Abstract
Background
Since ancient times the filamentous fungus Aspergillus oryzae has been used in the fermentation industry for the production of fermented sauces and the production of industrial enzymes. Recently, the genome sequence of A. oryzae with 12,074 annotated genes was released but the number of hypothetical proteins accounted for more than 50% of the annotated genes. Considering the industrial importance of this fungus, it is therefore valuable to improve the annotation and further integrate genomic information with biochemical and physiological information available for this microorganism and other related fungi. Here we proposed the gene prediction by construction of an A. oryzae Expressed Sequence Tag (EST) library, sequencing and assembly. We enhanced the function assignment by our developed annotation strategy. The resulting better annotation was used to reconstruct the metabolic network leading to a genome scale metabolic model of A. oryzae.
Results
Our assembled EST sequences we identified 1,046 newly predicted genes in the A. oryzae genome. Furthermore, it was possible to assign putative protein functions to 398 of the newly predicted genes. Noteworthy, our annotation strategy resulted in assignment of new putative functions to 1,469 hypothetical proteins already present in the A. oryzae genome database. Using the substantially improved annotated genome we reconstructed the metabolic network of A. oryzae. This network contains 729 enzymes, 1,314 enzyme-encoding genes, 1,073 metabolites and 1,846 (1,053 unique) biochemical reactions. The metabolic reactions are compartmentalized into the cytosol, the mitochondria, the peroxisome and the extracellular space. Transport steps between the compartments and the extracellular space represent 281 reactions, of which 161 are unique. The metabolic model was validated and shown to correctly describe the phenotypic behavior of A. oryzae grown on different carbon sources.
Conclusion
A much enhanced annotation of the A. oryzae genome was performed and a genome-scale metabolic model of A. oryzae was reconstructed. The model accurately predicted the growth and biomass yield on different carbon sources. The model serves as an important resource for gaining further insight into our understanding of A. oryzae physiology.
Similar content being viewed by others
Background
A. oryzae is a member of the diverse group of aspergilli that includes species that are important microbial cell factories, as well as species that are human and plant pathogens [1]. A. oryzae has been used safely in the fermentation industry for hundreds of years in the production of soy sauce, miso and sake. Today A. oryzae is also used for production of a wide range of different fungal enzymes such as α-amylase, glucoamylase, lipase and protease and it is regarded as an ideal host for the synthesis of proteins of eukaryotic origin [1]. In the post genome-sequencing era, various high-throughput technologies have been developed to characterize biological systems on the genome-scale [2]. Discovering new biological knowledge from high-throughput biological data and assigning biological functions to all the proteins encoded by the genome is, however, challenging and allowing systems level investigations of microbial cell factory. For fungi, several genome-sequencing and annotation projects have been presented, including Saccharomyces cerevisiae [3], A. nidulans [4], A. fumigatus [5], and A. niger [6, 7]. Recently, genome sequence of A. oryzae by Machida and his coworkers has been published [8]. Based on their sequence annotation using gene-finding software tools such as ALN [9], GlimmerM [10] and GeneDecoder [11], this analysis 12,074 genes encoding proteins were predicted to be present in the genome [8]. Despite this prediction many genes had not been assigned a definite function, and of the 12,074 genes, more than 50% were annotated as hypothetical proteins. Hence, there are clearly opportunities for refining the gene prediction and improving the annotation. However, the present one dimensional data does not allow for complete annotation of all genes and it would therefore be interesting and potentially fruitful to use integrative biological tools in the process of improving the annotation of fungal genomes [12]. In this process reconstruction of a genome-scale metabolic model is a good starting point as it allows for integration of various types of data. Nowadays, there are several open sources of fungal metabolic models, such as for S. cerevisiae [13], A. nidulans [14], A. niger [15] and a model for the central carbon metabolism of A. niger [16]. These models currently are prominent as one of the most promising approaches to achieve an in silico prediction of cellular function in terms of physiology [17].
The aim of this study is to improve the annotation of the genome sequence of A. oryzae and further integrate enhanced annotated data to construct a genome-scale metabolic model of A. oryzae. The first A. oryzae EST library, sequencing and assembly were performed in order to improve gene prediction. Then functional assignment was done by our developed annotation strategy and a combination of different bioinformatics tools and databases. The bioinformatics tools used were BLAST [18], HMMER [19], and PSI-BLAST [20]. Several databases used were namely the A. oryzae genome database [21], the EST database of A. flavus [22], the A. nidulans genome database [23], the A. fumigatus genome database [24], the S. cerevisiae genome database [25], the Pfam protein families database [26], the COG database [27], and the Non-Redundant (NR) protein database [28]. Subsequently, manual inspection was through in order to achieve a solid annotation for enzyme functions that were needed for reconstruction of the metabolic network. Based on the improved annotated genome, the genome-scale metabolic network was reconstructed. The network was built by comparison with other related metabolic models, namely models for S. cerevisiae [13], A. nidulans [14], and A. niger [15, 16], and biochemical pathway databases, literature, as well as experimental evidence for the presence of specific pathways. The biomass composition was taken from the literature, whereas, maintenance and growth-associated ATP consumption rates were estimated based on literature data on yields and growth rates. Finally, Flux Balance Analysis (FBA) was used to predict the flux distributions in the metabolic network, and the biomass yields as well as growth rates on different carbon sources were estimated to validate the metabolic model of A. oryzae.
Results and Discussion
Gene discovery and validation
The assembled EST sequences of A. oryzae were achieved from this study (see Additional file 1) where were deposited into Genbank database under accession numbers "EY424375–433412". Within our assembled EST data analysis of A. oryzae, we found 9,038 EST contig sequences with a GC content of 51.2% and an average EST length of 738 base pairs (bps). Based on analysis of sequences obtained from Machida and coworkers [8], the A. oryzae genome consists of eight chromosomes containing 37.2 Megabases (Mb) with a GC content of 48.2% and 12,074 annotated genes. According to the described strategy implemented for gene finding (See Methods), the 9,038 EST sequences were searched against the 12,074 previously identified genes [8] in the sequenced genome using various search parameters to create lists of predicted genes with different match stringencies. Using the criteria described in the Methods, many dissimilar sequences between the EST sequences and previously identified gene sequences of A. oryzae [8] were found. This suggests the presence of many newly predicted genes. Interestingly, approximately 12% (1,046 out of the 9,038 EST sequences) were categorized as newly predicted genes in the genome. Many homolog sequences were also found strongly validating previously identified genes [8], with approximately 75% of the total EST sequences (6,773 out of the 9,038 EST sequences) matching earlier identified genes (See Figure 1). To confirm that all the EST sequences do existed in the A. oryzae genome, the 9,038 EST sequences were searched by BLASTN [18] against the complete genome, and the results showed that only 20 EST sequences could not be found to be present in the genome. Therefore, this suggests that the assembled EST data of A. oryzae had very high quality and showed an excellent success rate for gene discovery and validation, even though approximately 13% (1,219 out of the 9,038 EST sequences) could not be used to predict genes, because 6% (582 out of the 9,038 EST sequences) were too short and about 7% (637 out of the 9,038 EST sequences) were too weakly validated in the original gene list using a conservative cut-off. In another attempt to predict new genes in A. oryzae genome, A. flavus EST contigs stored in the TIGR public database [22] were also used because A. flavus and A. oryzae are very closely related [29]. Also, there is a high degree of DNA homology between the two organisms (e.g. aflatoxin cluster > 96%) [29]. A. flavus EST library contained 7,218 sequences with a GC content of 49.7% and an average EST length of 636.2 bps. Using these A. flavus EST sequences to search against the genes in our new gene list for the A. oryzae genome, no new genes were predicted but 3,320 genes in the A. oryzae genome were validated by EST sequences (see Figure 1). Based on all the results of the gene finding a total of 13,120 protein-encoding genes were identified in the A. oryzae genome. This total number of genes derives from 12,074 previously annotated genes by Machida et al and 1,046 newly predicted genes from our assembled EST library.
Identification of protein functions by pairwise comparison
In order to assign protein functions to the 13,120 predicted genes, sequence alignment analysis by pairwise comparison between A. oryzae and closely related fungi was performed. These fungi included A. nidulans, A. fumigatus and S. cerevisiae. Table 1 shows some genome characteristics of the related fungi in comparison with A. oryzae. Initially pairwise comparison was done by similarity searching of the protein sequences of A. oryzae against the protein sequences of other related fungi as described in the Methods. With a chosen threshold of the alignment length (bps) and identity (%), a list of putative protein functions was created. The results are summarized in Table 1. Pairwise comparison shows that A. fumigatus has 6,274 homologs with A. oryzae sequences. It is the highest number of sequence homologs and this indicates the highest percentage (88%) of the homologs obtained between the three species tested. This result is consistent with the fact that A. oryzae and A. fumigatus are the phylogenetically closest species of those evaluated [4, 30]. Upon completion of the similarity searching, the results suggest that 7,161 genes in A. oryzae could be assigned as orthologous genes from the three fungi used for comparison. Of these 7,161 protein sequences, 5,836 sequences were assigned putative protein functions for A. oryzae. These functions were mainly obtained from A. fumigatus (Table 1). The remaining 1,325 sequences that have homologs in the three other fungi could not been assigned any function yet, and they are therefore classified as hypothetical proteins. The putative functions annotated here were classified using biological process (BP) type from the Gene Ontology (GO) database [31]. The genes and functions that have biological process terms involved in metabolism, including both biosynthesis and catabolism, were extracted and used for metabolic network reconstruction. The results of this process show that the A. oryzae genome contains 1,924 genes (15% of the 13,120 total genes) encoding 1,070 different protein functions involved in metabolism.
Metabolic pathway mapping
The metabolic models for S. cerevisiae [13], A. nidulans [14], and A. niger [15, 16] were combined to generate an initial reaction list for the construction of the A. oryzae metabolic network. Duplicated reactions were removed resulting in a list of 1,924 genes and 1,070 functions involved in metabolism. For each enzyme function involved in this reaction list it was searched in the above generated list of metabolic proteins present in A. oryzae. If an enzyme name matched, then the enzyme-encoding genes, enzyme functions and Enzyme Commission (EC) numbers of A. oryzae were selected and mapped onto this reaction list. Hereafter a classification system was established to divide reactions in the whole metabolic network of A. oryzae into 7 main metabolic pathways: carbohydrate metabolism, energy metabolism, amino acid metabolism, nucleotide metabolism, lipid metabolism, cofactor metabolism and secondary metabolism. It is hereby found that the highest number of enzyme-encoding genes is involved in carbohydrate metabolism, which is consistent with the fact that A. oryzae has the ability to use a wide range of carbohydrate substrates. For amino acid and lipid metabolisms, many enzyme-encoding genes were also found. A lower number of enzyme-encoding genes were found in nucleotide, cofactor and energy metabolisms. The lowest number of enzyme-encoding genes was found in secondary metabolism. In fact, the A. oryzae genome contains a lot of enzyme-encoding genes involved in secondary metabolism [29], but most of these genes are without EC numbers and could therefore not be mapped onto the metabolic network. The hereby resulting metabolic network contains several gaps, which means that there are metabolic reactions without corresponding enzymes.
Filling gaps in the metabolic network using an integrated bioinformatics tool
In order to identify genes encoding more enzyme functions and hereby reduce the number of gaps in the metabolic network, an integrated bioinformatics tool was developed and used to identify these missing enzymes. This tool called "G ap F iller for A spergillus o ryzae P athway (GFAOP)" was developed in- house by combining different bioinformatics tools (i.e. BLAST [18], HMMER [19], and PSI-BLAST [20]) and databases (i.e. A. oryzae genome [21], Pfam [26], COG [27], and NR [28]). GFAOP is similar to the McConkey searching algorithm which has been used for enzyme identification in eukaryote genomes [32]. The method is also related to Osterman's method for the identification of bacterial genes encoding metabolic functions [33]. An overview of GFAOP is shown in Figure 2. First, the tool was validated by searching for 441 known protein functions in A. oryzae using the information from the genome database [21]. The tool confirmed 100% accuracy of the prediction. This tool was then used to search for functional activity related to missing enzyme (Gap) in the metabolic reaction. To illustrate this approach, one of the missing enzymes ("D-xylose reductase" (EC: 1.1.1.21)) in the pathway of xylose degradation of A. oryzae is selected as an example. To answer the question of whether there is a gene encoding D-xylose reductase in A. oryzae. GFAOP was applied as follows. First the HMMER program generates a Hidden Markov Model (HMM) profile of this enzyme (D-xylose reductase) from the protein families databases (such as Pfam or COG). Second, a consensus sequence is generated. Third, the consensus sequence is searched against the A. oryzae genome by a PSI-BLAST [20]. Sequences where the hit has suitable statistical significance values are selected and extracted for protein function assignment by searching against the NR protein database [28] using BLAST [18] to verify its probable function.
The result clearly shows that there is a high probability for that the gene called "AO090003000859" encode D-xylose reductase. Based on searching of this gene in the A. oryzae genome database [21], the gene name AO090003000859 is only reported for general prediction and poorly characterized functions. Moreover, the exploration in other databases such as the Genbank, this gene name is only showed to have a region encoding aldo/keto reductase family proteins, but there is no evidence on the specific function of the gene. As a result from using GFAOP, the missing enzyme of D-xylose reductase is entered into the pathway. Our method results in an improved annotation of the genome using the context of the metabolic network. An iterative process was done for filling all the gaps in the whole metabolic network. Ultimately, 210 gaps in the metabolic network were closed using GFAOP. These gaps distributed with 86 gaps in lipid metabolism, 31 gaps in secondary metabolism, 34 gaps in amino acid metabolism, 23 gaps in nucleotide metabolism, 17 gaps in carbohydrate metabolism, 10 gaps in cofactor metabolism, and 9 gaps in energy metabolism.
Characteristics of the improved annotation and reconstructed metabolic network
The annotation process resulted in the improved annotated data shown in Table 2 where the data are compared with values in the A. oryzae genome database by Machida et al [21]. The results show that the number of improved annotated genes is 13,120 which are higher than the number of genes in the database [21]. Of these improved annotated data, the predicted genes and the putative functions are distributed into different groups. The first group contains new putative protein functions assigned to newly predicted genes, and it contains 398 new putative protein functions that are divided into 154 metabolic functions and 244 other functional groups. The second group contains hypothetical proteins assigned to newly predicted genes and it contains 648 hypothetical proteins. The third group is new putative protein functions assigned to proteins previously annotated as hypothetical proteins, and this group comprises 1,469 proteins of which 562 proteins have metabolic functions. The final group contains genes that is found to have the same putative protein function as previously reported in the database [21]. In total the hereby annotated genome of A. oryzae contains 5,391 protein functions of which 3,178 have metabolic functions. Even though the genome still contains 5,214 hypothetical proteins this is less than the 6,683 hypothetical proteins currently reported in the database [21], and our work therefore resulted in a substantial improvement of the genome annotation. An enhanced annotated data were mapped on the A. oryzae genome by using the Perl Scalable Vector Graphics (SVG) Module V2.33 [34]. Figure 3 shows an example of gene and EST mapping on the contig of AP007151 which is a part of chromosome 1 of the A. oryzae genome. The complete genomic map is available as Additional file 2. The list of all ESTs and genes contained on the genomic map is presented as Additional file 3.
As previously mentioned the improved annotation resulted in a final reconstructed metabolic network that contains 729 enzymes, 1,846 (1,053 unique) biochemical reactions and 1,073 metabolites (Table 2). The large number of isoenzymes (indicated by the difference between total biochemical reactions and unique biochemical reactions) points to a very high degree of flexibility in the metabolic network of A. oryzae. The 1,053 unique biochemical reactions are distributed into 832 cytosolic, 172 mitochondrial, 19 perosixomal, and 30 extracellular reactions. There are 281 (161 unique) reactions that function as transport processes, and of these 173 (53 unique) are included on the basis of gene assignments whereas there are no annotated genes for 108 of the transport reactions. All the genes and functions involved in metabolism were inspected manually. A total of 1,314 genes without duplication represented as enzyme-encoding genes are included in the reconstructed network. This corresponded to about 10% of the 13,120 total predicted genes of A. oryzae. For model comparison, the metabolic network of A. nidulans [14] was chosen, and it shows that the metabolism of A. oryzae is much larger than that of A. nidulans as it contains a higher number of genes, enzymes, metabolites and reactions (see Table 2). A list of the reactions in the reconstructed metabolic network that comprised the genes, EC numbers and enzymes was hereby obtained (see Additional file 4). To illustrate a whole network, overall metabolic map of A. oryzae was drawn as shown in Figure 4 (also see in Additional file 5 for full size) to represent a valuable link between genes, enzymes, metabolic reactions and metabolites. The complete metabolite list (with full name) is also given as Additional file 4.
Overall metabolic map of A. oryzae. A full size of metabolic map of A. oryzae is viewed in Additional file 5.
Biomass growth simulation
Using the reconstructed metabolic network, a stoichiometric model was developed and subsequently used to simulate growth. A list of the reactions that comprise the stoichiometric model is presented as Additional file 4. To describe growth, biomass production is regarded as a drain of macromolecules and building blocks required to produce cellular components. The demands on each of these compounds are estimated based on the biomass composition. No drain of free metabolites or dilution of the metabolite pool due to biomass growth is considered [35]. The cellular composition considered for A. oryzae is based on the contents of the main biomass components of A. oryzae [36] as shown in Table 3 (see also Additional file 4 for the original data used to perform this analysis). In addition, concerning the biomass composition, the only parameters that have to be estimated are key energetic parameters: ATP requirement for non-growth associated purposes (mATP), ATP requirement for synthesis of biomass from macromolecules (KATP) and the operational P/O ratio. These parameters can not be determined independently, but if one of the parameters is known the others can be estimated from experimental data. The operational P/O ratio was assumed to be 1.5 [35], and mATP (mmol/gDW) was estimated to be 1.9 and KATP (mmol/gDW) was estimated by fitting model simulation with experimental data obtained at a specific growth rate of 0.1 h-1 [36] with glucose as the sole carbon source. The value of KATP was hereby estimated to be 49 mmoles ATP/g DW.
Assessment of model validation of A. oryzae
The model was evaluated by simulating A. oryzae cell growth on different carbon sources and comparison of the simulated data to the experimentally determined growth rate and biomass yield from literature data [37, 38]. For each carbon source the substrate uptake rate was estimated from measurements of the substrate concentration in the medium, and this value is used as input to the model. From this input the flux distributions corresponding to optimal growth are calculated by maximizing the flux of the reaction leading to biomass. The validation results are shown in Figure 5 and Figure 6. From the results, Figure 5 indicates that the model can accurately predict the maximum specific growth rate (h-1) during batch cultivations on different carbon sources (when the uptake rate of the carbon source is given as input). The accuracy is on average about 98% of the experimentally determined value. Figure 6 also shows that the model can accurately predict the biomass yield (gDW/mmol substrate) during chemostat cultivations on different carbon sources. The small deviation can be explained by kinetic or genetic regulation within the metabolism, which is not accounted for in the model [17].
Conclusion
A strategy for the improved annotation of the genome sequence of A. oryzae was developed. Using our assembled EST library, 1,046 EST sequences (about 12% of 9,038 EST sequences) were discovered as newly predicted genes and about 75% (6,773 of 9,038 EST sequences) were used to validate previously annotated genes. This indicates that the developed annotation strategy is a very useful approach for gene prediction. Applying a combination of various bioinformatics tools and databases, this annotation strategy was successfully applied for function assignment of genes. A high number of newly predicted genes were assigned with 398 new putative functions, and with new putative functions to 1,469 proteins previously annotated as hypothetical proteins. Therefore our analysis results in a substantially reduced number of hypothetical proteins. In particular, more enzyme-encoding genes could be assigned functions and this led to filling of 210 missing enzymes in the metabolic network. Applying the enhanced annotated genome, biochemical pathway databases, other related metabolic models, and the literature, a metabolic network was reconstructed. The network contains 729 enzymes, 1,314 enzyme-encoding genes (10% of 13,120 total predicted genes), 1,073 metabolites and 1,846 (1,053 unique) biochemical reactions. The 1,053 unique reactions are distributed into different compartments, with 831 reactions located in the cytosol, 173 reactions located in the mitochondria, 19 reactions located in the perosixome, and 30 reactions located in the extracellular space. Transport reactions between the different compartments and the extracellular space represents 281 (161 unique) reactions. This metabolic network was formulated to a stoichiometric model. The model was applied for Flux Balance Analysis (FBA) to obtain the flux distributions corresponding to maximized growth. A physiological study on different carbon sources of A. oryzae was performed to validate the genome-scale model, and the model is found to accurately predict the maximum specific growth rate and the biomass yield on different carbon sources. This indicates that the A. oryzae metabolic model is able to simulate the phenotypic behavior and the model will hereby serve as an important resource for gaining further insight into our understanding of the important cell factory A. oryzae.
Methods
An overview of the approach employed here for improved genome annotation of A. oryzae is depicted in Figure 7. For gene discovery and validation, we constructed EST library and performed sequencing as well as assembly as described in following section.
EST library construction
The EST sequences of A. oryzae strain A1560 were constructed from a normalized library and an un-normalized library. The normalized library was constructed by inserting cDNA of A. oryzae in pCMV-Sport6 plasmids between the Mlu I and the Not I sites (Vector – Not I – poly A (3' of insert) – 5' of insert – Mlu I – Sal I- vector). The plasmids were amplified in Escherichia coli EMDH10B-TONA (a recA strain). The un-normalized library was made by inserting cDNA of A. oryzae between the EcoR1 and NotI sites in the vector pYES2. The plasmids were amplified in E. coli DH10B.
EST sequencing and assembly
The EST sequences were generated by sequencing on ABI 377 and ABI 3700 instruments from Applied Biosystems using BigDye terminators version 1 and 2. In total 23,072 EST sequences were produced. Quality clipping, vector removal, E. coli contamination removal and assembly were done with the phredPhrap package [39]. The sequences were assembled into 9,038 EST contigs.
Genome annotation process
The strategy of gene finding as shown in Figure 7A was carried out based on our assembled EST sequences of A. oryzae (see Additional file 1, also available online in Genbank database under accession number "EY424375–433412") together with public EST data of A. flavus [22]. Our assembled EST data of A. oryzae were compared to the genes previously identified [8] in the genome of A. oryzae strain RIB 40 by BLASTN [18]. The purpose of this comparison was to validate genes that were already annotated and to discover new genes that had not been annotated by Machida et al [8]. The 9,038 EST sequences were classified into four categories as outlined in Additional file 3 and described as follows. All sequences shorter than 300 bases were discarded from the analysis. If the length of an EST sequence was over 500 bps and the highest ranking hit had a score lower than 50 bits, then the EST sequence was categorized as a sequence that served as a newly predicted gene. If the length of the EST sequence was over 300 bps and the highest ranking hit had a score over 100 bits, then the EST sequence was categorized as validating an earlier identified gene [8]. If the highest ranking hit had a score lower than 100 bits, the EST sequence was classified as weakly validating a gene [8]. In the effort to predict new genes in the A. oryzae genome, A. flavus EST data from the TIGR database [22] was also used. The cut-off for gene discovery and validation was selected to be the identical as with our assembled EST data of A. oryzae. After performing gene finding, assignment of protein function was done. The main principle was performed based on sequence alignment analysis, metabolic pathway mapping, filling the gaps by integrated bioinformatics tool and lastly manual curation. The sequence alignment was done to assign putative function to newly predicted genes by BLASTX [18]. The newly predicted gene was searched against the NR protein database [28] and Protall_e protein database [Unpublished]. The assignment of putative protein function was transferred if the alignment length of the highest ranking hit was over 50 amino acids and the identity over 25%. The sequence alignment was done through pairwise comparison of protein sequences by BLASTP [18] between A. oryzae and other related fungi (i.e. A. nidulans strain FGSC-A4, A. fumigatus strain Af293, S. cerevisiae strain S288c) as shown in Figure 7B. The criteria for similarity searching were alignment length (bps) and identity (%), with the parameters depending on the type of fungus used for the comparison [40]. An estimated suitable cut-off for S. cerevisiae was an alignment length above 100 bps and an identity higher than 40%. For other related Aspergillus species, the cut-off was an alignment length above 200 bps and an identity higher than 40%. All cut-off values were determined by using sequences with known protein functions. After finishing the annotation process, the metabolic network of A. oryzae was reconstructed. At the beginning, an initial metabolic reaction list for A. oryzae was constructed by combination of S. cerevisiae [13], A. nidulans [14], and A. niger [15, 16] metabolic models. In addition, data collection from metabolic pathway databases, such as KEGG [41] and BioCyc [42], of other organisms was integrated into this reaction list. The improved annotated genomic data (i.e. enzyme-encoding genes, enzyme functions, and EC numbers) were then mapped into the reaction list. In order to visualize all the metabolic reactions, overall metabolic map was drawn (see Figure 4 and Additional file 5 for full size). The improved annotated data were placed onto this map. At the end, gaps that existed in the metabolic network were then filled using an integrated bioinformatics tool that allowed for automatic searching for specific enzyme functions. Finally, manual curation of the model was done for finalizing the reconstruction process.
Metabolic network reconstruction
The metabolic network reconstruction aimed at representing the whole metabolism of A. oryzae, which consists of primary catabolism of carbohydrates, biosynthesis of amino acids, nucleotides, lipids, cofactors and production of Gibbs free energy required for biosynthesis, as well as of secondary metabolism. Combination of different types of information was essential to carry out a solid reconstruction. Information was collected from the improved annotated data of A. oryzae, biochemical pathways, publications on specific enzymes, online protein databases (e.g. Swiss-Prot database [43]) and also literature. In addition, there was physiological evidence for the presence of a reaction or pathway in A. oryzae, e.g. when there was information of presence of a specific enzyme activity or presence of a pathway involved in consumption of a given substrate or formation of a given metabolic product, then the underlying reaction was added to the model, even if there was no annotated gene supporting the presence of the reaction. In the processes of stoichiometry for cofactors as well as the information on reversibility or irreversibility for each reaction, these were verified and added as information into the reconstructed network. Different cellular compartments were considered and consequently biochemical reactions were distributed into four different compartments: the extracellular space, the cytosol, the mitochondria, and the peroxisome [44]. Identification of localization of each biochemical reaction was analyzed according to enzyme localization, which was performed by applying protein localization predictors. Herein, pTARGET [45] and CELLO [46] were selected to predict sub-cellular protein localization because they contain databases of known eukaryotic protein localizations. If there is no information on localization of a biochemical reaction or its corresponding enzyme, then by default this reaction was considered to occur in the cytosol. In addition, the reconstructed metabolic network included transport steps between the different intracellular compartments and between the cell and the environment.
Modeling and simulation based Flux Balance Analysis (FBA)
After the metabolic network was reconstructed, this was transformed into a mathematical framework to perform Flux Balance Analysis (FBA) [47]. This approach is based on conservation of mass under steady-state conditions. This conversion requires stoichiometry of metabolic pathways, metabolic demands and a few specific parameters. An optimal flux distribution can be obtained within the feasible region by using linear programming [48]. A reaction is selected as an objective function that is to be maximized or minimized. For physiologically meaningful results, the objective functions must be defined as the ability to produce the required components of cellular biomass for a specified uptake rate of a selected carbon source. By maximizing the flux towards biomass formation, a flux is obtained for each reaction in the metabolic network.
Model validation of A. oryzae by physiological study on different carbon sources
Model validation is an important step in the reconstruction process. In this study, the model was validated by simulating the rate of biomass formation on different carbon sources in batch experiments. Here the uptake rate of the carbon source was given as input to the simulations. Different carbon sources namely glucose (C6), maltose (C12), glycerol (C3) and xylose (C5), which were selected as they result in widely different physiological responses and parameters. The strain used for generating these data was A. oryzae wild type strain A1560, which was obtained from Novozymes A/S, Denmark. Three biological replicates were done for each carbon source. The fermentations were performed using an in-house fermenter with a working volume of 1.2 L, and operated at 34°C and pH was kept constant at 6 by adding 10% of H3PO4 or 10% NH3 solution. The aeration flow rate was set at 1.2 L/min. The stirrer speed was controlled at 800 rpm for the first 4 hrs and later increased to 1100 rpm. The dissolved oxygen tension was initially calibrated at 100%. The concentrations of oxygen and carbon dioxide in the exhaust gas were measured by a gas analyzer (Magnos 4G for O2, Uras 3G for CO2, Hartmann & Braun, Germany). Biomass dry weight measurements were done as follows: A sample was filtered using nitrocellulose filters (pore size 0.45 μm, Munktell, Sweden), and the filter cake was therefore dried at 110°C overnight. Hereafter the filter was placed in a dessicator overnight, and subsequently, weighed. In addition, the extracellular concentration of sugars, organic acids, and polyols were measured by using high-performance liquid-chromatography (HPLC) on an Aminex HPX-87H, 300 mm*7.8 mm column. The column was kept at 45°C and eluted at 0.6 ml/min with 5 mM H2SO4.
References
Goldman GH, Osmani SA: The Aspergilli. Genomics, Medical Aspects, Biotechnology, and Research Methods (Mycology). 2008, Taylor and Francis group: CRC Press
Kitano H: Computational systems biology. Nature. 2002, 420: 206-210. 10.1038/nature01254.
Fisk DG, Ball CA, Dolinski K, Engel SR, Hong EL, Issel-Tarver L, Schwartz K, Sethuraman A, Botstein D, Cherry JM: Saccharomyces cerevisiae S288C genome annotation: a working hypothesis. Yeast. 2006, 23: 857-865. 10.1002/yea.1400.
Galagan JE, Calvo SE, Cuomo C, Ma LJ, Wortman JR, Batzoglou S, Lee SI, Basturkmen M, Spevak CC, Clutterbuck J, Kapitonov V, Jurka J, Scazzocchio C, Farman M, Butler J, Purcell S, Harris S, Braus GH, Draht O, Busch S, D'Enfert C, Bouchier C, Goldman GH, Bell-Pedersen D, Griffiths-Jones S, Doonan JH, Yu J, Vienken K, Pain A, Freitag M, Selker EU, Archer DB, Penalva MA, Oakley BR, Momany M, Tanaka T, Kumagai T, Asai K, Machida M, Nierman WC, Denning DW, Caddick M, Hynes M, Paoletti M, Fischer R, Miller B, Dyer P, Sachs MS, Osmani SA, Birren BW: Sequencing of Aspergillus nidulans and comparative analysis with A. fumigatus and A. oryzae. Nature. 2005, 438: 1105-1115. 10.1038/nature04341.
Nierman WC, Pain A, Anderson MJ, Wortman JR, Kim HS, Arroyo J, Berriman M, Abe K, Archer DB, Bermejo C, Bennett J, Bowyer P, Chen D, Collins M, Coulsen R, Davies R, Dyer PS, Farman M, Fedorova N, Fedorova N, Feldblyum TV, Fischer R, Fosker N, Fraser A, Garcia JL, Garcia MJ, Goble A, Goldman GH, Gomi K, Griffith-Jones S, Gwilliam R, Haas B, Haas H, Harris D, Horiuchi H, Huang JQ, Humphray S, Jimenez J, Keller N, Khouri H, Kitamoto K, Kobayashi T, Konzack S, Kulkarni R, Kumagai T, Lafon A, Latge JP, Li WX, Lord A, Lu C, Majoros WH, May GS, Miller BL, Mohamoud Y, Molina M, Monod M, Mouyna I, Mulligan S, Murphy L, O'Neil S, Paulsen I, Penalva MA, Pertea M, Price C, Pritchard BL, Quail MA, Rabbinowitsch E, Rawlins N, Rajandream MA, Reichard U, Renauld H, Robson GD, de Cordoba SR, Rodriguez-Pena JM, Ronning CM, Rutter S, Salzberg SL, Sanchez M, Sanchez-Ferrero JC, Saunders D, Seeger K, Squares R, Squares S, Takeuchi M, Tekaia F, Turner G, de Aldana CRV, Weidman J, White O, Woodward J, Yu JH, Fraser C, Galagan JE, Asai K, Machida M, Hall N, Barrell B, Denning DW: Genomic sequence of the pathogenic and allergenic filamentous fungus Aspergillus fumigatus. Nature. 2006, 439: 502-502. 10.1038/nature04572.
Baker SE: Aspergillus niger genomics: Past, present and into the future. Medical Mycology. 2006, 44: S17-S21. 10.1080/13693780600921037.
Pel HJ, de Winde JH, Archer DB, Dyer PS, Hofmann G, Schaap PJ, Turner G, de Vries RP, Albang R, Albermann K, Andersen MR, Bendtsen JD, Benen JAE, van den Berg M, Breestraat S, Caddick MX, Contreras R, Cornell M, Coutinho PM, Danchin EGJ, Debets AJM, Dekker P, van Dijck PWM, van Dijk A, Dijkhuizen L, Driessen AJM, d'Enfert C, Geysens S, Goosen C, Groot GSP, de Groot PWJ, Guillemette T, Henrissat B, Herweijer M, van den Hombergh J, van den Hondel C, van der Heijden R, van der Kaaij RM, Klis FM, Kools HJ, Kubicek CP, van Kuyk PA, Lauber J, Lu X, van der Maarel M, Meulenberg R, Menke H, Mortimer MA, Nielsen J, Oliver SG, Olsthoorn M, Pal K, van Peij N, Ram AFJ, Rinas U, Roubos JA, Sagt CMJ, Schmoll M, Sun JB, Ussery D, Varga J, Vervecken W, de Vondervoort P, Wedler H, Wosten HAB, Zeng AP, van Ooyen AJJ, Visser J, Stam H: Genome sequencing and analysis of the versatile cell factory Aspergillus niger CBS 513.88. Nature Biotechnology. 2007, 25: 221-231. 10.1038/nbt1282.
Machida M, Asai K, Sano M, Tanaka T, Kumagai T, Terai G, Kusumoto KI, Arima T, Akita O, Kashiwagi Y, Abe K, Gomi K, Horiuchi H, Kitamoto K, Kobayashi T, Takeuchi M, Denning DW, Galagan JE, Nierman WC, Yu JJ, Archer DB, Bennett JW, Bhatnagar D, Cleveland TE, Fedorova ND, Gotoh O, Horikawa H, Hosoyama A, Ichinomiya M, Igarashi R, Iwashita K, Juvvadi PR, Kato M, Kato Y, Kin T, Kokubun A, Maeda H, Maeyama N, Maruyama J, Nagasaki H, Nakajima T, Oda K, Okada K, Paulsen I, Sakamoto K, Sawano T, Takahashi M, Takase K, Terabayashi Y, Wortman JR, Yamada O, Yamagata Y, Anazawa H, Hata Y, Koide Y, Komori T, Koyama Y, Minetoki T, Suharnan S, Tanaka A, Isono K, Kuhara S, Ogasawara N, Kikuchi H: Genome sequencing and analysis of Aspergillus oryzae. Nature. 2005, 438: 1157-1161. 10.1038/nature04300.
Gotoh O: Homology-based gene structure prediction: simplified matching algorithm using a translated codon (tron) and improved accuracy by allowing for long gaps. Bioinformatics. 2000, 16 (3): 190-202. 10.1093/bioinformatics/16.3.190.
Majoros WH, Pertea M, Antonescu C, Salzberg SL: GlimmerM, Exonomy and Unveil: three ab initio eukaryotic genefinders. Nucleic Acids Research. 2003, 31 (13): 3601-3604. 10.1093/nar/gkg527.
Asai K, Itou K, Ueno Y, Yada T: Recognition of human genes by stochastic parsing. Pac Symp Biocomput. 1998, 3: 228-239.
Liu ET: Integrative biology and systems biology. Molecular Systems Biology. 2005
Forster J, Famili I, Fu P, Palsson BO, Nielsen J: Genome-scale reconstruction of the Saccharomyces cerevisiae metabolic network. Genome Research. 2003, 13: 244-253. 10.1101/gr.234503.
David H, Özçelik , Hofmann G, Nielsen J: Analysis of Aspergillus nidulans metabolism at the genome-scale. BMC Genomics. 2008, 9: 163-10.1186/1471-2164-9-163.
Andersen MR, Nielsen ML, Nielsen J: Metabolic model integration of the bibliome, genome,metabolome and reactome of Aspergillus niger. Molecular Systems Biology. 2008, 4: 178-10.1038/msb.2008.12.
David H, Akesson M, Nielsen J: Reconstruction of the central carbon metabolism of Aspergillus niger. European Journal of Biochemistry. 2003, 270: 4243-4253. 10.1046/j.1432-1033.2003.03798.x.
Borodina I, Nielsen J: From genomes to in silico cells via metabolic networks. Current Opinion in Biotechnology. 2005, 16: 350-355. 10.1016/j.copbio.2005.04.008.
Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ: Basic Local Alignment Search Tool. Journal of Molecular Biology. 1990, 215: 403-410.
Eddy SR: Profile hidden Markov models. Bioinformatics. 1998, 14: 755-763. 10.1093/bioinformatics/14.9.755.
Altschul S, Madden T, Schaffer A, Zhang JH, Zhang Z, Miller W, Lipman D: Gapped BLAST and PSI-BLAST: A new generation of protein database search programs. Faseb Journal. 1998, 12: A1326-A1326.
Aspergillus oryzae genome database . [http://www.bio.nite.go.jp/dogan/MicroTop?GENOME_ID=ao]
Aspergillus flavus Gene Index . [http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/gimain.pl?gudb=a_flavus]
Aspergillus nidulans genome database . [http://www.broad.mit.edu/annotation/genome/aspergillus_nidulans]
Aspergillus fumigatus genome database . [http://www.sanger.ac.uk/Projects/A_fumigatus/]
Saccharomyces genome database . [http://www.yeastgenome.org/index.shtml]
Pfam database . [http://www.sanger.ac.uk/Software/Pfam]
COG database . [http://www.ncbi.nih.gov/COG]
Non-redundant protein database . [ftp://ftp.ncbi.nih.gov/blast/db/FASTA/]
Payne GA, Nierman WC, Wortman JR, Pritchard BL, Brown D, Dean RA, Bhatnagar D, Cleveland TE, Machida M, Yu J: Whole genome comparison of Aspergillus flavus and A. oryzae. Medical Mycology. 2006, 44: S9-S11. 10.1080/13693780600835716.
Pain A, Böhme U, Berriman M: Hot and sexy moulds!. Nature reviews. 2006, 4: 244-245. 10.1038/nrmicro1388.
Gene Ontology Database . [http://www.geneontology.org/GO.annotation.shtml]
McConkey GA, Pinney JW, Westhead DR, Plueckhahn K, Fitzpatrick TB, Macheroux P, Kappes B: Annotating the Plasmodium genome and the enigma of the shikimate pathway. Trends in Parasitology. 2004, 20: 60-65. 10.1016/j.pt.2003.11.001.
Osterman A, Overbeek R: Missing genes in metabolic pathways: a comparative genomics approach. Current Opinion in Chemical Biology. 2003, 7: 238-251. 10.1016/S1367-5931(03)00027-9.
Perl Scalable Vector Graphics . [http://search.cpan.org/~ronan/]
Nielsen J: Physiological engineering aspects of Penicillium chrysogenum. World Scientific Pub Co Inc; 1996-
Pedersen H, Carlsen M, Nielsen J: Identification of enzymes and quantification of metabolic fluxes in the wild type and in a recombinant Aspergillus oryzae strain. Appl Environ Microbiol. 1999, 65 (1): 11-19.
Prathumpai W, Gabelgaard JB, Wanchanthuek P, van de Vondervoort PJI, de Groot MJL, McIntyre M, Nielsen J: Metabolic control analysis of xylose catabolism in Aspergillus. Biotechnology Progress. 2003, 19 (4): 1136-1141. 10.1021/bp034020r.
Carlsen M, Nielsen J: Influence of carbon source on alpha-amylase production by Aspergillus oryzae. Appl Microbiol Biotechnol. 2001, 57 (3): 346-349.
Ewing B, Green P: Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Research. 1998, 8: 186-194.
Rost B: Twilight zone of protein sequence alignments. Protein Engineering. 1999, 12: 85-94. 10.1093/protein/12.2.85.
KEGG pathway database . [http://www.kegg.com]
BioCyc database . [http://biocyc.org/server.html]
Swiss-Prot database . [http://www.expasy.ch/sprot/]
Carson BD: Microbodies in fungi. A review. Journal of industrial microbiology. 1990, 6: 1-10.1007/BF01576172.
Guda C, Subramaniam S: TARGET: a new method for predicting protein subcellular localization in eukaryotes. Bioinformatics. 2005, 21: 3963-3969. 10.1093/bioinformatics/bti650.
Yu CS, Chen YC, Lu CH, Hwang JK: Prediction of protein subcellular localization. Proteins-Structure Function and Bioinformatics. 2006, 64: 643-651. 10.1002/prot.21018.
Edwards JS, Covert M, Palsson B: Metabolic modelling of microbes: the flux-balance approach. Environmental Microbiology. 2002, 4: 133-140. 10.1046/j.1462-2920.2002.00282.x.
Bonarius HPJ, Schmid G, Tramper J: Flux analysis of underdetermined metabolic networks: The quest for the missing constraints. Trends in Biotechnology. 1997, 15: 308-314. 10.1016/S0167-7799(97)01067-6.
Acknowledgements
This research work was support by a stipend to Wanwipa Vongsangnak from Novozymes Bioprocess Academy (NBA) and Technical University of Denmark (DTU). The authors are grateful to José Manuel Otero, Roberto Olivares Hernandez, Chia-Wen Chang and Rawisara Ruenwai for discussions and critical revisions of the manuscript. Also, we thank Lone Vuholm and Jeanette Thomassen for valuable technical assistance.
Author information
Authors and Affiliations
Corresponding author
Additional information
Authors' contributions
WV carried out the improved annotation, performed the genome scale modeling and wrote the manuscript. PO performed the EST sequencing project, participated in the EST and gene mapping and supervised the bioinformatics work. KH supervised the fermentation process. SK provided the EST and gene mapping and supervised the bioinformatics work. JN supervised the whole work and assisted in manuscript preparation. All authors read and approved the final manuscript.
Electronic supplementary material
12864_2008_1438_MOESM1_ESM.fas
Additional file 1: The assembled EST sequences of A. oryzae. The FASTA file provides that 9038 assembled EST sequences. The title name of each sequence indicates dbEST ID, User ID and Genbank accession number. (FAS 7 MB)
12864_2008_1438_MOESM2_ESM.zip
Additional file 2: EST and gene mapping on A. oryzae genome. The ZIP file provides that 48 separated files with Perl Scalable Vector Graphics. (ZIP 2 MB)
12864_2008_1438_MOESM3_ESM.xls
Additional file 3: Improved annotated data in A. oryzae genome. This file is created to extract gene and EST list existed on genomic map (Additional file 2). (XLS 2 MB)
12864_2008_1438_MOESM4_ESM.pdf
Additional file 4: A reconstructed metabolic network of A. oryzae. A list of all metabolic reactions is shown with EC numbers, enzyme names, gene names and annotation methods. Besides, a list of metabolite abbreviations and biomass compositions are shown also. (PDF 382 KB)
12864_2008_1438_MOESM5_ESM.pdf
Additional file 5: Metabolic map of A. oryzae in Scalable Vector Graphics. EC numbers are shown for all reactions in the map. A list of abbreviations for the metabolite-names is available in Additional file 4. (PDF 3 MB)
Authors’ original submitted files for images
Below are the links to the authors’ original submitted files for images.
Rights and permissions
Open Access This article is published under license to BioMed Central Ltd. This is an Open Access article is distributed under the terms of the Creative Commons Attribution License ( https://creativecommons.org/licenses/by/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
About this article
Cite this article
Vongsangnak, W., Olsen, P., Hansen, K. et al. Improved annotation through genome-scale metabolic modeling of Aspergillus oryzae. BMC Genomics 9, 245 (2008). https://doi.org/10.1186/1471-2164-9-245
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/1471-2164-9-245