The reactions were monitored by thin-layer chromatography (TLC) on precoated silica GF₂₅₄ plates. High-resolution mass spectra (HRMS) were taken in the ESI mode using a Waters Q-TOF system. Mass spectra (MSs) were taken in the ESI mode using the Agilent 1100 LC-MS system (Agilent, Palo Alto, CA, United States). ¹H-NMR and ¹³C-NMR spectra were generated using Bruker AM-400 and 500 spectrometers (Bruker Bioscience, United States) with TMS as the internal standard. The purity of all targeted compounds was confirmed to be >95%, as determined by HPLC using a Waters HPLC instrument with a UV/visible detector by monitoring at 254 nm (column: 5 μm, 4.6 × 250 mm Hypersil ODS-2 column). The mobile phase of the potassium dihydrogen phosphate solution (pH 3.0) and methanol (v/v, 3:7) was used with a flow rate of 1.0 mL/min, the ratio was 3:7, and the injection volume was 10 μL. All other chemicals were analytical grade and used without further purification.

A series of target compounds and the relative intermediates were synthesized, and the synthetic routes are shown in Scheme 1 and Scheme 2. The structures of the compounds were confirmed by ¹H-NMR, ¹³C-NMR, and HRMS. Here, 2,4,6-trichloropyrimidine or 2,4,6-trichloro-1,3,5-triazine was allowed to react with the R¹ group through an S_N2 reaction to prepare intermediates 10a–10c (Borsari et al., 2019), followed by reacting with the R² group to yield 11a–11e through an S_N2 reaction (Miller et al., 2013). The products 12a–12f were prepared from 11a to 11e and different boric acid esters through a coupling reaction (Wu et al., 2022). The method is shown in Scheme 1.

First, 2,4,6-trichloropyrimidine-5-carbaldehyde was let to react with hydroxylamine hydrochloride to obtain intermediate 14 (Yang et al., 2023; Pecchi et al., 2010) and was treated with thionyl chloride at reflux to yield the key intermediate 15 (Zhang et al., 2016). The intermediate 16 was obtained by the substitution of chloride with morpholine (Li et al., 2010). Finally, 16 was coupled with substituted boric acid esters to obtain the desired compounds 17a–17p (Dugar et al., 2015). The method is shown in Scheme 2.

The in vitro inhibition assays of all final compounds against PI3K were performed by Shanghai ChemPartner Co., Ltd. (China). The antiproliferative activities of the compounds against A2780, U87MG, MCF7, and DU145 cell lines were tested by the standard MTT method. CHO cells stably expressing the transcript of hERG were investigated by the automated whole-cell patch clamp technique, using the QPatch system (Sophion). The compound was tested with Salmonella typhimurium strains TA98 and TA100 with the plate incorporation method at seven doses. Apoptosis was evaluated using the One-Step TUNEL Apoptosis Assay Kit (FITC) (cat. no. MA0223; Meilunbio, Dalian, China), according to the manufacturer’s protocol. Male and female SD rats (SLRC Laboratory Animal Inc., Shanghai, China) were used, and the detailed pharmacokinetic procedures are described in Supplementary Material.

First, the targeted compounds were assayed for their inhibitory activity against PI3Kα, and compounds with good inhibitory activity were further evaluated for their antiproliferative activity against A2780, U87MG, MCF7, and DU145 cancer cells, overexpressing PI3K, using BKM-120 as the positive control.

As shown in Table 1, the inhibitory activities of compounds 12a–12f were poor, and Table 2 shows that some compounds (17e, 17m, 17o, and 17p) had strong inhibitory activity against PI3Kα kinase (IC₅₀: 104.1–32.4 nM). In particular, 17o (IC₅₀: 34.7 ± 2.1 nM) and 17p (IC₅₀: 32.4 ± 4.1 nM) were as potent as BKM-120 (44.6 ± 3.6 nM), and 17e (IC₅₀: 88.5 ± 6.1 nM) and 17m (IC₅₀: 104.1 ± 12.5 nM) were less potent than BKM-120. These four compounds were further evaluated for their antiproliferative activity against cancer cells. As shown in Table 3, 17e effectively inhibited proliferation in all cell lines as efficiently as BKM-120, while 17o and 17p had a slightly weaker activity than BKM-120, and 17m had weaker antiproliferative activity against A2780 and MCF7 and no inhibitory effects in U87MG and DU145 cells. These in vitro results indicated that it is possible to develop novel and effective PI3K inhibitors using the 2,4-dimorpholinopyrimidine-5-carbonitrile group as the pharmacophore.

The pharmacological analysis indicated that the morpholine group played a critical role in PI3K inhibitory activities. The substitution of morpholine decreased PI3Kα inhibitory activity (12a and 12b); thus, the morpholine group was a pharmacophore and should be retained. Next, we designed different R³ groups to evaluate inhibitory activities; no activity was observed for 12c, and poor activity was observed for 12d, indicating that the corresponding substituted groups cannot bind with the protein, which may be due to steric hindrance. The triazine compound (12f) had some inhibitory activity (IC₅₀: 371.3 ± 45.4 nM), but 12e had no activity. Accordingly, the pyrimidine-5-carbonitrile group was next introduced as a new core structure.

When using a pyrazole or benzopyrazole structure as the R group, the inhibitory activity (IC₅₀) of pyrazole was better than that of benzopyrazole (17a > 17b), while the substituted pyrazole compound decreased the inhibitory activity (17c and 17d). When the R group was a substituted phenyl group, the inhibitory activity (IC₅₀) of 4-hydroxyl was comparable to that of a 4-amine (17f to 17h) and that of the meta position was better than at a para position (17e > 17f). The substitution of the meta-hydroxyl by a methoxy group decreased the activity (IC₅₀/17e > 17g), indicating that the substituted groups affected the binding affinity significantly. In addition, when the R group was a pyridine, the inhibitory activity (IC₅₀) of the compound with it at position 3 was better than at position 4 (17i > 17j). When the R group was a substituted pyridine, electron-donating groups decreased the inhibitory activity (IC₅₀), while electron-withdrawing groups did the opposite (17m > 17k > 17l). When the R group was a pyrimidine, 2-amine substitution decreased the inhibitory activity (IC₅₀), whereas electron-withdrawing groups in position 4 increased it (17p > 17o > 17n).

Table 2 shows the activity of compounds 17e, 17m, 17o, and 17p; all were better than others, suggesting their good affinities with PI3K and verifying the design concept. Figure 5 shows the SAR of the newly designed compounds.

A docking study of compounds 17e and 17p was conducted to assess the rationality of the designed strategy; we selected the co-crystal structure of BMK-120 with PI3Kα (PDB: 3ZIM) as the docking mode. In 17e, the position-2 morpholine group could form a hydrogen bond with Val851, the phenol group could form two hydrogen bonds with Tyr836 and Ash810, and the cyan group could form a hydrogen bond with Ser774 (Figure 6). Similarly, the position-2 morpholine group of 17p could form a hydrogen bond with Val851, and the amine–pyrimidine group could form two hydrogen bonds with Asp933 and Lys802 (Figure 6), which explains its potent activity. These results illustrated that the designed compounds had significant inhibitory potency and that using the 2,4-dimorpholinopyrimidine-5-carbonitrile group as the core structure could effectively inhibit PI3K kinase activity.

Compounds 17e, 17o, and 17p were further evaluated for their inhibitory activity against β, γ, and δ isoforms, using BKM-120 as a positive control. Table 4 shows that 17e and 17p could effectively inhibit PI3Kδ (IC₅₀: 55.6 ± 3.8 nM and 15.4 ± 1.9 nM, respectively), both better than the positive control (IC₅₀: 79.3 ± 11.0 nM). However, compound 17o was weaker (IC₅₀: 204.2 ± 22.2 nM). Furthermore, the inhibitory activities of the three compounds against PI3Kβ were comparable and better than that of the positive control. Regarding the inhibitory activity against PI3Kγ, 17e and 17p were worse than the positive control, and 17o was comparable. In summary, 17e and 17p could selectively inhibit PI3Kα and PI3Kδ, thus being selected for further study.

Several clinical PI3K candidates were severely intolerable to patients due to fatal hepatotoxicity and, thus, failed in clinical trials. Thus, evaluating the selectivity of compounds toward normal cells might reduce the risk of clinical failure of new compounds. The effects of 17e and 17p were assessed on MRC-5 and HL-7720 cells, using BKM-120 as a control (Table 5).

The antiproliferative activities of 17e and 17p on MRC-5 and HL-7702 were weaker than those of BKM-120 and also weaker than the inhibitory activities on A2780, which preliminarily indicated that 17e and 17p had weak toxicity toward normal cell lines MRC-5 and HL-7702, and 17p was selected as a candidate for further research.

To evaluate cardiotoxicity, it is necessary to test inhibitory activity on hERG potassium currents. Table 6 shows the IC₅₀ value of 17p, and no obvious inhibition of hERG potassium currents was observed, but PI3K activity was inhibited.

S. typhimurium TA98 and TA100 were selected to conduct the Ames experiment to investigate the genotoxicity risk of compound 17p. No obvious changes (Table 7) were observed in colony reversion mutation, so compound 17p exerted no mutagenic effect on TA98 and TA100 strains and had no potential genotoxicity risk.

The metabolic stability of 17p was evaluated in liver microsome assays, which measure clearance and half-life in rats and humans. The half-life and clearance rates of 17p in human hepatocytes were 127.9 min and 10.8 μL/min/mg, respectively, and 38.5 min and 36.0 μL/min/mg in rat hepatocytes, respectively, indicating that 17p had good in vitro liver microsome stability (Table 8).

A TUNEL assay showed the number of apoptotic ovarian cancer cells (A2780) under the indicated increasing dose of 17p treatment for 24 h (scale bar: 100 µM). The ED₅₀ value of 17p cytotoxicity was determined using a multiple logistic regression model.

As shown in Figure 7, 17p treatment induced dose-dependent cytotoxicity in A2780 cells. A logistic regression analysis showed a cytotoxicity ED₅₀ value of 5.345 µM for compound 17p.

According to the good in vitro properties of compound 17p, we performed pharmacokinetic analyses in male and female SD rats at doses of 10 (po) and 2.5 mg/kg (iv), respectively (Table 9). The pharmacokinetic properties in male and female SD rats were extremely variable. In female SD rats with 10 mg/kg through po administration, the area under the curve was 50,618.9 h/ng/mL, the half-life was 2.03 h, and the maximum concentration of 4,860.1 ng/mL was achieved in 3.33 h. The plasma clearance was 0.2 L/h/kg, and the oral bioavailability was 46.2%. Thus, compound 17p had favorable pharmacokinetic properties.