table of contents table of contents

Aspergillus nidulans is a model organism for aspergilli, which are an important …

Home » Biology Articles » Mycology » Metabolic network driven analysis of genome-wide transcription data from Aspergillus nidulans » Results

- Metabolic network driven analysis of genome-wide transcription data from Aspergillus nidulans

Reconstruction of the metabolic network and ORF annotation

The metabolic network of A. nidulans was reconstructed using a pathway-driven approach, which resulted in the assignment of metabolic roles to 472 ORFs that had not previously been annotated (Table 1). The reconstructed metabolic network linked a total of 666 genes to metabolic functions, including 194 previously annotated ORFs in the Aspergillus nidulans Database [9]. The resulting network comprises 1,213 metabolic reactions, of which 1095 are biochemical transformations and 118 are transport processes (Table 1), as well as 732 metabolites. Out of the 1,213 reactions there are 794 that are unique (681 unique biochemical conversions and 113 unique transport processes), indicating that 419 of the reactions in the metabolic network are redundant. All the reactions in the metabolic network are listed in Additional data file 7 (Table S1), as are the abbreviations assigned to the metabolite names (Table S2). The reconstructed metabolic network is to our knowledge the largest microbial network reported to date [24].

Transcriptional responses to changes in the carbon source

In order to be able to identify primarily the effect of carbon source on transcription, we grew the cells in well controlled bioreactors, which enabled us to perform very reproducible fermentations. Figure 1 shows the biomass and substrate profiles for growth on glucose, glycerol, and ethanol. For the fermentations with glucose and glycerol as the carbon sources, the carbon recoveries were above 90% (>98% for glycerol), whereas it was only about 64% for growth on ethanol because of evaporation of the substrate. The batch fermentations were carried out in three replicates on each of the carbon sources investigated (for standard deviations, see Figure 1). For all of the cultivations, the samples for transcriptome analysis were taken in the early exponential phase of growth, with the biomass concentration being in the range of 1 to 1.5 g dry weight/kg. At this stage, dispersed filamentous growth was observed in all cultivations.

Identification of differentially expressed genes in pair-wise comparisons

The expression data for the three biological replicates on the three carbon sources were normalized (Additional data file 8 [Tables S3 to S5]) and compared in a pair-wise manner, in order to detect genome-wide transcriptional changes in response to a change in carbon source. Differentially expressed genes for each of the comparisons were identified by applying a significance statistical test (see Materials and methods, below) and considering a significance level (or cutoff in P value) of 0.01. Table 2 shows the total number of significantly regulated genes within the genome of A. nidulans for the three possible pair-wise comparisons between carbon sources, as well as the number of upregulated and downregulated genes. Because the change in carbon source is expected to result in changes in carbon metabolism, the number of differentially expressed genes that were comprised in the metabolic reconstruction for A. nidulans is also presented for each case. It is observed that there is an over-representation of metabolic genes that exhibit significant changes in expression (metabolic genes only comprise about 7% of the total number of genes). The complete list of genes whose expression was significantly changed in the pair-wise comparisons can be found in Additional data file 9 (Tables S6 to S8; they are also partly illustrated in Figures S1 to S3 in Additional data files 1, 2 and 3, respectively). The differentially expressed genes were functionally classified based on Gene Ontology (GO) assignments provided by CADRE [25] (Additional data file 10 [Tables S9 and S10]).

Gene clustering

The genes were arranged in clusters, according to their expression profiles. In order to reduce the noise in the expression data before clustering analysis, an analysis of variance (ANOVA) test was performed that considered normalized transcriptome data from all of the replicated experiments on the different carbon sources (Additional data file 11 [Table S11]). The complete list of statistically significant genes for different significance levels is presented in Additional data file 11 (Table S12). For a significance level (or cutoff in P value) of 0.05, it was observed that the expression levels of 1,534 genes were significantly changed, of which 251 represented metabolic genes. Clustering analysis was applied to these 1,534 genes, and a total of eight clusters were identified (along with an additional cluster that included discarded genes). These clusters are represented in Figure 2, and the genes belonging to each group are listed in Additional data file 12 (Table S13). The GO annotation available in CADRE [25] was also used for functional classification of the genes included in the different clusters (Table 3). The transcriptional patterns of these 1,534 differentially expressed genes were also used for hierarchical cluster analysis (data not shown), and it was observed that the replicated experiments clustered together, as expected.

Identification of metabolic subnetworks

In order to map overall metabolic responses to alterations of the carbon source, we applied the algorithm proposed by Patil and Nielsen [23] to identify the so-called reporter metabolites and to search for highly correlated metabolic subnetworks for each of the three pair-wise comparisons. This analysis relied on the reconstructed genome-scale metabolic network of A. nidulans, and hence we demonstrated how this metabolic network could be used to map global regulatory structures in A. nidulans. The top 15 high-scoring reporter metabolites for each of the cases are listed in Table 4 (also see Additional data files 4, 5 and 6 [Figures S4 to S6, respectively]).

To identify metabolic subnetworks with co-regulated expression patterns we began by finding high-scoring subnetworks, using the whole reaction set in the reconstructed metabolic network for A. nidulans, and subsequently we repeated the algorithm to identify smaller subnetwork structures. The repetition of the algorithm resulted in more robust solutions and in the identification of smaller networks, as demonstrated earlier for yeast data [23]. Table 5 shows the list of enzymes and transporters comprising the 'small' subnetworks for each of the pair-wise comparisons between the three carbon sources investigated (also see Additional data files 4, 5 and 6 [Figures S4 to S6, respectively]). Figure 3 shows key enzymes and transporters comprising the 'small' subnetwork for the glucose versus ethanol comparison. The 'large' subnetworks are given in Additional data file 13 (Tables S14 to S16). The genes in each of the 'small' subnetworks were classified according to the GO-terms assigned, and the results are presented in Additional data file 14 (Table S17).

rating: 5.00 from 1 votes | updated on: 3 Sep 2007 | views: 15508 |

Rate article: