Transcriptional antitermination integrates the expression of loci of diverse phage origin in the chimeric Bartonella Gene Transfer Agent BaGTA

Reviewer #1 (Public review):

Summary:

Gene transfer agent (GTA) from Bartonella is a fascinating chimeric GTA that evolved from the domestication of two phages. Not much is known about how the expression of the BaGTA is regulated. In this manuscript, Korotaev et al noted the structural similarity between BrrG (a protein encoded by the ror locus of BaGTA) to a well-known transcriptional anti-termination factor, 21Q, from phage P21. This sparked the investigation into the possibility that BaGTA cluster is also regulated by anti-termination. Using a suite of cell biology, genetics, and genome-wide techniques (ChIP-seq), Korotaev et al convincingly showed that this is most likely the case. The findings offer the first insight into the regulation of GTA cluster (and GTA-mediated gene transfer) particularly in this pathogen Bartonella. Note that anti-termination is a well-known/studied mechanism of transcriptional control. Anti-termination is a very common mechanism for gene expression control of prophages, phages, bacterial gene clusters, and other GTAs, so in this sense, the impact of the findings in this study here is limited to Bartonella.

Strengths:

Convincing results that overall support the main claim of the manuscript.

Weaknesses:

A few important controls are missing.

Reviewer #2 (Public review):

Summary:

In this study, the authors identified and characterized a regulatory mechanism based on transcriptional anti-termination that connects the two gene clusters, capsid and run-off replication (ROR) locus, of the bipartite Bartonella gene transfer agent (GTA). Among genes essential for GTA functionality identified in a previous transposon sequencing project, they found a potential antiterminatior of phage origin within the ROR locus. They employed fluorescence reporter and gene transfer assays of overexpression and knockout strains in combination with ChiPSeq and promoter-fusions to convincingly show that this protein indeed acts as an antiterminator counteracting attenuation of the capsid gene cluster expression.

Impact on the field:

The results provide valuable insights into the evolution of the chimeric BaGTA, a unique example of phage co-domestication by bacteria. A similar system found in the other broadly studied Rhodobacterales/Caulobacterales GTA family suggests that antitermination could be a general mechanism for GTA control.

Strengths:

Results of the selected and carefully designed experiments support the main conclusions.

Weaknesses:

It remains open why overexpression of the antiterminator does not increase the gene transfer frequency.

Author response:

Reviewer 1:

(1) Provide Rsmd and DALI scores to show how similar the AlphaFold-predicted structures of BrrG are to other anti-termination factors. This should be done for Fig1B and also for Suppl. Fig 1 to support the claim that BrrG, GafA, GafZ, Q21 share structural features.

In the revised manuscript we will provide Rsmd and DALI scores.

(2) Throughout the manuscript, flow cytometry data of gfp expression was used and shown as single replicate. Korotaev et al wrote in the legends that error bars are shown (that is not true for e.g. Figs. 3, 4, and 5). It is difficult for reviewers/readers to gauge how reliable are their experiments.

As stated in the manuscript all flow cytometry data were performed in triplicate. In the revised manuscript we will include the two replicates not presented in the main figures as supplementary information.

(3) I am unsure how ChIP-seq in Fig. 2A was performed (with anti-FLAG or anti-HA antibodies? I cannot tell from the Materials & Methods). More importantly, I did not see the control for this ChIP-seq experiment. If a FLAG-tagged BrrG was used for ChIP-seq, then a WT non-tagged version should be used as a negative control (not sequencing INPUT DNA), this is especially important for anti-terminator that can co-travel with RNA polymerase. Please also report the number of replicates for ChIP-seq experiments.

Fig. 2A presents a coverage plot from the ChIP-Seq of ∆brrG +pTet:brrG-3xFLAG (N’). A replicate of this N-terminally tagged construct will be added as supplementary data in the revised version. As anticipated by the referee, we had used ∆brrG +pTet:brrG (untagged) as control.

(4) Korotaev et al mentioned that BrrG binds to DNA (as well as to RNA polymerase). With the availability of existing ChIP-seq data, the authors should be able to locate the DNA-binding element of BrrG, this additional information will be useful to the community.

We will mine the ChIP-Seq data to define the BrrG binding site as closely as possible and include the analysis in the revised version of the manuscript.

(5) Mutational experiments to break the potential hairpin structure are required to strengthen the claim that this putative hairpin is the potential transcriptional terminator.

We did not claim that the identified hairpin is a terminator but rather suggested it as a candidate terminator. We agree with the referee that the proposed experiment would be necessary to definitively prove its terminator function. However, our primary aim was to demonstrate that BrrG acts as a processive terminator, which we have shown by replacing the putative terminator with a well-characterized synthetic terminator that BrrG successfully overcame. Therefore, we prefer not to conduct the proposed experiment and will instead further tone down our conclusions regarding the putative terminator function of the identified hairpin structure.

Reviewer 2:

(1) The authors wrote "GTAs are not self-transmitting because the DNA packaging capacity of a GTA particle is too small to package the entire gene cluster encoding it" (page 3). I thought that at least the Bartonella capsid gene cluster should be self-transmissible within the 14 kb packaged DNA (https://doi.org/10.1371/journal.pgen.1003393, https://doi.org/10.1371/journal.pgen.1000546). This was also concluded by Lang et al (https://doi.org/10.1146/annurev-virology-101416-041624). In this case the presented results would have important implications. As the gene cluster and the anti-terminator required for its expression are separated on the chromosome, it would not be possible to transfer an active GTA gene cluster, although the DNA coding for the genes required for making the packaging agent itself, theoretically fits into a BaGTA particle. Could the authors comment on that? I think it would be helpful to add the sizes of the different gene clusters and the distance between them in Fig. 2A. The ROR amplified region spans 500kb, is the capsid gene cluster within this region?

We thank the reviewer for bringing up this interesting point. The bgt cluster (capsid cluster) is approximately 9.2 kb in size and could feasibly be packaged in its entirety into a GTA particle. In contrast, the ror gene cluster, which encodes the antiterminator BrrG, is approximately 20 kb in size—exceeding the packaging limit of GTA particles—and is separated from the bgt cluster by approximately 35 kb. Consequently, if the bgt cluster is transferred via a GTA particle into a recipient host that does not encode the ror gene cluster, the bgt cluster would not be expressed.

(2) Another side-note regarding the introduction: On page three the authors write: "GTAs encode bacteriophage-like particles and in contrast to phages transfer random pieces of host bacterial DNA". While packaging is not specific, certain biases in the packaging frequency are observed in both studied GTA families. For Bartonella this is ROR. In the two GTA-producing strains D. shibae and C. crescentus origin and terminus of replication are not packaged and certain regions are overrepresented (https://doi.org/10.1093/gbe/evy005, https://doi.org/10.1371/journal.pbio.3001790). Furthermore, D. shibae plasmids are not packaged but chromids are. I think the term "random" does not properly describe these observations. I would suggest using "not specific" instead.

We thank the reviewer for this suggestion and will adjust the working accordingly.

(3) Page 5: Remove "To address this". It is not needed as you already state "To test this hypothesis" in the previous sentence.

We will adjust the working accordingly.

(4) I think the manuscript would greatly benefit from a summary figure to visualize the Q-like antiterminator-dependent regulatory circuit for GTA control and its four components described on pages 15 and 16.

We thank the reviewer for this valuable suggestion and will include a summary figure illustrating the deduced regulatory mechanism in the revised manuscript.

(5) Page 17: It might be worth noting that GafA is highly conserved along GTAs in Rhodobacterales (https://doi.org/10.3389/fmicb.2021.662907) and so is probably regulatory integration into the ctrA network (https://doi.org/10.3389/fmicb.2019.00803). It's an old mechanism. It would be also interesting to know if it is a common feature of the two archetypical GTAs that the regulator is not part of the cluster itself.

We agree with the points raised by the reviewer and will address them in the revised manuscript. Specifically, we will highlight the high conservation of GafA in GTAs across Rhodobacterales and its regulatory integration within the ctrA network. Additionally, we will analyze whether the GafA-like antitermination regulator is typically located outside the regulated gene cluster, as we have demonstrated for BrrG of BaGTA in the Bartonellae.