Bayesian Modeling and Intrabacterial Drug Metabolism Applied to Drug-Resistant Staphylococcus aureus

Drug-Resistant Staphylococcus aureus

Article

Bayesian Modeling and Intrabacterial Drug Metabolism Applied to

Jimmy S. Patel, Javiera Norambuena, Hassan Al-Tameemi, Yong-Mo Ahn, Alexander L. Perryman,

Xin Wang, Samer S. Daher, James Occi, Riccardo Russo, Steven Park, Matthew Zimmerman,

Hsin-Pin Ho, David S. Perlin, V é ronique Dartois, Sean Ekins, Pradeep Kumar, Nancy Connell,

Je ﬀ rey M. Boyd,* and Joel S. Freundlich*

Cite This: https://doi.org/10.1021/acsinfecdis.1c00265

Read Online

ACCESS

Metrics & More

Article Recommendations

sı Supporting Information

ABSTRACT: We present the application of Bayesian modeling to identify chemical tools and/or drug discovery entities pertinent

to drug-resistant Staphylococcus aureus infections. The quinoline JSF-3151 was predicted by modeling and then empirically

demonstrated to be active against in vitro cultured clinical methicillin- and vancomycin-resistant strains while also exhibiting eﬃcacy

in a mouse peritonitis model of methicillin-resistant S. aureus infection. We highlight the utility of an intrabacterial drug metabolism

(

IBDM) approach to probe the mechanism by which JSF-3151 is transformed within the bacteria. We also identify and then validate

two mechanisms of resistance in S. aureus: one mechanism involves increased expression of a lipocalin protein, and the other arises

from the loss of function of an azoreductase. The computational and experimental approaches, discovery of an antibacterial agent,

and elucidated resistance mechanisms collectively hold promise to advance our understanding of therapeutic regimens for drug-

resistant S. aureus.

KEYWORDS: Bayesian modeling, Staphylococcus aureus, quinoline, intrabacterial drug metabolism, YceI, AzoR

S. aureus is a coagulase, catalase, and Gram-positive cocci

disease and burden for the population and for medical

practitioners.

residing on the skin and anterior nares of about 30% of healthy

individuals. Considered to be the most pathogenic of the

Staphylococci, S. aureus is responsible for a multitude of human

diseases, which are further complicated by its capability to

cause infection by at least two primary mechanisms: direct

tissue invasion and exotoxin production. Direct tissue invasion

is the primary cause of pathogenesis in humans, of which skin

and soft tissue infections, such as abscesses, have become the

Traditionally, treatment for S. aureus infections has included

β-lactams, such as penicillin G, nafcillin, and oxacillin. This

class of antibacterials targets transpeptidation (cross-linking) of

peptidoglycan stem peptides. In 1959, methicillinanother

antibiotic in this familywas licensed for use against S. aureus

in England. A year later, reports of methicillin-resistant S.

aureus (MRSA) surfaced in Europe, and in 1968, the ﬁrst

most common condition to be treated by physicians. Other

signiﬁcant diseases caused by S. aureus invasion of tissue

include osteomyelitis, endocarditis, and pneumonia. Metastasis

of any of these focal infections can lead to bacteremia and

sepsis. On the other hand, toxin-mediated diseases caused by S.

aureus include food poisoning, toxic shock syndrome, and

scalded skin syndrome. Thus, S. aureus is a major cause of

Received: May 14, 2021

XXXX American Chemical Society

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Article

Table 1. Biological Proﬁling of the Most Active Compounds versus MRSA

The MIC values are representative of three independent experiments. The model inherent in the selection of the molecule from the top-scoring 50

compounds is listed ﬁrst in the Model/Score column.

hospital outbreak of MRSA in the United States was

recognized. Resistance to methicillin is conferred by the

gonorrhoeae, is leveraged to discover and validate novel hit

compounds against drug-resistant S. aureus. We demonstrate

the successful prediction of in vitro growth inhibitory activity

for quinoline JSF-3151, we proceed to show its value as a

bactericidal agent against clinically relevant drug-resistant S.

aureus strains with in vivo eﬃcacy in a mouse model of

infection, and we utilize genetic methods to identify and

validate two mechanisms of resistance while probing their

eﬀects on JSF-3151 through intrabacterial drug metabolism

mecA gene, which encodes PBP2a, a penicillin-binding protein

with low aﬃnity for β-lactams. During the period of 1987 to

014, surveillance data have shown that S. aureus was a

signiﬁcant source of hospital-acquired infections, and nearly

0% of all hospital-acquired S. aureus infections were due to

,8−10

MRSA.

Nearly 20 years ago, treatment options for MRSA

included vancomycin and trimethoprim/sulfamethoxazole.

Unfortunately, therapeutic options for MRSA infections have

changed little, and reports of vancomycin-intermediate-

resistant and vancomycin-resistant S. aureus (VISA and

21,22

(

IBDM).

RESULTS

■

11−13

VRSA, respectively) make this notion more alarming.

Development of the Bayesian Model. The PubChem

thorough review by Kurosu et al. shows that the pipeline for

Bioassay database (https://pubchem.ncbi.nlm.nih.gov) was

queried with the keywords “aureus and MIC.” We ultimately

selected the studies with the largest number of compounds

tested against drug-resistant S. aureus to develop our training

sets. Inspection of each data set was conducted to remove

duplicates, choose the most conservative value of the MIC

(minimum concentration of compound to inhibit bacterial

growth in vitro by 90%), convert units from μM to μg/mL, and

ensure that the strain being tested was drug-resistant. We

found that the assays used S. aureus strains resistant to one or

more of the following drugs: methicillin, linezolid, vancomycin,

agents against MRSA contains few options. Therapies such

as dalbavancin, oritavancin, and tedizolid, though approved for

use, are not utilized in the inpatient setting due to cost and lack

of eﬃcacy data for bacteremia and sepsis.

The majority of chemical space has yet to be explored for

the seeds of new treatments for drug-resistant S. aureus

infections. Herein, we present the application of a computa-

tional approach to advance drug discovery. Bayesian modeling,

previously validated for use with data sets for the growth

6−19

inhibition of Mycobacterium tuberculosis

or Neisseria

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

structed with screening data extracted from ChemBank

Article

mammalian cell line) (Supplementary Table 3). Of the 15 top

scoring compounds assayed, three of the ﬁve most active

compounds exhibited an MIC ≤ 10 μg/mL versus the MRSA

strain (Table 1). All three hits were from the MRSA_1b

model. Interestingly, the highest scoring compound from

MRSA_1a, the ﬂuoroquinolone JSF-3144, displayed an MIC of

25 μg/mL (50 μM) against both MRSA and MSSA strains.

JSF-3144 and JSF-3145 did not score within the range (11.7−

15.3) for the top 50 compounds from the MRSA_1b model.

The three MRSA_1b selected actives either scored just outside

(JSF-3151, JSF-3157) of or within (JSF-3153) the range

(7.06−9.95) for the top 50 compounds from the MRSA_1a

model. Of the eight MRSA_1b bottom-scoring compounds in

the Enamine library, all were inactive with an MIC > 50 μg/

mL (Supplementary Table 4 ). As evidenced by our results with

the bottom- and top-scoring compounds, the Bayesian models

are able to enrich for active compounds while triaging

inactives. Furthermore, as seen with our previous Bayesian

24−26

models, the results also validate the utility of pruning.

chemotype should have a heightened likelihood of operating

via a novel mechanism of action. Utilizing this strategy, we

manually pruned the active tetracyclines, ﬂuoroquinolones,

rhodanines, other known PAINS, and β-lactams from the

actives of MRSA_1a to produce a new training set

Quinoline JSF-3151 proved to be the most promising of the

three hits based on its MIC values of 4.0 μg/mL (12 μM) and

1.6 μg/mL (4.8 μM) versus MRSA and MSSA strains,

respectively, and its Vero cell CC₅₀value of >50 μg/mL

(>150 μM). This gave JSF-3151 MRSA and MSSA selectivity

indices (a ratio of mammalian cell cytotoxicity to antibacterial

potency, SI = CC /MIC) of >12 and >31, respectively. The

(

MRSA_1b) that contained 1247 compounds, of which 657

compounds were classiﬁed as active (52.7%). These data sets

were used as independent training sets to create two Bayesian

models in Pipeline Pilot 9.1 (BIOVIA). We used the eight

standard physiochemical descriptors plus the structural ﬁnger-

prints FCFP6 (functional class ﬁngerprints of maximum

diameter 6, which utilize bits for diﬀerent functional groups).

For internal statistics obtained through 5-fold internal cross-

validation, MRSA_1a and MRSA_1b displayed Receiver

Operator Characteristic (ROC) scores of 0.903 and 0.917,

respectively. The sensitivity values were 93.5% and 96.5%,

respectively, while the speci ﬁ city metrics were 81.7% and

other two hits, JSF-3153 and JSF-3157, failed to exhibit

acceptable SI values (≥10). Searches of the SciFinder

(http://sci ﬁ nder.cas.org ), PubChem (https://pubchem.ncbi.

nlm.nih.gov ), and ChEMBL databases (https://www.ebi.ac.

uk/chembl ) supported the novelty of our ﬁnding of the

antibacterial eﬃcacy of JSF-3151. To further characterize the

compound, we determined the minimum bactericidal concen-

tration (MBC) against both MRSA and MSSA strains. The

MBC values of 0.39−0.78 μg/mL (1.2−2.3 μM) versus MSSA

and 1.6−3.1 μg/mL (4.8−9.3 μM) versus MRSA demon-

strated the compound is capable of killing ≥2 log colony-

1.0%, respectively (Supplementary Table 1). The models

were then assessed with an external test set of 374 in-house

compounds (synthesized by the Freundlich laboratory and

assayed versus MRSA ATCC 43300, of which 6.2% were

active). A comparison was also included with a methicillin-

sensitive S. aureus (MSSA) model (Broad_MSSA), con-

forming units (CFU) per mL (Supplementary Figure 2).

Furthermore, a time-kill curve showed that JSF-3151 at 1×

MIC sterilized a MRSA culture within 4 h, more than 6 times

faster than vancomycin at the same fold MIC (Supplementary

Figure 2). Because of the increasing emergence of vancomycin

12,13

resistance in the world,

we also quantiﬁed the MIC of JSF-

http://chembank.broadinstitute.org) (Supplementary Tables

and 2). The positive predictive value (Supplementary Table

) demonstrated the order of performance to be MRSA_1a >

3151 against VISA and VRSA strains. Gratifyingly, we found

that JSF-3151 exhibited even greater potency against six

clinical isolates of VISA and ﬁ ve clinical isolates of VRSA, as

compared with MRSA (Supplementary Table 5).

MRSA_1b ≫ Broad_MSSA (Supplementary Figure 1).

The MRSA_1a and MRSA_1b models were applied in a

prospective virtual screen of a library of three million

commercially available compounds from Enamine (http://

www.enamine.net) accessed through the ZINC server (http://

zinc.docking.org). For each compound in the library and each

of the two MRSA models, a Bayesian score was calculated,

such that the value scaled with the probability of the molecule

being eﬃcacious given our activity cutoﬀ of 10 μg/mL. The

top-scoring 15 candidates (10 from MRSA_1a and 5 from

MRSA_1b) and bottom-scoring 8 candidates from MRSA_1b

were assessed for in vitro growth inhibition activity. The MIC

values against MRSA (heretofore referring to ATCC 43300

unless noted otherwise) and MSSA (heretofore referring to

ATCC 25923 unless noted otherwise) strains were determined

along with the Vero cell CC₅₀(minimum concentration of

compound resulting in 50% growth inhibition of this model

In Vivo Study of JSF-3151 in a Mouse Peritonitis

Model. To support mouse eﬃcacy studies, we quantiﬁed the

mouse pharmacokinetic (PK) proﬁle of JSF-3151, demonstrat-

ing an inability to reach the MIC over a 5 h time period

(Supplementary Figure 3). This outcome is consistent with the

compound’s poor kinetic aqueous solubility (S ≤ 0.060 μM)

and low metabolic stability (half-life t of 1.46 min) in the

presence of mouse liver microsomes. Given its poor solubility,

JSF-3151 provided a suspension when formulated in sterile

saline. JSF-3151 was dosed in BALB/c mice at 24 mg/kgthe

highest achievable dose given its solubilityvia the intra-

peritoneal route. Nevertheless, 2 out of 12 (16.7%) mice

infected with the MRSA COL strain survived 3 days post-

administration of a single dose of JSF-3151, as compared with

0/12 with the vehicle-only control and 12/12 with the 100

mg/kg subcutaneous dose of vancomycin (Supplementary

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

The MIC values are representative of three independent experiments.

Article

Table 2. MIC Proﬁling of JSF-3151 versus Spontaneous Resistant Mutants

strain name

strain notes

JSF-3151

ampicillin

doxycycline

rifampicin

USA300_FPR3757

JMB1100

USA300_FPR3757 yceI1

USA300_FPR3757 yceI2

USA300_FPR3757 yceI3

ATCC 43300

parent

1.6

3.1

>50

3.1

>50

<0.25

USA300_LAC cured of pUSA03

T → G mutation at position −43 in yceI

adenosine base insertion at −37 in yceI

adenosine base deletion at position −41 in yceI

MRSA

ATCC 25923

MSSA

3.1

<0.25

Figure 1. IBDM studies of JSF-3151 and JSF-3640. (A) Time course study depicting the intrabacterial accumulation of JSF-3151 within S. aureus

strains JMB1100 (WT), yceI2, and azoR::Tn that were treated with JSF-3151. (B) Time course study depicting the intrabacterial accumulation of

JSF-3640 within S. aureus strains JMB1100 (WT), yceI2, and azoR::Tn that were treated with JSF-3151. (C) High-resolution mass spectrometry

supports assignment of JSF-3640 as the amine metabolite, formed from JSF-3151, based on its spectral signature.

Figure 3). This result, perhaps not surprising since the

Bayesian model was only trained to predict in vitro eﬃcacy

and not aqueous solubility, mouse liver microsomal stability,

and in vivo eﬃcacy, suggests that analogues of JSF-3151 with

optimized PK properties could demonstrate improved eﬃcacy

in this infection model.

accumulation of JSF-3151 within the bacterium (Figure 1A).

We also noted a time-dependent increase of a new species

labeled JSF-3640 (Figure 1B) whose high-resolution mass

spectrometry (HRMS) (Figure 1C) data were consistent with

the corresponding amine metabolite of JSF-3151:5-amino-N-

(6-chloroquinolin-8-yl)thiophene-2-carboxamide. Critically,

the synthetic metabolite coeluted with the authentic metabolite

on the LC (Supplementary Figure 4) and exhibited the same

low-resolution and high-resolution mass spectrum proﬁles.

Together these results support that JSF-3640 is, indeed, the

amine metabolite of JSF-3151. The amine metabolite was

whole-cell inactive (MIC > 50 μg/mL versus MRSA), despite

the ability of JSF-3640 to accumulate within MRSA in a time-

dependent fashion (Supplementary Figure 4 ).

Analysis of Drug Metabolism Demonstrates Intra-

bacterial Reduction of JSF-3151. We next sought to probe

the mechanism of action for JSF-3151 and began with the

determination as to whether JSF-3151 accumulates within the

bacterium using our recently developed approach to studying

21,22

intrabacterial drug metabolism (IBDM).

Brieﬂy, a MRSA

WT strain (JMB1100; Table 2) was incubated with 10× MIC

of JSF-3151 and was then sampled at 0, 15, 30, 45, and 90 min

post-treatment. Following the quenching of metabolism and

extraction of the cellular contents soluble in −80 °C 2:2:1

acetonitrile:methanol:water, analysis by liquid chromatogra-

phy−mass spectrometry (LC-MS) showed a time-dependent

Membrane Depolarization and NO^•Release Are

Ruled Out with Regard to the JSF-3151 Mechanism of

Action. Subsequently, we sought to rule out the potential for a

nonspeciﬁc mechanism of action such as membrane depola-

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Article

rization. MRSA was incubated with DiBAC , bis(1,3-

dibutylbarbituric acid)trimethine oxonol, which demonstrates

Table 3. The Eﬀect of Introducing the yceI Alleles into S.

aureus LAC

increased ﬂuorescence upon binding to proteins of a

8,29

genotype

pLL39

pLL39_yceI

MIC (μg/mL)

depolarized cell,

151. JSF-3151 failed to compromise the cell membrane, as

compared with the positive control sodium deoxycholate

and then treated with 4.5× MIC of JSF-

6.2

pLL39_yceI1

Supplementary Figure 5 ).

pLL39_yceI2

Given the propensity for nitroheterocycles to extrude nitric

•

30−32

pLL39_yceI3

oxide (NO ),

we examined the ability of JSF-3151 to

•

yceI::Tn pLL39

yceI::Tn pLL39_yceI

yceI::Tn pLL39_yceI1

yceI::Tn pLL39_yceI2

yceI::Tn pLL39_yceI3

0.78

1.6

produce NO . We assessed intrabacterial NO release via the

dye 4-amino-5-methylamino-2′,7′-diﬂuoroﬂuorescein diacetate

(

DAF-FM diacetate). In addition, the Griess reagent was

used to detect the presence of nitrite in the supernatant, which

•

is the oxidation product of NO that undergoes eﬄux from the

The MIC values are representative of three independent experi-

bacterium. Despite treatment of MRSA with 100 μM JSF-

ments. All strains are isogenic and constructed in the LAC

background.

151, neither DAF-FM (over 2 h) nor the Griess reagent (over

8 h) evidenced NO production (Supplementary Figure 5), as

•

compared with the DEA-NONOate positive control. Con-

sistent with these experiments was our observation that the

des-nitro JSF-3151 (N-(6-chloroquinolin-8-yl)thiophene-2-

carboxamide; synthesized independently as JSF-3587) did

not accumulate within the cell according to our above IBDM

isogenic yceI::Tn mutant (JMB9587) (Supplementary Table

). The strains aﬀorded two functional copies of yceI in strain

JMB1100 and only one functional copy in the yceI::Tn strain.

We next quantiﬁed the JSF-3151 MIC values for each of these

strains (Table 3; Supplementary Figure 6 ). As expected, the

strains containing any of the three yceI promoter mutations led

to high level resistance to JSF-3151. In addition, the yceI::Tn

mutant had a lower MIC than with JMB1100. A quantitative

real-time PCR (qPCR) study showed that relative to strain

LAC, each of the three mutants had increased transcription of

yceI (Supplementary Figure 6 ). Finally, we cloned a PCR

•

studies. The NO detection assays and our IBDM platform

suggest that the bactericidal eﬃcacy of JSF-3151 is not

•

attributed to NO formation.

A Bacterial Lipocalin and an Azoreductase Are

Linked to Drug Resistance. Using the MRSA strain LAC

(Table 2), we used a genetic screen to select for resistant

mutants using 8× MIC of JSF-3151. Single colonies were

fragment containing yceI into the pEPSA5. The pEPSA5_yceI

vector contained yceI under the transcriptional control of a

xylose-inducible promoter to enable gene overexpression.

Following overnight induction with xylose, we witnessed a

signiﬁcant resistance to JSF-3151 (>64 fold) as compared with

the strain carrying the empty vector (pEPSA5), as well as the

no induction controls (Table 4). We hypothesized that

apparent following 3 d of incubation with a frequency of 6 ×

−

0 . In comparison, resistance frequencies for front-line

agents such as vancomycin, ciproﬂoxacin, and linezolid at 8×

3,35,36

their respective MIC value have been reported

to range

_b_e_t_w_e_e_n₁_×₁₀₋9 and 1 × 10 . Ampicillin, erythromycin,

−6

doxycycline, rifampicin, and vancomycin exhibited the same

MIC versus both JSF-3151-resistant strains and the parental

strain. Whole-genome sequencing of three representative

mutants, demonstrating >16-fold loss of susceptibility to JSF-

Table 4. The Eﬀect of Overproducing yceI Gene Product on

JSF-3151 Resistance

151, exhibited three distinct mutations in the promoter region

of the gene SAUSA300_2620 (Table 2). Previously published

reports suggested that the SAUSA300_2620 gene product,

annotated as a YceI-like protein, may have a similar function to

the Burkholderia cenocepacia lipocalin (BcnA), which has been

demonstrated to confer resistance to hydrophobic small

strain ID

growth with xylose

MIC (μg/mL)

LAC pEPSA5

LAC pEPSA5_yceI

LAC pEPSA5

0.78

>50

0.78

3.1

LAC pEPSA5_yceI

37,38

molecules through their selective binding.

Sanger sequenc-

The MIC values are representative of three independent experi-

ing of these representative mutants, labeled as yceI1, yceI2, and

yceI3, further conﬁrmed the presence of these mutations in the

promoter region (Table 3). We hypothesized that the

mutations in the promoter region led to an increase in

production of YceI, which consequently aﬀords resistance to

JSF-3151.

ments.

increased expression of yceI would decrease the accumulation

of JSF-3151 in cells. To test this hypothesis, IBDM analysis

was performed with the yceI2 strain. Consistent with our

hypothesis, treatment of the yceI2 strain with JSF-3151 resulted

in a decreased rate of accumulation when compared with the

parent strain. We also noted a decrease in the rate of JSF-3640

accumulation.

A second genetic screen was performed using the LAC

yceI::Tn strain. Strains with heritable resistance to JSF-3151

arose at a frequency of approximately 3 × 10 . Whole-genome

sequencing found the four strains had single nucleotide

polymorphisms (SNP) in the gene SAUSA300_0206 (azoR).

The mutations in azoR resulted in the following protein

To explore this hypothesis, our ﬁrst goal was to observe the

eﬀect of each mutation with respect to JSF-3151 resistance. We

cloned a PCR fragment containing −290 base pairs upstream

of the predicted translational start site and yceI into a

chromosomal integration vector (pLL39). Five constructs

were generated: empty vector, WT promoter+gene, and each

of the three mutated promoter+gene sequences. Each of these

plasmids was individually integrated into S. aureus strain

−

39,40

RN4220 in the L54a attB locus.

Integrates were veriﬁed

using PCR (Supplementary Table 6). The individual

integrated episomes were transduced into two diﬀerent S.

aureus strains: parental USA300_LAC (JMB1100) and an

changes: Trp₁₀₀stop (azoR1), Trp stop (azoR2), Thr₁₂₁Ile

(azoR3), and Trp stop (azoR4); respectively, they exhibited

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Article

MIC shifts of 15-, 32-, 64-, and 25-fold as compared with the

yceI::Tn parental strain (Table 5). For comparison, a LAC

mutants, the intrabacterial levels of both compounds did not

reach those found in the WT strain until at least 45 min

postaddition of JSF-3151 to the culture.

Table 5. MIC Proﬁling of JSF-3151 versus Spontaneous

Resistant Mutants in the Background of theyceI::Tn

Mutant

DISCUSSION

■

Antibacterial resistance is a serious threat to global public

health, and its burden to healthcare increases every year. The

diseases caused by S. aureus infection are no exception, as they

are among the most commonly encountered cases in the clinic

today. Growing resistance to the therapeutic options currently

available, such as β-lactam resistance and vancomycin

resistance, necessitates new antibacterial candidates. Thus,

new approaches are required to aﬀord novel drug discovery

and tool compounds, especially with time- and cost-eﬃciency.

Here, our Bayesian modeling approach was shown to be

successful in screening millions of compounds and at a fraction

of the time traditional high-throughput screening would take.

The discovery of JSF-3151 via a machine learning approach

validates the use of this strategy for identifying active

genotype

LAC

yceI::Tn

MIC (μg/mL)

1.5

0.78

azoR::Tn

yceI::Tn azoR1

yceI::Tn azoR2

yceI::Tn azoR3

yceI::Tn azoR4

>50

The MIC values are representative of three independent experi-

ments. All strains are isogenic and constructed in the LAC

background.

strain with a transposon insertion in azoR was constructed and

demonstrated an increased MIC when compared with LAC

compounds versus drug-resistant S. aureus, extending eﬀorts

16−19

from our laboratories with M. tuberculosis

and Neisseria

(

50 versus 1.5 μg/mL). Moreover, returning azoR to the

gonorrhoeae and joins one publication we note that

mutant strains via the plasmid pCM28 decreased MIC values

to the levels of the yceI::Tn strain (Table 6). Overexpression of

azoR in LAC using the pEPSA5 vector increased sensitivity to

JSF-3151 (Table 6). These data verify that the lack of AzoR

increases JSF-3151 resistance.

speciﬁcally focused on drug-resistant S. aureus. This

publication from Wang et al. disclosed machine learning

models for the growth inhibition of MRSA that correctly

predicted (1/56 = 1.8% hit rate) an active (MIC = 4−8 μg/

mL) compound ((E)-2-isopropyl-5-styrylbenzene-1,3-diol)

versus only MRSA and MSSA strains and did not mention

critical downstream studies to assess in vivo eﬃcacy and

mechanism of action/resistance. Furthermore, our Bayesian

models have signiﬁcantly expanded beyond methicillin

resistance to include clinically relevant data sets with strains

resistant to linezolid, vancomycin, or ﬂuoroquinolones.

Our Bayesian models were built with 46 data sets pertaining

to assays against drug-resistant S. aureus. We utilized these

assays in particular because they examined the highest number

of compounds against drug-resistant S. aureus. These data sets

were crucial in developing the “prior distribution,” or rather

giving statistical weight to the chemical moieties and

physiochemical properties relating to activity and inactivity.

Our Bayesian models were successful in enriching active and

inactive small molecules. Through the use of the MRSA_1b

Table 6. The Eﬀect of Overproducing azoR Gene Product

on Resistance to JSF-3151

genotype

MIC (μg/mL)

LAC pEPSA5

1.5

0.39

0.78

>50

3.1

LAC pEPSA5_azoR

yceI::Tn pCM28

yceI::Tn pCM28_azoR

yceI::Tn azoR1 pCM28

yceI::Tn azoR1 pCM28_azoR

yceI::Tn azoR2 pCM28

yceI::Tn azoR2 pCM28_azoR

yceI::Tn azoR3 pCM28

yceI::Tn azoR3 pCM28_azoR

yceI::Tn azoR4 pCM28

1.5

>50

1.5

model constructed from the pruned training set, we were led

to JSF-3151, a small molecule that possesses anti-Staph-

ylococcal activity and a novel structure for an antibacterial.

Thus, we hypothesized that the unique structure of JSF-3151

could imply a new mechanism of antibacterial activity. While

nonspeciﬁc mechanisms of action were ruled out, the target

remains elusive. The in vitro bactericidal activity of JSF-3151

implies that it modulates the activity of one or more targets

that alone or together are essential for bacterial growth.

Spontaneous-resistant mutants were raised that increased

transcription of yceI, encoding a bacterial lipocalin protein

that proposedly binds JSF-3151 in analogy to the elegant

studies of Valvano and colleagues with Burkholderia

yceI::Tn azoR4 pCM28_azoR

1.5

The MIC values are representative of three independent experi-

^m^eⁿ^t^s^.c Strains containing pEPSA vectors were cultured with 1%

xylose. All strains were constructed in the LAC background.

The azoR gene is predicted to encode for an intracellular

FMN-dependent NADH-azoreductase, catalyzing reductive

43,44

cleavage of azo groups. Azoreductases from E. coli,

aeruginosa, B. wakoensis A01, and R. sphaeroides have also

been reported to reduce nitro groups. We hypothesized that

this enzyme is responsible for the biotransformation of JSF-

151 to JSF-3640. To test this hypothesis IBDM analysis was

37,49

performed with the azoR::Tn strain. Interestingly, the

accumulation of both JSF-3151 and metabolite JSF-3640

were decreased in the azoR::Tn strain as compared with the

WT strain (Figure 1, Supplementary Table 8 ). In comparison,

the intrabacterial accumulation of JSF-3151 was further

depressed in the yceI2 mutant (Figure 1, Supplementary

Table 8 ). The relative level of accumulation of metabolite JSF-

cenocepacia.

While their work involved complementation

of the B. cenocepacia lipocalin knockout with S. aureus yceI, we

are unaware of any reports in the literature detailing the role of

YceI in S. aureus drug resistance. Our genetic studies have

demonstrated that overexpression of yceI leads to loss of JSF-

3151 in vitro activity. The analysis of IBDM data with the yceI2

mutant evidenced decreased intrabacterial exposure to both

the parent JSF-3151 and identiﬁed metabolite JSF-3640 as

640 was similar in the yceI and azoR mutants. With both

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Article

compared with the WT strain. Taken together, these ﬁndings

support the hypothesis that YceI is decreasing the rate of JSF-

metabolite JSF-3640 within the bacterium. The chemical

entity, knowledge of resistance mechanisms, and platforms of

Bayesian modeling and intrabacterial drug metabolism should

inform further eﬀorts to impact next-generation therapeutic

strategies for drug-resistant S. aureus infections.

151 entry into S. aureus cells. Future studies will be required

to determine if YceI binds JSF-3151.

Raising JSF-3151 mutants in the yceI::Tn background

aﬀorded mutations in the gene azoR. Strains with null

mutations in azoR had increased resistance to JSF-3151. We

are unaware of previous reports of drug-resistant mutants in S.

aureus occurring through mutations in an azoreductase despite

elegant studies of the role of this protein family in

biotransformations of nitroaromatic and azo small mole-

METHODS

■

In the detailed methods below, no unexpected or unusually

high safety hazards were encountered.

Materials and Services. Quick ligase, Phusion polymer-

ase, and restriction enzymes were purchased from New

England Biolabs. Gel puriﬁcation kits were purchased from

Qiagen. Tryptic Soy broth (TSB) was purchased from MP

Biomedicals. Mueller−Hinton broth and agar were purchased

from Fisher Scientiﬁc. Whole-genome sequencing was

conducted by the Genomics Center, Rutgers University-New

Jersey Medical School, while Sanger sequencing was performed

at Genewiz, South Plainﬁeld, NJ.

3−47,50

cules.

It is possible that AzoR is involved in the

transformation of JSF-3151 into a yet to be identiﬁed

compound that has increased antibacterial activity against S.

aureus. Alternatively, while JSF-3640 did not exhibit anti-

bacterial activity despite its ability to enter cells, its formation

from JSF-3151 as promoted by AzoR could be cidal. In support

of the hypothesis, the azoR mutant had decreased intracellular

accumulation of JSF-3640.

The drug resistance and IBDM results are supportive of

decreased intrabacterial exposure to JSF-3151 correlating with

reduced growth inhibition of S. aureus. Overexpression of yceI

provides depressed accumulation of, and resistance to, JSF-

Mice. All animal experiments were approved by the Rutgers

Institutional Animal Care and Use Committee. All experiments

used 6-week old female outbred CD-1 or Swiss Webster mice

(

Charles River Laboratories, Wilmington, MA). Mice were

housed in ﬁlter-top cages and maintained in accordance with

American Association for Accreditation of Laboratory Care

criteria.

151. Loss-of-function mutations in azoR also decrease

intrabacterial levels of JSF-3151 and increase resistance to

JSF-3151. However, the degree of biotransformation of JSF-

Bacterial Strains, Culture Conditions, Primers, and

Plasmids. MRSA (43300) and MSSA (25923) strains used in

this study were purchased from ATCC (Manassas, VA). VRSA

and VISA strains were obtained from the Kreiswirth laboratory

at the Public Health Research Institute Center (Newark, NJ).

Other S. aureus strains used in this study were derivatives of

the S. aureus strain LAC, which is a community acquired

MRSA strain. JMB1100 is a derivative of LAC that has been

151 to JSF-3640 as quantiﬁed by IBDM in these strains and

in the WT does not appear to be linked to growth inhibition.

JSF-3640 is a whole-cell inactive metabolite that appears to be

formed by a pathway largely independent of the activity of

AzoR. The AzoR-dependent biotransformation product of JSF-

151 has yet to be identiﬁed and may be linked to the

mechanism of action of JSF-3151.

It is noteworthy that reports of S. aureus metabolomics

appear to primarily focus on the eﬀect of antibacterials on

cured of the pUSA03, which confers multidrug resistance. S.

aureus were routinely cultured in solid or liquid tryptic soy

broth. For experiments, S. aureus were cultured in 5 mL of

Mueller Hinton and grown at 37 °C in 30 mL capacity culture

tubes with shaking at 200 rpm. Overnight cultures were diluted

50−55

bacterial metabolite pools

and not on intracellular drug

metabolism. In vitro enzyme assays have been leveraged as a

major approach to study enzyme-mediated small-molecule

6,57

modiﬁcations

but are potentially limited by a lack of

1/1000 and transferred into 96-well plates containing serial

physiological relevance to intrabacterial metabolism. We are

aware of one elegant report that utilized LC-MS to quantify

drug accumulation in S. aureus, but it is noteworthy that drug

dilutions of JSF-3151 in Mueller−Hinton broth. Final

concentrations were calculated based on 200 μL ﬁnal volume.

Plates were wrapped with paraﬁlm from the sides to prevent

dehydration and incubated statically at 37 °C. Results were

measured after 18 h of incubation. When selecting for

plasmids, transposon, or episome insertions, antibiotics were

added at the ﬁnal following concentrations: 150 μg/mL

ampicillin; 30 μg/mL chloramphenicol (Cm); 10 μg/mL

erythromycin (Erm); 3 μg/mL tetracycline (Tet). For routine

plasmid maintenance, media were supplemented with 10 μg/

mL or 3.3 μg/mL of chloramphenicol.

Data Collection and Curation. The PubChem Bioassay

Database was queried with the phrase “aureus and MIC”,

which yielded 22 619 for results. After sorting the largest sets

with S. aureus results, the strain in each assay was identiﬁed and

sorted into “drug-resistant,” “drug-sensitive,” and “unknown.”

In total, 139 assay results were collected. For each of these

assays, the structural data ﬁles (SDF) and comma-separated

value (CSV) ﬁles were downloaded. The SDF’s contained

every tested compound’s PubChem CID and 2D structural

information (but no assay results). The CSV ﬁles contained

every tested compound’s CID and assay results (but no

structures). Additionally, for every AID, the corresponding

paper was retrieved. Every assay result from a CSV ﬁle was

metabolism was not considered. Increased emphasis on the

application of IBDM to S. aureus programs should further

enlighten our understanding of bacterial metabolic pathways,

and especially those engaged in xenobiotic biotransformations,

while also providing information critical to the design of more

potent analogues of JSF-3151 and, thus, furthering the impact

of antibacterial machine learning models.

CONCLUSIONS

■

The results presented herein disclose an antibacterial agent

JSF-3151 with both in vitro and in vivo eﬃcacy against clinically

relevant drug-resistant S. aureus strains. This molecule was

discovered by the successful application of a Bayesian

modeling approach. The mechanism of resistance to JSF-

151 has been shown to involve (1) overexpression of the

lipocalin-like protein YceI or (2) the loss of function of the

azoreductase AzoR. To the best of our knowledge, neither of

these resistance mechanisms has been described for staph-

ylococci and both are supported by extensive genetic analyses.

Furthermore, we have demonstrated how each type of

mutation perturbs accumulation of JSF-3151 and its identiﬁed

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Article

inspected and compared to its relevant paper. Using this

quality control process, any erroneous data could be noted or

discarded from our study. Following the comprehensive

manual inspection process, we found 72 data sets pertaining

to drug-sensitive S. aureus, 46 data sets pertaining to drug-

resistant S. aureus, and 21 data sets that were classiﬁed as

unknown. The drug-resistant sets contained assay results

primarily for MRSA but also included assay results against

ﬂuoroquinolone-resistant or linezolid-resistant S. aureus, as

well. Unfortunately, one of the CSV ﬁles from the drug-

resistant set, AID 548647, had erroneous data and was,

therefore, discarded. The data within the CSV ﬁles and SDF’s

for each of the drug-resistant sets were collated into a

spreadsheet in Discovery Studio 4.0 (BIOVIA, Inc., San Diego,

CA). For consistency, the μM sets were converted to μg/mL.

We then deleted duplicate compounds using a custom script in

Pipeline Pilot 9.1 (BIOVIA, Inc., San Diego, CA) and

compounds with a molecular weight >850 g/mol.

These descriptors were then combined in diﬀerent ways, with

diﬀerent weights applied to the sets of diﬀerent descriptor

combinations, until the most accurate model was produced,

according to 5-fold internal cross-validation.

The top 100 scoring compounds from MRSA_1a and the

top 50 from MRSA_1b versus an Enamine (www.enamine.net)

library of 3 million compounds, accessed through Zinc

(http://zinc.docking.org), were then visually inspected to

identify and discard any reactive compounds or promiscuous

inhibitors (e.g., Michael acceptors, rhodanines, etc.). Also

removed were compounds that were visually deemed too

similar to the training set to be novel (e.g., tetracyclines or

almost all ﬂuoroquinolones), or compounds with chemotypes

that were not of general interest (i.e., peptides or fatty

23,60

acids).

49 of the top 50 compounds from MRSA_1a did

not pass inspection. Therefore, we also inspected the next 50

top scoring compounds from this model. Unsurprisingly, the

highest-ranking compound from MRSA_1a was a ﬂuoroqui-

nolone which, instead of dismissing, we included in this

model’s predictions to be experimentally validated. We

selected an additional top 9 compounds from MRSA_1a, as

well as the top 5 ranking compounds from MRSA_1b that

passed inspection. All candidate compounds were purchased

and assessed by LC/MS for ≥95% purity and the expected

parent ion in the mass spectrum.

Training and Test Sets. In total, our full training set from

PubChem contained 1633 compounds, of which 1043

compounds (63.8%) were designated as active (MIC ≤ 10

μg/mL). This full training set (MRSA_1a) encompassed a

small yet fairly diverse portion of chemical space, which

included known antibacterial scaﬀolds such as tetracyclines,

ﬂuoroquinolones, and β-lactams. However, it also included

rhodanines (which are known to be promiscuous Pan Assay

Drug Susceptibility Assays. MICs were performed in 96-

well microtiter plates. A DMSO stock solution of each

compound was added to the ﬁrst column and serially diluted

across the columns of the plate. The last column of the plate

contained no drug and served as a no-drug control. Overnight

cultures of the bacterium being tested (MRSA ATCC 43300,

MSSA ATCC 25923, or the aforementioned VRSA/VISA

23,60

Interference compounds, called “PAINS”

). We hypothe-

sized that removing (or pruning) these problematic com-

pounds as well as known antibacterial chemotypes from the

training set would help teach the Bayesian to further focus on

new chemical scaﬀolds, which should increase the likelihood

that they display novel mechanisms of action. It is important

to note, however, that our M. tuberculosis Bayesian models have

routinely found hits which diﬀer signiﬁcantly from the model

testing set (pairwise Tanimoto similarity <0.7). Utilizing this

strategy, we manually pruned the tetracyclines, ﬂuoroquino-

lones, rhodanines, and β-lactams from the actives subset of

MRSA_1a to produce a new training set (MRSA_1b) that

contained 1247 compounds, of which 657 compounds were

classiﬁed as active (52.7%).

strains) were diluted 1000 fold (2 × 10 cells), and 100 μL was

used as an inoculum in each well. MICs were determined by

visual inspection for a pellet after 18 h incubation at 37 °C.

Compounds were tested for cytotoxicity against Vero cells

using the CellTiter 96 AQueous One Solution kit (Promega).

Vero cells were seeded in 96-well plates at a density of 2 × 10

cells per well, and the plates were incubated for 4 h at 37 °C to

allow attachments of the Vero cells. Compounds were then

added to the wells starting from a ﬁnal concentration of 50 μg/

mL and making 12 1:2 dilutions. Cells were incubated for 72 h

at 37 °C. Then 20 μL of freshly prepared MTS:PMS reagents

was added to each well. The plates were incubated for 2 h and

then read at an absorbance of 490 nm.

Another training set that we used as a reference was based

on the Broad Institute’s assay results against methicillin-

sensitive S. aureus (Broad_MSSA). This training set has 10 934

compounds, of which 193 compounds (1.77%) were deﬁned as

“active” according to the Z-factor threshold selected by the

researchers at the Broad Institute (Tali Mazor, assay names:

ChemBank BacTiterGlo 1064.0004 and 1064.0005; http://

chembank.broadinstitute.org). These 193 actives also included

known antibacterials, such as levoﬂoxacin, gatiﬂoxacin, β-

lactams, tetracyclines, and ﬂuoroquinolones.

Mouse Pharmacokinetic Study. Animals and ethics

assurance: Animal studies were carried out in accordance

with the guide for the care and use of Laboratory Animals of

the National Institutes of Health, with approval from the

Institutional Animal Care and Use Committee (IACUC) of

the New Jersey Medical School, Rutgers University, Newark.

All animals were maintained under speciﬁc pathogen-free

conditions and fed water and chow ad libitum, and all eﬀorts

were made to minimize suﬀering or discomfort. Two female

CD-1 mice received a single dose of experimental compound

administered orally at 25 mg/kg in 20% DMA/80% PEG300,

These data sets were used as independent training sets to

create two diﬀerent, new drug-resistant S. aureus machine

learning models and the reference MSSA model in Pipeline

Pilot 9.1. These Bayesian models utilized nine diﬀerent

descriptors: AlogP, molecular weight, number of rings, number

of aromatic rings, number of rotatable bonds, number of

hydrogen bond donors, number of hydrogen bond acceptors,

molecular fractional polar surface area, and molecular function

class ﬁngerprints of maximum diameter 6 (FCFP_6, character-

izing the 2D substructures up to and including 6 rings/zones/

and blood samples were collected in K EDTA coated tubes

predose, 0.5, 1, 3, and 5 h postdose. Blood was kept on ice and

centrifuged to recover plasma, which was stored at −80 °C

until analyzed by HPLC coupled to tandem mass spectrometry

(LC-MS/MS).

1,62

dimensions of topology).

Together, these descriptors

deﬁne the physiochemical properties of each compound as a

whole and the 2D substructures of diﬀerent regions within it.

LC/MS-MS analytical methods: LC/MS-MS quantitative

analysis for all molecules was performed on a Sciex Applied

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Article

Biosystems Qtrap 4000 triple-quadrupole mass spectrometer

coupled to an Agilent 1260 HPLC system, and chromatog-

raphy was performed on an Agilent Zorbax SB-C8 column (2.1

performed on an Agilent 1260 HPLC coupled to an Agilent

6120 MS. The run condition for the LC-MS was 10−100%

acetonitrile in water over 10 min (both acetonitrile and water

contained 0.1% formic acid). The column used for the run was

an EMD Millipore Chromolith Speedrod RP18-e, 50 × 4.6

mm.

30 mm; particle size, 3.5 μm) using a reverse phase gradient

elution. Milli-Q deionized water with 0.1% formic acid (A) was

used for the aqueous mobile phase and 0.1% formic acid in

acetonitrile (B) for the organic mobile phase. The gradient

was: 5−90% B over 2 min, 1 min at 90% B, followed by an

immediate drop to 5% B and 1 min at 5% B. Multiple-reaction

monitoring of parent/daughter transitions in electrospray

positive-ionization mode was used to quantify all molecules.

Sample analysis was accepted if the concentrations of the

quality control samples and standards were within 20% of the

nominal concentration. Data processing was performed using

Analyst software (version 1.6.2; Applied Biosystems Sciex).

Neat 1 mg/mL DMSO stocks for all compounds were ﬁrst

serial diluted in 50/50 acetonitrile/water and subsequently

Membrane Depolarization Assay. A single colony of S.

aureus (ATCC 43300) was inoculated in fresh Mueller−

Hinton media and grown to an OD of 0.5. DiBAC (bis(1,3-

dibutylbarbituric acid)trimethine oxonol) was added to the

resulting culture to aﬀord a ﬁnal concentration of 20 μg/mL

and incubated for 5 min. All reads were performed in triplicate

on a Biotek Synergy Neo 2 plate reader with excitation and

emission wavelengths set to 490 and 510 nm, respectively. In a

Costar 96-well, black-side, clear-bottom plate, 200 μL of

bacteria and dye was added and kept as a negative control. At a

separate location of the plate, 200 μL of bacteria and dye was

added along with 4.5× MIC of JSF-3151 (∼18 μg/mL, 54

μM). At yet another location, 200 μL of bacteria and dye was

added along with sodium deoxycholate, the positive control, to

a ﬁnal concentration of 0.05% w/v. Reads were performed for

10 min with 5 s intervals.

serial diluted in drug free CD-1 mouse plasma (K EDTA,

Bioreclamation IVT, NY) to create standard curves (linear

regression with 1/x weighting) and quality control (QC)

spiking solutions. Twenty μL of standards, QCs, control

plasma, and study samples were extracted by adding 200 μL of

acetonitrile/methanol 50/50 protein precipitation solvent

containing the internal standard (10 ng/mL verapamil).

Extracts were vortexed for 5 min and centrifuged at 4000

rpm for 5 min. 100 μL of supernatant was transferred for

HPLC-MS/MS analysis and diluted with 100 μL of Milli-Q

deionized water.

Nitric Oxide Release Assays. (A) DAF-FM: A single

colony of S. aureus (ATCC 43300) was inoculated in fresh

Mueller−Hinton media to an OD of 0.2. DAF-FM diacetate

(4-amino-5-methylamino-2′,7′-diﬂuoroﬂuorescein diacetate)

was added to a ﬁnal concentration of 10 μM and allowed to

incubate for 30 min in the dark. All reads were performed in

triplicate on a Biotek Synergy Neo 2 plate reader with

excitation and emission wavelengths set to 495 and 515 nm,

respectively. In the ﬁrst row of a Costar 96-well, black-side,

clear-bottom plate, 200 μL of Mueller−Hinton media alone

was added and kept as a negative control. In the second row,

200 μL of bacteria alone were added and also kept as a

negative control. In the third row, bacteria treated with DAF-

FM were added. In the fourth row, bacteria treated with DAF-

FM and JSF-3151 (∼33 μg/mL, 100 μM) for 120 min were

added. In the ﬁfth row, bacteria treated with DAF-FM and 100

μM diethylamine NONOate diethylammonium salt (DEA-

NONOate) for 120 min were added. (B) Griess: Sodium

nitrite stock solutions ranging from 50−0.78 μM (2-fold

dilutions) were prepared in Mueller−Hinton media. A single

colony of S. aureus (ATCC 43300) was inoculated in fresh

Mueller−Hinton media and grown to an OD of 0.4. The

Mouse Peritonitis-Sepsis Model. MRSA COL strain was

grown overnight in Mueller−Hinton media at 37 °C. The

resulting culture was diluted in 5% hog mucin and 0.9% NaCl

to a challenge inoculum of approximately 1.0 × 10 CFU per

mouse. The challenge inoculum was conﬁrmed by mannitol

salt agar plate spreads and administered to female outbred

Swiss Webster mice (Charles River Laboratories, Wilmington,

MA) via intraperitoneal injection (0.5 mL). Groups consisting

of 12 mice were given single doses of vehicle (10% DMSO in

sterile saline), JSF-3151 (24 mg/kg) or vancomycin (100 mg/

kg) 1 h postinfection. Vehicle and JSF-3151 were administered

via intraperitoneal injection while vancomycin was adminis-

tered by subcutaneous injection (ﬁnal injection volume = 0.2

mL). Mice were maintained in accordance with the American

Association for Accreditation of the Laboratory Care criteria.

All surviving animals were humanely euthanized at the end of

the study (3 d) by following the American Veterinary Medical

Association Guidelines for the Euthanasia of Animals. The

animal study was approved by Rutgers Institutional Animal

Care and Use Committee.

resulting culture was split into three equal volume batches. The

ﬁrst batch served as the negative control, the second volume

was treated with JSF-3151 (∼33 μg/mL, 100 μM), and the

third volume was treated with DEA-NONOate (100 μM).

Following incubations of 24 and 48 h, 1 mL of culture was

sampled and centrifuged. The supernatant was placed in the

same plate as mentioned above along with the nitrite standards

(180 μL per well). To each of these wells, 20 μL of freshly

prepared Griess reagent mix was added, and the plate was

allowed to sit at room temperature for 30 min. The plate was

read with the above reader at an absorbance of 548 nm. Reads

from the sodium nitrite standards were used to correlate

absorbance to concentration, which facilitated estimation of

nitrite production.

IBDM Studies. A single colony of S. aureus strain was

inoculated in fresh Mueller−Hinton media and grown

overnight at 37 °C. The resulting overnight was diluted

:100 in fresh Mueller−Hinton media, and 120 mL of the

culture was grown in a 250 mL ﬂask until OD was ∼0.5.

600

Then, JSF-3151 or JSF-3640 was added at the appropriate

concentration and 10 mL of the culture were sampled at the

relevant time intervals post-treatment. Each aliquot was

centrifuged. The supernatant fraction was immediately stored

while the pellet fraction was washed once with PBS and then

resuspended in ice cold 2:2:1 acetonitrile:methanol:water to

halt bacterial metabolism, and immediately placed on dry ice.

The cells in the pellet fraction were subjected to bead beating

and all of the fractions (6 pellet, 6 supernatant) were ﬁltered

Molecular Biology. Escherichia coli DH5α was used as a

cloning host throughout. Plasmids were passaged through

RN4220 and subsequently transduced into the appropriate

63,64

strains using bacteriophage 80α.

The amplicons used to

(

0.22 μm membrane) and analyzed by LC-MS. LC-MS was

create the pll39_yceI vectors were: 2620upBamHI and

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Article

620downSalI. The pLL39 vector was linearized with Sall and

added with stirring. After 3 h, the product A was concentrated

under reduced pressure to aﬀord a light yellow solid (88%

yield). 6-chloroquinolin-8-amine (1.03 g, 5.78 mmol) was

added in a round-bottom ﬂask followed by the addition of 15

mL dichloromethane. A (0.971 g, 5.09 mmol) was then added

followed by diisopropylethylamine (1.20 equiv, 6.94 mmol,

1.21 mL). The mixture was allowed to stir at room

temperature for about 6 h. The reaction mixture was washed

twice with water and saturated aqueous sodium bicarbonate

solution and once with saturated aqueous brine solution. The

organic layer was dried over anhydrous sodium sulfate and

then ﬁltered through a pad of Celite. The ﬁltrate was

concentrated in vacuo and the product was puriﬁed by ﬂash

column chromatography (10% methanol:dichloromethane) to

BamHI and combined with similarly digested PCR fragments.

The pLL39_yceI constructs were transformed into RN4220

containing pLL2787 and integrated onto the chromosome at

the L54a attB site as previously described. Episome

integration was veriﬁed using the Scv8 and Scv9 primers.

The integrated pLL39_yceI constructs were transduced

individually from RN4220 into JMB1100. The yceI::Tn and

azoR::Tn alleles were acquired from the NARSA collection

and were transduced individually into JMB1100 before further

strain construction. LAC genomic DNA was used as a template

and the following primer pairs were used to create amplicons

for cloning: 2620uppEPBamHI and 2620downSalI; pEPSA5-

206 EcoRI for and pEPSA5-0206 Bam HIrev; pCM28-0206

BamHI rev and pCM28-0206 PstI for (S6 Table). The pEPSA5

and pCM28 vectors were linearized with restriction enzymes

and combined with similarly digested PCR fragments before

ligation. Sanger sequencing was used to conﬁrm vectors

aﬀord the product as a yellow solid (1.67 g, 87% yield): H

NMR ((CD ) SO, 500 MHz) δ 10.9 (s, 1H), 9.06−9.00 (m,

3 2

1H), 8.51 (s, 1H), 8.47 (d, J = 8.2 Hz, 1H), 8.24 (d, J = 2.8

Hz, 1H), 8.16 (d, J = 3.0 Hz, 1H), 7.99 (s, 1H), 7.79−7.73 (m,

(

Genewiz; South Plainﬁeld, NJ). Single nucleotide poly-

1H). Also noted 5.8 (s, DCM), 1.2 (m). C NMR

morphisms were mapped as previously described.

((CD

) SO, 126 MHz) δ 158.8, 154.3, 150.4, 145.2, 137.9,

3 2

Total RNA Extraction and Quantitative Real-Time

136.6, 135.1, 131.4, 130.7, 129.4, 129.1, 124.0, 122.8, 119.5.

PCR (qPCR). The S. aureus strains, wild type and three JSF-

HRMS (ESI): Calculated for C14

ClN

S (M+H)

151 spontaneous-resistant mutants, were grown to mid log

334.0053; Observed 334.0069 (+4.8 ppm).

phase (OD₅₉₅= 0.4). Bacteria were harvested and the pellets

were lysed, post TRIzol (Invitrogen) addition, by bead beating.

The total RNAs were extracted in chloroform and puriﬁed via

RNeasy mini kit (Qiagen). The cDNA library was constructed

via reverse transcription using 2 μg total RNA as template with

SuperScript III First-Strand Synthesis System (Invitrogen).

The expressions of yceI and 16S rRNA were measured by

qPCR via SuperScript III Platinum SYBR Green One-Step

qPCR Kit with Rox (Invitrogen).

Compound 3 − JSF-3587.

JSF-3587 was synthesized following General Procedure A on

a 0.3 mmol scale. Following ﬂash chromatography, the product

Compound Synthesis. Reagents and solvents for syn-

thesis were purchased from Sigma-Aldrich, Fisher Scientiﬁc, or

Enamine Ltd., and used without further puriﬁcation unless

noted otherwise. Analytical TLC was performed with Merck

silica gel 60 F₂₅₄plates. Silica gel column chromatography was

conducted with Teledyne Isco CombiFlash Companion or Rf+

systems. H and C NMR spectra were acquired on a Bruker

Avance III HD 500 MHz instrument and are listed in parts per

million downﬁeld from tetramethylsilane. LC-MS was

performed on an Agilent 1260 HPLC coupled to an Agilent

was isolated as a white solid (82 mg, 95% yield): H NMR

(

(CD ) SO, 500 MHz) δ 10.6 (s, 1H), 9.01 (dd, J = 4.2, 1.5

3 2

Hz, 1H), 8.61 (d, J = 2.3 Hz, 1H), 8.45 (dd, J = 8.3, 1.6 Hz,

H), 8.03 (d, 1H), 7.99 (d, J = 5.0 Hz, 1H), 7.90 (d, J = 2.3

Hz, 1H), 7.75 (dd, J = 8.3, 4.2 Hz, 1H), 7.31 (t, J = 4.7, 3.9 Hz,

H). Also noted 5.7 (s, DCM), 3.3 (s, H O). LRMS (ESI):

Calculated for C H ClN OS (M+H) = 289.0; Observed

289.0.

General Procedure B: Synthesis of JSF-3640.

120 MS. All synthesized compounds were at least 95% pure

as judged by their HPLC trace at 220 or 250 nm and were

characterized by the expected parent ion(s) in the MS trace.

HRMS data were acquired on an Agilent 6220 Accurate-Mass

Time-of-Flight mass spectrometer.

General Procedure A: Synthesis of JSF-3151.

JSF-3151 (10 mg, 0.030 mmol) was added to a 20 mL

scintillation vial. This was followed by the addition of Zn

powder (9.7 mg, 0.15 mmol), ammonium formate (9.5 mg,

.15 mmol), and 5 mL of methanol. The reaction was allowed

to stir for 1 h at room temperature. Workup and puriﬁcation

was performed as outlined by General Procedure A. The

reaction mixture was washed twice with water and saturated

aqueous sodium bicarbonate solution and once with saturated

aqueous brine solution. The organic layer was dried over

anhydrous sodium sulfate and then ﬁltered through a pad of

Celite. The ﬁltrate was concentrated in vacuo and the product

was puriﬁed by ﬂash column chromatography (10% meth-

anol:dichloromethane) to aﬀord the product.

JSF-3151 was prepared in two steps. In short, a round-bottom

ﬂask was charged with 5-nitrothiophene-2-carboxylic acid

JSF-3640 was synthesized following General Procedure B.

Following ﬂash chromatography, the product was isolated as

(1.00 g, 5.78 mmol). Thionyl chloride (10 mL) was then

an oﬀ-white solid (8.1 mg, 88% yield): H NMR ((CD ) SO,

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

The Supporting Information is available free of charge at

Article

00 MHz) δ 10.1 (s, 1H), 8.96 (d, J = 4.1 Hz, 1H), 8.59 (d, J =

.1 Hz, 1H), 8.41 (d, J = 8.3 Hz, 1H), 7.78 (d, J = 2.1 Hz, 1H),

.72 (dd, J = 8.3, 4.2 Hz, 1H), 7.53 (d, J = 4.1 Hz, 1H), 6.75

Xin Wang − Department of Pharmacology, Physiology, and

Neuroscience, Rutgers University − New Jersey Medical

School, Newark, New Jersey 07103, United States; Present

Address: X.W.: Harvard University HSPH, Boston,

Massachusetts 02115, United States; orcid.org/0000-

0002-6930-3347

(

s, 2H), 5.99 (d, J = 4.1 Hz, 1H). Also noted 5.7 (s, DCM), 3.3

s, H O), 1.2 (s). HRMS (ESI): Calculated for C H ClN OS

M+H) = 304.0311; Observed 304.0304 (−2.3 ppm).

14 10

Samer S. Daher − Department of Pharmacology, Physiology,

and Neuroscience, Rutgers University − New Jersey Medical

School, Newark, New Jersey 07103, United States

James Occi − Department of Medicine, Center for Emerging

and Re-emerging Pathogens, Rutgers University − New Jersey

Medical School, Newark, New Jersey 07103, United States

Riccardo Russo − Department of Medicine, Center for

Emerging and Re-emerging Pathogens, Rutgers University −

New Jersey Medical School, Newark, New Jersey 07103,

United States

ASSOCIATED CONTENT

sı Supporting Information

https://pubs.acs.org/doi/10.1021/acsinfecdis.1c00265.

■

Comparative model external ROC curves; In vitro

bactericidal activity of JSF-3151; In vivo proﬁling of

JSF-3151; IBDM studies of JSF-3151 and its metabolite_•

JSF-3640; JSF-3151 membrane depolarization and NO

release assays; Quantiﬁcation of yceI transcription in JSF-

Steven Park − Public Health Research Institute, Rutgers

University − New Jersey Medical School, Newark, New Jersey

151-resistant strains; Internal Statistics for Bayesian

Models; External Statistics for Bayesian Models; S.

aureus MIC values for top-scoring compounds from

MRSA_1a and MRSA_1b predictions with the Enamine

library; MRSA MIC values for bottom-scoring com-

pounds from MRSA_1b predictions with the Enamine

library; MIC of JSF-3151 versus VRSA and VISA strains;

Primers used in this study; Strains and plasmids used in

this study; Accumulation metrics for JSF-3151 and JSF-

7103, United States; Present Address: S.P., M.Z., D.S.P.,

V.D.: Center for Discovery and Innovation, Hackensack

School of Medicine, 111 Ideation Way, Nutley, New

Jersey 07110, United States

Matthew Zimmerman − Public Health Research Institute,

Rutgers University − New Jersey Medical School, Newark,

New Jersey 07103, United States; Present Address: S.P.,

M.Z., D.S.P., V.D.: Center for Discovery and Innovation,

Hackensack School of Medicine, 111 Ideation Way,

Nutley, New Jersey 07110, United States

640 in select strains (PDF)

Hsin-Pin Ho − Public Health Research Institute, Rutgers

AUTHOR INFORMATION

Corresponding Authors

■

University − New Jersey Medical School, Newark, New Jersey

07103, United States

Jeﬀrey M. Boyd − Department of Biochemistry and

Microbiology, Rutgers, The State University of New Jersey,

New Brunswick, New Jersey 08901, United States;

orcid.org/0000-0001-7721-3926; Email: je ﬀ boyd@

SEBS.rutgers.edu

Joel S. Freundlich − Department of Pharmacology, Physiology,

and Neuroscience, Rutgers University − New Jersey Medical

School, Newark, New Jersey 07103, United States;

Department of Medicine, Center for Emerging and Re-

emerging Pathogens, Rutgers University − New Jersey

Medical School, Newark, New Jersey 07103, United States;

orcid.org/0000-0002-3411-3455; Email: freundjs@

rutgers.edu

David S. Perlin − Public Health Research Institute, Rutgers

University − New Jersey Medical School, Newark, New Jersey

7103, United States; Present Address: S.P., M.Z., D.S.P.,

V.D.: Center for Discovery and Innovation, Hackensack

School of Medicine, 111 Ideation Way, Nutley, New

Jersey 07110, United States

Véronique Dartois − Public Health Research Institute, Rutgers

University − New Jersey Medical School, Newark, New Jersey

7103, United States; Present Address: S.P., M.Z., D.S.P.,

V.D.: Center for Discovery and Innovation, Hackensack

School of Medicine, 111 Ideation Way, Nutley, New

Jersey 07110, United States; orcid.org/0000-0001-

9470-5009

Sean Ekins − Collaborations in Chemistry, Fuquay-Varina,

North Carolina 27526, United States; orcid.org/0000-

Authors

Jimmy S. Patel − Department of Pharmacology, Physiology,

and Neuroscience, Rutgers University − New Jersey Medical

School, Newark, New Jersey 07103, United States

Javiera Norambuena − Department of Biochemistry and

Microbiology, Rutgers, The State University of New Jersey,

New Brunswick, New Jersey 08901, United States

Hassan Al-Tameemi − Department of Biochemistry and

Microbiology, Rutgers, The State University of New Jersey,

New Brunswick, New Jersey 08901, United States

Yong-Mo Ahn − Department of Pharmacology, Physiology,

and Neuroscience, Rutgers University − New Jersey Medical

School, Newark, New Jersey 07103, United States

Alexander L. Perryman − Department of Pharmacology,

Physiology, and Neuroscience, Rutgers University − New

Jersey Medical School, Newark, New Jersey 07103, United

States; Present Address: A.L.P.: Repare Therapeutics,

002-5691-5790

Pradeep Kumar − Department of Medicine, Center for

Emerging and Re-emerging Pathogens, Rutgers University −

New Jersey Medical School, Newark, New Jersey 07103,

United States

Nancy Connell − Department of Medicine, Center for

Emerging and Re-emerging Pathogens, Rutgers University −

New Jersey Medical School, Newark, New Jersey 07103,

United States

Complete contact information is available at:

https://pubs.acs.org/10.1021/acsinfecdis.1c00265

Author Contributions

J.S.F. conceived of the project. J.S.P., A.L.P., and SE conducted

the Bayesian modeling. J.S.P., R.R., and J.O. carried out the in

vitro biological studies which N.C. helped analyze. J.S.P. and

210 Rue Frederick-Banting, Suite 100, Montreal, QC,

Canada, H4S 2A1

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

(6) Barrett, F. F., McGehee, R. F., Jr., and Finland, M. (1968)

Article

S.S.D. executed the synthetic chemistry. J.S.P., J.N., and Y.-

M.A. performed the IBDM studies. J.S.P, H.A.-T., and J.M.B.

and J.N. designed, conducted, and described the drug

resistance and genetic studies. M.D.Z. and H.-P.H. were

responsible for the mouse pharmacokinetic studies which were

analyzed by V.D.. S.P. conducted the mouse infection model

studies and the resulting data were analyzed by S.P., D.S.P.,

J.S.P., X.W., J.S.F., J.M.B., and J.N. analyzed the generated data

and wrote the manuscript. All authors contributed to editing

the manuscript.

(5) Lowy, F. D. (1998) Staphylococcus aureus infections. N. Engl. J.

Med. 339, 520−32.

Methicillin-resistant Staphylococcus aureus at Boston City Hospital.

Bacteriologic and epidemiologic observations. N. Engl. J. Med. 279,

(

41−8.

7) Matsuhashi, M., Song, M. D., Ishino, F., Wachi, M., Doi, M.,

Inoue, M., Ubukata, K., Yamashita, N., and Konno, M. (1986)

Molecular cloning of the gene of a penicillin-binding protein supposed

to cause high resistance to beta-lactam antibiotics in Staphylococcus

aureus. J. Bacteriol. 167, 975−80.

(8) National Nosocomial Infections Surveillance. (2004) National

Notes

Nosocomial Infections Surveillance (NNIS) System Report, data

summary from January 1992 through June 2004, issued October

The authors declare the following competing ﬁnancial

interest(s): S.E. is the CEO of Collaborations Pharmaceuticals

that is exploring the use of machine learning techniques to

discover, optimize, and commercialize novel anti-infective

agents.

(

004. Am. J. Infect. Control 32 (8), 470−485.

9) Hidron, A. I., Edwards, J. R., Patel, J., Horan, T. C., Sievert, D.

M., Pollock, D. A., Fridkin, S. K., National Healthcare Safety Network

Team, and Participating National Healthcare Safety Network

Facilities (2008) Antimicrobial-Resistant Pathogens Associated With

Healthcare-Associated Infections: Annual Summary of Data Reported

to the National Healthcare Safety Network at the Centers for Disease

Control and Prevention, 2006 −2007. Infect.Control Hosp. Epidemiol.

ACKNOWLEDGMENTS

■

This work was supported at Rutgers University − New Jersey

Medical School by the US National Institutes of Health grants

U19AI109713 (J.S.F.) and T32AI125185 (J.S.P.). The Boyd

lab was supported by NIAID grant R01AI139100-01 and

multistate research project project NE1028 from the USDA.

We thank Professor Barry Kreiswirth (Rutgers/PHRI) for

providing the VISA and VRSA strains. Dr. John Eng

(

9, 996−1011.

10) Weiner, L. M., Webb, A. K., Limbago, B., Dudeck, M. A., Patel,

J., Kallen, A. J., Edwards, J. R., and Sievert, D. M. (2016)

Antimicrobial-Resistant Pathogens Associated With Healthcare-

Associated Infections: Summary of Data Reported to the National

Healthcare Safety Network at the Centers for Disease Control and

Prevention, 2011 −2014. Infect.Control Hosp. Epidemiol. 37, 1288−

(

Princeton University) is thanked for guidance with the

11) Howden, B. P. (2005) Recognition and management of

301.

infections caused by vancomycin-intermediate Staphylococcus aureus

VISA) and heterogenous VISA (hVISA. Intern. Med. J. 35 (Suppl 2),

HRMS studies. We thank Dr. Ehud Zelzion (Rutgers

University) for genomic SNP analyses.

Y., Hori, S., Fukuchi, Y., and Kobayashi, I. (1997) Dissemination in

ABBREVIATIONS

■

S136−S140.

IBDM, intrabacterial drug metabolism; MRSA, methicillin-

resistant S. aureus; VISA, vancomycin-intermediate-resistant S.

aureus; VRSA, vancomycin-resistant S. aureus; MIC, minimum

concentration of compound to inhibit bacterial growth in vitro

by 90%; PAINS, Pan Assay Interference compounds; FCFP6,

functional class ﬁngerprints of maximum diameter 6; ROC,

Receiver Operator Characteristic; MSSA, methicillin-sensitive

(12) Centers for Disease Control and Prevention (1997) Update:

Staphylococcus aureus with Reduced Susceptibility to Vancomycin –

United States, 1997. MMWR Morb. Mortal. Wkly. Rep. 46, 813−815.

(13) Hiramatsu, K., Aritaka, N., Hanaki, H., Kawasaki, S., Hosoda,

Japanese hospitals of strains of Staphylococcus aureus heteroge-

neously resistant to vancomycin. Lancet 350, 1670−3.

(14) Kurosu, M., Siricilla, S., and Mitachi, K. (2013) Advances in

S. aureus; CC , minimum concentration of compound

MRSA drug discovery: where are we and where do we need to be?

Expert Opin. Drug Discovery 8, 1095−116.

(15) UpToDate: Evidence-based Clinical Decision Support. https://

www.uptodate.com.

resulting in 50% growth inhibition of this model mammalian

cell line; MBC, minimum bactericidal concentration; CFU,

colony-forming units; PK, pharmacokinetic; LC-MS, liquid

chromatography−mass spectrometry; HRMS, high-resolution

(16) Ekins, S., Perryman, A. L., Clark, A. M., Reynolds, R. C., and

Freundlich, J. S. (2016) Machine Learning Model Analysis and Data

Visualization with Small Molecules Tested in a Mouse Model of

Mycobacterium tuberculosis Infection (2014 −2015). J. Chem. Inf.

Model. 56, 1332−43.

mass spectrometry; DiBAC , bis(1,3-dibutylbarbituric acid)-

trimethine oxonol; DAF-FM, 4-amino-5-methylamino-2′,7′-

diﬂuoroﬂuorescein; qPCR, quantitative real-time PCR; SNP,

single nucleotide polymorphism/s.

18) Ekins, S., and Freundlich, J. S. (2011) Validating new

17) Ekins, S., and Freundlich, J. S. (2013) Computational models

for tuberculosis drug discovery. Methods Mol. Biol. 993 , 245 −62.

REFERENCES

tuberculosis computational models with public whole cell screening

■

(

1) Wertheim, H. F., Melles, D. C., Vos, M. C., van Leeuwen, W.,

van Belkum, A., Verbrugh, H. A., and Nouwen, J. L. (2005) The role

aerobic activity datasets. Pharm. Res. 28, 1859−69.

(19) Ekins, S., Reynolds, R. C., Kim, H., Koo, M. S., Ekonomidis, M.,

Talaue, M., Paget, S. D., Woolhiser, L. K., Lenaerts, A. J., Bunin, B. A.,

Connell, N., and Freundlich, J. S. (2013) Bayesian models leveraging

bioactivity and cytotoxicity information for drug discovery. Chem.

Biol. 20, 370−8.

(20) Pereira, J. C., Daher, S. S., Zorn, K. M., Sherwood, M., Russo,

R., Perryman, A. L., Wang, X., Freundlich, M. J., Ekins, S., and

Freundlich, J. S. (2020) Machine Learning Platform to Discover

Novel Growth Inhibitors of Neisseria gonorrhoeae. Pharm. Res. 37,

141.

of nasal carriage in Staphylococcus aureus infections. Lancet Infect. Dis.

(

, 751−62.

2) Tong, S. Y., Davis, J. S., Eichenberger, E., Holland, T. L., and

Fowler, V. G., Jr. (2015) Staphylococcus aureus infections:

epidemiology, pathophysiology, clinical manifestations, and manage-

ment. Clin. Microbiol. Rev. 28, 603−61.

(3) Moran, G. J., Krishnadasan, A., Gorwitz, R. J., Fosheim, G. E.,

McDougal, L. K., Carey, R. B., Talan, D. A., and for the EMERGEncy

ID Net Study Group (2006) Methicillin-resistant S. aureus infections

among patients in the emergency department. N. Engl. J. Med. 355,

66−674.

4) Talan, D. A., and Singer, A. J. (2014) Management of skin

abscesses. N. Engl. J. Med. 370, 2245−2246.

(21) Wang, X., Inoyama, D., Russo, R., Li, S. G., Jadhav, R., Stratton,

T. P., Mittal, N., Bilotta, J. A., Singleton, E., Kim, T., Paget, S. D.,

Pottorf, R. S., Ahn, Y. M., Davila-Pagan, A., Kandasamy, S., Grady, C.,

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Dartois, V., Ekins, S., Connell, N., Kumar, P., and Freundlich, J. S.

Article

Hussain, S., Soteropoulos, P., Zimmerman, M. D., Ho, H. P., Park, S.,

(38) Sisinni, L., Cendron, L., Favaro, G., and Zanotti, G. (2010)

Helicobacter pylori acidic stress response factor HP1286 is a YceI

homolog with new binding specificity. FEBS J. 277 , 1896 −905.

(39) Luong, T. T., and Lee, C. Y. (2007) Improved single-copy

integration vectors for Staphylococcus aureus. J. Microbiol. Methods

70, 186−90.

2020) Antitubercular Triazines: Optimization and Intrabacterial

Metabolism. Cell Chem. Biol. 27, 172−185.

22) Wang, X., Perryman, A. L., Li, S. G., Paget, S. D., Stratton, T. P.,

Lemenze, A., Olson, A. J., Ekins, S., Kumar, P., and Freundlich, J. S.

2019) Intrabacterial Metabolism Obscures the Successful Prediction

of an InhA Inhibitor of Mycobacterium tuberculosis. ACS Infect. Dis.

, 2148−2163.

23) Mendgen, T., Steuer, C., and Klein, C. D. (2012) Privileged

(

(40) Luong, T. T., and Lee, C. Y. (2006) The arl locus positively

regulates Staphylococcus aureus type 5 capsule via an mgrA-

dependent pathway. Microbiology 152, 3123−31.

J., Bose, J. L., and Bayles, K. W. (2013) A genetic resource for rapid

(41) Fey, P. D., Endres, J. L., Yajjala, V. K., Widhelm, T. J., Boissy, R.

and comprehensive phenotype screening of nonessential Staph-

ylococcus aureus genes. mBio 4, e00537-12.

scaffolds or promiscuous binders: a comparative study on rhodanines

(

43−53.

(42) Forsyth, R. A., Haselbeck, R. J., Ohlsen, K. L., Yamamoto, R. T.,

24) Perryman, A. L., Patel, J. S., Russo, R., Singleton, E., Connell,

Xu, H., Trawick, J. D., Wall, D., Wang, L., Brown-Driver, V., Froelich,

J. M., C, K. G., King, P., McCarthy, M., Malone, C., Misiner, B.,

Robbins, D., Tan, Z., Zhu Zy, Z. Y., Carr, G., Mosca, D. A., Zamudio,

C., Foulkes, J. G., and Zyskind, J. W. (2002) A genome-wide strategy

for the identification of essential genes in Staphylococcus aureus. Mol.

Microbiol. 43, 1387−400.

N., Ekins, S., and Freundlich, J. S. (2018) Naive Bayesian Models for

Vero Cell Cytotoxicity. Pharm. Res. 35, 170.

(25) Stratton, T. P., Perryman, A. L., Vilcheze, C., Russo, R., Li, S.

G., Patel, J. S., Singleton, E., Ekins, S., Connell, N., Jacobs, W. R., Jr.,

and Freundlich, J. S. (2017) Addressing the Metabolic Stability of

Antituberculars through Machine Learning. ACS Med. Chem. Lett. 8,

(

43) Mercier, C., Chalansonnet, V., Orenga, S., and Gilbert, C.

2013) Characteristics of major Escherichia coli reductases involved

in aerobic nitro and azo reduction. J. Appl. Microbiol. 115, 1012−22.

44) Nakanishi, M., Yatome, C., Ishida, N., and Kitade, Y. (2001)

(

099−1104.

26) Perryman, A. L., Stratton, T. P., Ekins, S., and Freundlich, J. S.

2016) Predicting Mouse Liver Microsomal Stability with ″Pruned ″

Machine Learning Models and Public Data. Pharm. Res. 33, 433−49.

27) Inoyama, D., Paget, S. D., Russo, R., Kandasamy, S., Kumar, P.,

Putative ACP phosphodiesterase gene (acpD) encodes an azor-

(

eductase. J. Biol. Chem. 276, 46394−9.

Singleton, E., Occi, J., Tuckman, M., Zimmerman, M. D., Ho, H. P.,

Perryman, A. L., Dartois, V., Connell, N., and Freundlich, J. S. (2018)

Novel Pyrimidines as Antitubercular Agents. Antimicrob. Agents

Chemother. 62, e02063−17.

(45) Wang, C. J., Hagemeier, C., Rahman, N., Lowe, E., Noble, M.,

Coughtrie, M., Sim, E., and Westwood, I. (2007) Molecular cloning,

characterisation and ligand-bound structure of an azoreductase from

Pseudomonas aeruginosa. J. Mol. Biol. 373, 1213−28.

(28) Suller, M. T., and Lloyd, D. (1998) Flow cytometric assessment

(46) Romero, E., Savino, S., Fraaije, M. W., and Loncar, N. (2020)

of the postantibiotic effect of methicillin on Staphylococcus aureus.

Mechanistic and Crystallographic Studies of Azoreductase AzoA from

Antimicrob. Agents Chemother. 42, 1195−9.

Bacillus wakoensis A01. ACS Chem. Biol. 15, 504−512.

(29) Marks, L. R., Clementi, E. A., and Hakansson, A. P. (2013)

(47) Liu, G., Zhou, J., Lv, H., Xiang, X., Wang, J., Zhou, M., and Qv,

Sensitization of Staphylococcus aureus to methicillin and other

antibiotics in vitro and in vivo in the presence of HAMLET. PLoS One

Y. (2007) Azoreductase from Rhodobacter sphaeroides AS1.1737 is a

flavodoxin that also functions as nitroreductase and flavin

mononucleotide reductase. Appl. Microbiol. Biotechnol. 76, 1271−9.

(

, e63158.

30) Singh, R., Manjunatha, U., Boshoff, H. I., Ha, Y. H.,

(48) Wang, L., Le, X., Li, L., Ju, Y., Lin, Z., Gu, Q., and Xu, J. (2014)

Niyomrattanakit, P., Ledwidge, R., Dowd, C. S., Lee, I. Y., Kim, P.,

Zhang, L., Kang, S., Keller, T. H., Jiricek, J., and Barry, C. E., 3rd

Discovering new agents active against methicillin-resistant Staph-

ylococcus aureus with ligand-based approaches. J. Chem. Inf. Model.

2008) PA-824 kills nonreplicating Mycobacterium tuberculosis by

intracellular NO release. Science 322, 1392−5.

31) Maia, L. B., and Moura, J. J. (2014) How biology handles

nitrite. Chem. Rev. 114, 5273−357.

32) Roldan, M. D., Perez-Reinado, E., Castillo, F., and Moreno-

49) El-Halfawy, O. M., and Valvano, M. A. (2013) Chemical

4, 3186−97.

communication of antibiotic resistance by a highly resistant

subpopulation of bacterial cells. PLoS One 8 (7), e68874.

(

(50) Chen, H., Hopper, S. L., and Cerniglia, C. E. (2005)

Vivian, C. (2008) Reduction of polynitroaromatic compounds: the

Biochemical and molecular characterization of an azoreductase from

Staphylococcus aureus, a tetrameric NADPH-dependent flavoprotein.

Microbiology (London, U. K.) 151, 1433−1441.

bacterial nitroreductases. FEMS Microbiol. Rev. 32, 474−500.

(33) Lewis, A. M., Matzdorf, S. S., and Rice, K. C. (2016)

Fluorescent Detection of Intracellular Nitric Oxide in Staphylococcus

aureus. Bio. Protoc. 6, e1878.

(51) Dorries, K., Schlueter, R., and Lalk, M. (2014) Impact of

antibiotics with various target sites on the metabolome of Staph-

(34) Sasaki, S., Miura, T., Nishikawa, S., Yamada, K., Hirasue, M.,

ylococcus aureus. Antimicrob. Agents Chemother. 58, 7151−63.

and Nakane, A. (1998) Protective role of nitric oxide in Staph-

(52) Aros-Calt, S., Castelli, F. A., Lamourette, P., Gervasi, G., Junot,

ylococcus aureus infection in mice. Infect. Immun. 66 , 1017 −22.

C., Muller, B. H., and Fenaille, F. (2019) Metabolomic Investigation

of Staphylococcus aureus Antibiotic Susceptibility by Liquid

Chromatography Coupled to High-Resolution Mass Spectrometry.

Methods Mol. Biol. 1871, 279−293.

(35) Ince, D., and Hooper, D. C. (2001) Mechanisms and frequency

of resistance to gatifloxacin in comparison to AM-1121 and

ciprofloxacin in Staphylococcus aureus. Antimicrob. Agents Chemother.

(

5, 2755−64.

(53) Liebeke, M., Meyer, H., Donat, S., Ohlsen, K., and Lalk, M.

36) Huband, M. D., Bradford, P. A., Otterson, L. G., Basarab, G. S.,

(2010) A metabolomic view of Staphylococcus aureus and its ser/thr

Kutschke, A. C., Giacobbe, R. A., Patey, S. A., Alm, R. A., Johnstone,

M. R., Potter, M. E., Miller, P. F., and Mueller, J. P. (2015) In vitro

antibacterial activity of AZD0914, a new spiropyrimidinetrione DNA

gyrase/topoisomerase inhibitor with potent activity against Gram-

positive, fastidious Gram-Negative, and atypical bacteria. Antimicrob.

Agents Chemother. 59, 467−74.

kinase and phosphatase deletion mutants: involvement in cell wall

biosynthesis. Chem. Biol. 17, 820−30.

(54) Meyer, H., Liebeke, M., and Lalk, M. (2010) A protocol for the

investigation of the intracellular Staphylococcus aureus metabolome.

Anal. Biochem. 401, 250−9.

(55) Hussein, M., Karas, J. A., Schneider-Futschik, E. K., Chen, F.,

Swarbrick, J., Paulin, O. K. A., Hoyer, D., Baker, M., Zhu, Y., Li, J., and

Velkov, T. (2020) The Killing Mechanism of Teixobactin against

Methicillin-Resistant Staphylococcus aureus: an Untargeted Metab-

olomics Study. mSystems 5, e00077−20.

(

37) El-Halfawy, O. M., Klett, J., Ingram, R. J., Loutet, S. A., Murphy,

M. E., Martin-Santamaria, S., and Valvano, M. A. (2017) Antibiotic

Capture by Bacterial Lipocalins Uncovers an Extracellular Mechanism

of Intrinsic Antibiotic Resistance. mBio 8, e00225-17.

ACS Infect. Dis. XXXX, XXX, XXX−XXX

ACS Infectious Diseases

Pilch, D. S. (2015) TXA709, an FtsZ-Targeting Benzamide Prodrug

Article

(56) Kaul, M., Mark, L., Zhang, Y., Parhi, A. K., Lyu, Y. L., Pawlak, J.,

Saravolatz, S., Saravolatz, L. D., Weinstein, M. P., LaVoie, E. J., and

with Improved Pharmacokinetics and Enhanced In Vivo Efficacy

against Methicillin-Resistant Staphylococcus aureus. Antimicrob.

Agents Chemother. 59, 4845−55.

(57) Brown, S., Zhang, Y. H., and Walker, S. (2008) A revised

pathway proposed for Staphylococcus aureus wall teichoic acid

biosynthesis based on in vitro reconstitution of the intracellular steps.

Chem. Biol. 15, 12−21.

(58) Farha, M. A., MacNair, C. R., Carfrae, L. A., El Zahed, S. S.,

Ellis, M. J., Tran, H. R., McArthur, A. G., and Brown, E. D. (2020)

Overcoming Acquired and Native Macrolide Resistance with

Bicarbonate. ACS Infect. Dis. 6, 2709−2718.

(

59) Boles, B. R., Thoendel, M., Roth, A. J., and Horswill, A. R.

2010) Identification of genes involved in polysaccharide-independ-

ent Staphylococcus aureus biofilm formation. PLoS One 5, e10146.

60) Zingle, C., Tritsch, D., Grosdemange-Billiard, C., and Rohmer,

(

M. (2014) Catechol-rhodanine derivatives: Specific and promiscuous

inhibitors of Escherichia coli deoxyxylulose phosphate reductoisomer-

ase (DXR). Bioorg. Med. Chem. 22, 3713−9.

(61) Clark, A. M., Sarker, M., and Ekins, S. (2014) New target

prediction and visualization tools incorporating open source

molecular fingerprints for TB Mobile 2.0. J. Cheminf. 6, 38.

(62) Clark, A. M., Dole, K., Coulon-Spektor, A., McNutt, A., Grass,

G., Freundlich, J. S., Reynolds, R. C., and Ekins, S. (2015) Open

Source Bayesian Models. 1. Application to ADME/Tox and Drug

Discovery Datasets. J. Chem. Inf. Model. 55 , 1231 −45.

(63) Novick, R. P. (1991) Genetic systems in staphylococci. Methods

Enzymol. 204, 587−636.

(64) Kreiswirth, B. N., Lofdahl, S., Betley, M. J., O’Reilly, M.,

Schlievert, P. M., Bergdoll, M. S., and Novick, R. P. (1983) The toxic

shock syndrome exotoxin structural gene is not detectably transmitted

by a prophage. Nature 305, 709−12.

(65) Mashruwala, A. A., Roberts, C. A., Bhatt, S., May, K. L., Carroll,

R. K., Shaw, L. N., and Boyd, J. M. (2016) Staphylococcus aureus

SufT: an essential iron-sulphur cluster assembly factor in cells

experiencing a high-demand for lipoic acid. Mol. Microbiol. 102,

1099−1119.