A diversity-oriented rhodamine library for wide-spectrum bactericidal agents with low inducible resistance against resistant pathogens

Antimicrobial resistance is a public health emergency and warrants coordinated global efforts. Challenge is that no alternative molecular platform has been identified for discovery of abundant antimicrobial hit compounds. Xanthene libraries have been screened for bioactive compounds. However, the potentially accessible chemistry space of xanthene dyes is limited by the existing xanthene synthesis. Herein we report a mild one-step synthesis, which permits late-stage introduction of a xanthene moiety onto i.e. natural products, pharmaceuticals, and bioactive compounds and construction of a focused library of rhodamine dyes exhibiting facile functional, topographical and stereochemical diversity. In vitro screening yields 37 analogs with mid-to-high bactericidal activity against WHO priority drug-resistant pathogens. These findings suggest that synthetic dye libraries exhibiting high structural diversity is a feasible chemical space combating antibacterial resistance, to complement the natural sources.

• The finding of some antibacterial activity with a subset of the analogs provides the best rationale for publication of this manuscript in a cross-discipline journal. The initial cutoff of 8 mg/L and 16 mg/L for MRSA and A. baumannii, respectively is a modest cutoff for potential starting points to identify lead molecules. The authors chose to further evaluate RD53 based on claims of best activity, but should explain their rationale for not exploring RD22 which exhibited the same initial profile as RD53. Having 2 structurally different exemplars from their library would have strengthened their claim that this chemistry platform could be useful for the identification of novel leads.
• The A. baumannii staining in Figure 7 (panel C) is at a resolution that is difficult to confirm the authors conclusions. After zooming in on the figure it appears that some of the cells in Figure 7C have intracellular staining whereas other at the top of the frame do not. Given the difference in shape between the species, and the thicker cell membrane of MRSA, the claim on lines 223-224 about the location of the target needs to be substantiated further. Either quantitative measurements of a statistically relevant number of individual cells, use of mutant strains, or even hybridization of non-denatured immobilized proteins after separation of cellular fractions with RD53 should be explored.
• The activity of RD53 against penicillin non-susceptible Streptococcus pneumoniae should be tested and reported -these are on both the WHO and CDC pathogen lists.
• The activity of RD53 against eukaryotic cells in a cytotoxicity assay should be provided to ensure that the antibacterial activity is not positively correlated with general cytotoxicity (and also erythrocyte assays to confirm absence of membrane disruption).
• Are the graphs in Figure 9 (and supplementary figure S132) simply the concentration of RD53 that supports growth upon passage? Individual colonies should be purified from each passage and characterized for absolute fold change with respect to RD53 (and control drugs). Multiple mutational lineages can co-exist and propagate in these types of population resistance experiments and the cultures should be purified to accurately understand independent mutations.
• The resistant mutants that emerged upon serial passage with A. baumannii and E. coli need to be characterized ( Figure 9C and S132A). What is their phenotypic profile with respect to susceptibility to other antibacterial drugs? Furthermore, the genome of these mutants should be sequenced to identify the source of the mutation that reduces the susceptibility as this may also provide significant insight to the identity of the target.
• Given the speed of resistance attainment in A. baumannii, spontaneous frequency of resistance experiments should be performed in this species, as well as the others, to better characterize the resistance risk.
• The authors claim that that RD53 does not acquire resistance in MRSA and use vancomycin as a surrogate (line274-275). This is not an appropriate comparison as the experimental conditions used do not enable the horizontal transfer of the van cluster that provides this resistance.
• Line 294 claims that RD53 has favorable properties for drug development. To make this claim the authors needs to provide data on some other characteristics that would be important in this regard, such as metabolic and chemical stability, plasma protein binding and solubility.
• Lines 301-307 -although the examples provided on antibacterial drugs coming from small focused libraries is tantalizing, there are just as many that require significant chemistry programs to optimize. The fact is true that RD53 could progress and coming from a small library is not an exclusionary factor, this claim would be significantly more compelling if additional data on drug-like properties are generated and presented as suggested.
Minor grammatical suggestions: Throughout -the tense should be standardized appropriately to be consistent. Line 32-33 -reword for clarity Line 35-36 -Change to something like "The genes encoding resistance to the natural products that are present within the original organisms can then be horizontally transferred to pathogenic microbes enabling resistance to emerge" Line 36 -adjuvants should be better defined. The submitted manuscript reports a Diversity-Oriented Rhodamine Library that applies to find new motif for bactericidal agents. Although there are interesting data especially for the diverse rhodamine library construction, the flow of the work does not support the logic of rhodamine as the key component for bactericidal activity. The main body of the work is construction of rhodamine library, whose diversity exceeds the previously reported library. This is an excellent achievement and very solid with spectral and computational analysis. After that, it is not clear why RD53 was picked from the library and examined for detailed bactericidal activity. There were some other similar level primary hits, such as RD22. The authors just claimed that RD53 was the best compound 'because of its substituent, i.e. adamantyl moiety.' Yes, it is true that adamatyl moiety has been known to have antibacterial effect. Then, what is the role of rhodamine part? Isn't it reasonable to make adamatyl library rather than rhodamine library? Other than the bactericidal effect, the authors did not suggest other applicable possibilities of the library. Therefore, an additional experiment either for adamantyl library or other application of rhodamine library to keep the current claim & title of the paper.
Minor comments: 1. If authors want to show novel bactericidal activity from the library member, the mechanism of the hit compounds should be added. If the activity is from the known motif, the comparison data with known drug should be included. For example, most of antibiotics have bacterial specific target (cell wall synthesis, folic acid metabolism and so on). However, there is no information about the mechanism of RD53 in antibiotic activity. Since the direction of this paper is not well-organized, authors should specify the purpose and importance of library, and then reorganize the experiments. 2. The authors mentioned the inhibitory effect of RD53 is correlated with cytosolic compartments of bacteria through 3 min staining data (Figure 7). The time scale is not suitable to observe the inhibitory effect. Authors should show the localization of RD53 depending on time until the cytotoxic point and confirm it with bacteria live/death marker. 3. In Table 1, most of MICs show too wide range against pathogen bacteria. It would be better to repeat the experiment and confine the values into smaller range of better defined number. 4. The authors showed that RD53 has a bactericidal activity comparing to three other standard antibiotics by colony counting and claimed that RD53 is suitable for clinical application. It is a little too premature claim and at least the sensitivity of RD53 on mammalian cell should be compared to bacteria. 5. There is too many information scattered in many figures and table apart from their explanation, so it is difficult to follow the contents. It would be helpful to rearrange them in more consistent manner, for example, by putting the description right in front of or behind figures. 6. The authors mentioned about low-yielding compounds. It would be better to improve the yield and include the compound into the library, or just leave it out. 7. It would be good to show the full name of abbreviation at the first time of the usage. For example, MIC was used without explanation when it first came out. 8. It would be better to change the word 'overmine', which is not common word and it is difficult to catch the meaning. 9. In figure 9, it would be better to number each figure for better understanding.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): (1) As most of the rhodamines are salts, the anion is required to calculate the yield. However, as noted by Sparr and Lavis, a standard work-up gives different salts or salt mixtures. These anions would therefore have to be either quantitatively determined, exchanged with a suitable anion exchange resin, or avoided by using the acid in the work-up that corresponds the precursor halogens.
Reply: In our original synthetic procedure, we dried the organic layer after extraction with MgSO 4 ·H 2 O. So, it is indeed possible that both chloride and sulfate may be present as the counterion.
We have revised purification procedure, which we resynthesized those rhodamine analogs insalt form. To be specific, we used CaCl 2 to dry the organic layer after extraction to avoid the introduction of sulfate anion. This way, we ensured that rhodamine dyes are paired with Clonly. We have accordingly recalculated the yields.
(2) Because of the diversity of the library, the physical properties of the products vary tremendously: the isolation and handling of the individual products should therefore be described in detail. Reply: We have updated the synthetic procedures of each compound in the SI, following the suggestion of the reviewer 1.
All the rhodamine analogs were prepared with the same protocol, i.e. stirring 100 mg of various substrates (S1-S70) with 1.3 equiv. of dilithium reagent (1) at -78 o C in anhydrous THF for 2 hrs. Each reaction was worked up with the same protocol, i.e. quenching with saturated NH 4 Cl solution, extraction with CH 2 Cl 2 , drying with CaCl 2 , evaporation under reduced pressure, and column chromatographic purification. We note that different eluent was used for different rhodamine dyes. Therefore, the composition of eluent was specified individually.
(3) The impact of these results would be however much improved by a specific description of the principal concepts. Reply: The primary challenge for the development of novel antibiotic drugs is the lack of an alternative molecular space capable of providing abundant hits exhibiting favorable antibiotic activities for further optimization after the soil actinomycetes has been over-exploited.
It is the objective of this manuscript to share with the community that the diversityoriented rhodamine library is potentially such a sought-after molecular space for discovery of antibiotic hit compounds. We have reorganized the entire manuscript to support the viability of this principle concept.
First, with the chemistry section, we report the facile construction of a focused rhodamine library and quantitatively analyzed the structural diversity of the library with a well-established chemometric protocol.
Second, with the phenotypic screening section, we confirm that (1) the existence of surprisingly abundant hits exhibits antibiotic activity, (2) few shortlisted hits (RD22 and RD53) exhibit high inhibitory activity and broad spectrum toward 11/12 WHO prioritized drugresistant pathogens, (3) RD22 and RD53 has potent bactericidal activities against Grampositive/negative pathogens, and (4) RD22 and RD53 do not readily induce bacterial resistance.
Third, the potential mechanism of action of RD53 was probed with SEM, cross-resistance experiments, and whole-genome sequencing of the RD53-resistant A. baumannii.
Fourth, the structure-activity relationship was investigated to further substantiate that the rhodamine core is the key pharmacologically active motif to render antibiotic activity.
We feel that the principle concept of this manuscript is presented clearly and convincingly with such an organization of the manuscript.
(4) To sufficiently describe the experimental results, detailed protocols for the individual product should be provided. Furthermore, for determining the yields, the salts require an unambiguous anion assignment. Reply: Among the 70 compounds (RD1-RD70) of this library, 10 of them, i.e. RD4, RD19, RD20, RD26, RD35, RD43, RD51, RD52, RD54, RD58, were found to predominantly exist in the ring-closed form by 1 H-NMR, And 7 of them were in their inner salt forms, i.e. RD6, RD18, RD28, RD50, RD55, RD57, RD59. The synthetic procedures and yields of these 17 compound do not need revision.
As for the other 43 compounds, we attempted to resynthesize them all, following a revised workup and purification procedure to avoid the presence of other anions but Cl -. To be specific, CaCl 2 was used in the new syntheses, instead of MgSO 4 , as the dehydrating agent of the organic layer after extraction to avoid potential introduction of SO 4 2as the counterion of the rhodamine dye.
Very interestingly, essentially identical amount of pure dyes were obtained with the first four dyes we resynthesized (RD1, RD2, RD3, and RD7), compared to the previous synthesis with the original workup and purification protocol. This makes us wonder if the chloride is the counterion of the rhodamine dyes, with either CaCl 2 or MgSO 4 as the dehydrating agent. We then tried another three (RD25, RD49, and RD60) to verify our hypothesis. Indeed, the same amount of dyes (within a very reasonable range) was obtained.
We updated the SI with the revised procedure. We acknowledge that our original yields were calculated with molecular weight without Cl -. Therefore, the original yields were indeed questionable. We have recalculated these yields with the correct molecular weight. We thank the reviewer for the help with this issue.

Reviewer #2 (Remarks to the Author):
(1) This reviewer feels that in a broader discipline journal such as Nature Communications, the synthetic aspects could be condensed and some of the information included in the Supplemental Information.

Reply:
We have greatly reduced the discussion of the synthesis by more than half, following the suggestion of the reviewer.
(2) The diversity analysis presented claims differentiation from the previous library but it would be useful to briefly include information, possibly scores from totally unrelated libraries, to enable the non-specialist reader to understand the relative significance of the 18.875 to 3.228 difference calculated between the 2 rhodamine libraries.

Reply:
The Extended-connectivity fingerprints (ECFPs) method is a commonly employed fingerprint topological methodology to capture precise atom environment substructural features. The ECFP feature count for a chemical library is widely used as a rapid assessment of its structural diversity and can be used to compare the structural diversity in different libraries, especially in literatures related to phenotypic screening and drug development. One of the original papers reporting this method has been cited over 1170 times (Reference 39: Rogers, D. & Hahn, M. Extended-connectivity fingerprints. J. Chem. Inf. Model. 50, 742-754 (2010)).
The ECFPs method can be used to different level of sophistication. In our study, ECFP_6 was employed, where 6 means the effective diameter of the largest feature. The total number of features is the sum of the features discovered at 6 iteration steps from all the members of a library. Then the diversity number fingerprint features (DNFPF) was defined as the total number of fingerprint features divided by the number of molecules.

DNFPF=
the total number of fingerprint features the number of molecules In reference 39, four representative existing chemical libraries were assessed with the ECFP_6 method. The first library contains 50,000 compounds, all of which were randomly selected from the Derwent World Drug Index. This library is a chemical library of high structural diversity and expected to yield a high DNFPF index. This library was calculated to contain over 750,000 ECFP features. Therefore, a DNFPF value of ca. 25 was found.
In the same manuscript, another library of 50,000 compounds was assessed. All the compounds of this library were selected from a combinatorial library with an indole core. The structural diversity of this library is low and a small DNFPF value is expected. Indeed, a DNFPF value less than 1.4 was found.
We hope that these two literature examples can help the audience to make sense the relative significance of 18.875 vs 3.228.
(3) The finding of some antibacterial activity with a subset of the analogs provides the best rationale for publication of this manuscript in a cross-discipline journal. The initial cutoff of 8 mg/L and 16 mg/L for MRSA and A. baumannii, respectively is a modest cutoff for potential starting points to identify lead molecules. The authors chose to further evaluate RD53 based on claims of best activity, but should explain their rationale for not exploring RD22 which exhibited the same initial profile as RD53. Having 2 structurally different examples from their library would have strengthened their claim that this chemistry platform could be useful for the identification of novel leads. Reply: After we finished preparation of this rhodamine library, we actually performed three rounds of phenotypic screening.
All analogs (RD1-RD70) were screened against MRSA and A. baumannii to check their MIC's, which are tabulated in Figure 4. Ones with the lowest MIC's, hence highest inhibitory activity, against either MRSA or A. baumannii were shortlisted.

Round 2.
The best few from the round 1 were screened against an additional list of pathogens, i.e. three Gram-positive and three Gram-negative pathogens, to have a preliminary touch of their antibiotic spectrum. RD22 and RD53 are the two best candidates based on their antibiotic potency and spectrum width. Round 3.
Originally, only RD53 but not RD22 was chosen for further widespectrum testing, because it does exhibit slightly overall better performance and because we wish to reduce the efforts/cost of biological evaluations. However, the reviewer 2 is absolutely correct that having more active analogs will greatly strengthen our claim that this chemical space is potentially a viable source of antibiotic compounds. Therefore, we further tested RD22 for its spectrum-width of antibiotic activity (Table 2), time-dependent killing curves against MRSA, VRE and polymyxin E resistant A. baumannii (Figure 5), and resistance acquisition of pathogens during serial passaging ( Figure  6). All the data suggest that RD22 is as promising as RD53 in all aspects of antibacterial properties, if not better.
(4) The A. baumannii staining in Figure 7 (panel C) is at a resolution that is difficult to confirm the authors conclusions. After zooming in on the figure it appears that some of the cells in Figure  7C have intracellular staining whereas other at the top of the frame do not. Given the difference in shape between the species, and the thicker cell membrane of MRSA, the claim on lines 223-224 about the location of the target needs to be substantiated further. Either quantitative measurements of a statistically relevant number of individual cells, use of mutant strains, or even hybridization of non-denatured immobilized proteins after separation of cellular fractions with RD53 should be explored. Reply: In our original manuscript, the fluorescence images of RD53 were included to offer some insights on the potential localization of target. However, we agree with the reviewer that the tentative conclusion that "the inhibitory effect of RD53 toward the bacteria likely involves the interaction of RD53 was cytosolic components of bacteria" needs further substantiation.
It is not routine practice to tentatively elucidate the potential antibacterial target via fluorescence imaging. However, these fluorescence images are viable data and could be of potential significance, therefore we simply moved the fluorescence imaging results to SI without further discussion.
After we received the comments and revision suggestions, we have carried out four biological studies to probe the potential mechanism of action of RD53, i.e. 1

) haemolysis of red blood cells, 2) SEM imaging of RD53 treated MRSA and A. baumannii, 3) the whole genome sequencing of RD53 resistant A. baumannii, and 4) cross-resistance of RD53 and five antibiotics against A. baumannii.
The aforementioned data are not yet sufficient to nail down the underlying mechanism of action of RD53. However, it is sure that the destruction of cell wall or membrane is not responsible. Interaction of RD53 with cytosolic components of bacteria does.
(5) The activity of RD53 against penicillin non-susceptible Streptococcus pneumoniae should be tested and reported; these are on both the WHO and CDC pathogen lists. Reply: The three clinical strains of S. pneumonia are penicillin-resistant strains, i.e. PRSP. We forgot to include this information in the original submitted manuscript.
RD53 has a MIC of 2-4 g/mL toward PRSP, and RD22 an even lower MIC of 0.5-1 g/mL. Therefore, both RD22 and RD53 exhibit high inhibitory potency toward PRSP.
(6) The activity of RD53 against eukaryotic cells in a cytotoxicity assay should be provided to ensure that the antibacterial activity is not positively correlated with general cytotoxicity (and also erythrocyte assays to confirm absence of membrane disruption).

Reply:
We assayed the haemolytic activity of both RD53 and RD22 against human red blood cells (RBCs). A Lysis20 value of above 100 g/mL was measured for both RD53 and RD22. Therefore a high therapeutic index, i.e. Lysis20/MIC of above 100 was calculated. Note: Lysis20 values are the concentration of a compound that lyses 20% or less of red blood cells (RBC).
The CC 50 value of RD53 was measured against human umbilical vein endothelial cells (HUVECs), to be 3.912 g/mL. We admit that this value is not yet optimal if RD53 is intended for systemic use. However, there's ample room for further structural modification to reduce its cytotoxicity during the in-depth lead optimization process. Second, RD53 itself could readily find practical use against dermal infections, in which case its potential cytotoxicity is less a concern. Therefore, we believe that RD53 along with other members of this rhodamine library is a major progress in the field of antibiotic development and still deserves the prestige of Nat. Comm.
(7) Are the graphs in Figure 9 (and supplementary figure S132) simply the concentration of RD53 that supports growth upon passage? Individual colonies should be purified from each passage and characterized for absolute fold change with respect to RD53 (and control drugs). Multiple mutational lineages can co-exist and propagate in these types of population resistance experiments and the cultures should be purified to accurately understand independent mutations. Reply: The concentration of RD53 in original Figure 9 (which is figure 4 in the revised manuscript) means the highest concentration (tested up to 128-256xMIC) that bacteria could grow to a minimum OD600 of 0.2 at 24 hour intervals. We agree that multiple mutational lineages can co-exist and propagate in these types of population resistance experiments. So we purified the 5 th and 10 th passage of heritable resistant bacteria such as Acinetobacter baumannii and Escherichia coli on MHA plates and randomly selected thirty individual colonies to test the MICs of control drugs, RD22 and RD53 to characterize for absolute fold change with respect to those compounds. Before measuring MIC, each individual colony was inoculated to antibiotic-free broth for three times. We found that the different colonies of bacteria population gave different MIC's. The MIC's of control drugs to almost all individual colonies of heritable resistant bacteria were higher than C x in Figure 4. In comparison, more colonies of RD22 and RD53 resistant bacteria could not grow in C x .
In conclusion, this experiment indicates that these bacteria had lower tendency of heritable resistance acquisition toward RD22 and RD53 than what Figure 4  (8) The resistant mutants that emerged upon serial passage with A. baumannii and E. coli need to be characterized ( Figure 9C and S132A). What is their phenotypic profile with respect to susceptibility to other antibacterial drugs? Furthermore, the genome of these mutants should be sequenced to identify the source of the mutation that reduces the susceptibility as this may also provide significant insight to the identity of the target. Reply: We have checked the phenotypic profile of RD53-resistant A. baumannii against five other clinical antibiotics, e.g. tigecycline (TGC), ciprofloxacin (CIP), polymyxin E (PmE), levofloksacin (LEV), ampicillin-sulbactam (SAM). As we can see from the following table, the MIC of A. baumannii enhanced by 32 fold from 4 mg/mL to 128 mg/mL. But, the MIC of this RD53-resistant strain remained essentially unchanged toward five other antibiotics. This suggests that RD53 likely works via a different mechanism of action than these five antibiotics. We further carried out the whole-genome sequencing of this RD53-resistant A. baumannii. Comparative analysis revealed the existence of eight nonsynonymous single nucleotide variants (SNV) and one nonframeshift deletion, which belong to resistance-associated genes encoding efflux pumps, metabolism DNA replication or transcriptional regulation related proteins. The mutation in rsmJ encoding 16S rRNA (guanine1516-N2)-methyltransferase is particularly interesting. Notably, the mutation in 16S rRNA methyltransferase has been shown to cause kasugamycin resistance in E.coli.

(10)
The authors claim that that RD53 does not acquire resistance in MRSA and use vancomycin as a surrogate (line274-275). This is not an appropriate comparison as the experimental conditions used do not enable the horizontal transfer of the van cluster that provides this resistance. Reply: We agreed that acquired resistance mainly contains two mechanisms, which is caused by importing foreign genetic elements and mutations in genome. So we correct acquire resistance to mutational resistance which is more accurate.

(11)
Line 294 claims that RD53 has favorable properties for drug development. To make this claim the authors need to provide data on some other characteristics that would be important in this regard, such as metabolic and chemical stability, plasma protein binding and solubility. Lines 301-307; although the examples provided on antibacterial drugs coming from small focused libraries is tantalizing, there are just as many that require significant chemistry programs to optimize. The fact is true that RD53 could progress and coming from a small library is not an exclusionary factor, this claim would be significantly more compelling if additional data on drug-like properties are generated and presented as suggested. Reply: We thank the reviewer for this comment. These two properties are favorable for drug development.
1. The heamolytic activity of RD53 and RD22 is low with a Lysis20 values above 100 g/mL. 2. Both RD22 and RD53 have a sufficient water solubility of above 5 mg/mL. We also checked the cytotoxicity of RD53 toward mammalian cells. For example, a CC 50 of 3 g/mL was recorded for human umbilical vein endothelial cells (HUVECs). We admit that this value is not yet optimal if RD53 is intended for systemic use. However, there's ample room for further structural modification to reduce its cytotoxicity during the in-depth lead optimization process. Second, RD53 itself could readily find practical use against dermal infections, in which case its potential cytotoxicity is less a concern.
We checked the chemical stability of RD53 and RD22. Both are stable toward nucleophiles (e.g. H 2 O, H 2 S, GSH, Cys), oxidants (e.g. H 2 O 2 ), and reductants (e.g. NADPH, GSH) up to their physiological levels, since the absorbance of their solution remain unchanged upon addition of these chemicals. These data are trivial and not included in the SI.
However, the metabolic stability of RD22 and RD53 were not studied since we feel this experiment should be performed only after we have confirmed the in vivo antibiotic activity in near future.
We were not able to provide all the requested data yet at this moment. However, we do believe that, with all the data we have accumulated, RD53 could be regarded as a feasible lead compound potentially suitable for further chemical optimization.
Therefore, we believe that RD53 along with other members of this rhodamine library is a major progress in the field of antibiotic development and still deserves the prestige of Nat. Comm.

(12)
Throughout; the tense should be standardized appropriately to be consistent. Reply: We have standardized the tense of the manuscript.
(13) Line 32-33; reword for clarity Reply: The sentence has been rewritten into "The resistant strains may readily mutate to resist these newer analogs if their existing resistance mechanisms do not already exhibit partial crosseffectiveness."

(14)
Line 35-36; Change to something like: "The genes encoding resistance to the natural products that are present within the original organisms can then be horizontally transferred to pathogenic microbes enabling resistance to emerge"; Reply: We deeply thank the reviewer for help with prosing of our manuscript.

(15)
Line 36; adjuvants should be better defined. Reply: The sentence regarding "adjuvants" is not very relevant to the current context and therefore removed from the text.  The original writing has been corrected.