Takeshi Izawa¹, Yoshiaki Nagamura²

¹Functional Plant Research Unit, ²Genome Resource Unit

［Abstract］

We gathered transcriptome data from the leaves of rice plants grown in a paddy field along with the corresponding meteorological data, and used them to develop statistical models for the endogenous and external influences on gene expression. Our results indicate that the transcriptome dynamics are predominantly governed by endogenous diurnal rhythms, ambient temperature, plant age, and solar radiation. The model also generated predictions for the influence of changing temperatures on transcriptome dynamics. We anticipate that our models will help translate the knowledge amassed in laboratories to problems in agriculture.

［Keywords］

field environments, rice, transcriptome, statistical modeling, prediction

［Background］

Since it is difficult to apply our knowledge of molecular biology to problems that occur in natural environments or agricultural fields, it is important to build on the knowledge collected under simple, controlled laboratory conditions to account for the more complex responses that occur outside of the laboratory, as currently our understanding of how organisms respond to natural conditions in the field remains limited. One of the major difficulties in analyzing data from field experiments is the complex and noisy character of various environmental factors; multiple factors that affect the transcriptome change simultaneously in the field. We focused on the transcriptome of rice (Oryza sativa) and employed a statistical modeling approach using a large collection of microarray and meteorological data to characterize the changes in transcriptome under natural field conditions. This approach enabled us to dissect how transcriptomes are controlled by environmental stimuli in the field, as well as by the plant’s entrained circadian clock and the plant’s age.

［Results and Discussion］

1. We collected mature leaf samples from rice plants growing in a paddy field in Tsukuba, Japan during a normal cultivation season from May to October in 2008 and used them for microarray analysis.
2. A total of 461 microarray data with distinct sampling time points and the corresponding meteorological data (wind speed, air temperature, precipitation, global solar radiation, relative humidity, and atmospheric pressure) measured at 1 min intervals were used to build the statistical model.
3. We independently modeled the transcriptional dynamics for all of the genes (27,201 genes). To define a gene set for further analysis of the modeling results, we used only the genes expressed in the majority of samples and for which residuals from the corresponding model were normally distributed among the 461 samples. As a result, a total of 17,193 genes met both of these criteria and were used in subsequent analyses. As an example, the result for modeling of Os01g0182600 is shown in Fig. 1.
4. Our results indicate that the transcriptome dynamics are predominantly governed by endogenous diurnal rhythms, ambient temperature, plant age, and solar radiation.
5. The data revealed diurnal gates for environmental stimuli to influence transcription and pointed to relative influences exerted by circadian and environmental factors on different metabolic genes.
6. The model also generated good predictions for transcriptome data obtained from another cultivation season (Fig. 2) and the influence of changing temperatures on transcriptome dynamics.

1. We anticipate that our models will help translate the knowledge amassed in laboratories to problems in agriculture.
2. Our approach in deciphering the transcriptome fluctuations in complex environments will be applicable to other organisms.

Fig.1. Expression of Os01g0182600. Expression in samples from June to September 2008. Samples were obtained at 2 hr intervals for a 48-hr period in nine data sets.

Fig.2. Scatterplot of observed vs. predicted transcriptome at a sample in 2009 in the field. A high correlation was obtained across all genes between the predicted transcriptome based on the 2008 statistical model with the meteorological data and transplanting data in 2009 and the observed transcriptome data from a sample obtained in 2009.

 

[Reference]

1. Nagano AJ, Sato Y, Mihara M, Antonio BA, Motoyama R, Itoh H, Nagamura Y, Izawa T (2012) Deciphering and prediction of transcriptome dynamics under fluctuating field conditions Cell 151(6):1358-1369