Development of New Thiophene-Containing Triaryl Pyrazoline Derivatives as PI3Kγ Inhibitors

A series of new thiophene-containing triaryl pyrazoline derivatives, 3a–3t, were synthesized and evaluated regarding PI3K inhibition activity and anti-tumor potency based on a trial of introducing significant moieties, including pyrazoline and thiophene, and simplifying the parallel ring structures. Most of the tested compounds indicated potent PI3K inhibitory potency, with this series of compounds showing more potency for PI3Kγ than PI3Kα. The top hit 3s seemed more potent than the positive control LY294002 on inhibiting PI3Kγ (IC50 values: 0.066 μM versus 0.777 μM) and more selective from PI3Kα (Index values: 645 versus 1.74). It could be inferred that the combination of para- and meta-, as well as the modification of the electron-donating moieties, led to the improvement in potency. The anti-proliferation inhibitory activity and the enzymatic inhibition potency indicated consistent tendencies. The top hit 3s could inhibit the phosphorylation of Akt by inhibiting PI3K through the PI3K-Akt-mTOR pathway. The molecular docking simulation indicated that the binding pattern of 3s into PI3Kγ was preferable than that of PI3Kα, with more hydrogen bond, more π-involved interactions, and fewer π-sulfur interactions. The information in this work is referable for the further development of selective inhibitors for specific isoforms of PI3K.


Introduction
Cancer is a great risk for human health in modern society and is associated with several major signaling pathways such as the Mitogen-Activated Protein Kinases (MAPK) family pathway, Akt pathway and so on [1][2][3]. Along with the Akt pathway, the upstream node and phosphatidylinositol 3-kinases (PI3Ks) have actually drawn more attention due to their essential roles in various cellular procedures, including motility, differentiation, proliferation, growth and intracellular trafficking [4][5][6]. In humans, PI3Ks and phosphatase and tensin homologue (PTEN) are reported as a complementary pair in the interconversion of the second messenger phosphatidylinositol 3,4,5-triphosphate (PIP3) and its precursor phosphatidylinositol (4,5) diphosphate (PIP2) [7][8][9]. As a primary process, PI3Ks can convert PIP2 into PIP3 and mediate downstream biological events, whereas PTEN can transform PIP3 into PIP2 as a feedback course. Based on the deeper understanding of the biochemistry, the PI3K family could be classified into three classes (I, II, and III) according to the features in structures and the specificity of substrates [10][11][12]. Among them, Class I PI3Ks have been associated more often with cancer therapies in recent decades and have become a research hotspot for investigators [13][14][15]. In detail, there are four known isoforms, named PI3Kα, β, γ and δ, in Class I PI3Ks which are usually assembled from the p110 catalytic subunit and p85 regulating subunit [16][17][18][19]. PI3Kα and PI3Kβ show common expression in many sites, while PI3Kδ and PI3Kγ are merely located in epithelial cells and the central nervous and hematopoietic systems [20][21][22]. Since Class I PI3Ks has been reported to participate in oncogenesis frequently, the development of corresponding inhibitors seems quite important in seeking the potential approaches for treating cancer.
A variety of PI3K inhibitors have been identified and exploited, among which several representatives have been promoted to early clinical trials [23][24][25]. A typical example was Wortmannin, which could inhibit both PI3K and polo-like kinase 1 (PLK1) with high potency. However, its antibiotic-like structure seemed difficult to modify. Other reported representatives included BEZ235, GSK2269557, GDC-0941, GDC-0980, and PI-3065 [26][27][28][29][30]. Except for the candidates in clinical trials, there were also other distinctive backbones such as chromeno [4,3-c] pyrazol-4(2H)-one [31,32]. From the above-mentioned cases, we identified the functional groups to organize the design concept in this work. As shown in Figure 1, the pyrazoline moiety was derived from BEZ235, GSK2269557 and GDC-0941; the thiophene was selected from GDC-0941, GDC-0980 and PI-3065; while the parallel ring structures were simplified to eliminate the colorimetric interference in the previous reports [31,32]. The parts similar to vanillic acyl were also reported in PI3K inhibition studies [33].

PI3K Inhibition Assay
Initially, we directly tested the inhibition potency of the synthesized compounds on PI3K by using competitive fluorescence polarization kinase kits, as mentioned in previous investigations [31,32]. The results were organized in Table 1. According to the significance and research maturity of the isoforms as referenced, the PI3Kα and PI3Kγ isoforms were preliminarily included here. The origin ligand of the protein crystal complex LY294002 was selected as the positive control. General speaking, the majority of the tested compounds indicated potent PI3K inhibitory potency, which set the basis for discussing the structure-activity relationship (SAR).
The results inferred that this series of compounds were more potent for PI3Kγ than PI3Kα. Compared with the positive control LY294002, the top hits indicated even better selectivity towards PI3Kγ. The top three hits for PI3Kγ were 3s, 3q and 3f with corre-

PI3K Inhibition Assay
Initially, we directly tested the inhibition potency of the synthesized compounds on PI3K by using competitive fluorescence polarization kinase kits, as mentioned in previous investigations [31,32]. The results were organized in Table 1. According to the significance and research maturity of the isoforms as referenced, the PI3Kα and PI3Kγ isoforms were preliminarily included here. The origin ligand of the protein crystal complex LY294002 was selected as the positive control. General speaking, the majority of the tested compounds indicated potent PI3K inhibitory potency, which set the basis for discussing the structureactivity relationship (SAR).
The results inferred that this series of compounds were more potent for PI3Kγ than PI3Kα. Compared with the positive control LY294002, the top hits indicated even better selectivity towards PI3Kγ. The top three hits for PI3Kγ were 3s, 3q and 3f with corresponding half inhibitory concentration (IC 50 ) values of 0.066 µM, 0.430 µM and 0.570 µM, respectively. Regarding the data of the selectivity index values, LY294002 was calculated as 1.74 (0.777/0.447), while 3s, 3q and 3f were calculated as 645 (42.600/0.066), 12.2 (5.240/0.430), and 46.7 (26.600/0.570), respectively. Accordingly, preliminarily, we could regard the most potent compound 3s as a potential candidate for selectively inhibiting PI3Kγ. Subsequently, the potency for inhibiting PI3Kγ was associated with the substituent group. First, if extending along with the symmetry axis from the linking bond (para-substitute) was an acceptable strategy, the tested compounds indicated the tendency that the steric factor was essential here and the suitable length seemed between one and two benzene rings. The corresponding data suggested that 3f (4-OCH 3 ) was among the top hits (IC 50 = 0.570 µM); other one-ring compounds (3a, 3i, 3l, 3m) indicated attractive potency (IC 50 < 10 µM), except for 3h (IC 50 = 18.100 µM), while two-ring compounds (3g, 3j, 3k) demonstrated no potential for further investigation (IC 50 > 25 µM). Then, the extending deviation from the symmetry axis (para-to metato othro-) was checked. Whether introducing othro-substitute alone (3b) was beneficial was not clear (IC 50 = 4.430 µM), while merely introducing metasubstitute (3c, 3d, 3e) seemed preferable for improving the potency (IC 50 = 4.210 µM, 2.220 µM, 1.120 µM, respectively). For a single substitute, it seemed that the electrondonating substitute (3f, 3i) resulted in better effects than the electron-withdrawing (3l, 3m) on the para-position, except for 3h. It was not obvious on the meta-position. Moreover, in the cases of multi-substituent, parallel or heterocyclic ring, the situation could be abstracted into the combination of paraand meta-(3n, 3p, 3q, 3r, 3s), metaand ortho-(3o), paraand othro-(3t). It seemed that only the combination of paraand metaled to the improvement of the potency, while, in detail, within a relatively narrow space, the tiny modification of the electron-donating moieties could bring better inhibitory activity. Afterwards, regarding inhibiting PI3Kα, the tendency could also be hinted according to the data in Table 1. The tendency could be summarized more simply. Except for 3e with nitro at the meta-position, the bulky moieties showed better potency than the single-substituent ones. Among the tested compounds, 3p indicated the potential to act as a pan blocker of both PI3Kγ than PI3Kα. In previous reports, many of the potential inhibitors were pan blockers or more potent for PI3Kα [31,32]. In this work, the top hits could be comparable in the potency, which was preferable for PI3Kγ instead. For the PI3K family, both pan blockers and specific inhibitors were essential; however, PI3Kγ seemed the most significant in the tumorigenesis, therefore the selectivity towards PI3Kγ in this work was meaningful. In this section, the anti-proliferation activity of the synthesized compounds 3a-3t was evaluated. Herein, HeLa (human epithelial cervical cancer cell line), HepG2 (human hepatoellular carcinoma cell line) and L02 (human normal hepatocyte line) cells were used. According to the affection on L02 cells, most of the tested compounds were low-toxic. For example, the cytotoxicity of the top hit 3s was much lower than the control (CC 50 values: 181.000 µM vs. 51.300 µM). Basically, many of the compounds in this series could inhibit the growth of HeLa and HepG2 cells effectively. The anti-proliferation inhibitory activity and the enzymatic inhibition potency indicated the consistent tendencies; the SAR of the observed anti-proliferation activity was also similar to that of the enzymatic inhibition, especially that of PI3Kγ inhibition; thus, the cellular inhibition might be caused by the effect on PI3K. In comparison with different cell lines, this series showed better potency on HepG2 than HeLa. This result agreed with the fact that the PI3K-associated metabolism was more frequent in liver than in model cancer cells (HeLa was from human epithelial cervical cancer but now more universal). Thus, HepG2 cells were selected to conduct further investigations at the protein level.

Western Blot
For the evaluation of PI3Kγ activity, the downstream signal of PI3Kγ should be checked. Commonly, in the PI3K-Akt-mTOR pathway, the activation of PI3K mediated the phosphorylation of Akt and then triggered the following biological events. Herein, Western blot was conducted to visualize the expression levels of Akt and p-Akt (phosphorylated Akt) in HepG2 cells after incubation with various concentrations of the top hit 3s (Figure 2). GAPDH was used as the internal reference to guarantee cell status. Along with the increase in 3s concentration, the p-Akt level indicated an obvious decrease, whereas the Akt level almost remained unchanged (even slightly increased). This meant that compound 3s has the potential to cause dose-dependent blocking on the PI3K-mediated Akt phosphorylation.

The Anti-Proliferation Assay
In this section, the anti-proliferation activity of the synthesized compounds 3a-3t was evaluated. Herein, HeLa (human epithelial cervical cancer cell line), HepG2 (human hepatoellular carcinoma cell line) and L02 (human normal hepatocyte line) cells were used. According to the affection on L02 cells, most of the tested compounds were low-toxic. For example, the cytotoxicity of the top hit 3s was much lower than the control (CC50 values: 181.000 μM vs. 51.300 μM). Basically, many of the compounds in this series could inhibit the growth of HeLa and HepG2 cells effectively. The anti-proliferation inhibitory activity and the enzymatic inhibition potency indicated the consistent tendencies; the SAR of the observed anti-proliferation activity was also similar to that of the enzymatic inhibition, especially that of PI3Kγ inhibition; thus, the cellular inhibition might be caused by the effect on PI3K. In comparison with different cell lines, this series showed better potency on HepG2 than HeLa. This result agreed with the fact that the PI3K-associated metabolism was more frequent in liver than in model cancer cells (HeLa was from human epithelial cervical cancer but now more universal). Thus, HepG2 cells were selected to conduct further investigations at the protein level.

Western Blot
For the evaluation of PI3Kγ activity, the downstream signal of PI3Kγ should be checked. Commonly, in the PI3K-Akt-mTOR pathway, the activation of PI3K mediated the phosphorylation of Akt and then triggered the following biological events. Herein, Western blot was conducted to visualize the expression levels of Akt and p-Akt (phosphorylated Akt) in HepG2 cells after incubation with various concentrations of the top hit 3s (Figure 2). GAPDH was used as the internal reference to guarantee cell status. Along with the increase in 3s concentration, the p-Akt level indicated an obvious decrease, whereas the Akt level almost remained unchanged (even slightly increased). This meant that compound 3s has the potential to cause dose-dependent blocking on the PI3K-mediated Akt phosphorylation.

Molecular Docking Simulation
Since the biological assay confirmed that 3s could act as a potential candidate for the selective inhibition of PI3Kγ, the molecular docking simulation was conducted to visualize the possible binding pattern of 3s into the active site of PI3Kα (PDB code: 1E7V) and PI3Kγ (PDB code: 3APF), respectively. The binding conformations were also compared with the original ligands, LY294002 and CH5039699, respectively. The maps of the binding patterns in both 2D and 3D are depicted in Figure 3. In Figure 3A, in the binding site of PI3Kα, 3s indicated possible hydrogen bonds with the key residues Val882 and Lys883, as well as π-alkyl and π-sulfur interactions with key residues Met804 and Met953. It was notable that several important interactions relied on sulfur, which was not typical and might be not strong enough for binding. This result inferred that the selectivity for PI3Kγ from PI3Kα might be caused by the affection of thiophene moiety. In Figure 3B, compound 3s could mimic part of the structural feature of LY294002, but the stretched dimethoxyphenyl (DMP) group differed from the control, thus leading to a new possible hydrogen bond. However, for embedding into the deeper site of PI3Kα, LY294002 seemed more suitable than 3s. On the other hand, with the binding pattern of 3s into PI3Kγ, as displayed in Figure 3C, there might be possible hydrogen bonds with the key residues Val882 and Lys883, π-π interactions with Tyr867, as well as π-alkyl interactions with Ile831, Ile879, Met953 and Ile963. Compared with the binding pattern into PI3Kα, this time one more hydrogen bond and much more π-involved interactions were introduced, while the π-sulfur interactions were not involved. Both the thiophene moiety and the aryl group might participate in the interaction of hydrogen bonds. These results agreed with the selectivity for PI3Kγ from PI3Kα because the tendency of interaction was stronger (Interaction Energy: −58.1403 Kcal/mol < −46.0220 Kcal/mol). Furthermore, in Figure 3D, 3s could mimic almost the whole molecule of CH5039699, except for a slight difference between the DMP group of 3s and the parallel ring of CH5039699. This result agreed with the initial design of simplifying the parallel ring structures and might lead to further inspiration on the modification of such inhibitory candidates, because, in this work, the potency of 3s was attractive.
ize the possible binding pattern of 3s into the active site of PI3Kα (PDB code: 1E7V) and PI3Kγ (PDB code: 3APF), respectively. The binding conformations were also compared with the original ligands, LY294002 and CH5039699, respectively. The maps of the binding patterns in both 2D and 3D are depicted in Figure 3. In Figure 3A, in the binding site of PI3Kα, 3s indicated possible hydrogen bonds with the key residues Val882 and Lys883, as well as π-alkyl and π-sulfur interactions with key residues Met804 and Met953. It was notable that several important interactions relied on sulfur, which was not typical and might be not strong enough for binding. This result inferred that the selectivity for PI3Kγ from PI3Kα might be caused by the affection of thiophene moiety. In Figure 3B, compound 3s could mimic part of the structural feature of LY294002, but the stretched dimethoxyphenyl (DMP) group differed from the control, thus leading to a new possible hydrogen bond. However, for embedding into the deeper site of PI3Kα, LY294002 seemed more suitable than 3s. On the other hand, with the binding pattern of 3s into PI3Kγ, as displayed in Figure 3C, there might be possible hydrogen bonds with the key residues Val882 and Lys883, π-π interactions with Tyr867, as well as π-alkyl interactions with Ile831, Ile879, Met953 and Ile963. Compared with the binding pattern into PI3Kα, this time one more hydrogen bond and much more π-involved interactions were introduced, while the π-sulfur interactions were not involved. Both the thiophene moiety and the aryl group might participate in the interaction of hydrogen bonds. These results agreed with the selectivity for PI3Kγ from PI3Kα because the tendency of interaction was stronger (Interaction Energy: −58.1403 Kcal/mol < −46.0220 Kcal/mol). Furthermore, in Figure 3D, 3s could mimic almost the whole molecule of CH5039699, except for a slight difference between the DMP group of 3s and the parallel ring of CH5039699. This result agreed with the initial design of simplifying the parallel ring structures and might lead to further inspiration on the modification of such inhibitory candidates, because, in this work, the potency of 3s was attractive.  The π-related interactions were shown as pink (π-π or π-alkyl) and yellow (π-sulfur) dotted lines.

Materials and Apparatus
Solvents and reagents with analytical grade were used without further purification. Chromatographic purification of products was performed on silica gel (200-300 mesh). The π-related interactions were shown as pink (π-π or π-alkyl) and yellow (π-sulfur) dotted lines.

Materials and Apparatus
Solvents and reagents with analytical grade were used without further purification. Chromatographic purification of products was performed on silica gel (200-300 mesh). Melting points were determined on a micro melting point apparatus (SGW X-4B, Shanghai, China) and uncorrected. 1 H and 13 C NMR spectra were recorded on a Bruker AVANCE III HD 400M spectrometer (Zurich, Switzerland) with Tetramethyl silane (TMS) as the internal standard. Chemical shifts (δ) were reported in ppm (parts per million) with respect to TMS. HRMS (High Resolution Mass Spectrometry) analyses were carried out using an AB Sciex TripleTOF 4600 System mass spectrometer (Framingham, MA, USA) with an ESI (electrospray ionization) source.

Chemical Syntheses
Chalcones 1a-1t were prepared from 2-furaldehyde and acetophenone derivatives through aldol reaction and dehydration reaction, then compounds 2a-2t were synthesized from chalcones 1a-1t and hydrazine hydrate by addition and condensation reaction, and, finally, the target molecules (3a-3t) were obtained from acylation reaction of the compounds 2a-2t.

Synthesis of Chalcones (1a-1t)
Chalcones 1a-1t were obtained according to the procedure as described previously [35,36]. 2-furaldehyde (10 mmol) was dissolved in EtOH (15 mL) and stirred, and then to the above solution was added an acetophenone derivative (10 mmol). The resulting mixture was cooled at 0 • C and 8 mL of 5% NaOH water solution was added drop wise. After reaction for 24 h, the resulting solution was poured into ice water (50 g) and stirred. The crude product precipitated from the solution, filtrated and washed with cold water and ethanol. The pure chalcones (1a-1t) were obtained from water-ethanol by recrystallization, yield 70-98%.

Western Blot
According to the results of the anti-proliferation assay, HepG2 cells were chosen to be incubated with the compound 3s for 24 h. The cells were then trypsinized and collected. Afterwards, the extracting of the total proteins was conducted with a 1 × RIPA lysis buffer (1% NP-40, 50 mM Tris-HCl, 150 mM NaCl, 0.25% deoxycholic acid, 1 mM EDTA containing protease inhibitors PMSF, pH 7.4,) (Amresco, Solon, OH, USA). Then, the protein extract was separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), transferred onto the PVDF membrane, blotted with primary antibodies, and marked with secondary isotype-specific antibodies tagging horseradish peroxidase. Finally, the immunocomplexes were pictured under the ChemiDOC™ XRS + system (BioRad Laboratories, Hercules, CA, USA).

Protocol of Docking Simulation
The depiction of the compounds was performed in Chemdraw 14.0 software (Cambridge Soft corporation, Cambridge, MA, USA (2012)). The synthesized compounds were defined as ligands, while the proteins were defined as the receptors. The high-resolution complexes of PI3Kα (PDB Code: 1E7V) and PI3Kγ (PDB Code: 3APF) were downloaded from RSCB Protein Data Bank (http://www.rcsb.org, accessed on 20 November 2021). Referring to the previous reports, all the proteins and ligands were prepared by minimization under the CHARMM force field [41,42]. Then, the molecular docking simulation was conducted using AutoDock 4.2 (The Scripps Research Institute) software [43]. The results were visualized by Discovery Studio Visualizer 2016 (BIOVIA, San Diego, CA, USA).

Conclusions
In summary, based on the candidates in clinical trials, a series of thiophene-containing triaryl pyrazoline derivatives, 3a-3t, were synthesized and evaluated regarding PI3K inhibition activity and anti-tumor potency. In addition to introducing the significant moieties, including pyrazoline and thiophene, we simplified the parallel ring structures to eliminate the colorimetric interference in previous reports. The majority of the tested compounds indicated potent PI3K inhibitory potency. The results inferred that this series of compounds were more potent for PI3Kγ than PI3Kα. The top hit 3s seemed more potent than the positive control LY294002 on inhibiting PI3Kγ (IC 50 values: 0.066 µM versus 0.777 µM) and more selective from PI3Kα (Index values: 645 versus 1.74). The preliminary SAR discussion indicated that the combination of paraand meta-, as well as the modification of the electron-donating moieties, led to an improvement in potency. Although both pan blockers and specific inhibitors were essential, the selectivity towards PI3Kγ in this work was meaningful. The anti-proliferation inhibitory activity and the enzymatic inhibition potency indicated the consistent tendencies; thus, the cellular inhibition might be caused by the effect on PI3K. In comparison with different cell lines, this series showed better potency on HepG2 than HeLa. This result agreed with the fact that the PI3K-associated metabolism was more frequent in liver than in model cancer cells. The top hit could inhibit the phosphorylation of Akt by inhibiting PI3K through the PI3K-Akt-mTOR pathway. In the molecular docking simulation, compared with the binding pattern into PI3Kα, it was indicated that more hydrogen bond and much more π-involved interactions were introduced to PI3Kγ, while the π-sulfur interactions were not involved. Both the thiophene moiety and the aryl group might participate in the interaction of hydrogen bonds. These results agreed with the selectivity for PI3Kγ from PI3Kα. The information in this work is referable for the further development of selective inhibitors for specific isoforms of PI3K.