3.1. General Bacterial Community Composition
454-pyrosequencing of the 16S rRNA gene amplicon library resulted in a total number of 110,751 raw sequences from the 50 samples, stemming from 13 domestic washing machines. After length/quality filtering, a total of 57,563 high quality forward reads and 44,564 high quality reverse reads were received. These data sets were combined and chimeric sequences (18,042) were removed. The remaining 81,206 sequences were then clustered de novo into 9211 OTUs that shared a 97% sequence similarity threshold. Further removal of mitochondrial and chloroplastic OTUs yielded 7080 bacterial OTUs, representing a total of 77,996 high quality sequences with 353–6802 sequences per sample (mean of 1560 reads per sample). After removal of singletons (4150), the whole data set was rarefied to 242 sequences per sample. Finally, 16 phyla, 36 classes, 67 orders, 124 families, 214 genera and 229 species-like OTUs were determined as components of the bacterial community inside the investigated washing machines.
At phylum level, Proteobacteria (85.8%) was by far the dominating phylum, followed by Actinobacteria (5.3%), Firmicutes (3.0%), Bacteroidetes (2.9%) and Acidobacteria (1.1%). At class level, most sequences were affiliated with Gammaproteobacteria (57.8%), followed by Alphaproteobacteria (17.5%) and Betaproteobacteria (10.3%). Further common classes inside the washing machines were Actinobacteria (5.2%), Bacilli (2.7%), Flavobacteria (2.2%) and Blastocatellia (1.1%), whereas the main identified orders were Pseudomonadales (50.9%), Rhizobiales (9.8%) and Burkholderiales (8.3%). Within the family level, most bacteria belonged to Pseudomonadaceae (30.9%), Moraxellaceae (21.5%), and Comamonadaceae (7.0%). The predominant genera could be identified as Pseudomonas (34.3%), Acinetobacter (17.4%) and Enhydrobacter (6.5%).
3.2. Site-Dependent Bacterial Community Composition
Differences in bacterial diversity were investigated by alpha diversity using observed OTUs, Chao1, Shannon and Simpson as parameters (Table 1
All diversity indices showed significant differences across the sampling sites (ANOVA: pObserved = 6.5 × 10−4, pChao1 = 5.6 × 10−3, pShannon = 9.3 × 10−4, pSimpson = 3.8 × 10−3). The highest alpha diversity was found for the detergent drawer, followed by the fibres isolated from the washing solution, and the sump. The lowest alpha diversity was found inside the door seal.
In order to visualize differences in community structure between the different sampling sites, principal component analysis using weighted und unweighted Unifrac measures was done (Figure 1
). Samples that originated from the detergent drawer were clearly distinct from the sump, which in turn were different from the door seal or the fibre samples. A segregation of the samples from door seal and the fibres becomes visible at the unweighted Unifrac distances, whereas the weighted analysis showed an overlay. The statistical analysis by means of PERMANOVA (p
= 1 × 10−4
for unweighted Unifrac and weighted Unifrac) and ANOSIM (unweighted Unifrac: R = 0.4; weighted Unifrac: R
= 0.3, p
= 1 × 10−4
for unweighted Unifrac and weighted Unifrac) showed that the structure of the bacterial community at the sampling sites was significantly different.
Consequently, also the distribution of the different taxa was found to be highly site-dependent (Figure 2
), in particular the phylum of Proteobacteria (Kruskal-Wallis: p
= 8.1 × 10−4
). Proteobacteria were found across all sampling sites, but in case of the door seal and the sump, this phylum accounted for 94.2% and 96.9% of all sequences, respectively, while the proportion in the detergent drawer (76.3%) and on the fibres from the washing solution (75.8%) was significantly lower. Firmicutes (Kruskal- Wallis: p
= 8.8 × 10−3
), however, were mainly found on the fibres isolated from the washing solution (9.3%) and in the door seal (2.2%). Furthermore, the relative abundance of Actinobacteria also depended strongly on the sampling site (Kruskal-Wallis: p
= 0.02). Here, we found frequencies of around one to two percent in the sump and the door seal. The washing solution fibres and the detergent drawer on the other hand showed relative abundances of ~9%. In addition to the most common phyla, other phyla also showed significant differences between sampling sites. For instance, the phyla Planctomycetes (Kruskal-Wallis: p
= 8.1 × 10−4
), Chloroflexi, (Kruskal-Wallis: p
= 8.8 × 10−3
) and Acidobacteria (Kruskal-Wallis: p
= 3.7 × 10−3
) were found mainly in the detergent drawer but rarely at the other sampling sites.
At genus level, the genera Pseudomonas (Kruskal-Wallis: p = 9.8 × 10−3), Acinetobacter (Kruskal-Wallis: p = 0.01) and Enhydrobacter (Kruskal-Wallis: p = 0.03) were found at all sampling sites. However, their relative abundances varied greatly. For instance, the relative abundance of Pseudomonas in the detergent drawer (13.7%) was much lower compared to the sump (56.7 %). On the other hand, the relative abundance of this genus for door seal (32.9%) and fibres (31.4%) was similar. In contrast, Enhydrobacter occurred mostly in the door seal (12.5%) and on the textile fibres (8.2%) but only barely in the detergent drawer (0.8%). Acinetobacter, in turn, occurred more often in the door seal (34.0%), followed by the washing solution fibres (21.7%). Its relative abundance, however, was significantly lower in the detergent drawer (6.9%) and the sump (2.7%).
In order to further identify the ten most abundant OTUs per site at species level, we calculated sequence similarity against the 16S rRNA gene sequences database from EzBioCloud (Table A1
). Notably, this analysis clearly revealed that the OTUs previously identified as Enhydrobacter showed a sequence similarity of 100% to the species Moraxella osloensis
Significant fractions (30–60%) of the 10 relatively most abundant OTUs per sampling site could be categorized as closely related to potentially pathogenic species based on the German TRBA #466, and many of these OTUs were detected at the majority of the investigated sites. For instance, OTUs closely related to Moraxella osloensis
were detected in up to 6 sump, 8 fibre and 9 door seal samples (Table A1
3.3. Effect of Environmental Factors on Community Composition
In addition to the clear effects of sampling site on bacterial community composition, we investigated the effects of further parameters with a potential influence on microbial community composition. Unexpectedly, the performed ANOVA analysis revealed that only the number of wash cycles per month at ≥60 °C seemed to have an impact on the microbial diversity (pObserverd
= 0.04, pChao1
= 0.06, pShannon
= 0.04, pSimpson
= 0.04). Surprisingly, there was a trend towards a higher alpha diversity with an increased number of wash cycles ≥60 °C compared to a lower number of wash cycles at high temperature (Table 1
). Furthermore, we also examined at which sampling site this factor had the strongest effect on microbial diversity. Figure 3
shows that there was a significantly higher alpha diversity in the detergent drawer from machines which undergo 6–10 washing cycles per month at ≥60 °C. At the other sampling sites, no clear influence of this parameter was seen.
Beta diversity revealed no clear differences between a higher and lower number of wash cycles at temperatures above ≥60 °C using PCoA or ANOSIM and PERMANOVA (data not shown). We therefore compared the relative abundances of single taxa between a high and a low number of wash cycles above 60 °C. A significant difference between a high and low numbers of wash cycles ≥60 °C was seen for the order of Xanthomonadales (Wilcoxon: p = 7.3 × 10−3). Its relative abundance increased with a higher number of wash cycles ≥60 °C from 0.6% to 4.8%. At the genus level, a borderline significant difference was determined for Paracoccus (Wilcoxon: p = 0.05). Its relative abundance increased from 0.2% at 1–5-wash cycles to 1.8% at 6–10 high-temperature wash cycles per month. Also, the minor abundant genera Kocuria, Dysgonomonas, Massilia (Wilcoxon: each p = 0.03) differed significantly between these two conditions.