Sequential forward and reverse transport of the Na+ Ca2+ exchanger generates Ca2+ oscillations within mitochondria

Mitochondrial Ca2+ homoeostasis regulates aerobic metabolism and cell survival. Ca2+ flux into mitochondria is mediated by the mitochondrial calcium uniporter (MCU) channel whereas Ca2+ export is often through an electrogenic Na+–Ca2+ exchanger. Here, we report remarkable functional versatility in mitochondrial Na+–Ca2+ exchange under conditions where mitochondria are depolarised. Following physiological stimulation of cell-surface receptors, mitochondrial Na+–Ca2+ exchange initially operates in reverse mode, transporting cytosolic Ca2+ into the matrix. As matrix Ca2+ rises, the exchanger reverts to its forward mode state, extruding Ca2+. Transitions between reverse and forward modes generate repetitive oscillations in matrix Ca2+. We further show that reverse mode Na+–Ca2+ activity is regulated by the mitochondrial fusion protein mitofusin 2. Our results demonstrate that reversible switching between transport modes of an ion exchanger molecule generates functionally relevant oscillations in the levels of the universal Ca2+ messenger within an organelle.

The manuscript by Samanta et al describes the role of the mitochondrial sodium-calcium exchanger (NC X) in shaping mitochondrial calcium oscillations. The authors show that the NC X switches between forward and reverse mode during calcium oscillations. This demonstrates that the same molecule can be responsible for calcium entry and release. Overall, the manuscript is extremely well written and addresses an important point in calcium signalling. The data provide compelling evidence for the tested hypothesis, and I particularly like the link between mitochondrial calcium oscillations and gene expression patterns. Once the minor comments listed below have been addressed, the manuscript can be accepted for publication.
-C omparing Figures 1C and 1G, it seems that there are fewer peaks in the mitochondria than in the cytosol. C ould the authors comment on whether this is an issue with the data analysis, or what mechanism would lead to the discrepancy in the number of peaks? -The authors identified mitofusin 2 as a key component in regulating mitochondrial calcium oscillations. Using pull downs, they show that mitofusin 2 interacts, directly or indirectly, with the mitochondrial NC X, but do not provide a further discussion of this point. C ould the authors expand on the potential direct or indirect interactions? - Figure 3I does not seem to add much to the manuscript. It basically shows that increasing the number of NC Xs increases peak mitochondrial calcium, and that increasing buffering decreases it.
Reviewer #3 (Remarks to the Author): In this manuscript, Samanta et al. explore a fascinating connection between mitochondrial Na+-C a2+-exchanger (NC LX) and mitofusin (MFN2), and the role of these proteins in mitochondrial/cytosolic C a2+ signaling. They demonstrate that knockdown of MFN2 rescues oscillatory cytosolic and matrix C a2+ signals in cells under the conditions where mitochondria are de-energized (with the use of FC C P and oligomycin). They further demonstrate that NC LX is essential in this process and that it involves sequential switching between forward and reverse mode of Na+-C a2+-exchange. They propose that NC LX mode switching is regulated by MFN2. Finally they show that functional NC LX supports gene expression and delays apoptosis in cells with de-energized mitochondria. Although majority of the experiments are performed in RBL-1 mast cells, the conclusions are likely to be applicable to other cell types.
The experiments are generally well thought out, and well executed. My only objection concerns the co-immunoprecipitation in Fig. 2I, which is presented as evidence for direct interaction between NC LX and MFN2. In fact, the ratio of MFN2 to NC LX in the precipitate is exceedingly low, far lower than in the extract. Although it is not unusual to see poor preservation of protein complexes under C o-IP conditions, the experiment shown is totally unconvincing. Furthermore, a proper control would require an unrelated IMM localized myc-tagged protein (not IgG as presented). Since the issue of direct NC LX-MFN2 interaction is marginal in the context of this manuscript, I would suggest to delete panels G,H and I in Fig. 2, and the relevant text in the manuscript.
The manuscript could be improved by including a brief discussion of the role of mitofusin in mitochondrial fusion dynamics and mitochondria-ER/SR connection, especially in the context of heart muscle contractility, which was the subject of several recent publications.
Reviewer #4 (Remarks to the Author): The present study manuscript by Krishna Samanta and coworkers describes the maintenance of cytosolic and mitochondrial C a oscillations after mitochondrial membrane potential (∆ψ) collapse when the mitochondrial fusion protein mitofusin2 is silenced in the RBL1 mast cell line. They provide reasonable evidence that that these oscillations originate from the activity of the mitochondrial Na/C a exchanger (NC LX) cycling between forward and reverse modes. While I have only a few problems with the experiments presented, my major concern is that the study is overly descriptive, provides few mechanistic insights and most importantly reports a phenomenon with no connection to physiology or pathophysiology. First, the conditions described in this study with membrane potential being completely collapsed with protophore, in combination with inhibition of ATP synthesis, together with the knockdown of mitofusin, will never occur under physiological conditions and thus the observations are purely phenomenological and thus can not be plausibly related to any cell function. Specifically, the activity of the Na/C a exchanger is absolutely predicable based on the electromotive driving forces which are artificially imposed over the mitochondrial membrane. Under these conditions, it is not surprising that any cytosolic signal would propagate to these mitochondria. Furthermore, this "reverse mode" of Na/C a exchange in mitochondria has been reported previously under similar conditions (Kim andMatsuoka. J. Physiol. 586 1683-97 (2008) and is thus, not even a novel observation.
The one potentially interesting aspect of this manuscript is the rescue of cytosolic C a signals and subsequently mitochondrial signals by knockdown of mitofusin2, apparently utilizing a fusionindependent mechanism. What is quite remarkable is that MFN2 KD seems to reverse the effects on mitochondrial potential collapse without affecting membrane potential, which begs the obvious and interesting question of how MFN2 KD achieves this? The study however, provides no real mechanistic clues on the action of mutofusin2 within this context, beyond the fact that the 2 tagged, overexpressed proteins appear to co-IP and their distribution overlaps within the cell. The experiments are wholly inadequate to discern mechanism and even experimentally they are superficial; there is no mention of how many times that the experiments were performed and no quantification. In addition, it is surprising that 2 presumably, appropriately targeted, overexpressed mitochondrial proteins do not overlap in distribution in cells at rest (2G) at the resolution of confocal microscopy.
Most experiments make use of protocols that are extremely difficult if not impossible to interpret where a double knockdown (MFN2 and NC LX or MFN2 and MC U) plus a combination of 2-4 drugs (FC C P, oligomycin, C GP, Ionomycin). See e.g. page 4 "…In cells in which we reduced both MC U and mitofusin2 expression and then depolarized mitochondria, oscillations in matrix C a to leukotriene receptor stimulation were rescued only if matrix C a had been increased slightly by Ionomycin prior to agonist exposure". The assertion by the authors that "C ollectively, these results show that FC C P-induced C ell death is delayed by functional mitochondrial Na/C a exchange" is not accurate; cell death would be delayed by MFN2 KD instead for which the mechanism is unresolved.

Other issues
The authors provide no data to confirm that NC LX knockdown is effective.
FC C P will collapse H+ gradients across all membranes. How do the Authors exclude the fact that cytosolic pH changes may contribute to this phenomenon.
The suggestion by the authors that their findings could explain why MC U KO mice show an unexpectedly mild phenotype is misleading. MC U KO mice are viable only on a mixed background and their mitochondrial membrane potential and matrix C a2+ are not significantly affected. Under these homeostatic conditions of hyperpolarized mitochondrial membrane potential, it is biophysically impossible for NC LX to work in reverse mode and underlie an alternative C a import pathway.

Reviewer #1 (Remarks to the Author):
Grundmann et al., investigate the independence and interdependence of G-protein and β-arrestin signaling downstream of GPCRs using a combination of "clean" KO background, inhibitors, receptor mutants and ligand variants. Unfortunately, the majority of conclusions drawn in this study do not provide any novel information, and in a very few places where they do, they refute the large body of data currently available from multiple laboratories.
Response authors: we thank the reviewer for his comments. Clearly, our original manuscript lacked sufficient clarity to unambiguously convey the novelty of our findings. In our revised manuscript, we paid particular attention to enhance the clarity of presentation and to highlight the novelty of our findings. Clearly, our findings are surprising and novel and should be of great interest for a broad readership. In agreement with this notion, one of the other referees considers our study as 'paradigm shifting'. This is what we feel as well. In addition, the quality of the manuscript and the strength of our conclusions have been further enhanced by incorporating a series of new data into the revised manuscript, as requested by the other reviewers.
As discussed in detail in the manuscript, our findings are novel and paradigm-shifting. Our perception of how GPCRs signal have taken intriguing twists over the last decades. I cite from a review of Robert Lefkowitz (Curr Opin Cell Biol 2014): ... it was quite surprising when it was first noted in the mid 90's that, at a single GPCR, different ligands could be biased or functionally selective toward one or another of these G proteins. Even more surprising were the discoveries a few years later that GPCRs could also signal through beta-arrestins and that ligands could be biased towards either a G protein or beta-arrestinmediated pathways ..." Beta-arrestin-mediated activation of ERK is one of the most prominent examples of arrestindependent, G protein-independent signaling. Yet, our work with G protein-and arrestin-depleted cells indicates that it is the other way round: G proteins drive ERK activation and arrestins do not, a clear paradigm shift.
Major points: i) For example, the lack of βarr dependent ERK activation for β2AR has been published recently (Science Signaling, 2017). Authors do not properly cite and discuss this thorough paper which basically reaches very similar conclusions at least for the β2AR.
Response authors: we apologize for not discussing this work in more detail. The study by the Gutkind lab reaches similar conclusions for the β2 receptor. However, our study greatly extends the work by Gutkind et al. by investigating a broad panel of GPCRs and using different experimental approaches including DREADD technology. It is also important to note that our work was carried out completely independently of the Gutkind study. Thus, the Science Signaling paper and the present study very nicely complement each other by providing converging novel insight into the role of betaarrestins in GPCR signaling.
We now refer to Science Signaling study in more detail in the revised version of the manuscript (please see Introduction, page 3, last sentence, and Discussion, first paragraph, yellow text).
ii) Using CRISPR based approach to generate "clean" background devoid of G proteins or βarrs to investigate GPCR signaling have been described previously (JBC, 2017; Science Signaling, 2017 and a few others).
Response authors: we are aware of several papers in which individual Gα proteins or beta arrestins have been eliminated by CRISPR/Cas9 genome editing and cite these papers in our manuscript (e.g. refs 22,23,24,28,67). We were also very careful not to claim any credit for generating these published CRISPR/Cas9 HEK cell lines. However, we do claim novelty for establishing cellular conditions that have not been created before: "zero functional G", which stands for the collective absence of functional Gα proteins from all four major G protein families. Arrestin signaling in general, or more precisely, how GPCRs activate ERK, has never been studied under these conditions. To emphasize the novelty of our findings, we also chose a new title for our manuscript: "Lack of beta-arrestin signaling at 'zero functional G'".
iii) There is pretty much only lab (or collaborators thereof) in the community which uses DMR based approach, and honestly, it is difficult to convincingly state what exactly it reports in cellular context. iv) There are multiple publications from different laboratories establishing the arrestin-biased nature of SII at AT1R; however, the authors cite the single paper which claims that SII can induce some G protein coupling (reference 40). It is fine to discuss this if it supports their finding but a balanced discussion should be included.
Response authors: we thank the reviewer for this comment. We fully agree with the arrestin-biased nature of SII, meaning that it clearly prefers arrestins over G proteins. SII was introduced into the literature as completely arrestin-biased by Bob Lefkowitz and coworkers (Wei et al., PNAS 2003). About a decade later, the very same lab provided a more accurate description of the 'arrestinbiased' nature of SII and refined the molecular mechanism of SII action to reveal its relative preference for arrestins over G proteins rather than an absolute bias for arrestins (Strachan et al., JBC 2014). We feel that this seminal publication is highly relevant for the field because it helped to clarify the mode of SII action. Therefore we cited this work in addition to the previous ref 40. We also rephrased the Discussion to emphasize that a thorough mechanistic understanding of how AT1R activates ERK requires consideration of the full spectrum of biological activities of SII, as revealed by the seminal work of the Lefkowitz group (Strachan et al., JBC2014) (page 11, third paragraph). v) Lines 330-339 -authors write that there is no study on tail vs. core conformation of GPCR-bound βarrs. This is not correct -there are three papers (Nature Communications, 2016;PNAS, 2017;Mol Biol Cell, 2017) that clearly establish the sufficiency of tail conformation to support receptor endocytosis and ERK activation.
Response authors: it was recommended by another referee to remove the above paragraph from the discussion due to its overly speculative nature. Still, we thank this reviewer for pointing out these recent elegant studies, two of which are also cited in our revised manuscript (ref 22,23).
vi) Control of surface population of GPCRs at "ZERO G" by βarrs is not only intuitive (owing to constitutive endocytosis) but also documented experimentally (PNAS, 2016).
Response authors: we thank the reviewer for this comment and assume that he/she refers to recent work of the Benovic lab (Carr et al., PNAS 2016), among other studies (the Carr et al. paper has been included in our reference list). Carr et al. utilized an arrestin-biased ligand to demonstrate arrestinbiased signaling to the ERK cascade. However, this study differs from our work in that it is not done at 'zero functional G'. Instead, an arrestin-biased ligand was employed in cells with functional G proteins. While we understand that this may be interpreted as zero G from the ligand's perspective, our study is at 'zero functional G' from the cell's perspective for which there is no precedence to date. Thus, the tools employed in both studies are vastly different, as are the resulting mechanistic conclusions.
Similarly novel and paradigm-shifting was a recent report from Gosh and Shukla (Nature Nanotechnology 2017). The authors developed an intrabody as generic inhibitor of GPCR endocytosis to realize that indeed, endocytosis and ERK signaling, can be separated for all receptors, contrary to the long-held view that internalization is necessary for βarr-driven ERK signaling.
vii) Interdependence of ERK signaling, at least for Gi coupled receptors, has been documented by many studies, and therefore, the requirement for some level of G protein for βarr dependent signaling is not surprising.
Response authors: we very much agree with this reviewer that the requirement for some level of G protein for βarr dependent signaling is not surprising. What is highly surprising is that -despite of this knowledge -G protein-independent βarr-dependent signaling is widely postulated, even for Gicoupled receptors, despite the fact that Gi-mediated ERK phosphorylation is entirely sensitive to PTX. For example, Walters et al. (JCI 2009) stated in the abstract "In a human cell line-based signaling assay, nicotinic acid stimulation led to pertussis toxin-sensitive lowering of cAMP, recruitment of beta-arrestins to the cell membrane, an activating conformational change in betaarrestin, and beta-arrestin-dependent signaling to ERK MAPK". This implies that cAMP formation is driven by Gi but ERK-MAPK signaling is not, although the results section shows complete dependence of ERK on PTX pretreatment. These statements suggest that arrestins drive ERK and that G proteins are dispensable. Yet, our work with G protein and arrestin-depleted cells indicates that it is the other way round: G proteins drive ERK activation and arrestins do not, a clear paradigm shift.
viii) Authors mention that incomplete suppression of βarrs in HEK cells in earlier studies might explain the discrepancies observed here. However, many of the studies have used KO MEFs which has as "clean" a background as CRISPR-CAS approach. At least for β2AR, Gs-null cells have also been used to document βarr dependent ERK activation (JBC, 2004).
Response authors: we thank the reviewer for raising this point. Mouse embryonic fibroblasts have been used in several studies to understand how beta receptors activate ERK. However, these studies have not led to a consensus view. Shenoy and Perkovska et al.

Reviewer #2 (Remarks to the Author):
Here the authors addressed one of the hottest topics in the GPCR field: the issue of G proteinindependent b-arrestin-dependent signaling. The authors should be commended for several aspects of their study. First, by using genetic elimination of Ga subunits of particular families and of both barrestins with CRISPR-Cas, they avoided ambiguities of siRNA or shRNA knockdown, which is virtually never complete and sometimes results in the knockdown of non-targeted proteins, misleading and undermining interpretation (e.g., compare EMBO Rep 2010, 11 (8) 605-11 and EMBO Rep 2013, 14 (2), 164-71). Second, they tested a large number and variety of GPCRs, including prostaglanding D2 receptor DP2, orphan receptor with synthetic agonist GPR17, free fatty acid receptor FFA2, as well as model receptors used in studies where b-arrestin-dependent signaling was first described, b2adrenergic receptor, vasopressin V2 receptor, and angiotensin AT1 receptor. Finally, the study was complemented with unbiased, Gq-biased, and b-arrestin-biased M3-based DREADDs. Third, wherever possible, they used ligands of these receptors that were reported to be biased. Fourth, in addition to commonly used readouts (cAMP, IP accumulation, ERK phosphorylation), they also used global mass redistribution (DMR) readout, which is not dependent on an individual pathway. Fifth, in addition to signaling readouts, they tested direct b-arrestin recruitment and b-arrestin-dependent GPCR internalization. Thus, in terms of comprehensiveness, this study has few peers. Finally, the authors present all necessary controls.
The results suggest that in all cases detectable signaling requires the activation of G proteins, whereas arrestin recruitment and receptor internalization does not. In some cases the presence of b-arrestins enhances ERK1/2 phosphorylation, suggesting scaffolding functions of these proteins downstream of G proteins. Thus, the authors conclude that there is b-arrestin-dependent GPCR signaling, but it is not G protein-independent.
The data presented are a lot less ambiguous than the results of most previous studies, where protein knockdown was employed, and therefore much more convincing. The authors logically explain previous reports in the context of their model. Overall, this work appears to be paradigm shifting. It would be of great interest to numerous researchers in the fields of GPCRs, cell signaling, and drug discovery. However, several minor changes can further improve the manuscript: Response authors: we thank this reviewer for the very favorable evaluation and constructive comments on our manuscript, and for regarding this work as paradigm-shifting. Thus, contrary to the long-held view that arrestins are signal transducers in their own right, they are not, at least not under the experimental conditions used in our study.
1. The data in Suppl Fig 8 are also consistent with a known fact that simple scaffolds (i.e., those that bring signaling proteins together but do not activate either of them) demonstrate bell-shaped dependence of signaling on scaffold concentrations (theory: PNAS 2000, 97 (11), 5818-23; experimental evidence: Biochemistry 2011, 50 (48), 10520-9; JBC 2013, 288 (52), 37332-42). At relatively low scaffold levels the signaling is enhanced via the formation of complete complexes scaffold-molecules A+B, whereas when the number of scaffold molecules exceeds the number of proteins they scaffold, the signaling is suppressed via preferential formation of incomplete complexes (scaffold-molecule A and scaffold-molecule B, but not scaffold-molecules A+B).
Response authors: we thank the reviewer for providing an alternative explanation to our data. In fact, we do prefer this reviewer's explanation over our own. Therefore both rationalization of our experimental results, as suggested by this reviewer, and the associated literature are now part of our revised manuscript.
2. References where modified M3-based DREADDs were characterized are necessary.
Response authors: we apologize for this omission. Three references are now given for the wildtype, Gq-biased, and arrestin-biased form of the M3-DREADD (refs 51,52,53).
3. Some minor editing is needed: lines 198 and 292, by "barr recruitment" the authors clearly mean "barr-dependent signaling"; line 337, the word "pendant" is inappropriate here and should be replaced.
Response authors: as suggested, we replaced "barr recruitment" by "barr-dependent signaling" or "barr signaling" (page 5, bottom, yellow highlighted text and page 7, bottom, yellow highlighted text). The term "pendant" was removed by deletion of a paragraph (previous lines 330-339) judged as 'overly speculative' by another referee.

Reviewer #3 (Remarks to the Author):
This paper deals with the question of involvement of G protein in arrestin-dependent signallingparticularly to pERK MAPK.
While I believe that the work addresses an important aspect of GPCR signalling and is of potential value to the field, there are multiple issues that need to be addressed.
While it is true that there are groups that talk about "G protein-independent" arrestin dependent signalling (and this tends to be motivated by the focus of these groups on arrestin-dependent signalling), this is not the consensus view for the field. I would argue that most competent researchers in the field have an expectation of an involvement of G protein (and in particular the bgsubunit) in regulation of arrestin recruitment/activation/signalling. Most groups would describe this signalling as simply "arrestin-dependent" rather than also G protein-independent. As such the introduction should be toned down to more appropriately reflect the broad spectrum of views in the field, though it is appropriate to acknowledge that G protein-independence is claimed by some groups.
Nonetheless, the authors are correct in that how these interplay is an area that has not been studied with rigour, and as such, the current paper is potentially important in helping to define this.
For this reviewer, the results are not surprising, but direct, robust experimental evidence is not readily available in the literature, and there are groups that inappropriately ascribe G proteinindependence to particular signalling, without the required evidence for this statement. Studies such as those reported in the current paper are therefore important to address this, but equally they need to be cautious in their description of their data, and additional work is required to support the current claims.
Response authors: we thank this reviewer for recognizing the importance of our manuscript and for her/his constructive comments.

Comments
Intro -while the DMR label free experiments measure integrated responses, it is not correct to term this "pathway-unbiased".

Response authors:
As requested by the reviewer, we deleted the term "pathway-unbiased" in the revised manuscript version.
Line 61, "how they interfere" -do you mean "how they interplay"? Response authors: "interfere" has been replaced by "interplay" Lines 65-66. Signaling should not be described as "unbiased", "G protein-biased" etc, as absolutes. These behaviours are always contextual, dependent upon cell background, require context of pathways studied (and normally the comparator ligands), and require reference as to what the "relative bias" actually describes.
Response authors: we thank the reviewer for raising this critical point. It is very common that ligands and receptors are simply referred to as 'balanced' or 'biased', although it is known that bias is strongly affected by various factors including cellular context. To address this issue, we made several changes throughput our manuscript: -we have deleted the term from the abstract and no longer refer to 'unbiased' and 'arrestin-biased ligands' -we placed the terms 'biased', 'unbiased' within quotation marks throughout the manuscript (this is frequently done by authors who are aware of the conditional nature of this term (see for example lit. -we explicitly mention the cellular context in which the 'unbiased', 'Gq-biased' and 'arrestin-biased' nature of M3-DREADDS was defined (page 8, lines 1-3, highlighted in yellow).
-we have rephrased the Discussion regarding the nature of SII bias to indicate that the authors are aware of the conditional nature of this term (page 11, third paragraph, text highlighted yellow).
Lines 67-68 -"true zero G" -I don't think this is an appropriate term to use, based on the current experiments. I am ok if they wish to use "zero functional G", at least in the context of a-subunit dependent function. However, unless the authors demonstrate that under conditions of over-night PTX treatment there is no G protein present, or minimally that the PTX treated Gi proteins are unable to be recruited to the receptors, then calling the condition "zero G" is misleading. For example, dominant -ve mutant Gi proteins are effectively recruited to activated receptors. This term should not be used.
Response authors: we fully agree with the reviewer. The term "zero functional G" is clearly more appropriate to describe the experimental conditions we have chosen. We replace the term 'zero G' with 'zero functional G' throughout the entire manuscript.
Line 73 -the end of this sentence should provide the caveat that G protein-independence has not been robustly demonstrated (which is the point of the current paper).
Response authors: we now raise the caveat that arrestin-dependence but not G proteinindependence has been extensively studied (page 3, last sentence of Introduction, highlighted yellow).
Lines 78-80 -I would suggest noting that the G protein partners identified for the studied receptors are "canonically studied partners", unless they have evidence that no other G protein coupling occurs (in any system).
Response authors: done, we added 'canonically studied signaling partners' (page 4, line 5) Results para 1. As noted above -the authors do not demonstrate "G protein-independent" recruitment of arrestin, only recruitment in the presence of functional disruption of the Gai subunit. The use of the term "Barr recruitment at zero G" should be revised to reflect the actual experimental evidence.
Response authors: As stated above, we have replaced 'zero G' with the term 'zero functional G' which is much more appropriate. In addition, we provide new experimental evidence with genetically encoded FRET-based biosensors that G proteins are not functional when cells are inhibitor-treated. This new set of data further strengthens the term 'zero functional G', as we demonstrated absence of G protein functionality using canonical second messenger assays (already present in the previous version), and, additionally, absence of activation-induced conformational changes within G protein heterotrimers (new data in Figure 1, panels b,f,g,l,m).

Results -general comment
With the exception of a limited subset of DMR experiments and cAMP, IP1 assays, there is no concentration-response data. This is true for b-arr recruitment (with the exception of Fig 1 B-side note -all figures showing BRET for b-arr should note that it is for arrestin on the Y-axis), and there is no concentration-response data for pERK. These latter 2 assays are the most critical for the arguments that are developed and require proper quantification (not just single high-concentration drug). The DMR may be easier (higher through-put) for the authors, but it only provides confirmatory data (as it measures a highly integrated response). All BRET and all pERK assays should be done as concentration-response.
Response authors: we thank the reviewer for this comment. We have replaced ALL BRET assays in the main body manuscript with concentration-effect data (Figure 1, panels h,I,n,o). We have performed all pERK assays of Figure 3 in concentration-response mode (Supplementary Figure  4d,e,f). Concerning proper quantification of pERK data: we chose to analyze one single high concentration on purpose to quantify pERK at various time points rather than cover a broad range of concentrations at a single time point. Our study was designed to interrogate the role of arrestins as signal transduction drivers using phenotypic integrated cell responses (DMR) and pERK assays. Independent G protein and arrestin pathways have been postulated to mediate angiotensin II AT1R-receptor-dependent ERK activation (Wei et al., PNAS2003;Ahn et al., JBC2004), with G proteins responsible for the rapid early and arrestins for the delayed component. With the goal of discriminating arrestins versus G proteins as ERK drivers, we considered it important to monitor responses over a wide time-frame. Therefore we chose a single high ligand concentration but varied the parameter time. We felt that this is the most sensible approach for identifying ERK drivers.
We did perform pERK assays for all receptors shown in Fig3 in concentration-effect mode, as requested by the reviewer. We arrived at the same conclusion obtained with the single high concentration-effect data: G proteins but not arrestins drive ERK signaling in their own right. To focus on driver but not scaffolding functions of arrestins, we performed ALL assays in kinetic mode.
Please see the below publication to name but one in support in of our experimental strategy: "Kumari et al., Nature Communications 2016: "Of particular interest is the ERK activation at late time points (10, 20, 30 min) which are well established to be mediated by barr-dependent and G proteinindependent pathway" (ref 23 in our manuscript).
Lines 150-154 -The text with respect to EGF-dependent stimulation of pERK is misleading, as it implies that this is not altered, but a comparison of measures on the Y-axes illustrates that this signalling is also attenuated in various treatment conditions (including b-arr-/-cells).
Response authors: we thank the reviewer for pointing out this misleading statement. In fact, EGF is very suitable as viability control to ascertain that inhibitor treatment is specific, yet it is not suited as reference for quantification of receptor stimuli or comparison of signal strength for receptor stimuli across different cell lines. In other words: larger EGF signals do not always coincide with larger receptor signals and EGF signals vary also with the nature of receptor expressed in a given cellular background. For these reasons, EGF is only used as viability control for inhibitor treatment and in cells in which receptor stimuli are undetectable, i.e. G protein-depleted cells.
Appropriate changes have been made in the revised manuscript (page 6, lines 11-13).
Lines 161-162 -as noted above, though it is correct that the described signalling has been described by some groups as "G protein-independent" arrestin-dependent -this is not universally true. This description promotes an impression of absolute opinion within the field. A more balanced statement would be appropriate.
Response authors: we thank the reviewer for raising this point. We realize that our chosen terminology may even imply that the three family A receptor prototypes exclusively signal in a G protein-independent, arrestin-dependent fashion. This is certainly not the case and does not reflect our intended meaning. We rephrased the misleading statement by a more balanced one (page 6, last paragraph, lines 1-5, altered wording highlighted yellow).
Line 164, isoprenaline cannot be described as a full "unbiased" agonist. This is not a sustainable statement as an absolute (see comments above). Similarly, carvedilol, although described by some authors as G protein-independent (line 165), this should not be presented as a universal description.
Response authors: isoprenaline is now referred to as 'synthetic full agonist'. Regarding carvedilol: we now state that only some groups refer to it as partial agonist for β-arrestin-mediated, G proteinindependent ERK1/2 signaling (page 6, last paragraph, last three lines).
Line 171, Ang II cannot be described as an "unbiased" agonist without detailed contextual statements that define the specific conditions under which this was determined.
Response authors: Ang II is now referred to as natural agonist rather than as 'unbiased' ligand and the cellular context in which the arrestin-biased nature of SII has been identified is now indicated (page 7, lines 5,6).

Response authors:
'unbiased' as well as 'Gq-biased' and 'arrestin-biased' DREADDs are placed within quotation marks to indicate that bias is not absolute, and cellular context is also given, along with three citations referring to generation and characterization of these DREADDs in the COS7 and HEK293 cell background (refs 51,52,53).
Line 205, DREADDs are poorly response to ACh rather than "unresponsive" Response authors: as suggested, unresponsive was replaced with poorly responsive (page 8, lines 2,3) Lines 213-214, "all cells without Ga retained EGF responsiveness" -this is true, but they need to note that it is also reduced relative to "WT".
Response authors: done, we added reduced pERK for EGF in relation to responses observed in WT cells (page 8, lines 11,12). As already stated above, we used EGF as viability control but not as reference ligand for normalization of receptor-mediated pERK responses. This is due to the lack of a strict correlation between EGF and GPCR signal strength.
Lines 218-219, also a general comment on analyses. Firstly, statistics -these are not described in methods, and regularly there is no description of what, specifically, is compared. Regularly, there is no error on WT, basal, control etc where this has been normalised. As such, if the statistical analyses is done on data shown in graphs, it is completely invalid as it removes the error within the control.
Response authors: thank you for making us aware of the necessity to correct statistics and to enhance clarity of presentation as to what specifically has been compared. In need of statistical advice and guidance for the revision we received guidance by Dr. John Spouge (NIH statistician; spouge@ncbi.nlm.nih.gov; available for any request on revised statistics).
With this revision, we provide the requested description of statistics in the Materials & Methods. We have furthermore performed numerous corrections/changes as specified below in detail: -In all figure legends, we describe what specifically is compared (Fig.1, Fig. S4, Fig. S6, Fig. S9).
- Supplementary Figures S3,S4,S9 were deleted (corresponding manuscript text does not claim significance) and statistics are not necessary. -Invalid or unnecessary statistics have been removed from all datasets where the manuscript text does not make claims on significance, and statistical tests have been retained exclusively for those datasets where significance is not obvious to the eye or is needed to support specific claims (Fig.1, Fig. S4, Fig. S6, Fig. S9). -Please note that we revised all statistics entirely based on advice given by Dr. Spouge.
In regard to comments in relation to higher expression of M3D-Gq. The data do not appear to support this. Eg. Fig S10 (no difference -even with no WT error). Fig. S12 reports a significant difference (no error for WT), and this also contrasts with data in Fig. S13 that shows raw data for expression and that it is not altered (lower in the M3D-Gq v WT if anything; panel C v panel D). Lines 218-227, the commentary related to reduced pERK in delta Barr1/2 ignores the fact that the EGF response is also attenuated.
Response authors: yes, this is correct. We do not consider EGF as appropriate control for quantification of signaling strength of GPCRs. Variability within and across cell lines and lack of consistent correlation of EGF and GPCR-mediated signals are the reason for this. Therefore we removed the statement made in the original manuscript (lines 218-227) which actually distracts from the main message of the paper: to identify drivers for pERK rather than comment on evidence supporting a scaffolding function for barrs which, in the view of these authors, is undisputed.
Lines 224-225 "… that Barr recruitment at zero G was a salient feature of M3D-Barr". It was not clear to me how the pERK and DMR assays demonstrated this… as Barr recruitment was not measured.
Response authors: Please note: Arrestin recruitment assays have been performed for M3D-barr in WT background and also for M3D-barr and M3D-wt in the DeltaGsix + PTX cells ( Figures Fig. S7B and S8B).
Para (lines 230-250) -the description implies that there is no impact on EGF signalling, however, there is considerable variation to this response (often in-line with other patterns of attenuation). Has this been evaluated statistically? Is this variance taken into consideration when comparing GPCR-dependent responses?
Response authors: we see the reviewer's point. The description states that variability of EGF responses in non-transfected clones is rather small (Fig. 6U-X), which was the case. However, upon enrichment of cells with different receptors, variability increases, and, importantly, does not correlate with the strength of receptor-dependent signals. For example, please compare panel k with l. We chose to share these data but did not use EGF for normalization of receptor stimuli, because there is no obvious correlation between EGF and GPCR-dependent signal amplitudes.
Lines 254-255, this should include the caveat "in this cellular background" Response authors: done, "in this cellular background" was added to the sentence as suggested (page 9, line 16, yellow highlighted text) Line 262, as noted above, the current experiments do not exclude Barr recruitment. This needs to be measured.
Response authors: we agree with this reviewer, our experiments don't exclude barr recruitment. In fact, the opposite is the case. Our experiments were done based on the notion that G proteins engaged and activated by individual receptors are in the OFF mode (based on lack of signaling studied with canonical methods and FRET biosensors based on G protein activation), while arrestins are still recruited. Thus, the last figure addressed the search for a functional correlate of arrestin recruitment in the absence of active G proteins. We chose conditions for each receptor that show inactivity of cognate G proteins but recruitment of arrestin to examine whether receptors still internalize. Clearly, barr recruitment takes place for all receptors when G proteins are inactive (shown for DP2, GPR17, and FFA2 in Fig.1; for M3D in Fig. S8). So yes, barr recruitment has been measured, and the first statement of this paragraph has been rephrased to clarify that we wanted to monitor a functional correlate of barr recruitment when G proteins are OFF (see page 9, beginning of second paragraph).
Lines 267-268, as noted above, treatment with PTX does not "eliminate" all Gi/o, unless they actually do experiments to demonstrate this.
Response authors: we thank the reviewer for this comment and replaced "eliminated" by "inhibited and/or eliminated" because we used PTX+FR for DP2 and GPR17 but required ΔG12/13KO cells + PTX+FR for FFA2 (page 9, line 7 from the bottom of the page).
Line 273 -the reported data do not establish that "G protein-independent" receptor internalisation occurs -though it does speak to effects when Ga-dependent activation is inhibited.
Response authors: thank you for noting this inaccuracy. We now use the term "internalization when G protein activation was inhibited".
Discussion -as noted above -the "G protein-independent" concept is not one that is universal and has been promoted by groups with a particular interest in arrestin-dependent signalling. Similarly, the current study is not "zero G" (unless this is specifically established). DMR is not "unbiased" (see note above). Elimination of G proteins and inactivation are not the same.
Response authors: "zero G" has been replaced by "zero functional G" throughout the manuscript; DMR is no longer referred to as unbiased; elimination and inactivation are used where appropriate (all replacements are boxed yellow).
Lines 312-313, is there any evidence that inactivation of Gq/11 by the inhibitors prevents recruitment of G protein. Also, when doing G protein recruitment experiments, this should be done as both GPCR to Galpha and GPCR to Gbg (as these are not equivalent when measured).
Response authors: we thank the reviewer for raising this issue. As we have not measured recruitment of PTX-treated Gi or FR-treated Gq to ligand-activated receptors, we followed the advice of this reviewer and replaced 'zero G' by 'zero functional G'. We agree that it would be a very interesting follow-up project to further define 'zero functional G' and to perform studies such as those suggested above.
Line 320, it is not clear to this reviewer why they find it "puzzling" that SII is applied to and referred to as arrestin-biased. The literature is littered with investigators who do not properly understand concepts and quantification of biased agonism (I would include the current authors based on some of the statements in the current paper).
Response authors: we have replaced "puzzling" by "unfortunately" because we are surprised that SII is treated in the literature by some as perfectly biased towards arrestin although it is not. We have rephrased this section to clarify that we understand the conditional nature of the term bias (page 11, third paragraph, yellow text).
Line 324, correctly notes that all data are in HEK 293 cells. In this context, it would be appropriate to discuss some of the background-dependent variables that are at play, including levels of GRKs, lipid rafts, etc.
Response authors: we have dedicated a short section to background-dependent variables (please see page 11, last paragraph).
Lines 328-329, do the authors really mean "arisen" or "proposed", the latter would be a more appropriate descriptor.
Lines 333-337, there is speculation about G protein-independent binding of arrestin and implications from structures in complex with receptor. In the limited studies with full receptor, the "pendant" binding that occurred was in presence of G protein. This section is overly speculative, in the view of this reviewer.
Response authors: we have deleted this overly speculative paragraph from the discussion.
Line 340, implies that "G protein-independent recruitment" of Barr has been established in the current work, whereas this is not true.
Response authors: we agree with the reviewer. The new wording 'arrestin recruitment in the absence of active G proteins' or 'barr recruitment at zero functional G' is much better supported by our experimental data. We used this expression as it is widely found in the scientific literature and adopted the term without questioning its accuracy.
Lines 345-346, while I agree that the gene deletion cell lines are an important tool, it needs to be noted that no fully G protein ablated cell line has been created (and indeed this may not be tolerated by cells). As such, there are still the limitations discussed above that need to be addressed.

Response authors:
To address this concern, the term "collective absence of G proteins" has been replaced with "collective absence of functional G proteins". Furthermore, we added to the Discussion that no fully G protein ablated cell line has been created (please see a new section on this topic on page 11, top paragraph).
In all pERK experiments, data is presented as fold-over-basal. In order for this to be meaningful, actual basal levels need to be reported.

Response authors:
We agree with the reviewer. Presenting data as fold over basal may become problematic if basal levels show significant variations. This was not the case in our study. Basal ERK levels varied to some extent between cell lines and inhibitor treatments, yet were in a very similar range to calculate fold over basal with great confidence. Please see below the distribution of basal values for the DP2, the first receptor shown in Fig.3 of the main manuscript. We have not prepared Figures or Tables with basal values (the paper contains so much information already) but would be happy to do so upon specific request by this referee. Under "Imaging" (p16), in these experiments PTX treatment is only for 1h. Under this condition, the effect of PTX is likely to be one reflected as activation of Gi. It is the O/N treatment that leads to depletion of the activateable pool. Why did the authors perform these experiments??
Response authors: thank you for making us aware of this typing error; PTX treatment was for 18 h. This is now corrected.
The reference list is not uniformly formatted.
Response authors: the reference list is now uniformly formatted.
In figures, DMR traces are "representative" but also apparently mean+SEM of 3 independent experiments (lines are all solid, and it is unclear what this means, or why, in this format only +SEM would be displayed).
Response authors: we have added a statement to the DMR method section to explain that each trace represents triplicate technical replicates over time presented as mean + SEM. Each experiment is repeated at least three times to obtain independent biological replicates.
Throughout the figure legends there is no detail on the specific measures that statistical tests are applied to.
Response authors: we are now more specific in all figure legends where statistics is applied (Fig. 1,  Fig. S4, Fig. S6, Fig. S9).
For DMR experiments, quantitative data is presented for only a limited subset of data (even where concentration-response curves have been performed). Where is the quantitative mean data?
Response authors: -quantitative mean data of Fig.1, panels A-F are in Fig.1, panels G-I; -quantitative mean data of Fig Response authors: we corrected the figure caption and replaced "statistical analysis" by concentration-effect curves… Fig. S15, a band in the Barr2 blots is claimed to be non-specific. What is the evidence that supports this statement?
Response authors: As pointed out by this referee, the anti-beta-arrestin2 monoclonal antibody exhibited two immunoreactive bands in the parental HEK293 cells, and the higher MW bands remain in the beta-arrestin1/2-mutant clones. For the following reasons, we concluded that the lower MW (major) band corresponds to beta-arrestin2 and the higher MW (minor) band reflects non-specific reactivity of the antibody.