De novo design of an intercellular signaling toolbox for multi-channel cell–cell communication and biological computation

Intercellular signaling is indispensable for single cells to form complex biological structures, such as biofilms, tissues and organs. The genetic tools available for engineering intercellular signaling, however, are quite limited. Here we exploit the chemical diversity of biological small molecules to de novo design a genetic toolbox for high-performance, multi-channel cell–cell communications and biological computations. By biosynthetic pathway design for signal molecules, rational engineering of sensing promoters and directed evolution of sensing transcription factors, we obtain six cell–cell signaling channels in bacteria with orthogonality far exceeding the conventional quorum sensing systems and successfully transfer some of them into yeast and human cells. For demonstration, they are applied in cell consortia to generate bacterial colony-patterns using up to four signaling channels simultaneously and to implement distributed bio-computation containing seven different strains as basic units. This intercellular signaling toolbox paves the way for engineering complex multicellularity including artificial ecosystems and smart tissues.


Comments
• Minor language corrections throughout; only a small subset of what has to be done is mentioned here. In general it is OK, but lacks the native English form.   Fig 11). o Sequences should be copiable text. o pg20: some explanation should be provided as to how the Sal system was implemented into S. cerevisiae. If it was done analogously to the DAPG--PhlF system, this should be mentioned.
Reviewer #2: Remarks to the Author: The authors have described an approach to develop new communication pathways that can be added to bacterial and eukaryotic cells that will enable molecular control over cells and cell networks. In general, the article "hits the spot" in that there is a tremendous need for advancements in this area and just a few innovative studies that have helped to define workable solutions. The methodologies proposed here are rational; the importance is high and the example methodology is noteworthy. My enthusiasm for this work is predicated on the need to solve difficult issues concerning cell networks that exist in nature, particularly those were there is limited analytical access and the need is great. I am thinking of the GI tract, for example. If we are to get a handle on understanding the influence of the GI microbiome on human physiology, we will need to engineer signaling among networks of cells and the community activities should be actuated in non-invasive ways. Approaches such as those provided by the authors, while to me a logical extension of earlier efforts, will be needed.
The experiments appear to be well organized, executed, and with a few grammatical issues, well described. This reviewer is of the opinion that the overall methodology is robust. I have provided some thoughts below that will help with the presentation, justification, and novelty.
Primary thoughts: The authors have based the work on requirements for (i) universality and (ii) orthogonality. The need for universality seems valid in that the synthesis of signal molecules should be such that exogenous precursors aren't needed, instead typical biosynthetic functions should enable their synthesis. In this way, the signals can be generated from a variety of organisms, in particular, those that will be resident in the niche community of interest. The orthogonality argument has been made by many and is reasonable here. In order to meet these requirements, constraints were imposed, however, that limit the potential application and appeal. These should be either discussed in more detail or otherwise validated.
1. One constraint is that the signal molecules freely diffuse across cell membranes. This dramatically reduces the number of viable signal molecules and also limits the mode by which they act in the host cells (i.e., binding to transcription factors). Was there are robust examination of this limitation? Some additional discussion is warranted. For example, the types of pathways that would be available for signal molecule generation is likely limited. Also, the gene circuits that can be controlled using aTF's will be limited. These constraints might have been buried, intentionally or unintentionally, in the experimental results. For example, the authors validated the orthogonality requirement by demonstrating minimal cross talk -they used the aTF's that were an intrinsic part of the study. There may have been other metabolic consequences that went undetected, but could have ramifications relative to end use. Some discussion is warranted here relative to the number of signal molecules, the types of pathways available for their synthesis and the limiting consequences of using aTFs.
2. The authors have reduced the experimental range of the study by limiting the number of enzymatic steps in the signal synthesis pathway. Specifically, they chose five enzymes as a maximum and justified this by "reducing cellular resource taxing". It would be important to show how this was conceived. Currently, this seems to be an arbitrary decision. Some enzymes for example, may represent a huge burden to the cells and could be in a two-step pathway, while a 7 step pathway might have minimal consequences based on flux through the pathway and the levels of enzymes needed. Also, they have not provided any data associated with this constraint. If these aren't shown, then there is no need to suggest that it is fundamental to the approach. Rather, it was a convenience.
3. Cross-kingdom universality. Why show this in HEK cells? Is there an intent to develop methodologies for urinary tract infections? It would seem to be more important to show Cross kingdom utility in a relevant cell line, such as an intestinal epithelial line. Provide some additional rationale here.
4. The logic gate circuits. It would be helpful if the authors provide an example where these complex circuits would be useful. The premise of the paper is on developing the molecules and pathways. What computations are performed using these strains in logic gate circuits. This section of the paper is limited to a paragraph and its utility could be strengthened.
Minor concerns (no particular order of importance): L 49. Short-medium-long-range…perhaps better terms are autocrine, paracrine, endocrine… L 54. This refers to "molecular communication" might cite this literature, especially as it relates to QS. L 91. Presumes that the signaling molecules must pass through cell membranes and find their way to aTFs. This is a considerable challenge, one met by the authors, and also a severe limitation on the approach in that it presumes that the receiving cells have no metabolic activity for the signal molecule (as noted above).
L 104. Some of them…state what it was…three?
L 126. Not sure I agree with the need to limit to 5 enzymes to reduce cellular taxing (as above). The load of the signal generation itself could be distributed among cells. Moreover, if one enzyme represents more of a burden than others, this would preclude this method from working. This is a significant limitation, seemingly arbitrary. Perhaps the method should include an in silico method for evaluating this criterion?
L 150. Two commas in a row. L 190-220. Co-Cultures. How did you stimulate the synthesis of the signal molecule. It would seem to be important to have a working system that opens a communication channel by providing a stimulus. In the co-culture, the stimulus is important, then the time over which the sender cells respond by making the signal molecule, the transport of the signal molecule to the receiver cells (autocrine, paracrine, endocrine) and the subsequent transport into a response of the receiver cells.
L251-L270. While this is a straightforward set of experiments and seems to have been expertly executed, what is the point? Also, did you screen using sequence searches to make sure that there aren't any aTFs in the genomes of the supposed non-responding receivers?
L 289. Why did you transfer "some" and not all 10?
L 299. The transfer to HEK cells is very nice. It should be presented with more data and its own figure.
L308-L350. Interesting work. This demonstrates the orthogonality of the signals and the sender and receiver cells. It is done on plates and depends on the diffusion of the signal molecules. I wonder what the application of this might ever be?

Figures. 2. It is important to see the background levels. Suggest in supplemental providing the FACS data.
Specifically, one of the most important issues in studies like this has to do with the diffusion and perception of the signal molecule. The fold induction, if performed using a fluorimeter, would give you the total fluorescence. But, the number of cells might be affected as well as, or instead of, the expression rate of the fluorophore in the receiving cell. Total fluorescence is the sum of both and these are important. Please specify what was used here. If the number of cells that fluoresce is the primary measure, then the transport of the signal molecule might have been a variable (that was not accounted for). 3. I'm not sure why the YFP is normalized by RFP. We need amplification from zero not from a different promoter.

Response:
We have followed the suggestion and measured the growth curves of the sender and receiver strains. All the experimental results of their growth curves are provided in Supplementary Figure 22 and 23, and additional discussion are added to Supplementary Information (Line 367-393). Briefly, none of the ten receiver parts exerted discernible burden on the host cell growth, while some of the sender parts resulted in severe growth defect for their host cells. Especially, for the Sal and DAPG sender systems, the growth of their host cells was dramatically hindered by the burden or toxicity of the biosynthetic enzymes and signal molecules involved.
Question 3: Detailed presentation of all genetic sequences and constructs. Minor, but potentially useful: include characterization of signaling modules in different growth media in E. coli (M9, LB, 2xYT).

Response:
We added all the sequences of the sender and the receiver parts in the supplementary files. As for robustness to growth conditions, we indeed evaluated the differences between LB and M9 media for several signaling systems. The results below show similar induction performances of four systems in M9 and LB, respectively, indicating that our characterization of the sensitivity and the induction fold are not significantly affected by the growth media.

Question 4: Expand referencing in Discussion.
Response: More references and expanded discussion have been added to the manuscript. Response: We are grateful for such helpful comments, and have fixed language issues throughout the text. We have also sent the manuscript to a native English-speaking editing service for polishing. Question 8: Include some measure of burden that the sensors presented in this study have on host cells; this can be growth of cells with fully active signaling and sensing modules; alternatively, it could be the expression levels of a constitutive genomic reporter in cells with fully active signaling and sensing modules.

Response:
In order to simplify the added experiments, we measured the growth curves of the host cells for all sender and receiver parts. A comprehensive analysis of cell burden has been provided in Supplementary Information, with the growth curves of senders and receivers shown in Supplementary Figures 22 and 23.
Question 9: pg8, line 298-299; not clear why you are referring to Supplementary Figure 11 here. Actually, there seems to be an error in the numbering of Supplementary Figures; please correct this.

Response:
We are sorry for our carelessness in editing and have corrected it and the following Figures.
Question 10: pg8, 'Cross-kingdom universality of the designed channels'; I am interested to see whether the same approach of promoter and strain design would work for the other communication channels that were presented (optional).

Response:
We thank the reviewer for the insightful question. Based on our design rules, we expected all the signal channels to be functional in all species. However, we did notice some sender parts which include multiple enzymes tended to cause burden to the host cells or were hard to be constructed in mammalian cells. In comparison, the signaling systems with simple and burden-free sender parts would be easily transferred. We have thus changed the section title to "Cross-kingdom capability of the designed channels". A copiable sequence for all genetic tools that were used must be easily available.

Response:
We have added all the sequences for ten sender and receiver parts of the signaling systems in Supplementary Information. Supplementary info: some explanation should be provided as to how the Sal system was implemented into S. cerevisiae. If it was done analogously to the DAPG-PhlF system, this should be mentioned.
Response: More detailed information of implementing Sal into S. cerevisiae system have been added to Supplementary Information (Line 252-257).
Questions/comments from Reviewer #2 (related comments may be grouped together): Question 1: The authors have described an approach to develop new communication pathways that can be added to bacterial and eukaryotic cells that will enable molecular control over cells and cell networks. In general, the article "hits the spot" in that there is a tremendous need for advancements in this area and just a few innovative studies that have helped to define workable solutions. The methodologies proposed here are rational; the importance is high and the example methodology is noteworthy. My enthusiasm for this work is predicated on the need to solve difficult issues concerning cell networks that exist in nature, particularly those were there is limited analytical access and the need is great. I am thinking of the GI tract, for example. If we are to get a handle on understanding the influence of the GI microbiome on human physiology, we will need to engineer signaling among networks of cells and the community activities should be actuated in non-invasive ways. Approaches such as those provided by the authors, while to me a logical extension of earlier efforts, will be needed. The experiments appear to be well organized, executed, and with a few grammatical issues, well described. This reviewer is of the opinion that the overall methodology is robust. I have provided some thoughts below that will help with the presentation, justification, and novelty.