Discovery and Exploration of Protein Kinase CK2 Binding Sites Using CK2 α (cid:48) Cys336Ser as an Exquisite Crystallographic Tool

: The structural knowledge about protein kinase CK2 is dominated by crystal structures of human CK2 α , the catalytic subunit of human CK2, and the product of the CSNK2A1 gene. In contrast, far fewer structures of CK2 α (cid:48) , its paralogous isoform and the product of the CSNK2A2 gene, have been published. However, according to a PDB survey, CK2 α (cid:48) is the superior alternative for crystallographic studies because of the inherent potential of the single mutant CK2 α (cid:48) Cys336Ser to provide crystal structures with atomic resolution. In particular, a triclinic crystal form of CK2 α (cid:48) Cys336Ser is a robust tool to determine high-quality enzyme-ligand complex structures via soaking. In this work, further high-resolution CK2 α (cid:48) Cys336Ser structures in complex with selected ligands emphasizing this trend are described. In one of these structures, the “N-terminal segment site”, a small-molecule binding region never found in any eukaryotic protein kinase and holding the potential for the development of highly selective substrate-competitive CK2 inhibitors, was discovered. In order to also address the binding site for the non-catalytic subunit CK2 β , which is inaccessible in these triclinic CK2 α (cid:48) Cys336Ser crystals for small molecules, a reliable path to a promising monoclinic crystal form of CK2 α (cid:48) Cys336Ser is presented. In summary, the quality of CK2 α (cid:48) Cys336Ser as an exquisite crystallographic tool is solidiﬁed.


Introduction
Towards the middle of 2023, the protein structure database PDB [1,2] contains 315 entries with homologues of CK2α, the catalytic subunit of protein kinase CK2.41 of them belong to CK2α from Zea mays (UNIPROT_ID P28523) [3,4], which provided the first CK2α crystal structure [5] and dominated the early years of CK2 structural biology.With the crystal structure of the CK2 holoenzyme (heterotetrameric complex of CK2α with the regulatory subunit CK2β) from Homo sapiens [6], human CK2α (UNIPROT_ID P68400; product of the CSNK2A1 gene) entered the PDB in 2001; it is now included in 231 entries, the majority of them being released since 2016 (Figure 1a) and belonging to complexes of human CK2α with ATP-competitive inhibitors [7][8][9][10].
In comparison, the PDB contains only 25 entries of human CK2α , the paralogous isoform of human CK2α (UNIPROT_ID P19784; product of the CSNK2A2 gene), albeit with an increasing trend (Figure 1a).A main reason for this backlog were solubility issues occurring with wild-type CK2α after recombinant expression in bacteria [11], but also in insect cells [12].The chemical background of this problem became apparent after an inspection of the sequences of human CK2α and CK2α -they are to 82% identical but differ completely in the C-terminal regions-and the construction of the chimeric variant CK2α_α containing its final 24 residues from CK2α [13].This chimera has a conspicuous tendency to form intermolecular disulphide bridges, which drew attention to Cys336 as part of the CK2α -derived tail; replacing this cysteine with serine led to the full-length variant tendency to form intermolecular disulphide bridges, which drew attention to Cys336 as part of the CK2α′-derived tail; replacing this cysteine with serine led to the full-length variant CK2α′ Cys336Ser , which was well soluble and applicable for enzymatic, calorimetric, and crystallographic studies [13].Alternatively, truncating the C-terminal segment, including Cys336, led to CK2α′ constructs that could be successfully crystallized as well [14,15].However, CK2α′ Cys336Ser was still somewhat problematic for the purpose of crystallization because of its propensity to crystallize as extremely tiny needles under a large variety of conditions.Although these crystal needles grow slightly thicker in the presence of 3-(4,5,6,7-tetrabromo-1H-benzotriazol-1-yl)propan-1-ol (MB002; Scheme 1) [13,16] and certain other ATP-competitive inhibitors [17,18], which led to the first crystal structures of CK2α′ Cys336Ser [13,17,18], a real breakthrough in growth behaviour only came with a significant increase of the lithium chloride concentration in the crystallization droplets [19].This-combined with a sophisticated seeding and soaking strategy to replace the originally bound MB002 ligand from the ATP site-resulted in large, compact, and robust CK2α′ Cys336Ser crystals.They led to a number of atomic-resolution structures of CK2α′ Cys336Ser [19], among them a 1.04-Å complex structure with the benchmark CK2 inhibitor CX-4945 [20][21][22][23].

•
For some members of a series of halogenated triazolo pyridines, the exact binding mode at the ATP site could only be clarified via high-resolution complex structures with CK2α′ Cys336Ser [24]; In the case of some 2-aminothiazole compounds supposed to occupy an allosteric binding site [25, 26], CK2α′ Cys336Ser complex structures revealed that in fact the ATP cavity harboured these inhibitory ligands [27]; with resolutions of 0.833 Å (PDB_ID 6TGU) and 0.922 Å (PDB_ID 6TE2), two of these structures are currently the best resolved among about 4500 X-ray diffraction entries in the PDB [1,2] with protein kinase chains;

•
The power of the approach became particularly obvious when the αD pocket, an allosteric site originally discovered with CK2α [28,29], could be occupied via soaking of CK2α′ Cys336Ser /MB002 crystals, although this required extensive local conformational rearrangements in the crystalline state of the protein [30].However, CK2α Cys336Ser was still somewhat problematic for the purpose of crystallization because of its propensity to crystallize as extremely tiny needles under a large variety of conditions.Although these crystal needles grow slightly thicker in the presence of 3-(4,5,6,7-tetrabromo-1H-benzotriazol-1-yl)propan-1-ol (MB002; Scheme 1) [13,16] and certain other ATP-competitive inhibitors [17,18], which led to the first crystal structures of CK2α Cys336Ser [13,17,18], a real breakthrough in growth behaviour only came with a significant increase of the lithium chloride concentration in the crystallization droplets [19].Thiscombined with a sophisticated seeding and soaking strategy to replace the originally bound MB002 ligand from the ATP site-resulted in large, compact, and robust CK2α Cys336Ser crystals.They led to a number of atomic-resolution structures of CK2α Cys336Ser [19], among them a 1.04-Å complex structure with the benchmark CK2 inhibitor CX-4945 [20][21][22][23].
Subsequently, these CK2α Cys336Ser crystals together with the optimized soaking protocol [19] were systematically exploited:

•
For some members of a series of halogenated triazolo pyridines, the exact binding mode at the ATP site could only be clarified via high-resolution complex structures with CK2α Cys336Ser [24]; In the case of some 2-aminothiazole compounds supposed to occupy an allosteric binding site [25,26], CK2α Cys336Ser complex structures revealed that in fact the ATP cavity harboured these inhibitory ligands [27]; with resolutions of 0.833 Å (PDB_ID 6TGU) and 0.922 Å (PDB_ID 6TE2), two of these structures are currently the best resolved among about 4500 X-ray diffraction entries in the PDB [1,2] with protein kinase chains;

•
The power of the approach became particularly obvious when the αD pocket, an allosteric site originally discovered with CK2α [28,29], could be occupied via soaking of CK2α Cys336Ser /MB002 crystals, although this required extensive local conformational rearrangements in the crystalline state of the protein [30].
We expected a similar performance for SGC-CK2-1 (Scheme 1), an ATP-competitive CK2 inhibitor with IC50 values for CK2α/CK2α′ inhibition in the low nanomolar range and with an extremely high selectivity [31,32].SGC-CK2-1 has the potential to compete with CX-4945 for the role of benchmark CK2 inhibitor and superior CK2-sensitive tool in cell biology studies, as already demonstrated [33][34][35].In fact, soaking of CK2α′ Cys336Ser /MB002 crystals in 2.5 mM SGC-CK2-1 led to a high-quality CK2α′ Cys336Ser /SGC-CK2-1 complex structure refined to 1.352 Å resolution (Table 1, Figure 2a).The excellent properties of CK2α Cys336Ser crystals are responsible for the fact that the resolution distribution of CK2α structures is significantly shifted to the high-resolution side compared to CK2α (Figure 1b).In this work, we supplement the collection of CK2α Cys336Ser /ligand complex structures with further cases.Partially, they were obtained via the crystallization and soaking protocol originally published [19], but we also present selected structural examples resulting from efforts to overcome certain limitations of those crystals and procedures.The size-optimized CK2α Cys336Ser /MB002 crystals used by Lindenblatt et al. [19] to determine atomic resolution complex structures belong to the triclinic crystal system (space group P1).In the corresponding crystal packing, the ATP-site is well accessible, so that in particular, high-affinity ligands like CX-4945 can easily displace the crystallization helper compound MB002 [19].

Results and Discussion
We expected a similar performance for SGC-CK2-1 (Scheme 1), an ATP-competitive CK2 inhibitor with IC 50 values for CK2α/CK2α inhibition in the low nanomolar range and with an extremely high selectivity [31,32].SGC-CK2-1 has the potential to compete with CX-4945 for the role of benchmark CK2 inhibitor and superior CK2-sensitive tool in cell biology studies, as already demonstrated [33][34][35].In fact, soaking of CK2α Cys336Ser /MB002 crystals in 2.5 mM SGC-CK2-1 led to a high-quality CK2α Cys336Ser /SGC-CK2-1 complex structure refined to 1.352 Å resolution (Table 1, Figure 2a).The cutoff level for electron density illustration in this figure is generally 1 σ.The structure pictures were produced with PyMol [36].(a) Overview of the CK2α′ Cys336Ser /SGC-CK2-1 complex structure.(b) SGC-CK2-1 occupying the ATP site in CK2α′ Cys336Ser (magenta-coloured C-atoms) and for comparison in CK2α (green C-atoms; PDB_ID 6Z83 [31]).(c) Overview of the CK2α′ Cys336Ser structure in complex with CX-4945 and DPA.(d) Hinge/helix αD region of the ternary CK2α′ Cys336Ser /CX-4945/DPA complex (this work) and an overlaid binary CK2α′ Cys336Ser /CX-4945 structure (PDB_ID 6HMB [19]).(e) Overview of the CK2α′ Cys336Ser structure in complex with two molecules of ARC-780.(f) ARC-780 ligand 1 occupying the ATP site of CK2α′ Cys336Ser ; the lengths of three H-bonds between the ligand and the enzyme s hinge region (black dotted lines) are given in Å; for comparison, CX-4945 is drawn from PDB_ID 6HMB [19] after 3D-fit of the protein matrices.(g) The ordered part of ARC-780 ligand 2 occupying the novel N-terminal segment site of CK2α′ Cys336Ser ; side chains of a neighbouring symmetry mate were drawn with yellow C-atoms; Ser18, His184, and Gln187 from unbound (ARC-780-free) CK2α′ Cys336Ser were depicted with black C-atoms.(h) Overview of the CK2α′ Cys336Ser /TBBt complex structure.(i) Zoom of TBBt bound to the ATP site of CK2α′ Cys336Ser (magenta-coloured C-atoms); for comparison, TBBt was drawn as bound to maize CK2α (orange-coloured C-atoms; PDB_ID 1J91 [37]) or to human CK2α at pH 8.5 (green C-atoms; PDB_ID 7QGE [38]).(j) Involvement of the CK2β binding site of CK2α′ Cys336Ser in a crystal contact.The cutoff level for electron density illustration in this figure is generally 1 σ.The structure pictures were produced with PyMol [36].(a) Overview of the CK2α Cys336Ser /SGC-CK2-1 complex structure.(b) SGC-CK2-1 occupying the ATP site in CK2α Cys336Ser (magenta-coloured C-atoms) and for comparison in CK2α (green C-atoms; PDB_ID 6Z83 [31]).(c) Overview of the CK2α Cys336Ser structure in complex with CX-4945 and DPA.(d) Hinge/helix αD region of the ternary CK2α Cys336Ser /CX-4945/DPA complex (this work) and an overlaid binary CK2α Cys336Ser /CX-4945 structure (PDB_ID 6HMB [19]).(e) Overview of the CK2α Cys336Ser structure in complex with two molecules of ARC-780.(f) ARC-780 ligand 1 occupying the ATP site of CK2α Cys336Ser ; the lengths of three H-bonds between the ligand and the enzyme's hinge region (black dotted lines) are given in Å; for comparison, CX-4945 is drawn from PDB_ID 6HMB [19] after 3D-fit of the protein matrices.(g) The ordered part of ARC-780 ligand 2 occupying the novel N-terminal segment site of CK2α Cys336Ser ; side chains of a neighbouring symmetry mate were drawn with yellow C-atoms; Ser18, His184, and Gln187 from unbound (ARC-780-free) CK2α Cys336Ser were depicted with black C-atoms.(h) Overview of the CK2α Cys336Ser /TBBt complex structure.(i) Zoom of TBBt bound to the ATP site of CK2α Cys336Ser (magenta-coloured C-atoms); for comparison, TBBt was drawn as bound to maize CK2α (orange-coloured C-atoms; PDB_ID 1J91 [37]) or to human CK2α at pH 8.5 (green C-atoms; PDB_ID 7QGE [38]).(j) Involvement of the CK2β binding site of CK2α Cys336Ser in a crystal contact.For comparison, a CK2α/SGC-CK2-1 complex structure was published with a resolution of 2.17 Å [31].An overlay of this structure on the novel CK2α Cys336Ser /SGC-CK2-1 structure (Figure 2b) reveals that SGC-CK2-1 is bound in a similar way to the two paralogs, albeit with a noteworthy conformational difference in the terminal propionamide moiety: the torsion angle around the final C-C-bond as indicated in Figure 2b is −23.14 • in CK2α Cys336Ser -bound SGC-CK2-1, but 35.24 • or 59.65 • , respectively, in the two crystallographically independent copies of the CK2α/SGC-CK2-1 structure [31].This difference is not caused by the direct protein environment around the propionamide moiety since it is identical in the two isoforms.Rather, Figure 2b suggests that the structural background lies in the interdomain hinge where CK2α and CK2α have their only sequence difference in the ATP-site region: Tyr116 of CK2α is more voluminous than its equivalent His115 of CK2α so that SGC-CK2-1 can approach the hinge less closely in CK2α .This causes a shift visible at the cyclopropyl group and a constraint at the other side of the molecule, namely at the aforementioned propionamide moiety, which penetrates deeper into the ATP site of CK2α than into that of CK2α (Figure 2b).This subtle effect correlates well with the fact that the IC 50 value of SGC-CK2-1 for CK2α inhibition is only half as large as for CK2α inhibition [31], and it could be a first step on the way to the attractive goal of generating isoform-specific CK2 inhibitors [40,41].

Ternary Complex Structure of CK2α Cys336Ser , CX-4945, and DPA
The αD pocket of CK2α and CK2α is a so-called "cryptic site" [42]; this designation was introduced for "hidden" pockets becoming visible and accessible only after conformational changes induced in the presence of a suitable ligand.In the case of the αD pocket, this ligand was 3,4-dichloro phenethylamine (DPA; Scheme 1), which led to the discovery of the αD pocket in 2016 in a fragment-based screen with CK2α crystals [28,29], while in all CK2α structures published before, it was blocked by either a phenylalanine or a tyrosine side chain from the eponymous helix αD.After its discovery, the αD pocket was used to develop the bivalent inhibitor CAM4066 [28,29].
For CK2α , the existence of an αD pocket was confirmed via soaking DPA into a CK2α Cys336Ser /MB002 crystal [30].In the resulting ternary complex structure, the two ligands are 6.9 Å away from each other, a distance that could be chemically bridged and led to the bivalent and highly selective inhibitor KN2 [30].This success induced the hypothesis that the ATP site and the αD pocket of crystalline CK2α Cys336Ser can be occupied independently of each other, so that CK2α Cys336Ser structures in complex with arbitrary pairs of ligands can be obtained via soaking CK2α Cys336Ser /MB002 crystals.
To test this, we soaked a CK2α Cys336Ser /MB002 crystal simultaneously with CX-4945 and DPA.In the resulting crystal structure with a resolution below 1.1 Å (Table 1), both ligands were well defined by electron density (Figure 2c), meaning CX-4945 had completely replaced MB002 and DPA had opened and occupied the αD pocket.A 3D-fit of this ternary CK2α Cys336Ser /CX-4945/DPA complex on the binary CK2α Cys336Ser /CX-4945 complex structure (Figure 2d) illustrates again the large conformational adaptations occurring in the αD region during soaking.In summary, this experiment confirms that the ATP site and the αD pocket of crystalline CK2α Cys336Ser cannot only be occupied independently but even simultaneously.

N-Terminal Segment Site: A Novel Ligand Binding Site Discovered in a CK2α Cys336Ser
Structure in Complex with the Bisubstrate Inhibitor ARC-780 ARC-780 (Scheme 1) is a follow-up compound of a series of CK2 bisubstrate inhibitors [43] occupying the ATP site and addressing the recognition region for peptide or protein substrates as well, so that the corresponding K i values are partly in the picomolar range [43].ARC-780 contains a long aliphatic linker of 11 methylen groups (Scheme 1), resulting in high hydrophobicity and many rotational degrees of freedom.Therefore, we extended the soaking time of initial CK2α Cys336Ser /MB002 crystals in ARC-780 containing solutions to eight months.In this time, the concentrations of salts were stepwise decreased from 900 mM LiCl, 500 mM NaCl, and 110 mM Tris/HCl to finally 50 mM LiCl and 10 mM Tris/HCl, while in parallel, the concentration of DMSO was increased from 0 to 30% (v/v), and the concentration of the precipitant PEG6000 was adopted to saturation.
In spite of these extreme alterations of the surrounding solution, the CK2α Cys336Ser crystals retained their integrity, and one of them led to a 1.255 Å crystal structure (structure no. 3 in Table 1).Surprisingly, in this structure, two large patches of non-protein electron density were visible.They could be interpreted unambiguously with two separate ARC-780 molecules, and in both cases, large, albeit different, parts of ARC-780 were not covered with significant electron density (Figure 2e).
The first ARC-780 ligand occupies the ATP site (Figure 2f, Video S1): its terminal carboxy phenyl moiety binds next to the Phe114 side chain, and after 3D-fit of the according protein matrices, it overlaps entirely with the corresponding sub-structure of CX-4945.Using its triazole ring and the two subsequent amide bonds, the first ARC-780 ligand forms three close hydrogen bonds with the enzyme's hinge region, while the aliphatic linker and the final peptidic part are completely disordered (Figure 2f).
The second ARC-780 ligand-or at least a considerable part of it-is bound to a region (Figure 2g, Video S1) that, to our knowledge, has never been described before as a small-molecule binding site in CK2α, CK2α , or any other member of the superfamily of eukaryotic protein kinases (EPKs).We call it the "N-terminal segment site" here because it is located close to an N-terminal extension of the EPK-typical N-lobe domain that attaches the likewise EPK-typical activation loop and folds back to the C-lobe domain (Figure 2e).This intramolecular interaction between the N-terminal segment and the activation loop is a distinguishing feature of CK2α/CK2α among the EPK superfamily and the main structural basis of CK2 s constitutive activity since it mimics the intermolecular coordination of the activation loop of cyclin-dependent kinases by activatory cyclin proteins [5].
Just like the αD site, the N-terminal segment site is a cryptic site [42].Its access for a ligand requires large conformational changes in the side chains of Ser18, His184, and Gln187, which are indicated in Figure 2g and visualized in Video S1.The four carboxy groups in the peptidic part of ARC-780 form ionic interactions with some positively charged side chains of a crystalline CK2α Cys336Ser neighbour (Figure 2g).Thus, the fact that this peptidic part is well defined by electron density in strong contrast to what has been observed in CK2α complex structures with precursor bisubstrate inhibitors [43] is most likely an artifact of the crystal packing.Insofar, it must remain open in the moment how far the peptidic part of ARC-780 contributes to the compound's affinity to the N-terminal segment site of the enzyme at all.
In contrast to that, the N-linked carboxy methyl group of ARC-780 (Scheme 1) is likely to be an important binding determinant because it interacts ionically with Arg156 (Figure 2g, time point 12 s in Video S1).Together with Arg81, Arg156 forms the P + 3 recognition site for CK2 substrates, which means that this carboxy methyl group would interfere with substrate binding.In the future, this fact could pave the way to developing highly selective, substrate-competitive CK2 inhibitors from the sub-structure of ARC-780 (Figure 2g) that binds to the novel N-terminal segment site with its unique molecular environment.
For TBBt, a pK A value of 4.78 was reported [45], so that it is monoanionic at pH 7 and above.Therefore, a "competition between electrostatic interactions and halogen bonding" [38] exists for the interaction of TBBt with the ATP site of CK2α.In fact, in a human CK2α/TBBt complex structure obtained at pH 8.5, two alternative binding modes of TBBt were observed [38] (Figure 2i).In one of them, which had been found before in a maize CK2α/TBBt complex structure as well [37] (Figure 2i), electrostatic forces dominate and direct TBBt via its anionic triazole moiety into the proximity of the highly conserved side chain of Lys68 (which is equivalent to Lys69 of CK2α drawn in Figure 2i) [46].In the other binding mode, however, halogen bonds force TBBt close to the interdomain hinge of human CK2α.It is this second binding mode that we observed exclusively in the Cys336Ser /TBBt complex (Figure 2i).Apparently, there, the three halogen bonds from the bromo substituents of TBBt to the carbonyl oxygen atoms of Glu115 and Ile117, as well as to the phenyl ring of Phe114, dominate clearly over the electrostatic term.
Noteworthy, the salt concentration during crystallization of the CK2α Cys336Ser /TBBt complex was much higher than that reported for crystallization of the CK2α/TBBt complex [38].Therefore, follow-up experiments are required to clarify if the difference in TBBt binding between human CK2α and CK2α is a crystallization artifact or if it reflects subtle differences between the isoenzymes themselves.In any case, atomic-resolution complex structures of CK2α Cys336Ser with halogenated inhibitors can help to refine the sophisticated knowledge about halogen bonding in the ATP site of CK2α subunits [38,44,47].

Complex Structures Determined with Monoclinic CK2α Cys336Ser Crystals
In order to test if the excellent diffraction of triclinic CK2α Cys336Ser crystals is restricted to this particular crystal packing, an alternative crystal packing of CK2α Cys336Ser had to be found.A further motivation to pursue this goal came from the observation that the outer surface of the enzyme's N-lobal anti-parallel CK2β-sheet, which serves as an interface for the binding of the non-catalytic subunit CK2β [6], is involved in a crystal contact in the triclinic CK2α Cys336Ser crystals (Figure 2j): it is thus inaccessible for small CK2βantagonistic molecules in soaking approaches and also for co-crystallization because the crystal packing is incompatible with any occupation of the CK2β binding site.
Small peptidic or non-peptidic molecules that compete with CK2β for its binding site and disturb the assembly of the CK2 holoenzyme have been repeatedly described [43,[48][49][50][51][52][53][54] and, in several cases, structurally characterized in recent years, but exclusively for CK2α.Structures of CK2α in complex with such ligands do not exist at present; in order to create the prerequisite for them, a reliable pathway to an alternative crystal form of CK2α Cys336Ser is desirable.
Since it was demonstrated with lithium sulfate [15] or lithium chloride [19] that the crystallization behavior of CK2α constructs is very sensitive to the content of lithium ions, we systematically varied the lithium chloride concentration in CK2α Cys336Ser crystallization experiments between 500 mM and 900 mM with otherwise unchanged conditions.Already 810 mM LiCl-i.e., a slight reduction of the LiCl concentration compared to the 900 mM in crystallization solutions of the triclinic CK2α Cys336Ser crystals (Table 1)-led to CK2α Cys336Ser crystals with monoclinic symmetry (Table 2).In this new crystal form, two CK2α Cys336Ser protomers per asymmetric unit are present.In one of them, the CK2β binding site is used to form the contact interface to a neighbouring symmetry mate (Figure 3a) in a similar way as it was observed for the triclinic crystals (Figure 2j).In the second protomer, however, the CK2β binding site, although also surrounded by two crystalline neighbours, is basically accessible to ligands of small and medium size (Figure 3b).The X-ray diffraction capacity of the monoclinic CK2α′ Cys336Ser crystals (Table 2) is similar to that of the triclinic crystals (Table 1), i.e., structures with atomic resolution can be achieved and expected.We demonstrate this here with two CK2α′ Cys336Ser complex structures with the co-substrate analogues AMPPNP or GMPPNP (Scheme 1), both of which are well visible in electron density at the ATP site (Figure 3c,d), at least in one of the two protomers, respectively.
The ability of CK2 to use GTP as well as ATP as cosubstrate for the kinase reaction has been known since the early 1960s [55].The first evidence that this dual-cosubstrate specificity is a special property of CK2 compared to other protein kinases emerged in the 1970s [56], but how exceptional this feature of CK2α and CK2α′ really is among EPKs has only been demonstrated in recent years by affinity profiling [57].Its structural background was clarified with maize CK2α structures in complex of AMPPNP or GMPPNP [58], which revealed essentially what is visible in Figure 3c,d again: the triphospho moieties of the two ligands overlap exactly, but in the region of the purine bases, the different hydrogenbonding potentials of adenine and guanine enforce a shift of GMPPNP compared to /AMPPNP complex structure (magenta-coloured C-atoms); for comparison, the CK2α Cys336Ser /GMPPNP complex structure was drawn with green C-atoms.(d) Cosubstrate binding site of the CK2α Cys336Ser /GMPPNP complex structure (green C-atoms); for comparison, the CK2α Cys336Ser /AMPPNP complex structure was drawn with magenta-coloured C-atoms.The cutoff level for electron density illustration in this figure is generally 1 σ.The structure pictures were produced with PyMol [36].
The X-ray diffraction capacity of the monoclinic CK2α Cys336Ser crystals (Table 2) is similar to that of the triclinic crystals (Table 1), i.e., structures with atomic resolution can be achieved and expected.We demonstrate this here with two CK2α Cys336Ser complex structures with the co-substrate analogues AMPPNP or GMPPNP (Scheme 1), both of which are well visible in electron density at the ATP site (Figure 3c,d), at least in one of the two protomers, respectively.
The ability of CK2 to use GTP as well as ATP as cosubstrate for the kinase reaction has been known since the early 1960s [55].The first evidence that this dual-cosubstrate specificity is a special property of CK2 compared to other protein kinases emerged in the 1970s [56], but how exceptional this feature of CK2α and CK2α really is among EPKs has only been demonstrated in recent years by affinity profiling [57].Its structural background was clarified with maize CK2α structures in complex of AMPPNP or GMPPNP [58], which revealed essentially what is visible in Figure 3c,d again: the triphospho moieties of the two ligands overlap exactly, but in the region of the purine bases, the different hydrogenbonding potentials of adenine and guanine enforce a shift of GMPPNP compared to AMPPNP so that finally the pattern of hydrogen bonds between the purine base and the hinge region is shifted by one peptide bond toward the C-terminus.

Protein Expression and Purification
For any crystallization experiment shown in this work, we used the point mutant CK2α Cys336Ser with an N-terminal His-tag [13].The corresponding construct in a pETDuet-1 vector (Merck) was transformed into either chemically competent E. coli Gen-X TM (DE3) or chemically competent BL21 Star (DE3) cells.A single clone, respectively, was selected for a preculture.For this, the cells were grown overnight at 30 • C in 100 mL LB medium (10 g/L yeast extract, 20 g/L tryptone, and 20 g/L NaCl) supplemented with 100 µg/mL ampicilin under constant agitation at 180 rpm.An amount of 10 mL of the preculture was used to inocculate 600 mL LB medium.
The subsequent main cultures were grown to an OD 600 of 0.6 at 37 • C under constant agitation at 180 rpm.Upon reaching an OD600 of 0.6, the temperature was lowered to 18 • C, and isopropyl-β-thiogalactosid (IPTG) was added to a final concentration of 0.5 mM.Expression was subsequently carried out for 18 h.Afterwards, the cells were harvested via centrifugation at 4 • C and 6200× g for 30 min.The pellet was washed with a 0.9% (w/v) sodium chloride solution and afterwards stored at −80 • C. For lysis, the pellet was thawed and then incubated with lysis buffer composed of 500 mM NaCl, 25 mM TRIS-HCl, pH 8.5, 10 µg/mL DNase1, and 1 mg/mL lysozyme for 1 h on ice.Sonification (40% amplitude, 2 s on/4 s off, 4 • C, 3 min total sonification) followed.The lysate was cleared by ultracentrifugation (186,000× g, 4 • C, 30 min).The supernatant was filtered and then filled into a superloop for application onto a 5 mL HisTrap FF column (Cytiva) mounted on an ÄKTA Explorer System.Application and washing (15 CV) were carried out using buffer A (40 mM imidazole, 500 mM NaCl, and 25 mM TRIS-HCl.pH 8.5).Afterwards, the protein was eluted using a linear gradient from 0 to 100% buffer B (250 mM imidazole, 500 mM NaCl, 25 mM TRIS-HCl, pH 8.5) over 15 CV.The flowrate was 0.8 mL/min.Elution was monitored via absorption at 280 nm.All fractions within a peak occurring during elution were tested using SDS-PAGE.Fractions containing pure CK2α Cys336Ser were pooled, rebuffered into a standard assay buffer (500 mM NaCl, 25 mM TRIS-HCl, pH 8.5), and concentrated to 5 mg/mL by ultrafiltration using AMICON ® tubes with a cut-off of 30 kDa.

Preparation of ARC-780
While all other CK2 ligands used in this work were either commercially available or (in the case of MB002) were provided by Prof. Maria Bretner, Warsaw, Poland, the substance ARC-780, which had been planned to be a bisubstrate inhibitor, had to be synthesized.For this, (L-Asp) 3 peptide with tert-butyl-protected side chains was synthesized on 2chlorotrityl chloride (2CTC) polystyrene resin (Iris Biotech GmbH) according to common Fmoc-peptide synthesis protocols.Succinic acid and the precursors of residues of 12-[(carboxymethyl)amino]dodecanoic acid and 4-(2-amino-1,3-thiazol-5-yl)benzoic acid were synthesized and attached to the peptide as it was described previously [59].The compound was cleaved from the resin with a 3 h treatment with TFA:TIPS:H 2 O (95:2.5:2.5, v:v:v); subsequent RP-HPLC purification [Shimadzu Prominence LC Solution HPLC system with SPD M20A PDA detector; Phenomenex Gemini C18 RP column, 250 × 4.6 mm, particle size 5 µm, eluted with MeCN/H 2 O gradient (0.1% TFA) at flow rate of 1 mL/min; gradient of 20-62% MeCN/14 min, R t of the title compound: 10.7 min, λ max = 322 nm; Figure S1] yielded ARC-780 with a purity of more than 96.5% (Peak Table insert in Figure S1).The structure was verified using high-resolution mass spectrometry (HRMS) with combined Varian 910-FT-ICR and Varian J-320 3Q spectrometers used in ESI positive ion mode (Figure S2  All crystals were grown by applying the sitting drop variant of the vapor diffusion method.For this, 24-well plates (Cryschem) were used.The volume of the reservoir solution was initially 700 µL.All plates were sealed with transparent adhesive film and kept at 20 • C. All microseeding steps were carried out by destroying one crystal from a previous experiment and streaking a loop first through the resulting suspension of crystal seeds and afterwards through the drop that should be inocculated.
SGC-CK2-1 (Sigma-Aldrich) was solved in DMSO at a concentration of 10 mM.This solution was added to CK2α Cys336Ser /MB002 complex crystals to a concentration of 2.5 mM for extensive soaking.
DPA (Sigma-Aldrich, St. Louis, MO, USA) was solved in DMSO at a concentration of 50 mM and CX-4945 (TargetMol, Wellesley Hills, MA, USA) to a concentration of 10 mM likewise in DMSO. 1 µL of each of these solutions, respectively, was added to a CK2α Cys336Ser /MB002 complex crystals for extensive soaking.

Preparation of CK2α Cys336Ser /ARC-780 Complex Crystals
Co-crystals of CK2α Cys336Ser and MB002 were grown, optimized by micro-and macroseeding, and purged as described above.Then, ARC-780 was added to an initial concentration of 1 mM by adding 1 µL of a 10 mM ARC-780 solution in DMSO in a 1:10 ratio to the droplet.After extensive initial soaking, the salt concentration was successively lowered over a period of 8 months, while in parallel, the concentrations of DMSO and PEG6000 were increased.This was conducted by gradually replacing the reservoir and the droplet solution until any NaCl (which came from the storage solution of the protein) was removed, the LiCl concentration was reduced to 50 mM, the PEG 6000 concentration was increased to saturation, and a DMSO content of 30% was reached.Meanwhile, the ARC-780 ligand was included in any desalting step reaching saturation after the final dilution step.

Co-Crystallization of CK2α Cys336Ser and TBBt
For this crystallization attempt, the initial co-crystallization with MB002 was avoided.Rather, a 5 mg/mL stock solution of CK2α Cys336Ser was mixed in a 1:10 ratio with a 20 mM solution of TBBt (Sigma-Aldrich) in DMSO.After incubation on ice for 30 min and centrifugation (16,100× g for 2 min, room temperature), 10 µL of the supernatant were mixed with 5 µL of the reservoir solution (900 mM LiCl, 28% (w/v) PEG 6000, 100 mM TRIS-HCl, pH 8.5).After equilibration, microseeding was carried out to induce crystal growth.Small crystals appeared after one week.One of them was used as a macroseed by transferring it to a second drop prepared in the same manner as the first one.

Growth of CK2α
Cys336Ser Crystals in Complex with AMPPNP or GMPPNP Monoclinic crystals of CK2α Cys336Ser were grown in a slightly altered manner compared to the triclinic CK2α Cys336Ser crystals.5 mg/mL CK2α Cys336Ser in storage buffer composed of 500 mM NaCl and 25 mM TRIS-HCl, pH 8.5, were centrifuged immediately after thawing and without mixing it with any inhibitor.10 µL of the supernatant were mixed with 5 µL of a reservoir solution containing a slightly lowered LiCl concentration of 810 mM, 28% (w/v) PEG 6000, 100 mM TRIS-HCl, pH 8.5.After equilibration, crystallization was initiated by microseeding, and then crystals were optimized via macroseeding.The ATP-analogue AMPPNP or the GTP-analogue GMPPNP were combined with these crystals by socking.In both cases, a 20 mM solution in 60 mM MgCl 2 was prepared.For soaking, 3 µL of the crystal mother liquor were removed and replaced with 3 µL of either 20 mM AMPPNP, 60 mM MgCl 2 or 20 mM GMPPNP, 60 mM MgCl 2 .

X-ray Diffraction Data Collection and Processing followed by Structure Determination
To prepare X-ray diffraction data collection, any CK2α Cys336Ser /ligand crystal except for the CK2α Cys336Ser /ARC-780 co-crystals was cryo-protected before vitrification.For this, 1.4 µL of the reservoir solution and 0.6 µL ethylene glycol were mixed, and the crystals were incubated in this solution for a few seconds.Next, the crystals were flash frozen in liquid nitrogen.Data collections were carried out at beamlines ID 23-1 and MASSIF-3 of the European Synchrotron Radiation Facility (ESRF) in Grenoble (France) or at beamline P13 of the European Molecular Biology Laboratory (EMBL) outstation at DESY in Hamburg (Germany).The temperature during diffraction data collection was 100 K.
The raw diffraction data were processed with the autoPROC toolbox [60] using default settings.The autoPROC pipeline comprises XDS [61] for indexing and integration, POINT-LESS [62] and AIMLESS [63] from the CCP4 suite [64] for symmetry determination and scaling, and finally STARANISO [39] for anisotropy analysis.In all cases, the ellipsoidal output data set of STARANISO [39] was used for the further procedure.
All structures were solved by molecular replacement using the CK2α Cys336Ser /MB002 structure [19] (PDB_ID 6HMQ) as a starting model.The search calculations were performed with PHASER [65] from Phenix [66].For the subsequent structure refinement, we used the phenix.refinemodule [67] of Phenix [66] in alternation with Coot [68] for manual model building.Small-molecule ligands were parametrized with eLBOW [69] within the Phenix [66] platform.

Conclusions
In this work, six crystal structures of CK2α Cys336Ser in complex with various ligands and obtained with two different crystal forms are described.In all cases, the high-resolution limit of the underlying X-ray diffraction data is in the range from 1.03 to 1.35 Å.This demonstrates the qualification of CK2α Cys336Ser as an exquisite tool for CK2 crystallography and emphasizes a tendency to atomic-resolution structures, as indicated already in a survey of the PDB (Figure 1b).In particular, the remarkable robustness of the longer-known triclinic CK2α Cys336Ser crystals is demonstrated here with an eight-month soaking experiment with ARC-780, a 920 Da molecule designed as a bisubstrate inhibitor, in which the crystals were acclimated to low-salt, DMSO-rich conditions by repeated replacement.In the end, this led to the discovery of the N-terminal segment site, a ligand binding site never observed before in the EPK superfamily that offers the potential to develop highly selective substratecompetitive CK2 inhibitors.

Figure 1 .
Figure 1.Distributions of CK2α and CK2α′ crystal structures based on data extracted from the PDB [1, 2].(a) Structures clustered into intervals of release years; (b) structures clustered into intervals of resolution.

Figure 1 .
Figure 1.Distributions of CK2α and CK2α crystal structures based on data extracted from the PDB [1,2].(a) Structures clustered into intervals of release years; (b) structures clustered into intervals of resolution.

Scheme 1 .
Scheme 1. Structures of CK2 ligands used in this study.Scheme 1. Structures of CK2 ligands used in this study.

Figure 2 .
Figure 2. CK2α Cys336Ser structures in complex with various ligands obtained in triclinic crystal form.The cutoff level for electron density illustration in this figure is generally 1 σ.The structure pictures were produced with PyMol[36].(a) Overview of the CK2α Cys336Ser /SGC-CK2-1 complex structure.(b) SGC-CK2-1 occupying the ATP site in CK2α Cys336Ser (magenta-coloured C-atoms) and for comparison in CK2α (green C-atoms; PDB_ID 6Z83[31]).(c) Overview of the CK2α Cys336Ser structure in complex with CX-4945 and DPA.(d) Hinge/helix αD region of the ternary CK2α Cys336Ser /CX-4945/DPA complex (this work) and an overlaid binary CK2α Cys336Ser /CX-4945 structure (PDB_ID 6HMB[19]).(e) Overview of the CK2α Cys336Ser structure in complex with two molecules of ARC-780.(f) ARC-780 ligand 1 occupying the ATP site of CK2α Cys336Ser ; the lengths of three H-bonds between the ligand and the enzyme's hinge region (black dotted lines) are given in Å; for comparison, CX-4945 is drawn from PDB_ID 6HMB[19] after 3D-fit of the protein matrices.(g) The ordered part of ARC-780 ligand 2 occupying the novel N-terminal segment site of CK2α Cys336Ser ; side chains of a neighbouring symmetry mate were drawn with yellow C-atoms; Ser18, His184, and Gln187 from unbound (ARC-780-free) CK2α Cys336Ser were depicted with black C-atoms.(h) Overview of the CK2α Cys336Ser /TBBt complex structure.(i) Zoom of TBBt bound to the ATP site of CK2α Cys336Ser (magenta-coloured C-atoms); for comparison, TBBt was drawn as bound to maize CK2α (orange-coloured C-atoms; PDB_ID 1J91[37]) or to human CK2α at pH 8.5 (green C-atoms; PDB_ID 7QGE[38]).(j) Involvement of the CK2β binding site of CK2α Cys336Ser in a crystal contact.

Figure 3 .
Figure 3. CK2α′ Cys336Ser structures obtained with monoclinic crystals.The asymmetric unit contains two CK2α′ Cys336Ser chains.(a) Crystal contact of the N-terminal domain of CK2α′ Cys336Ser chain A with a symmetry-related CK2α′ Cys336Ser protomer.(b) Crystal packing environment of the N-terminal domain of CK2α′ Cys336Ser chain B.(c) Cosubstrate binding site of the CK2α′ Cys336Ser /AMPPNP complex structure (magenta-coloured C-atoms); for comparison, the CK2α′ Cys336Ser /GMPPNP complex structure was drawn with green C-atoms.(d) Cosubstrate binding site of the CK2α′ Cys336Ser /GMPPNP complex structure (green C-atoms); for comparison, the CK2α′ Cys336Ser /AMPPNP complex structure was drawn with magenta-coloured C-atoms.The cutoff level for electron density illustration in this figure is generally 1 σ.The structure pictures were produced with PyMol[36].

Figure 3 .
Figure 3. CK2α Cys336Ser structures obtained with monoclinic crystals.The asymmetric unit contains two CK2α Cys336Ser chains.(a) Crystal contact of the N-terminal domain of CK2α Cys336Ser chain A with a symmetry-related CK2α Cys336Ser protomer.(b) Crystal packing environment of the N-terminal domain of CK2α Cys336Ser chain B. (c) Cosubstrate binding site of the CK2α Cys336Ser/AMPPNP complex structure (magenta-coloured C-atoms); for comparison, the CK2α Cys336Ser /GMPPNP complex structure was drawn with green C-atoms.(d) Cosubstrate binding site of the CK2α Cys336Ser /GMPPNP complex structure (green C-atoms); for comparison, the CK2α Cys336Ser /AMPPNP complex structure was drawn with magenta-coloured C-atoms.The cutoff level for electron density illustration in this figure is generally 1 σ.The structure pictures were produced with PyMol[36].

Table 1 .
X-ray diffraction data and refinement statistics for triclinic CK2α Cys336Ser crystals.

Table 2 .
X-ray diffraction data and refinement statistics for monoclinic CK2α Cys336Ser crystals.