IS900 PCR To Detect Mycobacterium Paratuberculosis in Retail Supplies of Whole Pasteurized Cows' Milk in England and Wales.

Applied and Environmental Microbiology, Sept 1996, p 3446-3452.

Douglas Millar, Jon Ford, Jeremy Sanderson, Simon Withey, Mark Tizard, Tim Doran, and John Hermon-Taylor.

Department of Surgery,
St. Georges Hospital Medical School,
London SW17 0RE,
United Kingdom.

Introduction.

Mycobacterium Paratuberculosis is a chronic enteric pathogen which can affect many different species of animals including primates ^8,26. The organism was first identified 100 years ago as the cause of chronic inflammation of the intestine in a German cow. Classically, Johne's disease or paratuberculosis in animals is characterized by the presence in the affected intestine of millions of bacillary-form acid-fast mycobacteria with macrophages but with little additional inflammatory cell infiltrate ^1,2. It is now becoming clear that there is a paucimicrobial form of this condition in animals in which abundant acid-fast M. paratuberculosis organisms cannot be seen but in which there is a florid chronic granulomatous inflammatory response ^3,11. M. paratuberculosis is rarely cultured in its bacillary form from these animals. This clinicopathological spectrum of pluribacillary-paucimicrobial paratuberculosis in animals is reminiscent of the extremes represented by the lepromatous and tuberculoid forms of leprosy in humans ³³. Progress in our understanding of M. paratuberculosis and of the disease it causes has been considerably retarded over the years by the sometimes great difficulty of identifying this agent by conventional culture in the laboratory ⁶.

Work carried out in our own laboratory in 1985 led to the discovery and characterization of IS900, an unusual DNA insertion element so far shown to be unique to M. paratuberculosis, with about 18 copies stably integrated into its genome ^20,27. IS900 PCR applied to long-term in vitro cultures and DNA extracts of intestinal tissues implicates M. paratuberculosis in chronic enteritis of the Crohn's disease type in humans ^{13,15,17,25,31,35,37,42}. This miserable condition has steadily increased in frequency since the 1940s, especially in Britain ^16,24,34. Apparently healthy subclinically infected cows are known to secrete the organism in their milk ^36,39,40. The present investigation was designed to determine whether M. paratuberculosis may be present in retail supplies of whole pasteurized cows' milk obtained over a period of 19 months throughout central and southern England and south Wales.

Materials and Methods.

Milk sample processing

A large number of preliminary experiments were carried out Jan-June 1991 to identify suitable conditions for the experimental processing of milk samples prior to IS900 PCR testing for M. paratuberculosis. The final procedure adopted was as follows: milk samples were brought into a dedicated unit and handled within a class 2 safety cabinet. The outside of the carton or bottle was cleaned with 100% ethanol. After shaking, 15ml of milk were aspirated using a sterile syringe and needle without opening the container. The aspirated sample was immediately transferred to a Beckman ultraclear quickseal centrifuge tube, sealed and centrifuged at 41000 x g for 1 hour at 4^oC. This generated three discrete fractions, cream, almost clear whey and pellet. Taking care not to disturb the fractions, a needle capped with a 0.2 µm bacterial filter was inserted into the tube to admit air and the whey fraction gently aspirated using a fresh syringe and needle. The whey was transferred into a 20 ml Sterilin universal and capped. The centrifuge tube containing the residual cream and pellet fractions was frozen -20^oC, cut in half, and the solidified fractions each expressed directly into separate 20ml Sterilin universal tubes. The cream and pellet fractions were then made up to the same volume as the whey fraction using 0.22 µM filtered molecular biology grade water. A 500 µl aliquot of each centrifugal fraction was then transferred to a sterile 1.5ml centrifuge tube, screw capped and boiled in a water bath for 20 minutes, followed by centrifugation in a microfuge. Three 5 µl portions of the resulting supernatants were added to each of three PCR premixes in capped 0.5ml tubes. Process controls consisting of molecular biology grade water taken through all steps in the preparative procedure, were included with each batch of milk samples.

IS900 PCR.

In a separate dedicated laboratory, PCR premixes were prepared containing the following components: primers p90⁺ (5'-GAAGGGTGTTCGGGGCCGTCGCTTAGG-3') and P91⁺ (5'-GGCGTTGAGGTCGATCGCCCACGTGAC-3', 6 ng/ µl of each; Taq polymerase, 2 units (Promega); four dNTPs 200 µM each; reaction buffer, 67mM Tris/HCl, 16.6 mM ammonium sulfate, 1.7 mg/ml bovine serum albumin, 10mM β-mercaptoethanol in T/E (10mM Tris/HCl pH 8.8 and 0.1mM EDTA) to a final volume of 40 µl, subsequently overlaid with 55 µl of light mineral oil. Capped tubes containing the PCR premixes were stored at -20^oC in the dedicated premix laboratory. In the test, the required number of premixes were transferred to the sample processing laboratory. To each was then added 5 µl of 15 mM MgCl₂ and 5 µl of the milk sample processed as described and the cap replaced. Reaction mixes were then centrifuged briefly and placed in a Perkin-Elmer 480 thermal cycler programmed for 94^oC for 5 min and then 93^oC for 1 min, 58^oC for 1 min and 72^oC for 3 min, the last 3 steps being repeated for 40 cycles. Cream, whey and pellet fractions from each milk sample were each tested in triplicate with a simultaneous no-template PCR negative control and process controls.

Southern blot hybridisation

The identity of IS900 PCR amplification products was confirmed by hybridisation with a 229bp internal probe generated by PCR using the primers p25 (5'-CCAGGGACGTCGGGTATGGC-3') and p26 (5'-GGTCGGCCTTACCGGCGTCC-3'). The 229bp probe was radiolabelled by a technique based upon random hexamer priming using the Multiprime DNA labelling system (Amersham U.K) according to the manufacture's protocol. For the screening of large numbers of samples the BioRad Biodot system was used. Transfer buffer (0.4M NaOH/0.6M NaCl) 400 µl, and 15 µl of PCR product were added to each well and the apparatus left for 20 min. at room temperature. The liquid was then sucked through the membrane under vacuum, the wells refilled with a further 400 µl of transfer buffer and left for a further 20 min., when the liquid was sucked through the membrane as before. The membrane (Amersham Hybond N+) was then removed from the apparatus, rinsed in 2 x SSC (1 x SSC contains 0.15M NaCl/ 0.15M sodium citrate), air dried and UV (320nm) cross-linked. Membranes were prehybridised at 65^oC in an orbital shaker using 3 x SSC containing 0.1% bovine serum albumin, 0.1% Ficoll (Pharmacia) 0.1% polyvinylpyrrolidine, 0.5% SDS and 100 mg/ml sheared and denatured salmon sperm DNA for one hour. The denatured, labelled 229bp probe was then added and hybridisation carried out overnight. The membranes were washed three times in 3 x SSC/0.1% SDS at 65^oC for 30 min. Washing was repeated using 1 x SSC/0.1% SDS and finally 0.1 x SSC/0.1% SDS. Membranes were then air-dried and autoradiographed. To confirm the correct 413bp IS900 PCR amplification product in positive wells, a further 10 µl of the corresponding PCR product was analysed by 1.5% agarose gel electrophoresis. The DNA was transferred to nylon membranes using a Posiblot apparatus (Stratagene Ltd), washed, hybridised as before and autoradiographed.

Spiking experiments

Stock cultures of a bovine isolate of M. paratuberculosis were grown in Dubos broth (Difco U.K.) without albumin but supplemented with 20% new-born calf serum (Gibco BRL) and 2mg/L Mycobactin J (Rhone Merieux). After incubation for 8 to 10 weeks, cells were harvested by centrifugation and washed in phosphate buffered saline. To prepare stock M. paratuberculosis DNA, cells were lysed overnight by incubation at 37^oC in 50 µM Tris/HCl pH8.0, 100mM EDTA, 150mM NaCl containing 1% SDS and 0.2mg/ml proteinase K (Sigma UK). After one cycle of phenol chloroform extraction the DNA was precipitated in 2M ammonium acetate with an excess of 100% ethanol. After centrifugation and aspiration of the supernatant, the DNA pellet was taken up in TE buffer and stored -20^oC in 100 µl aliquots. The concentration of DNA was determined spectrophotometrically.

In spiking experiments using free M. paratuberculosis DNA, 30pg DNA was added to 15ml of IS900-negative whole pasteurised cows' milk which was then subjected to processing and subsequent IS900 PCR as described. M. paratuberculosis cultured in liquid media by the method used, grows in tight clumps. Clumps of M. paratuberculosis harvested from these cultures were therefore bead-beaten for 1 min using 0.4mm glass beads (BDH) and a minibead beater (Biospec Products Inc. Bartlesville, OK). Microscopic examination showed that most of the clumps were disaggregated by this procedure and serial dilutions of a suspension of washed M. paratuberculosis could be counted with reasonable accuracy. In spiking experiments using intact M. paratuberculosis, 100 µl aliquots of a cell suspension were added to 15ml samples of milk to give a range of final concentrations 50 to 2000 M. paratuberculosis cells per ml. The spiked milk samples were then processed as previously described.

Raw Milk

Twenty ml samples of milk were obtained directly into screw-capped Sterilin sample tubes from the washed teats of 24 healthy Fresian cows in the University of London's Wye College herd, using fresh disposable gloves for each sampling. This inbred herd has been free of Johne's disease for many years. Six 20ml coded samples of raw milk taken in a similar manner, were obtained from the Institute of Animal Science and Health, The Netherlands (gift of Dr Douwe Bakker). These were centrifuged, processed and subjected to IS900PCR as described. Finally, samples of raw milk were obtained from two cows with clinical Johne's disease (confirmed by faecal culture) and from 4 apparently healthy animals from the same herd on a small farm in Hampshire. After cleansing the udders, washings from the outside of the teats of the Johne's diseased animals were also retained. In this case, total DNA extracts were prepared from these samples using 6M guanidinium thiocyanate followed by extraction of the crude DNA on Promega Wizard resin according to manufacturers' instructions. Aliquots 5 µl, of purified DNA were tested by IS900 PCR in triplicate.

Collection of retail milk samples.

Cartons and bottles of whole pasteurised cow's milk were obtained from retail outlets widely throughout central and southern England continuously week by week from 1st September 1991 to 31st March 1993. In general samples were obtained on Saturday or Sunday, kept in a domestic refrigerator 4^oC and brought directly into the dedicated unit in the department on Monday morning, where they were processed and tested as described.

Long term culture.

From the outset we recognised that a relationship between a positive IS900 PCR signal and the presence or absence of viable M. paratuberculosis in a retail milk sample would be difficult to establish for a species of mycobacterium which historically is slow growing or impossible to culture. Although other studies in our laboratory demonstrated the ability of reverse transcriptase (RT) PCR to identify IS900 mRNA in cultured M. paratuberculosis cells ²⁸, the overall sensitivity of that approach fell short by several orders of magnitude from that which would be required to be practically applicable to this problem. Reliance had therefore to be placed on the traditional but uncertain methods of sample decontamination followed by long term culture.

Aliquots of centrifugal pellets and in some cases cream fractions from 18 IS900 PCR positive and 36 IS900 PCR negative retail milks, were decontaminated by incubation overnight at room temperature with 0.1% benzalkonium chloride or 0.1% or 0.75% cetyl pyridinium chloride in 10ml PBS. After centrifugation and washing, samples were resuspended in a small volume of sterile PBS and 100 µl aliquots inoculated into 10ml Dubos broth supplemented with Mycobactin J as described, using 25cm² tissue culture flasks. Flasks were sealed and incubated at 37^oC. Groups of at least 6 and in a few cases up to 27 culture flasks were established for each milk sample together with 8 medium-only and 6 buffer controls. A 100 µl sample of culture fluid was withdrawn with great care not to introduce contamination, from an individual flask within the group of cultures derived from each of 4 of the IS900 PCR positive milks at intervals over the period of incubation. These were examined by microscopy and IS900 PCR.

Overall there were 575 liquid culture flasks which were incubated for between 13 and 40 months. Then the macroscopic appearance of each culture was recorded, the flasks opened and the contents transferred to a Quickseal centrifuge tube made up with sterile PBS and centrifuged 41,000xg for 30 minutes. The centrifugal culture pellets were resuspended in 1ml residual supernatant. A 5 µl aliquot of this was placed on a microscope slide with 5 µl of formol milk, air dried, fixed with formaldehyde vapour, stained by the Ziehl-Neelsen method and examined microscopically. A further 850 µl portion of resuspended culture pellet in a screw-capped tube was made 6M by the addition of solid guanidinium thiocyanate. To this was added 0.1% Tween 20, 10 mM EDTA, 1mM β-mercaptoethanol and incubated at 37^oC overnight. The sample was then boiled in a water bath for 20 min, microfuged to precipitate debris, and DNA extracted from 500 µl of the supernatant by the Promega Wizard system according to manufacturers instructions. Purified DNA was eluted from the resin using 50 µl preheated 80^oC molecular biology grade water. Two 5 µl aliquots of the eluted material were subjected to IS900 PCR as described.

PCR Contamination precautions.

Stringent precautions were continuously applied to exclude and monitor the exclusion of contamination artefact of the PCR. These included the use of one dedicated laboratory located in a remote part of the building for the preparation of PCR premixes. This facility was kept locked and entered by one person only (DM). Lab coat, overshoes, and gloves were worn and changed regularly. All manipulations were performed in a Class 2 safety cabinet UV irradiated between use. All surfaces were cleaned at the end of each work period with 1M HCl. All consumables for the preparation of premixes were taken directly to the premix laboratory on delivery. Premixes were stored frozen in capped centrifuge tubes, opened on one occasion only for the 5 µl sample and Mg²⁺ addition.

The second dedicated laboratory for milk sample processing also contained a Class 2 cabinet and all surfaces were regularly cleaned as described. Protective clothing was again committed to this laboratory. All pipettes were dismantled and washed after use and the barrel soaked overnight in 1M HCl. Plugged tips only were used. PCR amplification and product analysis was carried out in a third separate dedicated laboratory. A molecular biology grade water sample was simultaneously brought through all processing steps throughout the preparative procedure and tested by IS900 PCR in triplicate. An additional PCR no-template control was included with each run.

Results.

Tests on spiked and raw cows' milk. IS900 PCR showed that disaggregated intact M. paratuberculosis spiked into whole pasteurized cows' milk gave a signal which segregated into the cream and pellet fractions after centrifugation and processing as described above (Fig.1). The overall sensitivity of detection in spiking experiments was estimated to lie in the range of 200 to 300 M. paratuberculosis bacilli per ml. of milk. Spiking with free M. paratuberculosis DNA at 10fg/5 µl of milk (2 pg/ml) resulted in an even distribution of the IS900 PCR signal across the centrifugal fractions.

Application of these methods to raw milk from 24 individual healthy cows in the University of London herd demonstrated the absence of detectable M. paratuberculosis in all milk samples in the presence of correctly positive internal controls. Six representative results are shown in fig.2a. Positive IS900 PCR signals segregating principally to the cream and pellet fractions were observed in four of the six raw milk samples from the Netherlands tested blind (fig.2b). When the origin of these positive samples was decoded, one (from cow 3) was from an animal with heavy fecal shedding of M. paratuberculosis (by culture) which also showed early clinical signs of Johne's disease and whose milk tested positive in all three fractions. Of the other three positive milk samples, two (from cows 1 and 5) were from clinically quite healthy animals, nonetheless shown by culture to be light fecal shedders of M. paratuberculosis. The last positive milk sample in this group, which was weakly positive by IS900 PCR with signals in the cream and pellet fractions, came from an animal (cow 6) which was clinically quite healthy and fecal culture negative but was in a herd which also contained infected animals. The remaining two negative milk samples were from clinically healthy fecal culture-negative animals.

The total DNA samples extracted from the raw milk of both animals with Johne's disease tested in the Hampshire herd were strongly positive by IS900 PCR (Fig.3). Washings from the outside of the teats were negative. All process and PCR controls were correctly negative. Milk from one of the four apparently healthy animals tested in the same Hampshire herd was also strongly positive for M. paratuberculosis. This milk enters the pool submitted to the pasteurization process prior to retail distribution.

Tests on retail cows' milk. Figure 4 shows the regions of central and southern England and south Wales included in the study and the sites where the 312 samples of whole pasteurized cows' milk were obtained from retail outlets. Examples of the results of IS900 PCR tests on individual cartons and bottles are given in Fig. 5. Five of the milk samples shown in the figure exhibited strongly positive tests. In three of these, a signal was seen in the cream and pellet fractions. All simultaneous process controls and no-template PCR controls were correctly negative. All PCR-positive controls were correctly positive.

Figure 6 shows the percentage of retail milk samples testing positive by IS900 PCR, month by month from September 1991 to March 1993. Coming in surges especially in January, February, and March and in September, October , and November, when up to 25% of retail milk samples tested positive, an overall 22 of 312 samples (7.05%) contained M. paratuberculosis with an abundance equal to or greater than the detection limit of the system. In 18 of the 22 positive milk samples (81%), the PCR signal segregated to the cream or pellet fractions or both, consistent with the presence of intact organisms.

Figure 7A shows the results obtained with a 1-liter milk carton from a supermarket in Surrey, United Kingdom, for which the PCR amplification products were analyzed by agarose gel electrophoresis and autoradiography. Strongly positive signals are seen in the cream and pellet fractions, typical of the presence of intact mycobacterial cells. Figure 7B and C shows the appearance of a representative liquid culture from this milk sample after 4 weeks of incubation. Typical acid-fast organisms were seen in clumps and clusters, characteristic of the morphology of M. paratuberculosis growing in liquid media. This flask tested strongly positive by IS900 PCR (data not shown). Despite decontamination of the milk samples as described above, florid overgrowth by other organisms occurred in a substantial majority of the long-term liquid culture flasks. Although isolated clumps of acid-fast organisms were observed, none of the liquid cultures demonstrated an obvious abundant growth of clumped acid-fast mycobacteria when examined microscopically after 13 to 40 months of incubation. The centrifugal pellets from these cultures were, however, IS900 PCR positive for M. paratuberculosis from 9 of the 18 PCR-positive milk samples (50%) and from 6 of the 36 PCR-negative samples (16%). Strongly positive PCR results were observed in the centrifugal pellets of multiple culture flasks derived from four of the PCR-positive milk samples and one of the PCR-negative samples. All the simultaneous medium-only and buffer controls incubated over the same period were IS900 PCR negative.

Discussion.

Although the inclusion of a centrifugation step in milk sample processing followed by PCR testing of each of the cream, whey, and pellet fractions in the present study was laborious, it did permit some distinction to be drawn between the distribution of the PCR signal due to free target DNA and that due to intact lipophilic bacillary-form M. paratuberculosis. With free DNA, the PCR signal first occurred either in the whey or across all three centrifugal fractions. With intact M. paratuberculosis, the PCR signal characteristically segregated to the cream and pellet fractions. Some care must, however, be taken in extrapolating these results obtained by spiking with in vitro cultured bacillary-form M. paratuberculosis to those seen with raw or pasteurized cows' milk, since the range of phenotypes exhibited by M. paratuberculosis in native cows' milk has not yet been precisely defined but is likely to include intracellular forms present within Macrophages.

The methods of sample processing an IS900 PCR worked well when tested on raw milk samples obtained directly from known M. paratuberculosis-free and -infected cattle and gave results in close agreement with the predicted results. Twenty-four cows from the University of London Wye College herd known to be free of M. paratuberculosis over many generations all gave uniformly negative IS900 PCR tests in all three centrifugal milk fractions in the presence of correctly reporting positive and negative PCR controls and simultaneous process controls. Strongly positive IS900 PCR signals, particularly in the cream and pellet fractions, or both, of raw milk, correctly identified both clinically diseased and subclinically diseased cows shedding M. paratuberculosis in their feces. One healthy, fecal culture-negative cow from a herd with Johne's disease was weakly positive by IS900 PCR testing of cream and pellet fractions, suggesting that milk may be an appropriate test sample to detect latent M. paratuberculosis infections. PCR-positive and -negative controls and simultaneous process controls all gave the correct results. Results obtained with external washings showed that positive PCR tests on milk were not due to contamination from the outside of the udder.

In a recent telephone survey of veterinary practices and farms in southwestern England, 1% of farms were reported as having cattle with Johne's disease, with about 2% of the animals in the herds being clinically affected ⁴. Studies carried out in the 1950s showed that in herds containing clinically diseased animals, a substantial proportion of apparently healthy animals in the same herd were subclinically infected with M. paratuberculosis ¹⁴. A recent study in which IS900 PCR was performed on intestinal tissues and mesenteric lymph nodes obtained at slaughter, showed that 3.5% of cows throughout southwestern England tested positive for M. paratuberculosis in the absence of clinical or gross pathological signs of Johne's disease ⁵. Subclinically infected cows may shed M. paratuberculosis in their feces and milk ^36,39 as we also show here. Taken together, these data indicate that milk containing M. paratuberculosis shed by subclinically infected cows must be entering the pool presented to pasteurization processes in the United Kingdom and, indeed, elsewhere in Europe.

The present study found that an overall 7% of cartons or bottles of whole pasteurized cows' milk, obtained at random from retail outlets throughout central and southern England over a 19-month period, tested positive for M. paratuberculosis by IS900 PCR. Positive test results clustered in the peak periods from January to March and September to November, with up to 5 months of negative testing in between. At peak periods, up to 25% of retail cartons or bottles were affected. In most of the positive samples, the distribution of the PCR signal in centrifugal milk fractions was consistent with the presence of intact organisms. However, these could have been due to the presence of intact dead mycobacterial cells. Viability is difficult to determine for M. paratuberculosis in conventional culture, because of the requirement for sample decontamination and the very slow growth or unculturable nature of the organism, particularly of human isolates ^{9,10,12,18,21,32,41}. Further investigations of retail pasteurized milk, taking advantage of the advent of accurate quantitative PCR and recently developed processing methods ³⁰, are suggested.

The methods used here for the sample processing and detection of M. paratuberculosis in cows' milk by IS900 PCR were developed and optimized in the first half of 1991. They were applied throughout this study so that the results would be comparable. The overall lower limit of detection of these methods of 200 to 300 M. paratuberculosis organisms per ml of milk, established with reasonable accuracy by spiking with carefully counted disaggregated cultures, was not very sensitive but was more sensitive than an earlier estimate ²⁹. Simpler, faster, and much more sensitive DNA-based tests for M. paratuberculosis in milk and other samples could undoubtedly be developed from the solid-phase hybridization capture methods for sample processing which have since become available ³⁰.

Two studies have shown that laboratory conditions simulating pasteurization which ensure the destruction of M. bovis do not cause complete inactivation of M. paratuberculosis ^{7, 19}. In the most recent of these, residual viable M. paratuberculosis organisms were found in 55% of milk samples spiked with 10⁴ CFU/ml followed by exposure to 71.7 degrees C for 15 seconds, the conditions of the high-temperature short-time commercial pasteurization method. In the present study, 50% of the liquid cultures from IS900-positive milk samples and 16% of cultures from PCR-negative milk samples were IS900 positive in their centrifugal pellets after up to 40 months of incubation, despite florid overgrowth by other organisms. The subsequent occurrence of PCR-positive liquid cultures derived from milk samples which were originally PCR-negative for M. paratuberculosis suggests that the PCR test, when performed by the methods used, underreported the true incidence of contaminated retail milk supplies in this study. The presence of a very slowly growing residual population of M. paratuberculosis cells in retail milk in England, particularly during peak periods, cannot be excluded.

Studies at the University of Pennsylvania have shown that administration of M. paratuberculosis as an oral bolus to healthy Guermsey heifers resulted in a period of fecal excretion which peaked at days 3 to 4 and was over by day 7 ³⁸. At autopsy of the clinically unaffected animals 28 days later, the mesenteric lymph nodes were enlarged, with reactive follicles. However, M. paratuberculosis was not identified in these nodes by microscopy or culture. Although the gut itself appeared macroscopically and microscopically normal, M. paratuberculosis was cultured from the normal ileal mucosa of all the animals, suggesting that this enteric pathogen was retained by this region of the intestine. If a similar mechanism operates in other species, repeated exposure may result in the cumulative acquisition of a resident population of M. paratuberculosis in the human intestine. After months or years, this may lead to the emergence of a paucimicrobial chronic enteritis and clinical disease in people with an inherited or acquired susceptibility.

References.