Mechanism of phosphate sensing and signaling revealed by rice SPX1-PHR2 complex structure

Phosphate, a key plant nutrient, is perceived through inositol polyphosphates (InsPs) by SPX domain-containing proteins. SPX1 an inhibit the PHR2 transcription factor to maintain Pi homeostasis. How SPX1 recognizes an InsP molecule and represses transcription activation by PHR2 remains unclear. Here we show that, upon binding InsP6, SPX1 can disrupt PHR2 dimers and form a 1:1 SPX1-PHR2 complex. The complex structure reveals that SPX1 helix α1 can impose a steric hindrance when interacting with the PHR2 dimer. By stabilizing helix α1, InsP6 allosterically decouples the PHR2 dimer and stabilizes the SPX1-PHR2 interaction. In doing so, InsP6 further allows SPX1 to engage with the PHR2 MYB domain and sterically block its interaction with DNA. Taken together, our results suggest that, upon sensing the surrogate signals of phosphate, SPX1 inhibits PHR2 via a dual mechanism that attenuates dimerization and DNA binding activities of PHR2.

The manuscript by Zhou, et al. presents a new structure and interaction data with some genetics suggesting how InsP6 mediates biochemical interaction between SPX1 and PHR2. The InsP6-mediated alpha 1 helix stabilization mechanism proposed is very interesting, suggesting the interaction of SPX helix 1 with InsP6 interferes with PHR oligomerization. The new structure and mechanism warrant publication, and overall has potential to be a nice contribution. However, the authors mis-characterize the extent to which InsP6 is required for SPX-PHR biochemical interaction throughout the manuscript and some clarification of experimental details is required. Thus major revisions are required.  Fig 1A show about half the SPX1 co-migrates with PHR2 in the absence of InsP6, but the authors state "SPX1 and PHR2 proteins co-migrated on the column only in the presence of InsP6…". InsP6 generates a new peak, and the crystal structure defines what's in that peak, but it is not clear enough what's in the multiple SPX-PHR-alone peaks to say NONE of them is an SPX-PHR dimer, perhaps in a frustrated alternative conformation. The authors can either focus on the known dimer and their crystal structure describing it, or provide addition orthogonal evidence defining the two states (with and without Ins6P) if they want to talk about InsP6 induction of the dimerized state.
2. Given the dynamics of the Alpha 1 helix mechanism suggested by the authors, some kind of insolution dynamics study would provide orthogonal evidence that InsP6 stabilizes the alpha 1 helix, e.g. HDX, HSQC, SAXS, SANS, crosslinking as in Supp Fig 2A, engineered disulfide trapping or even just CD, limited protease protection or melting temp differential scanning fluorimetry. Any of these studies might also provide additional evidence on the structure / dimerization state of the InsP6-free complex mentioned above in #1.
3. Although the genetics are convincing in Fig 3D- be less PHR loaded on the SEC by OD and in the gels. The SPX mutant has a Y25F change which could (maybe) increase OD a little, but still doesn't full explain the increased SPX OD. The PHR OD should be the same in Fig 1 and Fig 3. Were equimolar amounts of SPX and PHR loaded? If so why the different PHR OD? If different amounts of protein were used the SEC should be re-run with equal molarities of PHR and SPX mutant. Further, since SPX mutant loses function with Ins6P, it would be easy to use Ins7P or Ins8P species to validate the result on different higher order InsP species, addressing two criticisms in one experiment.
Less Major: 4. Rama favored & allowed were in methods but the authors did not mention the 2 outliers in Chain D for some reason. All Ramachandran stats, including the two outliers, should be included in the crystallography table. And the outliers for A61 and R19 explained in the main text. There were also too many real space outliers so the authors should at least explain their attempts to fit those and/or why they had trouble doing so. 5. Methods say two different buffers were used for the ITC, but no mention if they were matched. Authors make statements about the Kd 's while the baseline energies are different. This just needs clarification that buffers were matched, but error for the Kd's need to be reported in order to say Kd's are same or different.
6. Line 173. The authors tested interface residues for SPX-PHR complex formation, but since all the experiments had InsP6 present, they did not test Ins6P dependence. Authors just need to cut out the "InsP6-induced" and then the statement is at least accurate. 7. Line 169, authors state L360A contributes to interface formation based on the crystal structure, but when they tested this residue in Supp Fig 5B, the L360A mutant does not affect interaction (note "the most orange" line in Supp Fig 5B). This phenomenon should be clearly acknowledged in the main text for clarity.

Minor:
8. Line 243, Authors say again the tested MYB interface is important for InsP6-induced interactions but they never test it. Just remove "Insp6-induced" and at least the statement will be accurate.  The PHR family transcription factors are the key players in Pi signaling. In the presence of InsP molecules, SPX family proteins can interact with PHR proteins, inhibiting their gene activation ability and avoiding the toxicity of high concentration of Pi. In this manuscript, Zhou et al. determined the crystal structure of InsP6-SPX1-PHR2 complex, revealed the detailed interaction between SPX1 and PHR2 CC domain. Instead of SPX1-PHR2 interacting interface, in vitro mutagenesis and gel filtration assays confirmed that InsP6 binds to SPX1 and disrupts the dimerization of PHR2 via the steric hindrance of SPX1 N-terminal helices. Unlike the apo-form PHR2, the MYB domain of PHR2 is incompatible with dsDNA binding upon InsP6-SPX1-PHR2 complex formation. All experiments were well designed and performed. This work advanced our understanding of the molecular mechanism how SPX1 inhibits PHR2 from binding to DNA upon sensing Pi signals.

Concerns:
1) Fig. 3a, please check the labels for the right panel. Are the labels K5 and K2 switched?
2) According to Table S1, the B-factors of InsP6 are higher than those of protein and water molecules. What is the occupancy of InsP6? What are the possible causes of the high B-factors? Instead of 2Fo-Fc electron density map, please provide simulated annealing omit map for InsP6 and InsP6-interacting residues. In addition, please include the Ramachandran parameter in Table S1.
3) MYB domain-containing proteins usually bind to dsDNAs. The sequence of P1BS (CGCGAATATTCCCA )is not perfectly self-complementary. Is the complementary strand of P1BS present in the ITC assay? 4) In addition to gel filtration assay (Fig. 4c), it will be helpful to confirm the oligomerization states of SPX1Δ17-PHR2-InsP6 complex by other methods, such as SEC-MALS assay used for other complex. Fig. S3b low panel, there are certain conformational difference between SPX-bound and CtGde1-bound InsP6. Are the InsP6-interacting residues conserved in these two proteins?

Comments and concerns from Reviewer 1
The manuscript by Zhou, et al. presents a new structure and interaction data with some genetics suggesting how InsP6 mediates biochemical interaction between SPX1 and PHR2. The InsP6-mediated alpha 1 helix stabilization mechanism proposed is very interesting, suggesting the interaction of SPX helix 1 with InsP6 interferes with PHR oligomerization. The new structure and mechanism warrant publication, and overall has potential to be a nice contribution. However, the authors mis-characterize the extent to which InsP6 is required for SPX-PHR biochemical interaction throughout the manuscript and some clarification of experimental details is required. Thus major revisions are required. Fig 1A show  Response: We agree with this comment. It is noticed that there is a positive charged patch in PHR2 and negative charged patches in SPX1. Additionally, we found PHR2 could bind to an acidic protein MsyB (a tag) by gel filtration assay. The PHR-MsyB interaction should be mediated by electrostatic force and be not related to their function. It is highly possible that PHR2 could nonspecifically interact with negative charged proteins. We noticed that SPX1 could nonspecifically interact PHR2 under low salt. When we increased ionic strength by changing buffer from 200mM NaCl to 200mM PBS, the non-specific interaction is much weaker. Further, as shown in Figure 1e，the possible nonspecific (dynamic) interaction between SPX1 and PHR2 could not inhibit DNA binding ability of PHR2. We believe the interaction should be dynamic and unspecific without InsP6, and is not related to their function. So, we changed the buffer (increase the ionic strength) and rerun this assay to alleviate the non-specific interaction. We replaced Figure 1a with a new figure to avoid misunderstanding. On the other hand, we emphasized the peak shift of PHR2 dimer in presence of InsP6, and removed "only" (line 86-91). We removed gel filtration data for MBP-tagged or H2A-tagged PHR2 248-380 and SPX1 1-198 since peak shift could be observed when fusion tag existed (line 109-113).

Major criticisms 2:
Given the dynamics of the Alpha 1 helix mechanism suggested by the authors, some kind of in-solution dynamics study would provide orthogonal evidence that InsP6 stabilizes the alpha 1 helix, e.g. HDX, HSQC, SAXS, SANS, crosslinking as in Supp Fig 2A, engineered disulfide trapping or even just CD, limited protease protection or melting temp differential scanning fluorimetry. Any of these studies might also provide additional evidence on the structure / dimerization state of the InsP6-free complex mentioned above in #1 Response: We really appreciate the suggestion. To visualize the conformational changes of SPX1 induced by InsP6, we obtained SAXS data for SPX1 and InsP6-SPX1. The comparison of scattering curve between SPX1 and InsP6-bound SPX1 shows that InsP6-bound state appears to be different via SAXS. Based on SAXS analysis, the Guinier radius of gyration (Rg) and Dmax for SPX1 is ~28.03 Å and 100 Å, larger than 25.33 Å and 89 Å of InsP6-bound SPX, calculated from their SAXS profiles ( Supplementary Fig.10). This result suggests that SPX1 is less compact than InsP6-bound SPX1. Further, we calculated respective molecular bead model using SAXS profiles, and reconstructed the SPX1 model by SASREF ( Supplementary Fig.11). SPX1 and InsP6-bound SPX1 show similar architecture except α1 helix. In addition, the Ab initio molecular envelope show difference around helix a1 area. Meanwhile, we performed thermal shift assay and compared Circular Dichroism (CD) spectra of SPX1 and InsP6-SPX1 ( Supplementary Fig.12). The thermal shift assay showed InsP6 could stabilize SPX1 greatly but could not change the stability of SPX1 △ N17 prominently. InsP6 could increase the content of a helix of SPX1 by the difference of CD spectra of SPX1 and InsP6-SPX1. We added these experiments and assays in our main text (line246-259) to support the function of InsP6. Fig 3D- Response: Thank you for pointing this out. We noticed that we did use less PHR2 than SPX1 in Figure 3c.

Major criticisms 3: Although the genetics are convincing in
We rerun this assay using equal amounts of SPX1 and PHR2, which is also similar to protein loading in Figure 1a. We replaced Figure 3c with a new figure. Regarding to higher order InsP species, we used InsP8 to validate the function of SPX1 mutant. We found that neither InsP6 or InsP8 could promote the interaction between PHR2 and SPX1 mutant. So, InsP8 probably binds to similar position of SPX1 to promote two protein interaction (line 216-219).
Less Major criticisms 4: Rama favored & allowed were in methods but the authors did not mention the 2 outliers in Chain D for some reason. All Ramachandran stats, including the two outliers, should be included in the crystallography table. And the outliers for A61 and R19 explained in the main text. There were also too many real space outliers so the authors should at least explain their attempts to fit those and/or why they had trouble doing so.
Response: Thank you for pointing this out. We use https://servicesn.mbi.ucla.edu/PROCHECK/ server to calculate Ramachandran data. The result is: 93.8% core 5.9% allow 0.4% gener 0.0% disall. Now we noticed it is different from the result calculated by PDB validation. Now we updated our Ramachandran data and mentioned these two outliers (located in the loop region) (line 147-149). For several real space outliers, we have tried serval times but failed to fix it. The probably reason for these outliers is that the quality of density is not high enough to restrain.

Less Major criticisms 5: Methods say two different buffers were used for the ITC, but no mention if they were matched. Authors make statements about the Kd 's while the baseline energies are different. This just needs clarification that buffers were matched, but error for the Kd's need to be reported in order to say
Kd's are same or different.
Response: Thank you for pointing this out. We used buffer F (20 mM HEPES/NaOH pH 7.5, 200 mM NaCl, 2 mM β-mercaptoethanol) for all ITC assay except the assay in Figure 1e (buffer E (200 mM PBS pH7.4) is used). These two buffers are both suitable for ITC assay. As we mentioned in our response to Major Criticisms 1 that 200 mM PBS could alleviate most of nonspecific SPX1-PHR2 interaction, it is more accurate to use buffer E (200 mM PBS pH7.4) than buffer F in ITC assay for evaluating the effect of SPX1 on PHR2 DNA binding. Actually，even using buffer F, SPX1 itself hardly affect PHR2 DNA binding，similar to the result using buffer E. For the difference of baseline energies，we realized that the InsP6 for our previous assay is made by resolving InsP6 powder into Tris buffer, which is not a suitable buffer for ITC assay . Now we made InsP6 solution using water and pH was adjusted to 8.0 by NaOH. Now

Less Major criticisms 6: Line 173. The authors tested interface residues for SPX-PHR complex formation, but since all the experiments had InsP6 present, they did not test Ins6P dependence. Authors just need to cut out the "InsP6-induced" and then the statement is at least accurate.
Response: Thank you for pointing this out. "InsP6-induced" has been deleted (line 188).
Less Major criticisms 7: Line 169, authors state L360A contributes to interface formation based on the crystal structure, but when they tested this residue in Supp Fig 5B, the L360A mutant does not affect interaction (note "the most orange" line in Supp Fig 5B). This phenomenon should be clearly acknowledged in the main text for clarity.
Response: Thank you for pointing this out. When we analyze the structure, The L360 of PHR2 looks like participate the hydrophobic patch formation, however, PHR2 single mutation L360A retained the ability to bind SPX1. Thus, PHR2 L360A does not play the major role in SPX1 interaction. We forgot to remove L360. Now, we mentioned the result of L360 in the maintext (line 181-186).

Minor criticisms 8: Line 243, Authors say again the tested MYB interface is important for InsP6-induced
interactions but they never test it. Just remove "Insp6-induced" and at least the statement will be accurate.
Response: Thank you for pointing this out. "InsP6-induced" has been deleted (line 277). We removed "only" as our response to Major criticisms 1 (line 87). For the discrepancy Line 107 vs. Lines 111-13, as we mentioned in our response to Major criticisms 1, SPX1 and PHR2 form stable, specific interaction under InsP6. The interaction should be dynamic and unspecific without InsP6. When we increase the ion strength in buffer, the non-specific interaction is much weaker. So, we claimed that if stable binding of SPX1 to PHR2 hinges on PHR2 monomerization. Fig Supp 2A Instead of SPX1-PHR2 interacting interface, in vitro mutagenesis and gel filtration assays confirmed that InsP6 binds to SPX1 and disrupts the dimerization of PHR2 via the steric hindrance of SPX1 N-terminal helices. Unlike the apo-form PHR2, the MYB domain of PHR2 is incompatible with dsDNA binding upon InsP6-SPX1-PHR2 complex formation. All experiments were well designed and performed. This work advanced our understanding of the molecular mechanism how SPX1 inhibits PHR2 from binding to DNA upon sensing Pi signals.

Minor criticisms 10:
Concerns 1: Fig. 3a, please check the labels for the right panel. Are the labels K5 and K2 switched?
Response: Thank you for pointing out the label errors. The label errors have been corrected. Table S1, the B-factors of InsP6 are higher than those of protein and water molecules. What is the occupancy of InsP6? What are the possible causes of the high B-factors? Instead of 2Fo-Fc electron density map, please provide simulated annealing omit map for InsP6 and InsP6interacting residues. In addition, please include the Ramachandran parameter in Table S1.

Concerns 2: According to
Response: Thank you for pointing this out. The occupancy of InsP6 is 1.0. It is possible that the density around InsP6 is not good enough to precisely define the position of InsP6. As a result，the B-factors of InsP6 are high. Thank you for the suggestion. We have replaced 2Fo-Fc electron density map with simulated annealing omit map (Figure 3a, supplementary Fig. 7) and put Ramachandran parameter in supplementary Table 1.

Concerns 3: MYB domain-containing proteins usually bind to dsDNAs. The sequence of P1BS
(CGCGAATATTCCCA)is not perfectly self-complementary. Is the complementary strand of P1BS present in the ITC assay?
Response: Thank you for pointing this out. PHR2 transcription factor, a central regulator of phosphate signaling, binds to the imperfect palindromic sequence (GNATATNC). According to Arabidopsis PHR1 and DNA structure 1 , (GNATATNC) the 5 blue colored deoxynucleotides are responsible for DNA binding specificity. It is expected that palindromic sequence will not affect the binding affinity. Therefore, we did not try the self-complementary sequence in our ITC assay.

Concerns 4:
In addition to gel filtration assay (Fig. 4c), it will be helpful to confirm the oligomerization states of SPX1Δ17-PHR2-InsP6 complex by other methods, such as SEC-MALS assay used for other complex.
Response: Thank you for the suggestion. Actually, SEC-MALS analysis indicated that SPX1 Δ17 -PHR2 complex formed by incubation SPX1 Δ17 , PHR2 and InsP6 together was eluted with a molecular mass of about 92.2 kDa (supplementary Fig. 9), consistent with a 2:2 SPX1 Δ17 -PHR2 complex (96.8kDa). Therefore, it will provide further evidence for important roles of a1 helix in dimer dissociation. Fig. S3b low panel, there are certain conformational difference between SPXbound and CtGde1-bound InsP6. Are the InsP6-interacting residues conserved in these two proteins?

Concerns 5: As depicted in
Response: Thank you for pointing this out. Most InsP6-binding residues are conserved. Residues M1, K2, F3, K5, Y25, K29, K147 and K151 in SPX1 participate in the interaction with InsP6. The corresponding residues in CtGde1 also contribute to InsP6-CtGde1 binding. In addition to these residues, 4 more lysine (K30, K128, K132 and K133) from CtGde1 also stabilize InsP6-CtGde1 interaction. Superimposition shows that the InsP6-contacting helixes are not totally fit, which may be the reason for certain conformational difference between SPX-bound and CtGde1-bound InsP6. Previous critiques: 1) Addressed -changed salt to decrease InsP6-free peaks and changed language.
2) Addressed but addition of Chi-square for SAXS data to formally compare datasets +/-InsP6 would be an improvement. Right now its just a qualitative comparison, but a formal, quantitative comparison would be very easy to do. Not required from me but would make it better.
3) Addressed -fixed very sloppy stoichiometry. 4) Addressed -fixed crystallography table and explained real space problems and rama outliers. 5) Authors still do not state syringe and sample cell buffers were MATCHED although other problems with ITC were fixed. What's being injected via syringe must be identical to buffer in the sample cell -I'm sure they did that but they have to state that is the case. Otherwise the ITC data should not be published.
6) Addressed -misleading language removed. 7) Addressed -mistake fixed. 8) Addressed -misleading language removed. 9) Cannot determine -The old AND new line numbers are needed to determine if new statements have been revised.
Reviewer #2 (Remarks to the Author): The authors have addressed all my concerns.
The authors have done an outstanding job improving this manuscript. barring small improvements there is now outstanding evidence supporting the authors claims. Only two minor corrections need to be added: A) Authors still do not say ITC syringe vs. sample cell buffers were matched. This must be stated or ITC should not be published. Response: Thank you for pointing this out. For each ITC assay, the buffer in syringe and sample cell are the same. Now, we mentioned that 'Samples in cell and syringe were both dialyzed against buffer' in the ITC assay method (lines 358-359).

Response:
For the discrepancy line 682 vs.169 in the first version of the manuscript (shown in left), it has been revised as shown in the right (line 158-162 in current version). When we analyze the structure, the L360 of PHR2 looks like participate the hydrophobic patch formation, however, PHR2 single mutation L360A retained the ability to bind SPX1. Thus, PHR2 L360A does not play the major role in SPX1 interaction.
For the discrepancy line85 vs. line 107 in the first version of the manuscript (shown in left), we believe that the weak interaction between SPX1-PHR2 should be dynamic and unspecific in absence of InsP6, and is not related to their function (we have elaborated the conclusion in our previous response to Major criticisms 1). Now, we emphasized the peak shift of PHR2 dimer in presence of InsP6, and removed "only" as shown in right (line lines 75-78, lines 95-96 in current version).
For the discrepancy Line 107 vs. Lines 111-13 in the first version of the manuscript (shown in left), we think they are not contradictious any more since we believe that SPX1 dynamically and unspecifically interacts with PHR2 in absence of InsP6. The specific and stable interaction between SPX1 and PHR2 relies on InsP6. We deleted "Interestingly" and mentioned the phenomena using sentence:'it is noticed that'(lines 95-101 in current version).

Previous critiques:
1) Addressed -changed salt to decrease InsP6-free peaks and changed language.
2) Addressed but addition of Chi-square for SAXS data to formally compare datasets +/-InsP6 would be an improvement. Right now its just a qualitative comparison, but a formal, quantitative comparison would be very easy to do. Not required from me but would make it better. Response: We performed quantitative comparisons of datasets +/-InsP6 as shown in Supplementary  Figure 9. The Rg and Dmax for SPX1 are ~28.03 Å and 100 Å, larger than the values of InsP6bound SPX (25.33 Å and 89 Å respectively). This result suggests that SPX1 is less compact than InsP6-bound SPX1. We also calculate Chi-square for SAXS datasets +/-InsP6 using formula as follows: The value is 11, indicative of obvious difference of two datasets. We also included Chi-square values for our atomic models and SAXS experiment data in the legend of Supplementary Figure 10. Since Rg and Dmax are widely used to quantitatively compare the difference of two datasets, we only put these parameters in Supplementary Figure 9.
3) Addressed -fixed very sloppy stoichiometry. 4) Addressed -fixed crystallography table and explained real space problems and rama outliers. 5) Authors still do not state syringe and sample cell buffers were MATCHED although other problems with ITC were fixed. What's being injected via syringe must be identical to buffer in the sample cell -I'm sure they did that but they have to state that is the case. Otherwise the ITC data should not be published. Response: Thank you for pointing this out. This has been answered in our response to A).