Estimating the Sensitivity and Specificity of Real‐Time Quantitative PCR of Fecal Samples for Diagnosis of Rhodococcus equi Pneumonia in Foals

Abstract

Background

Real‐time, quantitative PCR (qPCR) methods for detecting Rhodococcus equi in feces have been developed as a noninvasive, rapid diagnostic test for R. equi pneumonia, but have not been evaluated in a large population of foals.

Objective

The objective of this study was to evaluate the clinical utility of fecal PCR as a diagnostic test for R. equi pneumonia in foals using receiver operating characteristic (ROC) methods.

Animals

186 foals born in 2011 at an R. equi‐endemic ranch in Texas.

Methods

Fecal samples were collected at the time of onset of clinical signs for pneumonic foals (n = 31). Foals with pneumonia were matched by age and birth date to healthy (n = 31) and subclinical (n = 124) control foals; fecal samples were collected from these controls. DNA was extracted from feces using commercial kits and concentration of virulent R. equi in feces was determined by qPCR.

Results

Concentration of R. equi in feces differed significantly (P < .05) among groups. The area under the ROC curve for fecal qPCR for diagnosis of R. equi pneumonia was 89% (95% CI, 83–99), with a sensitivity of 94% and specificity of 72%.

Conclusions and Clinical Importance

qPCR of feces can be useful as an alternative to tracheobronchial aspiration for the diagnosis of R. equi in foals with clinical signs of pneumonia. Caution should be used in extrapolating results of this study to other populations because fecal concentration of R. equi might vary by geographic location or management practices.

Abbreviations

qPCR

quantitative polymerase chain reaction

ROC

receiver operating characteristic

TBA

tracheobronchial aspirate

TMD

total maximal diameter

TUS

thoracic ultrasound

Rhodococcus equi is a gram‐positive, facultative, intracellular bacterium and the most important cause of severe pneumonia in foals, leading to substantial animal suffering and economic losses.1 Both virulent and avirulent biotypes of R. equi have been identified.2 Virulent isolates contain an 85‐ to 90‐kb plasmid that encodes the virulence‐associated protein A (VapA) that is necessary to cause disease in foals. Virulent isolates can be identified by detection of the vapA gene using polymerase chain reaction (PCR).2, 3, 4, 5, 6, 7, 8 Real‐time qPCR is used for the detection of R. equi, because this bacterium is relatively slow growing and can be difficult to cultivate from biological samples such as feces or soil.9

Definitive diagnosis of R. equi infection in foals with signs of lower respiratory tract disease is based on bacterial culture or PCR amplification of DNA from a tracheobronchial aspirate (TBA), in combination with cytologic evidence of septic pneumonia in the TBA fluid.1, 8 The TBA procedure is invasive, labor‐intensive, requires skill, carries risks to foals, and is relatively expensive. In addition, it may take up to 3 days to obtain results. For these reasons, many practitioners avoid performing a TBA in foals in a farm setting and rely on a presumptive diagnosis of R. equi pneumonia on the basis of signalment, clinical signs, and farm history. Thus, a noninvasive diagnostic test for R. equi pneumonia would be clinically valuable. To date, however, a noninvasive test has remained elusive. Serologic detection has been demonstrated to be inaccurate,8, 10, 11 and accuracy of PCR of blood samples has not been reported.

Detection of virulent R. equi in feces by PCR was demonstrated to be highly sensitive in a small number of foals,12 but the need exists to further evaluate the accuracy of fecal PCR testing. Interpretation of fecal PCR results is complicated by the ecology of R. equi. Rhodococcus equi is shed in large quantities in the feces of both affected and unaffected foals, with up to 10⁶ colony‐forming units (CFU) of R. equi per gram detected as determined by quantitative bacterial culture, gel electrophoresis, and immunoblot analysis.2, 13, 14, 15, 16 Thus, there is a potential for false‐positive results of fecal PCR for R. equi.

The specificity of fecal testing is further complicated by the presence of foals with subclinical R. equi pneumonia that will not develop disease but may shed virulent R. equi in feces. Screening methods are used in hopes of earlier identification of foals that will develop R. equi pneumonia based on the rationale that earlier intervention will lead to higher success and shorter duration of treatment. Sequential thoracic ultrasonography (TUS) is a highly sensitive screening method used to detect peripheral lung abscesses.17 However, the use of TUS has indicated that many foals with findings consistent with R. equi infection will not develop clinical signs.18 ^{, 1}

To the best of our knowledge, real‐time qPCR of virulent R. equi in foal feces has not been used to differentiate between affected and unaffected foals in a population with known disease outcomes. The purpose of this study was to compare the results of qPCR for R. equi among foals with clinical pneumonia, subclinical pneumonia, or no evidence of clinical or subclinical pneumonia. We hypothesized that the quantity of virulent R. equi shed in the feces of foals at an R. equi‐endemic farm would differ significantly among foals of the 3 groups in an ordinal manner (ie, clinical > subclinical > unaffected). We further hypothesized that a cut‐point could be established for PCR testing that would be accurate for diagnosing foals with R. equi pneumonia.

Materials and Methods

Study Population

This study used a repository of fecal samples and existing data from a population of foals born on an R. equi‐endemic ranch in Guthrie, Texas during 2011. All 270 foals born on the farm during 2011 that survived to weaning and that participated in a prior study19, 20 evaluating screening tests for R. equi pneumonia were eligible for inclusion in this study. Thoracic ultrasonography was performed biweekly on every foal, from 3 to 16 weeks of age. The total maximal diameter (TMD) of all pulmonary consolidations was recorded for each examination. On the day of every ultrasound examination, fecal samples were collected if available, shipped chilled overnight, and stored frozen at −20°C. For this study, foals were categorized as affected, subclinical, or unaffected. Affected foals were those with clinical signs suggestive of pneumonia (eg, fever, lethargy, cough, nasal discharge, polysynovitis, tachypnea, increased respiratory effort, respiratory distress or a tracheal rattle), microbiologic isolation of R. equi from a TBA, cytologic evidence of sepsis from TBA fluid, and sonographically visible pulmonary consolidations or abscesses. Tracheobronchial aspirates were collected transendoscopically on the first day of apparent clinical signs of pneumonia and submitted for cytology and aerobic culture; R. equi was recovered from all affected foals. An additional fecal sample was obtained from affected foals on the day of onset of clinical signs of pneumonia. Subclinical foals had lung lesions (consolidation or abscess formation) identified by TUS but did not develop clinical signs of pneumonia from birth through weaning. Those performing diagnosis and treatment of affected foals were blinded to the results of TUS screening; no intervention was implemented for any foals in this study unless they developed clinical signs of pneumonia. Unaffected foals were defined as those having neither lesions detectable by TUS nor clinical signs of pneumonia at any point during the study. Forty‐six (17%) foals developed R. equi pneumonia, 54 foals (20%) were unaffected, and 170 foals (63%) were subclinical. Protocols for this study were reviewed and approved by the Clinical Research Review Committee (CRRC Protocol 10–12) of the College of Veterinary Medicine & Biomedical Sciences, Texas A&M University. At the time the samples were collected for this study, research involving client‐owned animals at Texas A&M University was not subject to review by the Institutional Animal Care and Use Committee.

Sample Selection

Fecal samples collected on the day of onset of clinical signs were available for 31 of the 46 affected foals. Each of these 31 affected foals was matched to an unaffected foal based on 2 criteria: age at sampling (+/− 7 days) and date of birth (+/− 30 days), to account for the effects of age and birth‐month on fecal shedding. The fecal sample selected from each control foal was obtained when the foal was closest in age to its index affected foal on the day of onset of clinical signs. In the event of a tie between 2 foals regarding their selection criteria, selection priority was given to age at sampling. Each affected foal also was matched to 4 subclinical foals on the basis of age and date of birth, as described above. For each affected foal, 2 subclinical foals were selected whose lesions had a TMD ≥ 200 mm at the time of sample collection and 2 subclinical foals with a TMD < 200 mm on all ultrasonographic examinations. The rationale for including the subclinical foals with different lesion sizes was that it was considered plausible that foals with larger subclinical lesions would shed larger numbers of virulent R. equi in feces than would foals with smaller subclinical lesions. The cut‐point of TMD = 200 mm was selected from receiver‐operator characteristic curve analysis conducted for the prior screening study performed using this population.19, 20

Fecal DNA Extraction

Frozen fecal samples were thawed and a 1‐g aliquot of feces was used for DNA extraction performed according to the manufacturer's recommendations.¹ DNA also was extracted from a vapA+ reference strain (American Tissue Culture Collection [ATCC] 33701p+) and an isogenic plasmid‐cured strain (ATCC 33701p‐) to be used as positive and negative amplification controls, respectively. Spectrophotometry was used to quantitate total DNA. Samples were diluted with nuclease‐free water to a standard concentration (20 μg/mL) and stored at −80°C until analyzed.

Standard Curve Construction

Absolute quantitation of DNA copy number required construction of a standard curve. The purified plasmid (pGEX‐2T; 4,948 bp) was obtained from an E. coli clone containing an R. equi strain 103+ plasmid vector with the vapA gene (570 bp), generously provided by Dr. Steeve Giguère, University of Georgia. Plasmid DNA was extracted using a commercially available kit² and the concentration was determined using spectrophotometry. Ten‐fold serial dilutions of plasmid DNA (approximate range, 10¹–10⁹ copies of pGEX‐2T/μg) were prepared in nuclease‐free water. Plasmid DNA standards were processed in duplicate using real‐time, qPCR. A standard curve was constructed using linear regression analysis of the log₁₀ quantity of pGEX‐2T copies per sample and the corresponding C _T values. A strong linear correlation (R ² = 0.999; slope = −3.489) existed between the initial copy number of vapA and the corresponding C _T value, and the amplification efficiency was 94%.

Real‐time, Quantitative PCR

Real‐time, qPCR was performed as previously described.5 Briefly, 2 μL of fecal DNA was added to 5 μL of a commercial mastermix,³ 0.5 μL of a custom premix,⁴ and 2.5 μL of buffered nuclease‐free water. The primer sequences used were previously designed5 based on the 564‐bp coding sequence of vapA for R. equi strain 33701 and chosen for their specificity for vapA.⁵ Samples were processed in a real‐time, qPCR unit⁶ using commercial software.⁷ Samples were tested in duplicate and measured against a standard curve of the vapA gene for absolute quantitation of copy numbers. Copy numbers were expressed as total copies per gram of feces. An assumption underlying our analysis was that each bacterial cell carried only 1 copy of the plasmid on which the vapA gene was encoded. Copy numbers of vapA also were calculated per g of feces based on the following formula:

$$\text{initial total DNA concentration}(ng/\mathrm{\mu }L)\times 100\mathrm{\mu }L/sampleweight(g).$$

where 40 ng was the total amount of DNA used per PCR reaction and 100 μL was the total volume for each DNA extraction procedure. Assuming that each R. equi has 1 copy of vapA, the copy number per g of feces is equivalent to the number of R. equi per g of feces.

The intra‐assay coefficients of variation (CV) determined from 10 replicates each of dilutions containing 10⁴, 10⁶, and 10⁸ CFU of R. equi as previously described5 were <5%. The interassay variation determined using 3 distinct assays as previously described5 had a mean overall CV <2% and a mean CV across all dilutions of <1%.

Data Analysis

Analyses were performed using S‐PLUS statistical software.⁸ For descriptive purposes, plots and summary statistics (including 95% confidence intervals) were used. Receiver‐operator characteristic (ROC) curves were plotted using the pROC procedure.21 Comparisons of fecal copy number of R. equi vapA among clinical status groups (ie, affected, subclinical, and unaffected) were made using generalized linear modeling with the log₁₀‐transformed concentrations of vapA detected in feces as the outcome variable and study group as the dependent variable. The healthy control foals were the reference group for the linear model. The method of Sidak was used for posthoc pair‐wise comparisons among groups in models with more than 2 groups. Model fit was assessed by graphical examination of diagnostic plots of residuals. A value of P < .05 was considered significant.

Results

The concentrations of R. equi in feces differed significantly among groups. The fecal R. equi concentrations of healthy foals and foals with lesions <200 mm did not differ significantly (P > .05). However, subclinical foals with lesions >200 mm had significantly (P < .05) higher concentrations of R. equi in feces than did foals in the healthy or subclinical <200 mm groups, and significantly (P < .05) fewer copies than age‐matched foals that had R. equi pneumonia (Fig 1). Results were essentially identical when the data were analyzed as the number of copies per g of feces. (data not shown).

Figure 1.

Boxplot of copy numbers of vapA per gram of feces in foals by study group. The horizontal line with the triangle is the median, and the top and the bottom of the boxed extend to the 75th and 25th percentiles, respectively. The thin vertical lines extending to thin horizontal lines represent multiples of 1.75 times their respective interquartile range, and the circles with horizontal lines represent outliers. Control – unaffected foals (n = 31); Subclinical <200 – subclinical foals with a TMD < 200 mm (n = 62); Subclinical >200 – subclinical foals with a TMD > 200 mm (n = 62); Affected – affected foals (R. equi pneumonia; n = 31). Different letters denote a statistically significant difference between groups (P < .05).

The utility of fecal PCR as a diagnostic test was assessed using ROC methods. The comparison of primary interest was between foals that developed R. equi pneumonia relative to those that did not develop pneumonia (ie, the R. equi group compared to the other 3 groups of foals; Fig 2). The area under the ROC curve for using fecal PCR diagnosis of R. equi pneumonia was 89% (95% CI, 83–99). The optimal cut‐point identified from the ROC curve was 10^2.82 copies (660 copies), and this cut‐point had a sensitivity of 94% (95% CI, 84–100) and specificity of 72% (95% CI, 65–79).

Figure 2.

ROC curve (affected foals versus the other 3 groups) of the sensitivity versus (1 – specificity), both as %. The thin diagonal line represents a completely uninformative test, wherein the area under the curve (AUC) is 50%. The area under the ROC curve represented by the thicker line was 89% (95% CI, 83–99) which was significantly (P < .05) >50%. The optimal cut‐point identified from the ROC curve was 10^2.82 copies (660 copies), and had a sensitivity of 94% (95% CI = 84–100) and specificity of 72% (95% CI = 65–79).

Discussion

This study demonstrated that the use of real‐time, qPCR for detection of R. equi in feces might be useful for the diagnosis of R. equi pneumonia. Fecal PCR for virulent R. equi is not proposed to replace the use of TBAs in clinical practice. Bacterial culture and cytologic interpretation of TBA fluid is advantageous in that one can document septic inflammation in the airways and obtain isolates for antimicrobial sensitivity testing, the latter being particularly important to guide appropriate treatment of R. equi pneumonia. Because TBAs are not routinely performed in the field before initiation of antimicrobial treatment, fecal PCR may help practitioners to substantiate a diagnosis for foal pneumonia and thus make more informed treatment decisions. The importance of the decision of whether or not to treat with antimicrobials is underscored by evidence that resistance to macrolide antimicrobials is linked to mass treatment of foals.22 In the present study, qPCR had a sensitivity of 94% and specificity of 72% for the diagnosis of R. equi pneumonia, and an area under the ROC curve of almost 90%. These data indicate that fecal PCR could be used as a diagnostic tool yielding rapid results to direct therapeutic choices for equine practitioners. As with any diagnostic test, the positive predictive value (PPV) will vary with the pretest probability of disease. For example, a positive test result in a 2‐month‐old foal with clinical signs of pneumonia from a farm with an annual cumulative incidence of 20% and estimated to have a pretest probability of disease of 80% would have a PPV of 93%, whereas the PPV for a randomly selected foal from this farm (ie, pretest odds of 20%) would be only 27%.

At the time of diagnosis, foals with R. equi pneumonia shed significantly more virulent R. equi than either subclinical or unaffected foals. Furthermore, subclinical foals with larger lesions (TMD > 200 mm) shed significantly higher concentrations of virulent R. equi than unaffected foals and foals with smaller lesions (TMD < 200 mm), but lower concentrations than affected foals. Differences between subclinical foals with a TMD < 200 mm and those that remained unaffected throughout the course of the study were not statistically significant. Virulent R. equi was detected in the feces of all but 1 foal (0.5%). The high prevalence of shedding was not unexpected, because foals are exposed to virulent R. equi early in life23 from a number of sources, including soil, shared air space, and the manure of dams.13, 24, 25 Clinically healthy dams shed 10² to >10³ CFU/g.13 In addition, R. equi uses volatile fatty acids in the feces and can multiply 10³‐fold in the environment.2 These data illustrate that identification of fecal shedding of virulent R. equi is not synonymous with a diagnosis of disease caused by the organism.

The sensitivity and specificity of fecal PCR for the diagnosis of R. equi previously has been described in a study12 that evaluated fecal shedding of virulent R. equi in foals with clinical signs of pneumonia. The sensitivity and specificity of fecal PCR for R. equi pneumonia in that study were 75% and 100%, respectively. The lower specificity in our study might be attributable to the population of subclinical foals with large pulmonary lesions. Foals with larger lesions, along with those that have clinical pneumonia, might have a relatively high concentration of virulent R. equi in their sputum that is being swallowed and passed through the gastrointestinal tract into their feces. Another explanation for the difference in estimated specificities is that a relatively small number of unaffected foals was used to calculate specificity in the aforementioned study and thus the precision of the estimated specificity was low.12

An important limitation of our study was that the population was restricted to a single, R. equi‐endemic breeding farm. Geography, the equine population, and management conditions all have the potential to affect fecal shedding of R. equi. Thus, the cut‐points determined in this study for the detection of R. equi pneumonia may not be applicable to other farms and further work is needed to determine the degree of variation in these cut‐points. In addition, many of the subclinical foals in this study had large pulmonary lesions, which may not be characteristic of R. equi infections on other farms.

There are a number of other factors that could affect the cut‐point for detection of R. equi pneumonia. The cut‐point identified in this study may not be applicable to a similar assay developed in another laboratory because of differences in methods of DNA extraction, sample handling, standard curve preparation, and PCR reagents. Furthermore, the use of frozen feces also may have affected our cut‐point. Freezing fecal samples may positively affect the extraction or stability of DNA from gram‐positive bacteria.26 Thus, a cut‐point for diagnosis could be lower if the assay was performed on fresh feces. These considerations emphasize the need for further evaluation of this method before it can be offered commercially by diagnostic laboratories for testing.

Another important limitation was the control groups used for estimating specificity. In this study, control groups included healthy foals and foals with subclinical pneumonia. Ideally, a control group comprised of foals from the same farm with clinical signs of pneumonia and pulmonary abscesses that cultured negative for R. equi and positive for another causative agent, such as Streptococcus equi subsp. zooepidemicus, would have been included. Unfortunately, such foals were not found within the study population during 2011. An advantage of the controls used in this study was the inclusion of subclinical foals that were not treated with antimicrobials. The large cohort of subclinical foals was evidence of widespread exposure to R. equi. This finding suggests that the specificity of this fecal qPCR assay for the diagnosis of R. equi pneumonia was good, even in the face of subclinical disease. Although TBAs were not performed on subclinical foals to rule out other infectious organisms, polymicrobial infection has been demonstrated in up to 83% of TBAs from foals with R. equi pneumonia and is not significantly associated with clinical outcome.27 Another limitation of our study is that we only obtained fecal samples from 31 of 46 foals (67%) with R. equi pneumonia. We do not believe that the absence of feces was nonrandom but we cannot exclude the possibility that results might have differed if we had obtained feces from all foals.

In conclusion, this study demonstrates that fecal qPCR might be useful for the diagnosis of R. equi pneumonia and may help guide treatment decisions when TBA samples are not collected. Fecal qPCR is a noninvasive technique with good diagnostic accuracy, making it a potential alternative to TBAs. Caution should be used in directly applying the results of this study to other populations because geographical and management factors may affect the magnitude of fecal shedding.