pYtags enable spatiotemporal measurements of receptor tyrosine kinase signaling in living cells

Article activity feed

1. 

   Version published to 10.7554/elife.82863 on eLife

   May 22, 2023
2. 

   eLife

   Mar 12, 2023

   Author Response

   Reviewer #2 (Public Review):

   The idea of using fluorescently labeled tandem SH2 domains to target tagged RTKs is brilliant and could potentially provide a powerful new way to assess the activation of RTKs in situ and in multiple physiological contexts. Thus, it was disappointing that there was insufficient characterization of the system to be able to interpret the data it generates. Although the paper shows that tagging the EGFR appears to have minimal impact on its biological activity, the readout for receptor kinase activity is % clearance of the fluorescent reporter tag from the cytosol. Such clearance is likely to depend on a variety of different factors, including the ratio of tagged receptors to probe, the number of functional pools in which the probe exists, the exchange rate between these pools, and the …

   Author Response

   Reviewer #2 (Public Review):

   The idea of using fluorescently labeled tandem SH2 domains to target tagged RTKs is brilliant and could potentially provide a powerful new way to assess the activation of RTKs in situ and in multiple physiological contexts. Thus, it was disappointing that there was insufficient characterization of the system to be able to interpret the data it generates. Although the paper shows that tagging the EGFR appears to have minimal impact on its biological activity, the readout for receptor kinase activity is % clearance of the fluorescent reporter tag from the cytosol. Such clearance is likely to depend on a variety of different factors, including the ratio of tagged receptors to probe, the number of functional pools in which the probe exists, the exchange rate between these pools, and the affinity of the probes for the tagged receptor. Without determining how each of these factors impacts % clearance, it is difficult to interpret either the dose-response curves or response kinetics.

   We appreciate the reviewer’s point that the paper would be improved by a thorough analysis of how membrane translocation depends on our biosensor’s expression levels. We have attempted to address this thoroughly in our response to the Editor’s summary comments above. Briefly, we have now added 3 new supplementary figures (Figures S2-S4) in which we quantify ZtSH2 translocation as a function of expression levels. We find that the ratio of EGFR/ZtSH2 expression predicts the extent of ZtSH2 translocation in both NIH3T3 and HEK293T cells, matching results from our computational model. We have also added a new section to the main text to clearly explain these results (Lines 190-235). We hope that these data clarify the design constraints for two-component biosensors of this type.

   For example, the difference in activation kinetics between EGFR and ErbB2 is very interesting, but the almost instantaneous rise (Fig S4B) is very surprising. The kinetics of activation of the EGFR have been extensively studied by mass-spectrometry and are generally limited by ligand binding, which has a characteristic time of several minutes, not seconds (pmid: 26929352; pmid: 1975591). Thus, such a response is suggestive of a freely exchanging ZtSH2 reporter pool that is mostly depleted in seconds with the slow secondary kinetics reflecting a slowly exchanging ZtSH2 reporter pool. Alternately, the cells could be accumulating an intracellular pool of activated receptors over time. That the authors are using concentrations of EGF >100-fold physiological levels (pmid: 29268862) further complicates the interpretation of these experiments.

   We thank the reviewer for bringing these papers to our attention. However, we strongly disagree with their interpretation of the results. In a paper cited by the reviewer (PMID:26929352), phosphotyrosine responses are extremely fast, with phosphorylation occurring within tens of seconds even in response to 20 nM EGF (see Figure 2 from Reddy et al PNAS 2016). Reddy et al further claim in their abstract “Significant changes were observed on proteins far downstream in the network as early as 10 s after stimulation.” While the timescale of EGFR phosphorylation may be of some debate, the response timescale we observe is consistent with previously published observations.

   It is also important to point out that the secondary gradual rise of ZtSH2 recruitment is only observed upon treatment with EGF, not EREG or EPGN (Figure 3A). The gradual rise can also be observed upon treatment with EREG in the presence of a GBM-associated EGFR mutation that alters receptor dimerization (Figure 3E). These data indicate that the secondary rise is not an intrinsic feature of the ZtSH2 reporter, and instead represents a feature of ligand-receptor activation itself.

   The reviewer suggests that perhaps there is some internal pool of ZtSH2 or EGF, but we find no evidence for such a pool in our microscopy imaging. To clarify this point to the reader, we have now added a new supplementary figure (Figure S6) showing representative cells for all stimulation conditions used in Figure 3A, showing consistent, high levels of EGFR and ZtSH2 enrichment at the plasma membrane and uniform cytosolic intensity for at least 30 min after stimulation across all ligands.

   Finally, while the reviewer mentions the use of high EGF doses in our paper, we would like to point out that we performed extensive experiments at other doses in the manuscript, testing 14 total doses of three EGFR ligands in Figure 3, and present additional data at 20 ng/mL EGF throughout Figures 2, S2, and S7. It is also very important to test high input doses for our negative controls to ensure that the ZtSH2 biosensor retains specificity for ITAM sequences and fails to show recruitment to untagged EGFR even under saturating conditions. It is also quite customary in the field: for example, the Erk KTR paper that the reviewer mentions in a later comment (Regot et al, Cell 2014) exclusively tests their biosensors using saturating doses of 50 ng/mL anisomycin, 100 ng/mL FGF, and 10 μM forskolin to characterize p38, Erk and PKA biosensor responses.

   There is also insufficient attention paid to either controlling or measuring important parameters, such as expression levels of tagged receptors or levels of endogenous receptors. 3T3 cells, contrary to the statement of the authors, do not have "negligible" numbers of EGFR: they have ~40K, which is typical for mouse fibroblasts. This is much higher than MCF7 cells, which are frequently used as a model system to study EGFR responses. Yet they do not see transactivation of their ErbB2 construct in 3T3 cells without expressing additional EGFR (Fig. 4C), suggesting low sensitivity of the assay. Conversely, they show a significant response mediated by endogenously tagged EGFR in HEK 293 cells, which are frequently used as an EGFR-negative cell line (PMID: 26368334). This indicates that their assay is extremely sensitive. Which is it? As mentioned above, it likely depends on the expression level and affinity of the different components of their system.

   After extensive searching we have not found any publications with an estimate as high as 40K EGFR receptors/cell in NIH3T3 cells. Livneh et al 1986 report that NIH3T3 cells express as little as 500 EGFR receptors per cell and do not respond mitogenically to EGF, and subsequent Schlessinger lab papers use NIH3T3 cells as an EGFR-null background for introduction of receptor variants. Eierhoff et al PLOS Pathogens 2010 use NIH3T3s as an EGFR-null control, showing immunoblot data of undetectable pEGFR responses. The paper we found with the highest stated EGFR expression per cell in NIH3T3 cells is Verbeek et al, FEBS Lett 1998, which reports a value of 3,000 receptors per cell, but does so without any literature citation or measurement. These references are consistent with our experience: over nearly a decade of MAPK signaling experiments in the lab, we have only seen weak or undetectable EGF-stimulated responses in unmodified NIH3T3s, depending on the assay. We are quite confident that more potent responses are elicited in HEK293T cells, where we observe EGFR expression by fluorescence imaging of CRISPR-tagged cells, immunofluorescence staining, and immunoblotting, and where we observe robust signaling responses using biosensors. We also now cite some of these references to support our claim (Line 144).

   The reviewer makes an excellent point in the last sentence of their comment: indeed, it is essential to match the expression level of our SH2-based biosensor to the expression level of EGFR in any system in order to observe potent membrane translocation! This was imperative for visualizing any translocation in our CRISPR-tagged HEK293Ts: we had to switch to an exceptionally bright fluorophore and select cells with very low ZtSH2 expression to observe translocation. The ZtSH2/EGFR ratio is a crucial design parameter, which we now present extensive data and modeling to support (Figure S2-S4; Lines 190-235). Our data suggests that quite sensitive biosensor responses are possible with appropriate balance between ZtSH2 and EGFR expression levels (Figure 6) and, in general, biosensor responses can be matched to a dynamic range of interest by scaling ZtSH2 expression with EGFR levels.

   A great advantage of using the EGFR system as a test case for the new system is that thousands of investigations have been performed over the last four decades. This provides a strong foundation for determining whether the new technology is working correctly. For example, the dynamics of EGFR activation and trafficking at the single cell level have been documented in many studies, which show a remarkable consistency (e.g. see pmid: 24259669; pmid: 11408594; pmid: 25650738). Unfortunately, instead of using differences between the new results and previously reported data as a basis for refining their technique, the authors attempt to apply their raw data to address complex questions of EGFR dynamics, with less than satisfactory results.

   For example, they attempt to use their technique to understand the basis of different signaling dynamics between EGFR ligands. Rather than being a relatively recent observation, differences in EGFR ligand signaling have been explored for over 30 years (pmcid: PMC361851), and are generally ascribed to differences in trafficking (pmid: 7876195). Based on these observations and resulting mathematical models, novel EGFR ligands have been designed with enhanced potency (pmid: 8195228 , pmid: 9634854 ). All this work was done over 20 years ago. Since then, new natural ligands for the EGFR have been discovered from sequence analysis and differences in their potency have similarly been ascribed to differences in their intracellular trafficking patterns (pmid: 19531065 - cited by the authors). An alternate hypothesis was proposed more recently by Freed et al (2017) as described by the authors, but that is what it is: an alternative hypothesis.

   We thank the reviewer for pointing out many excellent, classic studies on EGFR endocytosis and trafficking. We agree that this is a well-established field and that EGFR is certainly internalized, recycled, and degraded in a manner that depends on ligand affinity on the cell surface and in endosomes. These seminal studies lead the reviewer to propose an alternative hypothesis to explain our kinetic data in Figure 3: that differences in trafficking and maintenance of EGFR levels at the plasma membrane are the source of the altered kinetics between high- and low-affinity ligands. To address this question, we have now included new supplementary data examining endocytosis and trafficking in multiple contexts.

   First, we examine membrane EGFR levels in 3T3 cells overexpressing our EGFR-pYtag system (or ITAM-less EGFR as a control) after EGF stimulation (Figure S5A-C). We find that EGFR membrane intensity is virtually unchanged after 60 min of saturating EGF stimulation, a response that does not depend on whether ITAMs are appended to the receptor. We also now include still images of cells at every concentration examined in our dose-response experiments for all 3 ligands (Figure S6), which do not show clear differences in the subcellular distribution of EGFR before and after stimulation as a function of ligand identity. We also remind the reviewer that our interpretation is not simply an untested hypothesis – we experimentally tested a GBM-associated EGFR variant whose effect on receptor dimerization has been quantified, and observe EGF-like response kinetics even after EREG stimulation, a result predicted by our model (Figure 3D-E).

   We believe that the sustained membrane-localized signaling we observe might be ascribed to two factors: our choice of cell line and its expression level of EGFR. This conjecture is supported by some data: in contrast to our EGFR-overexpressing NIH3T3 cells, HEK293Ts harboring endogenous or low EGFR levels exhibit a dramatic redistribution of EGFR after EGF stimulation (Figure S3, Figure 6). This is clearly a context where transient versus sustained signaling might depend on the choice of ligand and its consequences on internalization.

   We also note that our data identify ligand-specific signaling differences that are distinct from prior studies, which focused on transient vs sustained signaling downstream of different EGFR ligands. In contrast, we identify a biphasic increase in EGFR activity after stimulation with EGF versus a rapid approach to steady state after stimulation with EREG or EPGN, despite the continued presence of high levels of membrane-localized EGFR in each case.

   Unfortunately, the model that the authors use to test this hypothesis does not even include endocytosis or receptor trafficking but instead uses variable "scaling" factors to see if the data can fit the dimerization hypothesis. In the supplement, they state that "Since our simulations were run on relatively short time scales (~30 min post-stimulation), we did not consider trafficking and degradation of receptors." However, the half-life of EGFR internalization is generally ~3-4min (pmid: 1975591) and degradation ~1hr, so most of the signal shown in Figure 3 is likely to come from internalized rather than surface-associated ligand-EGFR complexes. A further complication is that internalization rates are strongly influenced by receptor expression levels (pmid: 3262110), which are not controlled for here. Thus, the omission of trafficking in their model is not appropriate. This does not mean that the authors are wrong, it simply means that without validation or calibration, their new technology is not ready to resolve current problems in the field.

   We thank the reviewer for pointing out ways to improve our modeling (endocytosis) and discussion of its parameterization (scaling factors). We address both points below:

   Scaling factors: We thank the reviewer for their comments & agree that our discussion of model parameterization was lacking. To clarify: our base-case model for EGF includes 9 parameters, 6 of which are obtained from literature and 3 which reflect lumped kinetic processes of EGFR dimerization and activation and which we set to match our data. We then used experimentally-determined values to change the base-case model to simulate low-affinity ligand stimulation: a fold-change in ligand affinity and a fold-change in receptor dimerization. This is why we simulate EREG with β=50 and γ=100, reflecting the 10-to-100-fold differences in binding affinity and receptor dimerization that have been experimentally measured for this low-affinity ligand. Similar experimentally defined values constrain β and γ in the case of GBM-associated mutations. A more thorough explanation of our model and these scaling parameters is now included in Lines 334-362.

   Endocytosis: We wholeheartedly agree that our model is quite simplified, and a thorough treatment of endocytosis and trafficking would be essential for capturing nuances associated with these steps of the cascade. However, while we appreciate the 3-4 min rule of thumb for EGFR internalization that the reviewer mentions, it is simply not reflective of the membrane-associated EGFR levels we observe in our cells. Examples can be observed in Figure 1C, Figure 2A, Figure 5F, Figure S1B, Figure S2A-B, Figure S5A, and Figure S6, as well as in the quantification of membrane associated EGFR at 0 and 60 min in Figure S5B. It is quite likely that endocytosis and trafficking are operating throughout this time course, but are balanced to maintain similarly high level of EGFR at the cell surface. We wholeheartedly agree with the reviewer’s helpful note that EGFR expression levels heavily influence internalization, which our data also support, and may explain our results. For example, we also see rapid EGFR membrane clearance in HEK293T CRISPR cells (Figure 6) and in HEK293Ts that express low levels of EGFR but not high levels of EGFR (Figure S3A).

   In sum, we argue that our inclusion of additional data showing sustained EGFR protein levels and ZtSH2 recruitment at the plasma membrane should help justify our assumption of membrane-associated signaling in our model. However, we happily concede that this is a highly simplified model, and that endocytosis is a very important process that should be accounted for in future studies (e.g., Line 344-346: “However, we expect that internalization and trafficking can play a crucial role in EGFR dynamics in many contexts, which would need to be included in future models to adequately assess those scenarios”).

   Read the original source
3. 

   eLife

   Mar 10, 2023

   eLife assessment

   This is a well-explained and potentially useful study that describes the generation and use of pYtags, recombinant proteins that, if properly used, should allow spatiotemporal monitoring of the activation of different receptor tyrosine kinases in living cells. Although this study has generated new tools to evaluate receptor localization and activation in different cells, the broad concept showing that different receptor dimers generate specific stimuli, and downstream signaling pathways, is quite limited in terms of novelty. Although it is felt that the study is technologically innovative, the analysis of receptor spatial signaling is incomplete and should be improved.

   Read the original source
4. 

   eLife

   Mar 10, 2023

   Reviewer #1 (Public Review):

   Farahani et al. describe the generation of pYtags, recombinant RTKs, and reporters, that exploit phosphotyrosine/tandem SH2 interaction pairs from immune-specific signaling proteins to allow spatiotemporal monitoring of the activation of different ligand-binding (EGFR and FGFR1) or ligandless (ERBB2) RTKs in living cells stimulated with high and low-affinity ligands (e.g. EGF and EREG or EPGN respectively in the case of EGFR). The study is well-explained and the experiments are clear and clean. Although the authors expanded tool generation to different RTKs and different cells, the potential utility of the approach is limited because the broad concept that different receptor dimers activate different downstream signalling pathways is already well established. Additionally, the results only examine the …

   Reviewer #1 (Public Review):

   Farahani et al. describe the generation of pYtags, recombinant RTKs, and reporters, that exploit phosphotyrosine/tandem SH2 interaction pairs from immune-specific signaling proteins to allow spatiotemporal monitoring of the activation of different ligand-binding (EGFR and FGFR1) or ligandless (ERBB2) RTKs in living cells stimulated with high and low-affinity ligands (e.g. EGF and EREG or EPGN respectively in the case of EGFR). The study is well-explained and the experiments are clear and clean. Although the authors expanded tool generation to different RTKs and different cells, the potential utility of the approach is limited because the broad concept that different receptor dimers activate different downstream signalling pathways is already well established. Additionally, the results only examine the temporal kinetics of the receptors rather than their spatial organization, e.g. in different vesicular/endosomal compartments. The study also describes the use of CRISPR-Cas9 to generate a pYtag knock-in EGFR-expressing HEK 293T cell line to avoid complications arising from over-expression. There were significant differences in terms of receptor activation dynamics comparing knock-in and over-expressed cell lines.

   The study is technologically innovative, yet the analysis of RTK spatial signalling over time in ligand-stimulated cells should be improved.

   Read the original source
5. 

   eLife

   Mar 10, 2023

   Reviewer #2 (Public Review):

   The idea of using fluorescently labeled tandem SH2 domains to target tagged RTKs is brilliant and could potentially provide a powerful new way to assess the activation of RTKs in situ and in multiple physiological contexts. Thus, it was disappointing that there was insufficient characterization of the system to be able to interpret the data it generates. Although the paper shows that tagging the EGFR appears to have minimal impact on its biological activity, the readout for receptor kinase activity is % clearance of the fluorescent reporter tag from the cytosol. Such clearance is likely to depend on a variety of different factors, including the ratio of tagged receptors to probe, the number of functional pools in which the probe exists, the exchange rate between these pools, and the affinity of the probes …

   Reviewer #2 (Public Review):

   The idea of using fluorescently labeled tandem SH2 domains to target tagged RTKs is brilliant and could potentially provide a powerful new way to assess the activation of RTKs in situ and in multiple physiological contexts. Thus, it was disappointing that there was insufficient characterization of the system to be able to interpret the data it generates. Although the paper shows that tagging the EGFR appears to have minimal impact on its biological activity, the readout for receptor kinase activity is % clearance of the fluorescent reporter tag from the cytosol. Such clearance is likely to depend on a variety of different factors, including the ratio of tagged receptors to probe, the number of functional pools in which the probe exists, the exchange rate between these pools, and the affinity of the probes for the tagged receptor. Without determining how each of these factors impacts % clearance, it is difficult to interpret either the dose-response curves or response kinetics.

   For example, the difference in activation kinetics between EGFR and ErbB2 is very interesting, but the almost instantaneous rise (Fig S4B) is very surprising. The kinetics of activation of the EGFR have been extensively studied by mass-spectrometry and are generally limited by ligand binding, which has a characteristic time of several minutes, not seconds (pmid: 26929352; pmid: 1975591). Thus, such a response is suggestive of a freely exchanging ZtSH2 reporter pool that is mostly depleted in seconds with the slow secondary kinetics reflecting a slowly exchanging ZtSH2 reporter pool. Alternately, the cells could be accumulating an intracellular pool of activated receptors over time. That the authors are using concentrations of EGF >100-fold physiological levels (pmid: 29268862) further complicates the interpretation of these experiments.

   There is also insufficient attention paid to either controlling or measuring important parameters, such as expression levels of tagged receptors or levels of endogenous receptors. 3T3 cells, contrary to the statement of the authors, do not have "negligible" numbers of EGFR: they have ~40K, which is typical for mouse fibroblasts. This is much higher than MCF7 cells, which are frequently used as a model system to study EGFR responses. Yet they do not see transactivation of their ErbB2 construct in 3T3 cells without expressing additional EGFR (Fig. 4C), suggesting low sensitivity of the assay. Conversely, they show a significant response mediated by endogenously tagged EGFR in HEK 293 cells, which are frequently used as an EGFR-negative cell line (PMID: 26368334). This indicates that their assay is extremely sensitive. Which is it? As mentioned above, it likely depends on the expression level and affinity of the different components of their system.

   A great advantage of using the EGFR system as a test case for the new system is that thousands of investigations have been performed over the last four decades. This provides a strong foundation for determining whether the new technology is working correctly. For example, the dynamics of EGFR activation and trafficking at the single cell level have been documented in many studies, which show a remarkable consistency (e.g. see pmid: 24259669; pmid: 11408594; pmid: 25650738). Unfortunately, instead of using differences between the new results and previously reported data as a basis for refining their technique, the authors attempt to apply their raw data to address complex questions of EGFR dynamics, with less than satisfactory results.

   For example, they attempt to use their technique to understand the basis of different signaling dynamics between EGFR ligands. Rather than being a relatively recent observation, differences in EGFR ligand signaling have been explored for over 30 years (pmcid: PMC361851), and are generally ascribed to differences in trafficking (pmid: 7876195). Based on these observations and resulting mathematical models, novel EGFR ligands have been designed with enhanced potency (pmid: 8195228 , pmid: 9634854 ). All this work was done over 20 years ago. Since then, new natural ligands for the EGFR have been discovered from sequence analysis and differences in their potency have similarly been ascribed to differences in their intracellular trafficking patterns (pmid: 19531065 - cited by the authors). An alternate hypothesis was proposed more recently by Freed et al (2017) as described by the authors, but that is what it is: an alternative hypothesis.

   Unfortunately, the model that the authors use to test this hypothesis does not even include endocytosis or receptor trafficking but instead uses variable "scaling" factors to see if the data can fit the dimerization hypothesis. In the supplement, they state that "Since our simulations were run on relatively short time scales (~30 min post-stimulation), we did not consider trafficking and degradation of receptors." However, the half-life of EGFR internalization is generally ~3-4min (pmid: 1975591) and degradation ~1hr, so most of the signal shown in Figure 3 is likely to come from internalized rather than surface-associated ligand-EGFR complexes. A further complication is that internalization rates are strongly influenced by receptor expression levels (pmid: 3262110), which are not controlled for here. Thus, the omission of trafficking in their model is not appropriate. This does not mean that the authors are wrong, it simply means that without validation or calibration, their new technology is not ready to resolve current problems in the field.

   Read the original source
6. 

   eLife

   Mar 10, 2023

   Reviewer #3 (Public Review):

   Farahani et al. developed a novel biosensor, pYtag, to monitor receptor tyrosine kinase activity using live cell fluorescence microscopy. The approach to the sensor design relies on adding a tyrosine activation motif to a receptor tyrosine kinase of interest which when phosphorylated recruits a fluorescently-tagged SH2 domain protein. The sensor was used to monitor EGFR and ErbB2 activity and characterize their activity in the presence of different ligands, allowing for the kinetics of receptor activity to be determined in live cells with high temporal resolution.

   The design, characterization, and verification of the sensor with controls were rigorously done and the sensor appears to be a good approach to monitoring receptor tyrosine kinases. In addition to this, the biological characterization of RTK …

   Reviewer #3 (Public Review):

   Farahani et al. developed a novel biosensor, pYtag, to monitor receptor tyrosine kinase activity using live cell fluorescence microscopy. The approach to the sensor design relies on adding a tyrosine activation motif to a receptor tyrosine kinase of interest which when phosphorylated recruits a fluorescently-tagged SH2 domain protein. The sensor was used to monitor EGFR and ErbB2 activity and characterize their activity in the presence of different ligands, allowing for the kinetics of receptor activity to be determined in live cells with high temporal resolution.

   The design, characterization, and verification of the sensor with controls were rigorously done and the sensor appears to be a good approach to monitoring receptor tyrosine kinases. In addition to this, the biological characterization of RTK signaling kinetics allowed for mathematical modeling to determine the dimerization affinity of ligand-bound receptors is the rate-limiting step of receptor tyrosine kinase signaling dynamics. Proving these sensors can be used to monitor biological activities in live cells.

   Initial proof of principles of pYtag was demonstrated in cell lines where the tags were expressed, the authors went beyond this and showed the tagging system could be gene edited to endogenous proteins allowing for the function of receptor tyrosine kinase to be measured under physiological concentrations.

   Read the original source
7. 

   Version published to 10.1101/2022.08.13.503850v1 on bioRxiv

   Aug 15, 2022