Fluorescence lifetime imaging probes for cell-based measurements of enzyme activity

Summary

Post-translational modification (PTM) enzymes are important modulators of protein structure and function. They typically act by chemically modifying amino acids, often on side chain functional groups, to change the physiochemical landscape of the protein and thus its biophysical behavior. In particular, protein kinases are enzymes that transfer phosphate from ATP to serine, threonine, or tyrosine in protein substrates. They are key regulators of vital cellular pathways such as survival, proliferation and apoptosis, and their dysregulation in the context of cancer has been widely investigated for the purpose of development of anticancer drugs. However, several critical questions pertaining to their physiology, such as heterogeneity of kinase signaling within and between cells, and other factors that may play into the mechanisms of drug resistance, remain unanswered. Many of the current strategies to measure kinase activity lack the scope, subcellular resolution and real time monitoring ability needed to obtain the type of information needed about their dynamics and localization in cells. While FRET-based biosensors are capable of dynamic single cell imaging, their applications can be limited by difficulties in multiplexing and the inherent inadequacies of steady state measurements. In this chapter, we describe our fluorescence lifetime imaging microscopy (FLIM) probe technology in which peptide kinase substrates, linked to cell penetrating peptides and labeled with small molecule fluorophores, are used to report kinase activity through time-resolved fluorescence imaging to visualize and quantify changes to the probe’s fluorescence lifetime. These can be multiplexed for more than one kinase at a time, and interpretation is not affected by differences in local intensity due to probe uptake and distribution or photobleaching. With careful choice of peptide substrate(s), fluorophore label, and imaging set-up, high specificity and spatiotemporal resolution can be achieved. Due to the mechanism by which the lifetime change occurs, this approach is compatible with other PTMs (such as acetylation, methylation, etc.), and so the considerations for kinase FLIM probe design described in this chapter should be broadly applicable for other PTMs as well.

1. Introduction

1.1. Biosensors for kinases and cell-based activity analyses

1.1.1. Kinases in cancer.

Protein kinases are a family of enzymes that catalyze addition of a phosphate group to specific amino acid residues, typically serine, threonine or tyrosine, in target proteins. In doing so, they modulate signal transduction pathways governing key biochemical processes of the cell, such as proliferation, migration and apoptosis. Tight regulation of kinase signaling is critical to ensure maintenance of cellular homeostasis and sensitivity to apoptotic stimuli, otherwise cells can proliferate aberrantly and invade other tissues. Genetic and/or epigenetic alterations are known to cause over-expression and/or dysregulation of kinase activity in cancer cells, leading to a range of dysfunctional behavior from oncogenesis to metastasis. In particular, malfunctioning of signaling dynamics of tyrosine kinases has been implicated in cancer pathogenesis. The most well-studied example of this is chronic myeloid leukemia (CML), characterized by the Philadelphia chromosome 9:22 translocation mutation that creates the BCR-ABL oncogene (1). This mutation replaces regulatory regions of ABL with BCR, rendering the Bcr-Abl kinase protein that is expressed constitutively active to cause over-proliferation of immature myeloid leukemia cells (2). The kinase inhibitor imatinib (also known as Gleevec) was the first targeted, FDA-approved small molecule treatment for cancer, and stimulated a revolution in cancer biology and therapeutic development. Currently, there are 52 small molecule protein kinase inhibitor drugs approved by the FDA, the majority of which (75%) target tyrosine kinases (3). However, despite the apparent success of early kinase inhibitors like imatinib, achieving the full potential of these targeted approaches has been confounded by issues that are now being more widely recognized: the role of the tumor microenvironment in modulating kinase signaling, allosteric effects of regulatory proteins in the cellular context on the efficacy of kinase inhibitors in situ, and differential signaling in tumor-initiating stem-like cells relative to the rest of a tumor (4–6). All of these biological complexities highlight the need for analytical tools that will enable evaluation of spatio-temporal dynamics of kinase signaling in more complex settings such as live cells and tissues, particularly multiplexed tools that can measure more than one kinase activity at a time. Such tools could be used in kinase target validation, inhibitor activity characterization, and/or even as secondary screening or companion diagnostic applications.

1.1.2. Methods for kinase activity detection.

A number of techniques are traditionally used to measure phosphorylation activity of tyrosine kinases in cells and in vitro, as we have reviewed thoroughly elsewhere (7). However, most of these technologies lack the combination of subcellular multiplexability and real time monitoring that are needed to answer fundamental questions. For example, while cellular activity of kinases is often heterogeneous from cell to cell, most existing approaches for detection pool cell populations and extract their contents for analysis, forgoing valuable information on cell-to-cell variability and subcellular localization. As we have previously described, peptide artificial substrates delivered via cell penetrating peptides are a viable and attractive means to measure kinase activity in live cells (7), and they can circumvent some of the limitations of genetically-encoded sensors. Linear consensus substrate motifs recognized by kinases have been developed by e.g. screening synthetic peptide libraries (8–11), and mass spectrometry-based phosphoproteomics (12, 13). Using this kind of information, peptide substrates can be designed that are phosphorylated by a target kinase or kinase family and used in vitro or in cell-based experiments (7). However, in order to be useful for real-time cell-based analyses, readout methods that are capable of live cell imaging are needed.

Genetically engineered fluorescent protein biosensor probes have been reported and elegantly used for subcellular, live cell kinase activity imaging analyses (14–16). A number of peptide probes aimed at reporting kinase activity using fluorescence have also been documented (17). Their designs often share a common functional strategy: 1) The probe’s substrate sequence incorporates a motif which is recognized by the kinase, leading to the phosphorylation of a serine, threonine or tyrosine residue on the motif; 2) Phosphorylation of the probe induces a shift in a spectro-physical property of a fluorescent label, via intrinsic changes in the fluorescence signal output or effects on fluorophore quenching/Förster resonance energy transfer (FRET) via e.g. conformational changes to the probe (Figure 1A). Environmentally-sensitive fluorophores can provide intrinsic signal intensity shifts upon probe phosphorylation, while FRET- and quench-based probes typically employ a mechanism by which fluorescence is enhanced upon phosphorylation due to disruption or enhancement of a structure-induced FRET/quench mechanism. However, for most examples, the requirement for multiple fluorophores per probe means the number of kinases that can be measured simultaneously is limited to ~2 (Figure 1B). Genetically encoded biosensors are also limited by context—even those that use a single fluorophore can only be used in cells and tissues where expression of the construct is viable. Also, most of these methods and applications have relied on measuring the intensity of the output in some way—either intrinsic intensity or FRET-based intensity increases. Intensity measurements are inherently confounded by issues like photobleaching and differences in uptake and local concentration of the probe.

Figure 1:

A) Mechanism of phosphodetection in typical FRET biosensors requires two fluorophores, a donor and an acceptor. B) Limited scope of multiplexability due to use of two fluorophores for single read-out.

1.1.3. Fluorescence lifetime imaging probe technology.

Accordingly, we have employed fluorescence lifetime imaging for detecting kinase activity using cell-deliverable peptide substrates in a physiological context (Figure 2A). Fluorescence lifetime (further described in Section 1.2) is a property that is independent of intensity, so it is not affected by differences in the amount of probe or photobleaching that may occur. These substrates are comprised of a kinase activity reporter module (a sequence that is efficiently and selectively phosphorylated by a kinase or kinase family of interest), a transduction module (a cell penetrating peptide), and a fluorophore label typically attached via conjugation to a cysteine residue added to the peptide sequence. In this approach, the peptide is able to interact with endogenous cellular phosphorecognition domains (typically SH2 domains) in either the kinase itself or in other surrounding adaptor proteins, which changes the collisional quenching environment of the peptide’s fluorophore label and usually results in longer lifetimes being observed (Figure 2B). This contrasts with e.g. FRET biosensors, since 1. it requires only one fluorophore per probe and 2. even in time-resolved FRET, typically the output being measured is a decrease in the donor fluorophore’s lifetime (as described in section 1.2.2), as opposed to an increase. This approach is likely to be general for biosensors of post-translational modifications since such modifications almost always result in a change in the biophysical interactions of the molecule with reader domains. We have previously described proof-of-concept for this strategy, demonstrating that it depends on interaction with the kinase protein (18), and also reported examples of multiplexed, quantitative analysis for Abl and Src-family kinases (19). In this chapter, we describe the fundamentals and important considerations for this technology, including factors that should be considered during the design of these probes and the methods to detect and analyze them.

Figure 2:

A) Cell penetrating probes implemented for fluorescence lifetime imaging microscopy (FLIM) and cell-based detection of kinase activity. B) Interaction of the phosphorylated probe with a kinase regulatory domain causes a lifetime shift of the conjugated fluorophore due to reduced collisional quenching.

1.1.4. Probe design consideration

1.1.4.1. Peptide substrate selection.

For the purposes of kinase (or any other enzyme) activity measurement in cells, the correct choice of peptide probe is essential. It is critical that the kinase substrate sequence and cell-penetrating peptide are amenable to the cell type and biological activity to be analyzed. As we have reviewed extensively elsewhere (7), there are many confounding factors that affect the suitability of a given peptide sequence for this type of measurement. Despite being a common practice, it is almost never appropriate to simply take a truncated sequence from a known protein substrate and apply it as an enzyme activity probe. Protein substrates evolved with regulatory features well outside of the modification site that govern the recognition, specificity and selectivity of their modification by enzymes, such as protein-protein interaction domains and other allosteric sites, thus the sequences surrounding their modification sites usually do not confer sufficient selectivity or biochemical efficiency for cell-based assays (7). Accordingly, it is strongly recommended that researchers wishing to design and implement FLIM probes as described here use the guide that we published on “Assays for Tyrosine Phosphorylation in Cells” (7) as an essential companion to this chapter. This is necessary for any application of an enzyme FLIM probe other than the specific probes, stimuli and cell types described in the protocol in Section 3.

1.1.4.2. Fluorophore selection and labeling.

Peptide based biosensors can be fluorescently labeled at strategic locations by incorporating residues functionalized by free amine (lysine or arginine) or sulfhydryl (cysteine) groups. Cysteine residues may be conjugated to a synthetic dye through the maleimide linker via Michael type electrophilic addition reactions (20). While it is more selectivity formed at lower pH than an amino-maleimide bond, the thiol-maleimide bond is somewhat more chemically labile due to its propensity to hydrolyze in the presence of other free sulfhydryl containing molecules, especially in alkaline conditions (pH>8). Intramolecular hydrolysis or ring opening of the succinimide moiety (21) and N aryl substitutions on the maleimide (22) have been shown to enhance the stability of thiol-maleimide conjugates. Nevertheless, the ability to achieve mono-incorporation of fluorophores via cysteine-maleimide coupling makes it advantageous for probe design.

Certain fluorophores are more susceptible to photobleaching than others, owing to their structure. For example, among the green channel fluorophores, superior varieties such as Dylight 488 and Alexafluor 488 are significantly more photostable than fluorescein and Cy2 (23). In addition to photo-stability, other photophysical parameters of potential fluorophores should be considered to evaluate their applicability in lifetime measurements. Fluorophores with high structural flexibility of the skeleton such as carotenoids, retinoids, diazo dyes, diphenyl, triphenylmethanes, and styryl dyes undergo rapid internal conversion, mainly cis/trans isomerization. Their lifetime is known to be sensitive to environmental factors such as temperature, viscosity and polarity of the solvent. Such dyes may be useful for FLIM probes since phosphorylation of the biosensor may induce even larger shifts in the lifetime behavior of the dye. Conversely, dyes that are sensitive to pH (such as fluorescein and BCECF) or prone to aggregation (such as porphyrins), are less likely to be useful since their lifetime may be affected by cellular and biomolecular conditions independent of probe phosphorylation. Others (such as pyrene) self-associate in their excited state, to form dimers (called excimers) which are characterized by drastically shifted emission bandwidths and prolonged lifetimes, but it is unclear whether they would produce useful signal for FLIM probes of enzyme activity. For multiplexing of biosensors, especially careful selection of fluorophores is required to avoid spectral cross-talk across the excitation and detection channels. The use of far red, infrared and near infra-red dyes could minimize cytotoxic effects of photo-damage, as well as light scattering and autofluorescence, as long as the lifetime imaging instrumentation was compatible with their excitation and emission wavelengths.

1.2. Fluorescence lifetime analyses

1.2.1. Fluorescence Lifetime and Decay.

The lifetime of a fluorescent molecule is defined as the duration following excitation over which it returns (decays) to the ground state. There are several pathways by which a fluorophore might lose its excitation, illustrated in the Jablonski diagram (Figure 3). They are broadly divided into two categories: radiative and non-radiative. A radiative pathway is one in which a photon is emitted by the molecule and thus may be detected and quantified by a photo-detector. Non-radiative decay pathways, on the other hand, do not involve photon emission and are therefore difficult to detect. They are generally referred to as “quenching” mechanisms, and typically occur as a result of direct molecular interactions with other molecules. These quenching mechanisms are further subcategorized as either static or dynamic. Static (a.k.a. contact) quenching occurs when the fluorescent molecule forms a non-fluorescent complex in the ground state itself, i.e. before photo-excitation. An example of static quenching is hydrophobic interactions leading to aggregation of the fluorophore molecules, preventing excitation. Dynamic (collisional) quenching refers to loss of excitation due to rotational and translational diffusion of the fluorescent molecule in its excited state leading to collisions with quenchers that drop the fluorophore back into the ground state. Typically, the lifetime of a small molecule organic fluorophore is on the scale of 0.1–10 ns. In certain scenarios, a molecule may transition during its excited state to a hyper-stable triplet state whose lifetime can be as high as hundreds of nanoseconds to a few microseconds. The molecule may then return to its ground state radiatively (phosphorescence) or non-radiatively. When the decay occurs radiatively, it can be observed using time-resolved fluorescence analyses via either time domain or frequency domain measurements (as described below in Section 1.2.3).

Figure 3:

Jablonski diagram showing excitation of a molecule to the singlet state followed by decay through fluorescence, intersystem crossing or non-radiative quenching mechanisms.

1.2.2. Advantages of Lifetime Imaging over Steady State Measurements.

There are several major advantages to using time-resolved fluorescence lifetime measurements in biological experiments. First, the lifetime of a fluorophore is independent of the fluorophore concentration and/or intensity. This avoids confounding effects of things like different levels of analyte and/or photobleaching, which result in different intensities of fluorescence being detected even in situations where the biological change of interest is not occurring. For instance, intensity-based measurements for a phosphorylated biosensor would not be able to distinguish between higher uptake/concentration of the biosensor in that location vs. higher modification to phosphorylated product. Even FRET biosensors are subject to this limitation if acceptor emission intensity is measured as the output (although, as discussed below, time-resolved FRET could overcome it by providing information about the proportion of donor fluorophore that was experiencing FRET vs. not). Photobleaching of fluorophores over a long time-course experiment would also artificially decrease the intensity of a biosensor signal, even if overall the level of phosphorylation was increasing. Second, based on the mechanisms described above for the effects of collisional quenching and/or triplet state transition on a fluorophore, the lifetime of most fluorophores is strongly coupled to their physiochemical micro-environment, e.g. changes in pH, hydrophobicity, protein-protein interactions, etc. During steady state fluorescence measurements, most of the information on transient interactions, conformational changes and dynamics is lost due to averaging of the overall emission intensity. For example, sub-populations of fluorophores that are characterized by different macro-conformations and hence different fluorescence lifetimes cannot be distinguished by steady state methods if the dynamics of biophysical transitions are faster than the excited state lifetime—as is almost always the case. However, such sub-populations can typically be resolved and quantified by time resolved techniques. One application of this in biomedical imaging is in certain Ca²⁺ sensors, fluorescent probes that are quenched due to complex formation with Ca²⁺ ions and used to monitor Ca²⁺ influx within cells. Another example is time-resolved FRET, in which the non-radiative photon transfer to another fluorophore occurs, resulting in decreased lifetime of the donor fluorophore to a degree that is proportional to the distance between the donor and acceptor. Calculating the proportion of different donor lifetimes in the resulting distribution allows the relative amount of the biosensor that is phosphorylated and exhibiting FRET to be quantified. The caveat is that both a donor and an acceptor are required. The FLIM probes for kinase activity described in this chapter only require one fluorophore, where the interaction between the phosphorylated product of the kinase reaction and endogenous phosphotyrosine-binding SH2 domains gives rise to populations of fluorophores with longer lifetimes than the unphosphorylated probes, suggesting that the interaction results in protection of the fluorophores from collisional quenching and/or promotes a stable triplet state transition. Measuring the relative proportions of those populations enables quantification of the amount of probe that is phosphorylated vs. unphosphorylated to infer kinase activity.

1.2.3. Time Domain Acquisition.

Time-Correlated Single Photon Counting (TCSPC) is a high resolution technique used to determine fluorescence lifetime (24). In TCSPC, one measures the time between sample excitation by a pulsed laser and the arrival of the emitted photon at the detector. TCSPC requires a defined “start” signal, provided by the electronics steering the laser pulse or photodiode, and a defined “stop” signal, realized by detection with single-photon sensitive detectors (e.g., Single Photon Avalanche Diodes, SPADs). The measurement of this time delay is repeated many times with picosecond resolution to account for the stochastic nature of fluorophore emission. The delay times are sorted into a histogram that plots the time-gated emission intensities following each excitation pulse to generate a decay curve (Figure 4A).

Figure 4:

A) Time domain acquisition by Time Correlated Single Photon Counting (TCSPC). B) Frequency domain acquisition and quantification of lifetime.

In order to acquire a fluorescence lifetime image, the photons have to be attributed to the different pixels, which is done by storing the absolute arrival times of the photons additionally to the relative arrival time in respect to the laser pulse. To accomplish this, FLIM is often coupled to a laser scanning confocal microscope to take advantage of the superior optical sectioning provided in these instruments. The laser scans across the sample in the xy direction. Line and frame marker signals from the scanner of the confocal microscope are additionally recorded in order to sort the time stream of photons into the different pixels by the absolute arrival time. Thus, not only are photons resolved on the basis of relative emission time, but also spatially on a pixel by pixel basis.

Time domain acquisition via e.g. TSCPC is best suited for deconvoluting multi-exponential lifetime signals, detecting populations with fast decay, and imaging samples with weak fluorescence. However, it does have its own caveat that the relatively slow recording rate and limited dynamic range can be challenging for biological samples. The slow recording rate may be improved by implementing strategies such as the analog mean delay (AMD) method (25). In this scheme, fluorophores are excited with a picosecond laser pulse, and the time-domain waveform of the emitted fluorescence light intensity is measured using slow electronic devices whose 3-dB bandwidths are on the order of a few hundreds of megahertz. A fluorescence lifetime is obtained by calculating the expectation value of photon arrival time, treating the measured fluorescence intensity data as the probability density function of photon arrival in the time domain (26).

1.2.4. Frequency domain acquisition:

An alternative to time domain measurements is the frequency domain FLIM technique (Figure 4B), which requires a modulated light source and a modulated detector (27). The sample is excited by a source that is modulated at a certain frequency. The fluorescence emission mimics this modulation pattern but with a delay in time in the form of a phase-shift. The phase-shift and modulation-depth directly depend on the fluorescence lifetime and the known modulation frequency. To extract the phase shift and modulation depth from the fluorescence emission signal, a homodyne detection method is often used wherein the sensitivity of the detector is modulated (or gated) to match the frequency of the light source. The result is a frequency-domain FLIM signal as a function of the phase difference between light source and detector for each pixel in the image, which can be converted into a lifetime value. In the frequency domain an apparent lifetime (τφ) determined from the phase angle (φω) or the apparent lifetime (τm) determined from the modulation (mω). Frequency domain acquisition offers advantages for measuring many fast biological processes, however, resolution of multiple species via distributions in fluorophore lifetimes is not trivial, and thus the FLIM probes described in this chapter have so far only been applied in time-domain FLIM analyses.

1.3. Imaging hardware requirements

We recommend laser-scanning confocal TCSPC fluorescence lifetime imaging for FLIM probe applications based on its superior performance for per-pixel dissection of populations of fluorophores with different lifetimes. From a practical point of view, the integration of TCSPC requires the following hardware parts (28):

• 1.3.1.
  Pulsed light sources, for example:

  • A picosecond pulsed laser of the desired excitation wavelength. Pulsed laser diodes (e.g., the LDH Series) have the advantage that the laser repetition rate is adaptable to the lifetime of the dye through the laser driver (e.g., PDL 828 “Sepia II”).
  • A pulsed femtosecond laser (usually a Ti:Sa laser) as used in Two Photon Excitation. These lasers are typically operated at a fixed repetition rate of 80 MHz.

  Any laser used should provide a stable trigger output (SYNC) as a reference for the electronics.

• 1.3.2.
  Appropriate microscopic optics. Currently, confocal microscopes from virtually all major microscope manufacturers can be upgraded to provide FLIM functionality, e.g.:

  • Leica SP2, SP5, and SP8
  • Nikon C1, C1si, C2, C2+, and A1
  • Olympus FluoView FV300, and FV1000
  • Zeiss LSM510, LSM710, and LSM780

  The exact upgrading options will depend on the system configuration.

• 1.3.3.Single photon detection modules with appropriate sensitivity and time resolution. Detectors for FLIM imaging can be photomultiplier tubes (e.g., the PMA series), afterpulsing-free hybrid detectors (e.g., PMA Hybrid 40) or avalanche photodiodes (e.g., the PDM module from Micro Photon Devices or the PicoQuant’s t-SPAD). Hybrid detectors and SPADs are sufficiently sensitive even for fluorescence correlation measurements, where single molecule sensitivity is required. PMTs and hybrid detectors can furthermore be mounted in NDD fashion for two-photon microscopy applications.
• 1.3.4.Suitable timing electronics for data registration (e.g. PicoHarp 300).

1.4. Experimental design for fluorescence lifetime-based kinase activity measurements

1.4.1. Peptide sequences.

Peptides should contain an enzyme substrate sequence, a cell-penetrating peptide, and a fluorophore label. They can also include various linkers between the components. For more information about peptide sequence choices, see Section 1.1.4.1. above.

1.4.2. Fluorophore labels.

In our prior work (published and unpublished), we have characterized the fluorophores Cy3, Cy5, DyLight 488 and 550. Dyes Cy3, Cy5 and DyLight 550 have similar cyanine dye-like structures, but DyLight 488 has a fluorescein-like structure that is very different than the cyanine dyes. Based on this structural diversity, we predict that most fluorophores commonly used in biomedical applications should produce similar lifetime behavior when used in these applications. However, it will be crucial for each user to characterize the fluorophore behavior for any new dyes used via appropriate synthetic peptide positive and negative controls (i.e. synthetically phosphorylated, and Y→F mutant unphosphorylatable peptides). This is further discussed below in Section 3.1.2.1. Labeling chemistry is also likely to be general, however due to the cost of fluorophore reagents, labeling post peptide synthesis is usually desirable. Incorporating a cysteine amino acid into the peptide sequence can enable labeling of the synthesized peptide via maleimide-functionalized fluorophores, which are commonly commercially available (e.g. via Lumiprobe, ThermoFisher/Invitrogen, and other vendors). Further discussion of these considerations is provided in Section 1.1.4.2 above.

1.4.3. Cell culture consumables for cellular imaging.

In order to achieve high resolution cellular imaging, choice of culture dishes and/or slides is important. Many options are commercially available, and any culture dish/slide that provides sufficient performance for other types of confocal fluorescence imaging should be compatible with FLIM probe experiments as well. Frequently, however, researchers are limited by the amount of FLIM probe material available, since fluorophore reagents can be expensive. Multiple controls, conditions and replicates are also necessary in FLIM probe imaging experiments. Therefore, specialized dishes and slides with small culture chambers, such as those offered by Ibidi and Nunc, can be advantageous because they allow for smaller volumes to be used during FLIM probe incubation, and parallel preparation of controls and samples.

1.4.4. Positive and Negative Controls.

1.4.4.1. Synthetic peptide controls for fluorescence lifetime signal validation.

While we have observed several fluorophores on multiple tyrosine kinase substrate peptides to exhibit longer lifetimes when phosphorylated, it is still essential to validate any new fluorophore or peptide sequence’s behavior in this type of experiment. Synthetically-modified peptides that contain the modification of interest within the identical substrate sequence (i.e. synthetically-produced enzyme products) can be used as suitable positive controls. When introduced into the cell system of interest, these should exhibit longer lifetimes than their unmodified and/or non-modifiable counterparts. As negative controls, non-modifiable mutants can be used, for example sequences containing a phenylalanine (F) in place of the modification site tyrosine, or in the case of serine/threonine kinases, an alanine (A) can be substituted for the modified residue.

1.4.4.2. Biological controls.

In the context of the biological system, conditions of positive and negative control also need to be established, where the kinase is known to be either activated or inactive, in order to adequately attribute a signal change in the FLIM probe to the biological activity being examined. As we have previously described in great detail (7), this can include stimulation of kinase activity, constitutive activation, and/or inhibition of phosphatases (which dephosphorylate probes). Hyperactivating mutagenesis such as point mutations or deletions of regulatory domains can generate mutant versions that are constitutively active. In such cases, the need for extra-cellular activation is obviated. For example, deletion of select residues within the C2 domain of the phosphatidylinositol-4,5-bisphosphate 3-kinase destabilizes the interaction between its catalytic subunit (p110α), with the regulatory subunit (p85α). Point mutation of the C-terminal tyrosine of the Src kinase destabilizes the inactive conformation, resulting in hyperactivation (29). Conversely, it is also important to establish a system where the kinase is inactive which can be attained by using knock out cell lines or specific kinase inhibitors. Knock out cell lines are especially useful in evaluating the specificity of a biosensor. These can be generated by a variety of techniques, such as small interfering siRNA that post-transcriptionally targets and degrades messenger RNAs or CRISPR (30) by which the gene sequence encoding for the kinase can be deleted altogether from the host genome. However, development of these cell lines can be a time-consuming process. An alternative approach is to use an inhibitor specific to the kinase of interest, if available. This method can be quicker than generating a new cell line, with some optimization of inhibitor concentration and incubation time needed. However, off-target effects of inhibitors must be considered. Few inhibitors are truly specific to a single kinase, so this approach is less ideal for assessing biosensor specificity. If inhibitors are used, the potential for off-target effects must be carefully evaluated.

1.4.5. Biological conditions

1.4.5.1. Cell culture conditions.

While the kinase biosensors may be applied to any cell type typically cultured in labs, flat (2D) adherent epithelial cell types such as MDA-MB-231 are easiest to image. It is important to note that the cell penetrating peptide sequence may need to be optimized for each cell type to maximize cellular transduction, since TAT uptake and distribution varies among cell types. Moreover, it has been debated that 2D culture systems are unable to accurately mimic the complexity of a 3D tissue environment or allow proper modulation of cell shape and bio-mechanical properties (stiffness). Thus, embedding cells in 3D ECM matrices or more sophisticated 2D structures such as micro-patterns or ECM sandwiches have gained increasing popularity to emulate a physiologically relevant tissue environment (31). We have previously observed that peptide uptake and distribution into 3D cell cultures is possible (unpublished, personal communication from Dr. Andrew Lipchik). As long as sufficient optical resolution can be achieved, such 3D cultures should be compatible with FLIM probe imaging of kinase activity.

1.4.5.2. Cell preparation and Biosensor Incubation.

Optimal incubation time and biosensor concentration are tricky to determine since very little is known about their cellular transduction mechanisms and there is no existing physical model to predict their uptake rate. Our preliminary flow cytometry experiments have shown that concentrations of 10–20 μM and incubation periods of 1–2 hours generally ensure adequate signal-to-background ratio during imaging. Some cells may be sensitive either to the cell penetrating peptide or the substrate itself, and exposure to high concentrations of the biosensor for long periods of time could trigger apoptosis (32), therefore concentration and incubation time should always be optimized by the researcher. Post incubation, the cells need to be thoroughly washed (at least three times with PBS) to eliminate excess FLIM probe that may contribute to background/noise during imaging. Resources such as environment control chambers or enclosures equipped with precise temperature, humidity or CO2 level controls and microfluidics, should be exploited if accessible during live cell imaging in order to maximize the relevance of the biological activity observed. Based on the kinase and signaling pathway under scrutiny, prior overnight serum starvation of the cells may be necessary. Fetal bovine serum is known to activate many signaling pathways, including the Ras/Mitogen-Activated Protein (MAP)/Extracellular Signal-Regulated Kinase (ERK) pathway (33). Thus, when investigating kinases recruited and activated by such pathways, prior overnight serum deprivation is imperative in order to start from cells with unstimulated kinase activity.

1.4.5.3. Kinase Activation.

As further described in our comprehensive review (7), the choice of an activator for a specific kinase is dependent on the kinase itself and the cell system. Receptor kinases can be activated by treating with respective ligands, such as growth factors, which induce oligomerization and self-activation (34). Epidermal Growth Factor (EGF) and Platelet Derived Growth Factor (PDGF) can induce auto-phosphorylation and activation of their respective receptor kinases (EGFR and PDGFR) localized on the cell surface. Non-receptor kinases are usually activated by an upstream receptor (35). For example, crosslinking of immunoreceptors by ligands on B cell or T cell lymphocytes initiates downstream signaling, leading to subsequent recruitment and activation of the Syk kinase via ITAMS (36), and, as described in the protocol below, stimulation of EGF activates Abl and Src-family kinases via upstream activation of EGFR.

1.5. Lifetime image acquisition parameters:

1.5.1. Detection rate:

It is necessary to maintain a low probability of registering more than one photon per decay cycle to guarantee that the histogram of photon arrivals covers the entire temporal range of the decay kinetics. If the number of photons reaching the detector per cycle significantly exceeds one, there is an overrepresentation of early photons in the histogram, an effect called ‘pile-up’. In order to maintain single photon statistics, the average count rate at the detector should be limited to at most 1–5% of the excitation rate.

1.5.2. Pulsing Frequency.

FLIM systems employ pulsed laser sources with frequencies in the megahertz scale. The pulsing frequency dictates the interval duration between consecutive excitation pules, and hence should be optimized based on the specific application. For example, when the pulsing frequency is fixed at 40 MHz, the time window between consecutive pulses spans 25 ns, which is sufficient to capture the full decay curves of fluorophores such as DyLight 550 or Cy5 with lifetimes in the range 1–3 ns. However, when imaging fluorophores characterized by intrinsically longer lifetimes such as DyLight 488 and other fluorescein-based dyes, the same pulsing rate will clip out the trailing edge of the decay curve, complicating subsequent analysis. An added consideration is the dynamics of the target kinase and its signaling/phosphorylation kinetics. For example, when fast dynamics or a time course evolution is of interest, a higher pulsing frequency is preferred. Although higher pulsing frequencies yield larger photon counts per pixel for a fixed number of frames/imaging duration, the applied frequency should be limited to the lowest rate that achieves the required lifetime detection range to avoid large scale deviations from single photon detection statistics (described above in “Detection Rate”).

1.5.3. IRF Response:

The Instrument Response Function of a FLIM system serves as an indirect measure of its electronic accuracy. In an ideal system with infinitely accurate components (source and detectors), the IRF is infinitely narrow. However, in a realistic system, the IRF is broadened by several factors, including the response delay of detectors, error associated with the timing reference signal generated at the source during excitation and timing jitter in other components due to thermal noise and interference. Precise measurement of the IRF is therefore imperative to ensure accurate fitting of the decay curves for lifetime extraction (28).

1.5.4. Photon Economy:

Photon economy, which represents the detection efficiency of fluorescence photons in a FLIM system, is quantified by a figure of merit $\text{F}$ defined as $\mathrm{F}={\mathrm{N}}^{0.5}{\sigma }_{\tau }/\mathrm{\tau }$ where ${\mathrm{\sigma }}_{\tau }$ is the standard deviation in repeated measurements of a lifetime, N is the number of photons used, and $\mathrm{\sigma }$ is the lifetime (37). We have $\mathrm{F}\ge 1$ in all lifetime measurement methods, and the closer the value to unity, the better the performance of a FLIM system. As a rule of thumb, an average photon number of at least 100 per pixel is desired to ensure accuracy of biexponential fitting. Thus, the number of frames to be averaged should be optimized to attain the desired photon count. The measurement speed and the photon economy of a lifetime measurement is mostly determined by the quantum efficiency and the gain linearity of a photodetector. A recent study showed that a hybrid photodetector integrating both an avalanche photodiode and a photomultiplier tube in an AMD system demonstrated a high linearity for multiphoton detection and hence improved photon economy (38). Such factors should be taken into account when using a FLIM system for the types of analyses described in this chapter.

1.5.5. Background signal considerations.

There are many endogenous fluorescence sources, but the most prominent ones are the intrinsically fluorescent metabolic cofactors nicotinamide adenine dinucleotide (NAD+/NADH) and flavin adenine dinucleotide (FAD/FADH2) (39). Autofluorescence from cells may exhibit lifetimes that temporally overlap with lifetime distributions of the labeled FLIM probe and confound subsequent analysis. Careful selection of the FLIM probe fluorophore can help avoid this situation. Usually, autofluorescence emission is governed by the excitation bandwidth. For example, at excitation in the near ultraviolet region, the autofluorescence signals typically exhibit short lifetime distributions in the range 0–2 nm which may be easily resolved during analysis when using a dye with markedly different lifetime such as DyLight 488 (3.5–5 ns) (40). On the other hand, in the far-red region (600–700 nm), autofluorescence is greatly attenuated, making far red dyes such as Cy5 a good choice.

1.5.6. Photobleaching.

Photo-bleaching of the fluorophore is a critical point of consideration in fluorescence measurements, especially in laser scanning confocal FLIM readings that entail long exposure times to a high intensity laser source and repeated excitation and decay cycles. Photobleaching is caused when a fluorophore permanently loses the ability to fluoresce due to photon-induced chemical modification, mainly from free radical oxidative species such as peroxides and singlet oxygen (41). The average number of excitation and emission cycles that occur for a particular fluorophore before photobleaching is dependent upon the molecular structure and the local environment. Some fluorophores bleach quickly after emitting only a few photons, while others that are more robust (such as the Alexafluor and DyLight compounds) can undergo thousands or millions of cycles before bleaching. Treatment with anti-fade media, free radical scavengers and other such reagents loaded with antioxidants, substantially lowers the photo-damage (42).

1.5.7. Multiplexing.

Analyzing the activity of more than one kinase at a time is possible via simultaneous application of different FLIM probes, but requires careful discrimination between the various labels. In the classical spectral approach, the multiple targets may be tagged each with spectrally different fluorophores allowing for multicolor imaging. Careful selection of fluorophores is required to avoid spectral cross-talk across the excitation and detection channels, however sometimes spectral unmixing can help to raise the number of distinguishable labels (43), and in some cases, fluorescence lifetime itself can be exploited as an additional parameter even if the excitation and emission ranges of multiple fluorophores are the same: Fluorophores with overlapping excitation or emission spectra but distinct lifetimes may still be distinguished by multi-exponential fitting of their time resolved emission, so as long as the lifetime distribution ranges are sufficiently different, the two probes could be resolved. Consequently, multiple fluorophores can be monitored and separated simultaneously within a sample. However, depending on the setup of the FLIM system, it may be necessary to employ specialized control over the excitation and/or detection (as described in Section 1.5.7.1 below) in order to capture the kinase activity of more than one probe concurrently while avoiding cross-talk between signals.

1.5.7.1. Multiplexing with Time Course experiments.

When trying to capture the transient dynamics of phosphorylation for more than one kinase (which can occur on scales as short as milliseconds to seconds) using multiple FLIM probes, typical laser scanning confocal microscope setups may be subject to issues arising from the length of time it takes for the system to scan across the sample. If a scan across a sample takes on the order of 1–2 minutes for one probe’s excitation/emission combination, and then needs to repeat for another probe’s excitation/emission combination, the biological signaling behaviors for each probe may be offset in time and not directly comparable. Simultaneous measurement of both channels would therefore be preferable, however another potential issue is that in some FLIM setups, cross-talk between excitation lasers can cause background signal across channels. For both of these situations, the Pulsed Interleaved Excitation (PIE) technique can be implemented (Figure 5). PIE directs the system to sequentially alternate across the multiple sources and their corresponding detection channels between pulses while acquiring a single image. This enables near-simultaneous recording of lifetime behaviors for all the multiplexed fluorescent probes without losing valuable information on transient signaling dynamics. An added advantage of this technique is that it also switches between the detection channels, which avoids the signal cross-talk described above.

Figure 5.

Implementation of Pulsed Interleaved Excitation (PIE) for multiplexed imaging and analysis.

1.6. Data analysis

1.6.1. Lifetime fitting.

Decay curves are typically fitted with a multiexponential lifetime model, as shown in Eqn. 1. Most FLIM systems are accompanied by software that will perform the fitting, for example the PicoQuant software uses a Levenberg–Marquardt-based search algorithm to find the minimum of the weighted chi-square (24). As an alternative to fitting, the exponential curve can also be transformed into first-order Fourier coefficients and plotted as phasor coordinates (g, s). Phasor plots are obtained by the numerical calculation of the discrete Fourier transform of the time domain lifetime decay curve. The conversion between two different schemes has been described previously (44, 45). An alternative graphical plotting scheme can be used from the fitted values to generate these g and s values (46).

$$I={\sum }_{n=1}^N{A}_n{e}^{-t/{\tau }_n}$$ (Eqn 1)

Where n and A_n denote the number of exponentials and activation energies, respectively.

1.6.1.1. Dissecting different signal components and background.

Time domain FLIM images should contain both photon intensities and decay curves for all pixels. To obtain the decay parameters including lifetime and amplitude coefficients, the fitting algorithms use an iterative deconvolution process. This process requires the input of the Instrument Response Function (IRF) which constitutes the response of the detection system to the excitation pulse itself. An overall decay curve constructed from the intensity weighted averages of all decay curves in an image is usually fitted to initialize the parameters before the iterative fit process, followed by a pixel-by-pixel fit of the decay curves. This algorithm may be applied to complex multi-exponential decay profiles as well. In order to eliminate background contributions from the suspension media and other reagents, it is strongly recommended to create Regions of Interest (ROIs) encompassing only the cells and apply the final FLIM fit analyses to these ROIs. For instance, Epidermal Growth Factor (EGF) absorbs strongly under 485 nm excitation and gives some background fluorescence in the media with an average lifetime that is comparable to that of DyLight 488 (3–5 ns). Thus, emission from EGF is likely to overlap with the lifetime distribution of DyLight 488 labelled biosensors, unless appropriate ROIs are applied. In addition, a threshold should be applied to exclude pixels that do not meet the minimum photon count requirements (100 for bi-exponential fits, 1000 for tri-exponential fits and so on). Alternatively, spatial binning tools in the FLIM software package may be used to increase the photon count by merging adjoining pixels (in the order of 2×2 or 3×3 square matrices), while concomitantly reducing the spatial resolution. With implementation of the appropriate threshold and ROI, a 2-component (biexponential) fit should yield sufficiently low residuals. In the types of FLIM probe experiments described in this chapter, the shorter and longer lifetime components may be attributed to the unmodified (unphosphorylated) and modified (phosphorylated) probes, respectively. In order to quantify their relative proportions, however, a slightly different fitting model needs to be used (as described below in Section 1.6.1.2).

1.6.1.2. Relative quantification of phosphorylated probe.

Generally, phosphorylated and unphosphorylated probes are distinguishable due to differences in their mean lifetime spread. These sub-populations may be resolved from a single decay curve by employing bi-exponential fitting algorithms modeled as shown in Eqn. 2. The fitting algorithm also yields the amplitude values of both populations (A_long, A_short) which represent their relative enrichment (amount). Thus, the relative amount of the phosphorylated species may be determined on a pixel-by-pixel basis by calculating and mapping the quantity A_long/(A_long+ A_short) or I_long/(I_long+I_short). I_long and I_short denote the fraction of the photon counts reaching the detector from the phosphorylated and unphosphorylated subpopulations, respectively (Eqn. 3).

$${\mathrm{I}}_{\text{total}}={\mathrm{A}}_{\text{long}}{}^{*}\exp \left(-\mathrm{t}/{\mathrm{T}}_{\text{long}}\right)+{\mathrm{A}}_{\text{short}}{}^{*}\exp \left(-\mathrm{t}/{\mathrm{T}}_{\text{short}}\right)$$ (Eq. 2)

$${\mathrm{I}}_{\text{total}}={\mathrm{I}}_{\text{long}}+{\mathrm{I}}_{\text{short}}$$ (Eq. 3)

1.6.2. Image reconstruction for data visualization.

Upon full fitting analysis per pixel for the various images produced in the experiment, the average lifetime value corresponding to each pixel is spatially mapped using pseudo color scales (Figure 6A). The average lifetime distribution (from all the pixels) is also available as a histogram. Export each lifetime mapped image in the text and/or picture files and lifetime histograms as text files.

• 1.6.3.For phospho-quantification, photon intensities from the long (phospho) and short lifetime (unphospho) components (I_long and I_short) can be separately mapped and extracted (as text files) on a pixel by pixel basis (Figure 6B). The ratio of these intensities (F_long=I_long/(I_short+I_long)), representing the fraction of phosphorylated probes, may be computed for each pixel and mapped to the image matrix, and/or the relative proportion of the phosphorylated FLIM probe (F_long) can be replotted using pseudo color scales that illustrate the localization of longer lifetime species, interpreted as probe phosphorylation/kinase activity.

Figure 6:

(A) Lifetime Imaging of Dylight 488 labeled Abl probes in MDA-MB-231 cells following Abl activation and inhibition by treatment with Epidermal Growth Factor (EGF) and Imatinib, respectively. Average lifetime and fraction of phosphorylated probes (F_long) were mapped in subcellular images and average lifetime distributions were plotted as histograms. (B) Implementation of biexponential fitting algorithm for the deconvolution of shorter and longer lifetime components representing the unphospho and phosphorylated probe sub-populations. Photon intensity counts (I_long, I_short) and lifetime values (τ_long, τ _short) corresponding to each pixel can be compiled in text files using MATLAB and replotted later in ImageJ for visualization.

2. Materials

• 2.1.
  Imaging hardware/software

  • Nikon A1R laser scanning confocal microscope equipped with a FLIM, FRET, and FCS enabled system (LSM Upgrade Kit, PicoQuant, Germany); with 488 nm picosecond pulsed laser line excitation.
  • SymphoTime software (Picoquant, Germany)
  • ImageJ software (National Institutes of Health, open source, https://imagej.nih.gov/ij/index.html)
  • MATLAB software (MathWorks, https://www.mathworks.com/products/matlab.html)
• 2.2.
  Chemicals and supplies

  • Fluorophore-labeled cell-penetrating FLIM probe substrate for Abl kinase:

    • Recommended sequence: GGEAIYAAPC(DyLight 488)GGRKKRRQRRRPQ
    • Can be obtained via custom peptide synthesis (unlabeled) and labeled on cysteine using DyLight 488-maleimide (ThermoFisher Scientific) according to protocols reported previously (18).
  • Gibco EGF Recombinant Human Protein (PeproTech)
  • Imatinib, Methanesulfonate Salt (LC Laboratories)
  • D-PBS Buffer (w/o Ca & Mg) (Fisher Scientific)
  • DMEM, high glucose (ThermoFisher Scientific)
  • Gibco^™ Fetal Bovine Serum (FBS), qualified, USDA-approved regions (Fisher Scientific)
  • Gibco^™ Penicillin-Streptomycin (10,000 U/ml) (ThermoFisher Scientific)
  • DMEM, high glucose, HEPES, no phenol red (ThermoFisher Scientific)
  • MDA-MB-231 cells (ATCC)
  • 35 mm Dish, No. 1.5 Coverslip, 14 mm Glass Diameter, Poly-D-Lysine Coated (MatTek Corporation)

3. Methods (See Note 1) (Figure 7)

Figure 7.

Illustration of the steps of the protocol from seeded, confluent cells through imaging and data analysis.

• 3.1.Culture MDA-MB-231 cells in bulk in DMEM high glucose media supplemented with 10% FBS and 1% penicillin-streptomycin, using standard cell culture procedures, through at least two passages up to final 80% confluency. Seed into Poly-D-Lysine coated 35 mm cover slip dishes at 1:10 dilution, and allow to grow to confluency of 60–80% at 37°C, 5% CO₂. (see Note 2).
• 3.2.Remove media and serum starve the cells overnight (at least 12 hours) by adding serum-free DMEM growth media containing antibiotic (Pen/Strep, 1%) (see Note 3).
• 3.3.Incubate cells with fluorophore-labeled FLIM probe at a final concentration of 10 μM for 2 hours (see Note 4).
• 3.4.Following incubation, wash the cells three times with PBS to remove extra FLIM probe (see Note 5).
• 3.5.Suspend the cells in phenol red-free, serum-free growth media for imaging.
• 3.6.Choose a field of view that contains more than one cell if possible. Focus on the cells using their bright field or fluorescence intensity signal on the confocal-enabled FLIM microscope at 20X magnification (see Note 6).
• 3.7.In the Picoquant Symphotime software, select the required excitation laser (e.g. 473 nm laser line for DyLight 488). Set the pulsing frequency to any value between 20–30 MHz (based on the intrinsic lifetime of DyLight 488) and the TSCPC resolution to 25 ps for FLIM acquisition (see Note 7).
• 3.8.Capture a trial image with the selected pulsing frequency averaging over a hundred frames (test number). Using the reader tool, determine the average photon count per pixel in the test image and estimate the required number of frames to ensure an adequate photon count per pixel (100 for bi-exponential fit) in the final image. Capture the final image.
• 3.9.Add EGF (10 ng) and incubate for 15 min, then repeat the imaging process in steps 3.6–3.8 above (see Note 4).
• 3.10.Add imatinib (to a final concentration of 10 μM) and incubate for an additional 15 min, then repeat the imaging process in steps 3.6–3.8 above (see Note 4).
• 3.11.Switch to the analysis mode in the PicoQuant software. The corresponding IRF signal should be automatically calculated from the raw FLIM files. Select and apply the desired channel(s) for analysis, in this case the green channel for DyLight 488 signal.
• 3.12.For each image file from the initial image and different treatments, construct a region of interesting surrounding the cells, avoiding as much background as possible, using the “free ROI” or “Magic Wand ROI” tools. Using the reader tool in the PicoQuant software, determine the average number of photons per pixel in the image. For a biexponential fitting algorithm, a minimum of 100 photons is desired. Apply 2×2 spatial binning if photon counts are low. Apply a threshold of 100 photons across the image to exclude pixels with inadequate counts from the analysis.
• 3.13.Implement the Levenberg-Marquardt model in the software with 2 lifetime components for analysis. Restrict lifetime distribution values within a physically acceptable bound (i.e. within the reasonable range for the fluorophore being analyzed), if need be (see Notes 8 and 9). Apply an initial fit at first, to determine the residuals (single digit residual values imply a good fit), before a final pixel-by-pixel fit.
• 3.14.The fit algorithm should yield a color mapped image of the intensity weighted average lifetime in channel 2, the corresponding color mapped fluorescence (total intensity) counts in channel 1 (should not be confused with excitation/detection channels) as well as a histogram of the average lifetime distribution across all pixels. Export the lifetime-mapped and corresponding total intensity-mapped images as a single bi-channel image file, which can be accessed with ImageJ for further processing and visualization (see Note 10). Likewise, export the lifetime histograms as text files and replot with Excel or GraphPad Prism for data visualization.
• 3.15.To visualize the FLIM images, open saved picture files in ImageJ. Apply “Stack to Images” to stack channel 1 (intensity information) and channel 2 (average lifetime information) of the image in a single window. To set a color scheme for the lifetime mapped image, select channel 2 and apply any Look Up Table (LUT) settings from available options in ImageJ (16_colors, 5_ramps, 6_shades, etc). Set the minimum and maximum lifetime values in Brightness and Contrast Adjustment to match the range in lifetimes between the shorter and longer lifetime species observed in the experiment: 1e⁹ and 6e⁻⁹ (1–6 ns) for DyLight 488 (see Note 11), and calibrate the color scale.
• 3.16.For phospho-quantification, photon intensities from the long (phospho) and short lifetime (unphospho) components (I_long and I_short) need to be separately calculated on a pixel by pixel basis. Upon implementation of final FLIM fit in PicoQuant, select I1 (intensity counts from longer lifetime component) and I2 (intensity counts from shorter lifetime component) to be mapped and displayed in channels 1 and 2, instead of average lifetime and total intensity. Export the resulting I1 and I2 mapped images as text files. Calculate F=I1/(I1+I2), representing the fraction (F) of longer lifetime (i.e. phosphorylated) probes for each pixel using the following MATLAB code on the two exported text files (which were named Imatinib1 and Imatinib3 in the code below) to give an output file in ascii format (which was named Imatinib0 in the code below) (see Note 12 for details on how to change the code to match files in different experiments):

%% Data File 
data1=Imatinib3;
data2=Imatinib1; 
%% Data Process 
[row,col]=size(data1); %size of file 
for i=1:row 
for j=1:col 
dataSum(i,j)=data1(i,j)+data2(i,j);%sum of twofiles
if dataSum(i,j)>0 
dataAns(i,j)=data1(i,j)./dataSum(i,j); %divison 
elseif dataSum(i,j)==0 
dataAns(i,j)=0; 
else 
dataAns(i,j)=dataSum(i,j); 
end 
end 
end 
%% Save to ascii file 
save(‘Imatinib0.txt’,’dataAns’,’-ascii’) 

The computed F values will be saved as an ascii text file which can be imported and plotted using ImageJ (See example of output in Figure 5) as described for the base FLIM images in Step 3.13.

4. Notes

1. This protocol is described with a Nikon laser-scanning confocal microscope equipped with 488 nm-compatible excitation and detection, and PicoQuant TCSPC fluorescence analysis hardware and software modules. Other FLIM systems should also work for this protocol, but hardware and software aspects, including data analysis procedures, may be different.

2. General conditions for cell culture are described by ATCC: https://www.atcc.org/~/media/PDFs/Culture%20Guides/AnimCellCulture_Guide.ashx. For particular cell lines, similar but more specific guidelines including information about optimal media and any additional supplements required, are available on the ATCC website. Although we describe this protocol for adherent cells, it is also possible to perform these experiments on suspension cells. Subsequent imaging is a challenge due to the mobility of the cells on the surface, but it is possible to stabilize them using e.g. methylcellulose additives.

3. Some cells are highly sensitive to serum starvation and will not survive this treatment. In that case, other modes of stimulating the kinase activity should be employed as discussed above in the Introduction (section 1.4.5.3). Each biological application will need optimization of these aspects prior to beginning the experiment.

4. We strongly recommend using our review on kinase substrate probe design (7) as a companion to this chapter, as it covers many aspects of experimental design for peptide probes/cell lines that go beyond those described in this protocol. Different kinase substrate peptide probes and/or cell lines may need different conditions, including:

   • Concentrations and incubation times for optimal uptake. Typical ranges to test for different probes are 10–25 μM for a duration of 1–2 hours. The FLIM probe may be solubilized as a stock solution in phenol red-free, non-serum growth media or PBS. While FLIM probe may be dissolved in PBS as a stock solution prior to incubation, it is preferable to use serum free growth media during incubation to promote cell health and viability. Concentrations of the FLIM probe stock solutions may be determined from the absorbance values at λ_max of the dye and its extinction coefficient at λ_max using Beer’s law.
   • Different kinase activation and inhibition treatments, depending on the cells used and kinases of interest, as described above in the Introduction (Section 1.4.5.3).

5. For cells in suspension, spin the cells at 2500 rpm for 3 minutes to pellet and remove supernatant, for each wash.

6. Magnification of at least 20X is recommended; 40X or 60X objectives will provide better subcellular resolution.

7. The PicoQuant software package includes functions for performing steps 3.7 – 3.12. Other instrument control/analysis software packages may require these functions to be identified as options or performed manually.

8. In some cases, especially instances of high background signal, there may be more than two species detected. FLIM probe peptides labeled with most commonly used small molecule fluorophores, whether phosphorylated or unphosphorylated, typically will have lifetimes between 1 ns and 6 ns. When the correct positive and negative controls are included in your experiment, their lifetime distribution histograms should assist in identifying the lifetime ranges for probes vs. background, which typically has lifetimes below 1 ns. In that case, species with lifetimes below 1 ns can usually be excluded from further analysis.

9. For multiplexed analyses, perform the fit analysis for all relevant detection channels, separately. This will result in a different set of output files for each detection channel.

10. Be sure to export all files with informative filenames as you proceed. These exported files are necessary for keeping proper research records as well as in preparing figures for presentation and publication.

11. These min/max values are likely to be different depending on the fluorophore used for each probe, and should be determined from your data.

12. In the MATLAB code, “data1” and “data2” refer to the text files for the longer lifetime (data1) and shorter lifetime (data2) species exported in Step 3.14. The specific file names should be changed when using the MATLAB code to analyze your own data.