Serendipitous alkylation of a Plk1 ligand uncovers a new binding channel.

We obtained unanticipated synthetic byproducts from alkylation of the δ(1) nitrogen (N3) of the histidine imidazole ring of the polo-like kinase-1 (Plk1) polo-box domain (PBD)-binding peptide PLHSpT. For the highest-affinity byproduct, bearing a C(6)H(5)(CH(2))(8)- group, a Plk1 PBD cocrystal structure revealed a new binding channel that had previously been occluded. An N-terminal PEGylated version of this peptide containing a hydrolytically stable phosphothreonyl residue (pT) bound the Plk1 PBD with affinity equal to that of the non-PEGylated parent but showed markedly less interaction with the PBDs of the two closely related proteins Plk2 and Plk3. Treatment of cultured cells with this PEGylated peptide resulted in delocalization of Plk1 from centrosomes and kinetochores and in chromosome misalignment that effectively induced mitotic block and apoptotic cell death. This work provides insights that might advance efforts to develop Plk1 PBD-binding inhibitors as potential Plk1-specific anticancer agents.

The polo-like family of serine/threonine protein kinases (collectively, Plks) play crucial roles in cell cycle regulation and cell proliferation.[1-5] Of four human Plks (1 through 4), the ability of Plk1 to promote oncogenic transformation[6-8] has lead to the search for inhibitors of Plk1 function that could serve as clinically relevant anticancer therapeutics.[9-16] A potential drawback of classical inhibitors directed at the Plk1 kinase domain is a lack of specificity due to the high degree of similarity in the ATP binding clefts among kinases. This could present difficulties, since down regulation of Plk1 with concomitant inhibition of the closely-related Plk2 or Plk3 would be contraindicated because of the positive roles these latter kinases play in maintaining genetic stability.[17,18]

In addition to their kinase domains, Plks also contain C-terminal polo-box domains (PBDs) that recognize phospho-Ser (pS)/phospho-Thr (pT)-containing motifs having N-proximal serine residues [S-(pT/pS)].[19-21] PBD-mediated binding provides sub-cellular localization that is critical for proper Plk function. Blockade of PBD-dependent protein-protein interactions inhibits the mitotic functions of Plk1[22-24] and the uniqueness of PBDs to Plks makes disruption of PBD-dependent interactions an alternative and potentially highly specific means of inhibiting Plk function.[6,25-27] Screening small molecule or natural product libraries represents one approach to developing Plk1 PBD-binding antagonists.[22,24,28,29] Peptide-based antagonists, derived through an understanding of PBD-ligand interactions afford a complimentary approach.

A single PBD is composed of two highly homologous polo-boxes (PB1 and PB2), each of which consists of six anti-parallel β-sheet strands and an α-helix. The association of PB1 with PB2 provides a single functional PBD composed of a 12-stranded β-sandwich that is competent to recognize and bind phoshopeptide targets.[20,21] Crystal structures of the Plk1 PBD in complex with phosphopeptides [MQSpTPL (PDB: 1Q4K),[21] PMQSpTPL (PDB: 1UMW),[20] LLCSpTPN (PDB: 3BZI),[30] LHSpTA (PDB: 3FVH),[23] PLHSpT (PDB: 3HIK),[23] and PPHSpT (PDB: 3C5L)[23]] show that the peptides bind in similar fashion within a positively-charged groove formed between the PB1 and PB2 components. Experiments indicate that a pT residue is essential for high affinity binding. Calculations showing that this residue provides approximately one third of the overall peptide binding free energy support empirical observations of its importance.[31]

In an attempt to improve the pharmacological properties of the high affinity Plk1 PBD-binding peptide PLHSpT (1), we previously replaced without loss of Plk1 PBD binding affinity, the hydrolytically-labile pT residue with the phosphatase-stable pT mimetic, (2S,3R)-2-amino-3-methyl-4-phosphonobutanoic acid (Pmab)[32] (2).[23] More recently, we undertook the preparation of a series of pT phosphodiesters of peptide 1, with the intent of reducing the anionic charge of the pT residue. As reported herein, during the course of these latter studies, unexpected peptide byproducts were formed that exhibited exceptional binding affinities. A co-crystal structure of the highest affinity peptide byproduct bound to Plk1 PBD unambiguously identified the structure of the byproduct as a His-adduct and revealed a new mode of binding interaction. Treatment of cultured cells with a Pmab-containing PEGylated variant of this peptide resulted in Plk1 delocalization from centrosomes and kinetochores, and chromosome misalignment that effectively induced mitotic block and apoptotic cell death. These findings could potentially be useful in the design of PBD-binding ligands.

RESULTS

Initial peptide synthesis and evaluation

In order to construct a library of phosphodiesters based on 1,[23] we applied Mitsunobu coupling chemistries[33] to precursor peptides bound to acid-sensitive solid-phase resin. These peptides bore global protection of all reactive heteroatoms, except for a single free phosphoryl hydroxyl group, which was the intended site of condensation with substrate alcohols (Supplementary Methods, Supplementary Results, Supplementary Fig. 1). A variety of alcohols were employed for esterification, including short alkyl chains bearing terminal diol, carboxyl, alkenyl, thiofuranyl and phenyl substituents and progressively longer n-alkyl-1-ols having terminal phenyl rings. The peptides were cleaved from the resin under acidic conditions and the expected phosphodiesters (3a – 3l) were obtained as the main reaction products (Supplementary Fig. 1). Unexpectedly, in each case a faster eluting (HPLC) minor byproduct of unknown structure was obtained (designated as 4a – 4l, respectively) that exhibited a molecular weight identical to the expected product (Supplementary Fig. 1, Tables 1 and 2).

Evaluating the Plk1 PBD binding affinities of the synthetic products using an ELISA-based 96-well assay (Supplementary Methods), showed that all expected phosphodiester products 3a – 3l displayed measurable and in some cases, good affinity. However, the reaction byproducts (4c – 4l) (with the exception of 4a and 4b, which were tested as mixtures with the corresponding 3a and 3b) uniformly displayed significantly higher affinities than their corresponding phosphodiester counterparts (Table 1; Supplementary Figs. 5 and 6). The affinities of n-alkylphenyl byproducts increased roughly with lengthening of the alkyl chain (with the exception of n = 6 and n = 7) and reached a maximum for n = 8 (4j). Chain extension beyond this length resulted in a reduction in binding affinity (4k, 4l). Although some variation in IC50 values was observed from assay to assay, the affinity of the most potent analogue (4j) consistently exceeded that of 1 by a significant amount, that in some cases was approximately three orders-of-magnitude (1, IC50 = 36μM; 4j, IC50 = 17 nM, Table 1).

Table 1

Structures and Plk1 PBD binding IC50 valuesa,b

No.	X	R1	R2	R3	IC50 (μM)
1	O	H	H	CH3CO	36
2	CH2	H	H	CH3CO	174
3a	O		H	CH3CO	>200c
3b	O		H	CH3CO	>200d
3c	O		H	CH3CO	>200
3d	O		H	CH3CO	>200
3e	O		H	CH3CO	>200
3f	O		H	CH3CO	>200
3g	O		H	CH3CO	200
3h	O		H	CH3CO	66
3i	O		H	CH3CO	94
3j	O		H	CH3CO	11
3k	O		H	CH3CO	154
3l	O		H	CH3CO	70
4a	O	H		CH3CO	>200c
4b	O	H		CH3CO	>200d
4c	O	H		CH3CO	>200
4d	O	H		CH3CO	35
4e	O	H		CH3CO	19
4f	O	H		CH3CO	11
4g	O	H		CH3CO	4
4h	O	H		CH3CO	0.055
4i	O	H		CH3CO	0.12
4j	O	H		CH3CO	0.017
4j(S4A)	O	H		CH3CO	12
4k	O	H		CH3CO	0.10
4l	O	H		CH3CO	0.13
5	CH2	H	H		45
6	CH2	H			0.03
6(S4A)	CH2	H			9

Values were obtained from ELISA data presented in Supplementary Fig. 5 and 6. Due to assay variability, numerical IC50 values are for relative comparison.

1(pT5T), 1(S4A), 2(S4A), 3j(S4A) and 5(S4A) did not show detectable level inhibition at the concentrations tested.

Peptides 3a/4a were evaluated as a single mixture.

Peptides 3b/4b were evaluated as a single mixture.

Because adding long chain n-alkylphenyl groups to 1 introduces significant hydrophobic character, we considered the possibility that “promiscuous” mechanisms unrelated to specific interactions with the PBD could give rise to apparent high binding affinity of the byproducts.[34, 35] To address this question we made use of the fact that the “SpT” dipeptide motif is critical for specific high affinity Plk1 PBD-binding and that replacement of the serine residue by an alanine significantly reduces affinity.[20] Therefore, we prepared the corresponding analogues of 3j and 4j in which the serine residue was replaced with an alanine residue [3j(S4A) and 4j(S4A), respectively] and we observed that this resulted in a significant loss of affinity (Table 1). Previous work showing that high affinity peptides can retain some portion of binding affinity following serine/alanine replacement,[19] was consistent with the observed activity of the 4j(S4A) peptide, which could be attributable to the presence of substantially increased serine-independent interactions. Overall, the data argued strongly that binding of 3j and 4j was specific in nature.

Identification of peptide byproducts as histidine adducts

In order to identify the structure of the highest affinity byproduct (4j), tandem MS analyses were performed on both 4j and its associated phosphodiester product (3j) (Supplementary Methods, Supplementary Figs. 2 and 3, Tables 3 and 4). As had been anticipated, the mass spectral data for 3j was consistent with the intended phosphodiester. However, it was found that the fragmentation of the byproduct 4j was best explained by placement of the C6H5(CH2)8– group on the histidine residue. The histidine side chain consists of a (1H-imidazol-4-yl) ring that presents two nitrogen atoms as potential sites of alkylation. It was not possible from the tandem MS data to determine on which of the two histidine nitrogens alkylation had occurred.

Identification of a new PBD-binding channel

To unambiguously identify the site of the histidine alkylation and to understand the basis for the high binding affinity of 4j, the co-crystal structure of Plk1 PBD in complex with 4j was solved (Supplementary Methods, Supplementary Table 7, Supplementary Fig. 7). This structure confirmed the earlier tandem MS results, showing that alkylation had occurred on the histidine residue. It also showed that the C6H5(CH2)8– group was attached to the δ1 nitrogen (N3) on the imidazole ring. Based on this data, it is assumed that the remaining members of the 4-series of peptides also have placement of their respective alkyl groups at this position as well.

The PBD protein backbone in the PBD•4j complex as well as the peptide ligand were shown to be nearly superimposable with the previously reported[23] Plk1 PBD complexed to 1 (Fig. 1a). Differences in the two structures arose primarily from the binding of the C6H5(CH2)8– group of 4j, where the alkyl chain extended from the histidyl imidazole ring and traversed laterally across a series of antiparallel β-sheets (β1 – β4) of the PB1 unit. Binding interactions of the adduct occurred in a well-formed hydrophobic channel whose floor is comprised proximally by V415 (arising from the β1 sheet) and distally by F482 (arising from the αB helix) and whose opposing walls are defined by Y417 (arising from the β1 sheet) and Y485 (arising from the αB helix). The terminus of the channel is formed by L478 and Y481 (arising from the αB helix) (Supplementary Fig. 8a). Formation of this binding channel required very little movement in the side chain orientations of Y485 and F482 relative to the parent 3HIK structure and more pronounced, yet still modest movement in the side chain of Y417. However, the most dramatic movement occurred in the orientation of the Y481 aryl ring, which rotated significantly downward relative to the 3HIK structure (Fig. 1b). This movement had profound effects on the topology of the protein surface, resulting in the revelation of a new binding channel, which had previously been occluded (compare Fig. 1c and Supplementary Fig. 8b). The availability of this hydrophobic channel was unanticipated based on previous crystal structures.

Figure 1

X-ray co-crystal structures of Plk1 PBD complexed with peptides 1 and 4j. (a) PBD in complex with 1 (PBD 3HIK; protein backbone and peptide shown in red) superimposed on the complex with 4j (protein backbone in grey with peptide 4j colored by atom). Key protein structural features are labeled as indicated in reference 21. (b) Plk1 PBD complex with 4j (protein backbone in blue ribbon) showing residue side chains involved with the binding of the C6H5(CH2)8– group of 4j (ligand in yellow with protein carbons in grey) compared with the same residues in the 3HIK structure of PBD-bound 1 (shown in red). Displacements (in degrees) are shown for the Y417 and Y481 phenyl groups. (c) Electrostatic surface of PBD in complex with 4j with coloring based on an arbitrary electrostatic potential scale (positive = blue; negative = red). Peptide 4j is rendered as thick sticks and colored by atom (blue = nitrogen; yellow = carbon; tan = phosphorus and red = oxyben). Graphics were generated using ICM Chemist Pro by Molsoft, Inc. (www.molsoft.com).

Peptide PEGylation

Microinjection of the Pmab-containing peptide 2 into HeLa cells interferes with proper subcellular localization of Plk1 and induces apoptotic cell death as a result of prolonged mitotic arrest.[23, 27] However, in the current work direct incubation of 2 with cultured HeLa cells (at up to 200 μM concentration) failed to elicit a detectable cellular response. This failure was potentially due to limited intracellular bioavailability arising both from poor solubility and from low membrane transport.

Incorporation of polyethyleneglycol into a molecule (termed “PEGylation”) is known to be a valuable approach toward enhancing pharmaceutical properties.[36] Although historically, it has been applied to large constructs such as proteins (for example, see[37, 38]) and nanoparticles (for example, see[39]), the application of PEGylation to smaller entities, such as peptides (for example, see [40, 41]) and organic molecules (for example, see [42, 43]) is also known. Thus, we prepared N-terminal PEG conjugates of 2 (peptide 5) and the Pmab-containing variant of 4j (peptide 6) as well as their serine to alanine replacement analogues for use in whole cell studies (Supplementary Methods, Supplementary Table 5, Supplementary Fig. 4). We observed that both the non-PEGylated 4j and its PEGylated form (6) exhibited similar levels of PBD-binding affinities in ELISA assays (Table 1; Supplementary Fig. 6d) and in fluorescence polarization (FP) competition binding assays (Supplementary Methods, Supplementary Fig. 9, Supplementary Table 8).

Plk-binding specificity by FP techniques

To test for Plk1 specificity of the PEGylated peptides, we prepared appropriate FITC-labeled peptides (Supplementary Table 5) and performed direct FP binding assays, in which the simple construct, FITC-PEG-amide (7) served as a negative control (Supplementary Methods, Supplementary Fig. 10, Table 9). The data showed that relative to the FITC-containing version of PEGylated-1 (peptide 8, Kd = 59.8 ± 4.8 nM), the Plk1 PBD-binding affinity for FITC-derivatized 6 (peptide 9, Kd = 2.0 ± 0.2 nM) was markedly higher. (It should be noted that the Kd = 2.0 nM for peptide 9 was derived from the binding curve by using 2.0 nM of the peptide. This value is an over-estimation due to receptor depletion: The actual Kd of 9 is expected to be < 2 nM. However, precise experimental determination of this Kd would require the use of ligand concentrations < 2 nM, which are accompanied by an insufficient signal to noise ratio.) These assays also showed approximately two orders of magnitude less affinity for 9 against the Plk2 PBD (Kd = 194.2 ± 39.8 nM) and Plk3 PBD (Kd = 460.1 ± 99.2 nM) (Supplementary Table 9). Binding was PBD-specific, since deletion of the phosphoryl group in 8 [8(pT5T)] or introducing a serine to alanine replacement in 9 [9(S4A)] resulted in a significant loss of binding affinity. Specificity was also supported by results from FP competition assays using non-FITC-containing peptides, where 4j(S4A) and 6(S4A) showed significant affinity reductions relative to the parent peptides (Supplementary Table 8).

Cell lysate Plk pull-down assays

To compliment the FP binding results, we employed direct pull-down assays using cell lysates. For this work, we introduced an N-terminal Cys residue onto 1, 1(pT5T), 4j and 4j(S4A) via linkers (peptides 10, 10(pT5T), 11 and 11(S4A), respectively) as well as the PEGylated peptide 6 and its (S4A)-variant (peptides 12 and 12(S4A), respectively) (Supplementary Table 5) and then covalently conjugated the Cys residues to SulfoLink coupling resin (Supplementary Methods). Results showed that in PBD pull-down assays using transfected 293T cells, both constructs 11 and 12, made with peptides 4j and PEGylated 6, interacted with Plk1 approximately 40-fold better than the parent 10, made with control peptide 1 (Fig. 2a; Supplementary Fig. 15).

Figure 2

Specific inhibition of the function of Plk1 PBD by peptide 6. (a) Mitotic 293T cell lysates expressing kinase-inactive Flag-Plk1 (K82M), Flag-Plk2 (K108M), or Flag-Plk3 (K52R) were mixed and incubated with the indicated compounds covalently conjugated to SulfoLink coupling resin through Cys-(CH2)6-CO linker [for 10, 10(pT5T), 11, and 11(S4A)] or an N-terminal Cys residue [for 12 and 12(S4A)]. Precipitates were separated, immunoblotted, and stained with Coomassie (CBB). Arrows indicate Plk1, 2, and 3 proteins. Numbers indicate the relative amounts of precipitated proteins. (b–e). HeLa cells released from a thymidine block and treated with 200 μM of the indicated compounds were quantified to determine the fraction of mitotic cells with rounded-up morphology (b). Bright-field view (c) and fluorescence of immunostained cells (d) used to quantify aberrant mitotic cells with abnormal spindle/DAPI morphologies among total mitotic population (e). Symbols in (d): Asterisks, centrosomally-localized Plk1 signals; arrowed brackets, kineotchore-associated Plk1 signals; arrowheads, misaligned chromosomes. Note that Plk1 signals are almost completely delocalized from the centrosomes and congressed chromosomes, but rather accumulated at the kinetochores of misaligned chromosomes near the poles, as previously described (see[48]). (f) HeLa cells releasing synchronously from S phase were treated with BI 2536 and analyzed (Supplementary Fig. 14). The data in (b), (e), and (f) represent mean values +/− s.d. (bars) from three independent experiments.

Cell culture assays

In cultured HeLa cells, we observed that 6, but not 6(S4A), effectively inhibited cell proliferation in a dose-dependent manner with an IC50 value of 380 μM (Supplementary Fig. 11). The relatively high IC50 value may be due to low cell permeability, since the cellular uptake of the FITC-labeled construct (9) was less than 0.4% (Supplementary Fig. 12). PEGylation may have increased water solubility, thereby allowing higher concentrations of peptide to be used.

Subsequent experiments revealed that treatment of HeLa cells with 200 μM of 6, but not 6(S4A), effectively induced mitotic arrest and apoptotic cell death, while treatment of cells with low concentrations (50 μM and 100 μM) of 6 induced these defects weakly (Fig. 2b, c; Supplementary Fig. 13). Compound 5, but not 5(S4A), also induced a weak but significant level of mitotic arrest under these conditions. As a consequence of the increasing level of apoptotic cell death following mitotic arrest, the number of arrested cells shrank at later time points (Fig. 2b–c). In contrast to the biological activities of 5 and 6, non-PEGylated 2 and 4j failed to exhibit a detectable level of cellular effects under the same conditions (Supplementary Fig. 13), even though their in vitro PBD inhibitory activities were comparable to those of their respective PEGylated forms (Table 1; Supplementary Fig. 6d). As would be expected if the observed mitotic arrest was the result of inhibition of the function of PBD, treatment of HeLa cells with 6, but not with 6(S4A), induced drastic Plk1 delocalization from centrosomes and kinetochores, and severe misaligned chromosomes (Fig. 2d–e and data not shown).[44] Closely correlating with the degree of PBD binding, 5, but not the 5(S4A) variant, induced only mild Plk1 delocalization with a moderate level of misaligned chromosomes (Fig. 2d–e). Unlike the specific inhibition of mitotic progression by 6, treatment of HeLa cells with a previously characterized Plk1 catalytic inhibitor, BI 2536,[13] induced a greatly delayed, but pronounced mitotic arrest and apoptotic cell death (Fig. 2f; Supplementary Fig. 14). These observations suggest that, although more potent than 6, BI 2536 interferes with various uncharacterized processes during the early stages of the cell cycle.

DISCUSSION

The original intent of this study was to determine whether conversion of the dianonic pT phosphoryl group to monoanionic phosphodiesters could be achieved with retention of PBD-binding affinity. Although we did find that depending on the ester group (for example 3j, R = C6H5(CH2)8–), affinity equal to the parent pT-containing peptide was possible, the most significant aspect of the current work was the unanticipated finding that histidine residues bearing long chain alkylaryl groups on the δ1 nitrogen (N3) of the imidazole ring could impart exceptional binding affinity. This affinity enhancement was achieved through new PBD-ligand interactions that took advantage of a previously occluded hydrophobic binding channel on the surface of the PBD. (While this manuscript was under review, an independent approach was reported for identifying a related binding mode.[45]) We further found that N-terminal PEGylation of short (5-mer) peptides did not deleteriously affect PBD-binding affinity, and that PEGylated peptides exhibited enhanced activity when given to cells in culture. The low uptake of PEGylated peptide and the observation that the potency in cellular systems is less than would be expected based solely on PBD-binding affinity, indicates that the affect of PEGylation may be to increase water solubility, rather than to increase cellular bioavailability. Although further improvement in membrane permeability is likely required to increase the efficacy of the compounds, the unexpected new binding interactions identified in this work could impact the future design of PBD-binding antagonists. Our current results provide proof-of-principle that specific inhibition of the function of Plk1 PBD is sufficient to induce mitotic arrest and apoptotic cell death. Since Plk1 over-expression is closely associated with tumorigenesis in a wide range of cancers in humans[6-8] and PBD is essentially required for Plk1 function,[46-48] this study may provide a new paradigm for the design and discovery of PBD-specific Plk1 inhibitors.

METHODS

Methods and associated references are available in the Supporting Information available on the Nature Chemical Biology website.