*2.10. Expression Profiles of Tiller Related Genes and Their Network*

Many genes are previously reported to impact upon tiller numbers in their mutants [9]. To reveal their transcriptional responses to N rate, we extracted their FPKM value from the transcriptomes, and subjected it to statistical comparison (Table 2). Totally of 65 genes were selected to screen for their transcription levels. Among the 36 decently expressed genes (FPKM > 0.1), 27 showed significant varietal difference (*p* ≤ 0.05), 18 showed to be responsive to N rate, and three showed significant interaction between the variety and N rate. Of these significantly influenced tiller genes, only six of the 27 genes in the varietal comparison and nine of 18 genes in the N rate comparisons displayed more than twofold change in at least one comparison. Apparently, the expression level of these genes was more determined by a variety than to be affected by the N rate. A protein and protein interaction network (PPI) constructed from known connections seems to illustrate their complexity more (Figure 10). However, the change did not need to surpass a threshold of twofold to produce a significant impact on the phenotypes such as the tiller number.

**Figure 10.** The graphical visualization of protein and protein interaction (PPI) network show the relationship of tiller genes. The gene expression in their fold change are represented by the up and down half of the circle, for NPB and YD6, respectively; MN/LN denotes fold change of MN treatment to LN, HN/LN for HN treatment to LN, represented by the left and right half of the circle, respectively; red and pink mean induction (up regulation), blue and pale blue for suppression, and white for not detected; line colors denote that their connections are generated from curated database (blue), experimentally determined (purple), gene neighborhood (green), textmining (yellow green), co-expression (brown), or protein homology (light purple).

Meanwhile, through co-expression network analysis, these tiller related genes belong to different modules, and *D10*/*OsCCD8*, *HTD2*/*D14*/*D88*/*qPPB3*, and *NRR*/*CRCT* genes fall in the green module, which negatively controls the occurrence of the tiller number (Table 2). Surprisingly, these tiller genes are not part of the candidate hub genes. This suggests that these tiller related genes are not at the nodes

of the gene expression network, which may locate downstream of certain pathways, and changes of hub genes will regulate their expression accordingly. In addition, we found that other tiller-related genes belong to different network regulation modules, showing the complexity and diversity of the gene regulation at the early stage of tillering. It seems that many genes have other functions that are not yet discovered during this growth stage. For example, *TAC, LAX* and *IPA1* fell in the turquoise module are negatively correlated with dry weight in the leaf during this period.


**Table 2.** Fold change of the expression level change of tiller genes, their significance, and present module.

Note: The expression changes are in the form of log2, and significances at their respective probability level. Red and blue color represent scale of up-and down-regulation, respectively.
