Meriolin 16 and 36 are highly cytotoxic in different human leukemia and lymphoma cell lines and in malignant primary cells derived from leukemia and lymphoma patients
Based on the apoptotic capacity of Meriolin 31 [22], a novel derivative (termed Meriolin 16) with an additional methoxy group was synthesized with the intention of enhancing its cytotoxicity (Fig. 1B). We analyzed the cytotoxic potential of the novel derivative Meriolin 16 compared to the previously described derivative Meriolin 36 [22] in different human leukemia and lymphoma cell lines (i.e. HL60 (acute myeloid leukemia; AML), Jurkat (T cell acute lymphoblastic leukemia; T-ALL), HPBALL (T-ALL), K562 (chronic myeloid leukemia; CML), KOPTK1 (T-ALL), MOLT4 (T-ALL), SUPB15 (B cell acute lymphoblastic leukemia; B-ALL), and Ramos (B cell Burkitt lymphoma)). As shown in Fig. 2A, Meriolin 16 potently induced cell death at nanomolar range (IC50 values ranging from 10 to 40 nM) in all cell lines tested and was substantially more cytotoxic than Meriolin 36 (IC50 values ranging from 20 to 3500 nM). In addition, both Meriolin derivatives were highly cytotoxic in malignant primary cells derived from patients with diffuse large B cell lymphoma cells (DLBCL), follicular lymphoma cells (FL) or chronic lymphocytic leukemia cells (CLL) (Fig. 2B).
In vitro kinase assays and kinome screen reveal Meriolin 16 and 36 as potent inhibitors of cyclin-dependent kinases (CDKs)
Meriolin derivatives have been shown to inhibit a variety of CDKs [16, 17, 21]. Therefore, we used a luminescence-based kinase activity assay to investigate in how far Meriolin 16 and 36 might inhibit CDK1, CDK2 and CDK9. These three CDKs were selected due to their cell cycle regulatory function (CDK1 and 2) or transcriptional control (CDK9). The CDK 1, 2 and 4 inhibitor R547 was used as positive control [28]. Meriolin 16 and Meriolin 36 inhibited CDK1/Cyclin B1, CDK2/Cyclin A2 and CDK9/Cyclin T in a concentration-dependent manner and in a similar range as R547 – though the inhibitory effect on CDK2 was less pronounced (Fig. 3A-C).
In order to accomplish a comprehensive approach, we performed a kinome screening. In this screening, the inhibitory activity of Meriolin 16 and 36 was analyzed on a panel of 335 kinases (performed by 33PanQinase™ from Reaction Biology). Meriolin 16 and Meriolin 36 were tested in two concentrations according to their differing IC50 values (i.e. Meriolin 16 with 0.03 µM and 0.3 µM, Meriolin 36 with 0.3 µM and 3 µM). The results are shown as a kinome tree in Fig. 3D. We observed that the lower concentration of Meriolin 16 (0.03 µM; which equals the IC50 value in Ramos cells) showed a high specificity for kinases within the yellow area of the kinome tree, which represents the CMGC family. The CMGC family consists of CDKs, MAPKs (mitogen-activated protein kinases), GSKs (glycogen synthase kinases) and CLKs (CDC-like kinases) [19, 22]. Increasing the concentration to 0.3 µM, a more unspecific inhibition pattern for Meriolin 16 was apparent, with increasing diameter of the dots of the CMGC family correlating with increased inhibitory activity. For Meriolin 36, the concentration of 0.3 µM shows a comparable pattern of dots to the kinome tree with 0.3 µM of Meriolin 16. Again, a prevalence for the CMGC family but also a specificity for the AGC family (cAMP-dependent protein kinase (PKA), the cGMP-dependent protein kinase (PKG) and the protein kinase C (PKC); green area) consisting of 63 evolutionarily related serine/threonine protein kinases [29] could be observed. At 3 µM, Meriolin 36 inhibited almost all kinases tested, at least to a lower extent. In Fig. 3E, a heatmap of the inhibition profile of Meriolin 16 and 36 on all CDKs tested within the kinome screen is provided. Thus, in addition to CDK1, 2 and 9 (Fig. 3A-C) almost all tested CDKs in complex with their respective cyclins were inhibited (such as CDK1, 2, 3, 5, 7, 8, 9, 12, 13, 16, 17, 18, 19, 20) according to the heatmap in Fig. 3E. However, CDK4 and 6 were inhibited to a lesser extent – especially at the lower concentration (0.03 µM) of Meriolin 16. Intriguingly, CDK9, which regulates transcription via phosphorylation of RNA polymerase II, was also completely inhibited by both Meriolin derivatives.
In order to determine the overall specificity a selectivity score was assessed, whereby a high specificity is characterized by a low value. As shown in Fig. 3F, Meriolin 16 displayed the highest selectivity at 0.03 µM with a score of 0.140, while Meriolin 36 was less specific. A comprehensive inhibition profile of all tested 335 kinases is provided in Supplemental Table 1.
Drug-to-target modeling studies of the binding mode of Meriolin 16 and 36 to the ATP-pocket of CDK2
The kinome-wide assay results showed that both Meriolin derivatives are active against CDKs. To further investigate the structural basis of the binding mechanism, we performed molecular docking on CDKs. Since the active site across all the inhibited CDKs is fairly conserved (Supplemental Fig. 2), Meriolins should share a similar binding mode for each protein. To assess this, we docked both Meriolin 16 and 36 into the structures of the active forms of CDK1, 2, and 9. Our results show an overall similar binding mode across all different CDKs, with the indole moiety positioned next to the loop between the β5 strand and the α2 helix for both compounds. In Meriolin 36, small variations in the position of the benzene group can be observed (Supplemental Fig. 3). Since the binding mode is conserved among all CDKs, we chose CDK2 as a representative system to investigate the binding pose of both Meriolins in more detail.
In both derivatives, the indole moiety forms non-covalent interactions with the backbone of the conserved residues Leu83 and Glu81 (Fig. 4A,B). In case of Meriolin 16, we observed electrostatic interactions between the amino groups of the di-aminopyridine moiety and Glu145 and Glu51. Glu51 is part of the C-helix, a critical secondary structure element that undergoes major conformational rearrangements upon CDK activation [30]. The observed binding mode is in agreement with what has been experimentally observed for other indole-containing molecules through X-ray crystallography [31], where the indole binds to the same backbone atoms of the protein. Nonetheless, our initial docking results also showed an alternative possible binding mode with a better docking score for Meriolin 36 (-10.02 vs -10.81 kcal mol− 1). We termed this alternative pose "flipped mode", as in this case the aminopyridyl region is making interactions with the backbone atoms of Leu83, and the indole moiety of Meriolin 36 is placed deeper in the protein pocket surrounded by hydrophobic residues such as Val18, Phe80 and Leu134 (Supplemental Fig. 4A).
To further corroborate our binding mode predictions, we performed microsecond-long all-atom molecular dynamics simulations and assessed the overall stability of the docked poses. To quantify how tightly each ligand fits into the active site of CDK2, we calculated the B-factor of each atom of Meriolin 16 and 36, which measures how mobile the atoms are through the course of the simulations, and we also calculated the total volume that each ligand occupied within the protein through the simulations (Fig. 4C,D). The results show that Meriolin 16 remained tightly bound within the active site throughout the simulations as evidenced by the low B-factor values and the small effective volume occupied (Fig. 4C). Meriolin 36 also remains bound through the entire 5 µs sampled, however, different behaviors can be observed at different regions of the molecule. The indole moiety remains firmly placed in the loop between β5 strand and α2 helix, whereas the benzene group moves around the pocket due to the lack of prominent hydrophobic interactions, as can be seen by the high B-factor values of this moiety, and the shape of the volume used by the ligand (Fig. 4D).
Simulations of the flipped mode of Meriolin 36 show an overall more mobile pose compared to the canonical binding pose, with both the benzene and the indole group displaying higher B-factor values than in the standard mode and showing a bigger effective volume due to the constant tumbling of the ligand within the active site, suggesting weaker interactions with the protein site (Supplemental Fig. 4B). The exposed hydrophobic surface on ligands is commonly associated with higher desolvation penalties and a worse fit to the active site [32]. For this reason, we quantified the solvent-accessible surface area (SASA) of Meriolin 36 on its standard and flipped binding mode. The results show a significant increase of SASA when Meriolin 36 is bound in the putative flipped mode (Supplemental Fig. 4C).
CDKs undergo drastic conformational rearrangements upon activation by cyclins, which are essential for the protein function. Interestingly, crystal structures show that ATP can bind to CDK2 in its inactive state [33]. Since kinase inhibitors can be specific for active or inactive states [34], we tested whether Meriolin 16 and 36 could also bind to CDKs in their inactive conformational state by docking both compounds onto the inactive ATP-bound crystal structure of CDK2. Only Meriolin 16 yielded a binding pose with a score below − 5.0 kcal mol− 1. In this pose, the di-aminopyridyl moiety occupies the ribose binding region of the active site (Supplemental Fig. 4D, left panel). To assess the robustness of this alternative complex, we also performed molecular dynamics simulations starting from this docking pose. In four out of five replicas, we observed dissociation of the ligand within the first 500 ns of simulation (Supplemental Fig. 4D, right panel), which is indicative of a false-positive binding mode [35]. Therefore, the conformational changes triggered by the presence of a cyclin partner seem to be essential for the Meriolin binding. Thus, in contrast to Meriolin 36, which only binds to the active conformational state of CDK2, Meriolin 16 can also bind to the inactive state of CDK2.
Effect of Meriolin 16 and 36 on CDK-mediated phosphorylation of the Retinoblastoma (RB) protein
Since both Meriolin derivatives selectively inhibited CDKs at low concentrations (though also other kinases at higher concentrations; see Fig. 3D and Supplemental Table 1), we focused on the downstream signaling of cell cycle regulation in G1, S, G2 and M phase. In G1 phase, accumulation of CDK4/6/Cyclin D is necessary for cell cycle entry [5]. CDK4/6/Cyclin D are able to monophosphorylate the Retinoblastoma (RB) protein at any of its 14 known phosphosites and CDK1 and 2 are the main contributors to hyperphosphorylation of RB [6, 36]. The RB protein is the main regulator protein for gene expression, since in its hypo- or mono-phosphorylated form in complex with the transcription factor DP they suppress E2F-dependent gene expression during G1. This complex dissociates from E2F-regulated genes, as soon as RB gets phosphorylated by CDK4/6/Cyclin D on Ser249, Thr252 (for overview see Supplemental Fig. 1).
To investigate the possible effect of CDK inhibition by Meriolins on the RB protein, Ramos cells were treated with Meriolin 16 and 36 with 0.1 µM or 1 µM in a kinetics up to 24 h. Subsequently, the expression of Cyclin B1, Cyclin D3, phospho-RB (p-Ser249 and p-Thr252) and CDK1 was monitored by immunoblotting and quantified respectively. Thus, we observed that Meriolin 16 reduced the phosphorylation of RB at p-Ser249 and p-Thr252 after 24 h whereas Meriolin 36 had no effect on these RB-phosphosites. The CDK1, 2 and 4 inhibitor R547 however, completely abrogated the phosphorylation of RB within 4 h (Supplemental Fig. 5A,B). Since in contrast to R547, Meriolin 16 and 36 do not inhibit CDK4 at the applied concentrations, the effect on the CDK4-mediated phosphorylation of RB at p-Ser249, p-Thr252 was accordingly less pronounced. The Cyclin D3 levels were not impaired by Meriolins or R547 (Supplemental Fig. 5A,B). In contrast to Meriolin 36 and R547, which rather increased the expression of Cyclin B1 after 24 h, Meriolin 16 reduced the expression of Cyclin B1, whereas the expression of CDK1 was not affected by Meriolins or R547 (Supplemental Fig. 5C,D).
Next, we investigated the effect of Meriolin 16 and 36 on the CDK2-mediated phosphorylation of RB at Ser612 and Thr821. Sequential phosphorylation of the suppressor protein RB by CDKs ensures the inactivation of the suppressor activity of RB and thereby allows cell cycle progression. In early S phase, CDK2 is associated with Cyclin E and in late S/G2 phase in a complex with Cyclin A and plays a crucial role in genome replication [5]. At the end of the S phase, Cyclin A replaces Cyclin E by forming a new complex with CDK2, then Cyclin E is degraded [37]. The CDK2/Cyclin A complex is responsible for the termination of the S phase, driving the transition from S phase to G2, whereby the subsequent activation of CDK1 by Cyclin A enables the cell to undergo transition to M phase [37] (a schematic overview is given in Fig. 5A). To investigate the influence on the signaling in the S phase of the cell cycle, the protein levels of CDK2, Cyclin A2, Cyclin E, and phosphorylation of RB at Thr821 and Ser612 were analyzed via immunoblotting (Fig. 5B,C and Supplemental Fig. 6A,B). Quantification of the immunoblot kinetics revealed that CDK2 and RB protein levels remained stable upon Meriolin 16 and Meriolin 36 treatment. Greater variations were observed for Cyclin A2 and Cyclin E protein levels upon treatment with Meriolin 16 and 36 (Fig. 5C). The phosphosites p-Thr821 (which gets phosphorylated in early S phase by CDK2/Cyclin E) and p-Ser612 of RB (which gets phosphorylated in late S phase by CDK2/Cyclin A2) were no longer phosphorylated upon Meriolin 16 treatment. This effect was not observed upon Meriolin 36 treatment. The application of the CDK1, 2 and 4 inhibitor R547 resulted in an increased protein level of CDK2, Cyclin A2 and Cyclin E. The effect on the phosphosites p-Thr821 and p-Ser612 was different from Meriolins. The phosphorylation at p-Thr821 was lost upon R547 treatment, whereas an increase in the phosphorylation presence of p-Ser612 could be observed (Fig. 5C). Inhibition of CDK2 by Meriolins (either in complex with Cyclin A2 or with Cyclin E) did not result in a decrease of the respective protein level. However, the downstream target suppressor protein RB was affected by this inhibition. In general, the phosphorylation at p-Ser612-RB enhances cell cycle progression in S phase [38] and the phosphorylation of RB results in the release of E2F family members and enables the transcription of crucial E2F-responsive genes for the S phase [39]. Since Meriolin 16 substantially reduced the phosphorylation at p-Thr821 and p-Ser612 of RB (Fig. 5C), it most likely would affect cell cycle progression. The results shown in Fig. 5 were further supported by immunopurification of the RB protein after Meriolin 16 treatment (0.1 and 1 µM) for 4 and 24 h (Supplemental Fig. 7). The phosphosites p-Ser612 and p-Thr821 of RB were less phosphorylated after 4 h upon 1 µM Meriolin 16 treatment and completely abrogated after 24 h (Supplemental Fig. 7).
Meriolins and other known CDK inhibitors induce cell cycle arrest and reduce proliferation at sublethal doses
Next, the effect of Meriolins and other CDK inhibitors on the cell cycle was investigated. For this, we used established CDK inhibitors (such as Dinaciclib, Flavopiridol, Meriolin 3, R547, Roscovitine, SNS-032, and Zotiraciclib), some of which are presently undergoing clinical trials. Meriolin 3 was included since it was so far the most potent Meriolin derivative [16, 17, 25] and has already been tested in preclinical trials [12, 21, 40]. Since CDK inhibitors can induce apoptosis, Ramos cells were pretreated with the pan-caspase inhibitor Q-VD-OPh (QVD). This procedure should enable the differentiation between caspase-dependent (apoptotic) and non-caspase-dependent effects on the DNA content and cell cycle. Subsequently, cells were incubated with a non-lethal dose (IC25 values) of the respective CDK inhibitors. The broad kinase inhibitor and potent apoptotic stimulus Staurosporine (STS) was used as positive control. The detection of the different cell cycle phases was performed by flow-cytometric analysis of the DNA content of propidium iodide-stained nuclei [41]. Though non-lethal dosages (IC25 values) were applied, the different compounds induced residual apoptosis as indicated by the formation of hypodiploid apoptotic nuclei, which was completely abrogated upon addition of the pan-caspase inhibitor QVD (Fig. 6A). However, even in the presence of QVD, Meriolin 16, Meriolin 36, and most of the CDK inhibitors displayed only a slight shift to G2 phase. Only R547 induced a pronounced G2 arrest. This observation was further analyzed by measuring the proliferative activity using BrdU assay. As shown in Fig. 6B, treatment of Ramos cells with non-lethal dosages (IC25) of Meriolin 16, Meriolin 36, Meriolin 3, Dinaciclib and R547 induced a substantial decrease in proliferation after 24 h – which was not affected by caspase inhibition via QVD (Fig. 6B). In addition to the BrdU assay, EdU incorporation was analyzed by microscopy in HeLa cells. The cells were treated for 24 h with non-lethal doses of Meriolin 16 and 36 and residual EdU was incorporation detected via immunofluorescence and quantified as shown in Fig. 6C,D. After 24 h treatment, the EdU incorporation and thus proliferation were reduced to 10%. Hence, Meriolin 16 and 36 induce a proliferation and DNA replication stop at non-lethal doses.
Meriolin 16 and Meriolin 36 impair CDK9-mediated downstream phosphorylation of the transcription regulator RNA polymerase II
Since Meriolin 16 and 36 also potently inhibited CDK9 (Fig. 3D,E), we investigated in how far the downstream signaling of CDK9 was affected by both Meriolin derivatives. CDK9 is not primarily involved in cell cycle regulation but in transcriptional control by activating RNA polymerase II (a schematic overview is provided in Fig. 7A). Therefore, we examined whether Meriolin 16 and 36 inhibited the CDK9/Cyclin T mediated phosphorylation of RNA polymerase II at the transcriptional crucial phosphosite p-Ser2. For this, we analyzed the protein levels of CDK9/Cyclin T1 and of phosphorylation of RNA polymerase II (at p-Ser2) upon treatment with Meriolin 16, 36 or R547 over time (4, 8, 12, 16 and 24 h) via immunoblotting (Fig. 7B,C and Supplemental Fig. 6C). Meriolin 16 induced a decrease of CDK9 and Cyclin T1 protein levels over 4, 8, 12, 16 and 24 h at low (0.1 µM) and high (1 µM) concentrations. In contrast, these protein levels were relatively stable for Meriolin 36 (Fig. 7B,C and Supplemental Fig. 6C). Both Meriolin 16 concentrations resulted in a total loss of the phosphorylation at Ser2 within 4 h. In contrast, Meriolin 36 and R547 mediated reduction of the phosphorylation at Ser2 was less pronounced (Fig. 7B,C and Supplemental Fig. 6C). Comparing both Meriolins, Meriolin 16 treatment had a higher impact on this phosphorylation than Meriolin 36.
It has been shown that the CDK9/Cyclin T1 induced phosphorylation at Ser2 within the C-terminal domain (CTD) of RNA polymerase II mediates the transition from transcription initiation to elongation [42–44]. Since Meriolin 16 completely abrogated the transcription-initiating phosphorylation at Ser2 of RNA polymerase II (Fig. 7B,C and Supplemental Fig. 6C), it would consequently inhibit the transcriptional activity. Therefore, we measured the de novo RNA-synthesis by the incorporation of 5-ethynyl-uridin (EU) in HeLa cells treated with the respective IC25 values for Meriolin 16 (0.04 µM) and Meriolin 36 (0.4 µM) for 24 h. As shown in the immunofluorescence analysis in Fig. 7D,E, EU incorporation was reduced to ~ 60% by Meriolin 16 and to ~ 50% by Meriolin 36. Thus, both Meriolin derivatives impair de novo RNA-synthesis and transcription at non-lethal concentrations.
Meriolins are highly cytotoxic compared to other CDK inhibitors
Finally, we compared the cytotoxic potential of Meriolin 16 and 36 to known CDK inhibitors like Roscovitine, Flavopiridol, R547, Meriolin 3, Zotiraciclib, Dinaciclib and SNS-032. These CDK inhibitors were chosen based on their clinical development or therapeutic usage and CDK inhibition profile which are summarized in Supplemental Table 2. Meriolin 3 has been synthesized by others and included since it represents the most potent Meriolin derivative described so far [16, 17, 25]. For this, Ramos lymphoma cells were treated with Meriolins or other CDK inhibitors at increasing concentrations for 24 h and the cell viability was determined by AlamarBlue® assay. As shown in Fig. 8, Meriolin 16 was the most active derivative amongst all Meriolins tested – even more potent than Meriolin 3. With an IC50 value of 30 nM, Meriolin 16 was even in the range of Dinaciclib (IC50 value: 10 nM) and both compounds were by far more cytotoxic than all other CDK inhibitors tested Fig. 8B,C. Thus, CDK inhibition represents a promising approach for targeting B cell lymphoma.
Finally, we assessed the applicability of our Meriolin derivatives as drug candidates and performed ADME (absorption, distribution, metabolism and excretion) predictions using the Swiss-ADME tool [45]. The results are summarized in Supplemental Table 3. Thus, both compounds show a combination of lipophilicity and solubility that should allow gastrointestinal absorption and possess a drug-like scaffold suitable for further lead optimization processes.