Drug-dependent growth curve reshaping reveals mechanisms of antifungal resistance in Saccharomyces cerevisiae

Microbial drug resistance is an emerging global challenge. Current drug resistance assays tend to be simplistic, ignoring complexities of resistance manifestations and mechanisms, such as multicellularity. Here, we characterize multicellular and molecular sources of drug resistance upon deleting the AMN1 gene responsible for clumping multicellularity in a budding yeast strain, causing it to become unicellular. Computational analysis of growth curve changes upon drug treatment indicates that the unicellular strain is more sensitive to four common antifungals. Quantitative models uncover entwined multicellular and molecular processes underlying these differences in sensitivity and suggest AMN1 as an antifungal target in clumping pathogenic yeasts. Similar experimental and mathematical modeling pipelines could reveal multicellular and molecular drug resistance mechanisms, leading to more effective treatments against various microbial infections and possibly even cancers.

The authors first obtained a robustly unicellular strain by a single gene knockout (AMN1) of an otherwise multicellular clumping baker's yeast strain TBR1. Three different antifungals and hydrogen peroxide were used as stressors, and AMN1-deletion was generally shown to sensitize the cells against all of them in unique ways in the sense that no universal trend/effect was suggested by the work. Although their data seem to favor that AMN1-deletion exerts its sensitizing effect mostly through reducing clump formation, the work still cannot clearly rule out other potential roles of AMN1 that are independent of clump formation.
In terms of the technical details, they applied piece-wise linear fits to growth curves on the semilog scale under various stressors which helped them to break down the whole growth dynamics into distinct meaningful parts. In return, they reported some quantitative measurements on how different drugs and doses reshape growth dynamics with or without a functional AMN1 gene. Finally, their simple ODE model helped them to argue on how different drugs exert their effects on cell growth dynamics under stress such as by modulating drug influx, intracellular drug degradation, etc. In some control experiments, they also employed a standard non-clumping lab strain BY4742 and its AMN1-deletion. Their data suggested that AMN1 might be sensitizing this strain to the drugs in some cases. Hence, they argue that when clumping (which requires AMN1 in the TBR1 strain) is possible, it can counterbalance the otherwise sensitizing effects of AMN1.
Specific comments: 1. A major takeaway I got from the paper is that AMN1-deletion mediated clumping could contribute to increased drug tolerance/resistance in a manner that can overcome the pleiotropic effect of the gene. This message makes intuitive sense can could suggest a potential benefit of multicellularity. However, I feel this message was somewhat masked by the presentation of data, which sometimes appears a bit complicated (see below). By and large, the authors might consider elaborating on this point and present the modeling analysis in a way to better illustrate this point.
The authors might consider citing work on bacteria that exhibit collective antibiotic tolerance, though through different mechanisms (e.g Meredith et al, Science Advances 2018).
2. My impression is that some data analysis, while thorough, might be too complicated than needed. For example, the authors spent a significant amount of time talking about using crescement and piece-wise growth curve fitting to quantify growth under different conditions. I honestly found this to be somewhat an overkill. The authors seemed be using "crescement" and AUC interchangeably -if so, using "AUC" alone seems sufficient and its meaning is well understood.
3. Related to point 1, I thought it could be emphasized more in the text and modeling analysis. It is not clear how the effect of clumping is incorporated in the ODE model. As it stands, the modeling analysis gives the impression of data fitting and is somewhat disconnected from the central message (see point 1).
4. Aspects of the ODE model formulation are also somewhat unclear. For example, the authors assume growth of all their strains exhibit an Allee effect. What's the major basis for this assumption and how critical is this assumption for the central points? It appears that one rationale is the including the effect allows better fit -if this is the major rationale, it's good to make it clearer. Does the magnitude of the Allee term depend on the clumping capability of the strains tested?
Also, there appears to be some discrepancy in how it's modeled: in Equation 1, the densitydependent growth rate is modeled as (N+C), whereas in Section 2.3 (Supp), it's modeled as (N-C). Is this intended or the result of a typo? 5. In equation 6, why is drug uptake not considered? I would expect a term proportional to the first term in Equation 4, but with an opposite sign.
6. In Fig. 6E, it appears that 'f' (the drug influx rate) either does not change or gets smaller with increasing H2O2 or AmB and becomes larger with the increases in the other two drugs in the AMN1 deletion mutant. What should we have expected a priori? Was there prior evidence that clumping affects penetration of these different stressors differently? 7. 'Growth curve reshaping' is a vague term and repeatedly used without a clear introduction/definition of it.
8. Line 113: It would be helpful to provide at least a brief definition of the 'normal settings' in the main text.
9. Authors mention a base level growth even with 0% glucose in the experiments and attribute that to intracellularly stored sugar. That makes me wonder what the authors found for the Allee threshold (C) for 0% glucose.
10. For the equations along Lines 795 and 796, why doesn't the toxicity mediated killing of cells contribute to nutrients? 11. What would have they seen if they had tested higher concentrations of FLC? My sense is that the tested concentrations might be too low, although they are somewhat recommended by literature (as claimed by the authors).
12. FLC starts showing its "growth reshaping" effects only once after cells are resuspended in the same type of media before stationary phase. That resuspension also means reducing the cell density. So, might protection from FLC have a cell density-dependent component too in addition to an initial delay? Would it be possible to lower the initial cell density instead of resuspension in order to give cells more time under the FLC treatments tested? A related question, I see in Fig. S11 that resuspension was done after 10 hours and into fresh media containing the original concentration of the drug. Do the authors have any idea about how much drug degradation is going on during those 10 hours? 15. Figure 1A-C: it would be nice to have scale bars. 16. Figure 1 F and G: I could not figure out which data are for glucose and which data are for galactose. Probably, galactose data is only in Supp, and this caption needs to be corrected.
17. Figure 2A: legends for the strains are missing 18. Fig. 2F: Breakpoint labels are not displayed in the figure. Also, why was not a death phase considered in labeling the linear pieces of the log (OD600) curves, although it clearly appears in some of the conditions in Fig. 2A-D? 19. Fig. S15-18: In some many cases (but not all) the results show that a tiny and temporary increase in D (intracellular drug concentration) succeeded by a multiple fold higher and much longer lasting increase in toxicity. I wonder how justifiable this is. 20. Figs. S19: Please specify which data belong to which stressor. I assumed it is in the order of H2O2, AmB, CASP, and FLC from left to right.

Reviewer #2 (Remarks to the Author):
This is a very thorough and insightful investigation of how a clumping gene (AMN1) impacts adaptation to and growth in the presence of different stresses. A few suggestions would improve the quality of this manuscript: -the authors should discuss the fact that the 2 ergosterol-targeting antifungals have different effects on fungal cells (one is static, other is cidal) -the authors should verify that the 2 TBR1 strains investigated don't show different clumping/clustering patters upon exposure of the different stress agents -a quantification of clumping of the the 2 BY4742 strains (similar to that shown in Fig1A-E) should be included to strengthen the author's conclusions regarding the roles of AMN1 in different strain backgrounds Editor: 1. Expand the Methods to include more detail about how the model was constructed, and potentially include the additional variables suggested by Referee #1.
Response: As recommended, we included the variables suggested by Referee #1 that were reasonable. We replaced the corresponding figures (Figure 6, Figures S15-S21) with the updated model results, and updated our description of model construction with additional details in the corresponding Results and Methods sections, as well as in the Supporting Information.
Response: We have addressed this in the revised manuscript (Figure 1) as indicated in the response to comments 2 and 3 by Referee #2.
3. If feasible, we would strongly encourage you to test higher concentrations of FLC or whether there may be a density-dependent effect of FLC-resistance (in response to Referee #1's points 11-12).
Response: We appreciate this valuable recommendation, which we have addressed in the revised manuscript, as indicated in our response to comment 11 by Referee #1 (for higher FLC concentrations) and in our response to comment 12 by Referee #1 (for cell density-dependence of FLC effect on cells). Furthermore, we would encourage you to include additional example applications or case studies involving this mathematical model, to highlight its value as a resource for the research community.
Response: This was a useful recommendation. This mathematical model is original to this project and has not been used in other contexts yet, so we have mentioned some potential applications of the model in the Discussion section: "Similar models could be applicable to elucidate the mechanisms of stress-dependent growth curve reshaping for other microbes or even cancer cells." Reviewer #1, Specific comments: 1. A major takeaway I got from the paper is that AMN1-deletion mediated clumping could contribute to increased drug tolerance/resistance in a manner that can overcome the pleiotropic effect of the gene. This message makes intuitive sense and could suggest a potential benefit of multicellularity. However, I feel this message was somewhat masked by the presentation of data, which sometimes appears a bit complicated (see below). By and large, the authors might consider elaborating on this point and present the modeling analysis in a way to better illustrate this point.
Response: Thank you for pointing this out. We wish we could have a clearer message on how AMN1 affects drug resistance through clumping and pleiotropic effects. If AMN1 deletion would improve resistance of the unicellular lab strain BY4742 to all drugs, we could indeed argue generally that loss of clumping in TBR1 reverses this effect for all drugs. However, AMN1 deletion increases sensitivity to Caspofungin in both the lab strain and TBR1 strain, so AMN1's pleiotropic effect for Caspofungin is not reversed by the loss of clumping. Moreover, the AMN1 sequences and the genetic backgrounds of the two strains are not identical, which further complicates the picture. Despite these complexities, we tried to simplify the message while also expanding the description of parameter selection, to better reflect the dual effects of AMN1.
The authors might consider citing work on bacteria that exhibit collective antibiotic tolerance, though through different mechanisms (e.g Meredith et al, Science Advances 2018).
Response: Thank you for mentioning this interesting and relevant paper. We are citing it in the revised version.
2. My impression is that some data analysis, while thorough, might be too complicated than needed. For example, the authors spent a significant amount of time talking about using crescement and piece-wise growth curve fitting to quantify growth under different conditions. I honestly found this to be somewhat an overkill. The authors seemed be using "crescement" and AUC interchangeablyif so, using "AUC" alone seems sufficient and its meaning is well understood.
Response: Thank you for suggesting this simplification. We have replaced "crescement" with AUC throughout the manuscript.
3. Related to point 1, I thought it could be emphasized more in the text and modeling analysis. It is not clear how the effect of clumping is incorporated in the ODE model. As it stands, the modeling analysis gives the impression of data fitting and is somewhat disconnected from the central message (see point 1).
Response: Thank you for pointing this out. We did not include clumping explicitly in the model. However, from all parameters of the model, the rate of drug diffusion into/out of cells (parameter f) is most likely to depend on clumping. Indeed, clumping should reduce the average effective cellular surface area exposed to the extracellular environment, thereby lowering drug influx. Therefore, upon fitting the same model to both clumpy and unicellular data, we tried inferring the effect of clumping from changes in the parameter f versus other parameters. The fits indicated that (i) the drug in/outflux rate constant f was generally lower in the clumpy TBR1 versus the unicellular TBR1Δa strain, whereas the inferred f values were similar in the unicellular pair of BY4742 versus BY4742Δa lab strains; (ii) the threshold q that intracellular drug concentration needs to reach for growth inhibition (i.e., tolerance to the drugs) tended to be higher for Δa strains; iii) the detox rate p tended to be higher in the AMN1-containing versus AMN1-deleted strains. Other parameters changed in ways that were less consistent. Overall, this suggests that AMN1 affects resistance in a variety of ways, and clumping is only one of the multiple resistance mechanisms that further studies will need to disentangle. We added the above information into the revised Results section.
4. Aspects of the ODE model formulation are also somewhat unclear. For example, the authors assume growth of all their strains exhibit an Allee effect. What's the major basis for this assumption and how critical is this assumption for the central points? It appears that one rationale is the including the effect allows better fit -if this is the major rationale, it's good to make it clearer. Does the magnitude of the Allee term depend on the clumping capability of the strains tested?
Response: Thank you for raising this point. There is recently published evidence for an Allee effect in budding yeast, see Saurabh R. Gandhi, Kirill S. Korolev, and Jeff Gore, PNAS 116 (47) 23582-23587 (2019). Also, as the Reviewer surmised, including the Allee effect allowed better fits, especially for growth curves without drugs. The Allee effect indeed tends to be more apparent in the clumping strain, with a potential contribution from gravitational settling in the plate reader, but is noticeable even in more intensely shaken cultures. Nonetheless, the Allee effect is not the focus of this manuscript, so we did not engage in its detailed investigation. We have now clarified this in the main text and updated the Supporting Information accordingly.
Also, there appears to be some discrepancy in how it's modeled: in Equation 1, the densitydependent growth rate is modeled as (N+C), whereas in Section 2.3 (Supp), it's modeled as (N-C). Is this intended or the result of a typo?
Response: Thank you for catching this! It is indeed unintended (it is a typo). The fits can be done both ways, and they result in Allee parameters of opposite signs, as expected. We have corrected the equations to make them consistent. 5. In equation 6, why is drug uptake not considered? I would expect a term proportional to the first term in Equation 4, but with an opposite sign.
Response: This is an excellent point, which we appreciate very much. We omitted drug uptake from Equation 6 because we considered it negligible compared to the total amount of drug and its degradation within the whole extracellular growth medium, which we expected to have a much larger volume than the cell interiors. This approximation does indeed apply initially, when cell numbers are low, but probably not later, at high cell densities. We have redone the fits with the drug uptake included in Equation 6 (see Figure 6). These details are now reflected in the Supporting Information, Section 2.4. 6. In Fig. 6E, it appears that 'f' (the drug influx rate) either does not change or gets smaller with increasing H2O2 or AmB and becomes larger with the increases in the other two drugs in the AMN1 deletion mutant. What should we have expected a priori?
Response: Thank you for the comment. These trends are no longer apparent in the parameter fits based on the updated model ( Figure 6).
Was there prior evidence that clumping affects penetration of these different stressors differently?
Response: There was no prior evidence on penetration of these drugs into yeast clumps. Yeast cells in the interior of multicellular flocs were reported to have slower stressful chemical influx by . Therefore, we assumed that clumping might similarly reduce drug influx, pertly by reducing the effective cell surface area exposed to the extracellular environment. 7. 'Growth curve reshaping' is a vague term and repeatedly used without a clear introduction/definition of it.
Response: Thank you very much for the suggestion. We defined growth curve reshaping as a significant drop in the AUC versus stress-free conditions, due to the drugs "altering the number or slope of growth phases", in the "Loss of AMN1 impairs TBR1 growth in stressful conditions" section of the Results. 8. Line 113: It would be helpful to provide at least a brief definition of the 'normal settings' in the main text.
Response: Thank you for this comment. We have replaced "normal settings" with "without stress" in the revised manuscript. We use this term now to distinguish stress-free conditions from other growth conditions where stressors were added. 9. Authors mention a base level growth even with 0% glucose in the experiments and attribute that to intracellularly stored sugar. That makes me wonder what the authors found for the Allee threshold (C) for 0% glucose.
Response: We have addressed this comment by growing the cells in YPD and then resuspending them in 0% glucose (YP) medium. We still observed a weak Allee effect by growth curve fitting, see the revised Supporting Information, section 2.2. The cells managed to maintain growth for at least 24 hours after being transferred into YP.
10. For the equations along Lines 795 and 796, why doesn't the toxicity mediated killing of cells contribute to nutrients?