Structure-Guided Strategy for the Development of Potent Bivalent ERK Inhibitors

Abstract

ERK is the effector kinase of the RAS-RAF-MEK-ERK signaling cascade, which promotes cell transformation and malignancy in many cancers and is thus a major drug target in oncology. Kinase inhibitors targeting RAF or MEK are already used for the treatment of certain cancers, such as melanoma. Although the initial response to these drugs can be dramatic, development of drug resistance is a major challenge, even with combination therapies targeting both RAF and MEK. Importantly, most resistance mechanisms still rely on activation of the downstream effector kinase ERK, making it a promising target for drug development efforts. Here, we report the design and structural/functional characterization of a set of bivalent ERK inhibitors that combine a small molecule inhibitor that binds to the ATP-binding pocket with a peptide that selectively binds to an ERK protein interaction surface, the D-site recruitment site (DRS). Our studies show that the lead bivalent inhibitor, SBP3, has markedly improved potency compared to the small molecule inhibitor alone. Unexpectedly, we found that SBP3 also binds to several ERK-related kinases that contain a DRS, highlighting the importance of experimentally verifying the predicted specificity of bivalent inhibitors. However, SBP3 does not target any other kinases belonging to the same CMGC branch of the kinome. Additionally, our modular click chemistry inhibitor design facilitates the generation of different combinations of small molecule inhibitors with ERK-targeting peptides.

The RAS-RAF-MEK-ERK pathway is a major target for drug development in oncology.¹⁻³ Drugs have been mainly designed to target the upstream kinases within the cascade, particularly BRAF and MEK1/2. For example, the BRAF inhibitors vemurafenib and dabrafenib and the MEK1/2 inhibitor trametinib are approved for treatment of late-stage melanoma harboring the BRAF V600E oncogenic mutation.²⁻⁵ Initial melanoma patient responses to these drugs are often dramatic, but drug resistance mechanisms typically cause relapse within a year. Therapies simultaneously targeting different components of the RAS-RAF-MEK-ERK pathway can delay, but generally not overcome, the development of drug resistance.²⁻⁵ Most resistance mechanisms circumvent the inhibited BRAF and MEK1/2, but ultimately still rely on restoring activation of the effector kinases ERK1/2. Thus, recent efforts have focused on the development of direct ERK inhibitors, which should overcome many of the observed resistance mechanisms.² Additionally, ERK inhibitors may be used in combination therapies that could further delay the onset of resistance and in some cases reduce toxicity, as has been shown for combinations of BRAF and MEK1/2 inhibitors.¹⁻³

Reversible kinase inhibitors can be classified into five categories.⁶ Type I and type II inhibitors bind to the ATP-binding pocket in the active (“DFG-Asp in”) or inactive (“DFG-Asp out”) conformation of the kinase, respectively. These ATP competitive inhibitors are by far the most common, but because of the high structural conservation of the ATP binding pocket they generally have low selectivity for the target kinase and can inhibit even distantly related kinases.⁷⁻⁹ Type III and IV are allosteric inhibitors respectively binding to sites near or distant from the ATP binding pocket.⁶ These sites, which include kinase interfaces mediating the binding of substrates and regulatory proteins, can be targeted with higher specificity but often weak binding affinity.^1,6 An emerging new type of inhibitors (type V) involves bivalent targeting of the ATP binding pocket with a small molecule combined with targeting of a second site.^6,10,11 This site often is a protein interaction surface, typically targeted with a peptide.^6,10,11 These bivalent inhibitors are believed to yield a favorable combination of high binding affinity and specificity.

Two distinct protein interaction sites in ERK1/2 play an important functional role by promoting interaction with substrates such as RSK1, the upstream kinases MEK1/2, and regulatory proteins such as PEA15. One of these sites, known as the F-recruitment site (FRS) or DEF docking site, recognizes a short FXFP peptide motif and is accessible only in the active enzyme.^1,12,13 The other site, known as the D-site recruitment site (DRS), is accessible in both active and inactive ERK1/2.¹ It binds linear D-site peptide motifs of interacting proteins (Figure 1A), which are characterized by a sequence of hydrophobic and basic residues that can be found in both a N- to C-terminal and C- to N-terminal orientation in ERK1/2 binders.^1,12−14

Figure 1.

Structure of ERK2 and FR180204. (A) Structure of ERK2 (blue ribbon with surface) bound to ADP (green sticks) and the PEA15 D-site peptide (orange sticks, N- and C-termini are indicated). The ERK2 active site and DRS are indicated by dashed circles (PDB 4IZ5(12)). (B) Chemical structures of FR180204 and FR180204-CH₂–COOH and ERK2 inhibition assay measuring compound potency. The N1 position in the pyrazolopyridazine ring of FR180204 is highlighted in red color. Means with SEM from three independent experiments are shown.

To design a potent and selective ERK1/2 inhibitor, we sought to use a combinatorial approach by linking a small molecule targeting the ERK1/2 ATP binding site with a peptide that binds to the DRS. As the small molecule inhibitor we chose FR180204 because of its simple chemistry and the availability of structural information in complex with ERK2 at the time we started our studies.¹⁵ FR180204 is a Type I ATP-competitive inhibitor with an IC₅₀ value of ∼1 μM for ERK2 in in vitro activity assays and a 10–30-fold selectivity for ERK1/2 over the related p38α mitogen-activated protein kinase (MAPK).^15,16 To chemically link FR180204 to a DRS-targeting peptide, we modified the compound to include a reactive group. Structural and chemical analysis of FR180204 and its binding to ERK2 revealed that the nitrogen atom at the N1 position of the pyrazolopyridazine ring is oriented toward the DRS and is amenable to modification (Figure S1). Additionally, it is in only limited contact with ERK2 (Figure S1B), suggesting that modification of this group should not drastically impair binding, in agreement with previous data.¹⁶

We generated an FR180204-CH₂–COOH derivative in which a CH₂–COOH group is attached to the N1 position of the pyrazolopyridazine ring and measured its potency in an ERK activity assay (Figure 1B). We found that FR180204 inhibits ERK2 with an IC₅₀ value in the 1 μM range (Figure 1B), a value similar to those previously reported,^15,16 while the modified FR180204-CH₂–COOH compound has ∼10-fold weaker potency (Figure 1B). After establishing that we can chemically modify FR180204 while partially maintaining its ERK inhibitory function, we focused on identifying a suitable peptide for conjugation. Based on our previous structural work on the ERK2-PEA15 complex, we decided to use the C-terminal D-site peptide of PEA15 (residues 119 to 130), which we previously found to selectively bind to the DRS of ERK1/2 (Figure 1A) with a K_d of 18 μM.^12,13 To bridge the ∼15 Å gap between the ERK ATP-binding pocket and the DRS, we used a polyethylene glycol (PEG) linker. The combination of FR180204-CH₂–COOH, the PEG linker, and the PEA15 peptide yielded a bivalent ERK inhibitor designated SBP1 (Figure 2A). In the ERK activity assay SBP1 inhibits ERK2 with an IC₅₀ value of 0.7 μM (Figure 2B). This suggests that conjugation to the PEA15 D-site peptide increases the potency of the small molecule moiety, presumably due to the increased avidity caused by the bivalent binding to both the ATP binding pocket and the DRS.

Figure 2.

Structure and potency of bivalent ERK1/2 inhibitors. (A) Chemical structure of SBP1 and SBP2. SBP1 combines FR180204 and the PEA15 D-peptide (residues 119–130) with a PEG linker (shown in gray). SBP2 combines FR180204 and the RSK1 D-peptide (residues 713–729 with an N-terminal Thr-Ala linker) and the TAT peptide sequence (underlined) using a PEG linker. (B) Inhibition of ERK2 activity by the compounds. Means with SEM from three independent experiments are shown.

The PEA15 peptide binds the DRS in a “reverse” orientation with the peptide N-terminus oriented toward the ATP binding site,¹² which allowed linkage of the peptide N-terminus to the small molecule via a short PEG linker. Another peptide using this “reverse” binding mode is the D-peptide of the ERK substrate RSK1.¹⁷ Unlike PEA15, however, which uses a minimal peptide motif for binding to the ERK DRS, RSK1 utilizes an additional helical element that results in >35-fold tighter binding (K_d = 0.5 μM¹⁷) compared to the PEA15 peptide. Additionally, the RSK1 peptide binds to ERK2 with a ∼20-fold tighter binding affinity compared to p38α, attesting to its selectivity.¹⁷ Thus, we generated a new bivalent ERK inhibitor, designated SBP2, by linking the RSK1 D-site peptide (residues 713 to 729, with an N-terminal Thr-Ala linker) to FR180204-CH₂–COOH. We also included a C-terminal TAT sequence (Figure 2A) as an example of a cell-penetrating peptide.^18,19 The IC₅₀ value for inhibition of ERK activity by SBP2 was 14 nM (Figure 2B), representing a 50-fold increase in potency compared to SBP1 and demonstrating the importance of the peptide moiety for the potency of the bivalent inhibitor.

To confirm the expected bivalent binding mode and elucidate the detailed structural features of SBP2 bound to ERK2, we sought to solve the crystal structure of SBP2 bound to active ERK2 (phosphorylated on Thr185 and Tyr187). For this, we first crystallized phosphorylated active ERK2 bound to the ATP analogue AMP-PCP and solved its structure (Figure S2, Table S1). We then developed a method to exchange the AMP-PCP inside the crystals with SBP2 (see Supporting Experimental Procedures), which enabled us to solve the structure of SBP2 bound to ERK2 (Figure 3, Table S1).

Figure 3.

Structure of SBP2 bound to phosphorylated active ERK2 and comparison with bound FR180204 and RSK1 peptide. (A) Structure of ERK2 (blue ribbon with surface) bound to SBP2 (light green sticks, PDB 5V61). The phosphorylated side-chains of ERK2 (Thr185 and Tyr187) are shown as ball-and-stick model. (B) Detail of the ERK2 ATP binding pocket occupied by SBP2. Colors as in panel A. (C) Detail of the ATP binding pocket of unphosphorylated ERK2 (blue ribbon) occupied by FR180204 (light red, PDB 1TVO(15)). (D) Overlay of the peptide moiety of SBP2 (light green) and the RSK1 peptide (violet, PDB 3TEI(17)) bound to the ERK2 DRS (ERK2 from the SBP2-bound structure is shown as blue ribbon with surface).

The structure of the SBP2-ERK2 complex shows the inhibitor bound to the active (“DFG-Asp in”) conformation of ERK2. As expected, the compound bivalently binds to both the ATP-binding site and the DRS of ERK2 (Figure 3A). The FR180204 moiety of SBP2 occupies the ERK2 active site in a manner similar to unmodified FR180204, with small differences, probably due to the fact that we used dually phosphorylated ERK2 whereas the previous study used unphosphorylated ERK2 (Figure 3B,C). Interestingly, the glycine-rich ATP-binding loop adopts a new conformation in SBP2-bound ERK2 compared to the conformations observed in ERK2 bound to FR180204 (Figure 3B,C) or AMP-PCP (Figure S2B). In the ERK2-SBP2 structure, the glycine-rich loop is folded downward into the ATP binding pocket, with Tyr36 engaging in additional stacking interactions with the pyrazolopyridine ring of SBP2 (Figure 3B). A similar conformation of Tyr36 has been observed for ERK2 bound to other ATP-competitive inhibitors, including SCH772984 and several compounds of a series of pyrrolopyrazine-based inhibitors^20,21 and has been correlated to enhanced inhibitor efficiency and selectivity.²² Furthermore, the ERK2-SBP2 structure shows that the peptide moiety binds to the ERK2 DRS in a conformation very similar to that of the isolated RSK1 peptide (Figure 3D). The main differences are in the positively charged C-terminal residues of the peptide. These differences are probably due to conformational constraints caused by the TAT sequence. The TAT sequence, however, is not resolved in the crystal structure, suggesting flexibility and lack of significant interaction with ERK2. The linker between the FR180204 moiety and the RSK1 peptide is only weakly defined by electron density (Figure S3), indicating high conformational flexibility and lack of direct interactions with ERK2.

After establishing that the inhibitor binds to both the ATP binding site and the DRS, we focused on improving the design of the linker. In recent years, click chemistry has emerged as a tool to readily join different chemical entities using standardized and relatively gentle reaction conditions.²³ Thus, we chose a click chemistry approach, which will allow for straightforward linkage of different small molecule inhibitor and peptide combinations. For this, we generated an alkyne FR180204 derivative that contains a reactive alkyne group attached to the N1 position in the pyrazolopyridazine ring with a linker (Figure S4). For the peptide moiety, we used the RSK1 peptide (without TAT sequence) and included an N-terminal azidolysine residue followed by a Gly-Thr-Ala spacer (Figure S4). Cycloaddition of the FR180204 alkyne derivative with the azidolysine peptide successfully yielded SBP3 (Figure 4A). In our ERK2 activity assay, SBP3 showed a similar inhibitory effect on ERK2 as SBP2, with an IC₅₀ of 25 nM (Figure 4B), consistent with the observation that the linkers of SBP2 and SBP3 do not interact with ERK2 in the crystal structure (Figures 3 and 4). Moreover, since SBP3 does not contain the TAT sequence, the substantial increase in ERK inhibition by SBP2 and SBP3 over SBP1 can be attributed to the different affinities of the D-site peptides in these inhibitors and not to an additional effect of the TAT sequence. This is also consistent with the lack of interactions of the TAT moiety with ERK2 observed in the ERK2-SBP2 crystal structure. Importantly, the TAT sequence does not seem to negatively impact the potency of SBP2 compared to SBP3, suggesting that cell-penetrating sequences can be included in this type of bivalent ERK inhibitor without impairing its effectiveness.

Figure 4.

Potency and structure of SBP3 bound to phosphorylated active ERK2. (A) Chemical structure of SBP3. SBP3 combines FR180204 and the RSK1 D-peptide (residues 713–729 with an N-terminal Lys(N₃)-Gly-Thr-Ala peptide) with a click chemistry linker. (B) Inhibition of ERK2 activity by SBP3. Means with SEM from three independent experiments are shown. (C) Crystal structure of SBP3 (green sticks) bound to ERK2 (blue ribbon with surface, PDB 5V62). The phosphorylated side chains of ERK2 (Thr185 and Tyr187) are shown as ball-and-stick model. The SBP3 linker is mostly not resolved by electron density (Figure S5) and is in part indicated by a dashed line. (D) Overlay of the peptide moiety of SBP3 (green) and the isolated RSK1 peptide (violet, PDB 3TEI(17)) bound to the DRS of ERK2 (ERK2 from the SBP3-bound structure is shown as blue ribbon with surface).

The crystal structure of SBP3 bound to active phosphorylated ERK2 (Figure 4C, Table S1, Figure S5) shows that SBP3 essentially adopts the same binding mode as SBP2. Small changes are observed at the C-terminus of the peptide moiety, where SBP3 more closely resembles the conformation of the isolated RSK1 peptide (Figure 4D), probably due to the absence of the TAT sequence in SBP3. Interestingly, in the SBP3-ERK2 structure the glycine-rich loop including Tyr36 is present in two conformations; one resembling the conformation observed in the SBP2-ERK2 complex and the other resembling that in the AMP-PCP-ERK2 complex (Figure S6). This heterogeneity may be due to the shorter soaking time; SBP2 was soaked for ∼7 months, whereas SBP3 was soaked for ∼3 months, which perhaps was insufficient to allow for a complete shift in the position of Tyr36 since the packing of molecules in a crystal can make conformational changes very slow.

The bivalent inhibitors containing the RSK1 peptide show a 50–80-fold increase in potency compared to FR180204, demonstrating the validity of our approach to improve the potency of the small molecule. While bivalent inhibitors targeting other kinases have been reported,^10,11,24 to our knowledge this is the first report of a bivalent ERK inhibitor. Due to the reported specificity of both components of SBP3, namely, FR180204 and the RSK1 peptide, we expected a high selectivity of SBP3 for ERK1/2.^10,11 In a kinase selectivity assay⁷ against 55 kinases of the CMGC branch, which includes ERK1/2 and other MAPKs, we found that SBP3 strongly binds to ERK1/2. However, SBP3 also similarly targets the related MAPKs JNK1, JNK2, JNK3, and p38α as well as, slightly less effectively, p38β (Figure 5, Table S2). These MAPKs contain a DRS with some similarity to the ERK DRS.^14,17 In contrast, SBP3 does not effectively target any other kinase in the CMCG branch (Figure 5, Table S2). Affinity measurements of SBP3 for selected kinases, determined with the same assay used to measure SBP3 selectivity, confirmed that SBP3 tightly binds to ERK2 with a K_d of 16 nM and to ERK1, JNK1, and p38α with K_d values also in the low nanomolar range (Figure S7). As expected from the selectivity assay, SBP3 does not show any detectable interaction (K_d > 2.5 μM) with GSK3β, a kinase of the CMGC branch that does not contain a DRS. These data support the notion that bivalent binding can strongly increase the potency of MAPK inhibitors, in agreement with previous data.^10,11 However, they also suggest that even weak interactions of the two binding moieties of SBP3 with kinases closely related to ERK1/2 may be potentiated in a bivalent inhibitor, somewhat compromising specificity. We also cannot exclude a role of the linker in affecting binding to some of the targeted MAPKs. Thus, our data indicate that it is important to verify the expected selectivity of a bivalent inhibitor against kinases closely related to the target since factors difficult to predict can result in less stringent specificity than anticipated. Strategies to further increase selectivity as well as potency could include not only using a more selective ERK kinase inhibitor but also screening for modifications resulting in interaction of the linker with distinctive features present in ERK and not other MAPKs.

Figure 5.

Kinase specificity of SBP3. Kinase screen measuring the binding of SBP3 (at 2 μM) to 55 kinases of the CMGC branch. The circles indicate “percent of control” values (see Supporting Experimental Procedures), with red circles indicating high SBP3 binding (the circle size is proportional to binding strength) and green circles indicating low or no binding (scores >35%). The values for kinases that strongly interact with SBP3 (scores ≤35%) are listed. Values for all kinases tested are summarized in Table S2. The image was generated using the TREEspot compound profile visualization tool (KINOMEscan, DiscoverX).

In summary, we have generated a novel bivalent ERK inhibitor (SBP3) that combines a small molecule (FR180204) targeting the ATP binding site and a peptide (derived from the ERK substrate RSK1) targeting the DRS. SBP3 tightly binds to ERK1/2 with low nanomolar affinity and exhibits high inhibitory potency in ERK activity assays. While SBP3 can discriminate between ERK1/2 and most of the kinases in the CMGC branch of the kinome, it also efficiently targets several other closely related MAPKs containing a DRS. Moreover, the click chemistry design allows straightforward modular generation of similar inhibitors with different combinations of an ATP-competitive small molecule and an ERK-binding peptide. Several potent small molecule ERK inhibitors that have been recently reported^{20,21,25−27} can analogously be transformed into bidentate inhibitors and might represent more efficient building blocks than the relatively weak inhibitor FR180204. In addition, positively charged peptides, which have been shown to promote cell penetration of MAPK inhibitors,^19,24 could be included in the bivalent ERK inhibitors without loss of potency, as exemplified by our data with SBP2. The use of cell penetrating sequences that preferentially target tumors^28,29 could also enable more selective ERK inhibition in cancer tissue and thus less toxicity compared to indiscriminate ERK-targeting in all tissues.

Glossary

ABBREVIATIONS

AMP-PCP

β,γ-methyleneadenosine 5′-triphosphate

CMGC

cyclin-dependent kinases, mitogen-activated protein kinases, glycogen synthase kinases and cyclin-dependent kinase-like kinases

DRS

D-site recruitment site

ERK

extracellular signal-regulated kinase

FRS

F-recruitment site

JNK

c-Jun N-terminal kinase

MAPK

mitogen activated protein kinase

MEK

MAPK/ERK kinase

PEA15

phosphoprotein enriched in astrocytes of 15 kDa

PEG

polyethylene glycol

RAF

rapidly accelerated fibrosarcoma

RAS

rat sarcoma

RSK1

ribosomal protein S6 kinase alpha-1

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.7b00127.

• Supporting experimental procedures, references, tables, figures and figure legends (PDF)

Author Present Address

^∥ Biochemistry Department, School of Biomedical Sciences, University of Otago, Dunedin 9054, New Zealand.

Author Present Address

^⊥ PPG Industries, Pittsburgh, Pennsylvania 15222, United States.

Author Present Address

^# Cambridge Cancer Center, University of Cambridge, Cambridge, U.K.

Author Contributions

B.C.L. purified and crystallized proteins, solved and analyzed the structures and performed activity assays. P.D.M. established ERK2 protein expression and initial activity assays. E.H.S, R.W., R.S., and G.P.R. developed the click-chemistry approach and synthesized the SBP3 compound. Y.W. helped characterizing compounds. S.J.R and E.B.P. conceived the study and analyzed data. B.C.L, E.B.P., and S.J.R. wrote the manuscript with contributions from the other authors.

This work was supported by NIH grant R01CA160457 (to S.J.R. and E.B.P.), NCI Cancer Center Support grant P30CA030199 (to G.P.R., S.J.R., and E.B.P.), EMBO Long-Term Postdoctoral Fellowship (to B.C.L.), and DOD-BCRP Fellowship BC100466 (to P.D.M.). Data for this study were obtained at beamline X29 of the National Synchrotron Light Source with support from NIH grants P41RR012408 and P41GM103473, and at Stanford Synchrotron Radiation Lightsource beamline 11–1 with support from NIH grant P41GM103393 and U.S. Department of Energy Contract DE-AC02-76SF00515.

^∇ Deceased.

The authors declare no competing financial interest.