AcrIF9 tethers non-sequence specific dsDNA to the CRISPR RNA-guided surveillance complex

Bacteria have evolved sophisticated adaptive immune systems, called CRISPR-Cas, that provide sequence-specific protection against phage infection. In turn, phages have evolved a broad spectrum of anti-CRISPRs that suppress these immune systems. Here we report structures of anti-CRISPR protein IF9 (AcrIF9) in complex with the type I-F CRISPR RNA-guided surveillance complex (Csy). In addition to sterically blocking the hybridization of complementary dsDNA to the CRISPR RNA, our results show that AcrIF9 binding also promotes non-sequence-specific engagement with dsDNA, potentially sequestering the complex from target DNA. These findings highlight the versatility of anti-CRISPR mechanisms utilized by phages to suppress CRISPR-mediated immune systems.

A cryo structure was also determined for the complex incubated with dsDNA, revealing binding of the DNA to the surface of F9, allowing mapping of the interface as a basic area of F9 with conserved basic residues (see point 5). The authors speculate that F9 functions partly by sequestering Csy on random DNA as well as blocking target DNA binding.
Overall there is some decent cryo-EM data here paired with some preliminary, contradictory and variable quality EMSA data. This leaves any conclusion beyond "F9 probably functions like F1" to be on rather shaky ground.
Specific points: 1. AcrIF9 is described as a 5-stranded anti-parallel beta sheet cradling an alpha helix, but no structural comparison is offered. Are there any similar folds in PDB? Can the structure be compared more fully to AcrIF1, given their striking convergence in binding sites? Can the two binding sites be compared, beyond saying they are "similar". 2. The "triplicate" EMSAs shown in figure S3 are strikingly different from one another to a surprising degree. For example, the third gel in S3b is quite fundamentally different from the first gel. In the second and third replicates it looks like there is a significant proportion of shifted material half way up the gel over a wide protein range, suggesting an intermediate species, but the gel in figure 2 doesn't conform to this. These experiments should be repeated to give more confidence in the data and their interpretation. Ideally, an alternative technique such as fluorescence polarization should be considered, as binding or displacement of Csy will give rise to a large anisotropy change using a method that measures true equilibria. 3. Figure 2a suggests that one tenth molar ratio of AcrIF9 almost completely blocks binding of Csy to target DNA. Given that at least 90% of the Csy complex should be free and available to bind DNA (assuming non-cooperative binding of AcrIF9), what is the authors' view of the mechanism here? Notably, in the replicates, this effect is not observed at 20 nM F9 to nearly the same extent, and in the middle gel of S3a there appears to be 50:50 binding to target DNA at a molar ratio of 1:2, which could make more sense. 4. It would help to understand how Csy binds to target and non-target DNA in the absence of F9. Although this has no doubt been published previously it would be helpful to show this data for the DNA species used here. It is important to understand how F9 affects DNA binding for both target and non-target DNA. 5. A binding interface for F9 on dsDNA is observed that does not involve the Csy complex. The text states that "we attempted to make structure-guided mutations of these residues, but the F9 mutants failed to bind the Csy complex". Leaving aside the observation that there is no data in the paper to support this (were all 5 made? Did they purify as for the wild-type protein?), this is another puzzling observation as the binding interface between F9 and Csy has not been modified.

Reviewer #2 (Remarks to the Author):
Hirschi et al. present the cryo-EM structures of Csy-AcrIF9 complex and Csy-AcrIF9-dsDNA complex at ~3.9 Å and 6.9 Å respectively. In combination with biochemical data, the authors show AcrIF9 not only sterically blocks the binding of target DNA to crRNA, but also promotes nonspecific recruitment of dsDNA, potentially sequestering the complex from target DNA. Even without the high-resolution structure of Csy-AcrIF9-dsDNA complex, the biochemical data and the current density map support their conclusions. This paper significantly broadens the current view of the mechanisms for anti-CRISPR proteins mediated immune suppression, and will be of interest in the community given that CRISPR-Cas immunity is of broad interest.
During the revision session, the structure of Csy-AcrF9 complex has been described in a PNAS paper (PNAS first published March 13, 2020 https://doi.org/10.1073/pnas.1922638117). With 2.57 Å high-resolution Csy-AcrF9 complex structure, the authors of the PNAS paper proposed binding of AcrF9 to the Csy complex prevents target DNA binding to crRNA, similar to how AcrF1 works, also in line with part of Hirschi et al. 's conclusions. The author may briefly discuss the results in the PNAS paper. Even the high-resolution structure of Csy-AcrF9 complex recently has been reported, as far as I see, the non-sequence specific dsDNA bound to Csy-AcrF9 complex reported in Hirschi et al. paper significantly advances the understanding of how anti-CRISPR proteins work, I still support this paper to be published. 2) The authors should provide detailed information in the method session about how Csy-AcrIF9 and Csy-AcrIF9-dsDNA complexes are assembled. Fig.1a and Supplementary Fig.4a, it's confusing about the peaks they choose. The authors might use different columns to purify the complex, are the two major peaks in each file related to each other in these two gel filtration files? The authors present structural and biochemical analyses of the anti-CRISPR protein AcrIF9 that targets the type I-F CRISPR-Cas system. They have determined the 3D structure of AcrIF9 bound to the RNA-guided surveillance complex (Csy) by cryo-EM at an overall resolution of 3.9 Å. Two copies of AcrIF9 bind to the Csy complex at positions reminiscent of the AcrIF1 binding sites. AcrIF9, like AcrIF1, blocks target DNA binding in Electrophoretic Mobility Shift Assays with substoichiometric ratios of AcrIF9 to Csy complex. However, at higher stoichiometric ratios, AcrIF9 induces target and non-target DNA binding to Csy. The 3D reconstruction of the Csy-AcrIF9-dsDNA ternary complex, determined by cryo-EM at an overall resolution of 6.9 Å, shows that AcrIF9 tethers dsDNA to the Csy complex. As acknowledged by the authors, a similar phenomenon of anti-CRISPR-induced non-specific DNA binding to CRISPR-Cas surveillance complex has previously been reported for AcrIIA11 that targets the type II CRISPR-Cas9 system (https://doi.org/10.7554/eLife.46540). Although the manuscript under review presents the first example of a type I anti-CRISPR protein promoting non-specific DNA binding to the Csy complex in vitro, major revisions are required to strengthen the authors findings and provide an advance in understanding anti-CRISPR proteins. In particular, the role for dsDNA binding in AcrIF9's mode of inhibition remains elusive. What would be the benefit of a concentration-dependent two-pronged CRISPR-Cas inhibition mechanism while the inhibition of target DNA binding and cleavage is effective at low concentration of anti-CRISPR protein (substoichiometric to stoichiometric ratios of AcrIF9 to the Csy complex)? Noteworthy, the paper by Zhang K et al. entitled "Inhibition mechanisms of AcrF9, AcrF8, and AcrF6 against type I-F CRISPR-Cas complex revealed by cryo-EM", which has been published in the course of reviewing this manuscript (www.pnas.org/cgi/doi/10.1073/pnas.1922638117), presents a highly detailed cryo-EM structure of AcrIF9 bound to the Csy complex at an overall resolution of 2.6 Å and shows that AcrIF9 functions via the competitive binding to DNA binding sites on Cas7f subunits. Overall, further data interpretation and additional experiments could greatly improve the manuscript. Besides, all methodological details should be provided for a better understanding of the results and for the sake of reproducibility.

3) By comparing Supplementary
Major comments 1) Page 3: "To achieve this goal, we co-expressed and purified the crRNA-guided surveillance complex (Csy) with AcrIF9 and determined the structure […]" -Here one understands that AcrIF9 together with the crRNA-guided surveillance complex were coexpressed. However, there is no paragraph related to this procedure in the Methods. Alternatively, if AcrIF9 and the Csy complex have been produced independently, as described in the Methods, and then mixed to make the complex, this procedure should be described in the Methods. This point should be clarified.
2) Fig 1, panels d-e; page 4: "The two AcrIF9 binding sites are nearly identical. Residues interacting with each of the AcrIF9 molecules superimpose with an r.m.s.d. of 1.4 Å (see methods). The contacts include: a hydrogen bonding network involving AcrIF9 residues 11 to 15, a ß-sheet on the Cas7f thumb, and a nucleobase from the crRNA (Fig. 1c); interactions between AcrIF9 arginine 17 and residues in the Cas7f thumb (Fig. 1d); and the formation of an "aromatic clamp" around the Cas7f thumb loop by AcrIF9 residues 39 to 41 (Fig. 1d). AcrIF9.1, but not AcrIF9.2, also interacts with Cas8f (Fig 1e)." -This description of the intermolecular contacts between AcrIF9 and the Csy complex should be revised for a better reading and understanding. -The authors mention an "aromatic clamp" formed by AcrIF9 residues, which suggests hydrophobic or stacking interactions between AcrIF9 and Cas7f residues. However, panel d shows that these aromatic residues hydrogen bond with Cas7f. -The authors mention "interactions between AcrIF9 arginine 17 and residues in the Cas7f thumb". The term "interactions" is vague. According to panel d, these interactions are hydrogen bonds, as are all intermolecular contacts described in this section.
-The authors indicate that the two AcrIF9 binding sites are nearly identical. Is the only difference the interaction with Cas8f? This interaction between Cas8f and AcrIF9.1, shown as one hydrogen bond in panel e, should also be described in the text.
-The numbering of all interacting residues and nucleobases should be reported in the text and the figure panels.
3) Page 5: "Based on the similarity of the binding sites for AcrIF1 and AcrIF9, we predicted that AcrIF9 would sterically block target binding." -The authors have previously shown that AcrIF1 interacts with Cas7f lysine residues (K85, K254, K257) that are involved in target DNA binding (http://dx.doi.org/10.1016/j.cell.2017.03.012). Based on Fig 1 panel d, the lysine 85, which is located in the thumb of Cas7f, may also interact with AcrIF9. Testing the binding of AcrIF9 to wild type Csy and to the K85A mutant used in the paper mentioned above, could experimentally prove that AcrIF9 is a steric inhibitor of target DNA binding. 4) Page 6: Fig 2, panel A -In the EMSA with target dsDNA, the strong unbound DNA band in the 20 nM lane is puzzling. Indeed, it indicates that most of the Csy molecules are inhibited, which cannot be the case with a steric inhibitor at sub-stoichiometric concentration. The second EMSA presented in the Supplementary Fig 3 panel a would be more appropriate for the main figure.
-In the EMSA with target and non-target dsDNAs, how do you interpret the DNA migration profiles of the 200 nM lanes? At 200 nM of AcrIF9, and with two AcrIF9 binding sites per Csy molecule, not all Csy molecules are fully occupied by the inhibitor.
-Overall, how do you interpret these EMSA in terms of the AcrIF9 molecular mechanism? If a Csymediated DNA cleavage assay in the presence of a stoichiometric ratio of AcrIF9 shows an inhibition of the Csy activity (which is not the case for AcrIIA11), what would be the benefit of binding dsDNA at high concentrations of AcrIF9? 5) Page 5: "we next determined the structure of the Csy-AcrIF9 complex bound to dsDNA at a nominal resolution of ~6.9 Å (Fig. 2b and Supplementary Fig. 4)." -The procedure used to make the Csy-AcrIF9-dsDNA ternary complex is not described in the Methods. 6) Page 5: "The reconstruction reveals two stretches of non-contiguous helical density that are not accounted for by Csy or AcrIF9, but are consistent with the size and shape of B-form DNA (Fig.  2c)." -The EMSA with target and non-target dsDNAs show that non-specific DNA binding occurs at 400 nM of AcrIF9, which is a concentration of AcrIF9 allowing the occupation of all AcrIF9 binding sites on the Csy complexes. This indicates that two AcrIF9 per Csy complex are required to bind nonspecific dsDNA. Is this compatible with the 3D reconstruction showing two non-contiguous dsDNAs bound to the Csy-AcrIF9 complex? 7) Page 5: "The model reveals that the dsDNA segments are interacting with a positive patch on AcrIF9.
[…] Potential interactions are formed between dsDNA and the Csy complex (Fig. 2 c-d), however the resolution does not allow for reliable model building or structure-guided mutagenesis." -The authors have showed that AcrIF9 alone does not bind to dsDNA. They have also identified potential interactions between the Csy complex and dsDNA in the 3D reconstruction of the ternary Csy-AcrIF9-dsDNA complex. These results indicate that AcrIF9 and Csy surfaces both contribute to the dsDNA binding interfaces. Even though the resolution of the Csy-AcrIF9-dsDNA 3D reconstruction does not allow the modeling of amino acid side chains, the rigid and flexible fitting of the Csy-AcrIF9 model within the Csy-AcrIF9-dsDNA 3D reconstruction may enable the mapping of Csy positively charged residues and the identification of a positive patch at the Csy-AcrIF9 interfaces accommodating the dsDNAs. The authors have also identified some AcrIF9 basic residues that mediate DNA binding using the Csy-AcrIF9 model. 8) Page 6: "Collectively, the structures and biochemistry presented here reveal a two-pronged mechanism for AcrIF9 mediated immune suppression." -Based on the presented data, "suggest" would be more appropriate than "reveal".

Reviewer #1 (Remarks to the Author):
This short paper reports on the mechanism of the anti-CRISPR AcrIF9 (F9), which inhibits the type I-F (Csy) effector. This Acr was one of the first identified but has not been characterised structurally or mechanistically. This is achieved here using a combination of cryo-EM and some basic biochemical analyses. The Cryo structure reveals that AcrIF9 binds to the type I-F complex as two monomers by interacting with the Cas7 subunit in a manner that will block target RNA search and binding -a mechanism remarkably similar to AcrIF1.
As judged by EMSA, AcrIF9 does not bind dsDNA on its own up to a concentration of 3 µM, but does inhibit Csy binding to target DNA when present at 20 nM concentration (ten-fold less than the Csy concentration) -see point 3 below. However, at higher stoichiometries, F9 no longer prevents binding to DNA, which is also difficult to explain.
A cryo structure was also determined for the complex incubated with dsDNA, revealing binding of the DNA to the surface of F9, allowing mapping of the interface as a basic area of F9 with conserved basic residues (see point 5). The authors speculate that F9 functions partly by sequestering Csy on random DNA as well as blocking target DNA binding.
Overall there is some decent cryo-EM data here paired with some preliminary, contradictory and variable quality EMSA data. This leaves any conclusion beyond "F9 probably functions like F1" to be on rather shaky ground.
Response: Thanks for the critical read. We agree that the biochemistry is apparently misleading. However, we see no alternative explanation or contradictions for AcrIF9-mediated inhibition of crRNA-guide hybridization to a DNA target. We attempt to clarify the apparent contradiction in the biochemistry with the two points below and in the revised manuscript.
First, in our EMSA at low concentrations of AcrIF9, we see inhibition of Csy binding to target DNA. AcrIF9 has a similar binding site to what we observed previously for F1, and both Acrs physically block access to specific regions of the crRNA-guide. While the folds for AcrIF1 and AcrIF9 are distinct, it seems clear that they both sterically block crRNA-DNA interactions. Moreover, these results are consistent with those in the recently published PNAS paper by Zhang et al.
Second, at high concentrations of AcrIF9, we see binding of both target and non-target DNA to the Csy-AcrIF9 complex. Unlike F1, F9 triggers a non-sequence specific association with DNA, which restores DNA binding, but formation of the Csy-AcrIF9-dsDNA complex does not rely on base pairing. A structure of the Csy-AcrIF9-dsDNA complex reveals that the dsDNA is bound to a positive patch on AcrIF9. Bound in this fashion the target-DNA will not trigger recruitment of Cas2/3 and will not be degraded, which is consistent with the "cleavage assays" presented in Zhang et al., but the nature of their assay (i.e. detection of cleavage not binding) prevented them determining the mechanistic basis of this result.
Anti-CRISPR-induced non-sequence specific DNA associations have previously been observed for AcrIIA11 and Cas9 (Forsberg et al.), but structural information explaining the interaction is as yet unavailable. We are not suggesting that the mechanism for non-specific DNA association is conserved among different systems, but rather that this phenomenon may prove to be more general than has been previously appreciated.
Specific points: Point 1. AcrIF9 is described as a 5-stranded anti-parallel beta sheet cradling an alpha helix, but no structural comparison is offered. Are there any similar folds in PDB? Can the structure be compared more fully to AcrIF1, given their striking convergence in binding sites? Can the two binding sites be compared, beyond saying they are "similar".
Response: A DALI search revealed many structural homologs to AcrIF9 with significant Zscores. Most of these structural homologs possessed domains with alpha-beta plait or alphabeta roll topologies. These topologies are common superfolds for small domains (<100 residues) found in proteins with a broad array of functions and evolutionary origins (doi:10.1016/j.str.2009.06.015). Therefore, structural homology to AcrIF9 is unlikely to indicate shared functionality and we elected not to include this analysis in the manuscript. AcrIF1 and AcrIF9 have overlapping but not identical binding sites. To illustrate these differences, we have added text to the manuscript and edited Supplementary Fig. 2. Figure 1 and Supplementary Fig. 2 provide a full comparison of the structures and binding sites for AcrIF1 and AcrIF9.
On page 3: "AcrIF9 is a 7.9kDa protein made up of a five stranded, anti-parallel beta sheet cradling an alpha helix (Fig. 1b), a fold notably distinct from AcrIF1 (Supplementary Fig. 2)." Supplementary material page 10: " Supplementary Fig. 2: Comparison of AcrIF9 and AcrIF1. a-b. Schematic of the Csy-AcrIF9 (a) and Csy-AcrIF1 (b) complexes, with subunits colored as in Fig. 1. While AcrIF9 interacts with a single Cas7f, AcrIF1 interacts with residues from two neighboring Cas7f molecules. Intersubunit interactions are denoted by lines. c-d. Model of the Csy-AcrIF9 complex (c) and Csy-AcrIF1/AcrIF2 complex (d), Csy subunits shown as pipes and planks, Acrs shown in surface representation. e-f. The binding site of AcrIF9.2 (e) and AcrIF9.1 (f), residues interacting with the Acr highlighted in dark blue. AcrIF9 is composed of a five-stranded anti-parallel beta sheet, cradling an alpha helix. AcrIF1 is composed of a four-stranded anti-parallel beta sheet, flanked on one side by two alpha helices. The folds of the Acrs are notably different. In order to illustrate the difference between Acr binding sites, AcrIF1.2 is shown transparent in e. and AcrIF9.2 is shown transparent in f. Cas7f subunits shown in surface representation, crRNA and Acrs in cartoon representation. g-h. Detailed view of the interaction interface in the region of the Cas7.6f thumb helix for AcrIF9 (g) and AcrIF1 (h). While the AcrIF9 and AcrIF1 binding sites overlap, the majority of interactions with Csy residues are different, only interactions with S89 and N94 are common to both Acrs. Model shown in cartoon representation, interacting residues shown as sticks, hydrogen bonds are indicated by yellow dashes, hydrophobic interactions by blue dashes." Point 2. The "triplicate" EMSAs shown in figure S3 are strikingly different from one another to a surprising degree. For example, the third gel in S3b is quite fundamentally different from the first gel. In the second and third replicates it looks like there is a significant proportion of shifted material half way up the gel over a wide protein range, suggesting an intermediate species, but the gel in figure 2 doesn't conform to this. These experiments should be repeated to give more confidence in the data and their interpretation. Ideally, an alternative technique such as fluorescence polarization should be considered, as binding or displacement of Csy will give rise to a large anisotropy change using a method that measures true equilibria.

Response:
We thank the reviewer for this important point. The previous version of this paper included an outlier. We have now standardized conditions of the EMSA experiments and have resolved these issues. The intermediate species could be caused by dissociation of dsDNA alone or in combination with a subset of Csy components. However, we do not observe indications of Csy components dissociating in size exclusion chromatography or cryo-electron microscopy, so the latter explanations seem less likely. While we intend to further investigate the exact nature of this species, the conclusions we draw from these EMSA are not altered by the outcome. The data presented here show that AcrIF9 causes the Csy-AcrIF9 complex to bind dsDNA in a nonspecific manner and according to our cryo-EM data, the DNA remains uniformly bound to the complex in solution. Figure 2a suggests that one tenth molar ratio of AcrIF9 almost completely blocks binding of Csy to target DNA. Given that at least 90% of the Csy complex should be free and available to bind DNA (assuming non-cooperative binding of AcrIF9), what is the authors' view of the mechanism here? Notably, in the replicates, this effect is not observed at 20 nM F9 to nearly the same extent, and in the middle gel of S3a there appears to be 50:50 binding to target DNA at a molar ratio of 1:2, which could make more sense.

Point 3.
Point 4. It would help to understand how Csy binds to target and non-target DNA in the absence of F9. Although this has no doubt been published previously it would be helpful to show this data for the DNA species used here. It is important to understand how F9 affects DNA binding for both target and non-target DNA.
Response: As the reviewer suggests, the interaction between Csy and target DNA has previously been described in detail (Rollins et al. 2019, Guo et al. 2017, Rollins et al. 2015. We have added references to these papers. Additionally, we show how Csy interacts with target and non-target DNA oligonucleotides in absence of AcrIF9 by EMSA in the first lane of each gel (Figure 2a and Supplemental Figure 3a and 3b). We show that in the absence of AcrIF9, Csy forms a complex with target DNA, while non-target DNA does not associate with Csy.
On page 6: "This result suggests that formation of the ternary complex (i.e., Csy-AcrIF9-dsDNA) does not rely on base-pairing interactions, and is fundamentally different than target DNA hybridization to the crRNA guide (Rollins et al. 2015, Guo et al. 2017, Rollins et al. 2019." Point 5. A binding interface for F9 on dsDNA is observed that does not involve the Csy complex. The text states that "we attempted to make structure-guided mutations of these residues, but the F9 mutants failed to bind the Csy complex". Leaving aside the observation that there is no data in the paper to support this (were all 5 made? Did they purify as for the wildtype protein?), this is another puzzling observation as the binding interface between F9 and Csy has not been modified.

Response:
We designed and purified three mutants (i.e., AcrIF9 K31A,R32A,K36A , AcrIF9 K31E,R32E and AcrIF9 K31Q,R32Q ). These mutants exhibited altered size exclusion profiles and defects in binding to the Csy complex in EMSA experiments. The following text has been added to page 6 for clarification: "We attempted to make structure-guided mutations of these residues (AcrIF9 K31A,R32A,K36A , AcrIF9 K31E,R32E and AcrIF9 K31Q,R32Q ). While the mutants express and purify similar to wild-type AcrIF9, the size exclusion profiles are distinct. In addition, the mutants are defective for blocking crRNA-guided interactions with DNA targets (Supplementary Fig 6). Collectively, these observations suggest the mutations result in a folding defect that reduces the affinity for Csy."

Reviewer #2 (Remarks to the Author):
Hirschi et al. present the cryo-EM structures of Csy-AcrIF9 complex and Csy-AcrIF9-dsDNA complex at ~3.9 Å and 6.9 Å respectively. In combination with biochemical data, the authors show AcrIF9 not only sterically blocks the binding of target DNA to crRNA, but also promotes nonspecific recruitment of dsDNA, potentially sequestering the complex from target DNA. Even without the high-resolution structure of Csy-AcrIF9-dsDNA complex, the biochemical data and the current density map support their conclusions. This paper significantly broadens the current view of the mechanisms for anti-CRISPR proteins mediated immune suppression, and will be of interest in the community given that CRISPR-Cas immunity is of broad interest.
During the revision session, the structure of Csy-AcrF9 complex has been described in a PNAS paper (PNAS first published March 13, 2020 https://doi.org/10.1073/pnas.1922638117). With 2.57 Å high-resolution Csy-AcrF9 complex structure, the authors of the PNAS paper proposed binding of AcrF9 to the Csy complex prevents target DNA binding to crRNA, similar to how AcrF1 works, also in line with part of Hirschi et al. 's conclusions. The author may briefly discuss the results in the PNAS paper. Even the high-resolution structure of Csy-AcrF9 complex recently has been reported, as far as I see, the non-sequence specific dsDNA bound to Csy-AcrF9 complex reported in Hirschi et al. paper significantly advances the understanding of how anti-CRISPR proteins work, I still support this paper to be published.