Phylogenetic relationship between the species detected in the 16S rRNA analysis from all samples was depicted along with the division classification (Fig. 3). Figure 3 Phylogenetic dendrogram of bacterial species. Evolutionary neighbour-joining phylogenetic dendrogram of the 16S rRNA partial sequencing data derived from bacteria found in the shelf life experiment of cod loins. The sequence data generated from the samples was analysed and blasted on the NCBI server. The closest relative was used for construction of the dendrogram. The tree is composed of 668 bp fragments of the 16S rRNA gene and from

821 underlying sequences which were clustered in one group if the similarity Ku-0059436 nmr was greater or equal to 98%. Thermococcus onnurineus was used as an outgroup. The vertical bar on the right separates the different phylogenetic classes. A: Bacteroidetes, B: Alphaproteobacteria, C: Betaproteobacteria, D: Gammaproteobacteria. Bacterial diversity

during storage by t-RFLP The number of operating taxonomic units (OTU) was different between combinations of labelled primers and restriction enzymes. Analysis of forward Tfs coupled with HaeIII digestion resulted in 12 OTUs compared to 13 OTUs with AluI in all samples analysed. The reverse Tfs coupled with HaeIII Staurosporine cost or AluI digestion resulted in 12 and 17 OTUs, respectively. Principal component analysis (PCA) of Tf profiles showed a clear difference of the bacterial flora of Adenosine triphosphate newly packaged cod loins (LS, day 0) compared to other samples (data not shown). Excluding the d0-samples from the analysis, a better resolution between other samples in the study was established (Fig. 4). It showed that all samples stored under MA clustered tightly together regardless of storage time, temperature or salt content. The samples

stored under air showed a trend towards the MAP cluster with the exception of the 3 HS samples being situated at an opposite position in the score plot. Therefore, the first principal component (PC1) distinguishes the air-stored HS samples from other samples while PC2 seems to separate the air-stored LS samples from early storage time (D6 Air -2° LS) to late storage (D15 Air -2° LS), where the latter clusters with other air-stored and MAP samples (Fig. 4). Figure 4 Principal component analysis of t-RFLP datasets. Principal component analysis of t-RFLP datasets of the bacterial flora derived from periodic sampling of cod loins in a shelf life experiment. Labels correspond to days of storage (e.g. D6 for six days), packaging method (Air or MAP), storage temperature (0, -2 or -4°C) and salt content (low salt, LS and high salt, HS). A box has been added around the samples clustering tightly together for clarification. In this analysis PC1 explained 44% of the variability and PC2 26%. Except for the d0-samples, P. phosphoreum dominated the bacterial flora in all samples, with 40.