CK2 (casein kinase 2) is a very pleiotropic serine/threonine protein kinase whose abnormally high constitutive activity has often been correlated to pathological conditions with special reference to neoplasia. The two most widely used cell permeable CK2 inhibitors, TBB (4,5,6,7-tetrabromo-1H-benzotriazole) and DMAT (2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole), are marketed as quite specific CK2 blockers. In the present study we show, by using a panel of approx. 80 protein kinases, that DMAT and its parent compound TBI (or TBBz; 4,5,6,7-tetrabromo-1H-benzimidazole) are potent inhibitors of several other kinases, with special reference to PIM (provirus integration site for Moloney murine leukaemia virus)1, PIM2, PIM3, PKD1 (protein kinase D1), HIPK2 (homeodomain-interacting protein kinase 2) and DYRK1a (dual-specificity tyrosine-phosphorylated and -regulated kinase 1a). In contrast, TBB is significantly more selective toward CK2, although it also inhibits PIM1 and PIM3. In an attempt to improve selectivity towards CK2 a library of 68 TBB/TBI-related compounds have been tested for their ability to discriminate between CK2, PIM1, HIPK2 and DYRK1a, ending up with seven compounds whose efficacy toward CK2 is markedly higher than that toward the second most inhibited kinase. Two of these, K64 (3,4,5,6,7-pentabromo-1H-indazole) and K66 (1-carboxymethyl-2-dimethylamino-4,5,6,7-tetrabromo-benzimidazole), display an overall selectivity much higher than TBB and DMAT when tested on a panel of 80 kinases and display similar efficacy as inducers of apoptosis.
- ATP mimetics
- casein kinase 2 (CK2) inhibitor
- drug design
- homeodomain-interacting protein kinase 2 (HIPK2)
- provirus integration site for Moloney murine leukaemia virus 1 (PIM1)
CK2 (an acronym derived from the misnomer ‘casein kinase’ 2) denotes one of the most pleiotropic serine/threonine protein kinases, with more than 300 protein substrates already identified  and responsible for the generation of a substantial proportion of the eukaryotic phosphoproteome . Such an extreme pleiotropy may, at least partially, account for the observation that CK2 catalytic activity is not controlled by second messengers, phosphorylation events or association of its catalytic (α and/or α′) domain with its regulatory β-subunits to give a ‘constitutively active’ heterotetrameric holoenzyme [3,4]. In turn constitutive activity and lack of physiological down-regulatory devices may underlie the pathogenic potential of CK2, whose abnormally elevated levels have been correlated to a number of global diseases, with special reference to neoplasia. Abnormally high levels of CK2 have been documented in a number of cancers including kidney , mammary gland , lung , head and neck  and prostate , and a number of experimental models and coincidental arguments support the notion that CK2 promotes cell survival through the regulation of oncogenes and plays a global anti-apoptotic role (reviewed by Duncan and Litchfield ).
Studies aimed at dissecting the multifarious functions of CK2 are hampered by its constitutive activity and limitations in strategies exploited to inhibit the cellular activity of other kinases, notably usage of kinase-inactive dominant-negative mutants and RNAi (RNA interference) techniques. These techniques are hindered by long half-life, cellular localization and high expression of endogenous CK2 in cells . Indeed conflicting results regarding the effectiveness of down-regulation devices (e.g. compare Wang et al.  with Seeber et al. ) have been reported, suggesting that RNAi or antisense strategies sometimes may be inefficient, requiring continual treatment of cells in order to achieve effective CK2 depletion.
These limitations make the pharmacological approach the most promising tool to dissect the cellular functions of CK2 and explain why considerable efforts have been made to develop specific cell-permeable CK2 inhibitors (reviewed in Sarno et al. ). Some of these inhibitors have proved to be extremely helpful for the identification of CK2 cellular targets and the implication of CK2 in specific cellular functions, as well as in malignancy and defective apoptosis of cancer cells [10,13,14]. In a few cases the concept that the observed pharmacological effect was actually mediated by CK2 inhibition was validated by transfecting the cells with CK2 mutants refractory to inhibitors [15,16]. Generally, however, the implication of CK2 was inferred from the assumption that the compounds used are ‘specific’ inhibitors of this kinase. In this respect the most successful commercially available ‘specific’ CK2 inhibitors are: (i) TBB (4,5,6,7-tetrabromo-1H-benzotriazole) also termed TBBt to distinguish it from its analogue TBI (or TBBz; 4,5,6,7-tetrabromo-1H-benzimidazole), also a potent inhibitor of CK2 , and (ii) DMAT (2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole). The specificities of TBB and DMAT, although not that of TBI, have been assessed on a panel of approx. 30 protein kinases, none of which was inhibited as drastically as CK2 [18,19]. In contrast, other commercially available CK2 inhibitors, notably quercetin, apigenin and emodin, did not display a comparable selectivity [20,21]. Recently, however, the panel of protein kinases currently available for running selectivity profiles has been substantially enlarged, to include approx. 80 kinases representative of nearly all of the branches of the kinome (see Figure 1). This has provided a tool for updating the selectivity of kinase inhibitors and has led to the discovery of ‘new’ targets for ‘old’ inhibitors . This latter study  did not include any CK2 ‘specific inhibitor’. On the other hand the solution of the crystal structure of the complex between CK2 and TBI/TBBz  has allowed a comparison with the structures of the complexes with TBB  and with DMAT  leading to the conclusion that, despite their structural similarities, TBB binds in a quite different manner than TBI and DMAT. Taken together these observations prompted us to profile the selectivity of TBB, DMAT and TBI on the new large panel of protein kinases now available. The results reported in the present study support the view that the selectivity of TBB for CK2 is narrower than that of DMAT and TBI/TBBz, which are also powerful inhibitors of kinases belonging to the PIM (provirus integration site for Moloney murine leukaemia virus), HIPK (homeodomain-interacting protein kinase), DYRK (dual-specificity tyrosine-phosphorylated and -regulated kinase) and PKD (protein kinase D) families, and led to the identification of new and more selective inhibitors of CK2.
Native CK2 was purified from rat liver . Human recombinant α and β subunits of CK2 were expressed in Escherichia coli and the holoenzyme was reconstituted and purified as previously described . Single and double mutants of the CK2α subunit were generated as reported in [19,20]. The source of all of the other protein kinases used for specificity assays are as previously either described or referenced .
Cell culture, treatment and viability assay
The human leukaemia Jurkat T-cell line was maintained in RPMI-1640 medium, supplemented with 10% (v/v) foetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin. For the treatment, cells were suspended at a density of 106 cells/ml in a medium containing 1% (v/v) foetal calf serum, then incubated at 37 °C in the presence of the compounds at the indicated concentrations. Control cells were treated with equal amounts of solvent (0.5% DMSO). At the end of incubations, cells were lysed by the addition of hypo-osmotic buffer, as previously described . Cell viability was assessed by means of MTT [-(4,5-dimethylthiazol-2-yl)-3,5-diphenyltriazolium bromide] reagent, whereas apoptosis was evaluated by means of the Cell Detection Elisa kit (Roche), based on the quantification of nucleosomes present in the cytosol of the apoptotic cells, following the manufacturer's protocol. For each determination, 10000 cells were used.
CK2 activity was tested in a final volume of 25 μl containing 50 mM Tris/HCl (pH 7.5), 100 mM NaCl, 12 mM MgCl2, 100 μM synthetic peptide substrate RRRADDSDDDDD and 0.02 mM [γ-33P]ATP (500–1000 c.p.m./pmol), unless otherwise indicated, and incubated for 10 min at 37 °C. Native CK2 purified from rat liver (0.5–1 pmol) was usually the phosphorylating enzyme, unless otherwise indicated. Assays were stopped by the addition of 5 μl of 0.5 M orthophosphoric acid before spotting aliquots on to phosphocellulose filters. Filters were washed in 75 mM phosphoric acid (5–10 ml each) four times, followed by once in methanol and were dried before counting.
PIM1 activity was determined following the same procedure, by incubating the kinase in the presence of 50 mM Tris/HCl (pH 7.5), 0.1% 2-mercaptoethanol, 0.1 mM EGTA, 30 μM synthetic peptide substrate RKRRQTSMTD  and 100 μM [γ-33P]ATP. HIPK2 activity was determined under the same conditions used for PIM1 assays, except for ATP which was 20 μM and the use of 10 μg of MBP (myelin basic protein) as phosphorylatable substrate.
Conditions for the activity assays of all other protein kinases tested in selectivity experiments are as described or referenced previously .
Initial velocities were determined at each of the substrate concentrations tested. Km values were calculated from Lineweaver–Burk double-reciprocal plots of the data either in the absence or in the presence of increasing concentrations of inhibitor. Inhibition constants were then calculated by linear regression analysis of Km/Vmax versus inhibitor concentration plots. Alternatively, inhibition constants were also deduced from the IC50/Ki Cheng–Prusoff relationship  by determining the IC50 for each compound at a 1 μM ATP concentration, i.e. much below the Km value. The rationale underlying this approach, based on the assumption that inhibition is competitive with respect to ATP, is illustrated by Burlingham and Widlanski .
The human CK2α subunit and PIM1 catalytic subunit were retrieved from the PDB (PDB codes 1JWH and 3C4E respectively) and processed in order to remove the ligands and water molecules. Hydrogen atoms were added to the protein structure using standard geometries with the MOE program. To minimize contacts between hydrogens, the structures were subjected to Amber94 force-field minimization until the rms (root mean square) of conjugate gradient was <0.1 kcal·mol−1·Å−1 (1 Å = 0.1 nm) keeping the heavy atoms fixed at their crystallographic positions. To strictly validate the model generated and to calibrate the high-throughput docking protocol, a small database of known CK2 and PIM1 inhibitors was built and a set of docking runs was performed. After the calibration phase, flexible ligand-docking steps with three different programs, MOE-Dock, Glide and Gold, were performed essentially as described previously .
Synthesis of new analogues
All chemicals and solvents used for the syntheses were purchased from Sigma–Aldrich. Details for the synthesis of most benzimidazoles and benzotriazoles have been described previously [19,25,31,32]. New derivatives discussed in the present paper have been synthesized as described below.
The synthesis of 4,5,6,7-tetrabromo-1H-indazole (1, K59) was realized by the ring closure of the intermediate 3,4,5,6-tetrabromo-2-methyl-N-nitroso-N-acetylaniline . The bromination of 1 was performed using bromine in anhydrous acetic acid, at elevated temperature, yielding the pentabromoderivative 4 (K64). Alkylation of 1 with 2-bromoethanol and 2-dimethylaminoethylchloride provided, as expected, two regioisomers each, 2a and 2b (K60 and K62) and 3a and 3b (K61 and K63) respectively. The mixtures of N1- and N2-substituted derivatives were separated by flash chromatography and the structures of the derivatives were determined by comparing their 1H NMR- and UV-spectra with those of the respective known N-substituted indazoles [34,35]. Dissociation constants of 1 (10.6) and 4 (8.40) measured in 50% aqueous ethanol showed the inductive effect of bromine substituents by lowering the respective pKa value by several units as compared with that of unsubstituted indazole (14.0). A similar effect was observed for TBB (pKa = 4.95) and 4,5,6,7-tetrabromobenzimidazole (pKa=8.95) . In a similar manner to tetrabromobenzimidazole, the two brominated indazoles 1 and 4 described here are practically neutral molecules at physiological pH (Scheme 1).
The synthesis of 1-substituted acetic acid ethyl ester of 2-dimethylamino-4,5,6,7-tetrabromobenzimidazole 6 (K67) was performed by N-alkylation of DMAT (5) with bromoacetic acid ethyl ester in acetone in the presence of anhydrous potassium carbonate as base. The alkaline hydrolysis of 6 in ethanol/water solution gave the N1-carboxymethyl derivative of DMAT (7, K66) (Scheme 2).
The other modified tetrabromobenzimidazoles were obtained from 1-methyl-4,5,6,7-tetrabromobenzimidazole (8) as substrate. The heating of 8 with thiourea in propan-1-ol gave the respective 2-mercapto derivative 9 (K72). Alkylation of 9 with bromoacetic acid ethyl ester yielded 2-sulfanylacetic acid ethyl ester 10, which was substrate for further modifications. Aminolysis of 10 with methylamine gave the respective N-methylamide (11, K74), whereas alkaline hydrolysis yielded 1-methyl-4,5,6,7-tetrabromobenzimidazole-2-sulfanylacetic acid (12, K75). (Scheme 3).
A detailed description of the synthesis and analytical data for the newly obtained compounds discussed in the present study is shown in Supplementary Table S1 at http://www.BiochemJ.org/bj/415/bj4150353add.htm. The profiling of a library of compounds on CK2, PIM1 and HIPK2 is shown in Supplementary Table S2 at http://www.BiochemJ.org/bj/415/bj4150353add.htm.
Re-evaluation of TBB, DMAT and TBI selectivity
TBB and its 2-dimethylamino derivative DMAT are among the most widely used commercially available inhibitors of CK2. In contrast with other compounds used to inhibit CK2, notably DRB (5,6-dichloro-1-β-D-ribofuranosylbenzimidazole) and apigenin, both TBB and DMAT are believed to be very specific based on selectivity profiles run on panels of 30 and 33 protein kinases respectively [18,19]. More recently, however, the size of the panel of kinases available for testing the specificity of inhibitors has more than doubled , so that currently this includes representatives of nearly all subfamilies of the kinome (Figure 1). This prompted us to re-evaluate the selectivity of TBB and DMAT toward CK2. Since DMAT shares its benzimidazole scaffold with the parent compound TBI (or TBBz) (see Figure 2) which is also marketed as a CK2 inhibitor (although its selectivity has not been profiled before), TBI was also included in the analysis in the present study.
The residual activity of a panel of 76 protein kinases tested in the presence of 10 μM TBB, DMAT and TBI and expressed as the percentage of their activity in the absence of inhibitors is reported in Table 1 or in the legend of Table 1. Even a cursory scrutiny of the data reveals that CK2 is not the only protein kinase drastically affected, several other kinases (values shown in bold in Table 1) are more than 90% inhibited by one or more of the three compounds. A more detailed analysis allows the following conclusions to be reached. (i) Although the selectivity spectra of TBB, DMAT and TBI are similar they are not super-imposable, denoting higher similarity between DMAT and TBI (characterized by an imidazole ring), both of which also tend to be more promiscuous than TBB (composed instead of a triazole ring). This point is highlighted by drastic inhibition of PKD1 and CDK (cyclin-dependent kinase) 2/cyclin A by either DMAT or TBI, in contrast with fairly high residual activity observed in the presence of the same concentration of TBB. (ii) There are three subfamilies of protein kinases, none of which was represented in the 30 protein kinase panel originally used to define the TBB specificity , whose components tend to be inhibited as drastically as CK2 by all of the three compounds at a 10 μM concentration, with residual activities below 10%; these are DYRKs (1a, 2 and 3), PIMs (1, 2, 3) and HIPK2. A >90% inhibition of HIPK3 is also observed by DMAT and TBI, whereas TBB is significantly less effective in this respect.
More detailed information regarding the relative efficiency of each inhibitor toward protein kinases whose activity is 90% or more inhibited by a 10 μM concentration of inhibitor was obtained by determining their IC50 values at 50 μM ATP. These results are reported in Table 2 and they strengthen the notion that among the three inhibitors the most selective inhibitor towards CK2 was TBB with an IC50 value (0.15 μM) 3–5-fold lower than those of DYRK2, PIM1 and PIM3, and one order of magnitude or more below those of the other kinases. In contrast, TBI appears to be the least selective toward CK2, as it inhibits PIM3, PIM1, PIM2, DYRK2 and PDK1 more efficiently than CK2 itself, and HIPK2 and CDK2 with comparable efficiency. This is partially also due to its relatively high IC50 value with CK2 (0.60 μM).
The overall superiority of TBB over DMAT is also highlighted by selectivity profiles performed with a 1 μM instead of 10 μM inhibitor concentration (last two columns on the right in Table 1). Under these conditions the number of protein kinases inhibited by ≥50% are three and 12 with TBB and DMAT respectively (values shown as underlined in Table 1). Note that seven out of the 12 protein kinases that are ≥50% inhibited by DMAT are inhibited as drastically as CK2 itself.
DMAT as a powerful inhibitor of kinases other than CK2
Although the IC50 value with CK2 is quite low (0.13 μM), DMAT displays submicromolar IC50 values with almost all of the other kinases in Table 2, with special reference to PKD1, PIM3 and PIM1 whose IC50 value (148 nM) is comparable with that of the most potent PIM1 inhibitor [BIM-1 (bisindolylmaleimide)] reported to date .
In a similar manner to CK2 , PIM1 is also inhibited by DMAT by a mechanism which is competitive with respect to ATP, as shown by the double reciprocal plot shown in Figure 3(A). However, IC50 determinations at increasing ATP concentration denote weak competition by ATP which, at almost physiological concentration (0.6 mM), causes only a 5.3-fold decrease in DMAT inhibition, as compared with 1 μM ATP concentration, whereas in the same range of ATP concentration the IC50 with CK2 increases 22.1-fold, doubling the value calculated with PIM1 (1.2 μM) (Figure 3B). Therefore it is expected that in the cellular environment (where ATP ≥1 mM), PIM1 will represent the preferred target of DMAT, among all of the protein kinases probed in this respect.
Improvement of selectivity toward CK2
The observation that CK2 inhibitors tend to inhibit with comparable efficacy at least three other groups of kinases, notably PIMs, HIPK2 and DYRKs, prompted us to search for more selective compounds in the available library of TBB/TBI derivatives. This presently includes, besides several dozens of tetrabromobenzimidazole congeners previously described [19,32], also a number of newly developed compounds with a pyrazole penta-atomic ring replacing for the imidazole ring, whose prototype K59 is shown in Figure 2 together with its triazole (TBB) and imidazole (TBI) congeners. All of the compounds of the library (68 altogether) were tested at a 20 μM concentration on CK2, PIM1 and HIPK2. The assays were run at ATP concentrations at approx. the Km values of the kinases (20 μM for CK2 and HIPK2, 100 μM for PIM1) and the residual activities are shown in Supplementary Table S2.
Focusing our attention only on compounds highly effective on CK2 (residual activity ≤15%, shown in bold in Supplementary Table S2) we observed that most of these are equally, or even more, potent inhibitors of PIM1, and two actually inhibit HIPK2 even more efficiently than CK2. A few compounds, however, namely K36, K59, K60, K62, K64, K66, K74 and K75, whose structures are shown in Table 3, displayed an inhibitory efficacy significantly higher towards CK2 than those observed with either PIM1 or HIPK2.
The Ki values of these compounds with respect to CK2, PIM1, HIPK2 and DYRK1a were calculated (Table 3). In all cases CK2 displayed the lowest Ki value as compared with the other three protein kinases. In particular two compounds, K59 and K64 displayed Ki values below 100 nM with CK2, whereas their Ki values with the second most sensitive protein kinase, which is PIM1 in both cases, were several-fold higher. Both K59 and K64 are characterized by a pyrazole ring replaced for the triazole/imidazole rings of TBB and DMAT. K64 also contains a fifth bromine atom on the pyrazole ring in addition to the four bromine atoms on the benzene ring which are a common denominator of all TBB derivatives described so far. Three other promising compounds belong to the benzimidazole class: these are K66, K74 and K75. In this case the Ki calculated with CK2 was somewhat higher than 100 nM: this disadvantage however is amply compensated by proportionally higher Ki values with PIM1, HIPK2 and DYRK1a. We focused further attention on those compounds whose Ki values with CK2 were 10-fold or more lower than those with the second most inhibited kinase, notably K64, K66, K74 and K75.
We first investigated whether, and to what extent, the double mutation of CK2 residues Val66 and Ile174 to alanine residues, previously reported to decrease susceptibility to TBB and DMAT [19,20], also affected inhibition by the new compounds. This is clearly the case, as shown in Table 4; however, responsiveness to the double mutation was sharply variable: two inhibitors, K74 and K75, were somewhat more sensitive than TBB and DMAT to the double mutation, whereas K66 was almost insensitive to it, with a just 6-fold increase in IC50 value. By sharp contrast the pentabrominated inhibitor K64 had lost almost all of its efficacy toward the CK2 double mutant, with a 600-fold increase in IC50. These differences clearly reflect diverse modes of binding of these compounds, as further indicated by the variable efficacy of the two single mutations, that of Val66 and of Ile174. In particular the role of Ile174 is crucial for the binding of K64, whereas the binding of K66 relies more on interactions with Val66 as also confirmed by modelling (see Discussion below).
The behaviour of K66 attracted our attention because its scaffold is identical with that of DMAT, having in addition a short carboxylic side chain adjacent to the imidazole ring nitrogen (compare structures in Figure 2). Although this carboxylic adduct is still compatible with fairly potent inhibition of CK2 it makes its binding almost independent of Ile174, whereas it is more dependent on Val66 (Table 4), and, more importantly, it drastically reduces the ability of K66 to inhibit PIM1 and HIPK2 which are instead dramatically sensitive to DMAT (see Table 2).
Once we ascertained that the new CK2 inhibitors considered in Table 3, with special reference to K64, K66 and K75, had lost most of their potency toward those protein kinases which display the tendency of being susceptible to CK2 inhibitors, notably PIM1, HIPK2 and DYRK1a, we wanted to make sure that their overall selectivity was preserved. This analysis was performed on a panel of 78 protein kinases almost completely, but not entirely, overlapping the panel of 70 protein kinases used for the screening presented in Table 1. The results are shown in Figure 4. For the sake of comparison, the results relative to DMAT are reported. The most remarkable outcome is indeed the dramatic gain of selectivity resulting from the derivatization of DMAT with a carboxymethyl side chain to give K66 (Figure 4), whereas, as noted above, DMAT under these conditions inhibits by ≥50% 34 kinases out of 70 (i.e. almost 50%), this figure with K66 drops to just five kinases out of 78. Only two of these, not including CK2, are drastically inhibited (≥80%) by K66: ERK (extracellular-signal-regulated kinase) 8 and DYRK2. However, the IC50 values for inhibition of DYRK2 is far above that of CK2 (8 μM as compared with 0.25, see the legend of Figure 4). Consequently the only kinase whose inhibition by K66 is comparable with that of CK2 is ERK8 (IC50=0.50 μM), thus placing K66 in the category of ‘dual-kinase inhibitors’ able to block with equal efficiency two unrelated kinases. Note in this respect that it would be easy to discriminate between ERK8 and CK2 using the inhibitor BAY439006 which at 10 μM concentration suppresses ERK8 activity without significantly affecting CK2 .
The overall selectivity of K64 (Figure 4) is also good, with six protein kinases (including CK2) inhibited by ≥50% with 10 μM of inhibitor. One of these is again ERK8, but in this case it is inhibited significantly less than CK2. The same applies to PIM2, PIM3, HIPK2 and PKD1, whose IC50 values accordingly are at least one order of magnitude higher than that of CK2.
Owing to solubility problems, the selectivity of K75 was profiled at a 1 μM concentration (Figure 4). Under these conditions only three kinases were inhibited by ≥50%: CK2 and PIM3 (both 70% inhibited), and MSK1 (mitogen- and stress-activated protein kinase 1; 54% inhibited).
K74 turned out to be unstable when kept for >2 h in DMSO solution, a circumstance which prevented its robotic profiling on the panel of protein kinases and probably also seriously hampers its usage in cell experimentation.
Pro-apoptotic efficacy of K64 and K66
In light of the results presented above the first-choice novel CK2 inhibitors worthy to be tested on cell cultures were K64 and K66, whose selectivity is comparable. The former is somewhat more potent but the latter has the advantage that its only known target inhibited as drastically as CK2, ERK8, is absent in haematopoietic cells , which are those most frequently used for testing the pro-survival and oncogenic potential of CK2 (e.g. [16,38–40]). An added value of K64 is instead its almost complete loss of efficacy toward the CK2 double mutant V66A/I174A (see Table 4), a property which can be exploited to provide the proof-of-concept that the cellular effects of a given inhibitor are actually mediated by CK2 rather than by other cellular targets.
As shown in Figure 5(A) both K64 and K66 displayed cytotoxic activity comparable with that of TBB. Interestingly the cytotoxic efficacy of the three compounds approx. correlates to their potency as CK2 inhibitors calculated from their Ki values (0.048, 0.12 and 0.25 for K64, TBB and K66 respectively). It should also be noted that although in principle the cytotoxic effect of TBB could be accounted for by the inhibition of protein kinases of the PIM and DYRK families having IC50 values close to that of CK2 (Table 2), this does not apply to K64 and K66 (see above). As shown in Figure 5(B) the cytotoxic effect of K64, similar to that of TBB  and of K66 is, at least partially, due to apoptosis as judged from the quantification of nucleosomes present in the cytosol of the apoptotic cells: nucleosomes in cells treated with 5 μM K64 are comparable with those induced by 25 μM etoposide.
Considering that the human kinome is composed of more than 500 members whose catalytic site is highly conserved and that the most promising cell-permeable protein kinase inhibitors are competitive with respect to the universal kinase co-substrate ATP, it is understandable that the development of really specific protein kinase inhibitors is a formidable challenge. Although absolute selectivity may be not required for a kinase inhibitor to become a valuable drug and some promiscuity could even be an advantage in terms of therapeutic efficacy, as exemplified by clinical practice with Imatinib/Gleevec , lack of specificity is always a serious drawback whenever inhibitors are used to dissect the biological functions of a given protein kinase, where the unwanted blockage of other kinases may generate misleading results. This shortcoming can be attenuated by using more than one inhibitor of the same kinase and can be circumvented by the generation and cell transfection of kinase mutants which are refractory to a given inhibitor. Since, however, clusters of kinases often tend to be inhibited by the same classes of inhibitors and cell transfection is not always feasible and is likely to alter cell functionality anyway, the availability of pharmacological tools that are as selective as possible invariably represents a critical advantage for dissecting signalling pathways. This is especially true in the case of protein kinases such as CK2 which are constitutively active and cannot be easily up- or down-regulated within the cell using physiological effectors.
In the past considerable efforts have been made to develop CK2-specific inhibitors, belonging to different chemical classes (reviewed by Sarno et al. [13,14]). Apart from apigenin, a flavonoid devoid of narrow selectivity [20,42], the two most successful CK2 inhibitors were both derived from the nucleoside DRB, a rather weak CK2 inhibitor , by deleting its sugar moiety and replacing the chlorine and hydrogen atoms of its benzene ring with four bromine atoms. These were shown to be critical to encapsulate the inhibitor in a hydrophobic cavity which in CK2 is smaller than in the majority of other protein kinases [24,25]. This feature, which underlies the fairly narrow selectivity of these compounds toward CK2, has been exploited in many recent reports to probe the implication of this kinase in a variety of cellular functions . However, despite their structural similarity TBB and DMAT bind in different ways into the active site of CK2 , denoting a role of the negative charge present at neutral pH on the triazole ring of TBB, which is not present in DMAT or in the other TBB derivatives [17,23]. In the latter the N2 atom is replaced by a carbon, to give a tetrabromobenzimidazole derivative (TBI or TBBz) devoid of the negative charge at neutral pH . The relevance of this difference to determine kinase selectivity was overlooked in our original studies because the panel of kinases used at that time for selectivity profiles was not large enough to disclose significant differences. It was concluded therefore that, apart from some minor differences, TBB and DMAT exhibited quite super-imposable inhibition profiles and a similar, quite remarkable, selectivity toward CK2, although DMAT is somewhat more potent, especially if used in cell culture . Another advantage of DMAT, as well as of TBI and the other benzimidazole derivatives analysed in this respect  is that, owing to its high pKa value (8.9 for TBI and 10.6 for DMAT), at physiological pH the N-H of the imidazole ring is not dissociable and therefore these compounds are unable to depolarize mitochondria  by shuttling protons across the membrane. The same applies to the pyrazole ring of the indazole derivatives, e.g. K59 (pKa=10.6) and K64 (pKa=8.40) (see the Experimental section). In contrast, mitochondria depolarization is a drawback of TBB  whose triazole ring exhibits a much lower pH value, allowing proton shuttling under the effect of the pH gradient on the two sides of the mitochondrial membrane.
In the present study, exploiting a panel of kinases more than twice as large as the original one, and including members of nearly all of the kinome subfamily (Figure 1), we have reached the conclusion that the selectivity of TBB is definitely narrower than that of DMAT and TBI. Although in fact both DMAT and TBI drastically inhibit several kinases in addition to CK2, notably PIM1, 2 and 3, DYRK1a, 2 and 3, PKD1, HIPK2 and 3, and CDK/CyclinA, TBB is significantly less effective on these kinases than it is on CK2 as highlighted by comparing both IC50 values (Table 2) and selectivity profiles of TBB and DMAT performed at a 1 μM inhibitor concentration (see Table 1). On the other hand, the possibility should be considered that besides protein kinases non-kinase enzymes also interact with CK2 inhibitors: this applies to TBB, but, even more to TBI and DMAT .
The results of the present study provide valuable hints for the development of PIM and HIPK inhibitors more potent and/or selective than those currently available. In the case of PIM1, the product of a proto-oncogene whose dysfunction has been implicated in diverse cancers with special reference to DLCL (diffuse large cell lymphoma) , some inhibitors are available , whose potency is comparable with those of TBI and DMAT and whose selectivity has not been explored. Our results also provide the possibility to discriminate among PIM isoforms exploiting their different susceptibilities to TBI and DMAT (see Table 2). In contrast, inhibitors of HIPK2, a kinase required for normal haematopoiesis and blood vessel formation  which plays a critical role in triggering p53-dependent apoptosis in response to the anti-neoplastic drug cisplatin , are not available at all to the best of our knowledge, a circumstance which makes DMAT and TBI the best tools presently available to down-regulate HIPK2.
Once established that TBB and, even more, DMAT and TBI are CK2 inhibitors less selective than it was previously thought, our next aim was the identification of DMAT/TBB derivatives which are still effective on CK2 having lost most of their efficacy toward PIM1, HIPK2 and DYRK1a. This goal turned out to be hard to attain since most compounds tested, which are potent inhibitors of CK2, also tend to inhibit PIM1 and/or HIPK2 with similar potency (see Supplementary Table S2). A few notable exceptions, however, were provided by two compounds belonging to the class of indazoles (K59 and K64), having a pyrazole ring replaced for the triazole and imidazole pentatomic rings present in TBB and DMAT respectively, and three new benzimidazole derivatives (K66, K74 and K75). These are all characterized by the ability to inhibit CK2 more potently than PIM1, HIPK2 and DYRK1a, with Ki values which in the case of K64 and K66 are many-fold lower than those calculated with the second most inhibited kinase (PIM1 in both cases)
In this respect, the behaviour of K66 appears to be especially remarkable, considering its close structural relatedness with DMAT: while DMAT inhibits PIM1 with the same potency as CK2, K66 has lost most of its efficacy toward PIM1, as judged from big differences in Ki and IC50 values.
The structural similarity between DMAT and K66, in conjunction with the availability of three-dimensional structures of both PIM1 and CK2, the latter also in complex with DMAT , provide a key to rationalize the preferential binding of K66 to CK2. The structure of DMAT bound into the active site of CK2, together with the derived model in which DMAT has been replaced by K66 are shown in Figures 6(A) and 6(C). This accounts very well for the finding that Ile174 is less important for the binding of K66 (see Table 4). Note in fact that K66, owing to its carboxylic group, is making electrostatic interactions with the positive area close to Lys68 thus displacing the dimethylamino group from its apolar contacts with Ile174 and placing it closer to Val66. In the case of PIM1, DMAT is rotated approx. 60° with respect to the crystallographic data, owing to the presence of Phe49. In fact, unlike its CK2 homologue Tyr50 which is exposed toward the solvent, Phe49 is instead oriented toward the hydrophobic cavity forcing the inhibitor to a new pose. As shown in Figure 6(B) the dimethylamino group at position 2 is placed inside a new favourable hydrophobic pocket formed by PIM1 residues Val52, Leu120, Ile185 and Phe49, whereas in CK2 it is simply oriented towards the outside (Figure 6A). These structural observations may explain the similar activity of DMAT in CK2 and PIM1, even in the presence of Val66/Ala65 natural substitution. On the other hand, K66 interacts with the PIM1-binding cleft in the same way as shown for DMAT due to the presence of the dimethylamino group. Under these conditions its carboxy side chain at position 1 cannot interact with Lys67, homologous with CK2 Lys68, and therefore it is placed in an unfavourable hydrophobic zone (Figure 6D).
Importantly the improved selectivity of K66 as compared with DMAT is not restricted to its ability to discriminate between CK2 and PIM1, but it also reflects a reduced or abolished inhibitory efficiency towards a large number of kinases deeply affected by DMAT, as outlined by comparing their specificity profiles in Figure 4: although 34 kinases are inhibited by ≥50% at 10 μM DMAT, this figure drops to just five kinases with K66, under identical conditions. Only one of the latter (ERK8) is inhibited as drastically as CK2 (i.e. >80%), whereas 12 of those are >80% inhibited by DMAT.
In an attempt to quantify these differences a ‘promiscuity score’ has been devised, expressing the ‘average’ inhibition of all of the kinases of the whole panel, by a given concentration of the inhibitor, sufficient to drastically inhibit (approx. 90%) the kinase considered (in our case CK2) (see the legend of Table 5). Such a score, which reflects the promiscuity of the kinase and is inversely proportional to its ‘average’ selectivity, equals 18.36 and 46.66 with 10 μM K66 and DMAT respectively. Interestingly a promiscuity score similar to that of K66 has been calculated for K64 (16.8) reinforcing the concept that both K64 and K66 are markedly more selective than the three related commercially available CK2 inhibitors, TBB, DMAT and TBI, the latter having a score >50 (see Table 5). It should be also noted that, while at a 10 μM concentration of inhibitor the promiscuity scores of TBB and DMAT are similar, if the profiling is performed at a 1 μM concentration, where CK2 is approx. 80% inhibited by either TBB or DMAT (see Table 1), the promiscuity score of DMAT (23.1) doubles that of TBB (10.18) confirming also by this criterion the concept that TBB is more selective than DMAT. However, comparing the selectivity of TBB with those of K64 and K66 whose selectivity is much higher, and considering that the latter are equally effective as TBB as inducers of apoptosis it is predictable that K64 and K66 will replace TBB (although not DMAT and TBI) as the first choice tool for studying cellular functions of CK2. Since the chemical names of K64 and K66 are exceedingly long (3,4,5,6,7-pentabromo-1H-indazole and 4,5,6,7-tetrabromo-2-methylamino-1-carboxymethyl-benzimidazole respectively) we propose that, by analogy to TBB and DMAT, they are denoted by the acronyms PBIN and TMCB.
We thank Professor Sir Philip Cohen (MRC Protein Phosphorylation Unit, University of Dundee, Dundee, Scotland, U.K.) for critical reading of the manuscript. The Molecular Modelling Section co-ordinated by Professor Stefano Moro (University of Padova, Padova, Italy) is gratefully acknowledged. The work was supported by grants to L.A.P. from AIRC (Associazione Italiana per la Ricerca sul Cancro) and European Commission (PRO-KINASERESEARCH 503467) and by a grant to Z.K. and A.O. from the Ministry of Education and Science PBZ-MIN 014/P05/2004 (Poland).
Abbreviations: CDK, cyclin-dependent kinase; CK2, casein kinase 2; DMAT, 2-dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole; DRB, 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole; DYRK, dual-specificity tyrosine-phosphorylated and -regulated kinase; ERK, extracellular-signal-regulated kinase; HIPK, homeodomain-interacting protein kinase; MSK, mitogen- and stress-activated protein kinase; MTT, 3-(4,5-dimethylthiazol-2-yl)-3,5-diphenyltriazolium bromide; PKD, protein kinase D; PIM, provirus integration site for Moloney murine leukaemia virus; RNAi, RNA interference; TBB, 4,5,6,7-tetrabromo-1H-benzotriazole; TBI (or TBBz), 4,5,6,7-tetrabromo-1H-benzimidazole
- © The Authors Journal compilation © 2008 Biochemical Society