Lengthening the Guanidine-Aryl Linker of Phenylpyrimidinylguanidines Increases Their Potency as Inhibitors of FOXO3-Induced Gene Transcription.

Increased FOXO3 nuclear localization is involved in neuroblastoma chemoresistance and tumor angiogenesis. Accordingly, FOXO3 inhibition is a promising strategy for boosting antitumor immune responses and suppressing FOXO3-mediated therapy resistance in cancer cells. However, no FOXO3 inhibitors are currently available for clinical use. Nevertheless, we have recently identified (4-propoxy)phenylpyrimidinylguanidine as a FOXO3 inhibitor in cancer cells in the low micromolar range. Here, we report the synthesis and structure-activity relationship study of a small library of its derivatives, some of which inhibit FOXO3-induced gene transcription in cancer cells in a submicromolar range and are thus 1 order of magnitude more potent than their parent compound. By NMR and molecular docking, we showed that these compounds differ in their interactions with the DNA-binding domain of FOXO3. These results may provide a foundation for further optimizing (4-propoxy)phenylpyrimidinylguanidine and developing therapeutics for inhibiting the activity of forkhead box (FOX) transcription factors and their interactions with other binding partners.

Introduction

Forkhead box (FOX) transcription factors display high functional diversity and participate in development, proliferation, differentiation, stress resistance, apoptosis, and metabolic control processes.[1] This diverse group of transcriptional regulators shares a conserved, 110-amino-acid-long, DNA-binding domain (DBD) known as the Forkhead domain.[2,3] The O subclass of FOX transcription factors consists of four members, namely, FOXO1, FOXO3, FOXO4, and FOXO6, which are key regulators of cellular homeostasis, longevity, and stress responses.[4−7]

All FOXO proteins recognize the consensus DNA sequence 5′-TTGTTTAC-3′, and their transcriptional activity is negatively regulated by the proproliferative phosphoinositide 3-kinase (PI3K)/protein kinase B (PKB, also known as Akt) signaling pathway.[5,8] PKB/Akt phosphorylates three Ser/Thr residues and induces the nuclear exclusion of FOXO proteins in a process involving binding to the scaffolding 14-3-3 protein.[9−11] In addition to phosphorylation by PKB/Akt, FOXO proteins are further regulated through phosphorylation by other kinases, acetylation, ubiquitination, and methylation.[12] Phosphorylation by stress-induced signaling kinases, such as JNK and MST1, can override the effect of the PKB/Akt-mediated phosphorylation of FOXO3 and induce its nuclear localization and activation, thus promoting FOXO3-triggered therapy resistance in cancer cells.[13−15]

In cancer, FOXO3 stands out for its dual role. On the one hand, FOXO3 induces cell cycle arrest and apoptosis, thus functioning as a typical tumor suppressor.[16] On the other hand, FOXO3 activity can also promote tumor development and progression by inducing drug resistance,[7,17] cyclin transcription upregulation,[18] cellular detoxification,[19,20] and cancer stem cell maintenance[13] and by inhibiting apoptosis inducers such as the transcription factor p53.[14] Moreover, all FOXO proteins also regulate T cell differentiation, especially the pathway that leads to the development and function of regulatory T cells.[21,22] Therefore, pharmacological inhibition of FOXO3 transcriptional activity is considered a highly promising approach to boosting antitumor immune responses and suppressing FOXO3-mediated therapy resistance in cancer cells.

Strong chemotherapy resistance often hampers the treatment of neuroblastoma, a malignancy more prevalent among patients younger than 15 years.[23] As in other cancers, one of the factors responsible for increased neuroblastoma chemoresistance and tumor angiogenesis is FOXO3 nuclear localization.[14,15] However, no small-molecule FOXO3 inhibitors are currently available for clinical use. Nevertheless, we have recently identified three FOXO3 inhibitors, namely, carbenoxolone (a glycyrrhetinic acid derivative), repaglinide (an insulin secretagogue belonging to the meglitinide class), and (4-propoxy)phenylpyrimidinylguanidine (5ca, Figure ). These compounds inhibit the transcriptional program of FOXO3 and modulate its physiological function in the low micromolar range in cancer cells.[24,25] By NMR, we have also demonstrated that 5ca and its oxalate salt directly interact with the DNA-binding domain (DBD) of FOXO3, thereby blocking its DNA-binding surface in the α-helix H3 recognition region and in the N-terminus of β-strand S2.

Figure 1

General synthetic route to arylpyrimidylquanidines 5.

Here, we describe the synthesis and structure–activity relationship study of a small library of 5ca derivatives and show that some of these compounds inhibit FOXO3-induced gene transcription with an IC50 in the submicromolar range and thus 1 order of magnitude more potent than the parent 5ca. Our NMR measurements and molecular docking simulations showed that these compounds differ in their interactions with FOXO3-DBD and target various parts of the FOXO3 DNA-binding surface. Combined, these results may serve as a foundation for further optimizing 5ca and developing therapeutic agents for inhibiting the transcriptional activity of FOXO by interfering with target promotor binding.

Results

Synthesis of the Compounds

New bisarylguanidines for structure–activity relationship studies were synthesized by varying the pyrimidine and aromatic moieties of the original compound 5ca. First, we modified the pyrimidine ring system by introducing a more polar and electron-rich N-substituted amino substituent instead of the lipophilic methyl in position 4 on the 4,6-dimethylpyrimidine core. These compounds were synthesized by treating dicyandiamide (1) with ethyl acetoacetate (2a) under basic conditions followed by the nucleophilic addition of 4-propoxyaniline 4a (Figure and Table ).

Table 1

Properties of All Compounds Prepared in This Study

compound	molecular formula	molecular weight (g·mol–1)	Clog Pa	Clog Da(at pH 7.4)	solubility in water (mM)b
5aa	C15H19N5O2	301.35	3.03	3.01	0.1
5ba	C15H18ClN5O	319.79	3.56	3.55	20
5ca	C16H21N5O	299.38	2.87	2.84	2
5cb	C7H9N	107.16	1.45	1.45	4
5cc	C14H17N5	255.33	1.97	1.16	10
5cd	C15H19N5	269.35	2.48	1.70	7
5ce	C15H19N5	269.35	2.48	1.67	4
5cf	C15H19N5	269.35	2.48	1.64	2
5cg	C18H25N5	311.43	3.51	2.87	0.2
5ch	C14H16FN5	273.32	2.11	1.46	6
5ci	C18H19N5	305.39	2.96	2.09	0.7
5cj	C15H19N5	269.35	2.26	1.28	13
5da	C22H26N6O	390.49	4.53	4.50	0.2
5db	C23H28N6O	404.52	4.82	4.79	0.3
5dc	C19H28N6O	356.47	4.13	4.10	1.7

Clog P and Clog D values were estimated using the Chemicalize server (https://chemicalize.com/).

Solubility was estimated in 20 mM phosphate (pH 6.5) and 50 mM KCl buffer at 23 °C.

The hydroxy group of bisarylquanidine 5aa was directly converted into chlorine using a standard chlorination procedure.[26] Chlorine of 5ba was subsequently displaced with an N-nucleophile in a nucleophilic aromatic substitution reaction, yielding the desired analogues 5da–5dc in acceptable yields (35–59%). Unfortunately, these compounds showed a significantly decreased solubility in water. For this reason, we continued with the modification of the second part of 5ca, the aryl substituent.

The key intermediate, 3c, was prepared by dicyandiamide (1) condensation with acetoacetate (2) to 3c.[26] The following nucleophilic addition of aniline (4b) to 3c afforded compound 5cb in a low yield (31%). Moreover, 5cb showed insufficient solubility in water (Table ). When using, under the same conditions, the more flexible benzylamine (4c), the corresponding bisarylguanidine 5cc was obtained in a moderate yield (46%) and with a significantly increased solubility in water. Based on these results, we aimed to assess this effect using various benzylamines for nucleophilic addition.

Initially, the effect of substitution in the ortho, meta, and para positions of the benzene ring in benzylamine moiety was examined. For this purpose, the corresponding products 5cd–5cf were isolated in high yields (72–80%). The solubility of 5cd–5cf in water decreased in the ortho-substituted derivative, whereas the para-substituted derivative 5cd exhibited improved solubility, albeit slightly lower than that of 5cc. The reaction with the more sterically hindered and lipophilic tert-butyl in the para-position of benzylamine resulted in the formation of the derivate 5cg in a high yield (82%). Unsurprisingly, this change of substituent limited the solubility of 5cg in water.

Our attempts to prepare guanidines substituted with strong electron-deficient benzylamines (4l and 4m) were unsuccessful. Only with 4-fluorobenzylamine was the corresponding guanidine 5ch prepared in good yield (68%), also showing sufficient solubility in water. The following modifications, that is, extending the aromatic moiety and linker between guanidine and benzene moiety, led to 5ci and 5cj. These compounds were prepared using a standard procedure in high yields (61 and 82%, respectively). Bisarylguanidine 5cj was identified as the most water-soluble compound.

Biological Activity and Toxicity

The FOXO3 inhibition potency of the phenylpyrimidinylguanidine derivatives was first examined in a fluorescence polarization (FP) assay using the recombinant DNA-binding domain of human FOXO3 (residues 156–269, denoted as FOXO3-DBD) and FAM-labeled dsDNA containing the insulin-response element (IRE) consensus motif (Figure S1). As noted, all compounds tested at a concentration of 1 μM reduced FOXO3-DBD binding to FAM-IRE dsDNA by approximately 20%, similarly to the original 5ca, thus suggesting similar binding affinities in a micromolar range.

The ability of these compounds to inhibit FOXO3 transcriptional activity was further examined by assessing their effect on FOXO3 binding to the promoter region of decidual protein induced by progesterone 1 (DEPP1) after transfecting a DEPP1-luciferase reporter plasmid into SH-EP neuroblastoma cells stably expressing a 4-hydroxy-tamoxifen (4OHT)-regulated FOXO3(A3)ERtm transgene. DEPP1 modulates autophagy in a ROS-dependent manner and is strongly induced by FOXO3 via three functional FOXO consensus sequences in its promoter.[20,27] In parallel, the toxicity of the compounds was tested by assessing their effect on endogenous DEPP1 and actin promoters in SH-EP cells.

The initial screens performed with 50 μM compounds suggested that 5ba, 5cg, 5ci, 5da, 5db, and 5dc are more potent than 5ca in blocking FOXO3 binding to the promoter region of DEPP1 (Figure S2). Although the effect of 5cd, 5cf, and 5cj was similar to that of 5ca, compounds 5cb, 5cc, 5ce, and 5ch were less potent FOXO3 inhibitors than 5ca. Further experiments at concentrations of 25 and 12.5 μM revealed that 5cg, 5ci, 5cj, 5da, 5db, and 5dc were the most potent compounds among the set of 5ca derivatives (Figure ). However, all of these compounds exhibited significant toxicity at the concentrations tested (Figure S2). Therefore, the inhibitory assay was performed at lower concentrations, down to 0.78 μM (Figures and S3).

Figure 2

Effects of 5ca and its derivatives on target gene activation by FOXO3. SH-EP/FOXO3 cells transfected with either DEPP1-luciferase reporter plasmid (DEPP1) or β-actin-luciferase reporter (actin) (as toxicity control) were treated with 5ca derivatives 5ca, 5ci, 5cj, and 5db before adding 4-hydroxy-tamoxifen (4OHT) to activate ectopic FOXO3. The increase of firefly-luciferase activity/light emission was calculated as the percentage of dimethyl sulfoxide (DMSO)-only control. The values are expressed as the mean ± standard deviation (SD) of three independent experiments, each of which was performed in triplicate. The red line indicates the efficacy of 5ca in inhibiting FOXO3-induced relative light unit (RLU) at 12.5 μM. Significant differences between 4OHT treatment and substance + 4OHT: *p < 0.05, **p < 0.01, ***p < 0.001; significant differences between 5ca treatment and new derivatives at the same concentration: #p < 0.05, ##p < 0.01, ###p < 0.001 (Student t-test).

Compounds 5ci, 5cj, and 5db blocked FOXO3 binding to the promoter region of DEPP1 more efficiently than 5cg, 5da, and 5dc. Only compounds 5ci and 5cj significantly inhibited FOXO3-mediated transcription at the lowest concentration tested in this study (0.78 μM) without any effect on endogenous DEPP1 and actin promoters in SH-EP cells as toxicity markers. Moreover, we also attempted to determine the IC50 values of 5ci, 5cj, and 5db using a resazurin early cell death evaluation assay in cells expressing a conditional FOXO3 allele that triggers apoptotic cell death upon activation.

As shown in Figure S4, FOXO3-induced apoptosis was inhibited in a concentration-dependent manner by different compounds. The toxicity of these compounds at concentrations higher than 4 μM in 5ci, 5cj, and 5db and higher than 25 μM in 5ca prevented us from accurately determining their IC50 values. Nevertheless, the data suggest that compounds 5ci and 5db have similar IC50 values of ∼0.5 μM and that the IC50 of derivative 5cj is ∼ 1 μM and thus approximately 8–16× lower than that of 5ca (IC50 of ∼8 μM; Figure S4). Combined, these data suggest that derivatives 5ci, 5cj, and 5db are significantly more potent than 5ca in inhibiting FOXO3 transcriptional function in cancer cells.

Mapping of Interactions between FOXO3-DBD and Selected Compounds

Because no successful crystallization of apo FOXO-DBD has been reported so far (all available apo FOXO-DBD structures are solution structures), we characterized the interaction between FOXO3-DBD and compounds 5ci, 5cj, 5db, and 5dc by NMR spectroscopy. Our results showed a high potency in blocking the transcriptional function of FOXO3 and sufficient solubility in water (Figures and S3 and Table ). A similar approach was used in our previous study, wherein we characterized interactions between FOXO3-DBD and bisarylguanidine 5ca.[25]

Those interactions were first characterized by 1H saturation transfer difference (STD) NMR. However, given the low solubility of 5ci, 5db, and 5dc, an STD spectrum was successfully recorded only for 5cj (Figure a). STD signals were detected for protons from both aromatic moieties of 5cj, thus confirming its interaction with FOXO3-DBD. The binding site of 5cj in FOXO3-DBD was identified by analyzing 1H and 15N chemical shift perturbations (CSPs) of the backbone amide groups of 15N-labeled FOXO3-DBD in 1H–15N HSQC spectrum in the presence of 0.5 mM 5cj (Figures b,c and S5). The most significant CSPs (the chemical shift change was greater than 2σcorr0 above the mean) were observed mainly in residues located within the N-terminal extension (W157-S160), the helix H1 (Y162, D164, L165, T167, A169), the loop between H2 and H4 (V191), and the helix H4 (K195).

Figure 3

Interaction between FOXO3-DBD and compound 5cj. (a) 1H STD NMR experiment. The black line corresponds to the proton spectrum of 15 μM FOXO3-DBD in the presence of 1 mM 5cj. The green line corresponds to the 1H STD spectrum of 15 μM FOXO3-DBD without 5cj. The red line corresponds to the 1H STD spectrum of 15 μM FOXO3-DBD in the presence of 1 mM 5cj. (b) Distribution of CSPs observed in residues of 100 μM 15N-labeled FOXO3-DBD in the presence of 500 μM 5cj. Solid, dashed, and dotted lines correspond to the mean, mean + 1σcor0, and mean + 2σcorr0 values of CSPs, respectively. Gray bars represent unassigned residues in 1H–15N HSQC spectra. The secondary structure of FOXO3-DBD is indicated at the top. The structure of 5cj is shown in the inset, and equivalent protons visible in the 1H spectrum are numbered from 1 to 7. (c) Detailed view of selected peaks of 1H–15N HSQC spectra of 15N-labeled FOXO3-DBD in 5cj. (d) CSPs of 15N-labeled FOXO3-DBD in 5cj mapped onto the crystal structure of the FOXO3-DBD:DNA complex.[28] Residues with CSPs larger than the mean + 2σcorr0 and the mean + 1σcorr0 (from panel b) are highlighted in dark blue and light blue, respectively. The residues that could not be unambiguously assigned are highlighted in light pink. (e, f) Top-ranked HADDOCK model of the FOXO3-DBD:5cj complex; FOXO3-DBD is shown in either ribbon or surface representation. Residues located in close proximity to 5cj are shown as sticks.

Mapping of significant CSPs onto a solution structure of FOXO3-DBD[29] revealed a well-defined surface mainly involving the N-terminal tail and the N-terminus of α-helix H1, which likely forms the main region of the binding site of 5cj in FOXO3-DBD. NMR data-driven HADDOCK[30,31] docking and the solution structure of FOXO3-DBD (PDB ID: 2K86(29)) were used to predict a structural model of the FOXO3-DBD:5cj complex (Figure e,f). Residues with CSPs higher than the mean + 2σcorr0 were classified as “active”. The top-ranked resulting structures were clustered according to the root-mean-square deviation (RMSD) and ranked by HADDOCK scores.

The top-ranked pose indicated that the phenyl moiety of 5cj binds into the groove formed by residues G158, N159, and Y162 from the N-terminal segment, whereas the dimethylpyrimidinyl moiety interacts with Y193 and K195 from helix H4. These FOXO3-DBD regions either directly participate in DNA binding and/or undergo a conformational change upon FOXO3 binding to DNA, thus supporting the hypothesis that 5cj inhibits FOXO3 by blocking part of its DNA-binding surface.[28,29,32] Furthermore, the gradual shift in resonances of FOXO3-DBD residues during the titration (Figure c) suggests a fast exchange of bound compound on the NMR time scale, in line with the moderate IC50 values assessed by the resazurin early cell death evaluation assay (Figure S4).

A more scattered pattern of CSPs in residues of 15N-labeled FOXO3-DBD was observed in 0.5 mM 5ci (Figure a,b). Accordingly, this compound interacts with the FOXO3-DBD surface differently from the others. The most significant CSPs (CSPs higher than the mean + 2σcorr0) were observed for helices H2 (L180, S181, M187, V188), H3 (H212, N213), and H4 (F194, A204) residues (Figures c and S6). The top-ranked HADDOCK pose predicted that 5ci binds to the same region of FOXO3-DBD formed by the N-terminus of α-helix H1 and helices H3 and H4, albeit closer to helix H3 than 5cj. Although the naphthyl moiety of 5ci interacts with residues H212 and N213 of this helix, the dimethylpyrimidinyl moiety contacts helix H1 residues Y162 and L165 and helix H4 residue Y193 (Figure d,e). This pose, however, does not fully account for the CSP pattern, suggesting the presence of a second binding site. This CSP pattern can be explained, nevertheless, when also considering a HADDOCK pose from the second top-ranked cluster, which indicates that 5ci binds to the pocket formed by helix H2 residues L180, S181, and Y184 and H3 helix residues A204 and F194 (Figure f,g).

Figure 4

Interaction between FOXO3-DBD and compound 5ci. (a) Distribution of CSPs observed in residues of 100 μM 15N-labeled FOXO3-DBD in the presence of 500 μM 5ci. Solid, dashed, and dotted lines correspond to the mean, mean + 1σcorr0, and mean + 2σcorr0 values of CSPs, respectively. Gray bars represent unassigned residues in 1H–15N HSQC spectra. The salmon bar corresponds to a residue whose intensity was lost when adding 5ci. The secondary structure of FOXO3-DBD is indicated at the top. The structure of 5ci is shown in the inset. (b) Detailed view of selected peaks of 1H–15N HSQC spectra of 15N-labeled FOXO3-DBD in 5ci. (c) CSPs of 15N-labeled FOXO3-DBD in 5ci mapped onto the crystal structure of the FOXO3-DBD:DNA complex.[28] Residues with CSPs larger than mean + 2σcorr0 and mean + 1σcorr0 (from panel a) are highlighted in dark blue and light blue, respectively. Residues that could not be unambiguously assigned are highlighted in light pink. (d–g) HADDOCK models of the FOXO3-DBD:5ci complex from two top-ranked clusters; FOXO3-DBD is shown in either ribbon or surface representation. Residues located near 5ci are shown as sticks.

1H–15N HSQC measurements with 5db were performed at only 200 μM due to the very low solubility of this compound in water (Figure S7 and Table ). The most significant CSPs were observed for helices H2 (S181, E185), H3 (L214), and H4 (F194, D196, K197, S203) residues (Figures S7a–c and S8), as in 5ci (Figure a). HADDOCK docking suggested that the pyrimidinylguanidine moiety of 5db interacts with helix H2 residues S181 and E185, whereas the phenyl moiety binds to the pocket formed by H3 helix residues S203, A204, and K207 (Figure S7d,e). The (4-propoxy)phenyl moiety on the other end of 5db is positioned next to the side chain of S181. As such, this compound should bind to a region similar to the second binding site of 5ci (Figure f,g).

The last compound whose interactions with FOXO3-DBD were mapped was 5dc. Although this compound had a slightly lower inhibitory efficacy than 5cj, 5ci, and 5db (Figure S3), its solubility in water was higher than that of 5ci and 5db (Table ). 1H–15N HSQC measurements in the presence of 500 μM 5dc revealed the highest CSPs in residues of the N-terminal half of helix H1 and, especially, in helix H4 residues located between helices H2 and H3 (Figures S9a–c and S10). The HADDOCK pose predicted that 5dc binds to the same region as 5cj with the (4-propoxy)phenyl moiety interacting with helix H1 residue Y162, whereas its pyrimidinylguanidine moiety interacts with a surface formed by helix H4 residues Y193, K195, D196, and K197 (Figure S9d,e).

Taken together, the results from our detailed characterization indicate different interactions between FOXO3-DBD and 5ci, 5cj, 5db, and 5dc because these compounds target various parts of the FOXO DNA-binding surface formed by (1) the N-terminal tail, the N-terminus of helix H1 and helix H4 (5cj and 5dc), (2) the N-terminus of helix H1 and the C-terminus of helix H3 (5ci), (3) helices H2, H3, and H4 (5ci), and (4) the N-terminal halves of helices H2 and H3 (5db).

FOXO-DBDs Differ in Their Interactions with Phenylpyrimidinylguanidines

We investigated differences in the interactions of individual members of the FOXO family with 5ci, 5cj, 5db, and 5dc by performing 1H–15N HSQC with 15N-labeled FOXO1-DBD and FOXO4-DBD to compare CSP profiles of DBDs of all three FOXO variants (Figures and S11). 1H–15N HSQC spectra of 15N-labeled FOXO1-DBD and FOXO4-DBD with and without 5ci, 5cj, 5db, and 5dc are shown in Figures S12–S20. For all four compounds, the CSP profiles of FOXO1-DBD and FOXO3-DBD were highly similar, indicating similar interactions between them and DBDs of these FOXO proteins. In contrast, FOXO4-DBD CSP profiles differed from FOXO1-DBD and FOXO3-DBD profiles, especially in the presence of 5ci and 5db. In these two compounds, the most significant CSPs (the chemical shift change was greater than 2σcorr0 above the mean) were mainly observed in residues located in the N-terminus of FOXO4-DBD. Conversely, in FOXO1-DBD and FOXO3-DBD, the largest CSPs were observed in helices H2, H3, and H4 residues (Figures b and S7a).

Figure 5

Binding selectivity of compounds 5cj and 5ci. Distribution of CSPs observed in residues of 100 μM 15N-labeled FOXO-DBDs in the presence of 500 μM 5cj (a) and 400 μM 5ci (b). Solid, dashed, and dotted lines correspond to the mean, mean + 1σcorr0, and mean + 2σcorr0 values of CSPs, respectively. Gray bars represent unassigned residues in 1H–15N HSQC spectra. The salmon bar corresponds to a residue whose intensity was lost when adding 5ci. The secondary structure of FOXO-DBD is indicated at the top.

HADDOCK docking predicts that the naphthyl moiety of 5ci interacts with a FOXO4-DBD surface formed by W101, Q104, S105, and Y106, whereas the pyrimidinylguanidine moiety makes contacts with the residues S105, H156, N157, and H161 (Figure S21a). In turn, our docking simulation suggested that 5db is embedded between the N-terminal segment (residues N99, A100, and W101) and the loop between helices H2 and H4 (residues V135, Y137, and F138) and that the phenyl moiety also interacts with the side chain of Y105 in the N-terminus of helix H1 (Figure S21b). Taken together, these results indicate that phenylpyrimidinylguanidines may interact differently with different members of the FOXO subfamily, particularly FOXO4.

Discussion and Conclusions

Three (4-propoxy)phenylpyrimidinylguanidine (5ca, Figure ) derivatives, namely, 5ci, 5cj, and 5db, inhibited FOXO3 binding to the promoter region of DEPP1 in neuroblastoma cells considerably better than their parent compound 5ca, as shown by bioluminescence in a luciferase reporter assay (Figure ). These results were further corroborated in a resazurin early cell death evaluation assay, which showed that these compounds have significantly lower IC50 values (0.5–1 μM) than 5ca (Figure S4). The analysis of the effects of these compounds on target gene activation by FOXO3 suggested that their inhibitory potency depends on the length of the linker between the guanidine and aryl moieties (Figures , S2, and S3). Accordingly, 5cj exhibited the highest inhibitory potency, whereas compounds 5cb and 5cc, with a shorter linker between the guanidine and benzene moieties, showed a lower ability to inhibit FOXO3, presumably due to lower flexibility.

Although extending the aromatic moiety by adding a methyl group (5cd, 5ce, and 5cf) or a fluorine atom (5ch) did not increase the inhibitory potency, introducing a tert-butyl group in the para-position (5cg) and replacing benzyl by a naphthyl (5ci) group increased FOXO3 inhibition. These findings suggest that the overall size of the aromatic moiety on the (4-propoxy)phenyl side is yet another factor that affects the inhibitory potency of these compounds. In line with this inference, all three compounds with a modified dimethylpyrimidinyl moiety (5da–c) were better inhibitors than the parent 5ca. Of the three, 5db, which contains the ethylbenzyl moiety, was the strongest inhibitor, albeit more toxic than 5ci and 5cj at higher concentrations (Figure ). Therefore, combining further modifications of the dimethylpyrimidine moiety with changes in the (4-propoxy)phenyl side may open up new opportunities for improving the inhibitory efficacy of phenylpyrimidinylguanidines.

We have previously shown that the original 5ca binds to the surface formed by the C-terminal half of the DNA recognition helix H3 and the N-terminus of β-strand S2, thereby blocking the interaction between FOXO3 and the target DNA and thus inhibiting the physiological program activated by FOXO3 in cancer cells.[25] In this study, our structural characterization of the interaction between FOXO3-DBD and 5ci, 5cj, 5db, and 5dc by NMR and molecular docking revealed that these four compounds target different regions of FOXO3-DBD, but all binding sites directly participate in DNA binding and/or undergo a conformational change upon FOXO3 binding to the target DNA (Figure S22).[25,28,29]

The gradual shift in resonances of several FOXO3-DBD residues during the titrations (Figures c, 4b, S7b, and S9b) suggests a fast exchange of bound compounds on the NMR time scale, indicating weak binding affinity (in low mM range) under the conditions used. However, this contrasts with the IC50 values of 0.5–1 μM indicated by a resazurin early cell death evaluation assay (Figure S4). One factor responsible for this discrepancy may be the different conditions during NMR measurements compared to cell culture experiments, especially pH (pH 7.4 in cell cultures, pH 6.5 in NMR) and temperature (37 °C in cell cultures vs 25 °C in NMR). Furthermore, the IC50 values reflect the inhibition of FOXO3-induced cell death. To induce cell death, FOXO3 activation must reach a certain activation threshold; therefore, it is not necessary to inhibit all intracellular FOXO3 but only the fraction of active FOXO3 required to reach this threshold to trigger cell death.[15] In addition, these compounds may also inhibit FOXO3-induced transcription not only by blocking its DNA-binding surface but also through another, yet unidentified, mechanism, e.g., binding to the transactivation domain.

By comparing solution structures of individual FOXO DNA-binding domains, we identified differences in both their conformation and flexibility, specifically in the positions of helices H1, H2, and H3 and at the interface between the H2–H3 loop, the helix H3, and the N-terminal segment.[25] These differences likely reflect variations in protein sequences (Figure S23) and may explain why FOXO-DBDs, especially FOXO4-DBD, interact differently with phenylpyrimidinylguanidines (Figures and S11). In addition, the interaction between the N-terminal segment of FOXO4-DBD and the p53 transactivation domain is essential for the overall stability of the p53:FOXO4 complex.[33−35] This interaction between FOXO4 and p53 inhibits apoptosis in senescent cells, and its disruption by the D-retro-inverso peptide, which corresponds to the N-terminus of FOXO4-DBD, blocks the transcription of senescence-associated protein p21 and induces nuclear exclusion of active p53, thereby inducing death in senescent cells.[33] Our results showed that 5cj, 5ci, 5db, and 5dc binding significantly affect residues of the N-terminal segment of FOXO4-DBD (Figures and S11). Consequently, these compounds may also interfere with the interaction between FOXO4 and p53.

In conclusion, our phenylpyrimidinylguanidine derivatives inhibit gene transcription by FOXO3 with an IC50 in a submicromolar range and thus 1 order of magnitude more potent than the parent 5ca. The inhibitory potency of these compounds depends on the linker length between the guanidine group and aromatic moiety, the overall size of this aromatic moiety, and the modification of the dimethylpyrimidinyl moiety at the other side of the guanidine group. The compounds interact with FOXO3-DBD differently, suggesting the possibility to use these substances to selectively target various parts of the FOXO DNA-binding surface.

The DNA-binding surfaces of transcription factors outside the nuclear receptor family are difficult drug development targets for their lack of well-defined, small-molecule-binding pockets, high solvent exposure, and positively charged residues.[36−38] Nevertheless, FOXO3 inhibition by 5ca and other examples of successful targeting of DNA-binding surfaces, such as the inhibition of forkhead transcription factor FOXM1,[39] androgen receptor,[40] signal transducer and activator of transcription 3,[41] or transcription factor Gli1,[42] have demonstrated that DNA-binding surfaces are feasible drug targets.

The question as to whether these compounds can steer distinct target gene subsets/functions of FOXO transcription factors will undoubtedly be answered in subsequent studies. Further optimization is also required to improve the inhibitory efficacy and toxicity of these substances. Yet, targeting the FOXO3-specific transcription program and the interactions between FOXOs and their binding partners are excellent opportunities for drug development aimed at damaging cancer cells and boosting anticancer immunity. Thus, our results may serve as a foundation for further optimizing 5ca and developing therapeutic agents for inhibiting the activity of FOXO transcriptional factors and their interactions with other binding partners by identifying the requirements for enhancing the inhibitory potency of such compounds.

Materials and Methods

Synthesis of the Compounds

General

Chemicals with purity >98% were purchased from Merck KGaA (Darmstadt, Germany) and Fluorochem (Derbyshire, U.K.). Solvents of high-performance liquid chromatography (HPLC) purity were purchased from Lab-Scan (Gliwice, Poland) and Fisher Scientific (Hampton, New Hampshire) and dried by standard techniques. Thin-layer chromatography (TLC) was performed using silica gel plates 60 F254 (Merck KgaA, Darmstadt, Germany), and the compounds were visualized by irradiation with UV light and/or by treatment with a solution ninhydrin followed by heating. Column chromatography was performed on silica gel 60 (0.063–0.200 mm) (Merck KgaA, Darmstadt, Germany). Analytical HPLC analysis was carried out using an Agilent 6530 liquid chromatograph under the following conditions: Agilent Eclipse plus C18 column (3.5 μL, 4.6 mm × 100 mm), UV detection at λobs = 254 nm, flow rate: 0.4 mL/min, linear gradient elution method (5–100% of CH3CN in 0.1% aqueous formic acid over 8 min, then 100% CH3CN for 7 min). Samples for analysis were prepared by dissolving the compound in CH3CN/DMSO (15/1, v/v) mixture. All compounds are >95% pure by HPLC. 1H NMR and 13C NMR spectra were recorded on a Bruker Avance III HD 400 MHz spectrometer (Bruker, Billerica, Massachusetts). Chemical shifts for protons are given in δ, and they are referenced to residual protium in the NMR solvent (DMSO-d6: δ = 2.50 ppm, methanol-d4: δ = 4.87 ppm, chloroform-d3: δ = 7.26 ppm). Carbon chemical shifts are referenced to the carbon in the NMR solvent (DMSO-d6: δ = 39.52 ppm, methanol-d4: δ = 49.00 ppm, chloroform-d3: δ = 77.00 ppm). The coupling constants J are given in Hz. Structures of all prepared compounds (carbon signals) were verified using the CSEARCH-Robot-Referee server (https://nmrpredict.orc.univie.ac.at/c13robot/robot.php). IR DRIFT was recorded in cm–1 with a Nicolet AVATAR 370 FT-IR spectrometer (Thermo Fisher Scientific, Waltham, Massachusetts). High-resolution mass was recorded on an LCQ Fleet spectrometer (Thermo Fisher Scientific, Waltham, Massachusetts).

Preparation of Substrates

N-(4-Hydroxy-6-methylpyrimidin-2-yl)cyanamide (3a)

Dicyandiamide 1 (2.0 g, 23.8 mmol, 1.0 equiv) and ethyl acetoacetate (3.7 mL, 28.5 mmol, 1.2 equiv) were suspended in 15 mL of MeOH. Then, NaOMe (1.3 g, 23.8 mmol, 1.0 equiv) was added portionwise at room temperature. The reaction mixture was heated to 70 °C in an oil bath to reflux. With the full conversion of starting dicyandiamide in 24 h based on 1H NMR monitoring, the reaction mixture was cooled to room temperature. The formed precipitate was filtered and washed with 3 × 5 mL of MeOH. The filtrate cake was dissolved in 40 mL of water. This solution was neutralized with glacial acetic acid (pH ∼ 5), and the resulting white precipitate was filtered, washed with 3 × 40 mL of water, and dried under a reduced pressure of 0.4 mbar at room temperature. Physical and spectroscopic data were consistent with previously reported data.[26]

White amorphous solid. Yield = 33% (1.2 g). 1H NMR (400 MHz, DMSO-d6): δ 11.86 (s, 2H), 5.60 (s, 1H), 2.10 (s, 3H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 162.1, 155.2, 153.4, 115.0, 102.5, 18.8 ppm. IR (KBr): ν = 3103, 2995, 2941, 2902, 2839, 2591, 2316, 2199, 2125, 1912, 1853, 1748, 1661, 1494, 1470, 1440, 1428, 1392, 1359, 1263, 1192, 1156, 1096, 1039, 1009, 958, 887 cm–1. HRMS (ESI+) m/z: calcd for C6H7N4O [M + H]+: 151.0614, found: 151.0614.

1-(4-Hydroxy-6-methylpyrimidin-2-yl)-3-(4-propoxyphenyl)guanidine (5aa)

4-Propoxyaniline (400 mg, 2.64 mmol, 1.5 equiv) was added dropwise to a suspension of cyanamide 3a (264.1 mg, 1.76 mmol, 1.0 equiv) in 5.2 mL of EtOH at room temperature. The reaction mixture was heated in an oil bath at 80 °C to reflux. With the full conversion of 3a in 48 h based on TLC monitoring, the reaction mixture was cooled to −35 °C, followed by the addition of an aqueous solution of NaOH (10 mL, 0.1 w/w) dropwise. The resulting solids were filtered and washed with 4 × 20 mL of Et2O and dried under reduced pressure of 0.4 mbar at room temperature. Physical and spectroscopic data were consistent with previously reported data.[25]

White amorphous solid. Yield = 90% (710 mg). 1H NMR (400 MHz, DMSO-d6): δ 7.44 (d, J = 8.4 Hz, 2H), 6.83 (d, J = 8.9 Hz, 2H), 5.53 (s, 1H), 3.88 (t, J = 6.5 Hz, 2H), 2.06 (s, 3H), 1.71 (q, J = 7.0 Hz, 2H), 0.97 (t, J = 7.4 Hz, 3H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 158.6, 156.4, 154.7, 131.9, 123.0 (2C), 114.6 (2C), 114.4, 103.4, 79.2, 69.1, 23.7, 22.1, 10.5 ppm. IR (KBr): ν = 3345, 3267, 3252, 3103, 3046, 3016, 2968, 2866, 1649, 1607, 1577, 1512, 1458, 1422, 1404, 1350, 1264, 1234, 1201, 1168, 1147, 1114, 1075, 1030, 985, 970 cm–1. HRMS (ESI+) m/z: calcd for C15H20N5O [M + H]+: 302.1617, found: 302.1612.

1-(4-Chloro-6-methylpyrimidin-2-yl)-3-(4-propoxyphenyl)guanidine (5ba)

Guanidine 5aa (100.0 mg, 0.33 mmol, 1.0 equiv) was added portionwise to a stirred 1 mL of POCl3 at room temperature. At this temperature, the mixture was stirred to a full conversion of guanidine (20 h based on 1H NMR monitoring). Then, the reaction mixture was concentrated using a rotavap. Water (2.0 mL) was added to an oily residue. The resulting white solids were filtered, washed with 3 × 1 mL of water, and dried under a reduced pressure of 0.4 mbar at room temperature.

White amorphous solid. Quantitative yield (105 mg). 1H NMR (400 MHz, DMSO-d6): δ 11.38 (br s, 1H), 10.33 (s, 1H), 8.56 (br s, 2H), 7.44 (s, 1H), 7.30 (d, J = 8.6 Hz, 2H), 7.06 (d, J = 8.7 Hz, 2H), 3.97 (t, J = 6.4 Hz, 2H), 1.75 (h, J = 7.0 Hz, 2H), 0.99 (t, J = 7.4 Hz, 3H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 171.4, 161.0, 158.8, 157.0, 154.0, 128.2 (2C), 126.4, 116.7 (2C), 116.0, 69.8, 23.7, 22.4, 10.9 ppm. IR (KBr) ν = 3351, 3309, 3058, 2935, 2869, 2367, 1673, 1628, 1601, 1577, 1556, 1512, 1437, 1392, 1383, 1347, 1305, 1266, 1245, 1177, 1144, 1072, 982, 890, 863, 818, 779 cm–1. HRMS (ESI+) m/z: calcd for C15H19ClN5O [M + H]+: 320.1283, found: 320.1273.

General Procedure for the Preparation of 5da–dc

The corresponding amine 4 (1.55 mmol, 2.5 equiv) was added in one portion to a stirred suspension of 5ba (200 mg, 0.62 mmol, 1.0 equiv) and K2CO3 (214 mg, 1.55 mmol, 2.5 equiv) in 6.5 mL of MeCN. The reaction mixture was heated in an oil bath to 85 °C to reflux. With the full conversion of guanidine (typically overnight based on TLC monitoring), the reaction mixture was cooled to 0 °C (ice/water bath). The resulting solids were filtered and washed with 2 × 1 mL of MeCN followed by water (the pH of filtrate should be neutral). The product was dried under a reduced pressure of 0.4 mbar at room temperature.

1-(4-(Benzylamino)-6-methylpyrimidin-2-yl)-3-(4-propoxyphenyl)guanidine (5da)

The title compound was synthesized according to the general procedure using benzylamine (170 μL, 1.55 mmol, 2,5 equiv).

White amorphous solid. Yield = 59% (110 mg). HPLC purity = 99% (tR = 6.5 min). 1H NMR (400 MHz, CDCl3): δ 10.62 (br s, 1H), 9.27 (br s, 1H), 7.45–7.10 (m, 5H), 6.98–6.59 (m, 4H), 5.63 (s, 1H), 4.18 (s, 2H), 3.85 (t, J = 6.6 Hz, 2H), 2.12 (s, 3H), 1.78 (h, J = 7.1 Hz, 2H), 1.02 (t, J = 7.4 Hz, 3H) ppm. 13C{1H} NMR (101 MHz, CDCl3): δ = 164.8, 164.03, 159.5, 155.0, 152.3, 140.6, 138.87, 128.4 (2C), 126.9 (2C), 126.7 (2C), 124.6, 115.5 (2C), 94.7, 69.8, 45.7, 24.0, 22.7, 10.6 ppm. IR (KBr): ν = 3452, 3255, 3159, 3058, 3025, 2956, 2929, 2875, 2812, 1649, 1610, 1559, 1545, 1503, 1488, 1467, 1428, 1395, 1341, 1299, 1278, 1231, 1213, 1165, 1135, 1102, 1063, 1048, 1027, 979, 866, 833, 794, 734, 695, 668, 632, 552, 531 cm–1. HRMS (ESI+) m/z: calcd for C22H27N6O [M + H]+: 391.2241, found: 291.2241.

1-(4-Methyl-6-(phenethylamino)pyrimidin-2-yl)-3-(4-propoxyphenyl)guanidine (5db)

The title compound was synthesized according to the general procedure, using 2-phenylethylamine (200 μL, 1.55 mmol, 2,5 equiv).

White amorphous solid. Yield = 41% (100 mg). HPLC purity = 97% (tR = 6.8 min). 1H NMR (400 MHz, DMSO-d6): δ 7.69 (s, 2H), 7.24 (ddt, J = 22.0, 14.1, 7.3 Hz, 8H), 7.09 (s, 2H), 6.83 (d, J = 8.4 Hz, 2H), 5.85 (s, 1H), 3.86 (t, J = 6.5 Hz, 2H), 3.41 (s, 2H), 2.80 (t, J = 7.4 Hz, 2H), 2.11 (s, 3H), 1.71 (h, J = 7.1 Hz, 2H), 0.97 (t, J = 7.4 Hz, 4H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 163.3, 162.8, 154.1, 152.1, 140.0, 129.2 (2C), 128.7 (2C), 126.5 (2C), 123.2, 115.3 (2C), 97.5, 69.5, 42.3, 35.5, 23.9, 22.6, 10.9 ppm (two qC are missing). IR (KBr): ν = 3461, 3297, 3159, 3070, 3025, 2977, 2956, 2920, 2666, 1649, 1610, 1559, 1488, 1464, 1428, 1392, 1341, 1299, 1272, 1240, 1189, 1168, 1129, 1108, 1066, 1033, 1006, 979, 937, 860, 833, 812, 794, 746, 698, 644, 534, 519 cm–1. HRMS (ESI+) m/z: calcd for C23H29N6O [M + H]+: 405.2397, found: 405.2395.

1-(4-(Butylamino)-6-methylpyrimidin-2-yl)-3-(4-propoxyphenyl)guanidine (5dc)

The title compound was synthesized according to the general procedure, using n-butylamine (150 μL, 1.55 mmol, 2,5 equiv).

White amorphous solid. Yield = 35% (80 mg). HPLC purity ≥ 99% (tR = 6.6 min). H NMR (400 MHz, DMSO-d6): δ 7.62 (br s, 2H), 7.09 (s, 3H), 6.83 (d, J = 8.4 Hz, 2H), 5.83 (s, 1H), 3.86 (t, J = 6.5 Hz, 2H), 3.17 (s, 2H), 2.10 (s, 3H), 1.70 (q, J = 7.0 Hz, 2H), 1.47 (p, J = 7.2 Hz, 2H), 1.32 (p, J = 7.3 Hz, 2H), 1.02–0.90 (m, 3H), 0.87 (d, J = 7.3 Hz, 3H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 163.3, 154.1, 123.2 (2C), 115.2 (2C), 105.0, 97.1, 69.5, 40.4, 31.4, 23.9, 22.6, 20.1, 14.2, 10.9 ppm (three qC are missing). IR (KBr): ν = 3461, 3414, 3294, 3204, 3162, 3028, 2956, 2926, 2869, 1646, 1610, 1485, 1467, 1431, 1395, 1362, 1341, 1314, 1272, 1237, 1153, 1102, 1057, 1006, 967, 923, 863, 839, 800, 749, 701 cm–1. HRMS (ESI+) m/z: calcd for C19H29N6O [M + H]+: 357.2397, found: 357.2407.

N-(4,6-Dimethylpyrimidin-2-yl)cyanamide (3c)

Dicyandiamide 1 (5.0 g, 60 mmol) and acetylacetone (9.0 g, 90 mmol) were added portionwise to 40 mL of a stirred aqueous 0.3 M NaOH solution at room temperature. The resulting suspension was heated in an oil bath to 110 °C to reflux. With the full conversion of dicyandiamine (48 h based on 1H NMR monitoring), the reaction mixture was cooled to 0 °C (ice/water bath). The resulting solids were filtered and washed with 1 × 20 mL of water. The crude product was recrystallized from 120 mL of boiling ethanol and dried under a reduced pressure of 0.4 mbar at room temperature. Physical and spectroscopic data were consistent with previously reported data.[26]

White crystalline solid. Yield = 44% (3.9 g). Mp = 228–229 °C (ethanol). 1H NMR (400 MHz, DMSO-d6): δ 12.58 (s, 1H), 6.63 (s, 1H), 2.31 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.8, 160.2 (2C), 115.8, 109.7, 21.9 (2C) ppm. IR (KBr) ν = 3503, 3282, 3249, 3064, 3010, 2980, 2857, 2842, 2815, 2621, 2319, 2244, 2202, 2175, 2089, 1838, 1727, 1649, 1610, 1422, 1362, 1323, 1231, 1195, 1165, 1036, 1018, 985 cm–1. HRMS (ESI+) m/z: calcd for C7H9N4 [M + H]+: 149.0827, found: 149.0792.

General Procedure for the Preparation of Guanidines 5ca–cj

The corresponding amine 4 (2.0 mmol, 1.5 equiv) was added dropwise to a stirred suspension of cyanamide 3c (200 mg, 1.35 mmol, 1.0 equiv) in 4 mL of EtOH at room temperature. The reaction mixture was heated in an oil bath to 80 °C to reflux. With the full conversion of cyanamide (48 h based on TLC monitoring), the reaction mixture was cooled to −35 °C in a freezer. At this temperature, the solution of NaOH (10 mL, 0.1 w/w) was added dropwise. The resulting solids were filtered, washed with 4 × 20 mL of Et2O and dried under a reduced pressure of 0.4 mbar at room temperature.

1-(4,6-Dimethylpyrimidin-2-yl)-3-(4-propoxyphenyl)guanidine (5ca)

The title compound was synthesized according to the general procedure, using 4-propoxyaniline (300 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 47% (193 mg). HPLC purity = 99% (tR = 5.8 min). 1H NMR (400 MHz, MeOD-d4): δ 7.21–7.08 (m, 2H), 6.92–6.83 (m, 2H), 6.61 (s, 1H), 3.89 (t, J = 6.5 Hz, 2H), 3.27 (p, J = 1.6 Hz, 1H), 2.31 (s, 6H), 1.83–1.65 (m, 2H), 1.01 (t, J = 7.4 Hz, 3H) ppm. 13C{1H} NMR (101 MHz, MeOD-d4): δ 168.6 (2C), 165.2, 158.1, 157.3, 133.7, 126.6 (2C), 116.4 (2C), 112.9, 70.9, 23.74 (2C), 23.69, 10.9 ppm. IR (KBr) ν = 3312, 3106, 3088, 2959, 2893, 2869, 1631, 1577, 1527, 1509, 1419, 1383, 1344, 1237, 1171, 1117, 1075, 1048, 1024 cm–1. HRMS (ESI+) m/z: calcd for C16H22N5O [M + H]+: 300.1824, found: 300.1773.

1-(4,6-Dimethylpyrimidin-2-yl)-3-phenylguanidine (5cb)

The title compound was synthesized according to the general procedure, using aniline (180 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 31% (100 mg). 1H NMR (400 MHz, DMSO-d6): δ 7.34 (d, J = 7.9 Hz, 2H), 7.20–7.07 (m, 2H), 6.72 (tt, J = 7.2, 1.2 Hz, 1H), 6.22 (s, 1H), 2.16 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.2, 165.5 (2C), 157.7, 147.3, 128.5 (2C), 121.0, 119.3 (2C), 107.6, 24.1 (2C) ppm. IR (KBr): ν = 3306, 3273, 3231, 3150, 3114, 3052, 2962, 2920, 1637, 1604, 1577, 1524, 1500, 1449, 1416, 1407, 1377, 1332, 1299, 1245, 1204, 1174, 1159, 1114, 1102, 1027, 991, 967, 899, 830, 806, 746 cm–1. HRMS (ESI+) m/z: calcd for C13H16N5 [M + H]+: 241.1405, found: 242.1400.

1-Benzyl-3-(4,6-dimethylpyrimidin-2-yl)guanidine (5cc)

The title compound was synthesized according to the general procedure, using benzylamine (220 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 46% (155 mg). 1H NMR (400 MHz, DMSO-d6): δ 7.33 (d, J = 5.0 Hz, 4H), 7.29–7.17 (m, 1H), 6.46 (s, 1H), 4.48 (s, 2H), 2.21 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.5, 166.1 (2C), 158.1, 140.7, 128.7 (2C), 127.5 (3C), 127.1, 110.5, 43.9, 24.0 (2C) ppm. IR (KBr): ν = 3276, 3183, 3129, 3108, 3022, 2923, 2866, 1607, 1577, 1419, 1374, 1329, 1240, 1210, 1174, 1099, 1081, 1060, 988, 973, 946, 920, 887, 827, 809, 749, 737, 698 cm–1. HRMS (ESI+) m/z: calcd for C14H18N5 [M + H]+: 256.1554, found: 256.1557.

1-(4,6-Dimethylpyrimidin-2-yl)-3-(4-methylbenzyl)guanidine (5cd)

The title compound was synthesized according to the general procedure, using 4-methylbenzylamine (260 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 72% (260 mg). 1H NMR (400 MHz, DMSO-d6): δ 7.65 (br s, 1H), 7.21 (d, J = 8.0 Hz, 2H), 7.13 (d, J = 7.8 Hz, 2H), 6.46 (s, 1H), 4.43 (s, 2H), 2.27 (s, 3H), 2.21 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.5, 166.1 (2C), 158.2, 137.6, 136.1, 129.3 (2C), 127.5 (2C), 110.5, 43.8, 24.0 (2C), 21.1 ppm. IR (KBr): ν = 3270, 3252, 3103, 2920, 2884, 2155, 1900, 1607, 1592, 1565, 1521, 1416, 1377, 1338, 1207, 1159, 1135, 1069, 1030, 1021, 985, 964, 952, 911, 833, 812 cm–1. HRMS (ESI+) m/z: calcd for C15H20N5 [M + H]+: 270.1717, found: 270.1713.

1-(4,6-Dimethylpyrimidin-2-yl)-3-(3-methylbenzyl)guanidine (5ce)

The title compound was synthesized according to general procedure, using 3-methylbenzylamine (250 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 72% (263 mg). 1H NMR (400 MHz, DMSO-d6): δ 7.21 (t, J = 7.5 Hz, 1H), 7.16–7.09 (m, 2H), 7.05 (d, J = 7.5 Hz, 1H), 6.47 (s, 1H), 4.44 (s, 2H), 2.29 (s, 3H), 2.22 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.5, 166.1 (2C), 158.0, 140.5, 137.8, 128.7, 128.1, 127.8, 124.6, 110.6, 43.9, 24.0 (2C), 21.5 ppm. IR (KBr): ν = 3267, 3106, 3049, 3016, 2917, 1939, 1613, 1571, 1539, 1425, 1380, 1332, 1308, 1237, 1160, 1144, 1096, 1069, 1036, 1015, 991, 970, 943, 827, 812 cm–1. HRMS (ESI+) m/z: calcd for C15H20N5 [M + H]+: 270.1716, found: 270.1713.

1-(4,6-Dimethylpyrimidin-2-yl)-3-(2-methylbenzyl)guanidine (5cf)

The title compound was synthesized according to the general procedure, using 2-methylbenzylamine (250 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 80% (300 mg). 1H NMR (400 MHz, DMSO-d6): δ 7.65 (br s, 1H), 7.21 (d, J = 8.0 Hz, 2H), 7.13 (d, J = 7.8 Hz, 2H), 6.46 (s, 1H), 4.43 (s, 2H), 2.27 (s, 3H), 2.21 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.5, 165.9 (2C), 158.3, 138.6, 135.9, 130.3, 127.6, 127.0, 126.1, 109.9, 42.4, 23.5 (2C), 19.1 ppm. IR (KBr): ν = 3282, 3267, 3108, 3094, 3064, 2983, 1619, 1568, 1530, 1416, 1332, 1314, 1245, 1219, 1192, 1153, 1114, 1060, 1045, 1030, 991, 967, 931, 890, 806 cm–1. HRMS (ESI+) m/z: calcd for C15H20N5 [M + H]+: 270.1714, found: 270.1713.

1-(4-(tert-Butyl)benzyl)-3-(4,6-dimethylpyrimidin-2-yl)guanidine (5cg)

The title compound was synthesized according to the general procedure, using 4-tert-butylbenzylamine (350 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 82% (345 mg). HPLC purity = 99% (tR = 6.5 min). 1H NMR (400 MHz, CDCl3): δ 7.36–7.32 (m, 2H), 7.28 (s, 2H), 6.39 (s, 1H), 4.48 (s, 2H), 2.27 (s, 6H), 1.29 (s, 9H) ppm. 13C{1H} NMR (101 MHz, CDCl3): δ 166.4 (2C), 158.3, 150.6, 134.8, 126.7 (2C), 125.8 (2C), 111.1, 45.2, 34.5, 31.3 (3C), 24.00 (2C) ppm, one qC is missing. IR (KBr): ν = 3270, 3258, 3108, 3055, 2962, 2866, 1897, 1616, 1571, 1530, 1416, 1380, 1355, 1305, 1272, 1240, 1177, 1162, 1117, 1018, 988, 970, 934, 899, 824, 806 cm–1. HRMS (ESI+) m/z: calcd for C18H26N5 [M + H]+: 312.2187, found: 312.2183.

1-(4,6-Dimethylpyrimidin-2-yl)-3-(4-fluorobenzyl)guanidine (5ch)

The title compound was synthesized according to the general procedure, using 4-fluorobenzylamine (230 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 68% (250 mg). 1H NMR (400 MHz, DMSO-d6): δ 7.75 (br s, 2H), 7.41–7.30 (m, 2H), 7.19–7.08 (m, 2H), 6.47 (s, 1H), 4.46 (s, 2H), 2.21 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.1, 166.7, 162.3, 159.9, 157.8, 136.6, 129.0 (d, J = 8.0 Hz, 2C), 115.0 (d, J = 21.2 Hz, 2C), 110.2, 42.7, 23.6 (2C) ppm. 19F NMR (376 MHz, DMSO-d6): δ −116.51 ppm. IR (KBr): ν = 3258, 3103, 3088, 3067, 2962, 1885, 1622, 1601, 1577, 1542, 1509, 1419, 1383, 1338, 1216, 1180, 1156, 1096, 1015, 994, 976, 943, 931, 899, 827, 806 cm–1. HRMS (ESI+) m/z: calcd for C14H17FN5 [M + H]+: 274.1465, found: 274.1463.

1-(4,6-Dimethylpyrimidin-2-yl)-3-(naphthalen-1-ylmethyl)guanidine (5ci)

The title compound was synthesized according to the general procedure, using 1-naphthylmethylamine (290 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 82% (340 mg). HPLC purity = 99% (tR = 6.0 min). 1H NMR (400 MHz, DMSO-d6): δ 8.23–8.09 (m, 1H), 8.03–7.92 (m, 1H), 7.86 (dd, J = 7.4, 2.0 Hz, 1H), 7.70 (br s, 1H), 7.63–7.37 (m, 4H), 6.48 (s, 1H), 4.97 (s, 2H), 2.21 (s, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.5, 166.1 (2C), 158.3, 135.8, 133.7, 131.3, 128.9, 127.8, 126.6, 126.2, 125.9, 125.3, 124.0, 110.6, 42.2, 24.0 (2C) ppm. IR (KBr): ν = 3279, 3117, 3088, 3067, 3052, 2923, 2792, 1607, 1568, 1533, 1416, 1383, 1335, 1257, 1231, 1213, 1177, 1045, 1027, 994, 967, 937, 887, 830, 809, 776 cm–1. HRMS (ESI+) m/z: calcd for C18H20N5, [M + H]+: 306.1714, found: 306.1713.

1-(4,6-Dimethylpyrimidin-2-yl)-3-phenethylguanidine (5cj)

The title compound was synthesized according to the general procedure, using 2-phenylethylamine (250 μL, 2.0 mmol, 1.5 equiv).

White amorphous solid. Yield = 61% (300 mg). HPLC purity = 98% (tR = 5.4 min). 1H NMR (400 MHz, DMSO-d6): δ 7.34–7.24 (m, 5H), 7.24–7.17 (m, 1H), 6.45 (s, 1H), 3.46 (t, J = 7.2 Hz, 2H), 2.81 (t, J = 7.2 Hz, 2H), 2.20 (d, J = 6.5 Hz, 6H) ppm. 13C{1H} NMR (101 MHz, DMSO-d6): δ 166.5, 166.0 (2C), 158.0, 140.1, 129.2 (2C), 128.8 (2C), 126.5, 110.4, 42.0, 35.8, 24.1 (2C) ppm. IR (KBr): ν = 3282, 3117, 2998, 2947, 2920, 1619, 1571, 1542, 1419, 1377, 1332, 1287, 1263, 1216, 1177, 1093, 1033, 991, 967, 949, 926, 845, 830, 809, 719 cm–1. HRMS (ESI+) m/z: calcd for C15H20N5 [M + H]+: 270.1715, found: 270.1713.

Estimation of Compound Solubility in Water

To estimate the solubility, we prepared saturated solutions of the compounds by adding 5 mg of each selected compound to 700 μL of 20 mM phosphate buffer (pH 6.5) containing 50 mM KCl. After 48 h of gentle agitation at room temperature (RT), the solutions were spun down, and the supernatant was filtered through 0.2 μm of SPARTAN 13/0.2 RC filter units (Merck KgaA, Darmstadt, Germany) and diluted 100–200× in the same phosphate buffer. The diluted solutions were transferred into the quartz cuvette, recording UV–vis absorption spectra in the range from 200 to 600 nm on a UV-1800 UV/visible scanning spectrophotometer (Shimadzu, Kyoto, Japan) at room temperature. The compound concentrations were estimated by the single point calibration using the absorption at wavelengths higher than 240 nm outside of DMSO absorption of 2000× diluted 50 mM DMSO-dissolved stock solution in phosphate buffer.

Protein Preparation

FOXO3-DBD and FOXO1-DBD were purified, as described previously.[25,32] DNA encoding human FOXO3-DBD (residues 156–269) and mouse FOXO1-DBD (residues 156–269) were cloned into pGEX-6P-1 and expressed in E. coli BL21(DE3) as the N-terminal GST-tagged fusion proteins. Proteins for 1H–15N HSQC NMR experiments were expressed in minimal media supplemented with 1 g/L 15N-ammonium chloride (Cambridge Isotope Laboratories, Inc., Tewksbury, Massachusetts) as the sole nitrogen source and purified as described below. Protein expression was induced with 0.5 mM isopropyl β-d-1-thiogalactopyranoside (Sigma-Aldrich s.r.o., St. Louis, Missouri) for 16–20 h at 20 °C. The pellet from 1 L of culture was resuspended in 25 mL of lysis buffer containing 20 mM phosphate (pH 7.4), 250 mM NaCl, 1 mM ethylenediamine tetraacetic acid (EDTA), and 10 mM dithiothreitol (DTT). The cells were disrupted by sonication, and the fusion protein was purified by Glutathione Sepharose 4 Fast Flow (GE Healthcare, Chicago, Illinois) in a buffer containing 20 mM tris–HCl (pH 7.5), 0.5 M NaCl, 1 mM EDTA, and 10 mM DTT. After elution with 10 mM reduced glutathione (pH 8.0), the proteins were dialyzed against a buffer containing 20 mM tris–HCl buffer (pH 7.5), 150 mM NaCl, 1 mM DTT, 1 mM EDTA, and 10% (wt/vol) glycerol. The GST affinity tag was removed by PreScission protease (10 U/mg of recombinant protein) by incubation at 4 °C overnight. The last purification step was size-exclusion chromatography on a HiLoad Superdex 75 26/600 column (GE Healthcare, Chicago, Illinois) in a buffer containing 20 mM phosphate buffer (pH 6.5), 1 mM DTT, 50 mM KCl, and 10% (wt/vol) glycerol.

Human FOXO4-DBD (residues 86–211) was expressed as a His6-N-terminal tagged fusion protein and purified as described previously.[35] The affinity tag was removed by thrombin (Sigma-Aldrich s.r.o., St. Louis, Missouri) (10 U/mg recombinant proteins) by incubation at 4 °C overnight. The protein was then dialyzed against a 20 mM citric acid (pH 6.3) and 1 mM EDTA buffer. Subsequently, FOXO4-DBD was purified by cation-exchange chromatography on a HiTrap SP column (GE Healthcare, Chicago, Illinois). The final purification step was size-exclusion chromatography on a HiLoad Superdex 75 26/600 column (GE Healthcare, Chicago, Illinois) in a buffer containing 20 mM phosphate buffer (pH 6.5), 1 mM DTT, 50 mM KCl, and 10% (wt/vol) glycerol.

STD NMR Measurements

STD experiment was performed at 298 K on a Bruker Avance III HD 850 MHz spectrometer (Bruker, Billerica, Massachusetts) equipped with a 13C/1H/15N cryoprobe. The 550 μL sample contained 15 μM FOXO3-DBD and 1 mM 5cj in 20 mM phosphate buffer (pH 6.5) supplemented by 50 mM KCl and 10% 2H2O. 1H STD NMR was performed using a standard “stddiffesgp” pulse sequence with excitation sculpting and pulse-field gradients for water suppression. On-resonance irradiation was done at −0.54 ppm and off-resonance irradiation at 30 ppm, using a 50 ms shaped Eburp2.1000 pulse of power 40 dB for a saturation time of 3 s. The difference spectra were obtained by subtracting the on-resonance spectrum from the off-resonance spectrum. The spectra were analyzed using TopSpin software (v3.6).

2D 1H–15N HSQC NMR Measurements

2D 1H–15N HSQC spectra were acquired on both Bruker Avance III HD 850 MHz and Bruker Avance III 600 MHz (Bruker, Billerica, Massachusetts) spectrometers equipped with a 13C/1H/15N cryoprobe. All HSQC experiments were measured at 298 K in Shigemi NMR tubes (Shigemi Co., Ltd., Tokyo, Japan) containing 350 μL of sample in 100 mM phosphate buffer (pH 6.5) supplemented with 50 mM KCl and 10% 2H2O. The 100 mM phosphate buffer was used to avoid unwanted effects caused by pH changes induced by the presence of the tested compounds at high concentrations. 2D 1H–15N HSQC measurements were performed with 100 μM FOXO-DBDs and a given concentration of the compound. All spectra were processed in TopSpin software (v3.6) and evaluated in Sparky software (v3.1).[43] CSP values obtained from 2D 1H–15N HSQC experiments were calculated using the following formula[44]

HADDOCK Docking Calculations

Docking calculations were performed using HADDOCK2.4.[30,31] FOXO3-DBD and FOXO4-DBD coordinates were taken as the lowest-energy structure of the PDB entries 2K86 and 1E17. The first two N-terminal amino acid residues of FOXO3-DBD from the thrombin cleavage site were removed from the structure. For all docking runs, active residues of FOXO3-DBD were chosen as residues with CSPs higher than the mean + 2σcorr0. For FOXO3-DBD, residues 151–160 (N-terminus) and 241–251 (C-terminus) were set as flexible. For FOXO4-DBD, residues 96–105 (N-terminus) and 181–185 (C-terminus) were set as flexible. The ligand was set as active and flexible in both cases. Unmentioned running parameters were left on their default values for protein–ligand mode. The resulting structures were clustered by RMSD with a 1.5 Å cutoff and ranked based on their HADDOCK scores. The top-scoring cluster satisfactorily explained the NMR data of all compounds but 5ci. Its data were satisfactorily explained by combining the top and second clusters, as described in the Results section.

Cell Lines, Culture Conditions, and Reagents

The neuroblastoma cell line SH-EP/FOXO3 was cultured in RPMI1640 (Lonza, Basel, Switzerland) with 10% fetal bovine serum (GIBCO BRL, Paisley, U.K.), 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM l-glutamine (Sigma-Aldrich, Vienna, Austria) at 5% CO2. All cultures were routinely tested for mycoplasma contamination using the Venor GeM-mycoplasma detection kit (Minerva Biolabs, Berlin, Germany).

Promoter Activity

Transcriptional activation of the DEPP1 promoter by conditional FOXO3 was assessed after transfection of a DEPP1-luciferase reporter plasmid pGS c10orf10 Prom wt (DEPP1-LUC)[20] into SH-EP/FOXO3 cells using the JetPrime reagent (Polyplus, Berkeley). After 24 h of transfection, mTFP fluorescent protein expression from a cotransfected expression vector was assessed by live cell fluorescence microscopy to verify equal transfection rates before seeding cells into 24-well plates. After 48 h of transfection, the cells were pretreated with indicated concentrations of each compound (30 min) before adding 100 nM 4-hydroxy-tamoxifen (4OHT) for 3 h to activate ectopic FOXO3. Firefly-luciferase was analyzed using the Luciferase Assay System (Promega, Madison). Luminescence intensity was measured in a Hidex Chameleon—Multilabel Microplate Reader (Hidex, Turku, Finland).

The effects of these compounds on the endogenous DEPP1 promoter and on the actin promoter, which does not contain a FOXO consensus sequence, were assessed in SH-EP cells. For this purpose, 2 × 106 SH-EP cells were transfected with either a DEPP1-LUC-reporter plasmid or an actin-LUC reporter plasmid using JetPrime reagent (Polyplus, Berkeley). Actin-LUC was constructed by inserting a 1.3 kb fragment of the chicken β actin promoter into pGL3-luc reporter vector (Promega, Madison). After 24 h of transfection, the cells were seeded into 24-well plates. After another 24 h, the cells were pretreated with indicated concentrations of compound for 3.5 h. Firefly-luciferase was analyzed using the Luciferase Assay System (Promega, Madison). Luminescence intensity was measured in a Hidex Chameleon—Multilabel Microplate Reader (Hidex, Turku, Finland).

Fluorescence Polarization (FP) Analyses

The binding of compound 5ca and its derivatives to FOXO3-DBD was measured in vitro by FP in black, flat-bottom, 96-well plates (HVD Life Sciences, Vienna, Austria) in a Hidex Chameleon—Multilabel Microplate Reader (Hidex, Turku, Finland). Each compound (dissolved in DMSO, 1 μM final concentration) was added to 100 μL of reaction mix containing 125 nM FOXO3-DBD and 25 nM FAM-labeled double-strand FOXO3 consensus sequence oligonucleotides (5′-CTA TCA AAA CAA CGC-3′) in assay buffer (20 mM tris–HCl, 100 mM NaCl, 1 mM EDTA, pH 7.5). The final concentration of DMSO in all mixtures was adjusted to 5% (v/v). Positive (only assay buffer and FAM-labeled oligonucleotide) and negative (FOXO3-DBD and FAM-labeled oligonucleotide) controls were analyzed on each plate. Milli polarization values (mP) were measured at an excitation wavelength of 485 nm and an emission wavelength of 530 nm.

Resazurin Viability Measurement to Determine IC50 Values for FOXO3-Interacting 5ca Derivatives

Neuroblastoma cells SH-EP/FOXO3 were seeded into 96-well plates and treated with concentrations 0.05, 0.1, 0.2, 0.39, 0.78, 1.56, 3.125, 6.25, 12.5, 18.75, and 25 μM of 5ca and its derivatives 30 min before activating ectopic FOXO3 by 100 nM 4OHT. For 5ca derivatives, concentrations above 3.125 μM had to be excluded from the analysis due to spontaneous cell death induction. Resazurin salt (Sigma-Aldrich s.r.o., St. Louis, Missouri) was diluted to a final concentration of 0.01% (wt/vol), and 20 μL of this dilution was added to 200 μL of the cell culture in each well 24 h after FOXO3 activation. Resazurin is reduced in viable cells to red-fluorescent resoflurin, which was quantified in a Hidex Chameleon—Multilabel Microplate Reader (Hidex, Turku, Finland) at filter settings 544 nm Ex/616 nm Em.