Bacteria were collected in January 2009 at the dockside of Rouen harbour (Normandy, France) during loading from crop silo onto ship from stocking silo. All operations were done in the morning. Samples were obtained by aspiration of 30, 60, 90, or 190 L (under a flow rate 100 L.min^-1) in the dust cloud of particles generated by crop transfer using an AirTest Omega Biocollector (LCB, France). In this device, air is drawn in through a 0.5 mm grid to remove macroscopic particles and the air flow is directed towards the impact medium where micro-organisms are trapped. In order to limit the stress on bacteria, a semi-solid impact medium, Tryptipcase Soy Agar (TSA), was made with Tryptic Soy Broth (TSB) diluted 1:5 containing agar (10 g.L^-1 ) and amphotericin B (25 mg.L^-1) as a fungicide.

After collection, impacted TSA plates, presented in Figure 1, were incubated at 30°C for 72 h to reveal the colonies of cultivable bacteria. Agar plates showing well separated and countable colonies were selected and before other manipulation, a velvet copy of these plates was done. This velvet was laid on a new TSA plate which afterwards was incubated at 30°C for 72 h. All the formed colonies were transferred to and grown in the same culture in 1:5 diluted TSB and then frozen at - 80°C with a final glycerol concentration of 30%. This mixture of strains collected from the air impaction was designed as "Total Air Sample". Subsequently, from the original impacted TSA plate, individual colonies were removed and isolated as pure cultures on TSA for further analysis and storage at - 80°C as previously described.

Each isolated bacterial strain was initially characterized by morphological observation (colony aspect, cell shape, motility, endospore), Gram staining and biochemical tests (oxidase, catalase and oxidative metabolism). These observations allowed different bacterial groups to be defined. In each group including potential pathogens, one representative strain was chosen and submitted to further characterization on API tests galleries. API kits were operated according to the manufacturer's instructions (BioMérieux, France). For Gram-positive bacteria, API 50CH was used for the identification of Bacillus, API ID STAPH 32 for identification of germs of the genus Staphylococcus or Microccocus and API Coryne for Gram-positive catalase-positive non-sporing rods. For Gram-negative bacteria, API 20E was used for the identification of Enterobacteriaceae and API 20NE for non fermenting rods. Species identification of representative strains was confirmed by 16S RNA sequencing. For amplification of the complete 16S RNA gene, universal primers UNI_OL (5¢-AGAGTGTA GCGGTGAAATGCG-3¢,) and UNI_OR (5¢-ACGGGCGGTGTGTACAA-3¢,) were used as suggested by Sauer et al. 24 . Amplicons were then purified through migration on agarose gel and sequenced directly by using the amplification primers UNI_OL and UNI_OR (Qiagen, Germany). Afterwards 16S RNA fragments (+/- 750 pb) were analyzed (Qiagen, Germany) and homologies with sequences of other eubacteria were determined by searching the NCBI (http://www.ncbi. nlm.nih.gov/sutils/genom_table.cgi) and RDP (http://rdp.cme.msu.edu/seqmatch/seqmatch_intro.jsp) data banks using BLAST software. An isolate was positively identified when the full-length 16S RNA gene yielded a >98.3% sequence similarity with the closest bacterial species registered in the data banks.

The virulence of the airborne bacterial population, i.e. the "Total Air Samples", was compared to that of individual representatives of bacterial strains collected in the air using the slow killing tests on the nematode Caenorhabditis elegans, accordingly Figure 1.

Experiments were conducted using the wild-type Bristol strain N2 of C. elegans provided by the Caenorhabditis Genetics Center (Minneapolis, MN, USA). Worms were maintained under standard culturing conditions at 22°C on nematode growth medium (NGM) containing 3 g NaCl, 2.5 g peptone, 17 g agar, 5 mg cholesterol, 1 mL 1 M CaCl₂, 1 mL 1 M MgSO₄, 25 mL 1 M KH₂PO₄, H₂O for 1 L of medium. This medium was plated on Petri dishes and Escherichia coli OP50 as added as a normal food source 21 . For virulence tests, synchronized worms of the same development level were obtained by bleaching an adult population using sodium hypochlorite/sodium hydroxide solution 25 . The resulting eggs were incubated at 22°C on E. coli OP50 lawns until the worms reached the L4 life stage (48 h). The stage of development was confirmed by microscopic observation.

For bacterial virulence assays, "Total Air Samples" or individual air strain aliquots standardized by dilution (at OD₅₈₀ = 1) were prepared by spreading 100 μL on 35 mm Petri dishes containing NGM supplemented with 0.05 mg.mL^-1 5-fluoro-2'-deoxyuridine (FUDR), a eukaryote DNA synthesis inhibitor preventing C. elegans egg offspring during the experiments. Plates were incubated overnight at 29°C and then transferred at room temperature for 4 h. In each Petri dish a mean of 20 L4 synchronized worms, harvested in M9 solution (3 g KH₂PO₄, 6 g NaHPO₄, 5 g NaCl, 1 mL 1 M MgSO₄, H₂O in 1 L) were layered. Plates were incubated at 22°C and worm survival was scored every 24 h throughout a 22 days period using an Axiovert S100 optical microscope (Zeiss, Oberkochen, Germany) equipped with a Nikon digital Camera DXM 1200F (Nikon Instruments, Melville, NY, USA). A worm was considered dead when it remained static without grinder movements for 20 s. Results are expressed as percentage of worms surviving every day and were calculated as the mean of 3 independent assays in which each point was the average of 3 replicates. Nematode survival was calculated using the Kaplan-Meier method, and survival differences were tested for significance using the log-rank test (GraphPad Prism version 4.0; GraphPad Software, San Diego, California, USA).

Plates obtained with 30 L air presented a sufficient number of colonies which remained well separated. In cultures obtained with higher air volumes, bacterial colonies were too numerous for correct isolation.

As seen Figure 1, velvet replicates of impacted TSA plates were realized to obtain the "Total Air Sample". A total of 323 bacterial strains were isolated. The corresponding calculated bacterial concentration in the air was about 1.1 × 10⁴ CFU.m^-3 as shown in Table 1. This bacterial concentration in the dust cloud generated by crop loading was slightly higher than obtained in urban air, which ranged from 5.5 × 10² to 2.5 × 10³ CFU.m^-3 7 11 . The large quantity of dusts and particles, which represents the principal support of bacteria in the air 26 can explain this result. It is also interesting to note that the values of bacterial load measured at the centre of the dust cloud are markedly lower than previously measured 27 . This difference should be attributed to the application of more recent procedures aimed at reducing particle emission during ship loading (for instance covered conveyor belts).

Morphological characters, Gram staining and biochemical tests were carried out for the 323 bacterial strains. These orientation tests allowed to distribute the bacterial population in 8 groups, i.e. Gram-negative oxidase-negative strictly aerobic rods such as Acinetobacter, Pseudomonas... (Group 1), Gram-negative oxidase-positive facultatively anaerobic rods such as Aeromonas... (Group 2), Gram-negative oxydase-negative facultatively anaerobic rods mainly Enterobacteriaceae (Group 3), Gram-negative oxidase-positive strictly aerobic rods like Pseudomonas...(Group 4), Gram-positive sporing rods such as Bacillus...(Group 5), Gram-positive catalase-positive nonsporing rods such as Arthrobacter... (Group 6), Gram-positive catalase-positive strictly aerobic cocci such as Micrococcus... (Group 7), Gram-positive catalase-positive facultatively anaerobic cocci such as Staphylococcus... (Group 8). No Gram-positive catalase-negative were found. Gram-negative rods accounted for the greatest part of the bacterial population (72.4%) with the following distribution: 39.9% for Group 3, 17.6% for Group 1, 10.8% for Group 4 and 4.0% for Group 2. Concerning Gram-positive bacteria, the predominant group corresponded to Group 6 (11.1%) followed by Group 5 (5.3%), Group 7 (3.7%), and Group 8 (2.8%). The bacterial composition, characterized by a predominance of Gram-negative bacilli, appeared unchanged compared to a similar study 27 . This observation is logical with regards to the nature of the material at the origin of the particles (crops). The main bacterial group, i.e. Enterobacteriaceae, Group 3, (39.9%), corresponds to bacterial species frequently associated with plants, and behave as epiphyte, endophyte or even pathogens 28 . These bacteria colonize crops during field growth, survive and even proliferate during storage and then are present in the dust cloud 29 . The second group of bacteria in quantity is Group 1, represented by P. syringae, a germ present in the phyllosphere and generally adapted to growth at low temperature 30 . Catalase-positive non sporulated Gram-positive rods, Group 6, representing 11.1% of bacteria in the dust cloud, should correspond to bacteria of the genus Corynebacteria, Rhodococcus or Arthrobacter also usually present in the ground and already detected in urban air 11 and during crop manipulations 29 .

The aim of this study was to monitor the bacterial air quality, and the study was focused on the groups including potential pathogens: Group 1, 3-5, and 8. From each of these groups, a representative strain was randomly chosen among the isolated strains. They were submitted to API gallery identification and 16S RNA sequencing. As shown in Table 2, all the orientation results were confirmed by those of the identification. Indeed, P. syringae is an oxidase-negative Pseudomonas, then it was pooled with Acinetobacter. Brevundimonas is related to the genus Pseudomonas. The correlation between the API gallery identification and 16S RNA is quite good at the genus level.

The discrepancies between identification results, especially at the species level, could be explained by the incomplete covering of clinical API galleries for the identification of environmental bacteria. This imprecision could also result from the 16S RNA strategy of identification, especially when it is applied over a large diversity of environmental bacteria whose compilation is not complete in data bases such as NCBI/PUBMED.

This study deals with the impact of bacteria present in outdoor air on human health. A robust and practical experimental model was therefore needed in relation to respiratory diseases resulting from bacterial infection. The nematode C elegans model is known to model mammalian bacterial pathogenesis 21 and especially in human respiratory dysfunction 23 .

The practical experimental advantages of this free-living worm are its ability to feed solely on bacteria, its short life cycle, and its easy cultivation in large number 31 . The survival kinetics of C. elegans in the presence of each bacterial representative (P. syringae, Serratia marcescens, Brevundimonas, B. cereus, and S. aureus) was studied for a maximum of 22 days (Figure 2). Even in the presence of their normal food source (Escherichia coli OP50) worms progressively die of age, all died after 17 days of culture. It is important to remember that the DNA synthesis inhibitor (FUDR) present in the medium prevents C. elegans egg offspring and population renewal. The survival of worms in the presence of these bacteria was compared to the one measured using the opportunistic pathogen P. aeruginosa PAO1, frequently responsible for respiratory tract infections 32 . Worms died rapidly in the presence of P. aeruginosa PAO1 (10 days) and none of the environmental strains collected in the air had equivalent virulence. In fact, all the survival kinetics were highly significantly different of those obtained in the presence of P. aeruginosa PAO1 (P < 0.0001 log-rank test). P. syringae, S. marcescens, and B. cereus presented against C. elegans increased or similar virulence compared to E. Coli OP50.

Surprisingly, two strains: Brevundimonas sp., and S. aureus appeared less virulent than E. Coli OP50 (P < 0.0001 log-rank test). Even if S. aureus is accepted as usually pathogen and Brevundimonas considered as an opportunistic pathogen, we have to keep in mind that the bacterial virulence depends on the strains and is closely related to the bacterial adaptation to its microenvironment. But from the random selection of the representatives, no strain is known as strictly human pathogen. Also, according C. elegans test, no sanitary threat appeared because all representative strains were less virulent than the standard opportunistic pathogen P. aeruginosa PAO1. As isolation of all the bacterial spots on the impacted plate is rather fastidious work and complex populations should have virulence which differs from individual strains. Another approach is to test the virulence of the "global" impacted bacterial population against C. elegans, as shown in Figure 1. This was done with the bacterial mix, defined as "Total Air Sample". As with the representatives, virulence against nematodes was studied for the "Total Air Sample", corresponding to the bacterial population collected during crop loading on ships (plotted in Figure 2). The kinetics of the survival of worms with "Total Air Samples" was surrounded by most of the representative plots and was significantly different from that of P. aeruginosa PAO1 (P < 0.0001 log-rank test). Moreover, within the first 11 days of the study, the percentage of death of the worm population was greater using "Total Air Samples" than the avirulent standard E. coli OP50. This difference vanished afterwards and in both cases all worms were dead between 17 and 19 days.

The comparison of the lethal effects of the "Total Air Sample" with each environmental representative was clearly illustrated at Day 9 (Figure 3). This time corresponds to the last day of nematodes survival in the presence of P. aeruginosa PAO1. The survival of C. elegans exposed to P. syringae (50.00%), Pantoae spp (55.26%), and B. cereus (52.11%) was in the same range as for worms fed with "Total Air Sample" (47.73%). Nevertheless, as noted at Day 9, S. aureus and Brevundimonas favored the survival of nematodes with 79.55% and 91.66% of survival percentage respectively, compared to 47.73% for "Total Air Sample". Thus, nematode survival in the presence of the "Total Air Sample" seems pertinent to represent all bacteria synergies and individual virulence. The "Total Air Sample" can be considered without great sanitary risk for all healthy and non immuno-depressed people, as each representative strain.

Thence, any environmental sample with the help of a biocollector makes the assessment of the bacterial sanitary risk possible. After incubation of the softed impacted TSA plates, the total and cultivable bacterial population collected can directly feed the C. elegans nematode model, whose latest death shows the good quality and bacterial safety of the sample environment.