Dipeptidyl peptidase 9 triggers BRCA2 degradation by the N-degron pathway to promote DNA-damage repair

Article activity feed

1. 

   eLife

   Aug 26, 2020

   ###This manuscript is in revision at eLife

   The decision letter after peer review, sent to the authors on July 29 2020, follows.

   Summary

   The manuscript by Silva-Garcia and colleagues addresses an interesting aspect of BRCA2 biology: the regulation of its steady state levels in response to DNA damage. It has become clear that BRCA2 level are subjected to change in response to a number of different environmental challenges, including the induction of various types of DNA lesions. This manuscript identifies the serine protease DPP9 as cleaving off the first two N-terminal amino acids of BCRA2 to target BRCA2 towards the N-degron pathway. The DPP9-mediated turnover of BRCA2 regulates the BRCA2-RAD51 stoichiometry and appears to promote RAD51 focus formation.

   Overall Evaluation

   The biochemical data are generally solid and support the …

   ###This manuscript is in revision at eLife

   The decision letter after peer review, sent to the authors on July 29 2020, follows.

   Summary

   The manuscript by Silva-Garcia and colleagues addresses an interesting aspect of BRCA2 biology: the regulation of its steady state levels in response to DNA damage. It has become clear that BRCA2 level are subjected to change in response to a number of different environmental challenges, including the induction of various types of DNA lesions. This manuscript identifies the serine protease DPP9 as cleaving off the first two N-terminal amino acids of BCRA2 to target BRCA2 towards the N-degron pathway. The DPP9-mediated turnover of BRCA2 regulates the BRCA2-RAD51 stoichiometry and appears to promote RAD51 focus formation.

   Overall Evaluation

   The biochemical data are generally solid and support the conclusions, but the interaction has not been tested with the endogenous proteins and the affinity is low (~17 uM). The cell-based assays reveal a potentially significant problem in the BRCA2 construct. Overall, the physiological relevance is far from certain. To be conclusive, the PLA would need single primary antibody-only controls to ensure specificity. More importantly, although it seems possible that DPP9 has an effect on RAD51 foci, it is not clear from the results whether this is directly connected to BRCA2 or is an indirect effect. In particular, the results with the truncated form and full-length BRCA2 are done in overexpression; however, the levels of the two proteins are different and it is known that BRCA2 overexpression is toxic for the cells. Stable expression or equal transient expression would have been more convincing to draw these conclusions or at least rescuing the phenotype of full-length with the truncated form and/or with WT DPP9 would have been more adequate. The overall model needs to be significantly toned down in light of available data from cBioportal and DepMap that are inconsistent with the overall model that DPP9 cleavage is essential for BRCA2 function.

   Essential Revisions

   1. As the authors recognize, that there is a conceptual problem with the model that suggests that BRCA2 cleavage and degradation is required for RAD51 focus formation, when it is BRCA2 function that is required for RAD51 focus formation, as shown by genetic studies. How can the degradation of an essential factor for RAD51 focus formation be required for RAD51 focus formation? If, as the authors suggest, the BRCA2 : RAD51 ratio is critical, RAD51 overexpression should be detrimental. The authors may want to consult the literature on RAD51 overexpression and its effects and discuss this.

   2. Figure 1: Based on the data presented in Figure 1 panel A, it is concluded that DPP9-BRCA2 PLA signal is detected in the nucleus. However, that is hard to assess, as the visualized dots localizing in the DAPI stained areas could easily be on top of the nucleus. In addition, is there any evidence that DPP is actually a nuclear protein?

   The PLA experiments seem to lack the single antibody negative controls that require to account for nonspecific signal.

   Fig. 1D: This figure reports the results from the SPR as (partial) binding isotherms based on equilibrium analysis.

   Include the SPR sensorgrams (for example in supplemental figure 1) to show that equilibrium analysis can be performed on this dataset, thus if: • Observed response is due to specific binding of peptide to enzyme (methods section does not explain if reference surface subtraction has been performed and how much specific binding occurs). • Binding actually reaches equilibrium after 4.5 minutes association phase (rather than continuing to rise). • Dissociation occurs during dissociation phase of 7 minutes, at least for majority of signal, to make sure reaction is reversible (expected for micromolar range).

   Would it be possible to add inhibitor to the association phase to verify specificity of observed interaction?

   Saturation is far from complete. The uncertainty of the fits themselves (as reported in Graphpad analysis) are probably much larger than the reported variations between triplicate KD determinations, especially for the shorter peptide where not even 30% of the enzyme seems to become occupied at the highest peptide concentration. The uncertainty from extrapolation of maximum observed 50% and 25% binding for the different peptides towards 100% response plateau will have major influence on apparent KD. The uncertainty of the fits has to be considered in deciding whether the shorter peptide binds significantly weaker than the intact peptide.

   The enzyme on the chip surface will cleave the peptide and will retain the dipeptide but release the rest of the peptide (according to the observed electron density in the crystals). The intact peptide (40 amino acids) will results in an approximately 20-fold higher response in the SPR assay than the dipeptide (2 amino acids). Depending on the kinetics of peptide binding, the chemical conversion and product release, the observed signal in the sensorgrams will be mostly substrate, mostly product or a mixture of these, the composition of which changes during the experiment. Even if a steady-state binding level is observed in the raw data, the obtained parameters are probably not reflecting equilibrium binding constants. An active site mutant may be helpful here (and might return a much stronger affinity for the intact peptide). If that is not possible, a more qualitative reporting of the data could be considered rather than trying to obtain affinities.

   Fig. 1 S2: DPP9-BRCA2 PLA signal seems to increase with MMC in siBRCA2 treated cells. This potentially indicates that the PLA signal does not report on the DPP(-BRCA2 interaction.

   Why does removing FLNA reduce the BRCA2-DPP9 interaction? Is the BRCA2 pool bound to FLNA targeted by DPP9?

   Fig. 1B, G. It is unclear what is shown on the Y-axis, total fluorescence, number of foci?

   Fig. 1C, D: The affinity found for the interaction between BRCA2 and DPP9 is ~17 uM which seems too weak for a physiologically relevant interaction. Nonetheless the interaction appears to be real as they find it also with purified fragments. An immunoprecipitation with endogenous proteins with and without MMC treatment would be necessary to complement these findings in cells as the PLA is not sufficient.

   Fig. 1G: Please provide the statistical analysis of the (-) and (+) 1G244 samples under MMC treatment. This is the key control and not the comparison to -MMC.

   1. Figure 2: Fig. 2: If DPP8 and DPP9 crystals with the BCRA2 peptide are so similar, why is there a phenotype of DPP9 mutants? Why is DPP8 then not providing redundancy?

   Fig. 2: While the disorder in the DPP9 crystals is clear from the average B-factors (Table 1), what are the local B-factors for the active sites in both structures? Is the quality of the electron density in the active site of DPP8 actually good enough (~3 Å) to establish that the identity of the two amino acids are methionine and proline? Can it be ruled out that coincidentally one might be observing 2 residues from only partially ordered longer peptide? What other information from previously determine enzyme-inhibitor complexes (Ross et al 2018) and active site geometry is maybe used in concluding that the density corresponds to the N-terminal dipeptide?

   1. Figure 3: Fig. 3 A-D: The DNA damage sensitivity assays lack a control with siBRCA2 cells to show the sensitivity compared to knock-down of DPP9, without that it is difficult to interpret the results. This is especially relevant if the point is to show that DPP9 is required for the function of BRCA2 in DNA repair. The observed sensitivities to DNA damaging agents are very mild (plot axes are 1 log). Given that for example sensitivity of BRCA2 mutants to PARP inhibition is extreme, can the effect of DPP9 on survival be indirect?

   Fig. 3 E. The interpretation and description of these results do not appropriately reflect the data. HeLa DPP9 KD cells start with a higher constitutive level of gammaH2AX but the overall kinetics of the increase and decrease appears very similar. The statistical analysis, hence, is not appropriate to test a repair defect. How do normalized curves look like with 0 hr set to 100% and what is the statistical analysis of normalized curves?

   Fig. 3F. Same problem as in Figure 3E, although the increase in gammaH2AX signal is more dramatic. However, in the present illustration, it remains unclear, whether there is a defect in repair as measured by decrease of the gamma H2AX signal.

   1. Figure 4: Fig. 4A: Please provide statistical analysis of siDPP9 +/- MMC and siNT versus siDPP9 + MMC. The analysis provided -MMC siNT and +MMC siDPP9 is not helpful.

   Fig. 4 C and Fig. 4S1: When comparing the signal labeled BRCA2 in, for example Figure 4 panel C, with that in Figure 4 - figure supplement 1, it is difficult to understand that the signal represents the same protein, as in one blot the signal is a collection of bands, while in the other it appears to be a defined band.

   Fig. 4C: Is vinculin a proper control for quantitation of the levels of BRCA2, given that its signal appears out of linear range?

   Fig. 4I, J, line 373: The levels of BRCA2 1-1000 are different compared to BRCA2 3-1000. As this is a transient transfection the difference in the levels might be due to the quality of DNA of one plasmid versus the other. It is not clear that it can be concluded that BRCA2deltaMP is less stable than the unmodified N terminal BRCA variant. This is because the expression levels of the two variants is so different and both seem to decrease (relatively to their starting signal to the same extent). See overall evaluation.

   Fig. 4S2: What is the rational for using a different control protein to measure levels of RAD51 and BRCA2?

   Fig. 4S2: The scheme in G cannot relate to E and F, as RAD51 is analyzed there. Does it maybe relate to Figure 4 I and J. If so this should be indicated in the figure legend to Figure 4I/J and corrected. I suggest moving Figure 4S2G to Figure 4 as part K. What is the evidence that the ubiquitin is cleaved and how efficient is cleavage?

   1. Figure 5: It was very surprising that the WT BRCA21-3148 plasmid was not functional at all with the level of RAD51 foci being equal to the no plasmid negative control. Is it because of a construct problem? One would expect the WT construct can still rescue RAD51 focus formation but with a lower level than the mutant BRCA23-3148, as in Fig. 5C there are some RAD51 foci in siDPP9 cells under 300 nM MMC treatment as well as in Fig. 5-figure supplement 1F. If WT BRCA21-3148 is not functional at all without DPP9 catalysis (as suggested by Fig. 5F and 5G), how do the authors explain that N-terminally tagged BRCA2 variants are still functional, for example GFP-BRCA2, 2xMBP-BRCA2, which could not be processed by DPP9 as the dipeptide Met-Pro is not at the desired positions? This is a potential red flag and questions the validity of the central conclusion. The complementation activity of the constructs must be tested in a BRCA2-deficient background with proficient DPP9. The experiment in Figure 5F lacks a positive control to evaluate the level of complementation by the 3-3418 BRCA2 construct. This should be easy by omitting BRCA2 siRNA and using a scrambled control and no siRNA.

   Fig. 5A. The difference in RAD51 bound to chromatin is not clear and there is no quantification.

   Fig. 5B, please provide statistical analysis of the comparison of siNT versus siDPP9 + MMC. Also, the description in line 453 does not appropriately reflect the data, as there is a RAD51 focus signal in DPP9 depleted cells.

   Fig. 5B, D, F: it is important to show the real number of foci to determine whether the RAD51 foci per cells correspond to what is known from the literature and to find out the effect of the DPP9 KD alone to the number of RAD51 foci. For Fig. 5F this data is presented in Figure 5 suppl. 1 F.

   Fig. 5C: The detection of RAD51 foci is of poor quality. In addition, why under some conditions there is a strong cytoplasmic signal? The effect of different treatment on nuclear size is perplexing. The scale bar indicates 10 µm in every panel, yet nuclear size varies by ~2-4 fold.

   Fig.5. S1E. The levels of BRCA2 WT vs BRCA2delta MP are different, even if the amount of plasmid transfected is the same this does not mean the quality of the DNA is the same so the differences in the levels could be due to this. The quantitation is missing. Is this result reproducible?

   Fig 5 Suppl. 1F. This panel shows that the number of RAD51 foci is increased in cells complemented with a plasmid expressing 3-3418 compared to the full-length BRCA2. However, there are several possible issues in this experiment. A WB showing the control that the siBRCA2 worked in the same cells that overexpress BRCA2 and BRCA2deltaMP. The levels of BRCA2deltaMP is reduced compared to FL-BRCA2. The BRCA2 cDNA is big and its transfection is rather toxic for the cells so a reduced transfection efficiency could lead to higher survival and possibility of repair leading to increased RAD51 foci. What are the levels of gH2AX foci in both cells? If there is a difference in the amount of damage this could also lead to differences in RAD51 foci. The number of RAD51 foci in cells depleted of BRCA2 seems rather high compare to the ones reported in the literature. It remains unclear if the small difference in RAD51 foci between the two BRCA variants could be related to difference in their expression or heterogeneity in expression in the cell population.

   To assess the relationship of DPP9 and BRCA2 in DNA repair the phenotype of BRCA2 1-3418 should be rescued by WT DPP9.

   There is no convincing evidence to suggest that DPP9 regulates RAD51 filament formation by processing the BRCA2 N-terminus as stated in line 474. The authors examine RAD51 foci not filaments.

   1. In Figure 6 S1, the authors show a correlation between low DPP9 expression in breast cancer and patient survival. These data support the significance of DPP9 in breast cancer. However, a quick database analysis in cBioportal reveals that DPP9 and BRCA2 deficiency co-occur, which is not expected from the model presented here. Moreover, cBioportal also shows that DPP9 is often deleted in ovarian cancer but amplified in breast cancer. Again, these data are not consistent with the simple model presented here. Finally, analysis in DepMap shows that DPP9 in not essential whereas BRCA2 is an essential gene. These data do not support the model that DPP9 is essential for BRCA2 function. The authors should consider these available resources and refine their interpretation.

   2. There is a concern about the use of fragments. This appears acceptable for the structural analysis but in Figure 4I and J it is problematic for the stability experiments. In Fig. 5S1E, the full-length BRCA2 behaves consistently, but the analysis is very limited and not quantified. Is this finding reproducible? The reason this could be a bigger issue is the concern about the BRCA2 construct and the absence of complementation activity discussed above.

   Read the original source
2. 

   Version published to 10.1101/2020.08.24.265033 on bioRxiv

   Aug 25, 2020