TransDetect Identifies a New Regulatory Module Controlling Phosphate Accumulation

Abstract

Identifying transcription factor (TFs) cooperation controlling target gene expression is still an arduous challenge. The accuracy of current methods at genome scale significantly drops with the increase in number of genes, which limits their applicability to more complex genomes, like animals and plants. Here, we developed an algorithm, TransDetect, able to predict TF combinations controlling the expression level of a given gene. TransDetect was used to identify novel TF modules regulating the expression of Arabidopsis (Arabidopsis thaliana) phosphate transporter PHO1;H3 comprising MYB15, MYB84, bHLH35, and ICE1. These TFs were confirmed to interact between themselves and with the PHO1;H3 promoter. Phenotypic and genetic analyses of TF mutants enable the organization of these four TFs and PHO1;H3 in a new gene regulatory network controlling phosphate accumulation in zinc-dependent manner. This demonstrates the potential of TransDetect to extract directionality in nondynamic transcriptomes and to provide a blueprint to identify gene regulatory network involved in a given biological process.

Figure 1.

Scheme of TransDetect algorithm organization. The algorithm is built on an inference iterative process. First the target gene transcript level is fitted by a linear combination of the transcript levels of two TFs. The resulting model is kept only if the two TFs significantly participate in the fit. The learnt coefficients are then used to predict the transcript levels of the target gene in an independent data set. If the model can properly fit the transcript levels of the target gene in the first data set and predict them in the second data set, the corresponding TF pair is kept. A final list of selected TF pairs is generated, and TFs are ranked based on the number of times they appear in this list.

Figure 2.

TransDetect network potentially influencing PHO1;H3 gene expression. Each node represents a potential PHO1;H3 regulator based on the TransDetect criteria defined in the text. If a pair of TFs is predicted to explain PHO1;H3 expression, it is linked by an edge. The edge width is proportional to the sum of the R² for the fit and predict processes (values ranging from 1.6 to 1.76). The most influential factors are likely to be the most connected.

Figure 3.

MYB15, MYB84, and bHLH35 regulate both the expression of PHO1;H3 and the accumulation of Pi in the shoot under Zn deficiency. A, Relative PHO1;H3 (At1g14040) transcript accumulation in the roots of wild-type plants (Col-0) and mutant lines harboring loss-of-function mutations in the following transcription factors: At5g57150 (bHLH35), At3g23250 (MYB15), At3g49690 (MYB84), At4g37790 (HAT22), At2g46510 (bHLH17), At1g31050 (bHLH111), At5g04760 (MYB-type), At4g31800 (WRKY18), At4g24060 (Dof46), and At3g50060 (MYB77) in −Zn compared to +Zn conditions. Plants were grown for 18 d in +Zn or −Zn. PHO1;H3 transcript abundance was measured by qRT-PCR normalized against UBQ10 (At4g05320). B, MYB15, MYB84, and bHLH35 transcript accumulation in response to Zn deficiency. Relative MYB15, MYB84, and bHLH35 transcript accumulation was quantified in roots of wild-type plants (Col-0) grown for 18 d in the presence or absence of Zn by qRT-PCR and normalized against UBQ10. C, Y1H assay. Sequences of the Arabidopsis PHO1;H3 promoter fused to the HIS3 auxotrophic marker were stably transformed into yeast. These different yeast strains were then cotransfected with MYB15, MYB84, or bHLH35. Left, Growth of the different yeast strains on control media deprived of Trp (−W), allowing the selection of yeast cells expressing the selected TFs. Right, Growth of the different yeast strains on selective media deprived of Trp and His (−W −H). D, Pi concentrations measured in the shoots of wild-type, myb15, myb84, or bhlh35 plants grown for 18 d in the presence or absence of Zn. A, C, and D, Central lines in the boxes show the medians; box limits indicate the 25th and 75th percentiles; whiskers extend 15 times the interquartile range from the 25th and 75th percentiles. Letters a, b, and c indicate significantly different values at P < 0.05 determined by one-way ANOVA and Tukey’s HSD.

Figure 4.

Interactions between MYB15 and MYB84 and between MYB15 and bHLH35 influence the expression of PHO1;H3 Pi accumulation under Zn deficiency. A and B, TransDetect prediction of correlation between the expression of the MYB15 / MYB84 and MYB15 / bHLH35 TF pairs and the PHO1;H3 transcript level with R² = 0.75 and R² = 0.84, respectively. C, Y2H assay bHLH35, MYB15, and MYB84 were fused with either the GAL4 DNA binding domain (BD) or the GAL4 activation domain (AD) into appropriate expression vectors, which were then transferred into yeast. The different yeast strains were plated on nonselective medium (NS) or on selective media deprived of His (−His), adenine (−Ade), or both simultaneously (−His−Ade). D, Relative PHO1;H3 transcript accumulation in roots of wild type (Col-0), bhlh35, myb15, myb84, myb15/myb84, and myb15/bhlh35 mutant plants grown for 18 d in −Zn compared to +Zn. PHO1;H3 transcript abundance was measured by qRT-PCR and normalized against UBQ10. E, Shoot Pi concentrations measured in wild-type (Col-0), bhlh35, myb15, myb84, myb15xmyb84, and myb15xbhlh35 mutant plants grown on either +Zn or –Zn for 18 d. D and E, Box central lines show the medians; box limits indicate the 25th and 75th percentiles; whiskers extend 15 times the interquartile range from the 25th and 75th percentiles. Letters a, b, and c indicate significantly different values at P < 0.05 determined by one-way ANOVA and Tukey’s HSD.

Figure 5.

The ICE1 / MYB15 transcription factor pair regulates both the expression of PHO1;H3 and the accumulation Pi under −Zn. A, TransDetect’s prediction of the correlation between the expression of the TF pair MYB15 and ICE1 and the PHO1;H3 expression (R² = 0.73). B, ICE1 transcript accumulation. Expression of ICE1 was quantified in wild-type (Col-0) seedlings grown for 18 d in +Zn or −Zn. ICE1 transcript abundance was measured by qRT-PCR normalized against UBQ10. C, PHO1;H3 transcript accumulation. Expression of PHO1;H3 gene was quantified in wild-type (Col-0), ice1, myb15, and myb15/ ice1 seedlings grown for 18 d in +Zn or –Zn. PHO1;H3 transcript abundance was measured by qRT-PCR and normalized against UBQ10. D, Pi accumulations. Pi concentrations were measured from shoots of wild-type (Col-0), ice1, myb15, and myb15 /ice1 seedlings grown for 18 d in +Zn or –Zn. B, C, and D, Box center lines show the medians; box limits indicate the 25th and 75th percentiles; whiskers extend 15 times the interquartile range from the 25th and 75th percentiles. Letters a, b, and c indicate significantly different values at P < 0.05 determined by one-way ANOVA and Tukey’s HSD.

Figure 6.

Idealized model to explain how TransDetect extracts directionality in static data. A, Two transcription factors, TFA and TFB, positively control the expression of a target gene TA following a sort of an AND logic-gate. B, TA expression is induced only when TFA and TFB expression are both up-regulated. C, Linear modeling of TA expression, considering the ideal case where two transcription factors (TFA and TFB) control a target gene TA following a AND logic-gate. In this case, γ coefficient of the linear model will be highly significant, because it is the combination of TFA and TFB expression that is necessary to fully explain TA expression. On the other hand, it is not possible to infer TFA by a linear combination of TA and TFB, nor to explain TFB by a linear combination of TA and TFA. Thus, the term of the equation γTFA * TFB intrinsically possess some directionality explanatory power in this case where both TFs interact in the control of TA.

Figure 7.

Schematic representation of the MYB15, MYB84, bHLH35, and ICE1 regulatory module controlling PHO1;H3 gene expression and Pi accumulation in shoots under Zn deficiency. Pi increases in the shoots of plants exposed to Zn deficiency. PHO1;H3 plays a negative regulatory role in this process. Red solid lines indicate connections between MYB15, MYB84, bHLH35, and ICE1. Negative and positive regulatory effects of these transcription factors on PHO1;H3 expression under Zn deficiency are indicated by flat-ended dashed lines and arrowheads, respectively. a indicates previous knowledge on ICEI and MYB15 physical interaction.