Quantitative phosphoproteomics reveals novel roles of cAMP in plants

3′,5′‐cyclic adenosine monophosphate (cAMP) is finally recognized as an essential signaling molecule in plants where cAMP‐dependent processes include responses to hormones and environmental stimuli. To better understand the role of 3′,5′‐cAMP at the systems level, we have undertaken a phosphoproteomic analysis to elucidate the cAMP‐dependent response of tobacco BY‐2 cells. These cells overexpress a molecular “sponge” that buffers free intracellular cAMP level. The results show that, firstly, in vivo cAMP dampening profoundly affects the plant kinome and notably mitogen‐activated protein kinases, receptor‐like kinases, and calcium‐dependent protein kinases, thereby modulating the cellular responses at the systems level. Secondly, buffering cAMP levels also affects mRNA processing through the modulation of the phosphorylation status of several RNA‐binding proteins with roles in splicing, including many serine and arginine‐rich proteins. Thirdly, cAMP‐dependent phosphorylation targets appear to be conserved among plant species. Taken together, these findings are consistent with an ancient role of cAMP in mRNA processing and cellular programming and suggest that unperturbed cellular cAMP levels are essential for cellular homeostasis and signaling in plant cells.


INTRODUCTION
The cyclic adenosine monophosphate (3′,5′-cAMP) is an established signaling molecules, both in prokaryotes and eukaryotes. In plants, cAMP acts as second messenger in a number of processes including pollen tube growth [1], cell cycle regulation [2], auxin signaling [3], and stomatal closure [4]. Moreover, increasing evidence suggests that cAMP is also an essential component during the responses to abiotic and biotic stress [5][6][7]. Nevertheless, cAMP-dependent signaling in plants is still not well understood and molecular and cellular studies will be required to unravel the mechanisms of action as well as the systemic effects of cAMP.
cAMP signaling depends on the activation of adenylate cyclases (ACs) to rapidly increase cAMP levels [8] and phosphodiesterases (PDEs) to decrease these levels by converting cAMP it to AMP [5].
To-date only a few cAMP-dependent signal transduction mechanisms and pathways have been elucidated and they include direct activation of the cyclic nucleotide gated channels (CNGCs) [8,18,19]. It appears that cAMP-dependent signaling may be closely linked to cytosolic Ca 2+ as well as Na + and K + fluxes [20][21][22]. At the system level, rapid and reversible post-translational modification by phosphorylation is essential for plant development and adaptation to changing environmental conditions [23]. In mammals, increases in cAMP levels promote the phosphorylation of several intracellular enzymes via the activation of protein kinase A (PKA) [24].
While plant kinases that specifically respond to cAMP concentration changes remain elusive [25,26], cAMP-dependent changes of phosphorylation are likely given that cAMP, which promotes Ca 2+ influx, also initiates a protein kinase signaling cascade. This cascade, in turn, leads to changes in the protein phosphorylation status [27]. Furthermore, transcriptome analyses following AC stimulation also suggested that cAMP-dependent phosphorylation does occur in plants [26].
Here we propose the use of an established non-pharmacological approach, consisting of a genetically encoded tool based on the two cAMP-binding domains of the human PKA I regulatory subunit [7,28,29]. This cAMP-binding domain lowers the intracellular levels of cAMP in Nicotiana tabacum Bright Yellow-2 (BY-2) cells and thereby helps to uncover cAMP-dependent signaling events. This non-invasive modulation of cellular cAMP levels has previously revealed remarkable cAMP-dependent changes in the proteomic profile of tobacco BY-2 cells [6].
To follow on from these experiments, we now address the question of the role of 3′,5′-cAMP on the phosphoproteome with a view to infer mechanisms and systems-level responses that depend on this messenger.

Biological material
Wild type tobacco BY-2 (N. tabacum L. cv. Bright Yellow 2; TBY) cell suspensions were routinely propagated and cultured as described else- where [30]. The tobacco BY-2 line (termed cAS line) overexpressing a "cAMP sponge" based on the high-affinity cAMP-binding carboxyterminus of the regulatory subunit of a protein kinase A [28,29] was propagated in liquid selective medium containing 50 µg/mL kanamycin.
For the experiments, cAS cells were cultured in non-selective medium.
Two mL of both WT an cAS stationary phase cell suspensions (7 days) were diluted in 100 mL of fresh culture medium in 250-mL flasks and grown at 27 • C. After 5 days of culture, cells were collected by vacuum filtration on Whatman 3 MM paper, frozen in liquid nitrogen, and stored at −80 • C until the phosphoproteomic analyses. In total, four independent biological replicates were obtained and processed.

Phosphoproteomic workflow
Proteins were extracted following SDS/phenol method with minor adjustments [31].

Downstream bioinformatic analyses
In order to assess the quality of datasets, log 2 transformed and cen-  [36].
The analysis of significantly enriched phosphorylation motifs was performed by MOMO tool of MEME suite 5.1.1 (http://meme-suite. org/tools/momo) by using the Motif-X algorithm [37]. The peptide sequences (limited to 13 amino acids) were centered on aligned modification sites (phosphoserine or phosphothreonine). The number of occurrences was set to 20, and the probability threshold was set to p < 10 −6 . The dataset of unchanged peptides was uploaded as background data.
The scanning for occurrences of cAMP motifs was done by using The generation of sequence logos was done using the web-based application WebLogo (https://weblogo.berkeley.edu/).

Phosphosite conservation analysis
Phosphosite conservation analysis was conducted as detailed previously [44]. Essentially, a multiple sequence alignment (MSA) with each phosphoprotein and its paralogues was performed. For each protein, the best BLAST hit and its paralogues in reference organisms were selected by using PLAZA 5.0 dicots [45]. Conservation of the residues, as well as the window sequences around the residues (−6 or +6), was determined by remapping all residue positions within the N. tabacum protein. The percentage of conserved PPs was calculated for every species where the phosphorylated residue was present. The BLOSUM score was used to score the conservation of the sequence within the window around the residue.

PPI network construction and essential protein/hub analysis
The search tool for retrieval of interacting genes (STRING) database (https://string-db.org) was used to point to potential interactions between all phosphoregulated proteins in cAS versus WT comparisons [46]. Parameters were set as follows: co-expression as active interaction sources and medium confidence (>0.4). Disconnected nodes were hidden in the network. In order to visualize the protein-protein interaction (PPI) network the Cytoscape software version 3.6.1 was used [47]. The maximal clique centrality (MCC) algorithm of the CytoHubba plugin [48] was used to detect the top hub genes in co-expression networks. Proteins with the top 10 MCC values were considered hub genes/proteins.

Differentially abundant phosphosites (DAPPs) in response to cAMP depletion
In order to assess the dependence of cAMP on the phosphoproteome, we compared WT and cAS TBY cell lines grown for 3 days at 27 • C. Plant cell suspensions were chosen since they are a cytologically uniform and reproducible system that is suitable for an efficient induction of changes in the physical environment [49]. In the mutant line, the cAMP content was reduced by approximately 50% due to the sequestration by the sponge which is based on the a PKA I regulatory subunit that specifically binds free cAMP [6]. After 5 days of culture, cAS lines compared to WT showed an inhibition of cell growth equating to an about 35%-40% reduction in fresh weight [6].
By using a Ti-IMAC microsphere-based enrichment approach, we identified 2478 PPs on 2162 unique peptides (mapping to 1551 proteins) ( Table S1). The PCA showed that biological replicates plotted very closely in the PCA space, indicating a good correlation between them. It was furthermore noted that the two conditions tested resulted in distinct phosphoproteomic signatures dependent on the cellular levels of cAMP ( Figure S1).
The one-way ANOVA comparison test (FDR < 0.05) allowed the identification of 123 DAPPs between two conditions considered (Table   S2). Of those PPs sites, 80 showed increases while 43 decreased in their phosphorylation state. Over 80% of the total DAPPs were serine residues, approximately 15% were threonine residues, and less than 1% were tyrosine residues. Incidentally, this distribution is similar to the one previously reported in a large-scale in vivo phosphorylation site map of Arabidopsis cell suspensions [50]. The sequence windows of each of the detected PPs are shown in Table S2 (column AH) and the detected DAPPs were mapped to 115 phosphoproteins of which 105 (91.3%) showed phosphorylation change in one residue and 10 (8.7%) showed two changes.

cAMP-dependent phosphorylation is conserved among plant species
Cyclic AMP-sensitive PPs were compared between different species of higher and lower plants to determine the extent of conservation.
Overall, 20 (17%) phosphoproteins had phosphorylated orthologs in all eleven selected species. MSAs were performed with each phosphoprotein and its orthologues, and the conservation of the residues, as well as the flanking sequences around the residues (window sequences) were determined (Table S3). Perhaps not surprisingly, we noted the highest degree of conservation of PPs in the closely related Solanum lycopersicum (74% of total DAPPs). In the other dicots we found an average of 50 % of the DAPPs conserved and 42% in the early-diverged flowering plant Amborella trichopoda. In the monocot Oryza sativa the conservation was 37%, and in the moss Physcomitrella patens it was 31%. In the single-cell green alga Chlamydomonas reinhardti the conservation was markedly lower again (15% ; Table S3).
Similarly, the conservation of the of amino acids in the flanking sequences of the phosphorylated residues was also decreasing with increasing distance between the species and approximately 30 % of all flanking sequences showed higher conservation as compared to the whole protein (ratio > 1; Table S3 -column I), indicating that the areas flanking these PPs are likely functionally important. Overall, the presence of evolutionarily-conserved PPs is consistent with an evolutionary constraint on the cAMP-regulated PPs, and much like in animal species, this may be indicative for conserved functional roles.

In search of evidence for cAMP-dependent protein kinases
To date, the existence of cAMP-dependent kinases has remained elusive [25,26]. However, even if no bona fide cAMP-dependent protein kinases have been discovered in plants, a possible role for protein kinase cascades in cAMP-dependent signaling has been proposed [25] and this study lends further support to this notion.
In our dataset, we identified four DAPPs in four annotated kinases (A0A1S3YCC6, A0A1S3XTN9, A0A1S4AUV8, A0A1S4C3G8; Table   S2). No specific cyclic nucleotide-binding domain signatures were identified in these kinases that showed altered phosphorylation in cAS lines. This would exclude a direct interaction with a currently annotated cAMP-binding site. It is noteworthy that a serine/threonineprotein kinase, in which we observed a decreased phosphorylation level at S498 (A0A1S3XTN9) has an orthologue in the human proteome (Q96GX5; 59.5% of identity; E-value 2.2 e-47). Interestingly, this orthologue was been reported to be regulated by cAMP-dependent PDEs in human T cells [51] and this is further, albeit indirect evidence for cAMP-dependent plant kinases.
In order to discover further candidate phosphorylation targets of kinases, we searched for over-represented sequence motifs in the differentially phosphorylated phosphopeptides. The over-represented motifs, both in more and less phosphorylated peptides, were similar to a phosphorylation at Ser followed by Pro (. . . .SP. . . ; Figure 1).
Incidentally, this site is near identical to previously reported phosphorylation consensus motifs of MAP kinases (MAPKs), receptor-like kinases (RLKs), AFC2 kinases, AGC kinases (AGCKs), cyclin-dependent kinases (CDKs), SnRKs, and calcium-dependent kinases (CDPKs) [39,52]. Moreover, a PKA kinase motif (. . . R. . . S/T. . . ) occurred in seven dephosphorylated and three de novo phosphorylated proteins (Table   S4). Among the latter, we found a peptide that mapped on the KH domain-containing protein isoform X1 (A0A1S4D8B9). The human ortholog of this protein is a pre-mRNA-binding protein with a role in mRNA splicing (HNRPK; P61978) and alteration in its phosphorylation status was reported in human T cells [51].  (Table 1A). The serine/threonine kinase (AT2G34650; PID2), a member of the AGC family involved in ABA signaling and auxin transport [53], and the SNF1-related protein (AT3G01090), a SnRK1 family kinase with a role in the sugar signaling [54], are enzymes annotated as capable of phosphorylating multiple residues in the proteins of our dataset (Table 1A). In addition, 10 putative targets of kinases were also found in our dataset (Table 1B).
Among them are the auxin efflux carrier family protein (PIN2; AT5G57090, A0A0D4D8G6) and the trehalose phosphatase/synthase 5 (TPS5; AT4G17770, A0A1S4D237), both of which harbor multiple cAMP-dependent PPs. The SNF1-related protein was reported to alter phosphorylation of the TPS5 protein on a residue (S22; [55]) in A. thaliana and this site is close to the phosphorylation site altered by cAMP buffering in this study (S17).  [58], still, the binding to cAMP has yet to be experimentally proven [59].

Candidate cAMP binding-proteins
A cyclic nucleotide-binding domain was also found in the nuclearpore anchor-like protein (A0A1S3 × 5S0) and its human ortholog, the nucleoprotein TPR (P12270) showed a change in the phosphorylation status in T cells treated with selective inhibitors of cAMP-dependent PDEs [51]. Since PDEs degrade cAMP thereby regulating cellular cAMP levels [60], it is conceivable that the observed TPR phosphorylation  (Table S2). NPC components control the mRNA/protein nucleo-cytoplasmic trafficking [61]. Successful cell cycle completion requires the breakdown and reassembly of the nuclear envelope at the end of mitosis [62]. Many NPC proteins are phosphorylated during these processes [63,64]. Therefore, the alteration of the phosphorylation status reported here could explain the lower mitotic index previously recorded in cAS lines compared to WT lines under control condition [6].
Moreover, among kinases, CDKs have essential roles in exerting control of cell cycle progression and their activation requires the interaction with specific cyclin partners [65]. The cAMP-dependent phosphorylation of two cyclins (A0A1S3WZH8, A0A1S4A5E4) is further evidence for a role of CDK, as well as the critical role of cAMP in cell cycle regulation.

The roles of cAMP-dependent changes in phosphoproteomic signatures
To obtain functional insights into the biological processes affected by cellular cAMP buffering, a GO enrichment analysis was conducted.
Several processes associated with RNA processing and splicing are enriched in our dataset (Table 2). Overall, we identified 18 DAPPs assigned to proteins involved in RNA processing and splicing (Table 3).  representation of the spliceosome activation process including the phosphoregulated proteins, shows the extent of cAMP-dependence ( Figure 2A).
The RNA splicing is a pivotal step in gene expression which in turn modulates the messenger RNA population essential for cellular adap-tation to continuously changing conditions [66]. Many proteins have been reported to undergo phosphorylation and dephosphorylation during splicing events [67], and RNA splicing is affected by alterations in the phosphorylation status of components of the spliceosome, the molecular components that catalyzes the removal of introns from TA B L E 3 List of phosphoregulated proteins involved in RNA processing and splicing with their respective phosphosites, their conservation (%) and log fold change values.  nuclear pre-mRNA, as well as proteins with regulatory functions this process [68]. In mammalian cells, several stimuli that increase the intracellular cAMP level affect alternative splicing through phosphorylation of both SR proteins and heterogeneous nuclear ribonucleoproteins (hnRNPs) by PKA [24].

Phophosite
In this study, the phosphorylation status of several RNA-binding proteins (RBPs) is shown to be responsive to cAMP buffering (Table   S6; Table 3). Plant RBPs have a role in post-transcriptional processes essential for plant adaptation to environmental stimuli [69]. The activity and fate of RNA transcripts are critically dependent on RBPs [70] which enable and modulate RNA processing in the nucleus, export out of the nucleus, and metabolism in the cytoplasm.
The topological network of the RBP PPI ( Figure 2B) points to two serine/arginine-rich splicing factors, the RS40 (AT4G25500, A0A1S4A621) and the SC35-like (AT5G64200, A0A1S4DMI2) as top hub genes/proteins (Table S7). In the RS40 protein, the phosphorylated residue (S263) shows a high degree of conservation (66.67%) in the species investigated, and this also holds true for the S218 in the SC35-like protein (81.82%) (Table S3), again pointing to a conserved functional role of splice factor phosphorylation.
Among the RBPs, SR proteins are splicing factors rich in arginineserine dipeptide repeats (RS domains) involved in spliceosome formation and splice site recognition [71,72]. Changes in the phosphorylation status of RS domains can interfere with the ability of SR proteins to interact with RNA and other splicing-related proteins [73]. Despite this central function, mechanism by which SR protein phosphorylation is occurring and regulated in plant cell is not yet well understood.
In humans, several kinase families involved in the splicing-related phosphorylation have been reported and the growing list includes the SRPKs (SR-specific protein kinases) and CLKs (Cdc2-like kinases) which are the best characterized [74]. SRPKs phosphorylate the RS domains in SR proteins and in turn, RS domains contain prolines with flanking serines that are phosphorylated by CLKs. The proline-directed phosphorylation of these sites affects SR protein conformation and splicing [75]. CLKs are conserved kinases with members such as the human CLK1-4, the Saccharomyces cerevisiae KNS1, the Drosophila melanogaster DOA and A. thaliana AFC1-3 [76,77]. It is noteworthy that CLK homologs (AFC2s) were reported to participate in alternative splicing regulation under heat stress conditions [78].
We also investigated the amino acid sequence patterns flanking the phosphorylated residues in RNA splicing-associated proteins and found that the Ser-Pro motif was the most prevalent ( Figure S2) and again, this is consistent with a role of cAMP-dependent phosphorylation in the functionalization of AFC2 in the SR protein.
In support of a role of cAMP in the RNA splicing regulation is the finding that PPs appear conserved among 11 different plant species investigated (Table 3, Table S3, Figure S3). Overall, 11 out of 18 sites showed conservation of the central S/T of over 50%. The presence of several evolutionarily-conserved targets is also an indicator for the reliable quality of the data as well as supporting the functional assignment of the site [79,80].

Specificity of the cyclic nucleotide monophosphate (cNMP) response
In addition to cAMP, the plant cyclic nucleotide signaling system also includes 3′,5′-cyclic guanosine monophosphate (cGMP). Both cAMP and cGMP elicit different plant physiological processes, ranging from cell cycle progression to perception of external abiotic and biotic stimuli [8,16]. Both cNMP signals act through cellular effectors such as CNGCs and my affect Ca 2+ [81], Na + [82], and K + fluxes [83,84].
At the systems level, the case is less clear. Cyclic GMP-dependent protein phosphorylation has been demonstrated in A. thaliana cell suspension culture cells [42]. Exogenous administration of a membrane permeable cGMP analogue causes specific cGMP-dependent phosphorylation of spliceosome components and of proteins involved in cell-size regulation. Although different experimental approaches were adopted in the two studies, a comparison between them can provide some information on systemic roles of cNMPs and specificity of cAMP and cGMP. The comparative analysis reveals that, despite the limited number of common regulated phosphoproteins (9 proteins; Figure S4), RNA processing was the most enriched process in both studies. This finding therefore indicates that cAMP and cGMP signals may operate through phosphorylation of spliceosome components and further experiments will elucidate the nature of their complementary roles in the dynamic reorganizations of the spliceosome assembly system.
Among the common targets are two RNA-binding family proteins (AT3G56860, A0A1S4D8B9; AT5G15270, A0A1S4B2A6) that showed an increased phosphorylation both when the cAMP and cGMP levels decrease and increase, respectively. Therefore, we speculate that the two messengers may intervene in the splicing process regulating these proteins in an antagonistic way.

CONCLUDING REMARKS
Here we present a large-scale phosphoproteomic study to assess the role of 3′,5′-cAMP dampening in tobacco BY2 cells. Overall, the phosphoproteome is severely affected following cAMP buffer- Taken together, our results provide a repertoire of cAMPresponsive PPs in the proteome that will allow to infer the role of cAMP-dependent phosphorylation in plant responses at the molecular and systems level. The dataset will also serve as a useful baseline for the study of developmental and stimulus specific cAMP-dependent changes in protein phosphorylation plants.

ACKNOWLEDGMENTS
The authors have nothing to report.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
The mass spectrometry proteomics data were deposited in the Pro-teomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD040912.