Indoor dust and air concentrations of endotoxin in urban and rural environments



Gabriel Reboux, Department of Mycology, University Hospital Besançon, 2 Bd Fleming, 25030 Besançon Cedex, France. E-mail:


Studies in European children from a farming background have shown that these children have a reduced risk of asthma and atopic sensitization compared to their urban counterparts. It has been suggested that this might be due to exposure to high levels of endotoxin in the farming environment. The aim of this study was to compare indoor endotoxin concentrations in air and dust samples from randomly selected urban and rural dwellings. In the rural area, endotoxins were analysed in farmhouses and nonfarmhouses as well as housing characteristics, lifestyle factors and agricultural practices likely to influence air and dust endotoxin levels. Endotoxin levels were significantly higher in floor (6600 ± 6100 vs 3600 ± 5600 and 3800 ± 17 000 ng g−1; P < 0·001) and mattress dust (2900 ± 4100 vs 1100 ± 2400 and 800 ± 2600 ng g−1; P < 0·001) from farmhouses compared to other rural and urban homes. However, no difference was observed between endotoxin concentrations in the air of urban and rural houses, and airborne endotoxin levels did not correlate to dust levels. Lack of ventilation and direct entry into the house were correlated with an increase in dust endotoxin levels. These results confirm that dairy farming is associated with high exposure to endotoxins in indoor dust samples. No difference was observed between indoor airborne concentrations between urban and rural houses. These results suggest that measuring endotoxin in dust is the most relevant method to assess endotoxin exposure.

Significance and Impact of the Study

Rural dairy farming is associated with high exposure to indoor endotoxins as compared to rural nonfarming houses and urban houses. The time spent on the mattress (7 h for an adult) and of the proximity of the contaminated source should be taken into account with the other causes of exposure.


Studies in European children from a farming background have shown that these children have a reduced risk of asthma and atopic sensitization compared to their urban counterparts (Aberg 1989; Braun-Fahrlander et al. 1999; Riedler et al. 2000). It has been suggested that this might be due to exposure to high levels of endotoxin in the farming environment. Indeed, endotoxin concentrations were found to be higher in homes of children living on farms compared to homes of nonfarming families (Von Mutius et al. 2000). Endotoxins are lipopolysaccharides (LPS) derived from the cell wall of Gram-negative bacteria and are strong inducers of the immune response. It has been shown that LPS activate Th1-type immune responses and possibly interfere with the development of a Th2-type immune response (Martinez 1999). However, there is increasing evidence that stimulation of Treg cell numbers or function by microbial molecules such as LPS may also be involved (Van Oosterhout and Bloksma 2005).

House dust endotoxins were found to be significantly inversely related to the frequency of atopic asthma, hay fever and allergic sensitization in school-aged children from Austrian, Bavarian (Germany) and Swiss populations (Braun-Fahrlander et al. 2002; Eder and Von Mutius 2004). However, studies have shown that genetic background may also play a part. Children with the CC genotype of the CD14 gene had a lower risk of allergic sensitization and eczema when exposed to high levels of endotoxin (Simpson et al. 2006), and also had less specific IgE to aeroallergens (Eder et al. 2005).

Several studies have assessed indoor exposure to endotoxins, mainly through measurement of endotoxin levels in dust samples (Thorne et al. 2009; Chen et al. 2012). High levels of environmental indoor endotoxins were found in rural areas of developed countries and in farming communities (Gereda et al. 2000; Rylander 2006; Adhikari et al. 2011). Few studies have measured endotoxin levels in air samples (Wan and Li 1999; Park et al. 2000; Dales et al. 2006; Horick et al. 2006; Hyvarinen et al. 2006), and only a few groups have measured endotoxins simultaneously in dust and air samples (Park et al. 2000; Horick et al. 2006; Hyvarinen et al. 2006; Noss et al. 2008; Reponen et al. 2010; Adhikari et al. 2011). Only one study has compared endotoxin levels in air samples from urban and peri-rural areas, and no difference was found (Park et al. 2000). To our knowledge, there are no published reports comparing indoor endotoxin levels in air and dust samples from urban and rural environments.

The aim of the study was to compare indoor endotoxin concentrations in air and dust samples from randomly selected dwellings in an urban and rural environment. Both farmhouses and nonfarmhouses were analysed in the rural area, as well as housing characteristics, lifestyle factors and agricultural practices likely to influence air and dust endotoxin levels.


Mean endotoxin concentrations were significantly lower in floor dust samples from urban area compared to farmhouses from the rural area (3800 ± 17 000 vs 6600 ± 6100 ng g−1; P < 0·001). Interestingly, endotoxin concentrations in the dust from floors and mattress samples in the rural area were significantly higher in farmhouses than in nonfarmhouses (6600 ± 6100 vs 3600 ± 5600 ng g−1, P = 0·001 and 2900 ± 4100 vs 1100 ± 2400 ng g−1, P = 0·002, respectively) (Figs 1 and 2). In the urban and rural area, floor and mattress dust endotoxin levels were significantly correlated (r = 0·5; P = 0·002 and r = 0·23; P = 0·03, respectively). Endotoxin concentrations from floor and mattress dust samples did not differ between rural control dwellings and urban dwellings (3600 ± 5600 vs 3800 ± 17 000 ng g−1, P = ns; 1100 ± 2400 vs 800 ± 2600 ng g−1, P = ns, respectively).

Figure 1.

Comparison between urban and rural (farming and nonfarming) floor dust endotoxin concentrations (ng g−1) (mean ± SD).

Figure 2.

Comparison between urban and rural (farming and nonfarming) mattress dust endotoxin concentrations (ng g−1) (mean ± SD).

No difference was found between air samples from urban area and from nonfarming controls and farmhouses in the rural area (P > 0·05). Moreover, endotoxin concentrations in the air samples were not significantly different between farming and nonfarming environments (P = 0·06) (Fig. 3). No significant association was found between dust and airborne endotoxin levels in urban and rural samples (r = 0·05; P = 0·5 and r = 0·23; P = 0·3, respectively).

Figure 3.

Comparison between urban and rural (farming and nonfarming) air endotoxin concentrations (ng m−3) (mean ± SD).

In the urban area, lifestyle factors and environmental characteristics had no effect on endotoxin levels [see results in (Sohy et al. 2005)]. In the rural area, endotoxin concentrations in floor and mattress dust were significantly lower when houses were equipped with mechanical ventilation (2300 ± 2890 vs 6500 ± 8100 ng g−1 in floor dust, P = 0·004 and 1400 ± 3000 vs 3600 ± 4000 ng g−1 in mattress dust, P = 0·04). Endotoxin levels in rural area were also significantly lower in floor dust when the entrance was connected to the basement before going into the living room (2600 ± 2890 vs 5900 ± 6900 ng g−1, P = 0·03). Endotoxin concentrations in mattress dust were increased by the presence of pets (2600 ± 4000 vs 1300 ± 2600 ng g−1, P = 0·03).

Furthermore, levels of endotoxins in mattress dust were higher in nonmodern farms compared to modern farms (Table 1).

Table 1. Rural endotoxin concentration according to farm characteristics
  n Floor dust (ng g−1) P Mattress dust (ng g−1) P Airborne (ng m−3) P
Type of farm
Nonmodern277900 ± 72000·0513700 ± 49000·040·2 ± 0·20·48
Modern153500 ± 36001200 ± 16000·2 ± 0·2
Barn position
Adjacent to house177500 ± 59000·123400 ± 47000·320·3 ± 0·30·07
Separated to house204600 ± 39002200 ± 36000·1 ± 0·1
Storage hay mode
Round bales285900 ± 57000·973200 ± 47000·90·3 ± 0·30·06
In bulk64600 ± 23001100 ± 9000·2 ± 0·1
Other89100 ± 10 3002700 ± 36000·1 ± 0·1


Significantly higher levels of endotoxins were found in floor and mattress dust from farmhouses in the rural area. Dust sampled from farmhouses contained twice as much endotoxin as dust from control dwellings in the rural area and from urban homes. Endotoxin concentrations have already been reported to be higher in dust samples from kitchen floors and children's mattresses in Bavarian and Swiss farming families compared to control children from nonfarming families living in rural areas (Von Mutius 2000). Endotoxin concentrations from floor and mattress dust samples did not differ between rural references dwellings and urban dwellings. This suggests that only farming environment is associated with higher exposure to endotoxins. It has been suggested that clean living conditions and few bacterial contacts may explain intercommunity differences reported in the prevalence of atopic asthma (Weiss 2002). High indoor endotoxin exposure in floor and mattress dust from farm homes may therefore explain, in part, epidemiological observations that farm environment protects against the development of asthma (Riedler et al. 2000).

In our study, no difference was found between airborne concentrations of endotoxins in urban and rural houses. Only a few studies have assessed endotoxin exposure by measurement in indoor air samples (Wan and Li 1999; Park et al. 2000; Dales et al. 2006; Horick et al. 2006; Hyvarinen et al. 2006), although assays of airborne endotoxin may be more representative of true exposure as endotoxins could penetrate into the body through the lungs. Like other studies (Park et al. 2000), no significant association was found between dust and airborne endotoxin levels in urban and rural samples.

Measurement of endotoxins in air samples is usually carried out in workplaces such as waste industries (Smit et al. 2005; Liebers et al. 2006), greenhouses or farms (Radon et al. 2002) where endotoxin levels are very high (i.e. 14·35–855·16 EU m−3 in winter in greenhouses) (Adhikari et al. 2011). Conversely, airborne endotoxin concentrations in indoor environments are usually very low. In 15 Boston homes, airborne concentrations ranged from 0·02−19·82 EU m−3 with a mean of 0·64 EU m−3 (Park et al. 2000). Other studies have reported concentrations ranging from 0·01 to 30·23 EU m−3 with a mean of 0·7 EU m−3 (n = 116) (Korthals et al. 2008). Dales et al. (2006) found a mean of 0·49 EU m−3 and Dassonville et al. (2008) found a mean of 0·5 EU m−3 with a range from 0·005 to 17·74 EU m−3 in homes of Paris newborn babies (n = 140). Recently, Reponen et al. (2010) reported in infant home in Cincinnati (n = 184) a mean of 4·2 EU m−3 with 3·2–5·3 EU m−3 (95% Confidence Interval). We chose to express our results in ng m−3 corresponding to 10 EU m−3 [EC6, USP Reference Standard Endotoxin 10 EU (endotoxin unit) = 1 ng], and the mean concentrations of airborne endotoxins in our study ranged from 0·09 to 0·19 ng m−3 corresponding to 0·9–1·9 EU m−3.

Airborne endotoxin sampling is a complex procedure, and no standardized protocol currently exists. Significant seasonal variations have been described for outdoor endotoxin levels (Long et al. 2001; Madsen 2006), and therefore, season of sampling might also influence indoor airborne endotoxin levels. Dassonville et al. (2008) reported slightly higher indoor airborne endotoxin levels in cold season than in hot season. This was also reported for house dust in German dwelling (Heinrich et al. 2003). By contrast, Park et al., by performing monthly airborne measurement up to 13 months reported lower airborne endotoxin levels in winter than in spring and only little variability over time in settled dust while no seasonal influence on endotoxin levels in bed and bedroom floor dust (Park et al. 2000, 2001). In another study that compared urban and farming homes, there were also no seasonal variations in airborne, bed dust and floor dust samples (Hyvarinen et al. 2006). Therefore, indoor endotoxin sampling appears to be little influenced by seasonal variations.

A comparison of different studies is difficult because of the different methods used for sampling and for processing and analysis of samples. Furthermore, endotoxin units are also expressed heterogeneously. Because of the low endotoxin levels found in indoor air samples, the sampling procedure and extraction method have to be assessed carefully. In our study, air samples were collected during the daytime for 8 h with a pump at a flow rate of 2 l min−1 on two consecutive days. Other studies sampled air for 24 h (Park et al. 2000; Platts-Mills et al. 2005; Dassonville et al. 2008; Reponen et al. 2010) and found very similar concentrations, even though, unlike Dassonville et al. (2008), we did not use pyrogen-free filters and cassettes. Several other methods of airborne endotoxin sampling have been used, such as active airborne dust sampling with an ion-charged device (Platts-Mills et al. 2005) or dust fall collector (Hyvarinen et al. 2006; Noss et al. 2008).

One explanation for the very low levels of endotoxin in the air could be the size of the endotoxin-carrying particles. In previous studies, it was reported that endotoxins were associated with large particles in agricultural environments (Attwood et al. 1986) and outdoor air (Monn and Becker 1999). Consequently, endotoxins do not remain in the air in the absence of disturbance, explaining the very low airborne concentrations. Using a cascade impactor, we also observed that endotoxins were associated with particles >10 μm in size during disturbance in swineries (unpublished data). Kujundzic et al. (2006) found that 50% of endotoxin in the indoor environment was associated with airborne particulate matter <1 μm in size and 50% was associated with particles >1 μm.

In our study, pet ownership was not associated with significantly higher dust or air endotoxin levels in the urban area (Sohy et al. 2005). However, several studies found a correlation between pets and endotoxin exposure (Gereda et al. 2001; Bischof et al. 2002). Other authors failed to find any association between cat or dog ownership and airborne endotoxin levels (Platts-Mills et al. 2005; Dales et al. 2006; Dassonville et al. 2008), perhaps due to the small number of homes with pets in their sample (Dassonville et al. 2008).

In the present study, entry via the basement and presence of ventilation were associated with significantly lower endotoxin levels in floor dust. In the study of Lis et al. (2008), lack of ventilation and transport of bacteria from specific sources (animals, hay, straw, etc.) on clothes were main factors that explained the higher concentrations in air in farmhouses than in urban houses. It is likely that children carry dust containing bacteria on their hair, clothes and skin from farm buildings to the indoor environment (Korthals et al. 2008). In the same way, an increase in ventilation in agricultural buildings has been shown to be negatively correlated with airborne concentrations of dust, endotoxins and Gram-negative bacteria (Reynolds et al. 1994).

Materials and methods

Study design

In the urban area (Strasbourg), samples were collected from 100 dwellings (50 apartments and 50 houses), as described previously (Sohy et al. 2005). In the rural area (Haut-Doubs), samples were obtained from 50 farmhouses randomly selected from a register of farmers' insurance (Mutuelle Sociale Agricole). Fifty rural nonfarming references were matched to these farming samples. Nonfarming references consisted of neighbour houses of the randomly selected farmhouses in the same village belonging to individuals with no direct contact with cattle and who were willing to participate in the study. Domestic pets were allowed in both groups. Endotoxin sampling in urban and rural homes was conducted in the cold season (from December to end of March) during two consecutive years. No sampling was carried out during other seasons. Furthermore, matched farming and nonfarming houses were sampled on the same week.

Dust and air sampling procedure

Dust sampling

Dust samples were collected from the floor and mattress of one occupied bedroom in each dwelling. Reservoir dust samples were obtained using standardized methodology with a vacuum cleaner (Oxygen Z5530, 1500 watts; Electrolux®, Senlis, France) according to international consensus as described previously (Sohy et al. 2005). Briefly, the floor was vacuumed for 2 min m−2 and the mattress for 10 min (single-size bed) or 15 min (king-size bed).

Air sampling

Air samples were collected with a portable Gilair pump (Sensidyne®, Clearwater, FL, USA) connected to a IOM cassette (M 00003700; Millipore, St Quentin, France) containing a glass fibre filter (AP 4003705; Millipore) as described previously (Lieutier-Colas et al. 2001). The samplers were placed at the lapel of the person spending the longest time at home or close to him/her. Individuals were not allowed to wear the glass fibre filter outside the house. Two 8-h samples with an airflow of 2 l min−1 were collected during daytime on two consecutive days.

Endotoxin extraction and assay

All samples were analysed in the same laboratory, by the same technician (Strasbourg).

For dust samples, endotoxins were extracted from 100 mg dust in 2 ml of PBS-LAL (pH 7·2; Biomerieux, Craponne, France). For air measurements, the filter was eluted in 5 ml of PBS-LAL (Biomerieux®). The extract was shaken on a rotator for 60 min at room temperature. The fluid was collected and centrifuged at 1000 g for 15 min; supernatants were frozen at −80°C until endotoxin analysis.

Endotoxin concentrations were assessed using the Limulus Amebocyte Assay (Chromogenix AB®, Möndal, Sweden) as described previously and following the European Standard EN 14031 (2003) (Lieutier-Colas et al. 2001). Serial dilutions of each sample were run in duplicate. Endotoxin concentrations were expressed in ng g−1 of dust and ng m−3 of air, relative to the quantity of dust and to the pump airflow for airborne samples. The detection limit in dust and air was 0·0005 ng ml−1 (1 ng ml−1 = 10 EU ml−1), corresponding to 0·01 ng g−1 for dust samples and 0·003 ng m−3 for air samples. Inter-assay coefficient was 11·9% (n = 20), and intra-assay coefficient was 10·8% (n = 12). If endotoxin levels were below the limit of detection, the half-value of the detection limit was assigned.

Quality control of air sampling

Several quality controls of the air sampling procedure were performed. To evaluate endotoxin contamination of the filters before sampling, five new glass fibre filters were analysed. All were subjected to the same assay on the same day by the same technician. The mean concentration of endotoxin measured in the new filters was 0·01 ± 0·002 ng ml−1. Reproducibility was also assessed by measuring endotoxin levels in eight new filters on eight consecutive days in the same assay by the same technician. The intra-assay coefficient of variation (CV) was 21·8%, and the inter-assay CV was 139·6%.

To check possible contamination of the glass fibre filters in the laboratory during the endotoxin extraction procedure, 20 IOM cassettes were opened and closed in the laboratory, in one randomly selected dwelling and in a swinery. The mean values of airborne endotoxin concentration were 0·09, 0·16 and 0·14 ng m−3, respectively.

Quality of airborne endotoxin measurement was evaluated by carrying out duplicate assays in one home. Repeatability was determined by performing six measurements on six parallel pumps of airborne endotoxin on six consecutive days in one randomly selected dwelling. Intra-assay coefficient was 24·3%. Reproducibility was measured on samples obtained using two pumps in one randomly selected dwelling on five consecutive days. Inter-assay coefficient was 38%.

Environmental questionnaire

Data for housing characteristics and lifestyle factors were obtained by the technicians using a modified version of the Medical Indoor Environment Counsellor questionnaire (De Blay et al. 2003).

The questionnaire covered the following items: smoking habits, presence of pets, presence of house plants, presence and type of mechanical ventilation, type of living room suite, type of floor covering, presence of carpets and age of mattress. Additional questions were asked in the rural area concerning the removal of boots and work clothes at the entrance of the dwelling, wearing house shoes and presence of pets on the mattress.

Specific questions for farmers included the type of farm (modern or nonmodern), method of storage of hay (round bales, in bulk or other type) and distance between the farm buildings and house. In nonmodern farms, cowsheds had a stanchion-type stall, with or without a ventilation system. Modern farms had a milking room, and the cowshed was a free-stall type and was separate from the barn.

Statistical analysis

Statistical analysis was performed using spss 9·0 for Windows (SPSS Inc, Chicago, IL, USA) and stata Release 9 (Stata Corp LB, College Station, TX, USA). Data were expressed as mean and standard deviation. Because the data had a non-normal distribution, nonparametric tests were used to compare endotoxin levels in the urban and rural environments (Mann–Whitney U), in farmers and nonfarmers in the rural area (Wilcoxon test) and to test factors influencing air and dust endotoxin levels (Mann–Whitney U and Kruskal–Wallis tests). Spearman's correlation was used to assess correlations between dust and air endotoxin levels. Statistical significance was defined as a P value of <0·05.


We thank Ms Karine Humbert and Dr Bertrand Sudre for their meticulous sampling made on each part of a great rural area. The authors have no conflict of interest to declare concerning this paper.