Estimation of Mycobacterium avium subsp. paratuberculosis Growth Parameters: Strain Characterization and Comparison of Methods

Abstract

The growth rate of Mycobacterium avium subsp. paratuberculosis was assessed by different methods in 7H9 medium supplemented with OADC (oleic acid, albumin, dextrose, catalase), Tween 80, and mycobactin J. Generation times and maximum specific growth rates were determined by wet weight, turbidometric measurement, viable count, and quantitative PCR (ParaTB-Kuanti; F57 gene) for 8 M. avium subsp. paratuberculosis strains (K10, 2E, 316F, 81, 445, 764, 22G, and OVICAP 49). Strain-to-strain differences were observed in growth curves and calculated parameters. The quantification methods gave different results for each strain at specific time points. Generation times ranged from an average of 1.4 days for viable count and qPCR to approximately 10 days for wet weight and turbidometry. The wet-weight, turbidometry, and ParaTB-Kuanti qPCR methods correlated best with each other. Generally, viability has been assessed by viable count as a reference method; however, due to M. avium subsp. paratuberculosis clumping problems and the presence of noncultivable M. avium subsp. paratuberculosis cells, we conclude that qPCR of a single-copy gene may be used reliably for rapid estimation of M. avium subsp. paratuberculosis bacterial numbers in a sample.

INTRODUCTION

Mycobacterium avium subsp. paratuberculosis is the causal agent of a chronic granulomatous enteritis in ruminants called paratuberculosis or Johne's disease (JD). JD is prevalent in domestic animals worldwide, producing an important impact on the economy (14, 18, 27). It is transmitted mainly by the oral-fecal route but can also be vertically transmitted by intrauterine infection (29). M. avium subsp. paratuberculosis may be a human health concern, as well, since some studies have associated M. avium subsp. paratuberculosis with the etiology of Crohn's disease (6). Although M. avium subsp. paratuberculosis is defined as an obligate pathogen (31), the organism can survive for long periods outside the host, making pastures infective even after infected animals have been removed (34).

M. avium subsp. paratuberculosis is a slow-growing (>24-h generation time), acid-fast microorganism that has a strong tendency to form clumps (5). These characteristics make working with M. avium subsp. paratuberculosis challenging. Actually, quantification of M. avium subsp. paratuberculosis in a sample is one of the major problems encountered when working with the microorganism.

Viable-cell plate counting (VC) has been the reference method for M. avium subsp. paratuberculosis quantification because it is the standard method in bacteriology and because it assesses viability (17, 25). Thus, when new quantification methods are developed or more than one method is used in an experiment, a retrospective estimate of the number of CFU is always calculated by performing colony counts on plated serial dilutions. In M. avium subsp. paratuberculosis experimental infections, both the administered inoculum and bacteria recovered from feces, blood, or tissues should ideally be recorded as CFU. However, VC in the case of M. avium subsp. paratuberculosis is extremely time-consuming and does not give immediate results, since the data are obtained only after months because of M. avium subsp. paratuberculosis' typical slow growth. Actually, depending on the strain and culture conditions, up to 4 months may be required before visible colonies are detected on solid medium (3). Moreover, VC can also be subjective and dependent on the accurate enumeration of colonies that grow only under the provided conditions, introducing errors. Clumping of M. avium subsp. paratuberculosis increases the variability and reduces the accuracy of VC data, and many studies have demonstrated that the method is not reliable when working with M. avium subsp. paratuberculosis (20, 21, 22, 30). We must also bear in mind that M. avium subsp. paratuberculosis has been reported to exist in a dormant or viable-noncultivable state in the environment (34), and its close relative Mycobacterium avium subsp. avium enters a dormancy state in response to starvation (1). Bacteria in such a dormancy state are not cultivable and therefore not detectable by the VC method. For these reasons, it is desirable to find other methods that give faster and more accurate quantification results for M. avium subsp. paratuberculosis.

Optical density, or turbidometry (T), has also been used to quantify M. avium subsp. paratuberculosis (25, 28) or to study its growth (20). The advantages of this method are that it is rapid, nondestructive, and inexpensive compared to others. However, the lower detection limit for the T method is normally high (10⁷ bacteria/ml) (2), and it is only applicable to liquid cultures. M. avium subsp. paratuberculosis' tendency to form clumps can alter the results, and furthermore, the method is not specific, since other putative contaminating bacterial species in the sample will also be included in the estimated counts.

Quantification by wet weight of M. avium subsp. paratuberculosis harvested from broth cultures or from solid-medium plates has been used in many studies, and it has been recently recommended for experimental infection and challenge studies because of its ease and immediate results (8). The wet-weight standardized inoculum should also be quantified by VC. Clumping of bacteria does not interfere with the wet-weight method. However, considering that M. avium subsp. paratuberculosis varies between 1 and 2 μm in length (32) and that morphological differences have been observed, depending on the strain and growth phase, over- or underestimations also occur with this method, leading to errors.

A newer approach to quantification of bacteria is the use of molecular techniques. With this method, “genomic equivalents” (GE) are determined by quantitative real-time PCR (qPCR) amplification. Most M. avium subsp. paratuberculosis quantification in this area has been done using the IS900 sequence (13) and the F57 element (26). IS900 qPCR quantification of M. avium subsp. paratuberculosis has limitations because IS900 is a transposable element and the copy number varies depending on the strain. F57 is unique to M. avium subsp. paratuberculosis and has not been described in any other bacterial species (19), and each M. avium subsp. paratuberculosis cell harbors just one F57 copy, making quantification simple and reliable. We decided to use a ready-to-use qPCR kit based on element F57 amplification (ParaTB-Kuanti; Vacunek). The manufacturer states that, under optimum conditions, the detection limit is 10 copies per reaction and the correlation coefficient between the number of M. avium subsp. paratuberculosis copies and the number of PCR cycles is 0.998. ParaTB-Kuanti has been compared to another commercial qPCR kit, showing 100% sensitivity and 100% specificity by testing 43 positive and 160 negative cattle feces samples. After DNA extraction has been done, a sample is amplified, and results are obtained after 90 min. This method is therefore fast, meaning that results are obtained within 1 day.

In the present study, we wanted to determine the growth kinetics of 8 different strains of M. avium subsp. paratuberculosis. Several quantification methods were used in parallel to assess growth at different times, and the methods were compared. The goal of the study was to characterize selected M. avium subsp. paratuberculosis strains by their growth behavior and to find a method for fast and reliable quantification of M. avium subsp. paratuberculosis cells. This would, in turn, make M. avium subsp. paratuberculosis research easier, since it is desirable to have a precise and sensitive method for the preparation of experimental inocula for infection and challenge studies. The results of this study could also provide a reference, helping to quantify M. avium subsp. paratuberculosis counts in feces, milk, or tissues from infected animals or environmental samples.

MATERIALS AND METHODS

Isolate culture and maintenance.

The strains included in this study (Table 1) were either field isolates obtained from Neiker-Tecnalia (Basque Institute for Agricultural Research and Development, Derio, Spain) maintained as glycerol stocks at −80°C, vaccine strains, or strains from the ATCC (American Type Culture Collection, Manassas, VA). Some of the selected bovine and ovine strains represented the most common IS1311 PCR-restriction enzyme analysis (REA) and pulsed-field gel electrophoresis (PFGE) profiles found previously in Spanish M. avium subsp. paratuberculosis isolates (23). All strains were cultured in Middlebrook 7H9 broth supplemented with 10% OADC (oleic acid, albumin, dextrose, catalase) enrichment (Becton Dickinson and Company, MD), 0.05% Tween 80 (Panreac Quimica SA, Barcelona, Spain), and 2 mg/liter of mycobactin J (Allied Monitor, Inc., Fayette, MO), referred to below as 7H9-OADC-MJ-T.

Table 1.

M. avium subsp. paratuberculosis strains used in the study

Strain code	Origin	Profile
		IS1311 PCR-REAa	SnaBI-SpeI PFGE
K10 (ATCC BAA-968)	Bovine; reference		1-1
2E	Bovine; vaccine		ND
316F	Bovine; vaccine		ND
81	Bovine	C	2-58
445	Bovine	C	54-49
764	Bovine	C	2-1
22G	Sheep	S	69-50
OC 49	Sheep	S	57-57

^aC, cattle strain; S, sheep strain; ND, not done.

Growth kinetics experiments.

Inocula of each strain were prepared in 7H9-OADC-MJ-T. The wet-weight method described below (8) was used at this step to estimate the bacterial numbers for each strain. Briefly, after approximately 4 weeks of growth in T25 tissue culture flasks at 37 ± 1°C, bacterial suspensions were pelleted by centrifugation at 3,018 × g for 10 min in a Beckman Coulter Allegra X-12R centrifuge. The pellet was mechanically disrupted with a sterile rod to try to reduce bacterial clumps and to yield a homogeneous suspension by adding the mediun slowly while vortexing it. A suspension of 1.7 × 10⁵ CFU/ml in 7H9-OADC-MJ-T (final volume, 250 ml) for each strain was prepared based on the bacterial-content estimation done by wet weight. Ten milliliters of the previous adjusted suspension was transferred to each of 22 culture flasks (25 cm²; Corning). The flasks were labeled from T₀ to T₁₀ twice in order to have duplicates for each time point, that is, to be able to plot two separate curves, curve 1 and curve 2. The different time points when bacterial counts in the cultures were estimated were as follows: T₀, 0 days; T₁, 7 days; T₂, 14 days; T₃, 21 days; T₄, 28 days; T₅, 35 days; T₆, 42 days; T₇, 49 days; T₈, 63 days; T₉, 98 days; and T₁₀, 140 days.

Culture flasks were incubated and maintained unshaken in an incubator (Sanyo) at 37 ± 1°C during the whole experiment. Each flask was taken out to be processed at the time point it was labeled for.

Wet weight.

The procedure described by Hines et al. (8) was followed. Briefly, 15-ml conical tubes were weighed, and the values were recorded. The flasks were removed from the incubator and shaken manually, and 8 ml of culture broth was removed aseptically and added to the previously weighed conical tubes. The tubes were centrifuged at 3,018 × g for 10 min. The supernatant was removed, and the tubes were allowed to stand inverted for 5 min in racks over sterile Whatman paper. The tubes were then weighed, and the values were recorded. The wet weight was calculated by subtracting the empty-tube weight from the combined tube and pellet values.

Turbidity.

At each time point, the pellet obtained by the wet-weight method was disrupted mechanically with a sterile rod and by vortexing. Eight milliliters of 10 mM phosphate-buffered saline (PBS) (pH 7.2) was added slowly while vortexing in order to obtain a homogeneous cell suspension. Two to 3 ml was transferred to polystyrene turbidometric tubes (bioMérieux), and McFarland unit values were recorded (Densichek; bioMérieux). The remaining 5 to 6 ml was used for the rest of the methods.

Viable count.

Tenfold serial dilutions of the homogeneous suspension of bacteria were carried out in 10 mM PBS (pH 7.2) with vortexing after each dilution step. The estimate of cell numbers obtained by the wet-weight method was taken into account in order to decide what dilutions to plate. Three dilutions were chosen: one that would give 10 to 100 colonies and two that were 1 log₁₀ unit higher and lower, respectively. One hundred microliters from each of the selected dilutions was spread plated in each of two flasks with 7H9 agar, 10% OADC, 0.05% Tween 80, and micobactin J (2 mg/liter). Colony counting was performed after 4, 8, 12, and 16 weeks of incubation at 37°C.

Quantitative real-time PCR.

One milliliter of the homogeneous resuspended cell suspension was transferred to each of two vials. The vials were centrifuged at 16,000 × g for 5 min, and the supernatants were discarded. Samples were stored at −80°C. Once all the samples for each of the time points of the growth curve were collected, DNA extraction was done. The pellets were washed once with PBS to eliminate the presence of free DNA. Additional washing steps were not done in order to avoid bacterial loss. After the washing step, 100 μl of Tris-EDTA (TE) was added to the vials, and the pellet was disrupted. A homogeneous solution was achieved and passed to a 2-ml vial containing 300 μg of zirconium/silica beads. DNA was obtained by bead beating on a Ribolyser (Hybaid) for 3 cycles of 45 s each at 4 m/s. The vials were centrifuged at 16,000 × g for 5 min, and maximum volume was recovered. The amount of DNA was assessed by NanoDrop (ThermoFisher) at 260 nm.

qPCR was done by employing ParaTB-Kuanti-VK (Vacunek S. L.), and the manufacturer's instructions were followed. Briefly, 10 μl of DNA was added to 15 μl of master mix in 96-well plates. Triplicates of dilutions of positive-control standards were run in parallel to obtain data points for standard curves. Amplification and real-time measurement were performed in a 7500 Real Time PCR System (Applied Biosystems) with the following conditions: 95°C for 10 min, 95°C for 15 s, and 60°C for 60 s for 45 cycles. The results were analyzed with ABI Prism software version 7500 and SDS software v. 1.4.

Calculation of parameters.

The maximum specific growth rate (μ_max), defined as the increase in cell mass per time unit, was calculated as follows: μ_max = (ln N_f − ln N₀)/t_f − t₀, where N₀ is the initial number of bacteria at the initial time point considered (t₀) and N_f is the final number of bacteria at the final time point considered (t_f). Generation time (G), defined as the time it takes for cells to double, or the doubling time, was calculated using the following formula: G = t_f − t₀/n, where n is equal to (log N_f − log N₀)/log₂. G and μ_max are related by the following formula: G = ln 2/μ_max. Both μ and G were calculated with data from the exponential phase of each strain.

Strains were compared on the basis of their maximum production (mP) and the time at which this mP was achieved.

Statistical analyses.

Bacterial quantification methods were compared to each other. To place all quantification methods on the same scale, all categories (mg, McFarland units, CFU, and GE) were divided by the maximum value in a category. Comparisons between two methods were performed using Pearson correlations.

In order to calculate the equivalences between methods derived from our study, they were calculated as the mean of all strains at all times for each method or the mean of all strains for each method at each particular time point.

The coefficient of variation (CV) of the bacterial concentrations for each strain at each time point and each growth curve was calculated as the standard deviation divided by the mean. This was done in order to investigate the importance of strain variation and the quantification method for the VC and ParaTB-Kuanti methods. This analysis provided 396 observations for each method. To simplify, the mean of the coefficients of variation of all strains was calculated at each time point.

RESULTS

Strain characterization; growth parameters.

The growth curves plotted for all strains for each method are shown in Fig. 1. Sharp growth phase demarcation was not observed, and log phase was reached after 2 weeks of incubation in all methods. After 63 days, most of the strains were in stationary phase, as can be observed for wet weight (Fig. 1A), T (Fig. 1B), and qPCR (Fig. 1D). For the VC method (Fig. 1C), stationary phase started earlier, around 40 days. A slight bimodal distribution or biphasic loss of viability was observed with some strains and methods.

Fig. 1.

Growth curves obtained by quantification methods, wet weight (WW) (A), turbidometry (B), viable count (C), and qPCR (D), for vaccine and reference strains (■, 316F; ●, 2E; ▵, K10), bovine field isolates (○, 81; ♦, 445; □, 764), and ovine field isolates (▴, 22G; ♢, OC 49). McFar units, McFarland units; d, days. The error bars indicate standard deviations.

Maximum specific growth rates (μ_max) and generation times (G) for each strain and quantification method are shown in Table 2. VC and qPCR gave an average G value of 1.4 days for all strains and showed very little variability. Wet-weight and T measurements gave higher G values (9.6 and 11.5 days, respectively), with higher variability among strains.

Table 2.

Estimated maximum specific growth rate and generation times for wet weight, turbidometry, viable count, and qPCR for each strain

Strain	Wet wt		Turbidometry		Viable count		qPCR
	μmax (day−1)	G (days)	μmax (day−1)	G (days)	μmax (day−1)	G (days)	μmax (day−1)	G (days)
316F	0.077	8.99	0.067	10.27	0.403	1.72	0.451	1.54
2E	0.057	12.19	0.055	12.49	0.501	1.38	0.475	1.46
K10	0.080	8.70	0.050	14.00	0.451	1.54	0.569	1.22
81	0.064	10.91	0.045	15.28	0.513	1.35	0.517	1.34
445	0.079	8.75	0.056	12.34	0.499	1.39	0.525	1.32
764	0.065	10.67	0.080	8.65	0.497	1.40	0.527	1.32
22G	0.082	8.41	0.055	12.65	0.469	1.48	0.466	1.49
OC 49	0.083	8.34	0.113	6.12	0.350	1.98	0.450	1.54
All
Mean	0.073	9.619	0.065	11.475	0.460	1.53	0.497	1.40
SD	0.010	1.437	0.022	2.979	0.057	0.218	0.043	0.119
CV	13.67	14.94	34.07	25.96	12.42	14.28	8.66	8.50

Correlation of the time at which μ_max is reached for each strain and method is shown in Fig. 2. Ovine strains reached μ_max by day 14 for all methods. However, bovine strains (the reference strain, field isolates, and vaccine strains) showed higher variability, depending on the quantification method; when the qPCR method was used, μ_max was not reached until day 35 for almost all strains.

Fig. 2.

Time at which the maximum specific growth rate was achieved for each strain and quantification method. Black bars, wet weight; gray bars, turbidometry; hatched bars, viable count; open bars, qPCR.

Production was studied as the final or highest production for each strain and method. The highest-producing strains for most methods were 445, 81, 764, and K10, as seen in Table 3. By day 63, mP was reached in most cases. Strain production parameters were not identical for all methods. However, based on these production parameters, the strains could be divided into three groups: group I, composed of high-producing and fast-growing strains (445, 81, 764, and K10); group II, composed of intermediate-growing strains (316F and 2E); and group III, composed of slow-growing strains (22G and OC 49). Group I strains are basically bovine field isolates.

Table 3.

Production parameters

Strain	Wet wt		Turbidometry		Viable count		qPCR
	mP (mg)	TmPa (days)	mP (abs)b	TmP (days)	mP (CFU)	TmP (days)	mP (GE/day)	TmP (days)
316F	150.95	140	4.98	63	2.78 × 108	63	4.25 × 109	98
2E	114.40	63	4.88	63	1.80 × 108	63	2.83 × 109	63
K10	130.10	98	7.07	63	4.82 × 108	49	6.39 × 109	98
81	149.90	140	7.31	63	4.60 × 108	63	1.01 × 1010	98
445	151.25	63	7.55	63	7.28 × 108	63	1.42 × 1010	98
764	141.50	98	7.53	63	1.72 × 109	63	1.121 × 1010	63
22G	108.85	140	3.63	140	3.78 × 108	98	8.95 × 109	140
OC 49	103.05	140	4.00	98	8.76 × 108	98	7.51 × 109	140

^aT_mP, time at which mP was reached.

^babs, absorbance units.

Depending on the method, 50% production was reached at different time points (data not shown). The wet-weight data showed that 50% production was reached after 21 to 35 days, whereas T data showed 50% production after 35 days and qPCR after 42 days. The data obtained by the VC method showed that 50% production was achieved in the time interval of 28 to 42 days.

Quantification method comparison. (i) Equivalence between methods.

As stated in Materials and Methods, the inoculum size was adjusted to 1.7 × 10⁵ bacteria/ml estimated by wet weight. We wanted to compare this estimate with the time zero estimate by all methods in order to assess real overestimation or underestimation. At time zero, all flasks had received material from the same bacterial suspension, and the bacteria had had no time for replication; therefore, aside from pipetting errors, the rest of the variability could be attributed to the quantification method. The time zero quantity for wet weight, VC, and qPCR and their comparisons with the initial inoculum estimates are detailed in Table 4. At this time point, T does not give a reading because the bacterial concentration is below the minimal detection limit of the technique. The VC method and the qPCR ParaTB-Kuanti method estimated fewer bacteria than in the initial inoculum estimate, 54.71% ± 34.96% and 75.75% ± 14.65% fewer bacteria, respectively. However, the wet-weight method yielded values higher than the initial inoculum estimate. The mean in this case was 12,867.64% ± 4,029.52% higher than the initial inoculum, generally 2 log units more.

Table 4.

Time zero quantification by wet weight, viable count, and qPCR^a

Strain	Wet wt		Viable count			qPCR
	(Wet wt0) (bacteria/ml)	(Wet wt0/wet wti)	(VC0) (CFU/ml)	(VC0/wet wti)	(VC0/wet wt0) (GE/ml)	(qPCR0)	(qPCR0/wet wti)	(qPCR0/wet wt0)
316F	1.92 × 107	112.87	1.99 × 105	1.17	0.01036	7.97 × 104	0.47	0.00415
2E	3.33 × 107	195.96	1.90 × 104	0.11	0.00057	1.13 × 104	0.07	0.00034
K10	2.49 × 107	149.69	8.93 × 104	0.53	0.00358	1.91 × 104	0.11	0.00077
81	2.69 × 107	158.09	8.60 × 104	0.51	0.00320	3.92 × 104	0.23	0.00146
445	1.74 × 107	102.57	5.43 × 104	0.32	0.00311	3.19 × 104	0.19	0.00183
764	2.43 × 107	142.65	5.33 × 104	0.31	0.00220	4.21 × 104	0.25	0.00174
22G	1.78 × 107	104.78	3.78 × 104	0.22	0.00212	7.58 × 104	0.45	0.00425
OC 49	1.12 × 107	65.81	ND	ND	ND	2.88 × 104	0.17	0.00258

^aWet wt₀, wet weight value at time zero; wet wt_i, wet weight initial inoculum value; VC₀, viable count value at time zero; qPCR₀, qPCR value at time zero; ND, not determined. Turbidometry could not be compared because all the values were 0 or below. Wet wt_i = 1.7 × 10⁵ bacteria/ml.

A comparison of equivalences between methods is detailed in Tables 5 and 6. Table 5 shows equivalences estimated from this work compared to those of others. Our results established the following relations: 1 mg (wet weight) of pellet is approximately 3.75 × 10⁶ CFU obtained by the VC method and 3.39 × 10⁷ GE obtained by ParaTB-Kuanti qPCR, whereas 1 McFarland unit is about 1.3 × 10⁷ bacteria/ml by VC and 1.2 × 10⁸ GE/ml by ParaTB-Kuanti qPCR.

Table 5.

Equivalences between methods^a

Unit (method)	Equivalent to:		Reference
	VC (no. of bacteria)	qPCR (GE)
1 mg (wet wt)	3.75 × 106 ± 2.45 × 106	3.39 × 107 ± 2.35 × 107	Present study
	5 × 106		12
	3 × 107		9
	107		8
1 McFarland unit (T)b	1.3 × 107 ± 9.82 × 106	1.2 × 108 ± 4.5 × 107	Present study
	108		11

^aEquivalences from the present study calculated as the mean for all strains from all time points and published equivalences from other studies.

^bEquivalents are per milliliter.

Table 6.

Equivalences between methods^a

Unit (method)	Equivalent to:
	T (days)	qPCR (GE)	SD	CV (%)	VC (CFU)	SD	CV (%)
1 CFU (VC)	0	0.79	0.58	73.76	3.59 × 104	3.15 × 104	87.74
	7	0.23	0.16	72.27	2.62 × 106	2.95 × 106	112.67
	14	0.80	0.56	70.27	9.23 × 106	4.69 × 106	50.75
	21	5.70	6.28	110.32	5.53 × 106	6.99 × 106	126.36
	28	8.83	5.20	55.88	4.90 × 106	3.64 × 106	74.18
	35	17.23	9.53	55.34	2.60 × 106	1.54 × 106	59.24
	42	22.01	14.46	65.69	4.32 × 106	3.30 × 106	76.39
	49	19.64	12.70	64.69	3.15 × 106	2.53 × 106	80.38
	63	28.76	28.70	99.79	4.08 × 106	4.37 × 106	107.09
	98	24.51	19.31	78.78	3.88 × 106	2.88 × 106	74.38
	140	187.23	280.62	149.88	8.70 × 105	8.71 × 105	100.15
1 mg (wet wt)		2.03 × 104	1.39 × 104	68.59
		4.95 × 105	6.48 × 105	131.63
		6.78 × 106	3.84 × 106	56.59
		1.95 × 107	1.32 × 107	67.58
		3.36 × 107	1.31 × 107	38.90
		3.82 × 107	1.64 × 107	43.05
		5.95 × 107	1.82 × 107	30.56
		4.14 × 107	1.85 × 107	44.71
		5.87 × 107	2.15 × 107	36.61
		6.02 × 107	2.39 × 107	39.75
		5.39 × 107	2.67 × 107	49.55

^aEquivalences from the present study calculated as the mean for all strains for each time point.

The mean GE of all strains for each CFU and mg (wet weight) of bacteria for all time points are listed in Table 6. At early time points, 1 CFU by VC is not even 1 GE by qPCR. However, at intermediate time points (21 to 28 days), 1 CFU is approximately 10 GE. Subsequently, the number of GE in 1 CFU increases to a final value of approximately 187.23. At the beginning of the experiment, time point 0 days, 1 mg (wet weight) of M. avium subsp. paratuberculosis is approximately 10⁴ GE. Seven days later, 1 mg of M. avium subsp. paratuberculosis contains 10⁵ GE, and 14 days later, 1 mg (wet weight) of M. avium subsp. paratuberculosis is slightly over 10⁶ GE. After 14 days, the mean for the remaining time points is 4.56 × 10⁷ GE in 1 mg of bacteria. Table 6 also shows the mean CFU of all strains for each time point in relation to 1 mg of bacterial pellet. Most of the time points were in the same range, close to previously published equivalences (1 mg was equal to 10⁶ to 10⁷ CFU), whereas the initial and final time points showed lower values.

(ii) Correlation between methods.

The results of correlation between methods are shown in Fig. 3. The Pearson correlation coefficients between methods were 0.788 (qPCR versus VC; n = 87), 0.493 (VC versus wet weight; n = 87), 0.144 (VC versus T; n = 77), 0.881 (T versus wet weight; n = 77), 0.820 (qPCR versus wet weight; n = 88), and 0.865 (qPCR versus T; n = 77). All these correlations were statistically significant (P < 0.001), except for VC versus T, which was not significant (P = 0.212).

Fig. 3.

Correlation between quantification methods for all strains (■, 316F; ●, 2E; ▵, K10; ○, 81; ♦, 445; □, 764; ▴, 22G; ♢, OC 49). The trend lines represent least squares estimation. All values were divided by the maximum value for each strain and quantification method.

In order to compare the variabilities of the VC and qPCR methods, the coefficient of variation was calculated for bacterial concentrations for each strain at each time point and each growth curve. The strains behaved differently with each method, and no specific pattern was observed (data not shown). When the mean of the CV for all strains at each time point for each growth curve was calculated, as shown in Table 7, the values were lower for the qPCR method. This gives an indication of the variation associated with the experimental conditions and methods, suggesting that qPCR is more reliable.

Table 7.

Mean coefficients of variation for all strains at each time point for both growth curves

Time (days)	CV (%)
	Viable count (CFU/ml)	qPCR (GE/ml)
0	16.59	10.68
7	21.39	18.18
14	20.85	5.38
21	15.61	10.33
28	15.41	5.82
35	25.00	9.21
42	18.77	5.95
49	28.50	3.60
63	18.45	12.79
98	31.70	15.43
140	44.99	3.74

After DNA extraction was achieved, the total DNA concentration was measured with a NanoDrop and recorded. DNA concentration growth curves were plotted (Fig. 4A). The DNA concentration and qPCR correlated perfectly; the Pearson correlation coefficient was 0.947 (n = 88; P < 0.001), as shown in Fig. 4B.

Fig. 4.

(A) Growth curves monitored by DNA content for vaccine and reference strains (■, 316F; ●, 2E; ▵, K10), bovine field isolates (○, 81; ♦, 445; □, 764), and ovine field isolates (▴, 22G; ♢, OC 49). The error bars indicate standard deviations. (B) Correlation between qPCR (ParaTB-Kuanti) and DNA content. The trend line represents the least squares estimation. All values were divided by the maximum value for each strain and quantification method parameter.

DISCUSSION

Growth parameters and characteristics were determined and analyzed for selected M. avium subsp. paratuberculosis strains in the present study. Generation times for M. avium subsp. paratuberculosis have been reported previously to be over 24 h (25) and from 24 to 48 h (5). In our data, we observed different generation times for each strain for every method tested. However, VC and qPCR gave similar generation times for each strain, and they were in agreement with the results from previous studies (5, 25), since they were all included in the range from 25 to 47 h.

The wet-weight and T methods resulted in longer generation times. For wet weight, this could be due to the fact that at low bacterial concentrations, the error introduced by extra water left over in the tube and within the pellet can be greater, and therefore, variation between the number of bacteria at one time point and the next is not significant enough to make a difference. If that is the case, the growth rate is underestimated and the number of bacteria is overestimated because of the presence of water. In the T method, time points 0 and 7 days were not used because the turbidity could not be detected. This result was expected, since the detection limit of turbidometry readers is approximately 10⁷ viable cells (2), and at these initial time points in our experiment, there were fewer bacteria than that. G values were calculated for the exponential phase in all methods (Table 2), except the T method, which lacked early time point values. In this particular case, growth parameters were calculated using data in the late part of the exponential phase, where the growth rate had progressively slowed down, and as a result, G values were higher and the specific growth rate was underestimated.

The highest production as a mean of all methods was achieved by strains 445, 81, 764, and K10 (bovine isolates and the reference strain) in that order. For the wet-weight and T methods, intermediate production was observed for 316F and 2E, whereas low production was seen for 22G and OC 49. On the other hand, the VC and qPCR methods showed intermediate production for 22G and OC 49 and the lowest production for 316F and 2E. This could be explained by possible differences in the hydrophobic nature of ovine isolates compared to vaccines, which can influence size. The wet-weight method can be affected by the size of the bacteria and differences in water intake by each type of strain. The T method can also be affected by these parameters, causing differences in the amount of light absorbed or dispersed. In such case, 316F and 2E would accept more water, increasing their size and yielding higher production for wet weight and T than 22G and OC 49, whereas VC and qPCR show bacterial numbers independent of size.

The time points at which the highest production was reached were earlier for strains 445, 81, 764, and K10 than for the rest of the isolates. These results are partially in agreement with those reported by other groups (25), who have observed that recently isolated low-passage clinical isolates grow much faster than laboratory-adapted strains. On the other hand, our ovine field isolates (22G and OC 49) grew more slowly than bovine field isolates. This result was also partially expected, since differences in culturability among bovine and ovine strains have been observed before (23). Ovine isolates took 3 to 7 months to be detected in first-attempt cultures, whereas bovine isolates were detected in 2 to 4 months (23). Others have also reported differences in growth rates between bovine and ovine strains (4), and as a matter of fact, for many years, only a few laboratories worldwide were capable of culturing M. avium subsp. paratuberculosis strains of ovine origin. Genomic diversity between ovine and bovine strains has been reported (4, 23, 24), and these differences could be responsible in part for variations in growth due to different nutritional requirements. A recent study states that the differences in growth that have been reported historically among host species and M. avium subsp. paratuberculosis strain types have been strongly determined by the type of culture medium used (33).

The initial inoculum seeded was 1.7 × 10⁵ bacteria/ml, which expressed in log₁₀ units is an inoculum of 5.2. The generation time for this inoculum varies from 1.22 to 1.72 days as determined by VC or qPCR, depending on the strain. These results are in close agreement with those observed by Lambrecht et al. (17), who found that a log₁₀ inoculum size of 5.5 gives a generation time of 1.3 days.

At time zero, we should have at least obtained values resembling the inoculum size, but we did not. However, taking into account all the manipulations done before a sample is measured by the different methods, a reduction in bacterial numbers from the initial seed can be expected. Actually, we believe that this is what happened, since there was a loss of 54.71% and 75.75% for VC and qPCR, respectively, at time zero. Quantification at time zero was not affected in the same way in all methods, since the wet-weight method yielded 12,867.64% more than the initial inoculum estimate. This inconsistency can be explained by the fact that when bacterial suspensions are highly diluted, there is a higher proportion of water compared to bacteria in the pellets by wet weight. After centrifugation, there is always some water retained within the pellet and in the tube. The extra water increases the error when the number of bacteria is low, leading to a higher error in weight measurement, resulting in an overestimation. Therefore, we could conclude that at early stages measurements of diluted cultures or cultures at initial time points should be analyzed with caution, since all methods are prone to large errors. Actually, there is a 1-log-unit difference between the number of GE contained in 1 mg (wet weight) of M. avium subsp. paratuberculosis at day 0 compared to 7 days and at 7 days compared to 14 days.

In our study, qPCR and VC estimates were not equivalent, and the relationship varied with time. At early time points, 1 CFU by VC was not equivalent to 1 GE by qPCR, while after the 21-day time point, the ratio was inverted, with more GE observed per CFU. At initial time points, with low bacterial density, we believe that DNA might have been lost during the extraction process. Later, up to 98 days, clumping begins, so the imbalance at this stage could be attributed to the aggregation of M. avium subsp. paratuberculosis cells. The number of GE contained in 1 CFU could increase with time. The fact that clumping increases the growth rate of mycobacteria both in vitro and in vivo has been reported previously (10, 17). Also, some bacterial dormancy or death, or both, could be taking place due to nutrient depletion. Moreover, it has been well described that physical characteristics of M. avium subsp. paratuberculosis make the organism inappropriate for VC. Finally, at the last time point, 140 days, 1 CFU was 187.23 GE. Nevertheless, between the last two time points there was a longer time (42 days). It is possible that at this stage bacterial dormancy might be fully established.

In addition to dormancy and death, the presence of free DNA in culture may be responsible for overestimation by qPCR. Including washing steps prior to DNA extraction has been reported to reduce the amount of free DNA in suspension (7). In that report, 3 washing steps are recommended. However, we included only one washing step in order to avoid bacterial loss in the centrifugation steps, an event that we think would have had an effect on initial time point pellets that were very small.

The fact that 1 CFU of mycobacteria grown on agar can represent more than one bacterium has been previously reported (15, 21, 30). Klijn et al. (15) state that clumping also occurs in nature, as can be seen under the microscope in fecal samples, and therefore, VC may result in an underestimation of the total cell count by a factor of 100 ± 1,000. Taylor et al. (30) report that the clumping tendency of mycobacteria from the Mycobacterium avium complex may be the reason for 1 CFU representing more than 25 cells.

Alternative bacterial quantification and viability methods were also considered for this study. Direct microscope counting has been previously used in our laboratory and was found to be extremely time-consuming and unreliable due to bacterial clumping, which makes counting individual cells difficult. Bactec (Becton and Dickinson) radiometric and nonradiometric systems require large, expensive, and specific equipment that may not be available in all laboratories. Our previous experience with this method for the isolation of M. avium subsp. paratuberculosis from the blood of Crohn's disease patients and the feces of cattle was not satisfactory, and technical support from the manufacturer was denied repeatedly. For this reason, we did not consider using it when this study was conceived. Other groups have achieved good results with Bactec MGIT 960 (25), but although this method, involving liquid culture, would apparently be better than solid media for CFU estimation, it does not guarantee detection of cells in a dormant state (34). Fluorescence assays that permit quantification and assess the viability of bacteria and that have been validated with M. avium subsp. paratuberculosis need further development (16). In our experience, the fluorescent LIVE/DEAD BacLight (Invitrogen) method did not work well with M. avium subsp. paratuberculosis due to limitations in the sensitivity of the fluorimeter reader (data not shown). Finally, a colorimetry-based 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assay is frequently used in our laboratory for short-term quantification experiments. However, since this method reflects the metabolic status of the cell, we avoided it because we do not think it is appropriate for long-term experiments for a slow-growing bacterium with dormancy issues.

Correlation analyses between methods showed that each time VC was included, correlation coefficients were low. qPCR, wet weight, and T correlated very well. This is what can be expected, since VC is the only method representing viable bacteria, whereas all the other methods represent the whole spectrum: viable, nonculturable, and nonviable bacteria. On the other hand, this lack of correlation between VC and the other methods may reflect the fact that the VC method always underestimates bacterial numbers in a sample due to the clumping problem and the fact that 1 CFU is always much more than one bacterium. Correlations including VC were not satisfactory, except when it was compared to qPCR (0.623). Also, when the coefficients of variation for VC and qPCR were compared, the CV was lowest for qPCR and in most cases very good, ranging between 0 and 10%.

Correlation between the DNA concentration and qPCR was highest (0.947). This is expected, since the NanoDrop measurement and qPCR are performed on the same sample, that is, no manipulation steps take place after extraction, except for dilution of the final sample. From Fig. 4, one could estimate the GE by extrapolation, once the DNA concentration is known. Of course, this would be a rough estimate, and it would not be specific, since DNA contamination of other origin would not be identified.

In conclusion, all the methods examined can be used for M. avium subsp. paratuberculosis quantification, but the choice depends on the purpose of the research and the cell density of the bacterial suspension. From our results, we recommend that wet weight should not be used as an M. avium subsp. paratuberculosis quantification method at concentrations lower than 10⁶ bacteria/ml, since it will greatly overestimate the number of bacteria. T should not be used below 10⁷ bacteria/ml (the minimum detection limit of the apparatus). Even though VC is the gold standard in microbiology for bacterial quantification and cell viability assessment, in our hands, and also as reported by others, this method grossly underestimates the M. avium subsp. paratuberculosis concentration. qPCR is a method that can probably be used over a wider range of bacterial concentrations. It also shows more consistency among repeated samples, and it is fast and accurate. For this method, the lack of viability assessment may appear to be a caveat, but we think that for most purposes, including preparation of an inoculum for challenge experiments, this caveat can be avoided as long as established microbiology techniques are followed (a seed culture of less than 10%) and the bacteria are harvested when the culture is below 5 × 10⁹ GE/ml.