DOI: 10.1126/scitranslmed.3009815, 259ra146 (2014);6 Sci Transl Med

et al.Edward A. WeinsteinF-fluorodeoxysorbitol positron emission tomography

18Imaging Enterobacteriaceae infection in vivo with

Editor's Summary

treatment decisions.could provide clinicians with a sensitive and specific means of rapidly assessing infection and making−−tools

two common clinical−−F-labeled probes18time-consuming cultures. Whole-body imaging of bacteria using PET and Such imaging would allow noninvasive monitoring of disease progression or regression, without repeated

E. coli.FDS was used to monitor antimicrobial efficacy in animals infected with drug-susceptible or drug-resistant .Staphylococcus aureusinfection from general inflammation as well as from infection with the Gram-positive bacteria

E. coliadministered it to mice for positron emission tomography (PET) imaging. The probe was able to distinguish F-fluorodeoxysorbitol (FDS) and18could be used to image bacterial infections in vivo. The authors created

readily metabolize the sugar sorbitol; thus, Weinstein and colleagues reasoned that a radiolabeled version of sorbitol ,Klebsiella pneumoniae and Escherichia colithreat to human health. Pathogenic Enterobacteriaceae, including

Often acquired in hospitals and notoriously resistant to many drugs, Enterobacteriaceae represent an infectious

Probing Bacterial infections with PET Imaging

http://stm.sciencemag.org/content/6/259/259ra146.full.htmlcan be found at:

and other services, including high-resolution figures,A complete electronic version of this article

http://stm.sciencemag.org/content/suppl/2014/10/20/6.259.259ra146.DC1.html can be found in the online version of this article at: Supplementary Material

http://www.sciencemag.org/about/permissions.dtl in whole or in part can be found at: article

permission to reproduce this of this article or about obtaining reprintsInformation about obtaining

is a registered trademark of AAAS. Science Translational Medicinerights reserved. The title NW, Washington, DC 20005. Copyright 2014 by the American Association for the Advancement of Science; alllast week in December, by the American Association for the Advancement of Science, 1200 New York Avenue

(print ISSN 1946-6234; online ISSN 1946-6242) is published weekly, except theScience Translational Medicine

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

IMAG ING

Imaging Enterobacteriaceae infection in vivo with18F-fluorodeoxysorbitol positron emission tomographyEdward A. Weinstein,1,2* Alvaro A. Ordonez,1,3* Vincent P. DeMarco,3 Allison M. Murawski,3

Supriya Pokkali,3 Elizabeth M. MacDonald,1,3 Mariah Klunk,1,3 Ronnie C. Mease,4

Martin G. Pomper,1,4 Sanjay K. Jain1,3†

The Enterobacteriaceae are a family of rod-shaped Gram-negative bacteria that normally inhabit the gastrointestinaltract and are the most common cause of Gram-negative bacterial infections in humans. In addition to causing se-rious multidrug-resistant, hospital-acquired infections, a number of Enterobacteriaceae species are also recognizedas biothreat pathogens. As a consequence, new tools are urgently needed to specifically identify and localize in-fections due to Enterobacteriaceae and to monitor antimicrobial efficacy. In this report, we used commercially avail-able 2-[18F]-fluorodeoxyglucose (18F-FDG) to produce 2-[18F]-fluorodeoxysorbitol (18F-FDS), a radioactive probe forEnterobacteriaceae, in 30 min. 18F-FDS selectively accumulated in Enterobacteriaceae, but not in Gram-positivebacteria or healthy mammalian or cancer cells in vitro. In a murine myositis model, 18F-FDS positron emission to-mography (PET) rapidly differentiated true infection from sterile inflammation with a limit of detection of 6.2 ±0.2 log10 colony-forming units (CFU) for Escherichia coli. Our findings were extended to models of mixed Gram-positive and Gram-negative thigh co-infections, brain infection, Klebsiella pneumonia, and mice undergoing immu-nosuppressive chemotherapy. This technique rapidly and specifically localized infections due to Enterobacteriaceae,providing a three-dimensional holistic view within the animal. Last, 18F-FDS PET monitored the efficacy of antimi-crobial treatment, demonstrating a PET signal proportionate to the bacterial burden. Therapeutic failures associatedwith multidrug-resistant, extended-spectrum b-lactamase (ESBL)–producing E. coli infections were detected in realtime. Together, these data show that 18F-FDS is a candidate imaging probe for translation to human clinical cases ofknown or suspected infections owing to Enterobacteriaceae.

INTRODUCTIONThe Enterobacteriaceae are the most common cause of Gram-negativebacterial infections in humans and include prominent pathogens suchas Yersinia spp., Escherichia coli, Klebsiella spp., and Enterobacter spp.Several Enterobacteriaceae species are recognized as biothreat patho-gens by the U.S. Centers for Disease Control and Prevention (CDC),and also are a cause of serious multidrug-resistant (MDR) hospital-acquired infections (1). For example, E. coli, a prominent pathogenwithin this family, is the most commonly isolated bacteria in clinicallaboratories (2) and is a notorious source of life-threatening MDR in-fections in cancer patients.

When an infection is suspected, but occurs at an unknown locationor deep within the body, noninvasive anatomic imaging techniques areused to identify a site of interest for biopsy and microbial culture. Themost common approaches are computed tomography (CT), magneticresonance imaging (MRI), and nuclear medicine techniques such as111In-oxine–labeled white blood cell imaging and 2-[18F]-fluorodeoxyglucose(18F-FDG) positron emission tomography (PET). Each modality pro-vides limited clinical information. CT and MRI both reveal structuralabnormalities that are nonspecific and often occur late in a disease pro-cess. Labeled leukocytes, generated through a multistep process, producenonspecific signals and may have limited penetration into diseased tis-

sues (3). 18F-FDG PET has emerged as a highly sensitive imagingmethod, but lacks specificity and cannot differentiate among oncologic,inflammatory, or infectious processes. Moreover, these imaging tech-niques are dependent on host inflammatory responses to infection, whichmay be reduced or absent in immunosuppressed patients (for example,cancer chemotherapy, HIV/AIDS, or organ transplant) who are mostat risk for infection. Follow-up microbiological or molecular testing isstill required after performing any of these imaging techniques to es-tablish a diagnosis, resulting in a prolonged, uncertain process that takesseveral days. Last, current technologies fail to provide rapid feedbackabout the effect, or adequacy, of a selected antimicrobial regimen. There-fore, there is a need for rapid, whole-patient imaging techniques thatcould localize a pathogen with specificity and provide a quantitativereadout of disease burden in response to treatment.

In clinical microbiology, pathogenic Enterobacteriaceae are differ-entiated from other organisms by selective metabolism. Sorbitol is ametabolic substrate for Enterobacteriaceae, but is more commonly knownas a “sugar-free” sweetener. We hypothesized that a positron-emittinganalog of sorbitol, 2-[18F]-fluorodeoxysorbitol (18F-FDS), first describedfor tumor imaging (4), would be a suitable probe to selectively labeland tomographically image these bacteria in vivo. Using the methodsdescribed in (4), we derived 18F-FDS from commercial 18F-FDG to con-fer selectivity for Enterobacteriaceae for use as a diagnostic tool withbroad utility. We describe our development of the radioprobe inseveral preclinical models of infection by Enterobacteriaceae anddemonstrate that serial imaging over time can provide predictiveinformation on the success of a chosen antimicrobial treatment reg-imen. The pathogen-specific mechanism of 18F-FDS imaging could beused to identify and monitor known or suspected infections caused byEnterobacteriaceae.

1Center for Infection and Inflammation Imaging Research, Johns Hopkins UniversitySchool of Medicine, Baltimore, MD 21287, USA. 2Department of Medicine, Johns HopkinsUniversity School of Medicine, Baltimore, MD 21287, USA. 3Department of Pediatrics,Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA. 4Russell H. MorganDepartment of Radiology and Radiological Science, Johns Hopkins University School ofMedicine, Baltimore, MD 21287, USA.*These authors contributed equally to this work.†Corresponding author. E-mail: sjain5@jhmi.edu

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 1

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

RESULTS

18F-FDS is taken up rapidly and specifically by a rangeof EnterobacteriaceaeCultures of E. coli and Klebsiella pneumoniae were selected to assess18F-FDS accumulation in vitro. These two Gram-negative enteric speciesreadily incorporated 18F-FDS (Fig. 1A). Staphylococcus aureus, a Gram-positive bacterium, did not accumulate 18F-FDS but did accumulate 18F-FDG.Heat-killed bacteria did not incorporate either probe. E. coli cultures werealso co-incubated with 18F-FDS and increasing concentrations of unlabeled

sorbitol (Fig. 1B). 18F-FDS uptake was outcompeted by concentrations ofsorbitol above 40 mg/ml, suggesting that uptake does reach a point ofsaturation, presumably in a transporter-driven process.

The presence of the sorbitol-6-phosphate dehydrogenase (srlD)gene cassette responsible for sorbitol metabolism has been noted with-in the annotated genome of E. coli (5, 6). To predict the range of bacte-ria capable of 18F-FDS uptake and likely detection by PET, the srlD genewas used to query the UniProtKB database of genome-sequenced bac-terial species (Fig. 1C). Representative bacteria from that panel weretested to assess 18F-FDS uptake. Members of the Enterobacteriaceae

Fig. 1. In vitro uptake of 18F-FDS in bacterial pathogens and mamma-lian cell lines. (A) 18F-FDS uptake in E. coli (ATCC 25922), K. pneumoniae(ATCC 35657), and S. aureus (ATCC 29213) cultures incubated with 18F-FDS or18F-FDG. Heat-killed bacteria were negative controls. P values compared torespective heat-killed controls were determined by two-tailed Student’s ttest. (B) Competition of 18F-FDS uptake with increasing concentrations ofunlabeled (free) sorbitol in E. coli (normalized to E. coli uptake withoutsorbitol). (C) Homology of E. coli (ATCC 25922) sorbitol-6-phosphate de-hydrogenase gene, srlD, to other pathogenic bacteria in the UniProtKBdatabase. (D) Uptake of 18F-FDS in ATCC reference Gram-negative andGram-positive bacterial strains incubated for 120 min. (E) J774 (murinemacrophage), WEHI 164 (murine fibroblast), U87MG (human glioblastoma),and NCI-H460 (human large cell lung carcinoma) cultures were incubatedwith 18F-FDS or 18F-FDG. Uptake was measured at 120 min and is shown incomparison to the reference E. coli. Data in (A), (B), (D), and (E) are means ±SEM (n = 6).

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 2

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

family accumulated 18F-FDS, whereas Gram-positive bacteria, such asEnterococcus, Staphylococcus, and Streptococcus spp., and the aerobicGram-negative rod Pseudomonas aeruginosa did not accumulate theprobe (Fig. 1D). Additionally, mammalian cells and tumor cell linesdid not accumulate 18F-FDS, with an uptake almost 1000-fold higherin E. coli than in mammalian cells (Fig. 1E). Together, these data dem-onstrate the selectivity of 18F-FDS.

18F-FDS PET can rapidly differentiate infection sites fromsterile inflammationWe next investigated whether 18F-FDS PET could distinguish E. coliinfection from sterile inflammation in vivo. Mice were inoculatedwith live E. coli into the right thigh and a 10-fold higher burdenof heat-killed E. coli into the left thigh (sterile inflammation). 18F-FDSreadily concentrated in the infected right thigh, but not in the sterile,

Fig. 2. PET/CT imaging of E. colimyositisin immunocompetent mice. (A) 18F-FDSsignal is noted in the infected (yellow arrow)but not in the inflamed (control) sterile thigh(red arrow). 18F-FDG signal was noted inboth infected and inflamed thighs. H, heart;I, intestine; Bl, bladder. (B and C) 18F-FDS (B)and 18F-FDG (C) PET signals in vivo from theinfected (right) and inflamed (left) thighsand postmortem biodistribution in all or-gans. Data are medians with interquartileand ranges shown (n = 8 animals for 18F-FDSand n = 9 for 18F-FDG). P values were deter-mined by two-tailed Mann-Whitney U test.(D and E) Hematoxylin and eosin (H&E)and Gram stains with inflammatory cellsand Gram-negative bacteria in the E. coli–infected (right) thigh (D) and inflammatorycells without bacteria in the inflamed (left)thigh (E). Images are representative of fiveanimals.

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 3

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

inflamed left thigh (Fig. 2A). Reference imaging with 18F-FDG could notqualitatively distinguish the infected right thigh from the sterile, in-flamed left thigh (Fig. 2A).

Next, to quantify PET signal intensity, we drew spherical regions ofinterest (ROIs) within the thighs on the basis of anatomical localizationfrom CT. 18F-FDS produced 7.3-fold higher signal intensity in the in-fected right thigh compared to the sterile, inflamed left thigh (Fig. 2B).In contrast, 18F-FDG did not produce significantly different signal in-tensities between the thighs (Fig. 2C). We then surgically resected tis-sue from the mice postmortem to confirm the PET findings with a gcounter. In agreement with the PET signal, there was 7.9-fold higherg emission from the infected right thigh than the sterile, inflamed leftthigh (Fig. 2B). Similar to in vivo data, 18F-FDG did not discriminatethe infection from the sterile inflammation in the thighs (Fig. 2C). Itshould also be noted that 18F-FDG counts from both the E. coli–infected(1.7-fold) and sterile, inflamed control thighs (1.7-fold) were significant-ly higher than the uninflamed, uninfected deltoid muscle (Fig. 2C).However, no differences in 18F-FDS counts were noted between the ster-ile, inflamed control thighs and the uninflamed deltoid, further con-firming that 18F-FDS was specific for infection and not taken up at sitesof sterile inflammation. Postmortem tissue histology of the thighs ver-ified the presence inflammatory cells and Gram-negative bacteria in theE. coli–infected right thigh (Fig. 2D) and inflammatory cells withoutbacteria, in the uninfected (but inflamed) left thigh (Fig. 2E). More ex-tensive tissue histology is given in fig. S1. A three-dimensional 18F-FDSPET/CT image of the E. coli–infected mouse is shown in movie S1.

We then extended our findings to a mouse model of infection duringimmunosuppression. In this model, mice were infected with E. coli duringthe course of cancer chemotherapy. The same experiments as in Fig. 2were performed while treating mice with a neutropenia-inducing cyclo-phosphamide regimen (days 1 and 4), as described in (7). The absenceof a normal immune response did notimpede detection of infection in the rightthigh, but not the inflamed left thigh, by18F-FDS (fig. S2). 18F-FDG was not ex-amined in comparison for this study.

18F-FDS was also evaluated in braininfection and tumor models to confirmspecificity for infection versus oncologicprocesses. 18F-FDS PET intensities inE. coli–infected brains, human glioblasto-ma xenografts (U87MG), and normalbrains were compared. The data demon-strated a significantly higher uptake in theE. coli–infected brain versus brain tumoror normal brain (Fig. 3A). 18F-FDS PETcould localize the E. coli brain infection,but no signal was seen for the normal brainor glioblastoma (Fig. 3B). 18F-FDS was alsoevaluated in a dual brain tumor and E. colithigh infection in the same animal, whichdemonstrated a significantly higher PET sig-nal intensity in E. coli–infected thighs com-pared to U87MG brain tumors (Fig. 3C).Finally, dynamic 18F-FDS PET imaging(over 120 min) in mice with U87MG braintumors was performed (fig. S3). Althoughthere was some initial uptake by the tumor

tissues, the PET signal dissipated after 60 min. By comparison, signalsfrom infection could be seen consistently even at 120 min (Fig. 3B).

18F-FDS is selective for Enterobacteriaceae in mixed infectionsImmunosuppressed mice were inoculated with live E. coli and S. aureusin the right and left thighs, respectively. Consistent with in vitrouptake data, 18F-FDS readily concentrated in the E. coli–infected rightthigh, but not in the S. aureus–infected left thigh (Fig. 4A). 18F-FDGPET could not distinguish the E. coli infection from the S. aureus in-fection (Fig. 4A). The 18F-FDS PET signal intensity and tissue biodis-tribution were both higher at the site of the E. coli infection (right thigh)compared with those at the site of the S. aureus infection (left thigh)(Fig. 4B). There was no significant difference between thighs for18F-FDG (Fig. 4C). Postmortem tissue histology of the thighs verifiedthe presence of bacteria in the lesions (Fig. 4D). At the time of imaging,there were 30-fold higher numbers of S. aureus [8.3 ± 0.3 log10 colony-forming units (CFU)] than E. coli (6.8 ± 0.7 log10 CFU) in the respectivethighs. Moreover, we could reliably detect as few as 6.2 ± 0.2 log10 CFUof E. coli (at the time of imaging) with 18F-FDS PET in one cohort ofthe dually infected myositis mice (Fig. 4E).

Enterobacteriaceae, and in particularK. pneumoniae, are an importantcause of MDR hospital-acquired pneumonias, which cannot be easily dis-tinguished from other bacteria (8–10). In a pulmonary infection model,18F-FDS readily localized to the areas of lung infiltration observed by CT(Fig. 5A). PET signal intensity and corresponding postmortem tissue bio-distribution were significantly higher at the site of the K. pneumoniae pul-monary infection than in the uninfected control lungs (Fig. 5, B and C).

18F-FDS can monitor antimicrobial efficacy in MDR infectionsEnterobacteriaceae are becoming increasingly resistant to many classesof antimicrobials, and consequent treatment failuresare of great concern

Fig. 3. 18F-FDS PET/CT imaging of human brain tumorxenograft versus E. coli infection. (A) 18F-FDS PET signalfrom an E. coli–infected brain compared with a human braintumor (U87MG) xenograft and normal brain. Data are me-dian SUV ratios normalized to the uninflamed thigh muscleof each animal, with interquartile and ranges shown (n = 6animals for brain infection, n = 6 for normal brain, and n = 3for brain tumor). (B) 18F-FDS PET signal in the E. coli–infectedbrain, normal brain, or tumor brain tissues. (C) 18F-FDS PETsignal from right-thigh E. coli infection compared with hu-man brain tumor (U87MG) xenograft and infected left thigh

in the same animals. Data are median SUVs with interquartile and ranges shown (n = 4 animals). P valuesin (A) and (C) were determined by two-tailed Mann-Whitney U test.

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 4

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

to public health (11–14). For example, mortality can approach ashigh as 60% in infections associated with drug-resistant extended-spectrum b-lactamase (ESBL)–producing Enterobacteriaceae (14).18F-FDS was evaluated against 15 random ESBL-producing clinicalstrains, which demonstrated substantial uptake in all isolates (Fig. 6A).For comparison to Gram-positive cellular uptake (at 0 Bq per 106 cells),see Fig. 1D.

Immunosuppressed mice were inoculated with drug-susceptibleor drug-resistant (ESBL-producing) E. coli into the right thigh anda 10-fold higher burden of the corresponding heat-killed E. coli straininto the left thigh. 18F-FDS PET imaging was performed before andafter the administration of ceftriaxone, a commonly used antimicro-bial that is effective against drug-susceptible E. coli but ineffective againstESBL producers. Corresponding to the bacterial CFU, 18F-FDS PETsignal disappeared with treatment in mice infected with drug-susceptibleE. coli, but persisted in mice infected with the drug-resistant E. coli(Fig. 6B). An increase in CFU measured immediately after the com-pletion of imaging corresponded with disease progression. The in-adequately treated mice clinically appeared sick, manifesting ruffledfur, hunched posture, inactivity, and increased respiratory rate, allconsistent with sepsis, and were euthanized per protocol. Rapid assess-ment of a therapeutic failure would be highly valuable for the man-agement of seriously ill patients.

DISCUSSION

The Enterobacteriaceae produce a rangeof human disease, from lethal fulminantpneumonias to deep-seated MDR infec-tions. When serious infections are suspected,patients are often treated empirically witha combination of broad-spectrum anti-microbials while awaiting culture resultsthat provide information on the bacterialclass and species causing the infection, aswell as drug susceptibilities. These tradi-tional diagnostic techniques can take sev-eral days and may not provide reliableinformation, because blood cultures couldhave high contamination rates, prompt-ing several groups to propose validationrules (15, 16). The combination of timeand effort required for proper antimicro-bial selection has become a barrier leadingto uncertainty and indiscriminate broad-spectrum antimicrobial use (17). Judicioususe could reduce the emergence of drugresistance, increase patient safety, andsave billions of dollars for the U.S. healthcare system alone (18). Consequently, thereis a need to diagnose infections quickly, in-cluding deep-seated, difficult-to-access le-sions due to Enterobacteriaceae.

For oncological applications, 18F-FDGPET can detect on the order of 5 to 6 log10cells (19), which is near the limit of detec-tion for CT imaging. However, detectionof infection by current imaging techniquesis variable and dependent on host responses,

as well as the site and type of infection. In our study, 18F-FDS PETimaging accurately identified the magnitude and location of infectionscaused by Enterobacteriaceae in vivo in mice. 18F-FDS produced astrong signal with minimal background noise, allowing for the detectionof as few as 6.2 ± 0.2 log10 CFU of E. coli as enumerated from the in-fected tissues after imaging. This is a promising threshold becauseclinically relevant bacterial infections have high bacterial burdens[8.3 log10 CFU/ml (20)] and can be as large as several centimeters indiameter with volumes of tens to hundreds of milliliters (21, 22). There-fore, the 18F-FDS limit of detection suggests that a positive signal wouldprovide an early indication of infection in the host, although the sensi-tivity and specificity of this technique in the clinical setting remain to bedetermined. We predict that our results with E. coli and K. pneumoniaecould be extended to detect other members of the Enterobacteriaceaefamily, such as Yersinia spp. or Salmonella spp., which demonstratesubstantial 18F-FDS uptake in vitro (Fig. 1D).

We did not use background subtraction in any of our images. Insome images, an extraneous signal was noted in the mouse intestine,which corresponded to the biliary excretion of 18F-FDS and subsequentuptake by gut flora. The source of this signal was variably placed de-pending on peristalsis of the stool pellet in the colon. Anatomic local-ization with CT would be able to distinguish an intraluminal PET signalfrom an extraluminal PET signal in abdominal infections. Additionally,

Fig. 4. PET/CT imaging of mixed myositis thigh in-fection. (A) 18F-FDS PET signal was observed in theE. coli–infected right thigh (yellow arrow), but not in theS. aureus–infected left thigh (red arrow). 18F-FDG signalwas noted in both E. coli– and S. aureus–infected thighs.U, extravasated urine. (B and C) 18F-FDS (B) and 18F-FDG(C) PET signals in vivo from the infected thighs andpost-mortem biodistribution. Data are medians with in-terquartile and ranges shown (n = 7 animals each for18F-FDS and 18F-FDG). P values were determined bytwo-tailed Mann-Whitney U test. (D) H&E and Gramstains demonstrating Gram-negative (right thigh) andGram-positive (left thigh) bacteria. Images are represent-ative of five animals. (E) Bacterial burden immediatelyafter completion of 18F-FDS PET imaging in each mouse(n = 3) in one of the dually infected imaging cohorts.

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 5

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

there were signals from the kidneys and bladder owing to the urinaryclearance of 18F-FDS. Elimination of a probe by hepatic or renal routeswould be expected for any radiopharmaceutical, and we believe thatthis would not be a major limitation to the clinical use of 18F-FDS. Bydelaying PET image acquisition until a patient urinates (or by use of aurinary catheter), kidney infections could potentially be visualized. Forexample, prostate tissue has been successfully imaged with renally clearedPET probes using this approach (23).

Li et al. (4) were the first to chemically reduce 18F-FDG into 18F-FDSfor the molecular imaging of brain tumors. We therefore used the samebrain tumor model and demonstrated that the E. coli thigh infectionproduced a significantly higher 18F-FDS PET intensity compared tothe signal from the U87MG human brain tumor (Fig. 3). To considerthe possibility of differential uptake pharmacokinetics between brainand peripheral tissues, we developed an E. coli brain infection modelthat confirmed higher 18F-FDS uptake in an E. coli infection than in abrain tumor. We note that this finding is consistent with our in vitrodata indicating a lack of 18F-FDS uptake by eukaryotic cells (Fig. 1E),as well as data presented in the study by Li et al. (4). There is no se-lective transporter for 18F-FDS entry into eukaryotic cells, and the substi-tution of the hydroxyl group by fluorine at the C-2 position completelyabrogates the recognition by mammalian sorbitol dehydrogenase (24).Moreover, comparisons by Li et al.were made in the context of low signalwith high background at earlier time points (5 to 60 min). We thereforeperformed dynamic 18F-FDS PET imaging over 120 min in mice withU87MG brain tumors and demonstrated initial uptake, which dissipatedover time (fig. S3). This suggests that the results presented by Li et al.may be consistent with a nonspecific blood pool effect, that is, capil-lary leak at the site of inflammation or tumor. Detection of brain in-fections, however, is a new finding for 18F-FDS and, to our knowledge, hasnot been demonstrated for other probes in development.

Although several promising optical im-aging probes have been reported to detectbacterial infections (25–28), these probesare not yet clinically translatable. Despitethe excellent sensitivity of optical imaging,the approach is limited by the absorptionof light within deep tissues and, therefore,applicable to imaging of small animals orsuperficial sites only. The sensitivity ofoptical imaging (pico- to femtomolar) issignificantly higher than that of PET orsingle-photon emission computed tomog-raphy (SPECT) (nano- to picomolar) (29).Therefore, human translation (optical toPET or SPECT) will lower sensitivity byseveral orders of magnitude. Moreover, forconversion to PET or SPECT, an opticalprobe will need to be labeled with a high-energy emitting isotope, which maychange its chemical properties or targetbinding characteristics. PET or SPECTimaging intrinsically takes advantage ofhigh-energy photons that can radiate outfrom the deepest structures within the body.

Several radiolabeled antibiotics andpeptides, such as 99mTc-ciprofloxacin,have been evaluated in clinical studies

for diagnosing infections (30, 31). However, bacteria-specific anti-biotics and peptides are designed to kill or disable bacteria at thelowest possible concentration; therefore, they may not be ideal for im-aging purposes owing to the lack of signal amplification from probeaccumulation by the bacteria. Although safe, 99mTc-ciprofloxacin de-monstrated variable specificity in clinical studies and an inability toreliably differentiate infection from sterile inflammatory processes(32–35). Radiolabeled peptides similarly demonstrated varying degreesof specificity in their ability to detect infection (36). S. aureus endocarditishas been detected in mice with labeled prothrombin, which binds tothe staphylococcal coagulase (37). However, this probe is not specificfor S. aureus, because it is a prothrombin analog, which has the po-tential for activation and false-positive signaling by noninfectious pro-cesses or by other types of bacteria. 124I-FIAU is an example of a PETprobe for bacterial infections (38) but is at least one to two orders ofmagnitude less sensitive than 18F-FDS. Moreover, the synthesis of124I-FIAU is challenging because it uses I-124 iodide, which couldlimit its widespread use.

Because antimicrobial selection is dependent on the bacterial class andspecies causing the infection (for example, vancomycin is used to treatinfections with Gram-positive bacteria but has no activity against Gram-negative organisms), a general infection imaging probe, although valu-able, may not provide the necessary information to streamline antimicro-bial selection. However, imaging probes that are selective for importantand common classes of bacterial pathogens could not only diagnose in-fections noninvasively but also rapidly identify the causative bacterialclass and help in the selection of appropriate antimicrobial treatment.

PET is becoming a routine clinical tool, particularly for oncologyand neurology, and an increasingly useful array of radioprobes, partic-ularly those radiolabeled with 18F, are proliferating in the United Statesand abroad. The pharmacokinetics of 18F-FDS in humans have yet to be

Fig. 5. K. pneumoniae lung infection (pneumonia). (A) 18F-FDS PET signal colocalized with K. pneumoniaelung infection noted on CT (yellow arrow). (B and C) 18F-FDS PET signal in vivo (B) and postmortem bio-distribution (C) in infected versus uninfected lungs. Data are medians with interquartile and ranges shown(n = 5 animals for the infected group and n = 6 for the uninfected group).

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 6

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

established, but the rapid clearance of low molecular weight, short half-life probes, such as 18F-FDS, makes them favorable compared to radio-labeled antibodies, which sometimes require a week to circulate and clearfrom nontarget sites. Typically, a PET scan is obtained 1 to 2 hours afterinjection of the radioprobe and can take 15 to 60 min depending on thescan area. 18F-FDS PET could be applied clinically to patients with sus-pected deep-seated infections due to Enterobacteriaceae, such as intra-abdominal or implant infections or ventilator-acquired pneumonias, andin monitoring infections with drug-resistant Enterobacteriaceae.

Immunosuppressed patients would benefit from noninvasive imag-ing because a surgical biopsy in itself risks introducing infection.Whole-body scans may be used in patients whose source of infectionis not known, whereas targeted scans could be used for monitoringknown site(s) of infection. Repeat imaging would be useful if clinicalsigns and symptoms suggest persistent infection despite antibacterialtherapy. CT and MRI cannot monitor acute responses (hours to days),because they reveal structural abnormalities and inflammation, which

do not subside rapidly. 18F-FDG PET cannot detect the effect of anti-biotics on acute infection (39), and paradoxically, signal intensitiesmay increase acutely owing to inflammation upon effective antimicro-bial treatments (40, 41). 18F-FDS PET signal intensities, in contrast,correlate with bacterial burden and would support a clinical decisionto change the drug regimen, or proceed to surgical intervention.

In summary, we have developed a diagnostic tool for imaging in-fections due to Enterobacteriaceae, the most common Gram-negativehuman bacterial pathogens and a frequent cause of serious drug-resistant hospital-acquired infections. 18F-FDS is inexpensive andcan be easily synthesized anywhere in the world where 18F-FDG,the most widely used PET probe, is available. Sorbitol, the radiodecayproduct, is already approved for use in humans by the U.S. Food andDrug Administration. 18F-FDS could therefore be translated to hu-mans and provide a rapid, noninvasive diagnostic tool to identifyand localize infections from Enterobacteriaceae, guide antimicrobialselection, and increase patient safety. In our study, 18F-FDS was iden-tified through a systematic discovery program. Future efforts shouldfocus on discovering class-specific bacterial imaging probes that woulddetect and differentiate a wide range of pathogenic bacteria.

MATERIALS AND METHODS

Study designThe objective of this study was to test 18F-FDS as a noninvasive im-aging probe and diagnostic tool that identifies and localizes infectionsdue to Enterobacteriaceae in vivo. All protocols were approved by theJohns Hopkins Biosafety, Radiation Safety, and Animal Care and UseCommittees. Measurement of radioactivity using PET/CT imagingand postmortem biodistribution studies allowed for the quantificationof 18F-FDS uptake by Enterobacteriaceae. CFU confirmed the presenceof bacteria within the specified organ. Mice were purchased from thesame breeding source, housed similarly before infection, and randomlyassigned to infected and control groups for experiments. PET/CT im-ages were acquired using the same parameters. The study was not blinded.All data from the experiments were included in the analyses and arepresented in this article.

Synthesis of 18F-FDSAll chemicals used in the study were purchased from commercial ven-dors and used without purification except where stated. 18F-FDS wasgenerated from commercially available 18F-FDG (PETNET SolutionsInc. or IBA Molecular) using the method described in (4). Briefly,18F-FDG was reduced with sodium borohydride at 35°C for 30 minbefore quenching with acetic acid and pH correction to 7.4 with sodiumbicarbonate. The product was then passed through an n-aluminaSep-Pak and 0.2-mm filter (Waters Corporation). Chemical purity ofthe final product, 18F-FDS, was confirmed by high-performance liquidchromatography. Both 1H and 13C nuclear magnetic resonance andmass spectrometry confirmed the structure of the final product.

18F-FDS uptake assaysBacterial strains purchased from the American Type Culture Collec-tion (ATCC) or 15 drug-resistant (ESBL-producing) E. coli strains(random consecutive samples) from the clinical microbiology labora-tory (Johns Hopkins Hospital) were aerobically grown to absorbanceat 600 nm of 1.0 in Lysogeny Broth (LB). The number of CFU was

Fig. 6. 18F-FDS PET/CT monitoring of antimicrobial efficacy. (A) In vitro18F-FDS uptake in 15 MDR E. coli (ESBL-producing) clinical strains. Uptakewas measured at 120 min and is shown in comparison to the referenceE. coli (positive control). Data are means ± SEM (n = 6). (B) Animals wereinoculated with drug-susceptible (n = 5) or drug-resistant (ESBL-producing;n = 6) E. coli and treated with ceftriaxone for 24 hours. Antimicrobial treat-ment efficacy was measured as CFU and PET signal intensity. Yellow arrowindicates infected right thigh. K, kidney. Data are means ± SD (CFU) andmedians and ranges (SUV ratio).

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 7

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

enumerated by dilution and plated onto solidified LB medium. Probeuptake assays were performed by incubating bacterial cultures with18F-FDS (20 kBq/ml) at 37°C with rapid agitation. As a control, heat-killed (90°C for 30 min) bacteria were similarly incubated with eachprobe. Bacteria were pelleted by centrifugation and washed three timeswith PBS, and the activity for each pellet was measured using an auto-mated g counter (1282 Compugamma CS Universal gamma counter,LKB Wallac). Six replicates were used for each assay. Counts for eachsample were corrected for background and normalized to CFU or totalprotein. Protein estimation was performed using a Bradford assay (Sigma).For in silico analysis, E. coli srlD sequence was queried against theUniProtKB database of genome-sequenced bacterial species. Alignmentand percentage identity were then calculated using Clustal Omega (42).

Animal models and infectionCBA/J mice (female, 6 to 8 weeks old) were used for all experimentsexcept those bearing xenografts (athymic nude, male, 10 to 12 weeks).A subset of CBA/J mice were immunosuppressed with cyclophospha-mide as described in (7). Mice were injected with different strains oflive or heat-inactivated (90°C for 30 min) bacteria. Thigh infectionswere allowed to develop for 10 and 6 hours in immunocompetent andimmunosuppressed mice, respectively. Target implantation for E. coliwas 7 log10 CFU for all infections, except where indicated. For the dual-infection myositis model, target implantation for S. aureuswas 7 to 8 log10CFU, with 10- to 100-fold lower dose used for E. coli. The pneumoniamodel, adapted from (43), was made by intratracheal installation ofK. pneumoniae (6.5 log10 CFU) into the trachea of sedated, immuno-suppressed mice with 5 days of incubation. Brain tumor (5 × 105 U87MGhuman glioblastoma cells) and infection (7 log10 CFU of E. coli in im-munosuppressed mice) models were established by inoculations into theright frontal lobe (1.5 mm anterior and 1.34 mm lateral to the bregma,3.5 mm into the parenchyma) through a 50-mm glass capillary needleas described in (44). Brain tumor size at the time of imaging was 30 to40 mm3 (based on ex vivo measurements).

PET/CT imagingProbe (7.4 MBq) was injected via tail vein 120 min (18F-FDS) or45 min (18F-FDG) before acquiring a 15-min static PET frame usingthe Mosaic HP (Philips) as described before (40). CT was performedwith NanoSPECT/CT (Bioscan). A urinary catheter was placed toempty the bladder during imaging. For quantitative analysis, one totwo spherical (3-mm-diameter) ROIs were drawn manually in thethighs of each animal, using CT as a guide. The standardized uptakevalues (SUVs) were computed using Amide version 1.0.4 (http://www.amide.sourceforge.net), and Amira version 5.4.2 (Visualization ScienceGroup) was used to visualize the images. After imaging, mice weresacrificed to collect tissues for direct g counting (biodistribution).Some cohorts were also used to determine histology and/or CFU atthe time of imaging. The biodistribution data are presented as percentinjected dose per gram of tissue (%ID/g). H&E and Gram stains wereperformed to visualize tissue histology and bacteria, respectively. Tis-sues were also homogenized in PBS, plated onto solidified LB me-dium, and incubated overnight at 37°C to enumerate the CFU.

Antimicrobial efficacy studiesE. coli myositis was generated in immunosuppressed mice with drug-susceptible [ceftriaxone minimum inhibitory concentration (MIC),0.04 mg/ml] or ESBL-producing E. coli (ceftriaxone MIC >32 mg/ml).

Ceftriaxone (Sigma) was administered subcutaneously at a dose of5 mg/kg (every 2 hours) for 24 hours (45). Imaging was performed inthe same group of animals immediately before starting and after com-pletion of antimicrobial treatment. After the final imaging time point,the same mice were sacrificed to determine CFU. A separate group ofsimilarly infected mice were also sacrificed before the start of antimi-crobial treatment to determine the starting CFU. Tissues were homogenizedin PBS, plated onto solidified LB medium, and incubated overnight at37°C to enumerate the CFU. SUV ratios were obtained by normalizingthe PET SUV from the infected right thigh with the inflammationcontrol (heat-killed E. coli strain) left thigh for each animal.

Statistical analysesData from uptake assays are presented as means and SEM and analyzedusing unpaired two-tailed Student’s t test or two-way analysis of vari-ance (ANOVA) with Bonferroni correction (for multiple comparisons)with GraphPad Prism 6.03 (GraphPad Software Inc.). PET SUV andex vivo tissue biodistribution data are represented as box plots wherecentral bars represent median value, the edges of the boxes representquartile value, and whiskers show the upper and lower ranges. Statis-tical comparisons were performed using a two-tailed Mann-Whitney Uor Kruskal-Wallis (for multiple comparisons) test (GraphPad SoftwareInc.). Values of P < 0.05 were considered statistically significant. Dataare presented on a linear scale, except for CFU, which is presented ona logarithmic scale.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/6/259/259ra146/DC1Fig. S1. Tissue histology from E. coli myositis in immunocompetent mice.Fig. S2. 18F-FDS PET/CT imaging of E. coli myositis in immunosuppressed mice.Fig. S3. Dynamic 18F-FDS PET imaging in mice with human brain tumors.Movie S1. 18F-FDS PET/CT imaging of E. coli myositis in immunocompetent mouse.

REFERENCES AND NOTES

1. M. S. Donnenberg, in Mandell, Douglas, and Bennett’s Principles and Practice of InfectiousDiseases, J. E. Bennett, G. L. Mandell, R. Dolin, Eds. (Elsevier Inc., Philadelphia, PA, 2010),chap. 218, pp. 2815–2833.

2. M. N. Guentzel, in Medical Microbiology, S. Baron, Ed. (University of Texas Medical Branch atGalveston, Galveston, TX, 1996), chap. 26.

3. A. Signore, A. W. Glaudemans, The molecular imaging approach to image infections andinflammation by nuclear medicine techniques. Ann. Nucl. Med. 25, 681–700 (2011).

4. Z. B. Li, Z. Wu, Q. Cao, D. W. Dick, J. R. Tseng, S. S. Gambhir, X. Chen, The synthesis of 18F-FDSand its potential application in molecular imaging. Mol. Imaging Biol. 10, 92–98 (2008).

5. P. Lechat, L. Hummel, S. Rousseau, I. Moszer, GenoList: An integrated environment forcomparative analysis of microbial genomes. Nucleic Acids Res. 36, D469–D474 (2008).

6. M. Yamada, M. H. Saier Jr., Glucitol-specific enzymes of the phosphotransferase system inEscherichia coli. Nucleotide sequence of the gut operon. J. Biol. Chem. 262, 5455–5463 (1987).

7. A. F. Zuluaga, B. E. Salazar, C. A. Rodriguez, A. X. Zapata, M. Agudelo, O. Vesga, Neutropeniainduced in outbred mice by a simplified low-dose cyclophosphamide regimen: Characteriza-tion and applicability to diverse experimental models of infectious diseases. BMC Infect. Dis. 6,55 (2006).

8. J. D. Chalmers, J. K. Taylor, A. Singanayagam, G. B. Fleming, A. R. Akram, P. Mandal, G. Choudhury,A. T. Hill, Epidemiology, antibiotic therapy, and clinical outcomes in health care-associatedpneumonia: A UK cohort study. Clin. Infect. Dis. 53, 107–113 (2011).

9. M. H. Kollef, A. Shorr, Y. P. Tabak, V. Gupta, L. Z. Liu, R. S. Johannes, Epidemiology andoutcomes of health-care–associated pneumonia: Results from a large US database ofculture-positive pneumonia. Chest 128, 3854–3862 (2005).

10. M. A. Garbati, A. I. Al Godhair, The growing resistance of Klebsiella pneumoniae; the need toexpand our antibiogram: Case report and review of the literature. Afr. J. Infect. Dis. 7, 8–10 (2013).

11. M. McKenna, Antibiotic resistance: The last resort. Nature 499, 394–396 (2013).

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 8

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

12. P. Nordmann, L. Dortet, L. Poirel, Carbapenem resistance in Enterobacteriaceae: Here is thestorm!. Trends Mol. Med. 18, 263–272 (2012).

13. CDC, Vital signs: Carbapenem-resistant Enterobacteriaceae.MMWR Morb. Mortal. Wkly. Rep. 62,165–170 (2013).

14. M. Melzer, I. Petersen, Mortality following bacteraemic infection caused by extendedspectrum beta-lactamase (ESBL) producing E. coli compared to non-ESBL producing E. coli. J.Infect. 55, 254–259 (2007).

15. D. W. Bates, K. Sands, E. Miller, P. N. Lanken, P. L. Hibberd, P. S. Graman, J. S. Schwartz, K. Kahn,D. R. Snydman, J. Parsonnet, R. Moore, E. Black, B. L. Johnson, A. Jha, R. Platt, Predicting bac-teremia in patients with sepsis syndrome. Academic Medical Center Consortium Sepsis ProjectWorking Group. J. Infect. Dis. 176, 1538–1551 (1997).

16. T. Nakamura, O. Takahashi, K. Matsui, S. Shimizu, M. Setoyama, M. Nakagawa, T. Fukui, T. Morimoto,Clinical prediction rules for bacteremia and in-hospital death based on clinical data at the timeof blood withdrawal for culture: An evaluation of their development and use. J. Eval. Clin. Pract.12, 692–703 (2006).

17. R. C. Owens Jr., P. G. Ambrose, C. H. Nightingale, Eds., Antibiotic Optimization: Concepts andStrategies in Clinical Practice (Marcel Dekker, New York, ed. 1, 2005).

18. R. D. Scott, The Direct Medical Costs of Healthcare-Associated Infections in U.S. Hospitalsand the Benefits of Prevention (Centers for Disease Control and Prevention, 2009).

19. B. M. Fischer, M. W. Olsen, C. D. Ley, T. L. Klausen, J. Mortensen, L. Hojgaard, P. E. Kristjansen,How few cancer cells can be detected by positron emission tomography? A frequent questionaddressed by an in vitro study. Eur. J. Nucl. Med. Mol. Imaging 33, 697–702 (2006).

20. C. König, H. P. Simmen, J. Blaser, Bacterial concentrations in pus and infected peritoneal fluid—Implications for bactericidal activity of antibiotics. J. Antimicrob. Chemother. 42, 227–232 (1998).

21. K. Jang, D. H. Lee, S. H. Lee, B. H. Chung, Treatment of prostatic abscess: Case collectionand comparison of treatment methods. Korean J. Urol. 53, 860–864 (2012).

22. M. Yamamoto, T. Fukushima, K. Hirakawa, H. Kimura, M. Tomonaga, Treatment of bacterialbrain abscess by repeated aspiration—Follow up by serial computed tomography. Neurol.Med. Chir. 40, 98–104; discussion 104–105 (2000).

23. R. C. Mease, C. A. Foss, M. G. Pomper, PET imaging in prostate cancer: Focus on prostate-specific membrane antigen. Curr. Top. Med. Chem. 13, 951–962 (2013).

24. M. E. Scott, R. E. Viola, The use of fluoro- and deoxy-substrate analogs to examine bindingspecificity and catalysis in the enzymes of the sorbitol pathway. Carbohydr. Res. 313, 247–253(1998).

25. X. Ning, S. Lee, Z. Wang, D. Kim, B. Stubblefield, E. Gilbert, N. Murthy, Maltodextrin-basedimaging probes detect bacteria in vivo with high sensitivity and specificity. Nat. Mater. 10,602–607 (2011).

26. M. van Oosten, T. Schäfer, J. A. Gazendam, K. Ohlsen, E. Tsompanidou, M. C. de Goffau,H. J. Harmsen, L. M. Crane, E. Lim, K. P. Francis, L. Cheung, M. Olive, V. Ntziachristos, J. M. van Dijl,G. M. van Dam, Real-time in vivo imaging of invasive- and biomaterial-associated bacterialinfections using fluorescently labelled vancomycin. Nat. Commun. 4, 2584 (2013).

27. F. J. Hernandez, L. Huang, M. E. Olson, K. M. Powers, L. I. Hernandez, D. K. Meyerholz, D. R. Thedens,M. A. Behlke, A. R. Horswill, J. O. McNamara II, Noninvasive imaging of Staphylococcus aureusinfections with a nuclease-activated probe. Nat. Med. 20, 301–306 (2014).

28. Y. Kong, H. Yao, H. Ren, S. Subbian, S. L. Cirillo, J. C. Sacchettini, J. Rao, J. D. Cirillo, Imagingtuberculosis with endogenous b-lactamase reporter enzyme fluorescence in live mice.Proc. Natl. Acad. Sci. U.S.A. 107, 12239–12244 (2010).

29. M. L. James, S. S. Gambhir, A molecular imaging primer: Modalities, imaging agents, andapplications. Physiol. Rev. 92, 897–965 (2012).

30. S. Vinjamuri, A. V. Hall, K. K. Solanki, J. Bomanji, Q. Siraj, E. O’Shaughnessy, S. S. Das, K. E. Britton,Comparison of 99mTc infecton imaging with radiolabelled white-cell imaging in the evalua-tion of bacterial infection. Lancet 347, 233–235 (1996).

31. P. H. Nibbering, M. M. Welling, A. Paulusma-Annema, C. P. Brouwer, A. Lupetti, E. K. Pauwels,99mTc-Labeled UBI 29-41 peptide for monitoring the efficacy of antibacterial agents in miceinfected with Staphylococcus aureus. J. Nucl. Med. 45, 321–326 (2004).

32. L. Sarda, A. C. Crémieux, Y. Lebellec, A. Meulemans, R. Lebtahi, G. Hayem, R. Génin, N. Delahaye,D. Huten, D. Le Guludec, Inability of 99mTc-ciprofloxacin scintigraphy to discriminate betweenseptic and sterile osteoarticular diseases. J. Nucl. Med. 44, 920–926 (2003).

33. L. Sarda, A. Saleh-Mghir, C. Peker, A. Meulemans, A. C. Crémieux, D. Le Guludec, Evaluationof 99mTc-ciprofloxacin scintigraphy in a rabbit model of Staphylococcus aureus prostheticjoint infection. J. Nucl. Med. 43, 239–245 (2002).

34. N. Dumarey, A. Schoutens, Renal abscess: Filling in with Tc-99m ciprofloxacin of defectsseen on Tc-99m DMSA SPECT. Clin. Nucl. Med. 28, 68–69 (2003).

35. O. Langer, M. Brunner, M. Zeitlinger, S. Ziegler, U. Muller, G. Dobrozemsky, E. Lackner, C. Joukhadar,M. Mitterhauser, W. Wadsak, E. Minar, R. Dudczak, K. Kletter, M. Muller, In vitro and in vivo

evaluation of [18F]ciprofloxacin for the imaging of bacterial infections with PET. Eur. J. Nucl.Med. Mol. Imaging 32, 143–150 (2005).

36. M. Welling, M. Stokkel, J. Balter, L. Sarda-Mantel, A. Meulemans, D. Le Guludec, The manyroads to infection imaging. Eur. J. Nucl. Med. Mol. Imaging 35, 848–849 (2008).

37. P. Panizzi, M. Nahrendorf, J. L. Figueiredo, J. Panizzi, B. Marinelli, Y. Iwamoto, E. Keliher, A. A. Maddur,P. Waterman, H. K. Kroh, F. Leuschner, E. Aikawa, F. K. Swirski, M. J. Pittet, T. M. Hackeng,P. Fuentes-Prior, O. Schneewind, P. E. Bock, R. Weissleder, In vivo detection of Staphylococcusaureus endocarditis by targeting pathogen-specific prothrombin activation. Nat. Med. 17,1142–1146 (2011).

38. C. Bettegowda, C. A. Foss, I. Cheong, Y. Wang, L. Diaz, N. Agrawal, J. Fox, J. Dick, L. H. Dang,S. Zhou, K. W. Kinzler, B. Vogelstein, M. G. Pomper, Imaging bacterial infections with radio-labeled 1-(2¢-deoxy-2¢-fluoro-b-D-arabinofuranosyl)-5-iodouracil. Proc. Natl. Acad. Sci. U.S.A.102, 1145–1150 (2005).

39. M. T. Wyss, M. Honer, N. Späth, J. Gottschalk, S. M. Ametamey, B. Weber, G. K. von Schulthess,A. Buck, A. H. Kaim, Influence of ceftriaxone treatment on FDG uptake—An in vivo[18F]-fluorodeoxyglucose imaging study in soft tissue infections in rats. Nucl. Med. Biol.31, 875–882 (2004).

40. S. L. Davis, E. L. Nuermberger, P. K. Um, C. Vidal, B. Jedynak, M. G. Pomper, W. R. Bishai, S. K. Jain,Noninvasive pulmonary [18F]-2-fluoro-deoxy-D-glucose positron emission tomography corre-lates with bactericidal activity of tuberculosis drug treatment. Antimicrob. Agents Chemother.53, 4879–4884 (2009).

41. S. K. Jain, P. Kwon, W. J. Moss, Management and outcomes of intracranial tuberculomasdeveloping during antituberculous therapy: Case report and review. Clin. Pediatr. 44, 443–450(2005).

42. F. Sievers, A. Wilm, D. Dineen, T. J. Gibson, K. Karplus, W. Li, R. Lopez, H. McWilliam, M. Remmert,J. Söding, J. D. Thompson, D. G. Higgins, Fast, scalable generation of high-quality proteinmultiple sequence alignments using Clustal Omega. Mol. Syst. Biol. 7, 539 (2011).

43. M. Rouse, J. M. Steckelberg, in Handbook of Animal Models of Infection, O. Zak, M. A. Sande,Eds. (Academic Press, London, 1999), chap. 58, pp. 495–500.

44. O. Gonzalez-Perez, H. Guerrero-Cazares, A. Quiñones-Hinojosa, Targeting of deep brain structureswith microinjections for delivery of drugs, viral vectors, or cell transplants. J. Vis. Exp., 2082(2010).

45. M. L. van Ogtrop, H. Mattie, H. F. Guiot, E. van Strijen, A. M. Hazekamp-van Dokkum, R. van Furth,Comparative study of the effects of four cephalosporins against Escherichia coli in vitro andin vivo. Antimicrob. Agents Chemother. 34, 1932–1937 (1990).

Acknowledgments: We thank A. Woods and L. Muller (NIH/National Institute on Drug Abuse)for performance of mass spectrometric analysis; P. Mister and K. Carroll (Johns Hopkins Hos-pital) for providing the ESBL-producing clinical strains; and O. Wijesekera, H. Guerrero-Cazares, andA. Quinones-Hinojosa (Johns Hopkins Hospital) for the brain infection and tumor models. Wealso thank C. Sears and W. Bishai (Johns Hopkins Hospital) for comments. Funding: This studywas funded by the NIH Director’s New Innovator Award DP2-OD006492 (S.K.J.), as well as R01-HL116316 (S.K.J.), and a subcontract from Harvard University Center for AIDS Research5P30AI060354-09 (S.K.J.). Author contributions: E.A.W., A.A.O., and S.K.J. designed the research.E.A.W. and R.C.M. designed the protocols for synthesizing 18F-FDS. E.A.W., A.M.M., and E.M.M.synthesized 18F-FDS for the experiments. E.A.W., A.A.O., A.M.M., and E.M.M. performed the in vitroassays. A.A.O., V.P.D., and M.K. developed the animal models and performed imaging. A.A.O., V.P.D.,and A.M.M. performed biodistribution studies. A.A.O. and S.P. performed microbiological andhistological studies. S.P. performed protein estimations. A.A.O. performed the statistical analyses.E.A.W., A.A.O., M.G.P., and S.K.J. analyzed the data. All authors wrote the paper. S.K.J. providedfunding and supervised the project. Competing interests: E.A.W., A.A.O., M.K., M.G.P., andS.K.J. are inventors for an International patent PCT/US13/059897, “Bacteria-specific labeledsubstrates as imaging biomarkers to diagnose, locate and monitor infections” filed by JohnsHopkins University.

Submitted 17 June 2014Accepted 19 August 2014Published 22 October 201410.1126/scitranslmed.3009815

Citation: E. A. Weinstein, A. A. Ordonez, V. P. DeMarco, A. M. Murawski, S. Pokkali,E . M. MacDonald , M. K lunk, R . C . Mease , M. G. Pomper , S . K . Ja in , ImagingEnterobacteriaceae infection in vivo with 18F-fluorodeoxysorbitol positron emissiontomography. Sci. Transl. Med. 6, 259ra146 (2014).

R E S EARCH ART I C L E

www.ScienceTranslationalMedicine.org 22 October 2014 Vol 6 Issue 259 259ra146 9

on

Oct

ober

22,

201

4st

m.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m