Importance of phosphoinositide-dependent signaling pathways in the control of gene expression in resting cells and in response to phytohormones

Abstract

“Phosphoinositide” refers to phosphorylated forms of phosphatidylinositol, including phosphatidylinositol-4-phosphate and phosphatidylinositol-4,5-bisphosphate. Both of these molecules could be in vivo substrates of plant phospholipase C. These phosphoinositides can also be biologically active “per se,” by directly binding to proteins and thus altering their location and/or activity. The use of pharmacological agents in Arabidopsis suspension cells allowed us to identify genes whose expression was positively or negatively controlled, in the basal state, by products of phosphoinositide-dependent phospholipase C. In this basal state, it seems that no genes exhibit a phosphoinositide-dependent expression “per se.” However, many genes whose expression is altered in the presence of phospholipase C inhibitors appeared to be responsive to salicylic acid. This allowed us to show that salicylic acid acts both by increasing the phosphoinositide pool and by inhibiting the phospholipase C. In response to salicylic acid it is possible to identify genes whose expression is controlled by products of PI-PLC, but also genes whose expression is controlled by phosphoinositides “per se.” Our data highlight the importance of phosphoinositide-dependent pathways in gene expression in resting cells and in response to phytohormones.

Abbreviations

DAG

diacylglycerol (DAG); PA, phosphatidic acid; PI, Phosphatidylinositol; PI-4-P, phosphatidylinositol-4-phosphate; PI4K, PI-4-kinase; PI-4, 5-P₂, phosphatidylinositol-4, 5-bisphosphate; PI-PLC, phosphoinositide dependent-phospholipases C; PLD, phospholipase D; SA, salicylic acid

Phosphatidylinositol (PI) can be phosphorylated to give phosphatidylinositol-4-phosphate (PI-4-P) by PI-4-kinase (PI4K). PI-4-P can be phosphorylated further by PI4P-5-kinase to produce phosphatidylinositol-4,5-bisphosphate (PI-4,5-P₂). PI-4,5-P₂ and PI-4-P are the substrates of phosphoinositide dependent-phospholipase C (PI-PLC), whose products are phosphorylated inositol and diacylglycerol (DAG). DAG can be phosphorylated to make phosphatidic acid (PA) by DAG-kinase (DGK) (Fig. 1). Note that PA can also be produced by phospholipase D (PLDs) acting on structural phosphoglycerolipids such as phosphatidylcholine or phosphatidylethanolamine.

Figure 1.

Working model of the action of PI-PLC substrates and products on basal gene expression in Arabidopsis cells in suspension. Both the PI-PLC substrates (phosphoinositide pool) and products can have an impact on gene expression. This impact is represented either by an arrow (positive action) or a line with a bar (negative action). Micromolar wortmannin concentrations will decrease the level of both phosphoinositide pool and PI-PLC products, while inhibitors of PI-PLC will alter the balance between PI-PLC substrates and products. The genes of clusters 1 and 2 are positively and negatively controlled in the basal state, respectively, by the phosphoinositide pool, while genes of clusters 3 and 4 are positively and negatively controlled, respectively, by the PI-PLC products. IP₃, inositol triphosphate; IP_n, inositol polyphosphate.

In order to investigate the roles of these enzymes in the control of the basal transcriptome, a pharmacological approach was used. PI4Ks can be divided into type II and type III that differ by their primary structure and their sensitivity to inhibitors. The PI4Ks that are upstream from PI-PLC are all type III¹ and they are inhibited by micromolar concentrations of wortmannin. PI-PLC activity can be inhibited by edelfosine² or U73122.³ DGK can be inhibited by R59022.⁴ PA production by PLD can be inhibited by n-butanol.⁵

Transcriptomic studies of Arabidopsis cells in suspension using CATMA microarray chips,⁶ allowed us to establish a list of genes that showed an altered expression for each inhibitor.^7-9 Comparing the lists showed an over representation of genes for which wortmannin and edelfosine (or wortmannin and U73122) had the same effect.⁹ The common effect of inhibiting PI4K and inhibiting PI-PLC is to decrease the total amount of products generated by PI-PLC (Fig. 1). The basal expression of these genes is thus under the control of the products of basal PI-PLC activity. It is likely that PA production by DGK is one element of the control, since R59022 can also have the same effect as wortmannin.⁹ However, we cannot rule out an action of other products of the PI-PLC pathway such as phosphorylated inositols.¹⁰ We could not detect an over representation of genes for which PI4K inhibition (reducing the amount of phosphoinositides), on the one hand, and PI-PLC inhibition (increasing the amount of these lipids), on the other hand, had opposite effects.⁹ Such opposite effects are expected for genes that have their basal expression under the control of phosphoinositides per se, but not as substrates of PI-PLC. Indeed phosphoinositides can be cofactors of PLDs, or bind to proteins that have phosphoinositide-binding domains.¹¹

When the list of genes whose expression is altered by inhibitors of PI-PLC was used for similarity searches focusing on stress response microarray data, the top experiments retrieved dealt with either heat or drought⁹ (Fig. 2A). Among these genes was DREB2A,⁹ a transcription factor that is a major actor in the response of plants to drought and heat.¹² It thus appears that DREB2A expression in Arabidopsis cells in suspension is constitutively maintained at a low level by an active basal PI-PLC in control conditions. Does this mean that heat and drought induce DREB2A expression by alleviating the basal repression by PI-PLC? Perhaps not, since an activation does not necessarily mean an alleviation of a repression. More experiments are required to clarify this point.

Figure 2.

Similarity between the edelfosine-responsive transcriptome and public transcriptome data. The 200 genes the most up-regulated by edelfosine and the 200 genes the most down regulated by edelfosine were used as a signature to search for experiments with similar transcriptome changes. A similarity score, derived from Pearson's correlation, was calculated by Genevestigator¹⁷ between the edelfosine signature and each experiment of a set. Then a relative similarity score was calculated where a relative similarity score of 1 stands for a similarity between the input signature and an experiment that is the same as the average over all experiments of a set. (A) The similarity search was performed against the 225 experiments under the “stress” classification. (B) The similarity search was performed against the 236 experiments under the “hormone” classification. (C) The similarity search was performed against the experiments under the “stress” and “hormone” classifications. The relative similarity scores between our signature input and a particular experiment will be different in (A), (B), and (C) because the overall sets of experiments are different. When possible Arabidopsis accessions and the tissues used for the microarray experiments are indicated. ABA, abscissic acid; DFPM, 5-(3,4-dichlorophenyl)furan-2-yl]-piperidine-1-ylmethanethione. DFPM stimulates an effector-triggered immune signaling pathway and thereby rapidly disrupts ABA signal transduction.¹⁸

Recently, the list of genes whose expression is altered by PI-PLC inhibitors was used in another similarity search targeting microarray experiments dealing with hormone response.¹³ The most similar experiments thus retrieved dealt with salicylic acid (SA; Fig. 2). The fact that mining transcriptomic data with an edelfosine signature returns so many experiments in which SA was the treatment (9 out of the top 10; Fig. 2B) could mean that edelfosine is stimulating SA biosynthesis, thus increasing its cellular content. However, ICS1 expression, the key enzyme of SA biosynthesis, is not affected by this agent, indicating that this is unlikely. On the contrary, the similar effect of SA and PI-PLC inhibitors would suggest that SA could act in part by inhibiting basal PI-PLC. Indeed, we were able to show that SA inhibited in vivo PI-PLC activity in Arabidopsis cells in suspension. Analyzing the SA responsive genes, we could show that some of them exhibited an expression that was similarly altered by inhibitors of PI4K, of PI-PLC and by SA. It is likely that the expression of these genes is actively maintained at a basal level in control conditions by basal PI-PLC products. SA, by inhibiting basal PI-PLC, will alleviate this control. If the PI-PLC products have a positive control, SA will lead to an inhibition of gene expression, while if the PI-PLC products have a negative control, SA will lead to an increase of gene expression. Here again, DGK might have a role in the basal control since there is an over representation of genes for which SA and R59022 have a similar effect. Interestingly, we could also detect genes for which the response to SA was not mimicked by wortmannin but inhibited by this agent. These genes are controlled by phosphoinositides per se. Indeed, we had previously shown that SA leads to activation of type III-PI4K.⁷ PI-PLC inhibition coupled to PI4K activation leads to an increase of phosphoinositides that triggers signaling events leading to the specific expression of a cluster of SA responsive genes. Therefore the action of SA on PI-PLC is involved in the control of SA responsive genes by 2 mechanisms: alleviating the effect of PI-PLC products (or their phosphorylated forms) and increasing the cell content in phosphoinositides. It seems that alleviation of PI-PLC products concerns mostly SA repressed genes, that are positively controlled by basal PI-PLC.¹³

Therefore these 2 studies^9,13 clearly show the importance of the control of basal gene expression, and that the signaling pathways depending on phosphoinositides, either as substrates of PI-PLC or per se, are very much involved in this control. The genes whose basal expression is controlled by PI-PLC are mostly involved in the response to SA. For instance, when the list of genes whose expression is altered by edelfosine was used against microarray experiments dealing with “stress” or “hormone,” only one experiment dealing with heat stress appeared in the top 10 retrieved experiments, while 8 of them dealt with SA (Fig. 2C).

Following these data and conclusions, new questions need to be answered.

1/ Why is DREB2A not induced by SA? More generally, why are genes induced by inhibition of PI-PLC not necessarily induced by SA since SA inhibits PI-PLC?

Clearly SA does not only inhibit PI-PLC. Considering solely phosphoglycerolipid signaling, SA activates type III-PI4K⁷ and PLD.⁸ `Other signaling events are triggered by SA, such as changes in redox status, protein phosphorylation (reviewed by Janda and Ruelland).¹⁴ These signaling events other than PI-PLC inhibition might interfere with, or even counteract, the effect of PI-PLC inhibition. This might explain why SA does not induce DREB2A.

2/ Why is phosphoinositide level per se not important for basal expression but it is for SA response?

In Arabidopsis cells in suspension, this level results from the equilibrium between basal PI4K (source) and PI-PLC (sink) activities. Inhibition of PI-PLC should lead to an increase in phosphoinositides. But this change might not be sufficient to attain the threshold necessary for triggering other events leading to gene induction. On the contrary, the fact that for the response to SA, the phosphoinositide level per se can have a role is certainly due to the fact that SA not only inhibits PI-PLC, but also activates PI4K. These coupled actions would lead to an important phosphoinositide increase beyond the threshold necessary for activating downstream signaling events.

3/ Do genes controlled by PI-PLC products really exist?

It is always important to consider one's data with a critical eye. Genes whose basal expression is controlled by PI-PLC products are only identified because type III-PI4K and PI-PLC inhibitors have the same effects.^9,13 Similarly, the clusters of SA-responsive genes dependent on the products of PI-PLC are identified because SA and inhibitors of PI4K and of PI-PLC have the same effects¹³ These can be only similar, but independent, effects, and not due to the action of these agents on the corresponding enzymes, but due to side effects. It cannot be ruled out, even though unlikely, that the side effects -i.e. the effects not related to their action on enzymes of the PI-PLC pathway- of edelfosine, U73122, wortmannin and R59022 led to the same gene responses. We would like to be able to inhibit the PI-PLC inhibition during SA treatment, but this is probably impossible. However, the cluster of genes responsive to SA and dependent on PI-PLC products as defined above displays unique characteristics (such as enrichment in particular cis-elements in their promoters),¹³ strongly suggesting that there is a biological reality behind them.

In conclusion, our data allowed us to precise the action of SA on gene expression: SA acts by activating PI4Ks but also by inhibiting PI-PLC. However, our data also highlight the importance of phosphoinositide-dependent pathways in the control of basal gene expression in resting cells. Clearly the maintenance of a basal transcriptome in resting cells is an active process involving signaling pathways.¹⁵ Our analysis by inhibitors allowed us to identify the different roles of PI-PLC substrates and products. Yet, among the phosphoinositides, which one is important for the control of basal gene expression? Is it PI-4,5-P₂ or PI-4-P, or both? Similarly, among the PI-PLC products which ones are important for the control of gene expression? It could be DAG, PA and/or phosphorylated inositols. Indeed the over expression of type 1 inositol 5-phosphatase was shown to impact gene expression.¹⁰ The inhibition of DGK also impacts gene expression.^9,13 More work is required to fully understand the role of phosphoinositide-dependent pathways in the control of gene expression in resting cells.

Finally it can be discussed that our analysis is based on the use of pharmacological agents. Pharmacological inhibitors can have side effects. However, it is more likely that the common effect of wortmannin, edelfosine, U73122 and R59022 on gene expression is due to their common site of actions (i.e., on enzymes of the PI-PLC pathway) than to side effects that are a priori not related. Another approach would be to use mutants but the enzymes of the phosphoinositide pathways are encoded by multi gene family, which could make this approach tedious. Besides, mutants can also have side effects. We have recently shown that a PI4K double mutant constitutively accumulates high levels of SA.¹⁶

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

This research was supported by Center National de la Recherche Scientifique (ER) and Université Paris-Est Créteil (ER). TK has been awarded a cotutelle doctoral grant from the French Government.