AEM
Home Help [Feedback] [For Subscribers] [Archive] [Search] [Contents]
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental material
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Muscatello, G.
Right arrow Articles by Browning, G. F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Muscatello, G.
Right arrow Articles by Browning, G. F.
Agricola
Right arrow Articles by Muscatello, G.
Right arrow Articles by Browning, G. F.

 Previous Article  |  Next Article 

Applied and Environmental Microbiology, September 2006, p. 6152-6160, Vol. 72, No. 9
0099-2240/06/$08.00+0     doi:10.1128/AEM.00495-06
Copyright © 2006, American Society for Microbiology. All Rights Reserved.

Associations between the Ecology of Virulent Rhodococcus equi and the Epidemiology of R. equi Pneumonia on Australian Thoroughbred Farms{dagger}

G. Muscatello,* G. A. Anderson, J. R. Gilkerson, and G. F. Browning*

School of Veterinary Science, The University of Melbourne, Parkville, Victoria 3010, Australia

Received 2 March 2006/ Accepted 28 June 2006


   ABSTRACT
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
The ecology of virulent strains of Rhodococcus equi on horse farms is likely to influence the prevalence and severity of R. equi pneumonia in foals. This study examined the association between the ecology of virulent R. equi and the epidemiology of R. equi pneumonia by collecting air and soil samples over two breeding seasons (28 farm-year combinations) on Thoroughbred breeding farms with different reported prevalences of R. equi pneumonia. Colony blotting and DNA hybridization were used to detect and measure concentrations of virulent R. equi. The prevalence of R. equi pneumonia was associated with the airborne burden of virulent R. equi (both the concentration and the proportion of R. equi bacteria that were virulent) but was not associated with the burden of virulent R. equi in the soil. Univariable screening and multivariable model building were used to evaluate the effect of environmental and management factors on virulent R. equi burdens. Lower soil moisture concentrations and lower pasture heights were significantly associated with elevated airborne concentrations of virulent R. equi, as were the holding pens and lanes, which typically were sandy, dry, and devoid of pasture cover. Few variables appeared to influence concentrations of virulent R. equi in soil. Acidic soil conditions may have contributed to an elevated proportion of virulent strains within the R. equi population. Environmental management strategies that aim to reduce the level of exposure of susceptible foals to airborne virulent R. equi are most likely to reduce the impact of R. equi pneumonia on endemically affected farms.


   INTRODUCTION
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Rhodococcus equi has been recognized as a pulmonary pathogen of horses for over 80 years (16), causing significant morbidity and some deaths in foals, typically between the ages of 1 and 4 months. R. equi may be avirulent or virulent in foals. Virulent isolates carry an 85- to 90-kb plasmid containing virulence-associated protein (vap) genes and have been isolated from the lungs of diseased foals (11, 26, 29). The virulence plasmid is required to produce disease in foals. Strains lacking these plasmids are regarded as avirulent or intermediately virulent in foals and are not associated with disease, (13, 28). Serological testing of horses has demonstrated a high prevalence of exposure to R. equi (30, 34). R. equi bacteria have been isolated from the soil of many horse farms. Virulent R. equi has been isolated on farms where R. equi pneumonia was considered endemic as well as on farms with no history of R. equi disease (17, 31). It has been suggested that the prevalence of virulent R. equi within the soil R. equi population and the concentration of virulent R. equi in soil affect the prevalence of R. equi pneumonia (24). However, factors that influence the survival and proliferation of virulent R. equi on a farm and its propensity to infect the lungs of foals are unknown. Assessment of the environmental virulent R. equi burden on farms has relied on isolation of bacteria from soil and subsequent assessment of the presence of the virulence-associated protein A gene (vapA) by PCR or of the VapA protein by immunoblotting (25, 33). The primary route of infection is thought to be inhalation of aerosolized virulent R. equi, so measurement of the airborne concentration of virulent R. equi, rather than of the concentration in soil, may yield a more meaningful assessment of the risk of disease on a farm.

We have developed a novel technique to measure airborne and soil concentrations of R. equi on farms (19), allowing quantitative assessment of environmental R. equi populations on farms. This work used techniques developed previously to examine the relationships between the concentration of virulent R. equi, the proportion of R. equi isolates that were virulent, environmental and management variables, and the prevalence of R. equi pneumonia on farms.


   MATERIALS AND METHODS
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Description of farms and sampling period.
Samples were collected from Thoroughbred breeding farms in the 2000 and 2001 horse breeding seasons. In the 2000 season, samples were collected from six farms in Victoria, Australia. In the 2001 season, samples were collected from 10 farms in Victoria and 12 farms in New South Wales (NSW), Australia. In total, three state-year combinations comprising 28 farm-year combinations were studied in the two seasons. R. equi pneumonia was endemic on some of the farms included in the study group, while other farms had no recent history of R. equi pneumonia prior to the study and did not report a single case of disease caused by R. equi during the study period. The area of the farm used for horse husbandry (farm size) ranged from 36 to 182 ha, and the maximum number of foals on the farms ranged from 42 to 488. The six Victorian farms sampled in 2000 were also sampled in 2001.

Samples collected.
In Victoria during the 2000 breeding season, monthly samples were collected on each farm between September 2000 and February 2001 inclusive. Four paddocks (fenced, spacious, outdoor fields covered with grass that are used to keep grazing horses) that contained mares and foals (mare-foal paddocks) were selected at random on each farm. Air and soil samples were collected from these paddocks each month throughout the study period. Foals on these farms ranged in age at the time of sampling from newborn to 4 to 5 months. Air and soil samples were collected around feeding stations, as these were areas where horses congregated to be fed, and consequently these areas had lower pasture cover and were dusty, conditions considered to be important in infection by many investigators (2, 10, 14, 20, 21, 32). Air samples were also collected monthly from holding pens (small outdoor fenced areas used to confine horses for convenience while procedures such as breeding of mares, veterinary treatment, and farriery are completed) and lanes (paths along which horses are moved to and from pens and paddocks), starting in October 2000. No soil samples were collected from pens or lanes in 2000.

During the 2001 season, air and soil samples were collected every 2 weeks from farms in Victoria. Samples were collected from four mare-foal paddocks, a holding pen, and a lane on each farm throughout the season. On farms sampled in the 2000 season, the same or similar paddocks were sampled in 2001, when possible. During the 2001 season, these samples were collected fortnightly from the middle of October 2001 to the end of January 2002. No samples were collected during the last 2 weeks of December 2001.

On farms in NSW, air and soil samples were collected only during November and December 2001, for logistical reasons. Samples were collected from four mare-foal paddocks, a holding pen, and a lane on each of the 12 farms. The paddocks were randomly selected in November, and these paddocks were sampled again the following month.

Foals.
The number of foals in each of the sampled paddocks was recorded at the time of sampling as a measure of paddock group (or mob) size. The total number of foals on the farm and their ages at the time of each sample collection were obtained from the farm manager. Each farm manager provided information on foaling dates and the number of foals on the property between specific dates coinciding with the times of sample collection. They also provided data on cases of R. equi pneumonia, including the age of the foal at diagnosis, the date of diagnosis, and the clinical outcome of each case. Cases of R. equi pneumonia had two or more of the following diagnostic features: clinical signs of pneumonia, ultrasonographically detectable pulmonary abscesses, leukocytosis with a neutrophilia and fibrinogenemia, and positive R. equi culture results from tracheal lavages.

Environmental samples.
Air samples were collected using a portable air monitoring system (M Air T; Millipore, Saint-Quentin-Yveline, France) loaded with a cassette containing NANAT medium (36). The air monitoring system was placed approximately 5 cm above the soil surface, and two or three samples of increasing volume were taken at each location (250 liters, 500 liters, and 1,000 liters). The sieve of the air monitoring system was immediately adjacent to the agar and contained perforations that facilitated delivery of air onto the agar surface. The sieve was disinfected with an isopropanol wipe (Isowipe; Kimberly-Clark, Milson's Point, NSW) before the collection of each sample. After incubation at 37°C for 48 h the plates from the largest volume of air sampled that had well-spaced colonies and did not have any fungal contamination were used for enumeration of environmental R. equi (representative cultures).

Adjacent to each air sampling site in the paddock, holding pen, or lane, two superficial soil samples (down to a depth of 5 cm) were taken using a soil auger. The auger was disinfected with an isopropanol wipe before each sample was collected. Samples were placed in sterile containers and sealed with an airtight lid for transfer to the laboratory. One sample was used for analysis of soil characteristics (pH, moisture, and texture), while the other was used for microbiological analysis. The samples for microbiological testing were air dried and placed at –70°C for storage. Diluted soil samples were later cultured to quantify R. equi in soil by dilution of 1 g in 9 ml phosphate-buffered saline and spreading 50 or 100 µl onto duplicate NANAT agar plates. After incubation at 37°C for 48 h, representative cultures were selected for enumeration of environmental R. equi, as for the air samples.

At least 10 g (wet weight) of soil from each sampling site was weighed and then incubated at 200°C for approximately 2 h. The samples were then reweighed, and the soil moisture was determined and expressed as a percentage of the wet weight. Approximately 10 g of the oven-dried soil was added to 25 ml of distilled water. Samples were then mixed thoroughly by vortexing or placed in a shaker at 200 rpm for 15 to 30 min. The pH was then determined using a pH meter (35). Soil that was not used for measuring moisture or pH was used to assess soil texture (sand or clay) based on methods described previously (18).

At the time of air sample collection from each paddock, the pasture (grass and/or clover) height was measured at 20 randomly chosen sites within a 10-m radius of the sampling site. The mean of the 20 measurements was used as an estimate of pasture height in the sampled area of the paddock. The lanes and pens were generally devoid of grass, and consequently no pasture height measurement could be made in these areas.

Mean temperature, humidity, and wind speed data for the period of the day when samples were collected were obtained for the Australian Bureau of Meteorology weather station nearest to each farm (within 30 km) from www.bom.gov.au.

Colony blotting and DNA hybridization.
Single representative air and soil samples from each paddock, pen, and lane on each farm at each time of sampling were selected for colony blotting and DNA hybridization to identify and differentiate R. equi colonies (19). In total, 768 air samples and 708 soil samples were blotted and probed. Blots were probed using radiolabeled PCR product amplified from the R. equi rrnA gene to identify R. equi colonies. The blot was then stripped and reprobed using radiolabeled PCR product amplified from the virulent R. equi vapA gene. This allowed quantitative evaluation of the R. equi population. Colony blotting was performed as described previously (19) with a virulent R. equi control (isolate number 7) and an avirulent R. equi control (isolate number 128) and a negative control (Corynebacterium ammoniagenes) incorporated into each blot.

Statistical analysis.
The number of CFU of R. equi bacteria per 1,000 liters was calculated for each air sample, and the number of CFU per milligram was calculated for each soil sample. The geometric means of the concentrations of R. equi and the concentrations of virulent R. equi were determined for each farm-year combination. A value of 1 was added to all data before the calculation of the geometric mean. The proportions of R. equi bacteria that were virulent were derived from these geometric means.

The Mann-Whitney test was used to evaluate the significance of associations between categories of R. equi concentration and the prevalence of disease caused by R. equi at the farm level. The farm levels were defined as the 28 farm-year combinations. The geometric mean concentration and the proportion of R. equi bacteria that were virulent in air and soil samples were categorized according to the approximate upper quartile points over the 28 farm-year units.

As the concentrations of R. equi and virulent R. equi in the soil and air samples were not normally distributed, Spearman's rank correlation coefficient (rs) was used to establish the strength of the correlation between paired soil and air samples for both the concentration of R. equi and the concentration of virulent R. equi, with an rs of >0.5 regarded as indicative of a strong correlation. McNemar's test was used to compare the proportion of soil samples that were positive for R. equi or for virulent R. equi to the proportion of air samples that were positive for R. equi or for virulent R. equi. An adjustment to McNemar's test to account for clustering of observations within farm-year was performed according to the method of Eliasziw and Donner (9) by use of the program "PAIRSetc" (1).

The concentrations of R. equi in air or soil were not normally distributed, with an excess of 0 counts. The variance was greater than the mean, so the distribution was also overdispersed compared with a Poisson distribution, which has a variance equal to the mean. A visual comparison of the distribution of the counts with both a Poisson distribution and a negative binomial distribution was made using the nbvargr macro of Stata (Stata Corporation, College Station, TX). The negative binomial distribution was a better fit to the data, and this distribution was used in statistical models.

There were 28 farm-year combinations. Within each of these combinations there were between 12 and 42 observations, so the observations were clustered. To take account of this clustering, a random-effects negative binomial model (xtnbreg) was used (6, 12) (Stata cross-sectional time-series reference manual, release 8; Stata Corporation, College Station, TX). The model-building strategy involved two stages. All predictor variables were screened in a univariable random-effects negative binomial regression. Four outcome variables were used individually when screening the predictor variables. These were the concentrations of R. equi in air or soil and the concentrations of virulent R. equi in air or soil. Only predictor variables with a P of <0.25 as determined using a likelihood ratio test were included in the second stage. For the second stage, we used a backward stepwise method to arrive at a final model that only included statistically significant variables. The criterion for exclusion of a variable was a P of ≥0.05 and for inclusion was a P of <0.05 as determined using a likelihood ratio test.

Each environmental and stocking variable (excluding state-year and date) was assigned to one of two categories based on the median, lower or upper quartile, and biological relevance. Soil moisture, pasture height, and soil pH cutoff values were based on the median. As a pH of <6 favors the expression of virulence genes by R. equi, this cutoff has biological relevance (20, 24). The remaining environmental variables were categorized based on either the lower or upper quartile points of the data. The categories or cutoff values for the predictor variables are shown in Table 1.


View this table:
[in this window]
[in a new window]
  		TABLE 1. Categories used to assess the relationship between predictor variables and the concentration of environmental R. equi, the concentration of virulent R. equi, and the proportion of R. equi bacteria that were virulent in the environment

 
Colinearity between predictor variables (environmental and stocking [i.e., group size], excluding state-year and date) was assessed by calculating Spearman's rank correlation coefficient and the odds ratio (OR) between binary predictor variables. A rank correlation coefficient of more than 0.8 was regarded as suggestive of colinearity (8).

Missing values for predictor variables were taken into account in the second stage by only including observations that did not have any missing values for the variables that entered the second stage. If only one variable remained in the final model and it had some missing values, the number of observations used for this final model was equal to the number of observations at the beginning of the backward stepwise process when other variables were eligible for deletion. The majority of predictor variables did not have missing values, and the two variables (wind speed and soil moisture content) with the most missing values had less than 8% of observations missing.

The exponential of a coefficient from the random-effects negative binomial regression model was interpreted as a count ratio (CR) for a unit change in the predictor variable (8). When the predictor variable was categorical, with two levels or groups, this represented the count ratio of being associated with the factor (the level that was coded as 1) compared to the reference group (the level coded as 0). The count ratio was the ratio between the mean counts in the two groups. For example, if the count ratio was 1.5, this would indicate that the concentration of R. equi increased by 50% for a unit change in the predictor variable.

Two analyses were conducted, because pasture height and foal group size were not applicable to the pen or lane areas of a farm. One analysis included pasture height and the group size as well as other variables, and so was only applicable to the paddocks. The other analysis excluded pasture height and numbers of foals on the paddock in the screening process and so was applicable to all locations (paddocks, pens, and lanes) on the farm.

The change in the estimate of the CR of the variables in the final model when a new variable was added to the final model was used to assess potential confounding variables. A change of over 10% in the CR was regarded as confounding. If the number of valid observations was decreased when adding the single variable, the assessment of confounding was done on a final model that included only observations where the single added variable was not missing. This occurred only when wind speed was the single variable that was added. The statistical significance of the added variable was also noted. This allowed assessment of variables that were not included in the final model.

A random-effects logistic regression analysis was performed using the generalized linear latent and mixed model (gllamm) macro of Stata to assess predictor variables for the proportion of R. equi bacteria that were virulent in air and soil separately. The proportion of R. equi bacteria that were virulent in an air or soil sample was defined as the concentration of virulent R. equi divided by the total concentration of R. equi. Only samples containing more than 4 CFU of R. equi per unit of measurement in the denominator of the proportion were used in the regression analyses.

The univariable screening and multivariable model-building processes were performed as described above, and the same predictor variables were used. The unit of analysis was the proportion, with a binomial denominator of 1 used in the gllamm command. The adapt option (adaptive quadrature) of gllamm, instead of ordinary quadrature, was used to fit the models.

McNemar's test, xtnbreg (Stata cross-sectional time-series reference manual, release 8; Stata Corporation, College Station, TX), and random-effects logistic regression (gllamm) (22) were performed using Stata 8.2. All other statistical analyses were performed using Minitab for Windows version 12. Statistical significance was regarded as represented by a P of <0.05.


   RESULTS
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
R. equi pneumonia on farms.
Cases of R. equi pneumonia were reported for 24 of the 28 farm-year combinations. In total, 296 cases of R. equi pneumonia were reported on participating farms during the study period. The prevalence of R. equi pneumonia on these farms ranged from 0% (in four farm-year combinations) to 20.8%. Cases were first reported in September, and the prevalence increased to a maximum during November and December and declined subsequently (Table 2). A total of 15 deaths due to R. equi pneumonia were reported for 11 farm-year combinations. The overall prevalence of R. equi pneumonia on NSW farms was approximately double that seen on Victorian farms.


View this table:
[in this window]
[in a new window]
  		TABLE 2. Distribution of cases of R. equi pneumonia on Thoroughbred farms in the 2000 and 2001 seasons

 
The age at diagnosis of R. equi pneumonia was provided for 269 cases. The majority (175/269) of cases were foals between 31 and 60 days of age (see Table S1 in the supplemental material). The youngest foal diagnosed with R. equi pneumonia was 16 days of age, and the oldest was 158 days of age. Of the 15 deaths only 1 was reported for a foal diagnosed at an age of more than 60 days.

Environmental R. equi.
The concentration of R. equi in air samples ranged from 0 to 124 CFU/1,000 liters and in soil samples from 0 to 136 CFU/mg. The concentration of virulent R. equi in air samples ranged from 0 to 72 CFU/1,000 liters and in soil samples from 0 to 28 CFU/mg. The range and median concentrations of R. equi and virulent R. equi in environmental samples over the 2000 and 2001 seasons are summarized in Table 3.


View this table:
[in this window]
[in a new window]
  		TABLE 3. Concentrations of R. equi and virulent R. equi in air and soil samples collected from 22 Thoroughbred farms in the 2000 and 2001 seasons

 
The correlation between the concentrations of R. equi in the paired soil and air samples was weakly positive (rs = 0.11), as was that between the concentrations of virulent R. equi in the paired soil and air samples (rs = 0.12).

A greater proportion of soil samples than air samples yielded both R. equi and virulent R. equi. The odds of recovery of R. equi from soil samples were 2.8 times greater than the odds of recovery from air samples (P < 0.001), but the odds of recovery of virulent R. equi from soil samples were only 1.5 times greater than the odds of recovery from air samples (P = 0.02) (Table 4).


View this table:
[in this window]
[in a new window]
  		TABLE 4. Proportions of soil and air samples positive and negative for all R. equi bacteria detected compared to the proportions positive and negative for virulent R. equi bacteria

 
The variances of the concentrations of R. equi and virulent R. equi in air samples were both about 25 times greater than the means. The variances of the concentrations of R. equi and virulent R. equi in soil were at least 5 times greater than the means. The concentrations of avirulent R. equi and the concentrations of virulent R. equi in air and soil both followed a negative binomial distribution with a degree of overrepresentation of even-numbered concentrations, especially in the air samples, due to the effect of using 500-liter and 250-liter samples to estimate concentrations per 1,000 liters (see Fig. S1 in the supplemental material).

Although logarithmic transformation did not normalize the data, its use was deemed appropriate (because bacterial growth is exponential) to describe the concentrations of R. equi and virulent R. equi by use of the geometric means of the data after a value of 1 had been added to each value to allow the inclusion of all data in the calculations.

Virulent R. equi on farms.
The associations between the prevalence of R. equi pneumonia and the geometric mean of the environmental burden of R. equi and virulent R. equi and the proportion of virulent R. equi were examined using the Mann-Whitney test (Table 5).


View this table:
[in this window]
[in a new window]
  		TABLE 5. Associations between categories of geometric mean concentrations of R. equi and virulent R. equi in air and soil samples, proportions of R. equi that were virulent, and the median prevalence of R. equi pneumonia on farms in the 2000 and 2001 seasons

 
The median prevalence of R. equi pneumonia on farms with a geometric mean airborne concentration of virulent R. equi of ≥1 CFU/1,000 liters was significantly greater (P = 0.008) than on farms with airborne concentrations of virulent R. equi of <1 CFU/1,000 liters. Similarly, the median prevalence of disease due to R. equi on farms on which ≥35% of airborne R. equi were virulent was significantly greater (P = 0.001) than on farms on which a lower proportion of airborne R. equi were virulent. There were no associations between the concentrations of R. equi, concentrations of virulent R. equi, and proportion of R. equi bacteria that were virulent in the soil and the prevalence of R. equi pneumonia.

Foal population dynamics.
In the 2000 and 2001 seasons, most farms had the maximal number of foals <16 weeks of age in the middle of the season (November and December) (see Table S2 in the supplemental material). The proportions of foals between the ages of 4 and 12 weeks was also highest on farms in the middle of the season and then dropped rapidly by the end of the season (January and February). Farms where R. equi pneumonia was not reported had fewer than 50 foals <16 weeks of age and fewer than 30 foals between 4 and 12 weeks of age throughout the season or had peak numbers of foals in the 4- to 12-week age group occurring earlier in the season (September and October).

Univariable and multivariable analyses.
The outcome variables of the concentration of R. equi and concentration of virulent R. equi in both air and soil had multiple variables associated with them (P < 0.25) in the univariable analyses, so these were used in the multivariable analyses.

Date of sample collection, location within the farm, and temperature were significantly associated (P < 0.05) with the airborne concentrations of R. equi at all locations (Table 6). The greatest deviation of the count ratio (CR) from 1.0 was for the association between airborne concentrations of R. equi and the location within the farm. The airborne concentrations of R. equi in the pens and lanes were almost double (CR = 1.93) those in the paddocks.


View this table:
[in this window]
[in a new window]
  		TABLE 6. Variables identified (using backward stepwise random-effect negative binomial regression analysis) as significantly affecting the airborne concentrations of R. equi and virulent R. equi on farms

 
Temperature and date of sample collection were the only variables significantly associated with the airborne concentrations of R. equi in paddocks (Table 6). The airborne concentrations of R. equi in paddocks in the middle of the season were almost double (CR = 1.97) those seen in the earlier part of the season.

Variables that were significantly associated with the airborne concentrations of virulent R. equi at all locations were the state-year, date, location within the farm, and soil moisture (Table 6). The greatest deviations of the CR from 1.0 were for the associations between the airborne concentrations of virulent R. equi and state-year, date, and location within farm. The airborne concentrations of virulent R. equi were 62% lower on Victorian farms in the 2000 season and 54% lower in the 2001 season than on NSW farms in 2001. The airborne concentrations of virulent R. equi were 75% greater in the middle of the season and 112% greater late in the season than early in the season, and in the pens and lanes the airborne concentrations of virulent R. equi were 91% greater than in the paddocks.

The state-year, pasture height, soil moisture, and mean ambient temperature were significantly associated with the airborne concentrations of virulent R. equi in paddocks (Table 6). The airborne concentrations of virulent R. equi were 63% greater in paddocks with a low pasture height (≤10 cm) than in paddocks with a greater pasture height, 52% greater in paddocks with low soil moisture (≤10% water) than in those with higher soil moisture, and 64% greater in paddocks when the mean ambient temperature at sampling was high (>25°C) than when the average ambient temperature was lower (≤25°C). The paddocks on Victorian farms in the 2000 and 2001 seasons had 74% and 46% lower concentrations of airborne virulent R. equi, respectively, than those on NSW farms in 2001 (Table 6).

Soil moisture, humidity, and date were significantly associated with the concentrations of R. equi in soil at all locations (Table 7). Dry soils (≤10%) had 23% lower concentrations of R. equi than moist soils (>10%), while a low mean air humidity at sampling (≤70%) was associated with concentrations of R. equi in soil 31% higher than in samples collected in more humid conditions. Soil samples collected in the middle and late parts of the season had higher concentrations of R. equi than those collected early in the season.


View this table:
[in this window]
[in a new window]
  		TABLE 7. Variables identified (using backward stepwise random-effect negative binomial regression analysis) as significantly affecting the concentrations of R. equi and virulent R. equi in soil on farms

 
The foal group size in a paddock and air humidity were significantly associated with the concentrations of R. equi in soil on paddocks (Table 7). Soil samples collected when the mean air humidity was low (≤70%) had concentrations of R. equi 27% higher than soil samples collected in more humid conditions, while the concentrations of R. equi in soil samples collected from paddocks with a greater foal group size (containing more than seven foals) were 23% lower than in samples from paddocks with fewer foals (seven or fewer foals).

State-year was the only variable significantly associated with the concentration of virulent R. equi in soil at all locations (Table 7). The concentrations of virulent R. equi were 54% lower in soil samples from Victorian farms in the 2000 season than in those from NSW farms in the 2001 season.

No variables were significantly associated with the concentration of virulent R. equi in soil from paddocks alone (Table 7).

All but one of the CRs for variables in the final models changed by less than 10% when different variables were added, suggesting that there was no confounding. The exception was the final model for airborne concentrations of virulent R. equi, where the CR for the mean ambient temperature at the time of sampling increased from 1.64 to 1.92 when humidity was added to the model. There were no occasions when the added variable was significant (P ≥ 0.05).

No significant (P < 0.05) associations were found in the univariable or multivariable analyses between any variable and the proportions of R. equi bacteria that were virulent in air or in soil samples. Some weak associations (P < 0.25) were seen in the univariable analyses between soil texture and soil moisture and the proportions of airborne R. equi bacteria that were virulent. The largest OR was for the association with soil moisture, with the odds of an R. equi recovered from an air sample being virulent in drier soil samples (≤10% water) being 1.77 times the odds seen with moister soil samples (>10% water) (P = 0.10). A weak association (P < 0.25) was also seen in the univariable analysis between soil pH and the proportion of R. equi bacteria in soil that were virulent. The odds that an R. equi organism recovered from soil was virulent in acidic soil (pH < 6) were 1.57 times the odds seen with more alkaline soil (pH > 6) (P = 0.15).

Correlations between predictor variables.
To evaluate the associations between binary environmental and stocking predictor variables used in the regression analyses, Spearman's correlation coefficients and odds ratios were calculated.

There was no significant colinearity (rs > 0.8) between any of the predictor variables. However, a strong positive relationship was seen between temperature and humidity (rs > 0.5) when assessing the binary relationship between high temperature (>25°C) and low humidity (≤70%) (OR = 22.2). Other variables with positive relationships (0.3 < rs < 0.4) were sandy soils and pen and lane locations (OR = 6.1), low soil moisture (≤10% water) and sandy soil (OR = 4.1), and low soil moisture (≤10% water) and pen and lane locations (OR = 5.1).


   DISCUSSION
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
This study showed that both the airborne concentration of virulent R. equi and the proportion of airborne R. equi bacteria that were virulent were associated with the prevalence of R. equi pneumonia on Thoroughbred horse breeding farms (Table 5). These results confirm the anecdotal and experimental evidence that inhalation of virulent R. equi is the main route of pulmonary infection of the foal. Furthermore, these observations support the hypothesis that it is the size of the challenge with aerosolized virulent R. equi that influences the likelihood that a pulmonary infection will develop into clinical pneumonia (14, 20, 21). No association was observed between the concentration of virulent R. equi or the proportion of R. equi bacteria that were virulent in soil samples and the prevalence of R. equi pneumonia. This is in contrast to previous studies (24, 31) that found an association between soil contamination and disease. These findings suggest that it is the capacity for aerosolization of virulent R. equi that is the most critical aspect of the epidemiology of the disease.

Multiple environmental and management factors influenced the concentration of virulent R. equi on farms in this study. Warm ambient temperature and low pasture height were associated with elevated airborne concentrations, as was dry soil. Dry soil was also associated with elevated proportions of airborne R. equi bacteria that were virulent. This supports the use of farm management practices based upon conclusions of previous investigators that increased soil moisture and good pasture cover on paddocks may reduce the level of exposure to aerosolized virulent R. equi (14, 20, 21). There were very few associations between environmental factors and the concentration of virulent R. equi in soil. Increased proportions of R. equi bacteria that were virulent were common in acidic soil, and while this association was not statistically significant in this study, it may be worthy of further, more focused investigation, as laboratory studies have shown that a variety of R. equi virulence genes are upregulated under acidic conditions (3, 23).

The lanes and pens were associated with higher concentrations of airborne virulent R. equi. Lanes and pens had low soil moisture and poor grass cover. Consequently, greater time spent in these areas would be expected to increase the exposure of foals to higher levels of airborne virulent R. equi, which in turn is likely to increase the relative risk of foals contracting R. equi pneumonia. The high airborne concentrations of virulent R. equi in the lanes and pens make these areas dangerous places to allow foals to congregate for any extended period of time. Such areas could be considered potential infection "hot spots."

Aerosolization of virulent R. equi appeared to be influenced by the soil texture. The most prominent environmental factor influencing airborne concentrations of virulent R. equi and the proportions of airborne R. equi bacteria that were virulent was soil moisture, and this variable was associated with soil texture. Sandy soils tended to be drier than clay soils. The soils in holding pens and lanes were sandy on many farms. The reduced capacity of sandy soils to hold water and their generally lower concentrations of nutrients (15) may favor survival, replication, and/or aerosolization of virulent R. equi. The upregulation of vap gene expression within the endosome of the macrophage is considered part of the mechanism by which the organism survives intracellular killing (4, 23). In vitro studies have found that expression of some vap genes is upregulated when concentrations of micronutrients are restricted, mimicking the endosome environment (23). Even though upregulation of vap genes is most potent in the intracellular environment of the macrophage, it may occur at reduced levels under harsh environmental conditions, possibly contributing to the survival and replication of virulent R. equi bacteria outside the host. R. equi bacteria have been shown to be acid tolerant (5), and acidic conditions and high temperatures have been shown to be positive regulators of the expression of vap genes, with vapA optimally expressed at 38°C in a mildly acidic (pH = 6.5) environment (5, 23, 27). The association between acidic soils and elevated proportions of virulent R. equi in soil was not statistically significant but did suggest that the soil pH may have some influence on the survival of virulent R. equi outside the host. However, the lack of statistically significant associations between environmental variables and the concentration of virulent R. equi in soil suggests that key environmental factors, other than fecal contamination, influencing the proliferation of virulent R. equi in the soil may still need to be identified. Investigating the expression of vap genes and other plasmid-encoded genes in various soil environments may be worth considering in future studies exploring mechanisms that may influence the environmental survival and replication of virulent R. equi.

The stronger association between predictor variables and the concentrations of virulent R. equi in air compared to soil may reflect the habitat of virulent R. equi in the surface soil. As the soil samples that were analyzed were core samples taken down to a depth of 5 cm, they were unlikely to fully reflect the bacterial population in the most superficial soil, which was the soil layer most likely to be aerosolized and thus more likely to be represented in air samples. The air samples were also likely to reflect a relatively greater area than the soil samples. Thus, the greater relative concentrations of virulent than of avirulent R. equi in air, as indicated by the generally higher proportions of R. equi bacteria that were virulent in air samples compared to soil, suggest that virulent R. equi may predominate in the most superficial soil layers.

Previous studies have noted that the likelihood of a farm reporting cases of R. equi pneumonia increases with increased foal numbers and stocking density (7). This association appeared to be affected by the age dynamics of the foal population. Farms that did not report R. equi pneumonia in the 2001 season all had low numbers of foals <16 weeks of age during the sampling period and between 4 and 12 weeks of age throughout the season or had peak numbers of foals between 4 and 12 weeks of age occurring earlier in the season. Reducing the number of foals under 3 to 4 months of age throughout the season, but especially when environmental conditions become warmer and drier, may reduce the risk of R. equi pneumonia. As stocking density was not measured directly in this study, we cannot comment in detail about the effect of stocking density on burdens of environmental R. equi and the risk of disease in foals, but we suggest that farmers may wish to avoid stocking foals less than 3 to 4 months of age at high densities during the warmer and drier months of the season.

The southern hemisphere Thoroughbred breeding season spans the spring and summer months, with the first foals being born in August and more than 95% of mares foaling by the end of December. The tightly regulated Thoroughbred breeding season results in a close relationship between the age of foals and calendar month of the year on most farms, so variations in environmental factors between and within years may affect the prevalence of disease. The majority of cases occurred in the middle of the season, in association with significant increases in the airborne concentrations of virulent R. equi and the peak in the number of foals between 4 and 12 weeks of age. Airborne virulent R. equi burdens continued to be high later in the season, but this was not associated with a continued high prevalence of R. equi pneumonia, probably because the number of foals aged between 4 and 12 weeks had begun to fall on most farms.

This work has highlighted the importance of the airborne R. equi population in determining the prevalence of R. equi pneumonia and identified important ecological factors that facilitate aerosolization of virulent R. equi (i.e., soil moisture, sandy soils, poor pasture cover). These factors affected the concentration of airborne virulent R. equi and the proportion of airborne R. equi bacteria that were virulent in the environment of susceptible foals and appeared to influence the prevalence of R. equi pneumonia and contribute to the severity of disease. Environmental management strategies focusing on reduction of the level of aerosol exposure to susceptible foals through the use of selective irrigation in dry areas such as holding pens and lanes, avoidance of areas with soils with a low water-holding capacity (i.e., sandy areas), and maintenance of good pasture cover where foals between the ages of 4 and 12 weeks are kept are likely to substantially reduce the impact of R. equi pneumonia on the horse-breeding industry.


   ACKNOWLEDGMENTS
 
This work was supported by the Rural Industries Research and Development Corporation of Australia.

We thank the Statistical Consulting Centre of the University of Melbourne and specifically Graham Hepworth for assistance with statistical analysis.


   FOOTNOTES
 
* Corresponding author. Mailing address: Faculty of Veterinary Science, The University of Melbourne, Parkville, Victoria 3010, Australia. Phone for G. Muscatello: 61 3 8344 7368. Fax: 61 3 8344 7374. E-mail: mug{at}unimelb.edu.au. Phone for G. F. Browning: 61 3 8344 7342. Fax: 61 3 8344 7374. E-mail: glenfb{at}unimelb.edu.au. Back

{dagger} Supplemental material for this article may be found at http://aem.asm.org/. Back


   REFERENCES
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Abramson, J. H. 2004. WINPEPI (PEPI-for-Windows): computer programs for epidemiologists. Epidemiol. Perspect. Innov. 1:6.[CrossRef][Medline]
  2. Barton, M. D., and K. L. Hughes. 1984. Ecology of Rhodococcus equi. Vet. Microbiol. 9:65-76.[CrossRef][Medline]
  3. Benoit, S., A. Benachour, S. Taouji, Y. Auffray, and A. Hartke. 2001. Induction of vap genes encoded by the virulence plasmid of Rhodococcus equi during acid tolerance response. Res. Microbiol. 152:439-449.[Medline]
  4. Benoit, S., A. Benachour, S. Taouji, Y. Auffray, and A. Hartke. 2002. H2O2, which causes macrophage-related stress, triggers induction of expression of virulence-associated plasmid determinants in Rhodococcus equi. Infect. Immun. 70:3768-3776.[Abstract/Free Full Text]
  5. Benoit, S., S. Taouji, A. Benachour, and A. Hartke. 2000. Resistance of Rhodococcus equi to acid pH. Int. J. Food Microbiol. 55:295-298.[CrossRef][Medline]
  6. Cameron, A., and P. Trivedi. 1998. Regression analysis of count data. Cambridge University Press, Cambridge, United Kingdom.
  7. Chaffin, K. M., N. D. Cohen, and R. J. Martens. 2003. Evaluation of equine breeding farm characteristics as risk factors for development of Rhodococcus equi pneumonia in foals. J. Am. Vet. Med. Assoc. 222:467-474.[CrossRef][Medline]
  8. Dohoo, I., W. Martin, and H. Stryhn. 2003. Veterinary epidemiological research. AVC Inc., Prince Edward Island, Canada.
  9. Eliasziw, M., and A. Donner. 1991. Application of the McNemar test to non-independent matched pair data. Stat. Med. 10:1981-1991.[Medline]
  10. Falcon, J., B. P. Smith, T. R. O'Brien, G. P. Carlson, and E. Biberstein. 1985. Clinical and radiographic findings in Corynebacterium equi pneumonia in foals. J. Am. Vet. Med. Assoc. 186:593-599.[Medline]
  11. Haites, R. E., G. Muscatello, A. P. Begg, and G. F. Browning. 1997. Prevalence of the virulence-associated gene of Rhodococcus equi in isolates from infected foals. J. Clin. Microbiol. 35:1642-1644.[Abstract]
  12. Hausman, J. A., B. H. Hall, and Z. Griliches. 1984. Econometric models for count data with an application to the patients-R&D relationship. Econometrica 52:909-938.[CrossRef]
  13. Giguere, S., M. K. Hondalus, J. A. Yager, P. Darrah, D. M. Mosser, and J. F. Prescott. 1999. Role of the 85 kilobase plasmid and plasmid-encoded virulence-associated protein A in intracellular survival and virulence of Rhodococcus equi. Infect. Immun. 67:3548-3557.[Abstract/Free Full Text]
  14. Giguere, S., and J. F. Prescott. 1997. Clinical manifestation, diagnosis, treatment and prevention of Rhodococcus equi infections in foals. Vet. Microbiol. 56:313-334.[CrossRef][Medline]
  15. Laegreid, M., O. C. Bockman, and O. Kaarstad. 1999. Some descriptions of fundamental soil properties and functions, p. 92-93. In M. Lægreid, O. C. Bøckman, and O. Kaarstad (ed.), Agriculture, fertilizers and the environment. Norsk Hydro ASA, Porsgrunn, Norway.
  16. Magnusson, H. 1923. Spezifische infektiose Pneumonie beim Fohlen. Ein neuer Eiterregar beim Pferd. Arch. Wiss. Prakt. Tierhelkd. 50:22-38.
  17. Martens, R. J., S. Takai, N. D. Cohen, K. M. Chaffin, H. Liu, K. Sakurai, H. Sugimoto, and S. W. Lingsweiler. 2000. Association of disease with isolation and virulence of Rhodococcus equi from farm soil and foals with pneumonia. J. Am. Vet. Med. Assoc. 217:220-225.[CrossRef][Medline]
  18. McLaren, R. G., and K. C. Cameron. 1996. The storage of water in soil, p. 2-4. In Soil science: sustainable production and environmental protection, 2nd ed. Oxford University Press, Oxford, United Kingdom.
  19. Muscatello, G., and G. F. Browning. 2004. Identification and differentiation of avirulent and virulent Rhodococcus equi using selective media and colony blotting DNA hybridization to determine their concentrations in the environment. Vet. Microbiol. 100:121-127.[CrossRef][Medline]
  20. Prescott, J. F., and J. A. Yager. 1991. The control of Rhodococcus equi pneumonia in foals, p. 21-25. In W. Plowright, P. D. Rossdale, and J. F. Wade (ed.), Proceedings of the 6th International Conference on Equine Infectious Diseases. R&W Publications, Newmarket, United Kingdom.
  21. Prescott, J. F., M. Travers, and J. A. Yager-Johnson. 1984. Epidemiological survey of Corynebacterium equi infections on five Ontario horse farms. Can. J. Comp. Med. 48:10-13.[Medline]
  22. Rabe-Hesketh, S., A. Skrondal, and A. Pickles. 2004. GLLAMM manual. University of California, Berkeley. [Online.] http://www.gllamm.org/docum.html.
  23. Ren, J., and J. F. Prescott. 2003. Analysis of virulence plasmid gene expression of intra-macrophage and in vitro grown Rhodococcus equi ATCC 33701. Vet. Microbiol. 94:167-182.[CrossRef][Medline]
  24. Takai, S. 1997. Epidemiology of Rhodococcus equi infections: review. Vet. Microbiol. 56:167-176.[CrossRef][Medline]
  25. Takai, S., M. Iie, C. Kobayahi, T. Morishita, T. Nishio, T. Fujimura, Y. Sasaki, and S. Tsubaki. 1993. Monoclonal antibody specific to virulence-associated 15- to 17-kilodalton antigens of Rhodococcus equi. J. Clin. Microbiol. 31:2780-2782.[Abstract/Free Full Text]
  26. Takai, S., M. Shoda, Y. Sasaki, S. Tsubaki, G. Fortier, S. Pronost, K. Rahal, T. Becu, A. P. Begg, G. F. Browning, V. N. Nicholson, and J. F. Prescott. 1999. Restriction fragment length polymorphisms of virulent plasmids in Rhodococcus equi. J. Clin. Microbiol. 37:3417-3420.[Abstract/Free Full Text]
  27. Takai, S., N. Fukunaga, K. Kamisawa, Y. Imai, Y. Sasaki, and S. Tsubaki. 1996. Expression of virulence-associated antigens of Rhodococcus equi is regulated by temperature and pH. Microbiol. Immunol. 40:591-594.[Medline]
  28. Takai, S., N. Fukunaga, S. Ochiai, Y. Imai, Y. Sasaki, S. Tsubaki, and T. Sekizaki. 1996. Identification of intermediately virulent Rhodococcus equi isolates from pigs. J. Clin. Microbiol. 34:1034-1037.[Abstract]
  29. Takai, S., S. A. Hines, T. Sekizaki, V. N. Nicholson, D. C. Alperin, M. Osaki, D. Takamatus, M. Nakamura, K. Suzuki, N. Ogino, T. Kakuda, H. Dan, and J. F. Prescott. 2000. DNA sequence and comparison of virulence plasmids from Rhodococcus equi ATCC 33701 and 103. Infect. Immun. 68:6840-6847.[Abstract/Free Full Text]
  30. Takai, S., S. Kawazu, and S. Tsubaki. 1985. Enzyme-linked immunosorbent assay for diagnosis of Corynebacterium (Rhodococcus) equi infection in foals. Am. J. Vet. Res. 46:2166-2170.[Medline]
  31. Takai, S., S. Ohbushi, K. Koike, S. Tsubaki, H. Oishi, and M. Kamada. 1991. Prevalence of virulent Rhodococcus equi in isolates from soil and feces of horse-breeding farms with and without endemic infections. J. Clin. Microbiol. 29:2887-2889.[Abstract/Free Full Text]
  32. Takai, S., T. Fujimori, K. Katsuzuaki, and S. Tsubaki. 1987. Ecology of Rhodococcus equi in horses and their environment on horse-breeding farms. Vet. Microbiol. 14:233-239.[CrossRef][Medline]
  33. Takai, S., T. Ikeda, Y. Sasaki, Y. Watanabe, T. Ozawa, S. Tsubaki, and T. Sekizaki. 1995. Identification of virulent Rhodococcus equi by amplification of gene coding for 15- to 17-kilodalton antigens. J. Clin. Microbiol. 33:1624-1627.[Abstract]
  34. Takai, S., Y. Sasak, and S. Tsubaki. 1995. Rhodococcus equi infection in foals—current concepts and implications for future research. J. Equine Sci. 6:105-119.
  35. Tan, K. H. 1998. Soil reaction, p. 384-386. In Principles of soil chemistry, 3rd ed. Marcel Dekker Inc., New York, N.Y.
  36. Woolcock, J. B., A. M. Farmer, and M. D. Mutimer. 1979. Selective medium for Corynebacterium equi isolation. J. Clin. Microbiol. 9:640-642.[Abstract/Free Full Text]

Applied and Environmental Microbiology, September 2006, p. 6152-6160, Vol. 72, No. 9
0099-2240/06/$08.00+0     doi:10.1128/AEM.00495-06
Copyright © 2006, American Society for Microbiology. All Rights Reserved.



This article has been cited by other articles:


This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental material
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Muscatello, G.
Right arrow Articles by Browning, G. F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Muscatello, G.
Right arrow Articles by Browning, G. F.
Agricola
Right arrow Articles by Muscatello, G.
Right arrow Articles by Browning, G. F.

Home Help [Feedback] [For Subscribers] [Archive] [Search] [Contents]
J. Bacteriol. Microbiol. Mol. Biol. Rev. Eukaryot. Cell All ASM Journals