Molecules. 2015;20(7):13384‐13421.
85. Walters WP, Stahl MT, Murcko MA. Virtual screening—an overview. Drug Discov Today. 1998;3(4):160‐178.
86. Tang J, Zhu J, Yu Y, et al. A virtual screen identified C96 as a novel inhibitor of phosphatidylinositol 3‐kinase that displays potent preclinical activity against multiple myeloma in vitro and in vivo. Oncotarget. 2014;5(11):3836‐3848.
87. Ohwada J, Ebiike H, Kawada H, et al. Discovery and biological activity of a novel class I PI3K inhibitor, CH5132799.
Bioorg Med Chem Lett. 2011;21(6):1767‐1772.
88. Lipinski CA. Lead‐ and drug‐like compounds: the rule‐of‐five revolution. Drug Discov Today Technol. 2004;1(4): 337‐341.
89. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experimental and computational approaches to estimate solu- bility and permeability in drug discovery and development settings. Adv Drug Deliv Rev. 2001;46(1‐3):3‐26.
90. Taha MO, Al‐Sha’er MA, Khanfar MA, Al‐Nadaf AH. Discovery of nanomolar phosphoinositide 3‐kinase gamma
(PI3Kgamma) inhibitors using ligand‐based modeling and virtual screening followed by in vitro analysis. Eur J Med Chem. 2014;84:454‐465.
91. Veber DF, Johnson SR, Cheng HY, Smith BR, Ward KW, Kopple KD. Molecular properties that influence the oral bioavailability of drug candidates. J Med Chem. 2002;45(12):2615‐2623.
92. Guha R. On exploring structure‐activity relationships. Methods Mol Biol. 2013;993:81‐94.

93. Tong W, Welsh WJ, Shi L, Fang H, Perkins R. Structure‐activity relationship approaches and applications. Environ Toxicol Chem. 2003;22(8):1680‐1695.
94. Bergamini G, Bell K, Shimamura S, et al. A selective inhibitor reveals PI3Kγ dependence of T(H)17 cell differ- entiation. Nat Chem Biol. 2012;8(6):576‐582.
95. Zhu J, Li K, Xu L, Jin J. Insight into the selective mechanism of phosphoinositide 3‐kinase gamma with benzothiazole
and thiazolopiperidine gamma‐specific inhibitors by in silico approaches. Chem Biol Drug Des. 2019;93(5):818‐831.
96. Friesner RA, Banks JL, Murphy RB, et al. Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. J Med Chem. 47, 2004:1739‐1749. 7.
97. Cho AE, Guallar V, Berne BJ, Friesner R. Importance of accurate charges in molecular docking: Quantum
mechanical/molecular mechanical (QM/MM) approach. J Comput Chem. 2005;26(9):915‐931.
98. Gohlke H, Kiel C, Case DA. Insights into protein‐protein binding by binding free energy calculation and free energy decomposition for the Ras‐Raf and Ras‐RalGDS complexes. J Mol Biol. 2003;330(4):891‐913.
99. Genheden S, Ryde U. How to obtain statistically converged MM/GBSA results. J Comput Chem. 2010;31(4):837‐846.
100. Leahy JW, Buhr CA, Johnson HWB, et al. Discovery of a novel series of potent and orally bioavailable phosphoi- nositide 3‐kinase γ inhibitors. J Med Chem. 2012;55(11):5467‐5482.IPI-549