Exploring temporal transcription regulation structure of Aspergillus fumigatus in heat shock by state space model

Abstract

Background: The thermotolerance of Aspergillus fumigatus plays a critical role in mammalian and avian infections. Thus, the identification of its adaptation mechanism to higher temperature is very important for an efficient anti-fungal drug development as well as fundamental understanding of its pathogenesis. We explored the temporal transcription regulation structure of this pathogenic fungus under heat shock conditions using the time series microarray data reported by Nierman et al. (Nature 2005, 438:1151-1156).

Results: The estimated transcription regulation structure of A. fumigatus shows that the heat shock proteins are strongly negatively associated with central metabolic pathway genes such as the tricarboxylic acid cycle (TCA cycle) and carbohydrate metabolism. It was 60 min and 120 min, respectively, after the growth temperature changes from 30 degrees C (corresponding to environments of tropical soil) to 37 degrees C and 48 degrees C (corresponding to temperatures in the human body and compost, respectively) that some of genes in TCA cycle were started to be upregulated. In these points, most of heat shock proteins showed lowest expression level after heat shocks. Among the heat shock proteins, the HSP30 (AFU6G06470), a single integral plasma membrane heat shock protein, presented most active role in transcription regulation structure in both heat shock conditions of 37 degrees C and 48 degrees C. The metabolic genes associated with multiple genes in the gene regulation network showed a tendency to have opposite expression patterns of heat shock proteins. The role of those metabolic genes was second regulator in the coherent feed-forward loop type of regulation structure having heat shock protein as its first regulator. This type of regulation structure might be very advantageous for the thermal adaptation of A. fumigatus under heat shock because a small amount of heat shock proteins can rapidly magnify their regulation effect on target genes. However, the coherent feed-forward loop type of regulation of heat shock proteins with metabolic genes became less frequent with increasing temperature. This might be the reason for dramatic increase in the expression of heat shock proteins and the number of heat shock response genes at heat shock of 48 degrees C.

Conclusion: We systemically analysed the thermal adaption mechanism of A. fumigatus by state space model with times series microarray data in terms of transcription regulation structure. We suggest for the first time that heat shock proteins might efficiently regulate metabolic genes using the coherent feed-forward loop type of regulation structure. This type of regulation structure would also be efficient for adjustment to the other stresses requiring rapid change of metabolic mode as well as thermal adaptation.

Figure 1

The metabolic distribution of common genes (596 genes) between heat shock response genes of 37°C and 48°C.

Figure 2

The average gene expression profiles of heat shock proteins. (A) expression profiles of heat shock proteins at 37°C, (B) expression profiles of heat shock proteins at 48°C.

Figure 3

The heat map of estimated temporal relationship between top-ranked 20 genes in each sub-module and temporal relationship between modules. The upper part represents the heat map of temporal relationship from previous time to current time between top ranked 20 genes in each sub-module. The gene pairs associated with strong negative regulation are presented as deep green while the white regions represent gene pairs with strong positive regulation. The lower part shows the temporal relationship between modules and the number on arrow represents regulation coefficient estimated by system model (equation 3 in Methods section). (A) heat shock of 37°C, (B) heat shock of 48°C.

Figure 4

The dependence of edge density on threshold in networks of 37°C and 48°C.

Figure 5

The visualization of estimated transcriptional network at the threshold of 0.015. The node in the network stands for gene. The yellow and red nodes represent heat shock proteins and transcription factors, respectively. The hub node is shown as grey and green node, respectively, at network of 37°C and 48°C. The red and blue arrows represent positive and negative regulation, respectively. The direction of arrow indicates the regulation direction, i.e., from regulator gene to regulated or target gene. The gene information corresponding to yellow node (heat shock protein): 293 (AFU1G11180), 1059 (AFU3G14540), 1465 (AFU5G04170), 1807 (AFU6G06460), 1808 (AFU6G06470). The gene information corresponding to red node (transcription factor): 723 (AFU3G02000), 909 (AFU3G09670), 1060 (AFU3G14550), 1208 (AFU4G09710), 1957 (AFU6G12160). The gene information corresponding grey and green nodes (hub nodes) : 1062 (AFU3G14590), 1172 (AFU4G08340), 1553 (AFU5G08750), 1554 (AFU5G08800), 1806 (AFU6G06430), 1289 (AFU4G12010), 1931 (AFU6G10610), 1933 (AFU6G10650), 1934 (AFU6G10660), 2038 (AFU7G00170), 2036 (AFU7G00120), 2039 (AFU7G00200), 2050 (AFU7G01000), 2077 (AFU7G01920), 2078 (AFU7G01930), 2311 (AFU8G06340), 2312 (AFU8G06350). (A) network of 37°C, (B) network of 48°C. Gene information corresponding to each node in the network is shown at additional file 3.

Figure 6

The average expression profiles of hub nodes negatively regulated by heat shock proteins. (A) expression profiles of hub nodes at network of 37°C, (B) expression profiles of hub nodes at network of 48°C.

Figure 7

The extracted regulation network from Figure 5 consisting of heat shock proteins, transcription factors and hub metabolic genes. The gene corresponding to each node index is the same as Figure 5. The red and yellow nodes represent transcription factor and heat shock protein, respectively. (A) Regulation structure at heat shock of 37°C, (B) Regulation structure at heat shock of 48°C.

Figure 8

Types of three-node coherent feed-forward loop. Arrow and ⊣ symbol denote positive and negative regulation, respectively.

Figure 9

The schematic diagram for pre-processing of microarray data.

Figure 10

The BIC curves for the model-based clustering of microarray data including heat response genes. (A) BIC curve for model-based clustering of gene expression data at 37°C, (B) BIC curve for model-based clustering of gene expression data at 48°C. The numerical values of BIC in each model are shown at additional file 4. The model is divided according to the covariance matrix equation form. EII: spherical, equal volume; VII: spherical, unequal volume; EEI: diagonal, equal volume and shape; VEI: diagonal, varying volume, equal shape; EVI: diagonal, equal volume, varying shape; VVI: diagonal, varying volume and shape; EEE: ellipsoidal, equal volume, shape, and orientation; EEV: ellipsoidal, equal volume and equal shape; VEV: ellipsoidal, equal shape; VVV: ellipsoidal, varying volume, shape, and orientation.