The rolB plant oncogene affects multiple signaling protein modules related to hormone signaling and plant defense

The rolB plant oncogene of Agrobacterium rhizogenes perturbs many biochemical processes in transformed plant cells, thereby causing their neoplastic reprogramming. The oncogene renders the cells more tolerant to environmental stresses and herbicides and inhibits ROS elevation and programmed cell death. In the present work, we performed a proteomic analysis of Arabidopsis thaliana rolB-expressing callus line AtB-2, which represents a line with moderate expression of the oncogene. Our results show that under these conditions rolB greatly perturbs the expression of some chaperone-type proteins such as heat-shock proteins and cyclophilins. Heat-shock proteins of the DnaK subfamily were overexpressed in rolB-transformed calli, whereas the abundance of cyclophilins, members of the closely related single-domain cyclophilin family was decreased. Real-time PCR analysis of corresponding genes confirmed the reliability of proteomics data because gene expression correlated well with the expression of proteins. Bioinformatics analysis indicates that rolB can potentially affect several levels of signaling protein modules, including effector-triggered immunity (via the RPM1-RPS2 signaling module), the miRNA processing machinery, auxin and cytokinin signaling, the calcium signaling system and secondary metabolism.


2-D Gel Electrophoresis and quantification of protein expression.
Proteins were isolated from 0.5 g fresh weight of calli using a phenol extraction methanol/ammonium acetate precipitation method 29 . The phenolic phase was collected and precipitated overnight in five volumes of 100 mM ammonium acetate in ethanol at −20 °C. After centrifugation (10 min, 6000 g, 4 °C), the pellet was washed twice with ice-cold acetone.
For isoelectric focusing, dried protein pellets were dissolved in IPG buffer (9.5 M urea, 4% w/v CHAPS, 2% Pharmalyte pH 3-10 (GE Healthcare, Uppsala, Sweden), DeStreak Reagent (GE Healthcare) and 0.01% w/v bromophenol blue). Protein concentration was determined using an RC/DC kit (Bio-Rad Laboratories Inc., Hercules, CA, USA). A total of 500 µg of whole protein sample in 350 ml IPG buffer was applied to 18-cm Immobiline DryStrip pH 3-10 NL (GE Healthcare) by passive rehydration for 12 h at 20 °C according to the manufacturer's recommendations. IEF was performed in a Protean IEF Cell (Bio-Rad) for 60,000 V-h. Before separation in the second dimension, the Immobiline DryStrip was equilibrated in buffer (6 M urea, 0.375 M Tris-HCl, pH 8.8, 2% SDS, 20% glycerol, and 2% DTT) for 10 min. For SDS-PAGE, 12% polyacrylamide gels with 4% stacking gels were run in a Protean II xi cell (Bio-Rad). The gels were stained with Coomassie Brilliant Blue G-250. Three control and three experimental gels were used in the analysis.
Protein Expression. Gels were scanned using the PharosFX Plus System (Bio-Rad). PDQuest 8.0.1 Advanced software (Bio-Rad) was used for image and analysis of protein maps. The Spot Detection Wizard was used to select the parameters for spot detection, such as a faint spot and a large spot cluster. The results of automated spot detection were checked and manually corrected. On average, 1,500 protein spots were detected on gels of Arabidopsis calli. A local regression model (Loess) was used for normalization of spot intensity. The protein expression was accessed using PDQuest 8.0.1 Advanced software and was presented as mean total intensity of a defined spot in a replicate gel group. Spot quantity is the sum of the intensities of pixels inside the boundary. Fold of protein expression change was accessed based on mean protein intensity. For quantitative differentiation, a 1.5-fold change or higher in the average spot intensity was regarded as significant. Statistical significance of differences was assessed using Student's t test at a significance level of 0.05 in three replicates. All identified proteins in qualitatively different spots were considered. Mean expression values and standard deviations were calculated from three biological experiments.
Mass spectrometry. The total number of samples analyzed by MALDI was 203. The number of technical replicates was 2-3 (up to 8 for important proteins). Individual protein spots selected on the basis of image-analysis output were excised and digested in-gel with trypsin (Trypsin V511, Promega, Madison, WI, USA) as previously described 30 . For MALDI-TOF identification, 0.5-1 μl of the sample (50% solution of acetonitrile in water, 0.1% TFA) was placed on a ground steel MALDI target plate or AnchorChip or SmallAnchor (depending on the Scientific RepoRTs | (2018) 8:2285 | DOI: 10.1038/s41598-018-20694-6 protein quantity; also see Supplementary Dataset S1), and 0.5-1 μl of the matrix (α-cyano-4-hydroxycinnamic acid) was added. For LC-ESI-MS/MS, 10-μl protein samples dissolved in water containing 0.1% TFA were used.

MALDI-TOF Mass Spectrometry and Protein Identification.
All mass spectra were acquired with an Autoflex (Bruker Daltonics, Bremen, Germany) MALDI-TOF mass spectrometer with a nitrogen laser operated in the positive reflector mode (standard method RP 700-3500 Da.par) under the control of FlexControl software (version 3.4; Bruker Daltonics). The analysis was performed in the automatic mode (AutoXecute -automatic Run). The spectra were externally calibrated using the CalibratePeptideStandards.FAMSMethod and a standard calibration mixture (Protein Calibration Standard I, Bruker Daltonics). The data files were transferred to Flexanalysis software version 3.4 (Bruker Daltonics) for automated peak extraction. Assignment of the first monoisotopic signals in the spectra was performed automatically using the signal detection algorithm SNAP (Bruker Daltonics). For MS and MS/MS analyses, we used the PMF.FAMSMethod and SNAP_full_process.FALIFTMethod, respectively. Each spectrum was obtained by averaging 1500-5000 laser shots (300 shots in a step) acquired at the minimum laser power. The data were analyzed using BioTools (version 3.2; Bruker Daltonics). A peptide mass tolerance of 0.5 Da and a fragment mass tolerance of 0.5 Da were adopted for database searches. The m/z spectra were searched against the Arabidopsis thaliana NCBInr and SwissProt databases using the Mascot search engine. Threshold score was 40. Further data were analyzed using UniProt (http://www.uniprot.org/uniprot/) and other specialized databases and programs as indicated below. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 31 partner repository with the dataset identifier PXD005889 (DOI: 10.6019/PXD005889).

LC-ESI-MS/MS.
For determination of proteins of low abundance, we used (in additional to MALDI analysis) an HCTultra PTM Discovery System (Bruker Daltonik GmbH, Germany) equipped with a Proxeon EASY-nLC ultra-performance liquid chromatograph and a nanoFlow ESI sprayer. The coupling of Proxeon EASY-nLC to the Bruker HCT ion trap was performed using the program HyStar v3.2 (Bruker Daltonik GmbH). The HCTultra is equipped with a high-capacity ion trap that enables the acquisition of MS/MS data on low-abundance precursor ions. For the LC studies, Buffer A (0.1% formic acid in water) and Buffer B (0.1% formic acid in acetonitrile/10% water) (Acetonitrile G Chromasolv for HPLC, super gradient grade; Sigma-Aldrich, Steinheim, Germany) were used. Separation was carried out on a C18-reversed phase EASY-Column (10 cm × 75 µm i.d., 3-µm beads, 120-Å pore size, Thermo Fisher Scientific). The flow rate was 300 nl min −1 with the following gradient: 5% Buffer B at 0 min, linearly increased to 35% B at 10 min and to 100% B from 10 to 25 min followed by washing at 100% B from 25 to 40 min. The ion trap capillary temperature was set to 300 °C, and the dry gas flow was 5 l min −1 . The ion trap was set to acquire in positive ion mode, scanning in the manufacturer-specified standard enhanced mode (8,100 m/z/s) between m/z 300 and 2,000 for MS, averaging five spectra, and accumulated 200,000 charges (by ion charge control). Collision-induced dissociation fragmentation was performed on the four most intense ions with the threshold for precursor ion selection at an absolute intensity of 20,000. The strict active exclusion was used; a precursor ion was excluded after one spectrum and released after 0.1 min. MS-MS spectra were scanned from m/z 300-2,000, averaging three spectra. Data were analyzed using BioTools (version 3.2; Bruker Daltonics). The following parameters were used for database searches: peptide mass tolerance 0.1% and fragment mass tolerance 0.5 Da.
RNA isolation, cDNA synthesis and real-time PCR. The isolation of total RNA and first-strand cDNA synthesis were carried out as described previously 9 . RNA samples were isolated from callus cultures during the linear phase of growth (20-22 days). RNA concentration and 28S/18S ratios were determined using an RNA StdSens LabChip® kit and Experion TM Automated Electrophoresis Station (Bio-Rad Laboratories Inc., Hercules, CA, USA) with Experion TM Software System Operation and Data Analysis Tools (version 3.0) following the manufacturer's protocol and recommendations. The samples with 28S/18S ribosomal RNA between 1.5-2.0 and an RNA Quality Indicator (RQI) above 9.0 were used for real-time PCR analysis. Quantitative real-time PCR (qPCR) analysis was performed using a CFX96 (Bio-Rad Laboratories, Inc., Hercules, CA, USA) with 2.5 × SYBR green PCR master mix containing ROX as a passive reference dye (Syntol, Russia) as described 9 . Two biological replicates, resulting from two different RNA extractions, were used for analysis, and three technical replicates were analysed for each biological replicate. The gene-specific primer pairs used in the qPCR were as follows in the Supplementary Table S1. A. thaliana actin (AtAct2) and ubiquitin (AtUBQ10) genes (GenBank ac. no. NM_112764 and NM_001084884, respectively) were used as endogenous controls to normalize variance in the quality and the amount of cDNA used in each real-time RT-PCR experiment 9 . The highest expressing sample assigned the value 1 in the relative mRNA calculation using the formula 2 −ΔΔCT . Data were analyzed using CFX Manager Software (Version 1.5; Bio-Rad Laboratories, Inc.). For comparison among multiple data, analysis of variance (ANOVA) followed by a multiple comparison procedure was employed. Fisher's protected least significant difference (PLSD) post-hoc test was employed for the inter-group comparison. A difference of P < 0.05 was considered significant.
Protein Network Visualization. The network was built using the program Cytoscape as previously described 32 . The data loaded into the program were obtained from the PAIR [PAIR-V3.3 33 , http://www.cls.zju. edu.cn/pair/]. The protein-protein interactions presented in PAIR were compared with the databases BioGRID 34 (http://thebiogrid.org/). The size of each circle is correlated with the "betweenness centrality" metric, which describes the global position ("centrality") of the protein in the interactome. Betweenness centrality was calculated by Cytoscape. Information about protein-protein interactions was also obtained by using UniProt and by linking Cytoscape with external databases (IntAct and STRING). The network was validated using recently developed algorithms 35

Results
Characterization of the rolB-expressing line AtB-2. Expression of rolB in AtB-2 callus line was tested by qPCR before proteomic analysis. The rolB-expressing callus line AtB-2 was shown to be a line with moderate rolB expression (0.56 ± 0.04 relative expression level units, see Supplementary Table 2). The presence of the RolB protein in callus extracts was confirmed by mass spectrometry (Supplementary Dataset S3).
Proteomics Analysis. Total protein fractions were isolated from Arabidopsis thaliana vector control and rolB-transgenic callus cultures as described in Materials and Methods. Overall, 1,500 proteins were resolved on 2-D gels (Supplementary Figure S1). Of these, over 200 were identified using MALDI MS. Proteins whose determinations represented reliable data meeting the requirements of precision mass-spectrometric analysis and quantitative differences for proteins are included in Tables 1 and 2 and were considered in further analysis (see also Supplementary Dataset S1 and Dataset S2). Three differentially expressed proteins remain undetermined because the search of databases yielded no results. Thus, 31 proteins were upregulated in rolB-expressing cells compared with control cells (Table 1), and 29 proteins were down-regulated ( Table 2). We also performed a targeted search of chaperone-type proteins using their predicted masses and isoelectric points. These data are presented in Table 3. To identify low-abundance proteins such as VH1-interacting kinase and heat stress tolerant DWD1 (DWD1/HTD1), we used an Anchor chip or SmallAnchor chip (otherwise, it was not possible to identify these proteins). Examples are shown in Supplementary Dataset S1. RolB itself was also detected (Supplementary Dataset S3).

Proteins Upregulated in rolB-expressing Cells. Primary metabolism and ROS-detoxifying enzymes.
Several proteins involved in various biosynthetic processes of primary metabolism were strongly activated; these included alanine aminotransferase, carbamoyl phosphate synthase, malate dehydrogenase, threonine synthase, pyruvate dehydrogenase and others (Table 1). Another subset of upregulated proteins was represented by defensive enzymes involved in ROS metabolism. Among them were peroxidases, the activation of which in rolB-expressing cells was previously demonstrated at the level of gene expression 37 . The increase in expression of antioxidant enzymes determined in the present work was essentially the same as determined previously by other methods 38 , thus confirming the reliability of the proteomics experiments. The previously found induction of ascorbate peroxidase genes in rolB-transformed cells 9 was also confirmed ( Table 1). New data were obtained regarding glutathione S-transferases. Glutathione S-transferases F6 and F7, as well as glutathione S-transferase DHAR1, a key component of the ascorbate recycling system 39 , were upregulated in rolB-transformed cells. These transferases are involved in redox homeostasis and especially in the scavenging of ROS under oxidative stress conditions subsequent to induction by biotic or abiotic inducers 39 . Taken together, our data confirm the hypothesis 9 that rolB affects ROS metabolism by participating in a cellular process that resembles the process of stress acclimation.
Heat-shock proteins and chaperonins. Heat-shock 70-kDa proteins 6 and 7 (Hsp70-6 and Hsp70-7), Hsp90-5, 20-kDa chaperonin (Cpn10) and chaperonin 60 subunit α1 were activated in rolB-expressing Arabidopsis cells (Table 1). It is known that in cooperation with other chaperones, Hsp70s stabilize preexisting proteins against aggregation and mediate the folding of newly translated polypeptides in the cytosol as well as within organelles 40 . Transgenic Arabidopsis plants expressing a fungal hsp70 gene exhibited enhanced tolerance to heat stress and to osmotic, salt and oxidative stresses 40 .
Other proteins. The abundance of the VH1-interacting kinase (VIK, VH1-interacting tetratricopeptide repeat (TPR)-containing protein) in rolB-transformed cells was also increased. Another protein upregulated in the transformed cells was DWD1/HTD1. This protein was shown to participate in heat stress responses, possibly by interacting with Hsp90-1 42 . Enzymes that participate in secondary metabolism, such as chalcone-flavonone isomerase 1 and ATP sulfurylase 4, were upregulated.
Proteins Down-Regulated in rolB-expressing Cells. Expression of the 40S ribosomal proteins S7-2 and S7-1 was significantly inhibited in rolB-expressing cells ( Table 2). These proteins are structural constituents of the ribosome and participate in ribosomal RNA processing, ribosomal small subunit biogenesis and translation (BioGrid). Initiation factor 3 g was also down-regulated. This factor is involved in protein synthesis; together with other initiation factors, it stimulates binding of mRNA and methionyl-tRNAi to the 40S ribosome. These data indicate that rolB can potentially inhibit protein biosynthesis.

Analysis of Gene Expression.
To confirm the results of the proteomic analysis, qPCR was performed to detect expression of genes corresponding to six up-regulated proteins, six down-regulated proteins and five proteins which abundance was not changed in rolB-expressing Arabidopsis calli (Fig. 1). The activation of VIK, Hsp70-6, Hsp70-7, Hsp90-5, Cpn60 and Cpn10 genes in AtB callus culture was consistent with the proteomic data ( Table 1, Fig. 1A). Expression of RACK1A, ROC2, ROC4, ROC1, ROC6 and CYP20-2 was decreased in rolB-expressing cells (Fig. 1B). In agreement with the proteomic data, no significant differences were observed in expression levels of the Hsp70-10, Hsp70-14, Hsp70-15, Hsp90-2 and TCP-1 genes in At and AtB calli (Fig. 1C). Thus, the gene-expression data were in accordance with proteomics data.

Network Reconstruction and Analysis.
To create a subnetwork of signaling components affected by RolB, we used our previous reconstruction of the Arabidopsis interactome 32 , as well as algorithms for construction of subgraphs and validation of small subnetworks 35,36 . Our analysis indicated high level of integrity of the subnetwork presented in Fig. 2. Deleting the individual nodes indicated by octagons (affected by RolB) eliminated the subnetwork. Removing nodes that are not directly related to octagons does not destroy the network (in this case, the network turns out to be simpler). As can be seen from Fig. 2, the network shows the perturbations of the proteome but does not show input nodes. Although it is impossible at present to determine the primary targets of the oncoprotein, the reconstruction is useful for creating a working model.
Cyclophilins. ROC1 (CYP18-3): As a first step in the reconstruction of signaling components affected by RolB, we began to reconcile ROC1 interactions (Fig. 2). The abundance of ROC1 in rolB-transformed cells is significantly decreased ( Table 2). Via RIN4, ROC1 is connected to the RPM1-RPS2 signaling module 46 that controls effector-triggered immunity. Another consequence of ROC1 deficiency might be perturbations in the expression of ROC1-associated proteins such as RAN1 and RAN2 (RAN GTPase-activating proteins, Fig. 2) as well as the 14-3-3 proteins GRF1 and GRF8 (general regulatory factors). RANs mediate protein import into nuclei and the cellular response to salt stress. The interaction of ROC1 with GRFs was demonstrated previously 47 , but neither the exact mechanism of this interaction nor its outcome is known. ROC2 (CYP19-3): ROC2 physically interacts with calmodulins (CAMs) and thus affects a broad array of reactions controlled by CAMs 48 . These include the response to stress (mediated by the CBL-interacting serine/ threonine-protein kinase 6 (SIP3), BTB and TAZ domain protein 4 (BT4) and basic leucine-zipper proteins; Fig. 2) and induced systemic resistance (mediated by TGA6 and NPR1). It is likely that in this manner, i.e., via ROC2-CAM interactions, rolB also exerts its modulating effect on calcium-dependent protein kinases 49 .
ROC4 (CYP20-3): ROC4 connects redox and light signals to cysteine biosynthesis and stress responses in chloroplasts 50 and is known to be a key effector protein that links hormone signaling to amino acid biosynthesis and redox homeostasis during stress responses 51 . The important interactions of ROC4 include its interaction with the 26S proteasome subunit RPT2A and with RACK1A (Fig. 2). RPT2A controls the meristematic activity in roots and shoots 52 . Both RPT2A and RACK1A mediate crosstalk between developmental and defense signaling pathways in plants 44,45,53,54 . Reduction of the concentration of ROC4 in transformed cells should ultimately lead to a change in the immune status of the cells. Unfortunately, the precise mechanism of the ROC4-RPT2A interaction or ROC4-RACK1A interaction is unknown.
RACK1 is a WD-40-type scaffold protein that is conserved in eukaryotes and plays regulatory roles in diverse signal transduction and stress response pathways 44 . RACK1A ensures the accumulation and processing of some pri-miRNAs, directly interacting with SERRATE and the AGO1 complex 55 . These interactions explain recent data indicating the active involvement of rolB in the modulation of expression of core components of the miRNA processing machinery, including SERRATE and AGO1 56 .
ROC4 also interacts with TRX3 (thioredoxin) 57 . TRX3 controls the abundance of numerous proteins that are involved in a wide variety of processes including the Calvin cycle, metabolism, photosynthesis, defense against oxidative stress and amino acid synthesis 57 . Again, the precise function of the ROC4-TRX3 interaction is unknown (see also the interaction data presented at http://www.ebi.ac.uk/intact/interaction/EBI-449668;jsessioni d=BC14D657B31C0FA84F73F7E9DC43F683). However, because TRX3 has a dual function as a disulfide reductase and a molecular chaperone 58 , decreased ROC4 abundance could diminish ROC4-TRX3 interactions and thus TRX activity. Indeed, many of the proteins whose expression is increased in rolB-expressing cells (Table 1) are under the control of TRX3 59 . These include ascorbate peroxidases, glutathione S-transferase DHAR, glutathione S-transferase F6, alanine aminotransferase and others.
ROC6 (CYP19-2) and CYP20-2: The cyclophilins ROC6 (CYP19-2) and CYP20-2 interact with the transcriptional repressor BZR1 and the cytokinin signaling system (Fig. 2). The interaction between CYP20-2 and BZR1 is presently considered important in the regulation of flowering 60 . BZR1 modulates ovule initiation and development by monitoring the expression of genes related to ovule development. The HK2 (histidine kinase 2) cytokinin receptor, together with the histidine-containing phosphotransferase protein AHP1 and the histidine kinase WOL,   regulates many developmental processes including meristematic activity, cell division, chlorophyll content, root growth and shoot promotion (TAIR annotation). Reprogrammed reproductive fate of the ovule, decreased chlorophyll content, lateral root growth and shoot promotion are characteristic traits of rolB-transformed plants 7,8,61 . Therefore, the CYP19-2:CYP20-2-HK2/BZR1 interactions provide evidence in favor of the involvement of rolB in cytokinin signaling and may explain the numerous cytokinin-dependent morphological alterations observed in A. thaliana rolB-transformed plants 61 .

VH1-interacting kinase (VIK). Expression of rolB in
Arabidopsis calli led to the activation of several regulatory proteins. We found increased expression of the VH1-interacting kinase (VIK) in rolB-transformed cells (Table 1). VIK participates in the regulation of the hub protein VH1/BRL2, facilitating the diversification and amplification of signals perceived by VH1/BRL2 62 . VH1/BRL2, in turn, interacts with TCTP (translationally controlled tumor protein), a general regulator required for the development of the entire plant, and with IAA7 (auxin-responsive protein IAA7), one of the members of the AUX/IAA family of auxin-induced transcriptional regulators. VIK is involved in the auxin-activated signaling pathway, the defense response to fungi, the negative regulation of programmed cell death, regulation of the plant-type hypersensitive response and responses to cold and water deprivation (TAIR annotation). Many of these responses have previously been shown to occur in rolB-expressing cells. RolB perturbs the auxin signaling pathway 1 , activates the defense response to fungi 12 , negatively regulates programmed cell death 10 , ensures higher resistance to salinity, cold and water deprivation 11 and causes symptoms that closely resemble systemic acquired acclimation 9 . However, we could not find a relationship between VIK and cyclophilins either in our reconstructions or the literature.
Auxins and cytokinin signaling. It is generally accepted that rolB-induced modification of hormone signaling causes developmental abnormalities in transformed plants. The interaction of rolB with the protein module VIK-VH1/BRL2-(TCTP; IAA7) (Fig. 2) offers a plausible explanation of the mechanism by which rolB modulates auxin signaling. TCTP is a central mediator of auxin homeostasis and root development 63 ; modification of its activity might be essential for the manifestation of many rolB-induced traits. Moreover, the function of TCTP in regulating cell division is part of a conserved growth regulatory pathway that is shared by plants and animals 64 , further confirming the idea that plant oncogenes affect ancient regulatory mechanisms. TCTP interacts with GRF1; modulation of its activity by rolB might also occur more directly via ROC1-GRF1-TCTP interaction (Fig. 2). IAA7 is connected with the expanded auxin subnetwork (27 proteins; the complete auxin network is presented in ref. 32 ). IAA7 mediates not only the response to auxin but also gravitropism. Lessening of gravitropism is a well-known effect of rolB 4 . It is clear that modification of auxin signaling in rolB-expressing cells is closely connected to the modification of cytokinin signaling. One pathway by which rolB might affect cytokinin signaling involves the interaction of ROC6 and CYP20-2 with HK2. This interaction affects the central cytokinin signaling module HK2-AHP1-WOL (indicated by the violet circles in Fig. 2). Thus, promising interactions for further investigation of the modification of auxin/cytokinin pathways in rolB-transformed cells include VIK-VH1/BRL2-(TCTP; IAA7), ROC1-GRF1-TCTP and ROC6;CYP20-2-HK2. It is very likely that auxin signaling is affected by cyclophilins in rolB-transformed cells; recent studies have shown a pivotal role of cyclophilins in auxin signaling and lateral root formation that includes perturbation of the activity of auxin-responsive Aux/IAA family proteins 65,66 . Certainly, these predictions must be further confirmed by experimental evidence.

Discussion
Primary Metabolism. Some enzymes of primary metabolism were highly activated in rolB-transformed cells, whereas some decreased in abundance. At first, these observations seem contradictory. However, we found that most of the enzymes that were hyper-activated in rolB-transformed cells have been shown to be highly responsive to various types of stress 59 (Table 1). In general, rolB suppresses primary metabolism and activates anti-stress defense pathways in cells.
Chaperonin Family Proteins. Hsp70s are highly conserved in eukaryotes, and some their functions are conserved in animals and plants. In animals, overexpression of Hsp70 was found to confer tumorigenicity and provide a selective survival advantage to tumor cells due to its ability to inhibit multiple pathways of cell death, including apoptosis 67 . In the case of the rolB gene, we can see a similar picture, i.e., increased abundance of some Hsp70 proteins ( Table 1) and inhibition of programmed cell death 10 . Therefore, rolB may function to provide favorable conditions for tumor growth after T-DNA integration. Only chloroplastic forms of Hsp proteins such as Hsp70-6, Hsp70-7, Hsp90-5/CR88 (synonym: Hsp88.1), 20-kDa chaperonin and chaperonin 60 subunit α1 were upregulated in rolB-transformed cells. Expression of genes encoding these proteins was also upregulated (Fig. 1). Indeed, recent data have shown the higher expression of genes encoding chloroplast heat-shock proteins in rolB-transformed tomato plants, compared with normal plants 68 .
It is presently unclear which reactions represent the direct action of RolB and which reactions compensate for this action. Presently, we assume that increased expression of chloroplastic heat shock proteins (Hsp70-6 and Hsp70-7, Hsp90-5, 20-kDa chaperonin and chaperonin 60 subunit α1) in rolB-transformed calli represents some kind of compensatory reaction. We propose the following development of events after the transformation. Basal levels of chaperones facilitate normal protein folding and guard the proteome against misfolding and aggregation. Increased expression of chaperones in normal Arabidopsis cells subjected to stress, which has been reported many times previously, is an adaptive response that enhances cell survival. The increased expression of chaperone proteins in rolB-transformed cells reflects the efforts of these cells to maintain homeostasis. These chaperone proteins also help tumor cells balance changes in cell biochemistry.
The enhanced expression of chaperonin family proteins in rolB-transformed calli can be linked with the decreased expression of cyclophilins CYP18-3 (ROC1), CYP19-2 (ROC6), CYP19-3 (ROC2), CYP20-2 and CYP20-3 (ROC4). Little is known about the functional connection of heat shock proteins with cyclophilins in plants 69 , but in animal and human studies, connections of this type have been demonstrated 20 . These interactions are critical in establishing tumor phenotypes through the disturbance of processes involved in protein folding, trafficking and degradation. Whereas these investigations are of high importance for human biology 20 , they are presently almost unknown for plant biology and represent an emerging (and intriguing) topic for understanding the formation of tumor phenotypes in plants.
Plant cells transformed with the rolB gene tolerate high temperatures 11 . Many properties of rolB-transformed cells resemble those of heat-acclimated plants, including inhibition of plant cell death, Hsp activation and induction of ascorbate peroxidases and other defense enzymes 70 . However, a fundamental difference is that the expression of cyclophilins is increased in heat-acclimated plants 70 but decreased in rolB-expressing cells. Taken together, our results indicate that rolB affects the expression of chaperone-type proteins such as heat-shock proteins and cyclophilins. These chaperones seem to regulate several layers of developmental and defense processes and potentially can affect many components of the Arabidopsis signaling system, including the RPM1-RPS2 signaling module, auxin and cytokinin signaling, the calcium signaling system and secondary metabolism.
Effector-Triggered Immunity. According to the zig-zag model of the plant immune system 71 , pathogens have evolved virulence factors that promote pathogen growth by suppressing pattern-triggered immunity (PTI). To counteract the action of specific pathogen effectors, plants have evolved effector-triggered immunity (ETI) 72 . In Arabidopsis, the ETI receptor RPM1 is activated by phosphorylation of the RPM1-interacting protein RIN4. During activation of the RPS2 pathway, RPS2 physically interacts with RIN4 73 . RPS2 initiates signaling based upon perception of RIN4 disappearance and induces plant resistance 73 .
The most probable scenario for rolB action is its primary effect which is inhibition of ROS, apoptosis and eventually cell immunity. However, rolB-transformed cells counteract this action in various ways. The first way is ROC1 suppression. Because ROC1 suppresses RPM1/RIN4 immunity in a PPIase-dependent manner 46 , it can be assumed that rolB-transformed cells, by suppressing ROC1, attempt to maintain a constitutively activated process that resembles ETI. Therefore, the final effects of rolB gene expression resemble ETI more than PTI. It is likely that RolB partially mimics the action of nucleotide-binding/leucine-rich-repeat (NLR) receptors that are necessary for ETI 72,74 .
On the other hand, RPM1 is an Hsp90-5 client protein 75 (Fig. 2). Hsp90-5, together with cofactors, ensures dynamic interactions in the module Hsp90-5-PBS2/RAR1-SGT1, which regulates the stability and function of RPM1 75 . Therefore, the current hypothesis is that rolB controls the RPM1-RPS2 signaling module in two ways: via ROC1-RIN4 and via Hsp90-5-RPM1 interactions. RolB, Cyclophilins and RACK1A. Ito and Machida recently suggested that plant T-DNA oncogenes change the epigenetic status of the host chromatin through intrinsic histone chaperone activity 17 . Indeed, in both plants and animals, cyclophilins acting as PPIases and chaperones alter transcription by altering chromatin structure and by other mechanisms that include the recruiting of chromatin-and histone-modifying enzymes 76 . Another possible effect of cyclophilin silencing in rolB-expressing cells is silencing of RACK1A, an important protein that regulates the small RNA (miRNA and short interfering RNA)-processing machinery. Therefore, the action of the rolB gene could be similar to that of the 6b gene, the product of which targets key components of the small RNA processing machinery, namely both the DCL1-SE-HYL1 and RISC/AGO1 complexes 77 . Intriguingly, RACK1 suppression promotes gastric cancer by modulating the expression of miRNAs 78 . RACK1 inhibition may be important for rolB-mediated tumor progression in plants.
Secondary Metabolism. The mysterious ability of rolB to greatly activate secondary metabolism in transformed cells has been known for many years 13 . It was recently shown that expression of rolB in Arabidopsis thaliana calli leads to the activation of genes encoding secondary metabolism-specific MYB and bHLH transcription factors 15 . Accordingly, a higher transcript abundance of corresponding biosynthetic genes related to these factors was detected. The effect of rolB on the expression of transcription factors was highly specific; for example, rolB did not induce MYB111 or PAP1 expression and caused the conversion of MYB expression from cotyledon-specific to root-specific patterns 15 .
It should be noted that none of the regulatory proteins described in the present work whose expression was changed by rolB gene activity can be attributed to the common secondary metabolism activator pathways described earlier for Arabidopsis 32 . The rolB gene most likely does not affect secondary metabolism directly; its effect is more likely a part of general defense reactions. We suggested three signaling modules by which rolB might influence secondary metabolism: ROC4-RACK1A → MYC2 (MYB2-TT8; JAZ1-TT8); (VIK-HAI1-HAB1-ABI2)-MYB12 and ROC2-(CAM-CDPK) (Fig. 2). The first of these is based on the MYB2 signaling module, which connects secondary metabolism with hormone (JA, auxin, cytokinin and ethylene) signaling 32 . The second represents the connection between secondary metabolism and abscisic acid, which is mediated by HAI1-MYB12 interactions 79 . The third, ROC2-(CAM-CDPK) module, represents a pathway of secondary metabolism activation known as activation through calcium-dependent protein kinases 80 . Considering the observation that rolB is a more powerful activator of secondary metabolism than a constitutively expressed CDPK gene 81 , we suggest that more than one mechanism is involved in its activator function.