MRI Biosensor for Protein Kinase A Encoded by a Single Synthetic Gene

Abstract

PURPOSE

Protein kinases including Protein Kinase A (PKA) underlie myriad important signaling pathways. The ability to monitor kinase activity in vivo and in real-time with high spatial resolution in genetically-specified cellular populations is a yet unmet need, crucial for understanding complex biological systems as well as for preclinical development and screening of novel therapeutics.

METHODS

Using the hypothesis that the natural recognition sequences of protein kinases may be detected using chemical exchange saturation transfer (CEST) magnetic resonance imaging (MRI), we designed a genetically encoded biosensor composed of eight tandem repeats of the peptide LRRASLG, a natural target of PKA.

RESULTS

This sensor displays a measurable change in CEST signal following phosphorylation by PKA. The natural PKA substrate LRRASLG exhibits a CEST-MRI contrast at +1.8 and +3.6 ppm, with a >50% change after phosphorylation with minutes-scale temporal resolution. Expression of a synthetic gene encoding eight monomers of LRRASLG yielded two peaks at these CEST frequencies.

CONCLUSION

Taken together, these results suggest that this gene may be used to assay PKA levels in a biologically relevant system. Importantly, the design strategy used for this specific sensor may be adapted for a host of clinically interesting protein kinases.

INTRODUCTION

Protein kinases play an instrumental role in almost every signaling process in the living cell, and yet, there is no means to visualize these enzymes in vivo with high spatial and temporal resolution. Kinases are drug development targets of great importance for a wide range of diseases including cancer, diabetes, and inflammation, as each disease is linked to the perturbation of protein kinase–mediated cell signaling pathways (1). There are more than 500 different protein kinases identifies in the human genome (2). Protein Kinase A (PKA), also known as cAMP-dependent protein kinase, was the first one discovered in this large family of proteins (3). Since then, PKA has been extensively studied in innumerable systems, mostly in tissue extracts using antibodies or pharmacological inhibitors, which are not always specific (4). Protein kinases catalyze the energetically irreversible step of adding a phosphoryl group (-PO₄^-2) to hydroxyl (-OH) targets. In particular, PKA phosphorylates the hydroxyl group of serines or threonines having the pattern RRX(S/T)Y with the highest specificity towards the peptide sequence LRRASLG (5). This sequence has allowed the development of a genetically encoded sensor for PKA based on fluorescence resonance energy transfer (FRET) (6), which exhibits extraordinary spatial and temporal resolution for cells in culture when using optical microscopy (7). However, due to limited visible light penetration and scattering in tissue, FRET cannot be easily applied in vivo. As an alternative, we sought to develop an MRI sensor that can report on phosphorylation in real-time.

Molecular MRI, with tools like chemical exchange saturation transfer (CEST) reporter genes (8,9), is attractive as it can track metabolite dynamics in vivo with better spatial resolution and anatomic co-registration than traditional techniques. In CEST MRI, a chemical group of interest is identified as one whose resonance frequency is separated sufficiently from water via a chemical shift (10-12). For example, the guanidyl and amide protons of arginine residues resonate at +1.8 and +3.6 ppm from water, respectively, with a separation from water that is sufficient for performing CEST MRI. During CEST MRI, these proton spins are saturated with radiofrequency pulses tuned to their resonance frequency. When these protons undergo chemical exchange with protons of the surrounding water, they transfer saturation to the water protons, thereby significantly reducing the signal generated by the water protons. This new contrast mechanism has been used in a range of applications, including measuring temperature (13), pH (14), enzymatic activity (15), metal ions (16), gene expression (8), glycogen (17), glycosaminoglycan (18), glutamate (19), drug delivery (20,21) and kidney function (22).

Since LRRASLG contains two arginine residues coupled with a peptide (amide) bond, we hypothesized that this peptide could serve as a good CEST agent. Moreover, a polypeptide containing multiple repeats of this sequence should yield amplification of the contrast with an overall detection sensitivity appropriate for in vivo use and a size that would yield ease of handling with standard biochemical techniques. Given that CEST contrast is highly dependent on the proton exchange rate, the presence of a strong negatively charged group, (i.e. PO₄^-2) should slow the exchange rate and consequently reduce the contrast, making this candidate sensor able to detect phosphorylation by PKA. Based on this hypothesis, we designed a genetically encoded biosensor whose proton exchange rates should change sufficiently with phosphorylation by PKA to be detected with CEST MRI. Eight tandem repeats of the base peptide LRRASLG was chosen for the design of the sensor based on the above considerations. Specifically, the gene is translated to a 6.8 kDa protein that allows both high CEST contrast as well as better protein detection and purification with standard biochemical techniques. Such an MRI biosensor may have applications in monitoring stem cell and organ transplants, and screening therapeutics that target this important signaling pathway.

MATERIALS AND METHODS

Compound preparation

Peptides were synthesized by NeoBioSci (Cambridge, MA) and provided as lyophilized powder. All compounds were dissolved in 10 mM PBS (pH=7.4) at the indicated concentration.

CEST-MRI

CEST MRI experiments were performed on a vertical 11.7T Bruker Avance system. Approximately 20 μl of each sample was loaded into a capillary tube and up to 20 samples were scanned at each session to minimize variability as described before (23). A modified RARE (TR/TE=8000/9.4 ms, RARE factor=16, 1 mm slice thickness, FOV=11×11 mm, matrix size=64×32, resolution=0.17×0.34mm, and NA=1-2) including a magnetization transfer (MT) module (B₁=5.0μT/4000ms) was used to acquire CEST weighted images from -5 to +5 ppm (step=0.2 ppm) around the water resonance (0 ppm). To account for inhomogeneity of the B₀ field, the absolute water resonant frequency shift was measured using a modified Water Saturation Shift Reference (WASSR) method (24), using the same parameters as in CEST imaging except TR=1.5 sec, saturation pulse=500 ms, B₁=0.5 μT, and a sweep range from -2 to +2 ppm (step= 0.1 ppm). Data processing was performed using custom-written scripts in MATLAB. After B₀ correction for each voxel using the WASSR data, mean CEST-spectra were calculated from an ROI for each sample. The MTR asymmetry was quantified as MTR_asym=(S^-Δω – S^+Δω)/ S⁰ (S^-Δω and S^+Δω are the water peak differentials due to saturation at the target frequency +ω and its negative –ω, respectively. S⁰ is the unsaturated signal. By using the difference in signal between saturation at the target frequency and its opposite (negative) frequency cancels any effects due to a direct, non-exchange mediated saturation of water.

Phosphorylation assay

PKA and ATP were purchased from New England Biolabs (NEB, Ipswich, MA) and provided in solution buffer. The reactions were completed with a peptide concentration of 1 mM in 1x PKA buffer and 4 mM ATP. Samples were prepared on ice. Then, following the manifacturer's protocol, samples were reacted for the indicated time at 30 °C using a bench-top S1000 Thermal Cycler (BioRad, Hercules, CA) and then heat-shocked at 65 °C for 20 minutes to inactivate the enzyme. Following heat inactivation, samples were stored at 4 °C until further analysis.

Plasmid construction

Plasmids were generated by custom synthesis of a gene encoding 8 tandem repeats of the base sequence LRRASLG, with a 5-histidine C-terminal tag, whose coding sequence is optimized for expression in E. coli (Genscript, Piscataway, NJ). This gene was then subcloned into the pET expression system using TOPO cloning (Invitrogen, Carlsbad, CA).

Protein expression

E. coli: BL21(DE3) chemically competent cells (Invitrogen) were transformed with pET-LRRASLG8-His6. After induction in Magic Media™ (Invitrogen) at 30 °C for 18 hours, the total protein was extracted using Cell Lytic B reagent (Sigma, St. Louis, MO). Lysate buffer was exchanged by passage through a Zeba 7K MWCO spin column (Pierce, Rockford, IL) with PBS (pH=7.4) as the elution buffer.

RESULTS

CEST contrast was measured for the PKA consensus sequences LRRASLG (Figure 1a) and its phosphorylated form, LRRApSLG, which has a phosphoryl group synthetically conjugated to the hydroxyl group of the serine, modeling the product of PKA action on the unphosphorylated substrate (Figure 1b). In Figure 1c, the CEST-spectra – i.e., the normalized signal intensity of water protons – is plotted as a function of the off-resonance saturation frequency for these two peptides at a concentration of 1 mM. For both peptides, the CEST spectra are asymmetric, indicating that these peptides generate CEST contrast, with the unphosphorylated form generating significantly higher contrast. There is a considerable difference in MTR_asym between the unphosphorylated and the phosphorylated peptide, with two distinguishable peaks at 1.8 ppm and 3.6 ppm (Figure 1d) frequency offsets. This difference is directly concentration-dependent down to micromolar concentrations (Figure 1e,f).

Figure 1.

Sequence visualization of (a) unphosphorylated and (b) phosphorylated LRRASLG peptide (phosphoryl group in orange). (c) CEST-spectra and (d) and MTR_asym plots of both peptides at 1 mM. MTR_asym at (e) 1.8 ppm and (f) 3.6 ppm. L=leucine; R=arginine; A=alanine; S=serine and G=glycine.

To test whether this effect can resolve the temporal action of PKA, recombinant PKA was incubated with the peptide LRRASLG for varying amounts of time (Figure 2). A >50% decrease in CEST contrast was seen with kinetics similar to that expected for recombinant PKA activity, with the majority of the signal difference generated within the first 15 minutes of incubation.

Figure 2.

Biosensor phosphorylation by PKA. Shown is the time course of PKA phosphorylation measured with CEST MRI for 1 mM peptide.

Next, an artificial gene that encodes eight tandems repeats of LRRASLG sequence was designed. The gene was cloned into a bacterial expression vector and was introduced into E. coli. Figure 3b shows that a lysate of E. coli expressing the genetically encoded PKA biosensor provides higher MTR_asym compared to a lysate of E. coli expressing a control protein, cytosine deaminase, which shows no measurable CEST contrast of its own. This difference in MTR_asym is more pronounced at the expected peaks of 1.8 ppm and 3.6 ppm (Figure 3c). Taken together, these results demonstrate the feasibility of detecting this genetically encoded candidate PKA biosensor in a biologically relevant environment.

Figure 3.

(a) Design of a genetically encoded biosensor for measuring PKA activity. (b) MTR_asym of lysates of E. coli expressing the PKA biosensor gene (red) and a non-CEST-contrast-generating control (cytosine deaminase, black). (c) The difference between the two spectra in b (dashed-line), with arrows pointing to peaks at 1.8 and 3.6 ppm. MTR_asym maps at (d) 1.8 ppm and (e) 3.6 ppm of cell extracts of E. coli overexpressing cytosine deaminase (left capillary) and extracts of E. coli overexpressing the PKA biosensor (right capillary).

DISCUSSION

Magnetic resonance has been used for detection of different kinases using genetically encoded reporters. Transgenic overexpression of creatine kinase (CK) by Koretsky et. al. (25) and of arginine kinase by Walter et. al. (26) was detected via ³¹P MR spectroscopy. In addition, an innovative design for an MRI sensor for PKA was described based on a transverse relaxation (T₂) contrast mechanism, which requires aggregation of several proteins to generate MRI contrast (27). Here, we describe a biosensor that is encoded by a single gene that may respond directly to PKA activity. We anticipate that these reporters are uniquely suited to non-invasively monitor treatment response to small molecule and cellular therapeutics.

This sensor could potentially be further optimized using molecular tools, such as systematic mutagenesis, to improve the MRI sensitivity and the contrast between the phosphorylated and unphosphorylated state. Similar methods have been used to develop an MRI sensor for dopamine (28). This biosensor can be expressed in numerous target tissues using viral vectors for transduction (29) or in transgenic animals as was demonstrated with other genetically encoded MRI reporters (30,31).

In this study, we have capitalized on the arginine and serine content of consensus recognition sequences of proteins phosphorylated by PKA. Both of these residues are expected to provide high CEST contrast (32). In particular, peptides with basic, positively-charged lysine and arginine residues show high CEST contrast as the exchange rate of amide protons (NH) and amine/guanidyl protons (-NH3⁺/=NH2⁺) is sufficiently fast to generate contrast, without being too fast to merge their resonance peak with that of water. Hydroxyl protons (-OH) of polar serine and threonine residues can also contribute to CEST signal and perhaps enhance contrast by inducing water coordination with nearby basic residues. This suggests that phosphorylation by serine/threonine kinases could measurably alter the exchange rates of substrate protons by 1) altering the exchange of nearby guanidyl and amide protons through coordination by the negatively-charged phosphate group and 2) removing the CEST contrast generating hydroxyl proton of the phosphorylated serine or threonine. Indeed, the peptide described in our experiments exhibited a high CEST contrast in the unphosphorylated state, that is quenched by >50% in the phosphorylated state. Importantly, this effect is conserved across a range of sub-millimolar concentrations that are similar to the expected intracellular concentration of functional reporter gene expression.

As can be seen in Figures 1 and 2, the change in the CEST contrast displays both a linear and non-linear range. Thus, we anticipate that protocols for in vivo applications will have to include acquisition of baseline data prior to manipulation of PKA and are only quantitative in a linear range of concentrations. Our in vitro studies show that the sensor responds to phosphorylation on a timescale of minutes (Figure 2). This is the same timescale that is required for acquisition of the CEST data. However, in vivo, enzymes such as phosphatases can remove the phosphate group from the substrate, and may thus reduce the measured PKA activity before the end of the acquisition. In the future, faster CEST acquisition schemes on a timescale of seconds (33) should dramatically improve the temporal resolution. It also should be noted that the accuracy of in vivo quantification of CEST agents at chemical shifts close to water resonance frequency may be greatly influenced by B₀ and B₁ inhomogeneity. Therefore, the same WASSR protocol used here can be used to correct for B₀ inhomogeneity for in vivo studies, as it is suitable for correcting inhomogeneities of up to 1000 Hz (34). The B₁ inhomogeneity should also be measured separately and corrected if needed as shown previously (35).

PKA is not the only protein kinase that phosphorylates consensus sequences rich in arginine, lysine and serine residues. Among such kinases of biological importance are protein kinase C (PKC) (36) and Ca²⁺/cAMP-dependent protein kinase IIα (CaMKIIα) (37). Even though many kinases phosphorylate sequences with high similarity in amino acid composition, the specific ordering of these amino acids determines the kinase specificity. Thus, for instance PKC and CaMKII will not phosphorylate the substrate for PKA (6) and vice versa (36). This opens the possibility to engineer an array of CEST MRI-based sensors for detecting multiple variants of protein kinases in vivo.