A fluorescent reporter of ATP binding-competent receptor kinases

Abstract

ERBB receptor kinases play a crucial role in normal development and cancer malignancies. A broad range of modifications creates receptor subpopulations with distinct functional properties in live cells. Their apparent activation state, typically assayed by tyrosine phosphorylation of substrates, reflects a complex equilibrium of competing reactions. With the aim of developing optical tools to investigate ERBB populations and their state of activation, we have synthesized a fluorescent ‘turn-on’ probe, DMAQ, targeting the ERBB ATP binding pocket. Upon binding, probe emission increases due to the hydrophobic environment and restricted geometry of the ERBB2 kinase domain, facilitating the analysis of receptor states at low occupancy and without the removal of unbound probes. Cellular ERBB2 autophosphorylation is inhibited with saturation kinetics that correlate with the increase in probe fluorescence. Thus, DMAQ is an example of a new generation of ‘turn-on’ probes with potential applications in querying receptor kinase populations both in vitro and in live cells.

Dysregulation of the family of epidermal growth factor receptors tyrosine kinases, ERBB1 (also EGFR or ErbB1), ERBB2, ERBB3 and ERBB4, plays a key role in a broad range of solid tumor malignancies making them the subject of intense ongoing research.^1–3 Over the past decade, several small molecule drugs aimed at the ERBB family have been developed, yet their ability to inhibit their target is difficult to reconcile with their predicted binding mode. A well-documented example with clinical relevance is the pan-ERBB kinase inhibitor, lapatinib, for which inhibition constants differ by more than two orders of magnitude between some of the ERBB receptors despite identical molecular contacts in co-crystal structures of the inhibitor and kinase domains.⁴ Lapatinib is a member of the quinazoline class of inhibitors (that also includes gefitinib and erlotinib), which target the ATP binding fold of the ERBB family. The variation in their efficacy amongst patient populations and development of resistance is a major challenge towards effective treatment of ERBB-linked cancers and highlights the need to develop a complete picture of ERBB activation both in solution and in live cell settings.⁵ In addition, dysregulated variants of ERBB2, such as N-terminally truncated species,⁶ have emerged as indicators of poor disease prognosis, and it is thought that such hyperactive subpopulations contribute to the disease phenotype in ways that are very disproportional to their abundance.

The readily accessible and commonly used surrogate to study the levels of activated ERBB is the measurement of tyrosine phosphorylated substrates, including the tyrosine phosphorylation of the receptors themselves. However, this approach cannot differentiate between contributions from high levels of marginally activated receptors and small populations of hyperactive receptors, a distinction with significant mechanistic and ultimately clinical relevance. In addition, ‘pseudoactivation’ occurs when the balance of tyrosine phosphorylation and dephosphorylation is upset. This is for example the case when reactive oxygen species (ROS) inhibit cellular tyrosine phosphatases. As a consequence, elevated ROS can trigger a rapid ‘pseudoactivation’ of ERBB receptors that rivals ligand stimulation based on phosphotyrosine levels, but with fundamentally different signaling properties and modes of signal propagation.⁷ Many aspects of this complex interplay are not readily accessible with current methodologies and new methods to obtain direct information on the state of the receptors under different cellular conditions would be very desirable.

Fluorescent probes have proven to be invaluable tools for assessing biomolecular structure, function and activity.^8–10 So-called ‘turn-on’ probes are of particular interest as they can report specific binding events without the need for rinsing steps to remove unbound probe. This dramatically simplifies assays employing such probes and enables their use in more complex environments such as live cells or tissue. In addition, the lack of a requirement for free probe removal should facilitate the use of high specificity but low affinity probes that report on the equilibrium of receptor states with minimal perturbation of the equilibrium under investigation. As a first and critical step towards such probes, we have synthesized a ‘turn-on’ probe, 6-(4-dimethyl-aminostyryl)-N-benzylquinazolin-4-amine (DMAQ, Fig. 1), which targets the ATP binding site of ERBB2 with desirable micromolar affinity while retaining the specificity of the 4-aminoquinazoline family of derivatives that constitute a large class of ERBB inhibitors with well-characterized pharmacology.^1,11–13 N-phenyl derivatives such as gefitinib or erlotinib as well as N-benzyl derivatives such as 1¹² that exhibit micromolar to nanomolar affinities for ERBB members. In a general sense, DMAQ may be viewed as a fluorescent adenosine analog yet lacks the ribose unit found in other fluorescent adenosine mimics.^13–16 In designing DMAQ, we considered the need to combine both an optical reporting element and a pharmacophore with demonstrated affinity towards members of the ERBB family. Incorporation of the optical reporter was achieved by extending the conjugation of the quinazoline core via a dimethylamino-substituted styryl arm. This modification is not expected to alter the binding mode of the quinazoline core based on the molecular orientation of several quinazoline-based inhibitors observed in ERBB kinase crystal structures.¹⁷ The electron withdrawing nature of the quinazoline functionality coupled opposite the electron rich styryl substituent creates a compact donor-π-acceptor system with attractive physical and optical properties similar to 4-N,N-dimethylamino-4′-cyanostilbene (DCS).¹⁸ The inherent fluorescence of DMAQ avoids the need to couple an external fluorophore eliminating additional steric bulk of the fluorophore and tether. Synthesis of DMAQ was achieved in two steps via condensation of benzyl amine with 4-chloro-6-iodoquinazoline followed by the palladium catalyzed coupling of N,N-dimethyl-4-vinylani-line (characterization data provided in the ESI).

Figure 1.

Design strategy for a ‘turn-on’ fluorescent ligand: incorporation of a push–pull chromophore similar to DCS and an ERBB-targeting pharmacophore resembling 1, yields DMAQ, which is comparable in overall size to gefitinib.

To gain an understanding of how DMAQ might behave in solution and the bound states, UV–vis and fluorescence spectroscopy were carried out in PBS, octanol and PEG (Fig. 2A). Bound ligands may experience a more rigid and potentially less polar environment that can reasonably be approximated for solution spectroscopy by polyethylene glycol (PEG) and octanol.¹⁹ The lowest energy optical transition of DMAQ in PEG is centered at 380 nm with ε = 22,300 M⁻¹ cm⁻¹, while in octanol a slight blue shift (λ_{max, abs} = 374 nm) and hyperchromic shift (ε = 24,300 M⁻¹ cm⁻¹) was observed. Thus, DMAQ is comparable to commonly used fluorescent probes such as DAPI or coumarin derivatives and may take advantage of existing filter sets or laser lines for imaging purposes; at 405 nm, the excitation of commonly available violet lasers, ε ≈ 15,000 M⁻¹ cm⁻¹. In PBS, a marked hypsochromic shift is observed with λ_{max, abs} = 324 nm, which is indicative of H-aggregate formation.²⁰ The emission intensity in PEG is dramatically enhanced relative to PBS with quantum yields of photoemission (Φ_em) of 0.39 and 0.017, respectively; an identical enhancement was observed in octanol. This represents a 23-fold increase in emission intensity upon going from aqueous solution to a more viscous environment that limits the nonradiative relaxation of DMAQ. This is comparable to ON–OFF ratios observed for molecular beacon constructs.^21,22 This emission enhancement may be the result of two simultaneous phenomena. First, the absence of ground state H-aggregates reduces the likelihood of intermolecular quenching. Additionally, stilbene-like dyes possessing both electron rich and electron deficient substituents are highly sensitive to solvent polarity and viscosity due to the formation of intramolecular charge transfer (ICT) excited states.²³ In the case of DMAQ, the dimethylamino group serves as the electron donor while the quinazoline moiety functions as the acceptor due to the presence of two electron-withdrawing, sp²-hybridized nitrogen atoms. An additional mode of emission enhancement was anticipated if probe DMAQ docked in the catalytic site of ERBB2; the confined space of the binding pocket could serve to reduce the likelihood of a twisted intramolecular charge transfer (TICT) excited state, leading to an enhancement of emission. Similar effects have been observed for photoexcited stilbenes such as DCS confined to the binding site of antibodies.¹⁸ Inspection of the frontier molecular orbitals (FMO) of DMAQ and DCS reveals that these two molecules share a similar FMO topology (Fig. 2B) with nearly identical energies for the HOMO → LUMO transition suggesting that DMAQ would exhibit similar photophysical behavior to DCS.

Figure 2.

(A) UV–vis and emission spectra of DMAQ in PEG and PBS. Emission is enhanced 23-fold in PEG relative to PBS. The lowest energy electronic transition correlates well with the predicted energy (DF B3LYP 6–31G*). (B) The frontier orbital distribution and energies of DMAQ compare are remarkably similar to DCS, a well-studied stilbene that functions as a ‘turn-on’ fluorophore, suggesting similar photophysical behavior for DMAQ.

In order to evaluate whether the solvent dependent increase in quantum yield is also reflected in increases in fluorescence in receptor binding, we incubated DMAQ with full length, His-tagged ERBB2 receptors which were extracted under mild lysis conditions and subsequently enriched on and eluted from Ni-NTA beads in PBS/EDTA. Differential fluorescence 3D scans of samples identified 375 nm excitation and 450 nm emission as the conditions for the best differential signal after addition of DMAQ. Under these physiological salt conditions, DMAQ gives minimal fluorescence in the absence of affinity purified receptors but exhibits a peak emission that is increased 23-fold over the free probe in PBS. Characterization of the emission and excitation spectra of ERBB2 bound DMAQ shows that its fluorescence characteristics resemble those of DMAQ in PEG with an excitation maximum of 375 nm for maximum emission at 450 nm (Fig. 2A).

In order to correlate the ERBB2 dependent fluorescence of DMAQ in vitro with its ability to bind ERBB2 receptors in a cellular setting, we tested the potency of DMAQ as an inhibitor of ERBB2 autophosphorylation in live cells. ERBB2 overexpressing BT474 breast cancer cells were incubated with DMAQ at various concentrations for 12 h (Fig. 3A and B). The autophosphorylation of ERBB2 at Tyr-1139 (normalized for ERBB2 levels) was evaluated by Western Blotting. The K_i for the inhibition of autophosphorylation (3.1 μM) correlates well with the saturation behavior of fluorescence on enriched ERBB2 preparations (K_d = 2.9 μM). Since DMAQ targets the ATP binding pocket of ERBB receptors, we first evaluated the extent to which it may target ATP binding pockets in a relatively generic manner. Tyrosine and serine/threonine kinases, while functionally relatively distinct share most common structural characteristics in terms of essential regulatory and catalytic elements,²⁴ a fact that is reflected in the potency of well established pan-kinase inhibitors such as stauroporine which binds to many protein kinases with high affinity but low specificity.²⁵ We tested the impact of DMAQ on the activity of PIK3/AKT/mTOR pathway in the setting of an MCF7 cell with constitutively active and ERBB independent AKT using the feedback phosphorylation of AKT as an established readout for AKT pathway activation. The specicity of the detection system was validated using the bifunctional (mTOR/PIK3) inhibitor BEZ235. At concentrations in which the autophosphorylation of cellular ERBB2 is fully inhibited by DMAQ, no inhibition of AKT signaling was observed (Fig. 3B). To further evaluate the specificity of DMAQ we tested its ability to inhibit the more closely related SRC tyrosine kinase. SRC is located downstream of many receptor initiated pathways but reciprocally phosphorylates and modulates the activity of receptor tyrosine kinases. We therefore evaluated SRC activity in vitro using recombinant purified SRC and a recombinant biotinylated substrate, the heterogeneous ribonucleoprotein K (HNRNPK). Following incubation, HNRNPK-biotin is removed from the reaction and evaluated for its level of SRC dependent tyrosine phosphorylation. While the known SRC inhibitor PP1 inhibits the tyrosine phosphorylation of HNRNPK by SRC, DMAQ showed no inhibitory effect up to a concentration of 6 μM.

Figure 3.

(A) Saturating titration of DMAQ binding to ERBB2 (λ_ex = 375 nm, λ_em = 450 nm) compared with the relative inhibition of ERBB2 autophosphorylation in BT474 cells. B Representative Western blot of ERBB2 inhibition by DMAQ and Canertinib (CI) and constitutively active AKT after treatment with DMAQ or the dual PI3K/mTOR inhibitor: BEZ-235 (BEZ). (B) The impact on SRC activity was tested in vitro using biotinylated HNRNPK, a known SRC client; PP1 was used as positive control. (C) Titration of DMAQ dependent fluorescence in whole cell lysate containing a fusion protein of soluble ERBB2 kinase domain and mCherry with and without Canertinib pretreatment (CI) prior to lysis. Soluble kinase domain fusions in both samples were standardized by mCherry fluorescence.

System wide screens of inhibitor binding as well as structure comparisons of bound inhibitors to ERBB kinase domains have shown that binding specificity and affinity is not readily predictable based on the sequence conservation. This is also evident by the nature of off-target effects of existing kinase inhibitors.²⁶ The above assays provide some insight into the degree of specificity in binding by DMAQ. However, an even more exhaustive evaluation of potential individual kinase targets does not readily address a key concern that is unique to optical probes, namely the extent to which a large number of incremental and functionally inconsequential binding events to a large number of kinase folds adds up to a non-specific fluorescence signal that rivals the on-target signal. We therefore evaluated whether DMAQ would selectively bind to the truncated kinase domain of ERBB2 when expressed as a soluble mCherry fusion and presented in the context of whole cell lysate. The contribution of ERBB2 kinase domain to the fluorescence was evaluated through pretreatment and subsequent removal of Canertinib. Canertinib is considered specific for ERBB receptors, and furthermore reversibly modifies a unique, catalytic site proximal cysteine in ERBB1, ERBB2 and ERBB4.²⁷ Thus DMAQ induced fluorescence that is suppressed by pretreatment with Canertinib prior to lysis can with a high degree of certainty be assigned to ERBB receptors. The equivalent recovery of soluble ERBB2 kinase domains from Canertinib treated and non-treated cells was confirmed using the fluorescence of mCherry. Despite the low abundance of ERBB2 kinase domains in lysate from less than 10⁶ cells, the characteristic fluorescence of DMAQ could be detected. In Canertinib pretreated cells, emission was suppressed by approximately 80% (Fig. 3C), indicating that at a minimum, 80% of the obtained signal was derived from the ERBB2 kinase domain. The remainder may represent cumulative fluorescence from alternative targets, or incomplete covalent modification of ERBB2 kinase domains by Canertinib, given that it is labile chemically reactive group and inhibitory binding moiety are distinct entities. Alternatively, we cannot exclude an unfavorable but detectable mode of binding that involves covalently coupled and tethered Canertinib but binding pocket localized DMAQ. The saturation behavior for the fluorescence of Canertinib treated samples would argue against fluorescence from a large number of low affinity targets.

In summary, we have successfully demonstrated the fusion of an optical reporting element with an ERBB-targeted ATP mimic. DMAQ possesses several attractive properties as a novel biochemical tool for investigating dynamics of receptor tyrosine kinases. First, the large increase in fluorescence upon binding facilitates measurements without the need of removal of free probe. Second, DMAQ is cell permeable, which makes it a suitable starting point for the design of state specific probes for both in vivo and in vitro usage. Additionally, the quinazoline core is readily modified through either the optically active styryl arm or the substitution at the pharmacologically relevant 4-amino position. This versatility should enable development of spectroscopically unique probes with tunable affinities or specificities for related receptor tyrosine kinases. Although designed on the structural framework of ERBB ‘specific’ kinase inhibitors, the degree of binding specificity of DMAQ requires further analysis. Additional targets may be hit at higher concentrations and indeed ‘off target’ effects have in many cases emerged as some of the primary discriminating factors in clinical efficacy. Furthermore, large numbers of distantly related molecular ATP binding scaffold could bind collectively with low affinity contributing to a significant response. However, screens against an enriched target, AKT, and against whole cell lysate containing multiple ATP binding sites suggests that DMAQ retains the selectivity seen for related 4-aminoquinazoline inhibitors. While initial tests indicate that the cumulative signal from off target effects may be low compared to the primary, ERBB-derived signal, this ratio will determine the threshold for the number of receptor molecules that can be studied. The concept of an optically responsive pharmacophore may be readily extended to other classes of biomolecules and ligands providing a complementary spectroscopic tool and bioanalytical technique for investigating biomacromolecular structure and dynamics.