Introduction

Most cellular processes are controlled by the coordinated action of protein kinases. However, the multifaceted actions of over 500 kinases in human cells pose major challenges for the study of kinase function and the deconvolution of signaling networks in cells. A range of techniques are available for investigating in vivo kinase activity, although each technology suffers from important intrinsic limitations. Mass spectrometry-based phosphoproteomic technologies are capable of mapping and quantitatively analyzing tens of thousands of phosphorylation events in cells. However, without meticulous subcellular fractionation, they lack the ability to probe kinase action with spatial resolution, a key variable in the study of kinase biology1,2. Phosphoproteomic approaches are also not easily parallelizable, making them less amenable to high throughput analyses. Moreover, phosphoproteomic data do not inform on direct kinase-substrate relationships unless combined with computational motif analyses, whose assignments are often ambiguous and biased towards well-characterized kinases3. Phosphorylation site-specific antibodies have been traditionally used to image the location of a phosphorylation event in cells using microscopy4,5. Unfortunately, antibodies displaying high specificity and low background for a given phosphorylation site are difficult to generate and not widely available. Therefore, experiments based on phospho-specific antibodies commonly suffer from low signal-to-noise ratio, limited dynamic range, and tend to yield semi-quantitative results that require extensive and complex image analyses and are not compatible with high-throughput analytical purposes.

Genetically encoded fluorescent biosensors offer a solution for the direct visualization of kinase activity toward specific target peptides in living cells6,7. Fluorescent kinase sensors, typically designed with a pair of fluorescent proteins for Förster Resonance Energy Transfer (FRET) or a conformationally changing circularly permuted fluorescent protein (cpFP), integrate a kinase-specific peptide substrate with a phospho-motif binding domain7,8,9. Upon phosphorylation by the kinase of interest, the recognition of the phospho-peptide by the binding domain induces a conformational change, triggering changes in energy transfer in FRET-based sensors or fluorescence in cpFP-based sensors9. Alternatively, the phosphorylation of the peptide substrate sequence can cause translocation or degradation of the fluorophore7. Cell-penetrating fluorescent (FLIM) probes for Abl and Src-family kinases have also been designed and implemented without the need for FRET10. Despite the successful application of fluorescent kinase sensors in a range of biological contexts, these systems have limitations, both in terms of general applicability as well as technical implementation. The specificity required for recognition by the phospho-motif domain limits the range of kinases that can be monitored, and the biosensor design process is labor-intensive, often requiring extensive optimization11. Moreover, excessive binding affinity between the phospho-motif and the interaction domain can lead to sensor saturation, restricting dynamic range and potentially impeding dephosphorylation by phosphatases. This was observed with a designed ATM sensor, which showed reduced efficiency in monitoring phosphorylation decay following kinase inhibition due to excessive binding strength12. Additionally, multiplexing these sensors is constrained by fluorescence signal overlap, limiting the ability to track multiple kinase activities concurrently11.

Damage to DNA or stress during DNA replication trigger the activation of phosphatidylinositol 3′ kinase (PI3K)-related kinases (PIKKs) ATR, ATM and DNA-PKcs13,14,15. Once activated, PIKKs orchestrate elaborate signaling responses that regulate a range of cellular processes such as DNA repair, DNA replication, the cell cycle, transcription, and apoptosis. The downstream kinases CHK1 and CHK2 are activated by ATR and ATM, respectively, and mediate key aspects of the overall signaling response, including replication fork stability and cell cycle progression16,17. Despite extensive studies on PIKKs and downstream signaling responses and the mapping of the signaling network controlled by these kinases, our understanding of the spatial organization of kinase signaling within distinct subnuclear domains and cellular compartments remains elusive. There is currently a need for quantitative tools capable of rigorously and systematically monitoring locations and kinetics of DNA damage signaling in a lesion- and cell type-dependent manner, and with high dynamic range and throughput.

Here, we develop the Proteomic Kinase Activity Sensor (ProKAS) technique that leverages MS for the multiplexed, spatially resolved, and quantitative monitoring of kinase activity in living cells. ProKAS is based on a tandem array of peptide sensors that allows simultaneous tracking of multiple kinases within a single polypeptide module. The introduction of amino acid barcodes into these peptide substrates enables the multiplexed monitoring of kinase activities across different cellular locations or under varying experimental conditions. The multiplexing capabilities make ProKAS compatible with high-throughput analyses and screening purposes. We engineered a ProKAS module with sensors specifically designed to sense the activity of DNA damage response kinases ATR, ATM, and CHK1, with an expanded version also including a pan-CDK sensor and a sensor for the PPM1D phosphatase. Our results demonstrate the ability of the ProKAS sensor to capture kinase activity with high specificity and spatio-temporal resolution, uncovering kinetics of kinase signaling during DNA damage responses. Additionally, we developed a de novo approach for the rational design of substrate peptides, which is expected to be broadly applicable to most kinases within the human kinome. Overall, ProKAS offers a versatile method for probing entire kinase signaling networks, opening additional avenues for investigating kinase action in cells.

Results

ProKAS biosensor design

We designed the Proteomic Kinase Activity Sensor (ProKAS) technique to allow for quantitative, multiplexable and spatially resolved monitoring of kinase activity in cells using mass spectrometry (MS) (Fig. 1). The core of ProKAS lies in the Multiplexed Kinase Sensor (MKS) module, a tandem array of 10-15 amino acid peptide sensors representing preferred substrate motifs for selected kinases of interest (KOIs) (Fig. 1A, B). Each sensor has a serine or threonine at its center and three or more small amino acid residues at the edge for barcoding purposes (Fig. 1B). The ProKAS polypeptide also features an N-terminal enhanced Green Fluorescent Protein (eGFP) for imaging and a Targeting Element (TE) that directs the sensor to a specific subcellular location (e.g. nuclear localization signal (NLS), nuclear export signal (NES), or a protein of interest) (Supplementary Data 1). To enable proteomic analysis, we incorporated an affinity tag (e.g., ALFA tag) and also flanked each kinase substrate motif in the MKS with arginine residues, ensuring that trypsin digestion during sample preparation would result in kinase sensor peptides with distinct masses readily identifiable and quantifiable via MS18. The sensor peptides were also checked to ensure the target serine or threonine was not immediately flanked by lysines or arginines, which would result in tryptic peptides too small for conventional detection. The experimental workflow involves transfecting cells in culture with plasmids expressing the ProKAS biosensor with kinase sensor peptides and targeting elements of choice, after which cells are treated with a specific stimulus (Fig. 1C). Following cell lysis and affinity purification, tryptic digestion generates a mixture of modified and unmodified sensor peptides that are quantified via MS, assessing the level to which each kinase sensor peptide became phosphorylated, normalized by the abundance of the unmodified version of each sensor peptide. The use of barcodes allows the generation of ProKAS biosensor libraries and multiplexed analyses where each code can be linked to a specific targeting element, enabling analysis with spatial resolution (Fig. 1D). Alternatively, barcodes can be linked to other variables, such as distinct drug treatments or genotypes, enabling comprehensive profiling of kinase signaling networks and facilitating high-throughput screens for drug effects or genetic perturbations.

Fig. 1: Design and rationale of ProKAS, a modular technique for multiplexed analysis of kinase activity using mass spectrometry.

A Schematic representation of the ProKAS construct for expression of the biosensor polypeptide containing fluorescent tags for visualization, a targeting element (TE) for subcellular localization, a multiplexed kinase sensor (MKS) module for detecting kinase activity, and an ALFA tag for high-affinity purification. B Detailed view of the MKS module consisted of multiple kinase sensors, each featuring a kinase substrate motif, a barcode, and a flanking arginine residue allowing for tryptic digestion and independent MS-based quantification. C Overview of the ProKAS workflow, starting with expression of biosensors via plasmid transfection. Expressing cell populations are then treated with a kinase stimulus of choice, after which ProKAS biosensors are purified from cell lysates, digested with trypsin, and the individual kinase sensors are quantified via MS in both their unphosphorylated and phosphorylated forms. D Application of ProKAS for multiplexed spatial analysis of kinase activity. Multiple plasmids with distinct TEs matched to a specific barcode are co-transfected. MS analysis distinguishes the sensors based on the barcode mass, allowing matching signal intensity of the specific peptide-probe to the respective cellular location.

Generation of ProKAS sensors based on endogenous kinase substrates

The success of ProKAS relies on high-quality sensors used in the MKS module, wherein each peptide substrate must report the activity of one kinase with sensitivity and specificity. We reasoned that the sequence of the substrate peptides could be based on ~10-15 amino acid residues flanking phosphorylation sites on endogenous substrates, which could provide enough specificity for preferential targeting by a KOI. Phosphoproteomic data from large-scale experiments using kinase inhibitors or loss-of-function mutations would provide initial lists of phosphorylation events uniquely dependent on the KOI that are also compatible with MS detection. Once cloned into ProKAS biosensors, experiments should validate: (1) the ability of the MS to detect the phosphorylated peptide sensor; (2) the stimulus-dependent regulation of the probe phosphorylation; and (3) the specificity of the probe phosphorylation by the KOI (Fig. 2A).

Fig. 2: Development and validation of a ProKAS sensor specific for ATR using phosphoproteomic data.

A Schematic outlining of the strategy for designing kinase sensors based on phosphoproteomic data and known endogenous substrates. 10-15 sequences surrounding phosphorylation events detected to be dependent on a kinase of interest (KOI) are cloned into ProKAS constructs, which are then expressed in cells to test for kinase sensor detectability, inducibility, and specificity. B Diagram showcasing that ATR is activated by single-stranded DNA damage, after which it preferentially phosphorylates substrates at the S/T-Q motif. Selective ATR inhibitors, such as AZD6738, are used in phosphoproteomic analyses to determine ATR-dependence of each detected phosphorylation event. C Identification of FANCD2 S717 as an ATR-specific phosphorylation site that is also induced by genotoxin camptothecin (CPT) through phosphoproteomic analysis. Phosphoproteome results are included as Supplementary Data 3 and 4. D Cloning of sequence surrounding FANCD2 S717 as an ATR kinase sensor candidate into a ProKAS biosensor featuring a nuclear localization signal. E MS/MS spectrum gathered for phosphorylated ATR sensor, showcasing successful detection of the sensor candidate by MS. F MS analysis showing inducibility and specificity of the ATR sensor candidate after treatment with genotoxin and selective ATR inhibition, respectively. Cells were treated with 1 millimolar HU for 2 hours, and ATR-inhibited cells were treated with 5 micromolar AZD6738 15 minutes prior to HU addition. Error bars in F indicate the mean and standard deviation of triplicate independent experiments. Source data are provided as a Source Data file.

As a proof of principle, we utilized the pipeline depicted in Fig. 2A to obtain MS-detectable ATR substrates suitable for use as ATR kinase sensor peptides (Fig. 2B). Large-scale phosphoproteomic analyses of camptothecin (CPT)-treated cells revealed dozens of phosphorylation sites in established ATR substrates that were strongly impaired by the ATR inhibitor, indicating high specificity for ATR and no predominant targeting by related kinases ATM or DNA-PKcs. Among these sites was serine 717 of FANCD2, a previously known ATR substrate, which our data indicated was highly dependent on ATR and readily detectable via mass spectrometry19 (Fig. 2C, Supplementary Data 3, 4). We therefore selected the 13 amino acid sequence surrounding FANCD2 S717 and designed the ATR sensor by also adding a barcode and flanking R residues. This design results in the generation of a tryptic phosphopeptide slightly different from the endogenous FANCD2 counterpart. We then cloned this sequence into a ProKAS vector bearing an NLS as the targeting element, intending to monitor ATR activity in the nucleus (Fig. 2D). After transfection into HEK293T cells and treatment with CPT, the ATR ProKAS biosensor was pulled down with anti-ALFA beads, trypsinized, and analyzed by MS. PRM quantitation was used to compare peak areas for both the unphosphorylated and phosphorylated versions of the ATR sensor candidate in conditions treated with drug vehicle and hydroxyurea (HU). This revealed a significant increase in the abundance of phosphorylated ATR sensor upon genotoxin treatment after normalizing by the abundance of the unphosphorylated form of the sensor (Fig. 2E). Sensor specificity was also confirmed by treating cells with ATRi prior to the addition of HU, which prevented HU-induced phosphorylation of the sensor and further reduced its phosphorylation status to below that of the untreated cells, likely due to inhibition of basal ATR activity (Fig. 2F). Overall, these results validate the strategy of extracting small (10-15 amino acids) sequences from endogenous substrates to design a ProKAS sensor that exhibits high specificity for a given KOI.

De novo generation of a kinase-specific sensor peptide for ProKAS

We also developed a computational approach for the design of specific kinase probes by leveraging a dataset derived from positional scanning peptide array (PSPA) analyses of 303 human Ser/Thr kinases20. The pipeline shown in Fig. 3 scans PSPA-based kinase preference scores in a space of billions of 10-residue sequences using a genetic algorithm approach to identify those exhibiting high specificity for a KOI, while minimizing cross-reactivity with the broader kinome (Fig. 3A). After selection of 10 sequences with highest predicted specificity and lowest cross-reactivity, the candidate sequences are then cloned into and expressed as one single ProKAS biosensor for experimental validation (Fig. 3B). This allows for the identification of the sensor sequences that are detectable in the MS and that demonstrate the best kinase inducibility and specificity, ultimately guiding the selection of the top KOI-specific peptide substrate sensor to be used for ProKAS applications.

Fig. 3: Pipeline for the computational design and experimental validation of a CHK1-specific ProKAS sensor.

A Overview of the computational approach for generating kinase-specific motifs based on PSPA data. A genetic algorithm approach was utilized to arrive at a cohort of optimized sensor candidates out of the billions of possible sensor peptide sequences. B Workflow for multiplexed screening and selection of MS-detectable kinase sensors from candidates generated and filtered in silico. C Schematic illustrating CHK1 as a downstream effector kinase of ATR, and that CHK1 can be selectively inhibited by CHIR-124. D Selection of 10 candidate sequences for a CHK1 sensor based on PSPA scores as shown in (A). E MS/MS spectra gathered for the phosphorylated form of the indicated CHK1 sensor candidates. Screening sensor with the candidates indicated in (D) were expressed in HEK293T cells treated with HU. F MS analysis showing inducibility and specificity of the CHK1-4 sensor candidate after treatment with genotoxin and selective CHK1 inhibition, respectively. Cells were treated with 1 millimolar HU for 2 hours, and CHK1-inhibited cells were treated with 500 nanomolar CHIR-124 15 minutes prior to HU addition. Error bars in G indicate the mean and standard deviation of triplicate independent experiments. Source data are provided as a Source Data file.

We applied this pipeline to develop a probe for CHK1, a key downstream effector kinase activated by ATR in response to DNA damage21 (Fig. 3C). While CHK1 is known to preferentially phosphorylate substrates containing Arg or Lys residues at the −3 position relative to the phosphosite, the PSPA data confirmed this preference and further revealed that CHK1 exhibits strong preferences for bulky hydrophobic residues at the position −5 (e.g., Leu, Iso, and Phe), as well as a moderate preference for Pro and Asn at the position −120,22,23,24. We selected sequences exhibiting high PSPA-bases scores for CHK1, while minimizing scores for the other 301 kinases (Fig. 3D). These contained Arg/Lys at position −3, but also incorporated variations at this and other positions to explore the impact on CHK1 specificity and sensor performance. A CHK1 ProKAS biosensor was generated with an MKS module containing 10 candidate sensor peptides and an NLS targeting element. MS analysis could readily detect tryptic peptides for all 10 candidate sensors (Supplementary Data 5). However, upon genotoxic stress, only two phosphorylated peptide substrates were identified, CHK1-1 (LERHNSDQGGGAGR) and CHK1-4 (IEKYPSDGDQGAGR), both containing Arg or Lys at the position −3 (Fig. 3E). The CHK1-4 peptide, selected as our final CHK1 sensor, demonstrated high inducibility upon cell treatment with HU, which is known to generate long stretches of ssDNA and thereby robustly activate ATR and consequently CHK113,25 (Fig. 3F). We also confirmed that the CHK1-4 sensor was specifically phosphorylated by CHK1 through the use of a selective CHK1 inhibitor CHIR-12426 (Fig. 3F). Overall, these results confirm that Arg/Lys at the position −3 is a strong determinant of CHK1 specificity, and is consistent with the in vitro specificity assay showing that Leu/Iso at the position −5 favors CHK1 phosphorylation recognition in vivo. Importantly, in addition to generating preferential sequences for phosphorylation by the kinase of interest, this strategy is expected to also generate sequences that are less prone to be targeted by other kinases. The fact that phosphorylation of the sensor was impaired by CHK1 inhibitor supports the notion that the sensor is not being predominantly targeted by another kinase. These findings also highlight the need for an experimental screening step, given that the majority of in silico-generated sequences did not behave as effective sensors for our kinase of interest, whether due to a lack of phosphorylation or poor ionization efficiency of phosphopeptides. To compare with phosphoproteome-derived sensors, this approach was also applied to ATM, resulting in kinase sensor peptide sequences that proved to be MS-detectable, inducible, and specific to a similar degree when compared to the ATM sensor derived from our phosphoproteomic data (Supplementary Fig. 1).

Multiplexed monitoring of ATR, ATM, and CHK1 activity with ProKAS

We next assessed the use of ProKAS for monitoring the activity of multiple kinases simultaneously, circumventing the need for conducting separate experiments for each kinase of interest. We generated a multiplexed “DDR” ProKAS biosensor to simultaneously monitor the activity of ATR, CHK1, and ATM kinases (Fig. 4A, B). For the ATM ProKAS sensor, we employed the phosphoproteome-guided strategy, revealing a phosphorylation event in the known ATM substrate 53BP1, which was ultimately selected as our sensor peptide27 (Supplementary Fig. 2). The DDR ProKAS containing a tandem array of the ATR, ATM, and CHK1 sensors (Fig. 4B) was expressed in cells that were mock-treated or treated with either CPT or HU to predominantly generate DSBs or stalled forks, respectively28,29. We first confirmed that swapping substrate peptide positions within the MKS module did not alter their level of inducibility upon genotoxin-induced kinase activation (Supplementary Fig. 2). We next assessed whether each of the sensors retained the specificity for the respective KOI. As shown in Fig. 4C, the use of specific kinase inhibitors confirmed that the sensors within the context of the ProKAS peptide array were specifically phosphorylated by the expected kinase. Inhibition of ATR under both genotoxin conditions revealed a reduction in the phosphorylation of the CHK1 sensor, consistent with the activation of CHK1 being canonically controlled by ATR13,30. Inhibition of CHK1 alone did not reduce the phosphorylation of ATR or ATM kinase sensors in cells treated with CPT. However, phosphorylation of the ATM sensor increased in HU-treated cells upon CHK1 inhibition, consistent with CHK1 playing key roles in preventing the collapse of stalled replication forks in HU, which causes DSBs and ATM activation31. As expected, ATM inhibition robustly and specifically ablated phosphorylation of the ATM sensor upon CPT treatment, and also displayed minor inhibition of the ATR and CHK1 sensors likely due to effects in inhibiting DNA end resection during the DSB response32,33. We also confirmed that expression of these kinase sensors did not impair the endogenous DNA damage response by blotting for markers of ATR and ATM activation after treatment with CPT and HU34,35 (Supplementary Fig. 4). Expression of kinase sensors also had no significant impact on cell viability when measured via MTS assay (Supplementary Fig. 5). Additionally, large changes in expression level of the biosensors did not have a major effect on the quantitative readouts obtained from the kinase sensor peptides (Supplementary Fig. 6).

Fig. 4: Multiplexed analysis of DDR kinase activities using ProKAS.

A Schematic illustrating the canonical mechanisms of activation for DDR kinases ATR, ATM, and CHK1. B Design of a triplexed ProKAS construct containing kinase sensors for ATR, ATM, and CHK1. C Validation of kinase specificity of each sensor within the triplexed ProKAS construct via treatment with selective kinase inhibitors for each KOI in the presence of either 1 micromolar CPT or 1 millimolar HU. ATR, ATM, and CHK1 were inhibited with 5 micromolar AZD6738, 50 nanomolar AZD0156, and 500 nanomolar CHIR-124, respectively. D Flowchart showing primary components of the semi-automated high-throughput pipeline enabling larger-scale ProKAS experiments. E MS analysis showing kinase activation dynamics in response to 1 micromolar CPT over 6 hours in HEK293T cells expressing the triplexed ProKAS construct. F MS analysis showing kinase activation dynamics in response to 1 millimolar HU over 6 hours in HEK293T cells expressing the triplexed ProKAS construct. G MS analysis showing inhibitor titration using ProKAS to demonstrate the potency and selectivity of ATM inhibitor AZD0156. Cells were treated with 1 micromolar CPT for 30 minutes to activate ATM. Error bars/envelopes in E, F and G indicate the mean and standard deviation of triplicate independent experiments, except for CHK1 sensor quantification in G which comprises duplicate independent experiments from which no statistical significance has been calculated or displayed. Source data are provided as a Source Data file.

In addition to multiplexing, other advantages of ProKAS include the throughput and highly quantitative nature of the analyses, which should allow for precise kinetic studies on kinase signaling. We implemented a semi-automated pipeline for processing dozens of samples in 24-well cell culture plates (Fig. 4D), enabling temporal kinetic analysis of ATR, ATM, and CHK1 signaling in response to CPT and HU (Fig. 4E, F). Consistent with the expected behavior of these kinases upon treatment with CPT, which induces DSBs, ATM showed rapid and robust activation within 10 minutes, reflecting its role as a primary responder to DSBs. ATR activation followed, ultimately matching ATM activation levels at later timepoints, consistent with the longer time required for DNA ends to be resected and to support ssDNA-mediated ATR activation36. The slower kinetics of CHK1 activation were consistent with it being a kinase downstream of ATR. In contrast, HU treatment revealed distinct activation kinetics for these DDR kinases. Consistent with HU causing stalled replication forks and rapid ssDNA exposure, ATR was rapidly activated whereas the ATM activation did not reach levels as high as ATR activation (Fig. 4F). CHK1 phosphorylation was slower initially, but ultimately surpassed ATM. These observations align with the canonical behaviors of these kinases in response to DNA damage and highlight the ability of ProKAS to generate highly quantitative data on the kinetics of kinase activation. Notably, raw peak areas showed no significant drop in unmodified sensor abundance as treatment durations increased, suggesting that phosphorylation of the kinase sensor peptides does not lead to proteolytic degradation of the biosensor (Supplementary Figs. 7, 8 and Source Data). These results also showed that unmodified sensor peptides were several orders of magnitude more abundant than their phosphorylated counterparts, consistent with no perceptible depletion of the unmodified sensors as phosphorylation changes remain sub-stoichiometric2. Finally, we leveraged the multiplexed and throughput capabilities of ProKAS to assess the inhibitory potency of an ATM-specific inhibitor (AZD0156) in vivo. We performed an inhibitor titration experiment in HEK293T cells expressing the triplexed DDR ProKAS sensor, monitoring the phosphorylation levels of all three sensors simultaneously. As expected, AZD0156 potently inhibited ATM activity with an IC50 of 9 nanomolar, somewhat higher than what has been measured in HT29 cells via ATM autophophorylation37 (Fig. 4G). Importantly, the inhibitor exhibited high selectivity, showing no significant effect on CHK1 activity and minimal impact on ATR activity even at high concentrations (IC50 > 5 micromolar), aligning with the known degree of selectivity for this inhibitor. This experiment showcases the power of ProKAS in enabling precise, multiplexed, and in vivo assessment of kinase inhibitor efficacy and selectivity, offering a valuable tool for drug discovery and development.

Spatial analysis of kinase signaling using location-barcoded ProKAS

In addition to its quantitative and multiplexed capabilities, ProKAS was designed to also enable the spatial analysis of kinase activity in cells. As shown in Fig. 1D, the ProKAS construct features a targeting element that controls cellular localization and can be linked to a specific amino acid barcode embedded within each sensor peptide. Upon co-transfection of cells with ProKAS constructs containing different targeting elements, detection of distinct barcoded peptides by MS should reveal the relative levels of kinase activity in different cellular locations. As a proof of concept, we used an NLS or NES as targeting elements to direct the ProKAS polypeptide to the nucleus or cytosol, respectively (Fig. 5A, B). The AGA barcode was used in the construct containing the NLS and a GAG barcode was used in the construct with the NES. Importantly, addition of different barcodes to the substrate sensor peptides within the MKS module did not alter their efficiency of phosphorylation upon kinase activation (Supplementary Fig. 9).

Fig. 5: Spatially-encoded ProKAS for analysis of DDR kinase activities.

A Illustration of nuclear and cytosolic ProKAS biosensors featuring distinct codes for simultaneous monitoring of DDR kinase activities in both locations. B Microscopy showing the localization of ProKAS biosensors containing an NLS or an NES. Experiment was repeated 3 times. C Peak area ratios for DDR kinase sensors showing the proportion of phosphorylated sensor peak area to unphosphorylated peak area. Data based on MS analysis of cells not treated with any genotoxin. D MS analysis of co-expressed NLS- and NES-containing ProKAS biosensors, simultaneously monitoring DDR kinase signaling kinetics in both the nucleus and cytosol in response to 1 millimolar HU and 1 micromolar CPT. The experiments were quantified via SILAC. E Nuclear and cytosolic DDR kinase signaling kinetics were also monitored after treating cells with 10 micromolar gemcitabine (GEM) and 5 micromolar doxorubicin (DOXR). The experiments were quantified via label-free quantification (LFQ). F Microscopy showing the co-localization of the ProKAS biosensor (utilizing PCNA as a targeting element) with EdU foci. The experiment was repeated 3 times. G Densitometry analysis of EdU and ProKAS-PCNA signal across the white line drawn across the nucleus in panel E, showing signal coincidence between the EdU and ProKAS-PCNA foci formed. H MS analysis simultaneously monitoring the effect of HU on ProKAS biosensors containing an NLS or PCNA as the targeting element. I Schematic illustration of different spatial distributions of ATR, CHK1, and ATM kinase activity observed upon replication stress or DSBs. Error bars/envelopes in C, D, and F indicate the mean and standard deviation of triplicate independent experiments. Source data are provided as a Source Data file.

ATR, ATM, and CHK1 are DNA damage signaling kinases known to operate primarily inside the nucleus. While cytosolic roles for these kinases have been proposed38,39,40, potential non-nuclear functions remain elusive due, in part, to the need of more quantitative tools capable of monitoring the activity of these kinases with rigorous specificity and spatial resolution. Using ProKAS, we analyzed cells co-transfected and thereby co-expressing nuclear and cytosolic versions of the multiplexed ATR, ATM, and CHK1 sensors differentially barcoded based on their location (Fig. 5C–H). The intensity of the phosphorylated peptide in each compartment was divided by the intensity of the corresponding unphosphorylated peptide in that compartment, yielding normalized values of peptide phosphorylation, a rough measure of stoichiometry of phosphorylation. As shown in Fig. 5C, while we were able to detect phosphorylation of all three sensors in both nucleus and cytosol in the absence of any genotoxic treatment, the constitutive levels of kinase sensor phosphorylation were significantly lower for the cytosol-localized peptide sensors compared to their nucleus-localized counterparts. Such high levels of constitutive kinase signaling in the nucleus are consistent with the high endogenous levels of DNA replication stress in HEK293T cells41. We next monitored the kinetics of ATR, ATM, and CHK1 signaling in the nucleus and in the cytosol following treatment with CPT and HU (Fig. 5D). Once again, the two drugs exhibited distinct patterns of kinase activation over time in the nucleus. Interestingly, while HU did not result in major changes in kinase activity in the cytosol, ATM activity was specifically detected in the cytosol when treating cells with CPT. We proceeded to perform similar kinetics experiments utilizing two more genotoxic agents, gemcitabine (GEM) and doxorubicin (DOXR), which again revealed distinct kinase activation dynamics (Fig. 5E). Despite being an inhibitor of ribonucleotide reductase like HU, GEM showed a slower pattern of kinase activation in the nucleus while still exhibiting a similar lack of cytosolic kinase activity changes. DOXR also showed different dynamics from CPT, even though both genotoxic agents are DSB-inducing toposiomerase inhibitors, inducing a strong increase in nuclear ATM signaling that was largely uncoupled from ATR-CHK1 signaling. However, as seen with CPT, an increase in cytosolic ATM signaling was once again detected, which was then shown to be prevented via ATM inhibition in the presence of both drugs (Supplementary Fig. 10). This cytosolic ATM sensor phosphorylation was verified via cytosolic fractionation followed by ALFA purification, which also confirmed that the ATR and CHK1 sensors featured no induced phosphorylation in the cytosol (Supplementary Fig. 11). ATM has been shown to localize to various cytosolic locations in specific situations42,43, but its level of activation outside the nucleus has been difficult to measure. While the increase in cytosolic ATM signaling was relatively mild compared to the detected increase in nuclear ATM signaling, it is important to note that the larger volume of the cytosol is likely diluting the observed induction, especially if the source of the non-nuclear ATM activation is at a specific cellular location, such as the external membrane of a specific organelle. Since we did not detect an increase in cytosolic ATR or CHK1 signaling in response to any of the treatments tested, the increase in cytosolic ATM signaling induced by DSB is likely a specific feature of ATM activation or signaling propagation, and is unlikely to represent a non-selective leak of nuclear kinases into the cytosol.

To demonstrate the ability of ProKAS to probe differences in kinase activity within nuclear sub-compartments, we focused on sites of DNA synthesis. Upon HU treatment, sites of DNA synthesis are expected to accumulate stalled replication forks and single-stranded DNA, resulting in the selective induction of ATR-CHK1 signaling15. We localized the ProKAS biosensor to sites of DNA synthesis by using the processivity factor PCNA clamp, a key component of replication forks, as the targeting element44. Microscopy analysis confirmed that the PCNA-containing biosensors co-localized with nascent DNA, as visualized via incorporation of EdU (Fig. 5F, E). Using HCT116 cells, we then co-expressed ProKAS biosensors using PCNA with an NLS or only the NLS as the targeting element, allowing comparison of kinase signaling diffused in the nucleoplasm with signaling localized to sites of DNA synthesis. Upon treatment with 1 millimolar HU for 1 hour, biosensors localized to the nucleus with an NLS alone showed induced phosphorylation of all three kinase sensors (Fig. [5H](https://w