METHODS

Farms

Fifteen farms were selected for this study. Ten of them were considered as farms at risk: five were owned by subjects with a previous history of FLD and the other five, located in the same area, were owned by asymptomatic farmers with similar agricultural practices. These farms were located at an altitude of 830–1100 metres, in a rainy area where the prevalence of FLD is maximal.⁵ Their cowsheds and main barns were made of stone and poorly ventilated. Half of these barns were attached to the house. The last five farms, selected as controls, were modern and located at an altitude of 360–580 metres, in a better weather area. Their cowsheds and barns were made of metal, were well ventilated, and were not attached to the house.

Hay sampling

Data related to farming methods was collected during interviews using a standardised questionnaire which included mode of hay packing (bulk or cylindrical bales), storage location (beside or above the cowshed), hay type (first or second crop of hay), and harvesting conditions which could prevent drying. These conditions, defined a priori, were: terrain liable to flood, rain during harvest, and soil in the hay due to numerous tumuli caused by the proliferation of voles. If any of these conditions was indicated for a batch of hay, it was categorised under “bad harvesting conditions”. A batch was defined as a set of bales or bulk collected from the same meadow at the same period and harvested and stored using the same procedure.

Samples were collected three times during the indoor feeding season (November 2000, January and March 2001) from farms at risk and once (in January) from modern farms. For each batch, two samples were collected in sterile bags, using a standardised method, and the mean value of the number of colonies obtained from cultures was taken into account for analysis in colony forming units per gram of hay (cfu/g). Each sampling was associated with a measure of relative humidity using a mobile humidometer (Wile 25, Tripette & Renaud Agro, Villeneuve-la-Garenne, France).

Microbiological analysis

Each sample was frozen at −18°C overnight to kill the mites. Samples were weighed, rinsed with 20 ml of sterile distilled water, shaken vigorously for one minute, and cultured on Petri dishes. Samples were cultured on five culture media as follows: dichloran-glycerol (Oxoid, Unipath, Basingstoke, UK) with 0.5% chloramphenicol (Merck, Darmstat, Germany) at 30°C for mesophilic mould isolation; Actinomycete isolation agar Bacto medium (Difco, Detroit, MI) at 30°C for mesophilic actinomycetes and at 52°C for thermophilic actinomycetes; 3% malt agar (Oxoid, Unipath, Basingstoke, UK) with 10% salt and 0.5% chloramphenicol, at 20°C for osmophilic fungal species; and R8 medium according to Amner and colleagues at 52°C to detect Sacharomonospora viridis.¹¹ The number of cfu per plate was counted after three and seven days of incubation.

Exposure indexes, representing a theoretical amount of inhalable spores, were calculated every two months for A corymbifera, E amstelodami, and W sebi with the following formula: I = (H*Σ(Xi*Ti))/V, with Xi = the concentration of each microorganism in every batch of hay handled during the two month period (in cfu/g), Ti = the weight of each batch handled during this period (in grams), H = the number of hours of exposure during this period, and V = the cowshed volume (in cubic metres).

Serological analysis

Serological tests were performed only for farmers working on farms at risk and already treated at the time of a previous study:⁷ one at the beginning of the indoor feeding season (October) and one at the end (April). Serum precipitins were measured by electosyneresis on cellulose acetate (Sartorius, Goettingen, Germany) according to a previously described procedure using antigens obtained from species isolated from fodder.⁷ Informed written consent was obtained from each subject. The protocol was approved by the local review board for research involving human subjects.

Statistical analysis

The batches which were sampled three times were used to measure the evolution of microorganism concentration and relative humidity during storage (n = 48). The batches which were sampled twice were used to measure harvesting and hay packing conditions leading to the proliferation of microorganisms (n = 71).

Environmental data are known not to show normal distribution. In this case, classic linear models are usually more likely to generate both type I and II errors. Moreover, logarithmic transformation often fails to normalise these right skewed distributions. Consequently, a non-parametric Mann and Withney test (test U) was used to correlate continuous variables (microorganism concentrations and hygrometric measures) and qualitative variables (farm type, harvest, and hay packing conditions). We used Statview 5.0 software (SAS, North Carolina, USA). Probability values less than 0.05 were considered significant.

RESULTS

The number of sampled batches per farm varied from 5 to 15. One hundred and ninety five batches were collected from farms at risk: of the 76 collected in November, 71 were still available for sampling in January and 48 in March. Thirty six samples were collected from modern farms in January and were used to compare the proliferation of microorganisms between modern farms and those at risk.

Microflora of hay and its evolution during feeding season

More than 24 microorganisms were isolated on the 15 farms. Six of them accounted for more than 75% of the total microorganism count: E amstelodami, W sebi, A corymbifera, mesophilic and thermophilic Streptomyces spp, and S viridis. Others species, listed by percentage of positive samples, were: thermophilic actinomycetes (61%), Aspergillus umbrosus (43%), Cladosporium spp (30%), Aspergillus nidulans (26%), Penicillium spp (22%), Aspergillus fumigatus (20%), Aspergillus ochraceus (18%), Alternaria spp (17%), Aspergillus niger (11%), Aspergillus flavus (8%), S rectivirgula (7%), Rhizopus spp (7%), Aspergillus versicolor (6%), Mucor spp.(4%), Scopulariopsis brevicaulis (2%), Epicoccum purpuracens (0.7%), Blastobotrys nivea (0.4%), and Aspergillus terreus (0.4%).

The concentration of microorganisms in hay was not identical during the indoor feeding season (table 1). In most farms, A corymbifera, E amstelodami, and mesophilic and thermophilic Streptomyces spp reached a peak in January, whereas W sebi reached a high concentration in January and stayed at same level. S viridis showed two peaks, one in November and the second in March. S rectivirgula was rarely isolated in these farms; it was also rarely found in modern farms. In farms at risk, most positive samples for S rectivirugula were collected at the end of the indoor feeding season (10 positive samples in March versus only 4 between November and February).

Harvesting and hay packing conditions leading to the proliferation of microorganisms

On modern farms, where samples were collected in January only, the amount of microorganisms was significantly lower than on farms at risk for the same period, particularly for A corymbifera, E. amstelodami, and W. sebi (table 2). On farms at risk, the amount of these microorganisms was related mainly to harvesting conditions. Samples of hay harvested under poor conditions were more contaminated than the others by A corymbifera and E amstelodami concentrations (table 3).

The impact of hay packing technique seems to be lower in terms of mould proliferation. However, during the storage period, mean relative humidity increased in cylindrical bales from 18.2 to 19.6 degrees and decreased in bulk from 18.9 to 18.5 degrees. Consequently, the difference in relative degrees of humidity between hay packing techniques, which was not significant for the first sampling set (p = 0.99), did become significant for the second (p = 0.02) and the third (p<0.01) sampling sets. This evolution could have consequences on mould proliferation. Indeed, A corymbifera concentration was three times higher when hay was stored in high density cylindrical bales rather than in bulk, but the difference was not significant (p = 0.06) (table 3).

Observation of three cases of FLD relapse and relation with exposure

Among the five subjects with a previous diagnosis of FLD, three complained of clinical symptoms suggestive of a relapse.

Patient 1, who had been suffering from FLD for eight years, presented with chronic cough and dyspnoea from November to April with an improvement in the spring. He had been exposed to the highest amounts of both A corymbifera and E amstelodami antigens during the entire indoor feeding season (table 4). He did not notice which batches provoked respiratory symptoms because he usually handled hay from several batches on the same day.

Patient 2 was also exposed to a very high amount of A corymbifera, but only in November and December, when he distributed a mouldy batch which he wanted to dispose of (table 4). This batch was highly contaminated with A corymbifera (110 000 cfu/g, the highest amount of A corymbifera we have ever counted in hay) and E amstelodami (80 000 cfu/g). At the same time, he presented with episodes of coughing and fever a few hours after exposure.

Patient 3 was moderately exposed to moulds, except in January when he showed new signs of FLD (cough, dyspnoea, fatigue, and a weight loss of 15 kg). At this time, he was handling 1.5 tons of batches, samples of which he gave to the laboratory investigators. This hay was strongly contaminated by A corymbifera and E amstelodami (45 000 cfu/g and 168 000 cfu/g, respectively).

In the spring, all three patients presented with positive serology for A corymbifera. Two other subjects (subjects 4 and 5 with no history of FLD) were exposed to fairly high amounts of A corymbifera, though less than that of subjects 1 and 2, but experienced no symptoms. One of them, however, showed a positive serology in the spring (three arcs with A corymbifera antigen) and underwent further medical examination, including spirometry, which revealed no respiratory symptoms or anomaly in respiratory function.

On the whole, exposure indexes to A corymbifera measured in the farms and periods when patients presented with symptoms compatible with FLD (periods 1, 2, and 3 for patient 1; period 1 for patient 2; and period 2 for patient 3) were significantly higher than exposure indexes measured in other farms and periods (median 239, range 52–565 versus median 11, range 2–190; p = 0.001). The same tendencies were noted for E amstelodami, albeit to a lesser degree (median 777, range 462–2709 versus median 140, range 5–1749; p = 0.01).

Exposure to W sebi was also very high on some farms. However, no relation could be established with the occurrence of respiratory symptoms or the appearance of precipitin arcs using specific antigens (data not shown).

DISCUSSION

The amount of microorganisms in hay was significantly lower on modern farms than on farms at risk. On the latter, the main microorganisms reached a peak of proliferation in January. The most contaminated hay had been harvested wet. Hay packing techniques for cylindrical bales may have contributed to the proliferation of microorganisms by maintaining a high humidity level in hay. Three working farmers with a history of FLD suffered a relapse and presented a positive serology for A corymbifera during winter after having been exposed to a significantly higher amount of this microorganism. Similar, but less significant, results were found for E amstelodami but not for W sebi.

To perform this study, we decided to monitor the evolution of hay prospectively during the distribution period. This allowed us to quantify exposure by exposure indexes, which include the level of contamination of all batches handled by farmers during the three periods as well as other specific variables, such as work time. This way, it was possible to take into account the putative variability in microorganism populations within batches of hay and working conditions, both principal factors of exposure. We chose not to perform a sequential analysis of the air in the cowsheds because the amount of microorganisms in the air varies considerably in a given cowshed from one day to another, partially as a result of handling different batches of hay. Indeed, a single measure of air contamination at a given time is not representative of exposure for the entire period studied.

In order to heighten the probability of recording the proliferation of moulds in hay, we selected farms either in which FLD episodes had previously been recorded or which were very similar with respect to location, farming methods, and construction. Because these farms were selected, they were not representative of all farms in our region, either for agricultural practice or for microbiological findings. To widen the scope of the investigation, we included five modern farms located in an adjacent area.

The first outcome of this study is the identification of some of the harvesting and storage conditions which lead to the concomitant proliferation of A corymbifera, E amstelodami, and W sebi. Since the 1960s, fewer than 20 microbiological studies have been performed in dairy farms. They measured all airbound microorganisms globally, sometimes in large categories (fungi, actinomycetes, other bacteria, etc). These one-time measures were taken over a short period of time and linked to a specific task (hay and straw handling, drying, unbaling, etc). Hence, they do not reflect the long term evolution of hay.^12–15 In this study, we show that the concentration of the majority of microorganisms in hay increases between October and January, then slowly decreases between January and March. The peak in the middle of winter corresponds to the period when the number of FLD cases was highest in our region. We were able to ascertain the onset of the disease for 24 of the 29 cases of FLD diagnosed and registered in our hospital between 1997 and 2002. For 16 of them, the first symptoms were felt between December and February (data not published). The microbiological evolution of hay was not limited to two periods, as is commonly supposed (harvest flora followed by storage flora after fermentation).^16,17 Regarding relative humidity, hay packed in cylindrical bales became more humid at the end of the winter fodder season, whereas hay packed in bulk became drier. The reasons for this increase in humidity were not clear, but we presume that it was linked to the metabolism of the microorganisms. Gregory and colleagues, reporting in-depth studies of the evolution of hay, attributed the role of A corymbifera to furthering the increase of S rectivirgula.^18,19 It is worth noting that these studies focused essentially on the two months following harvest. These authors suggested that excess humidity—more than 30% water content in the field—leads to increased heating of the hay as well as a selection of microorganisms. First, thermo-tolerant fungi (that is, A corymbifera) cause biochemical changes and an increase in temperature, which provide the optimum conditions for S rectivirgula and other thermophilic actinomycetes to grow. In our area, humidity levels are lower in hay at harvest time (mean 23%, range 18–35%),²⁰ so it would seem that this process is interrupted or slowed down. This would partly explain why the isolation of S rectivirgula occurs later, during the indoor feeding season.

The influence of hay packing techniques or storage building type was not clearly shown in our study, even if cylindrical bale storage was associated with a higher, albeit not significantly higher, level of A corymbifera proliferation. The influence of cylindrical bale hay packing techniques has been observed on the proliferation of S rectivirgula, also a thermotolerant species.²¹ In another study performed in our region, limitation of the proliferation of microorganisms was associated with ventilated bulk.⁴

Secondly, despite the low number of subjects included in this study, three subjects presented with episodes of respiratory attacks, which were confirmed by a pulmonologist to be symptoms of FLD. These subjects had a previous history and confirmed diagnosis of FLD (exposure to moulds, respiratory symptoms suggestive of the diagnosis, inspiratory crackles, low CO diffusing capacity, suggestive high resolution thoracic computed tomography scan features and lymphocytic proliferation observed in bronchoalveolar lavage). Serological tests performed at the end of the indoor feeding season were positive for the three patients. In two of these cases, we were able to identify the bales of hay which were at the root of contamination; the concentration of A corymbifera was very high (45 000 cfu/g) in one case and exceptionally high (110 000 cfu/g) in the other. Although the concentration of E amstelodami was also very high in these samples, it was not that exceptional. The presence of these microorganisms may be indicative of the risk for a given agricultural practice or it may be associated to other more pathogenic microorganisms. Nevertheless, the presence of seric precipitins shows that the farmers who experienced relapses had an immunological reaction against these antigens. These data confirm the results of our previous study,⁷ by means of a different methodology, and provide new arguments for considering A corymbifera and possibly E amstelodami as causative agents of FLD. In the former study, in which FLD patients were matched and compared with healthy controls, we showed that the farms of patients suffering from FLD contained more A corymbifera than those of healthy farmers, and electrosyneresis, performed with A corymbifera and E amstelodami antigens, showed more arcs of precipitins in FLD patients than in control subjects. W sebi has also been suspected, but to a lesser degree than the other two microorganisms. Our current results argue in favour of exonerating W sebi.

Finally, even if these two studies do provide strong arguments for attributing FLD attacks to high levels of exposure to mould, the role of susceptibility and the effect of long term exposure cannot be excluded.²² For these reasons, it is difficult to establish a level of exposure which correlates perfectly with the symptoms of FLD.

From a practical standpoint, classical measures such as the use of protective masks are seldom accepted, not even by patients who experience iterative relapses due to persistent exposure. Consequently, some patients stop working in order to preserve their respiratory function.²³ The fact that the proliferation of A corymbifera specifically concerns only some batches of hay, harvested in poor conditions, and that the most contaminated hay continues to deteriorate during storage may be of interest in the framework of implementing a prevention programme for patients with a history of FLD.

In conclusion, this study allows better characterisation of the circumstances which lead to the proliferation of FLD antigens in hay. It also strongly suggests a close relation between the concentration of moulds—especially A corymbifera—and the risk of occurrence of FLD. Our results open pathways for the implementation of preventive strategies for FLD patients.