Quantitative PCR and culture evaluation for enterotoxigenic Escherichia coli (ETEC) associated diarrhea in volunteers
Brianna R. Lindsay,
Department of Epidemiology and Public Health, University of Maryland, School of Medicine, Baltimore, MD, USA
Correspondence: Brianna R. Lindsay, University of Maryland School of Medicine, 660 W. Redwood St. Howard Hall Rm 133A, Baltimore, MD 21201, USA. Tel.: +1 410 706 6323; fax: +1 410 706 1644; e-mail: firstname.lastname@example.org
Recent evidence suggests that the abundance of enteric pathogens in the stool correlates with the presence of clinical diarrhea. We quantified the fecal pathogen after feeding enterotoxigenic Escherichia coli (ETEC) strain H10407 to 30 adult volunteers. Stools were collected daily and examined using qualitative and quantitative (Q) culture. DNA was isolated, and quantitative (Q) PCR targeting the heat-labile toxin (LT) gene was performed. Nine volunteers developed diarrhea. Among 131 stool specimens with complete data, pathogen abundance by QPCR was strongly correlated with Qculture, ρ = 0.61, P < 0.0001. Receiver operating characteristic curve analysis comparing quantitative data against diarrhea status suggested cut-points, based on a maximum Youden Index, of 2.8 × 104 LT gene copies and 1.8 × 107 CFU. Based on these cut-points, QPCR had a sensitivity and specificity compared with diarrheal status of 0.75 and 0.87, respectively, and an OR of 20.0 (95% CI 5.7–70.2), whereas Qculture had a sensitivity and specificity of 0.73 and 0.91, respectively, and an OR of 28.6 (95% CI 7.7–106.6). Qculture had a sensitivity and specificity of 0.82 and 0.48, respectively and an OR of 4.4 (95% CI 1.2–16.0). The correlation between Qculture and QPCR was highest in diarrheal specimens, and both quantitative methods demonstrated stronger association with diarrhea than qualitative culture.
When attempting to understand the cause of a diarrheal episode, recent evidence suggests that it is important to examine not only whether pathogenic bacteria are present in the stool specimen, but also to estimate the quantity of that pathogen (Barletta et al., 2011; Lima et al., 2013; Lindsay et al., 2013). This understanding relates to the nature of immunity which can limit the extent of colonization of enteric bacteria, even though immunity may not completely prevent colonization. This was shown recently when volunteers were challenged and then some were rechallenged with an enterotoxigenic Escherichia coli (ETEC) strain H10407 (Harro et al., 2011). About 70% of naïve volunteers developed diarrheal illness. The concentration of the challenge strain averaged 1 × 108 CFU in fecal specimens from these ill volunteers. When these volunteers were rechallenged, none of them developed moderate or severe diarrhea. Interestingly, however, although the challenge strain was cultured from nearly all the volunteers, the concentration of the challenge strain was c. 100-fold lower.
Children in developing countries encounter many episodes of ETEC, and previous infections with these bacteria expressing certain antigens stimulate immune responses which provide partial protection from subsequent infections (Steinsland et al., 2003; Qadri et al., 2007; Darsley et al., 2012). In a study of volunteers who had been immunized with an experimental ETEC vaccine, ACE527, and then challenged with strain H10407, immune volunteers had lower concentrations of the challenge strain in their feces than nonimmune volunteers (Porter et al., 2011). These studies suggest that immunity can reduce the quantity of ETEC, but that immunity does not prevent colonization; the pathogen is therefore passed in the stool. In the course of epidemiologic studies in developing countries, therefore, it may be valuable to utilize pathogen quantification to distinguish which cases of diarrhea may be caused by a certain pathogen, particularly in episodes yielding more than one pathogen. To appreciate the value of quantifying the pathogen, we compared qualitative and quantitative methods in volunteers who ingested a known diarrheal pathogen.
Materials and methods
The study utilized fecal specimens from an ongoing volunteer study of ETEC diarrhea (ClinicalTrial.gov Identifier#NCT00844493) (Harro et al., 2011). Thirty subjects were recruited and randomized 1 : 1 in two groups to receive a single oral dose of strain H10407 in bicarbonate buffer at an approximate dose of 1 × 105 CFU or 1 × 106 CFU using methods described previously (Harro et al., 2011). The clinical protocol was approved by the Institutional Review Board of Johns Hopkins Bloomberg School of Public Health & Institutional Biosafety Committee, and the Western Institutional Review Board (Olympia, WA). Use of the challenge strain was approved under BB-IND#12 234.
Healthy, 18- to 45-year-old, male or female subjects (8 females and 22 males, average age 33) were recruited for the study using print and electronic media advertisement and by word of mouth. The prechallenge health status of subjects was assessed by written and oral medical history, physical examination, complete blood count, urinalysis, urine toxicology, blood chemistries, and tests for liver and kidney function, HIV-1, hepatitis B, and hepatitis C. Subjects were excluded if they had significant medical problems detected by history, physical examination, or screening laboratory tests, abnormal stool pattern on a regular basis, a history of diarrhea in the 2 weeks prior to planned inpatient phase, regular use of laxatives, antacids, or other agents to lower stomach acidity, use of antibiotics during the 7 days before dosing, travel to countries where ETEC or cholera infection is endemic within 2 years prior to dosing, a history of vaccination for or ingestion of ETEC, cholera, or LT toxin, or a stool culture positive for ETEC or other bacterial enteric pathogens. Volunteers were admitted to the Center for Immunization Research Isolation Unit the day prior to challenge and were not allowed any oral intake for 9 h before challenge and the following 90 min after challenge. On the day of challenge, subjects drank 120 mL of 1.33% sodium bicarbonate buffer prior to ingesting the ETEC inoculum with 30 mL of the same buffer. All subjects were treated starting on study day 5 with ciprofloxacin (500 mg by mouth twice daily for 3 days). Early antibiotic treatment was provided to subjects who had severe diarrhea, moderate diarrhea for 2 days, and mild or moderate diarrhea with two or more of the following symptoms: fever (100.4 °F), vomiting, and certain severe constitutional symptoms, including abdominal pain/cramping, headache, myalgias, or nausea. Diarrhea was defined as one loose/liquid stool of 300 g or two loose/liquid stools totaling 200 g during any 48-h period within 120 h of challenge. Stools were collected on days 0–7, 9, 28, and 84.
Strain H10407 is a wild-type virulent ETEC, serotype O78:H11 and produces both heat-labile toxin (LT), heat-stable toxin (ST), and colonization factor antigen 1 (CFA/I). It is sensitive to ampicillin, trimethoprim-sulfamethoxazole, and ciprofloxacin. It has been used in previous challenge studies and is considered to be more virulent than other ETEC strains which have been used (Porter et al., 2011).
After challenge, from day 0 to day 4, before administering antibiotics, up to two stool specimens were collected per day. Colonization was defined as detection of strain H10407 in two stool specimens collected at least 24 h after challenge. For quantitative (Q) culture, 1 gram of fecal sample was first diluted 10-fold up to 10−5 in PBS. Aliquots, consisting of 0.1 mL, of these dilutions were spread onto MacConkey agar. After overnight incubation, the colonies of bacteria which appeared to be E. coli were counted and the proportion of these colonies (of 5 colonies tested), which agglutinated with anti-H10407 antiserum was recorded. The polyclonal anti-H10407 antiserum was prepared with rabbits at the International Centre for Diarrheal Disease Research, Bangladesh, using formalinized ETEC H10407 cells as the immunogen. Quantity of H10407 was expressed as CFU per gram of stool.
DNA isolation and quantitative polymerase chain reaction
DNA was isolated from frozen stools using a bead beater with 3-mm-diameter solid-glass beads (Sigma-Aldrich) and subsequently 0.1 mm zirconium beads (BIO-SPEC Inc.) to disrupt cells. The cell slurry was then centrifuged at 16 000 g for 1 min, the supernatant processed using the Qiagen QIAamp® DNA stool extraction kit. Extracted DNA was precipitated and shipped to the University of Maryland, Baltimore.
QPCR was conducted on 301 samples using the Applied Biosystems 7500/700 Fast Real-Time PCR System with software V2.0.5 and SYBR Green-Based fluorescent dye, by a single technician who was blinded to diarrheal status. Each DNA was run at a minimum in triplicate, and results were averaged. The assay contained 10–200 ng of stool DNA, 8.5 uL of water, 1.5 uL each of 5 uM forward and reverse primers designed for LT, and 12.5 uL of SYBR Green PCR Master Mix (Applied Biosystems) (Taniuchi et al., 2012). PCR was carried out for 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Gene copies were determined by absolute quantification using standard curve fitted for each 96-well plate. The standard curve was constructed from E. coli strain H10407, ATCC#35401. R2 values of a linear model fit to the standard curves of cycle threshold (Ct) vs. log dilution of DNA in the standard ranged from 1.00 to 0.988. To convert quantity given by the QPCR output to number of target gene copies, we computed that one nanogram of DNA is approximately equal to 2 × 105 gene copies. We used the original concentration of DNA, determined spectrophotometrically (Nanodrop 1000, ThermoScientific) to estimate the number of gene copies based on 100 ng of total stool DNA.
All statistical analyses were performed using sas Version 9.2 (SAS Institute, Cary, NC). Means and standard deviations were calculated using the log base 10 transformed values. Statistical comparisons were made between the mean quantity of ETEC measured by Qculture and QPCR using a Wilcoxon rank-sum test. Dependence among multiple samples from each individual was taken into account using a repeated-measures anova on means that included more than one specimen per subject. For the repeated-measures analysis, we only included individuals with complete data on days 0 through 4 (N = 25). The correlation between QPCR and Qculture was estimated by Spearman correlation coefficient.
Sensitivity and specificity were calculated for i) qualitative culture, Qculture given a cut-point, and QPCR given a cut-point compared with whether the specimens were collected from volunteers with diarrhea on the day of collection and ii) qualitative culture (Qculture) given a cut-point and QPCR given a cut-point compared with whether the volunteer that provided the specimens had diarrhea on any day. Sensitivity and specificity were calculated by comparing the results of the reference (diarrheal status) to QPCR or culture. Sensitivity was calculated as the number of true positives (positive by both classifications) divided by the number of reference positives. Specificity was calculated as the number of true negatives (negative by both classifications) divided by the number of reference negatives. We constructed receiver operating characteristic (ROC) curves to determine the cut-point from the continuous measurements of the number of gene copies per sample or the number of CFU from culture by plotting the estimated sensitivity by 1-specificity at each increment increase in the number of gene copies or number of CFU's. The optimal cut-point for Qculture and QPCR was determined by the maximum Youden Index, J = (Sensitivity + Specificity−1; Akobeng, 2007). Using these cut-points, the odds of ETEC identification in cases compared with the odds in controls were computed using logistic regression.
Of the 30 participants receiving ETEC, nine reported any diarrhea within 4 days. Two participants reported initial diarrhea on the first day following challenge, three on the second day, and four on the third. Fifteen volunteers received a dose of 105, of which three developed diarrhea, while the remaining 15 received a dose of 106, of which six went on to develop diarrhea. There did not appear to be a difference between the day the subject developed diarrhea and the dose given (Fig. 1). For the following analysis, we excluded 19 time points because either there was no stool (10 patients on day zero; 2, day one; 1, day two; 2, day three) or the patient received antibiotics (Ab) before day 4 (one given Ab on day two; 1 Ab, day three). This resulted in a final sample size of 30 subjects and 131 stool specimens.
Seventy-three of 131 specimens tested positive for H10407 by qualitative culture, and of these, 14 (19%) corresponded with a report of diarrhea on the day they were collected. If a specimen tested positive qualitatively for ETEC, there were 4.35 times (95% CI 1.18–16.0) the odds that diarrhea would be reported on the same day compared with a sample testing negative. Twenty-five (34%) qualitative-culture-positive specimens corresponded with a report of diarrhea on any days from 0 through 4. There was a significant correlation between the proportion of colonies out of five picked that agglutinated with H10407 serum and any diarrhea over 4 days (Pearson ρ = 0.23, P = 0.03).
Complete data for QPCR and Qculture were available for 131 specimens. The maximum estimated from Qculture was 1.0 × 109 CFU per gram of stool and from QPCR was 1.3 × 109 LT copies per 100 ng of stool DNA. Both methods detected a greater quantity of ETEC in diarrheal compared with nondiarrheal specimens collected on the same day. As shown in Fig. 2, in both diarrheal specimens and specimens from subjects not reporting diarrhea, the mean on each day measured by QPCR increased, except on day 4 in nondiarrheal specimens where it decreased; in contrast, using Qculture, the mean did not have a consistent trend. There was a significant difference in the means over time within subjects for QPCR performed on diarrheal samples compared with controls (P = 0.0001), but no such significant difference measured by Qculture (P = 0.52).
Qculture and QPCR results were strongly correlated (Fig. 3). The overall Spearman correlation was 0.61 (P < 0.0001), and a higher correlation was seen in diarrheal specimens, 0.89 (P < 0.0001), than among those not reporting diarrhea, 0.51 (P < 0.0001). An alternative analysis splits the samples into groups according to whether or not diarrhea was reported on the following day. This analysis showed that the correlation between QPCR and Qculture is high (next day diarrhea ρ = 0.95, P < 0.0001 vs. no diarrhea next day ρ = 0.54, P < 0.0001). In addition, the overall mean of ETEC measured by both Qculture and QPCR seems to differ between these two groups more than if the groups are identified by diarrhea on the day the sample was taken. For culture, the overall geometric mean for specimens from volunteers with diarrhea the next day was 1.8 × 106 CFU, whereas the geometric mean for volunteers with no diarrhea the next day was 1 × 103 CFU (P = 0.0005). For QPCR, the overall geometric mean for volunteers with diarrhea the next day was 1 × 104 LT copies; while for specimens with no diarrhea the next day, it was 5 × 101 LT copies (P = 0.0006). The difference between the geometric mean in those with diarrhea the next day compared with those without diarrhea the next day was approximately three orders of magnitude for both Qculture and QPCR.
ROC curves were constructed from the sensitivity and 1-specificity for quantity of ETEC estimated by Qculture and number of gene copies determined by QPCR using diarrhea status on each day as the reference standard (Fig. 4). The optimal cut-point for predicting diarrhea for Qculture was 1.8 × 107 CFU, while for QPCR, it was 2.8 × 104 LT copies. Using these cut-points, the sensitivity, specificity, and odds ratio for each method were calculated. The sensitivity for Qculture (0.73; 95% CI 0.51–0.96) was similar to that for QPCR (0.75; 95% CI 0.54–0.96), as were the specificities for Qculture (0.91; 95% CI 0.86–0.96) and QPCR (0.87; 95% CI 0.81–0.93). The odds ratio for Qculture was 28.6 (95% CI 7.7–107) and QPCR, 20.0 (95% CI 5.7–70.2). Qualitative culture has the highest sensitivity 0.82 (95% CI 0.64–1.00), lowest specificity 0.48 (95% CI 0.39–0.57), and lowest odds ratio 4.35 (95% CI 1.2–16.0).
Complete QPCR results on 29 subjects at days 0–7, 9, 28, and 84 are shown in Fig. 5. The mean QPCR LT abundance at each day in individuals who had diarrhea that were given an antibiotic on day 3 (n = 2), those who had diarrhea and were given an antibiotic on day 4 (n = 6), and those who did not have diarrhea and were given antibiotic on day 4 (n = 21; the one individual given antibiotic at day 2 who had no QPCR data for day 1 was excluded). Individuals given an antibiotic on day 3 had a much higher mean peak number of LT gene copies at day 3 than those given an antibiotic on day 4. Decreasing levels of LT were detected after the administration of antibiotic. Those with no diarrhea had the lowest measured levels of LT gene copies by QPCR.
Population-based studies which compare quantitative and qualitative estimates of pathogens in stools are complicated by the fact that many pathogens or even a combination of pathogens may be truly responsible for symptoms associated with infection (Kotloff et al., 2013). This study evaluated the use of quantitative technologies following the administration of a known pathogen to volunteers, where we can be more certain that the relationship between the administered pathogen and diarrhea is causal. We found that Qculture and QPCR produce similar results. The geometric mean in samples that were collected on a day where the subject reported diarrhea was higher than the mean in samples that were collected on days where no diarrhea was reported. The correlation between Qculture and QPCR was highest in the diarrheal specimens, and both quantitative methods demonstrated stronger association between measurement and clinical presentation than qualitative culture.
Dichotomous cut-points of the Qculture and QPCR were calculated using whether the volunteers had diarrhea as the reference. Using diarrhea on that measurement day as the reference, Qculture had a similar sensitivity and specificity to that of QPCR. Given these results, we can conclude that Qculture and QPCR are comparable, but neither method was clearly superior to the other. Considerations such as laboratory capabilities, processing throughput and ease of use may be some issues to take into consideration when determining which quantitative method may be most appropriate (Giocoli et al., 2009).
A strength of our study was that we were able to perform analyses longitudinally because samples were collected on multiple days. A repeated-measures anova found that the difference over time within an individual between the number of LT gene copies measured by QPCR varied between those reporting diarrhea and those not reporting diarrhea. A disadvantage of a volunteer study is the small sample size which limits our power to observe statistically significant results. Using an ROC analysis to determine a threshold for clinical relevance has both limitations and advantages. This gives a clear dichotomization into clinically relevant groups; however, the applicability of this threshold to other settings may be limited. Our experimental study design has the advantage of knowing the time and magnitude of the exposure to ETEC, and we can be certain that it was the cause of diarrhea. In observational settings, this is not the case.
Previously, we examined the utility of QPCR in an observational setting using Shigella as our pathogen of interest, and it appears to result in measurements that are comparable to culture (Lindsay et al., 2013). In an observational epidemiologic scenario, the threshold for a clinically relevant infection due to ETEC may be less clear and obfuscated by the presence of other pathogens. A limitation of our methodology is that ROC analysis does not take into account dependence due to repeated measurements for each subject. We have accounted for this dependence in our repeated-measures analyses and logistic regression, but concede that in the approximation of these thresholds dependence has not been taken into account. Additionally, given our limited sample size, we were unable to observe differences due to ETEC dose and further research is warranted examining a dose–response relationship in pathogen quantitation.
Quantifying ETEC using Qculture or QPCR and the use of a clinically relevant threshold gives a more definite diagnosis of ETEC-associated diarrhea compared with qualitative culture. Qualitative culture, with the highest sensitivity and lowest specificity, produced the most false positives. Our analysis examined the occurrence of diarrhea on the same day, and on the following day, the sample was collected, and on any of the days, measurements were taken. Each analysis seemed to suggest that detection in the stool by Qculture or by QPCR is increased in diarrheal specimens and subjects that developed diarrhea. Further, our results may suggest that the greatest difference in the stool specimens may occur a day prior to developing diarrhea.