Clinical Signicance of Rapid and Sensitive Quantication of Pseudomonas aeruginosa by Quantitative Reverse Transcription PCR Targeting of rRNA Molecules

Background: For Pseudomonas aeruginosa, nosocomial infection control and appropriate antimicrobial treatment have become important issues. Diagnosis is critical in managing P. aeruginosa infection, but conventional methods are not highly accurate or rapid. A novel P. aeruginosa quantication system based on 23S rRNA-targeted quantitative reverse transcription PCR (qRT-PCR) is a candidate diagnostic tool for managing P. aeruginosa infection. Here, we assessed the performance and potential impact of qRT-PCR on antibiotic therapy administered to ICU patients. Methods: We rst evaluated specicity and detection sensitivity of the 23S rRNA-targeted qRT-PCR system in vitro and determined whether P. aeruginosa viable counts detected by this system reect the inammatory response of infected cells. Next, we utilized this system on fecal samples collected from 65 septic ICU patients and 44 healthy volunteers to identify the ICU infection status. Additionally, we monitored drug-resistant P. aeruginosa in 4 ICU patients. The trend was compared with trends in fecal microbiota composition, antibiotic use, and mechanical ventilator use. Results: The 23S rRNA-targeted qRT-PCR system quantied P. aeruginosa directly from clinical samples with high sensitivity (blood, 1 cell/mL; stool, 100 cells/g) without cross-reaction (within 6 h). Additionally, P. aeruginosa numbers detected under antibacterial treatment correlated well with the inammatory response compared to other detection methods such as culturing and qPCR. Using this system, we conrmed that the P. aeruginosa detection ratio in ICU patients was signicantly higher than that in healthy volunteers (49.2% vs. 13.6%, P<0.05). While some ICU patients had high P. aeruginosa numbers in feces (10 8 cells/g of feces), some patients had very low P. aeruginosa numbers (10 2 cells/g of feces) as observed in healthy volunteers. P. aeruginosa counts in feces of ICU patients monitored by this system correlated negatively with the proportion of total obligate anaerobes

effective for treating P. aeruginosa infections, has become a serious problem [9,10]. This resistance in P. aeruginosa is attributed to a variety of mechanisms, including the production of antibiotic-inactivating enzymes such as cephalosporinases, the constitutive expression of various e ux pumps, and the low permeability of the outer membrane [11]. Consequently, infection with MDRP strains can limit antibiotic therapy, resulting in inappropriate empirical therapy or delays in initiating appropriate therapy, which can lead to high mortality [8,12,13]. Thus, in order to select appropriate antibacterial agents for treating P. aeruginosa infection, it is necessary to regularly monitor and rapidly and accurately identify P. aeruginosa including MDRP strains in clinical specimens.
Currently, clinical assays for P. aeruginosa are typically conducted by bacterial culture. Generally, bacterial culture methods are the gold standard for detecting pathogens; however, culture methods have several limitations: (i) the detection of pathogens by culture methods is generally time-consuming (takes 2-3 days) [14,15]; (ii) culture methods only detect culturable pathogens that have the ability to form colonies [16]; and (iii) culture methods can underestimate the number of pathogen due to selection bias induced by administered antibiotics that are present in human blood [17]. Accordingly, culture-based diagnostic procedures can produce false-negative results. Culture-negative sepsis is commonly observed in ICU patients: previous studies have shown negative blood cultures in 30% of septic shock patient [18]. It has also been reported that blood cultures from febrile neutropenic patients can identify the pathogens in only 17 to 42% of cases [19,20]. Thus, to overcome the problems associated with bacterial culture, we adopted Yakult Intestinal Flora-SCAN (YIF-SCAN®), a highly sensitive and rapid system based on quantitative reverse transcription PCR (qRT-PCR) [21][22][23]. YIF-SCAN® targets rRNA molecules that exist in bacteria abundantly (approximately 10 4 copies per actively growing cell), the sensitivity of qRT-PCR is 100 times higher than that of qPCR assays that target rRNA genes (more than 10 copies/bacterial genome) [21]. Therefore, qRT-PCR is a more sensitive assay for detecting occult bacteria. Also, it requires only 6 h for the quanti cation of bacteria and detects only live bacteria from specimens [23]. Based on this, we applied this system to rapid and accurate diagnosis of P. aeruginosa infection.
In the present study, in order to establish a method that is useful in the early diagnosis of P. aeruginosa infection, to guide infection control measures, and to assess the results of antimicrobial treatment, we developed a novel P. aeruginosa quanti cation system based on 23S rRNA-targeted qRT-PCR. We demonstrated the detection sensitivity and speci city of the developed method and its applicability for monitoring enteric P. aeruginosa in ICU patients.

Methods
Strains and culture conditions.
The bacterial strains used are listed in Table 1. Each strain was routinely grown under the conditions listed in Additional File 1: Supplementary Table 1. The colony forming units (CFU) of P. aeruginosa were determined by culturing each strain aerobically on trypticase soy agar (TSA, Beckton Dickinson, Franklin Lakes, NJ, USA) or NCC agar at 37 °C for 24 h. NCC agar is a newly developed selective medium for the detection of P. aeruginosa in this study, and contains NAC (nalidixic acid, cetrimide) medium (Nissui Pharmaceutical, Tokyo, Japan) and TSA at a ratio of 1:20, and 0.5 µg/ml ceftizoxime sodium (Chem-Impex International, Inc., Wood Dale, IL, USA) and 64 µg/ml cephalothin sodium crystalline (Sigma-Aldrich Co., St. Louis, MO, USA). Total bacterial cell counts of fresh cultures were determined using 4', 6diamidino-2-phenylindole (DAPI) staining according to a previously described method [24]. Based on total cell counts, each fresh culture was diluted to obtain 1 ml of bacterial suspension containing 10 9 cells. In the case of P. aeruginosa, the cells in the culture solution were dispersed using a 27 G syringe with a needle (Terumo, Tokyo, Japan) prior to use in subsequent experiments. Development of 23S rRNA gene-targeted primers speci c to P. aeruginosa.
Multiple alignment of the 6 P. aeruginosa strains and 8 related species was performed with the CLUSTAL_X program [25] using the 23S rRNA gene sequences obtained from the previous experiment.
After comparison of the sequences in silico, target sites for P. aeruginosa species-speci c detection were identi ed, and a primer set, s-Pa-F (5 -GTC TTT TAG ATG ACG AAG TGG-3 ) and s-Pa-R (5 -TGG TAT CTT CGA CCA GCC AGA-3 ), was newly constructed. The product size and annealing temperature were 234 bp and 60 °C, respectively. The speci city of the designed primer pair was con rmed by submitting the sequences to the BLAST program of the National Center for Biotechnology Information (NCBI) (http://blast.ncbi.nlm.nih.gov/Blast.cgi).
Total RNA extraction.
For RNA stabilization, 2 volumes of RNAprotect Bacteria Reagent (QIAGEN, Hilden, Germany) were added to samples of each bacterial strain. After being kept for 10 min at room temperature, the bacterial suspensions were centrifuged at 13,000 × g for 10 min. The supernatant was discarded, and the pellet was stored at -80 °C until used for RNA extraction. RNA extraction was performed using a previously described method [21]. Brie y, each thawed sample was resuspended in a solution containing 346.5 µl of RLT buffer (QIAGEN, Hilden, Germany), 3.5 µl of β-mercaptoethanol (Sigma-Aldrich Co., St. Louis, MO, USA), and 100 µl of Tris-EDTA buffer (Wako Pure Chemical Industries, Osaka, Japan). Glass beads (300 mg; diameter, 0.1 mm) (TOMY Seiko, Tokyo, Japan) were added to the suspension, and the mixture was vortexed vigorously for 5 min with a ShakeMaster Auto machine (BioMedical Science Inc., Tokyo, Japan). Then, 500 µl of water-saturated phenol (Wako Pure Chemical Industries, Osaka, Japan) was added to the mixture, which was incubated at 60 °C for 10 min. Next, 100 µl of chloroform-isoamyl alcohol (24:1) was added to the mixture. After centrifugation of the mixture at 13,000 × g for 10 min at 4 °C, the supernatant (470 µl) was collected, and an equal volume of chloroform-isoamyl alcohol was added to it. After centrifugation at 12,000 × g and 4 °C for 5 min, the supernatant (400 µl) was collected and subjected to isopropanol precipitation. Finally, the nucleic acid fraction was suspended in nucleasefree water (Ambion Inc., Waltham, MA, USA). qRT-PCR.
qRT-PCR was performed using a previously described method [21]. Brie y, in the case of bacterial cultures or blood samples, qRT-PCR was conducted in a one-step reaction using a Qiagen OneStep RT-PCR kit (QIAGEN, Hilden, Germany aeruginosa ATCC 10145 T were generated by using the threshold cycle (C T ) values and the corresponding cell counts, which were determined microscopically with DAPI staining as described elsewhere. C T values in the linear range of the assay were applied to the analytical curve generated in the same experiment to obtain the corresponding bacterial count in each nucleic acid sample; this count was converted to the count per sample.
DNA extraction and qPCR.
Fresh cultures of each bacterial strain (1 ml) were centrifuged at 13,000 × g for 10 min. Then, the supernatant (800 µl) was discarded and the pellet was stored at -80 °C until used for DNA extraction. The bacterial suspension (200 µl) was subjected to DNA extraction. RNA extraction and qPCR were performed using a previously described method [26]. Brie y, each thawed sample was mixed with 250 µl of extraction buffer (100 mM Tris-HCl, 40 mM EDTA; pH 9.0) and 50 µl of 10% sodium dodecyl sulfate. Glass beads (300 mg; diameter, 0.1 mm) and 500 µl of Tris-EDTA (TE)-saturated phenol were added to the suspension, and the mixture was subjected to vigorous vortexing for 10 min on a ShakeMaster Auto apparatus (Bio Medical Science Inc., Tokyo, Japan). After phenol-chloroform puri cation and isopropanol precipitation, the nucleic acid fraction was suspended in 100 µl of nuclease-free water. qPCR was carried out using a Qiagen OneStep RT-PCR kit (QIAGEN, Hilden, Germany). Each reaction mixture (20 µl) contained the same components as those for qRT-PCR, except for the replacement of 2 µl template RNA with the same amount of template DNA. The reaction mixture was incubated at 95 °C for 15 min and 45 cycles at 94 °C for 20 s, 60 °C for 20 s, and 72 °C for 35 s. The subsequent procedures were the same as those for qRT-PCR.
Determination of primer speci city.
Total RNA fractions extracted from the bacterial cells of each strain (shown in Table 1) at a dosorresponding to 10 5 cells were assessed by qRT-PCR using the primer set of s-Pa-F and s-Pa-R. Using the standard curve for P. aeruginosa as described above, we judged the ampli ed signal to be positive (+) when it was more than that of 10 4 standard cells and negative (−) when it was less than that of 10 1 standard cells. The ampli ed signal was also de ned as negative (−) when the corresponding melting curve had a peak different from that of the standard strain.
Determination of qRT-PCR sensitivity.
Total RNA and DNA fractions of the 6 strains of P. aeruginosa were extracted from culture samples, and bacterial counts were determined microscopically with DAPI staining as described elsewhere. Serial RNA and DNA dilutions corresponding to bacterial counts ranging from 10 − 2 to 10 4 cells were assessed by qRT-PCR and qPCR assays, respectively. The range of RNA and DNA concentrations at which there was linearity with the C T value was con rmed. Results are expressed as the mean and standard deviation of the results from triplicate samples.
Quanti cation of P. aeruginosa spiked to human blood by qRT-PCR and culture methods.
Commercially available human blood type A (Kohjin Bio Co., Saitama, Japan) was used in this study to determine the detection limit of qRT-PCR and to compare the bacterial counts determined by qRT-PCR and the culture method. P. aeruginosa ATCC 10145 T prepared as described above was serially diluted with 1% Tween 20 (Wako Pure Chemical Industries, Osaka, Japan) in trypticase soy broth (TSB, Beckton Dickinson Co., Franklin Lakes, NJ, USA) and then spiked to make nal concentrations ranging from 10 0 to 10 4 cells/ml of blood. For qRT-PCR assays, 1 ml of the blood sample was added to 2 ml of RNAprotect bacterial reagent (QIAGEN, Hilden, Germany). After being kept for 10 min at room temperature and centrifuged at 15, 000 × g for 10 min, the pellet was stored at -80 °C until used for RNA extraction. Then, RNA fractions extracted from the mixture were assessed by qRT-PCR assay. The obtained C T values were applied to the standard curve generated with the RNA dilution series for P. aeruginosa ATCC 10145 T to determine the qRT-PCR counts with the primer set of s-Pa-F and s-Pa-R. For the culture method, to avoid the effect of reduced colony-forming capacity by the presence of antimicrobial substances in human blood such as lysozyme and complement proteins, the P. aeruginosa-spiked human blood sample was immediately cultured on NCC agar plates. One ml of each appropriate dilution series of blood sample was cultured on 10 culture plates (100 µl each), at 37 °C for 24 h. The CFU counts of P. aeruginosa in 1 ml of blood was calculated by summed up the results of 10 culture plates.
Quanti cation of P. aeruginosa added to human feces by qRT-PCR and culture methods.
Fecal samples collected from three healthy adult volunteers who had been con rmed in advance by qRT-PCR not to include P. aeruginosa in their indigenous intestinal populations were used in this study. Each fecal sample was weighed and suspended in 9 volumes of TSB. P. aeruginosa ATCC 10145 T prepared as described above was serially diluted with 1% Tween 20 in TSB and then spiked at nal concentrations ranging from 10 2 to 10 7 cells/g of feces. In preparation for total RNA extraction, 400 µl of each fecal homogenate was added to 800 µl of RNA later (Ambion Inc., Waltham, MA, USA). After being kept for 10 min at room temperature, 120 µl of the mixture was added to 1 ml of Dulbecco's PBS (-) (Nissui Pharmaceutical Co., Tokyo, Japan) Then, after centrifugation of the mixture at 15,000 × g for 10 min, the pellet was stored at -80 °C until used for RNA extraction. RNA fractions extracted from the pellet were assessed by qRT-PCR assays. The C T values obtained were applied to the standard curve generated with the RNA dilution series for P. aeruginosa ATCC 10145 T to determine the qRT-PCR counts with the primer set of s-Pa-F and s-Pa-R. For the culture method, 100 µl of each appropriate dilution series of fecal homogenate was cultured on NCC agar plates at 37 °C for 24 h.
Monitoring of MDRP exposed to antibiotics by three different methods. For RNA extraction, 1 ml of the dispersed culture solution was added to 2 ml of RNA later. After being kept for 10 min at room temperature and centrifuged at 15,000 × g for 10 min, the pellet was stored at -80 °C until used for RNA extraction. RNA extraction and qRT-PCR with the primer set s-Pa-F/s-Pa-R were performed as described above. Also, for DNA extraction, 1 ml of the dispersed culture solution was centrifuged at 15,000 × g for 10 min, and the pellet was stored at -80 °C until used for DNA extraction.
DNA extraction and qPCR with the primer set s-Pa-F/s-Pa-R were performed as described above. For the culture method, 1 ml of the appropriate dilution series was cultured on TSA plates at 37 °C for 24 h.
Infection of Caco-2 cells with MDRP exposed to antibiotics.
Caco-2 cells (86010202) obtained from Public Health England (London, UK) were grown in DMEM supplemented with 10% FBS, 1% MEM nonessential amino acids, 1% penicillin-streptomycin (10,000 units/ml penicillin, 10,000 µg/ml streptomycin) (Gibco, Waltham, MA, USA) at 37 °C in the presence of humidi ed 5% CO 2 in air. Cells were plated on 6-well plates (Costar, Corning, NY, USA) at a density of 3.0 × 10 5 cells/well and incubated for 24 h for the infection assays. After incubation, the cells were washed three times with D-PBS (Thermo Fisher Scienti c Inc., Waltham, MA, USA) to remove penicillin G and streptomycin contained in the cell culture. For cell infection, 2 ml of MDRP suspension treated with antibiotics and washed as described above was applied to the cells, and cells were incubated at 37 °C with 5% CO 2 in air for 6 h. To compare with a reference value of uninfected cells, fresh antibiotic-free culture medium was also applied to the cells as a control, and then the cells were incubated at 37 °C with 5% CO 2 in air for 6 h.
Quanti cation of in ammatory response in infected Caco-2 cells.
The bacterial proin ammatory effect was assessed by measuring the transcription level of IL-8 and IL-8 secretion in infected Caco-2 cells. After a 6-hour incubation, 2 ml of the cell culture medium was sampled to measure IL-8 production. The medium was centrifuged at 2,300 g for 5 min, and the supernatant was stored at -80 °C until used for determinations. Then, total RNA was extracted from infected Caco-2 cells Observation and acquisition of uorescent images of MDRP.
Fluorescence in situ hybridization (FISH) analyses were performed as described previously with minor modi cations [29]. Brie y, the bacterial suspension for acquisition of uorescent images, which was sampled at 0, 48, and 336 h after cultivation, was centrifuged at 13, 000 × g for 10 min and condensed to 100 times concentration. The suspension was xed with 3 volumes of 4% paraformaldehyde and left at 4 °C for 16 h. Ten microliters of the xed sample, after being condensed to 4 times the concentration by centrifugation, was smeared on a MAS-coated slide glass (Matsunami Glass Ind., Ltd., Osaka, Japan), which was then hybridized with the TAMRA-labeled 16S rRNA probe Eub338 qRT-PCR quanti cation of P. aeruginosa and monitoring of drug-resistant P. aeruginosa in clinical samples.
For the analysis of clinical samples, we used total RNA isolated from fecal samples collected in a previous study [30], which were stored at -80 °C until used in this study. In the previous study, fecal samples were acquired from 65 patients who were more than 16 years old and had been placed on a ventilator within 3 days after admission to the ICU, and who were diagnosed as having sepsis in the Department of Traumatology and Acute Critical Medicine, Osaka University Medical School, and Osaka General Medical Center during the period from November 2011 to September 2016. If sepsis occurred, patients were initially treated empirically for the underlying clinical syndrome and then according to the results of antibiotic susceptibility testing of the bacterial isolate causing the sepsis. Antibiotics were administered under the same policy during the entire study period. Also, we collected fecal samples from 44 healthy Japanese adults and prepared total RNA in the same manner as for the clinical samples, as described above. Then, three serial dilutions of each extracted RNA sample were used for qRT-PCR, and the C T values obtained were applied to the standard curve generated with the RNA dilution series for P.
The microbiota compositions of 4 ICU patients were analyzed using the YIF-SCAN® version of a 16S and 23S rRNA-targeted qRT-PCR system using speci c primers for the 6 most prevalent obligate anaerobic bacterial groups (Clostridium coccoides group, C. leptum subgroup, Bacteroides fragilis group, Bi dobacterium, Atopobium cluster, and Prevotella) and 5 facultative anaerobic and aerobic bacterial groups (Lactobacillus, Enterobacteriaceae, Enterococcus, and Staphylococcus] as described previously [21,22]. A standard curve was generated with qRT-PCR using the C T value, i.e., the cycle number when the threshold uorescence was reached, and the corresponding cell count was determined microscopically with DAPI staining for a dilution series of the standard strains as described elsewhere. To determine the types of bacteria present in the samples, three serial dilutions of an extracted RNA sample were used for qRT-PCR, and the C T values in the linear range of the assay were applied to the standard curve to obtain the corresponding bacterial cell counts in each nucleic acid sample. These data were then used to determine the number of bacteria per sample.

Statistical analysis.
Statistical analyses were performed using IBM SPSS Statistics Desktop version 22.0 software (IBM Japan Ltd., Tokyo, Japan). Fisher's exact test was used to compare the detection rate of P. aeruginosa between 2 groups. The Mann-Whitney U test (2-tailed) was used to compare the average P. aeruginosa counts between 2 groups. Pearson's correlation coe cient was used to analyze the relationship between the P. aeruginosa count quanti ed with each measurement method and the cytokine levels of the infected Caco-2 cells. P values < 0.05 were considered statistically signi cant. Spearman's rank correlation coe cient was used to analyze the association between the number of P. aeruginosa and fecal occupation rate of representative microorganisms.

Results
The newly designed primer set for the detection of P. aeruginosa shows excellent speci city.
The speci city of the newly designed primer set, s-Pa-F/s-Pa-R, was evaluated by qRT-PCR with total RNA (corresponding to 10 5 cells) template extracted from each pure culture of 39 strains, including 6 strains of P. aeruginosa, 12 species related to P. aeruginosa, and 21 strains of enteric bacteria or causative agents of human intestinal tract infectious disease ( Table 1). The s-Pa-F/s-Pa-R primer set reacted strongly with the 6 strains of P. aeruginosa, but showed no reaction with the other 33 bacterial strains investigated. Therefore, the newly designed primer set targeting P. aeruginosa was speci c for the target strains.
Speci c quanti cation of cultured P. aeruginosa by qRT-PCR has a lower detection limit compared to qPCR.
The counts of P. aeruginosa in pure cultures obtained by DAPI staining and C T values obtained by qRT-PCR showed good correlation for the 6 strains of P. aeruginosa (R 2 = 0.9977-0.9999) (Fig. 1). The counts of P. aeruginosa also correlated well with the C T values obtained by qPCR (R 2 = 0.9979-0.9998); however, the detection limit was 10 0 cells/reaction for the 6 P. aeruginosa strains (Fig. 1). These results indicated that qRT-PCR is approximately 100-fold more sensitive than qPCR for the quanti cation of P. aeruginosa. The parameters from the linear regression analyses for the 6 different P. aeruginosa strains were highly similar, indicating they had the same quantitative performance (Fig. 1G).
qRT-PCR can quantitate the number of P. aeruginosa spiked in human blood or feces with good correlation with the culture method.
The spiked bacterial counts could be detected by qRT-PCR using the primer set s-Pa-F/s-Pa-R, at the lowest concentration of 10 0.3 cells/ml of blood and 10 2.4 cells/g of feces. In addition, the bacterial counts detected by the culture method and those determined by qRT-PCR showed good correlation ( Fig. 2A, 2B). These results indicated that our qRT-PCR method enabled quantitative detection of P. aeruginosa in blood with the lowest detection limit of 10 0 cells/ml of blood and in feces with the lowest detection limit of 10 2 cells/g of feces.
In vitro experiments revealed that the number of MDRP obtained by qRT-PCR accurately correlated with in ammatory cytokine production in Caco-2 cells.
Changes over time in the counts of MDRP treated with antibiotics were measured with qRT-PCR using the P. aeruginosa-speci c primers, bacterial culture, and qPCR. The detected counts differed greatly depending on the measurement method (Fig. 3A). The counts detected with qRT-PCR gradually decreased until 336 h after antibiotic treatment. In contrast, the counts detected with qPCR showed almost no change from soon after antibiotic treatment (6 h after antibiotic treatment) to 336 h after antibiotic treatment. The counts detected with the culture method decreased rapidly after antibiotic treatment.
Bacteria were no longer detected 48 h after treatment (detection limit: <1 cell/ml).
When the above-mentioned P. aeruginosa solution was observed with FISH, a large number of viable bacteria were observed at the start of antibiotic treatment; however, the number of bacteria dramatically decreased after 48 h, and almost no viable bacteria could be observed after 336 h (Fig. 3B).
Next, P. aeruginosa solutions treated with antibiotics were sampled over time and used to infect Caco-2 cells. The relationship between the in ammatory response and the P. aeruginosa counts obtained with each measurement method was then evaluated (Fig. 3C). Results showed that while the IL-8 gene expression level and production level in Caco-2 cells increased more than 10-fold and 9-fold, respectively, when infected with P. aeruginosa at 6 h after antibiotic treatment, these decreased as the duration of antibiotic treatment increased. At 336 h after treatment, they had decreased to less than half and onetenth, respectively, compared to the levels at 6 h after antibiotic treatment. Even at 48 h after antibiotic treatment, when P. aeruginosa was no longer detectable by the culture method, the IL-8 gene expression and secretion levels were very high (gene expression level: 4-6 times higher than when not infected; production: 7-10 times higher than when not infected). At the point when no decrease in cell counts from antibiotic treatment was seen with qPCR at 336 h, IL-8 gene expression and production had decreased remarkably. Trend in the P. aeruginosa cell count monitored by qRT-PCR is similar with the level of the Caco-2 cell Il-8 production; however, the results of the culture method and qPCR for the cell count did not similarly re ect the level of the Caco-2 cell IL-8 production (Fig. 3C). qRT-PCR can detect P. aeruginosa from human fecal samples.
To examine whether the qRT-PCR system using the newly designed primer set can be applied to clinical samples, qRT-PCR analysis was performed using fecal samples collected from 65 patients who were placed on a ventilator within 3 days after admission to the ICU ( SD, 39 ± 10 years]) was also tested. As a result, 6 out of 44 healthy volunteer samples (13.6%) were positive for P. aeruginosa, and the average count was 10 4.38±1.21 cells/g of feces ( Table 2). The detection ratio in ICU patients was signi cantly higher than that in healthy volunteers (P < 0.05). As for bacterial counts, while some patients in ICU showed high levels of P. aeruginosa in the feces (10 8 cells/g of feces), some patients had it at very low levels (10 2 cells/g of feces) as observed in healthy volunteers. The combined use of primers targeting P. aeruginosaspeci c 23S rRNA and drug resistance genes can monitor the presence of drug-resistant P. aeruginosa in human fecal samples.
When P. aeruginosa counts in the feces of ICU patients undergoing antibiotic treatment (4 patients) were measured with qRT-PCR using the new P. aeruginosa-speci c primers, we were able to monitor abnormal proliferation and excessive decreases of P. aeruginosa in the intestines within a range of < 10 2 cells/g to 10 9 cells/g (Fig. 4). Monitoring of drug-resistant P. aeruginosa was also possible with the combined use of drug-resistance gene ampli cation primers and the P. aeruginosa-speci c primers (Fig. 4). In the four patients in whom P. aeruginosa was detected from the feces, a drug e ux pump gene (mexA) and the AmpC β-lactamase gene (ampC) were detected from the feces of all patients (Fig. 4, A-D). From Patient B, who was receiving meropenem (a carbapenem antibiotic), a metallo-β-lactamase encoding gene (bla IMP ) was detected (Fig. 4B).
Trend in P. aeruginosa counts in the feces of ICU patients monitored by the qRT-PCR method were related to whether or not antibiotics were being administered, whether or not an arti cial ventilator was being used, and the composition of the patients' intestinal microbiota (Fig. 4, Additional le 3: Supplementary   Table 3). In Patient A, explosive proliferation of P. aeruginosa in the intestines was induced at the time (Day 30) when obligate anaerobe groups in the intestines rapidly decreased with the administration of antibiotics (Fig. 4A). Similar trends were seen in Patient B (Day 15, Day 41) and Patient D (Day 16) (Fig. 4B, D). In contrast, when the administration of antibiotics was ceased, the P. aeruginosa count in patients' intestines decreased (Patient A: Day 38, Patient C: Day 31, Patient D: Day 23) (Fig. 4A, C, D). In two patients taken off from arti cial ventilators, intestinal counts of drug-resistant P. aeruginosa decreased to below the detection limit at the time when the counts of the most prevalent obligate anaerobes in the intestines recovered after ventilator withdrawal (Patient C: Day 31, D: Day 23) (Fig. 4C, Negative correlations were seen between the P. aeruginosa counts in the intestines of ICU patients and the proportion of total obligate anaerobes. In contrast, positive correlations were seen between P. aeruginosa counts in the feces of ICU patients and the proportions of total facultative anaerobes and aerobes. Speci cally, the proportions of Clostridium coccoides group, C. leptum subgroup, and Atopobium cluster were negatively correlated, whereas the proportion of Enterobacteriaceae was positively correlated with the P. aeruginosa counts (Fig. 4E).

Discussion
For P. aeruginosa infection, nosocomial infection control and appropriate antimicrobial treatment have become important issues. Diagnosis plays an important role in managing P. aeruginosa infection, but conventional inspection methods such as bacterial culture and qPCR are limited by low sensitivity and inaccuracy [35,36]. However, using qRT-PCR with the newly developed 23S rRNA-targeted primers speci c to P. aeruginosa, P. aeruginosa was detected from 1 cell/ml of blood and 10 2 cells/g of feces in in vitro experiments using spiked samples (equivalent to a sensitivity more than 100 times greater than qPCR; Fig. 1), with high speci city and quanti ability. When this qRT-PCR method was applied to fecal samples from ICU patients, it was demonstrated that P. aeruginosa had colonized some patients at very low levels (10 2 cells/g of feces) in the intestine, as observed in healthy subjects. qRT-PCR was shown to be useful in the screening of patients who carry minute amounts of P. aeruginosa, which are easy to miss with conventional methods. Measurements of P. aeruginosa in clinical samples other than feces and environmental samples have also been reported to be important in the control of P. aeruginosa infection in ICU patients [37,38]. The current qRT-PCR method was able to accurately quantify P. aeruginosa even in fecal samples that contain large numbers of bacteria and PCR inhibitors. This suggests that the qRT-PCR method can be applied to a wide range of samples, including bronchoalveolar lavage uid, sputum, intraperitoneal drainage uid, devices, and environmental samples such as hospital sewage.
The proper timing of antibiotic administration is very important in the treatment of P. aeruginosa infection. The results of microorganism tests such as culturing and qPCR and patients' clinical conditions do not necessarily coincide [18][19][20]. In our experimental system using P. aeruginosa-infected cells, the counts of P. aeruginosa by qRT-PCR correlated well with the in ammatory response (IL-8 gene expression and production level) of infected cells compared to culturing and qPCR. IL-8 is a biomarker that re ects the severity of infection, including the in ammatory response and risk of death [39,40]. Also, IL-8 production is almost entirely dependent on the type III secretion of exotoxin [41,42] which is a major determinant of virulence and is associated with disease severity in the infected host [43,44]. Given this, the detection of viable P. aeruginosa with infectivity would seem to be essential in understanding the infectiveness, in ammation induction, and pathogenesis in the host. The qRT-PCR method can measure only live bacteria, as it targets rRNA that is rapidly degraded in dead cells [21]. Given this, we think that the bacterial count as measured by the qRT-PCR method closely re ected the extent of the in ammatory response of infected cells. In contrast, with qPCR, which targets DNA, the correlation between the detected P. aeruginosa number and the in ammatory response of infected cells was weak. Moreover, we observed that DAPI staining, which targets DNA, and uorescence microscopy images remained fairly constant even at the death phase, so dead P. aeruginosa continued to be observed after antibiotic treatment (Fig. 3B). Therefore, with qPCR, the poor correlation between the detected P. aeruginosa number and the in ammatory response may be attributed to the detection of both viable and dead cells. The discrepancy between bacterial count results with culture methods and the in ammatory response of infected cells is thought to be due mainly to a decrease in the ability of P. aeruginosa to form colonies in a viable but nonculturable state [45,46] under antibiotic treatment.
Previous reports indicated that more than 70% of ICU patients receive antimicrobial treatment and are therefore susceptible to development of drug-resistant P. aeruginosa infection [47,48]. In this study, we demonstrated that intestinal infections of drug-resistant P. aeruginosa in ICU patients can be quickly detected using qRT-PCR with the combined use of drug-resistance gene ampli cation primers. It has been reported that among antipseudomonas antibiotics, meropenem is associated with the highest risk of resistance emergence [49,50]. In Patient B in this study, who had received meropenem for more than 2 weeks, an elevated fecal P. aeruginosa count was observed and upregulation of drug resistance genes such as the bla IMP , ampC, and mexA gene was detected. About 80% of carbapenem-resistant P.
aeruginosa have been reported to be multidrug resistant [51]. Furthermore, P. aeruginosa actually proliferates in the intestines during the administration of levo oxacin [52]. Therefore, there is a high likelihood that the bacterium detected in the feces was MDRP. This indicates that qRT-PCR method may also be applicable for rapid and highly sensitive monitoring of such MDRP.
The P. aeruginosa counts with qRT-PCR system closely re ected the treatment background of ICU patients. Speci cally, the induction of abnormal proliferation of P. aeruginosa in the intestines was observed with the long-term use of various types of antibiotics. It has been shown that the eradication of P. aeruginosa is actually facilitated by discontinuing the administration of sulbactam/ampicillin, which is not active against P. aeruginosa [53,54], or broad spectrum cephem antibiotics (cefepime), for which drug-resistant P. aeruginosa has low sensitivity [55]. The eradication of P. aeruginosa in the intestine was also observed after the withdrawal of mechanical ventilator use. Thus, we demonstrated that the timely monitoring of fecal P. aeruginosa accompanying medical interventions is possible with this system.
In this study, abnormal proliferation of P. aeruginosa in the intestines occurred in ICU patients with decreased obligate anaerobes in the intestines after antibiotic administration, and a negative correlation was observed between obligate anaerobe counts and P. aeruginosa in the intestines. A relationship has been shown between dysbiosis [56] and intestinal infection of P. aeruginosa in ICU patients. When the intestinal microbiota of ICU patients in whom ventilator-associated pneumonia occurred was analyzed using YIF-SCAN® [21][22][23], there was a dramatic decrease in the counts of obligate anaerobes, and dysbiosis was shown to have occurred [30,57]. We have also previously shown that dysbiosis is a high-risk factor for mortality in critically-ill ICU patients [58]. Moreover, Robak et al. showed that microbiotadependent IgA production is required for antibacterial immunity during acute bacterial pneumonia [59].
These ndings indicate the importance of intestinal microbiota in ICU patients. Since both intestinal microbiota disturbances and P. aeruginosa infection dynamics can be understood simultaneously, the qRT-PCR method may be a very effective means of controlling the risk of infection in ICU patients.
There remain some limitations to our study. First, while we attempted to monitor drug-resistant P. aeruginosa with the concurrent use of 3 drug-resistance gene primers, P. aeruginosa has various other drug-resistance mechanisms. In the future, we anticipate that de nitive diagnosis and monitoring of MDRP will also be possible with the generation of aminoglycoside resistance and new quinolone resistance gene detection primers. Second, horizontal transfer between multiple pathogens containing the bla IMP gene has been reported; thus, the possibility that bacteria such as Enterobacteriaceae or Acinetobacter baumannii, which show carbapenem resistance, have the bla IMP gene cannot be ruled out.
We think it will also be possible in the future to estimate gene origins by preparing detection primers for those genes and monitoring uctuations in bacterial counts. Third, in the study of monitoring drug resistant P. aeruginosa in the feces of ICU patients, the sample size was small. Further studies with a larger sample size are necessary to clarify the relationship between the uctuations of P. aeruginosa detected by qRT-PCR method and clinical course.

Conclusion
The features of this new P. aeruginosa quanti cation system based on 23S rRNA-targeted qRT-PCR are as follows: (1) it has very high speci city and sensitivity; (2) since it can quantify bacteria directly from clinical specimens, the time required for the measurement is about 6 h, making early diagnosis of P. aeruginosa infection possible; (3) compared with conventional detection methods, the number of viable P. aeruginosa can be accurately determined in patients receiving antibiotics; and (4] the results obtained accurately re ect the various treatment backgrounds in ICU patients. The present study shows the potential of the method described herein as an effective tool for controlling nosocomial infections of P. aeruginosa as well as for judging the e cacy of antibiotic treatment and determining whether to continue or withdraw other medical actions.  Quanti cation of P. aeruginosa spiked to human blood or feces using qRT-PCR and culture method. (A) Commercially available human peripheral blood was spiked with P. aeruginosa ATCC 10145T at nal concentrations ranging from 100 to 104 cells/ml. The total counts of P. aeruginosa were determined by qRT-PCR and the culture method. For qRT-PCR counts, the newly developed primer set for P. aeruginosa was used. For the culture method, an appropriate dilution series of each blood sample was cultured on Comparison of detection methods in generating time-kill curves for MDRP exposed to antimicrobials. (A) Time-kill curves for P. aeruginosa ATCC BAA-2108TM exposed to colistin and doripenem combinations (n = 4). Bacterial counts in the culture solutions sampled over time were determined by qRT-PCR, culture method, and qPCR. For qRT-PCR and qPCR, the primer set targeting P. aeruginosa-speci c 23S rRNA was used. For the culture method, the appropriate dilution series of cultures were cultured on TSA plates. (B) A part of the culture solutions sampled at 0, 48, and 336 h after cultivation were subjected to measurement of bacterial counts and assayed for FISH with FITC-labeled Eub338. Green and red images of P.
aeruginosa were obtained by DAPI and FISH, respectively. Scale bars: 10 μm. (C) After Caco-2 cells were infected with a part of the culture solution used for determination of bacterial counts, relative amounts of IL-8-encoding transcripts were measured by qRT-PCR. Expression was normalized to GAPDH, and the relative gene expression values were compared to those of uninfected cells (n = 4). Also, the production of IL-8 was measured by ELISA (n = 4). (D) The relationship between P. aeruginosa counts by three different methods and the in ammatory response in infected Caco-2 cells were investigated. The upper diagram shows the correlation between the P. aeruginosa counts detected by qRT-PCR (left), culture method (center), and qPCR (right) and the relative mRNA expression of IL-8 in Caco-2 cells. The lower diagram shows the correlation between the P. aeruginosa counts and IL-8 production. Pearson's correlation coe cient was applied to the dataset. Data indicate the mean ± SEM. When the bacterial count was not detected (< 1 cell/ml), the plot was outlined in (A) and (D).

Figure 4
Monitoring of drug-resistant P. aeruginosa by qRT-PCR with the newly designed primer set re ects the clinical treatment. qRT-PCR quanti cation of P. aeruginosa using the P. aeruginosa-speci c 23S rRNA primer set (brown) and simultaneous detection of drug resistance genes (blaIMP, ampC, mexA) using