Modulation of the hormone setting by Rhodococcus fascians results in ectopic KNOX activation in Arabidopsis.

The biotrophic actinomycete Rhodococcus fascians has a profound impact on plant development and a common aspect of the symptomatology is the deformation of infected leaves. In Arabidopsis (Arabidopsis thaliana), the serrated leaf margins formed upon infection resemble the leaf phenotype of transgenic plants with ectopic expression of KNOTTED-like homeobox (KNOX) genes. Through transcript profiling, we demonstrate that class-I KNOX genes are transcribed in symptomatic leaves. Functional analysis revealed that BREVIPEDICELLUS/KNOTTED-LIKE1 and mainly SHOOT MERISTEMLESS were essential for the observed leaf dissection. However, these results also positioned the KNOX genes downstream in the signaling cascade triggered by R. fascians infection. The much faster activation of ARABIDOPSIS RESPONSE REGULATOR5 and the establishment of homeostatic and feedback mechanisms to control cytokinin (CK) levels support the overrepresentation of this hormone in infected plants due to the secretion by the pathogen, thereby placing the CK response high up in the cascade. Hormone measurements show a net decrease of tested CKs, indicating either that secretion by the bacterium and degradation by the plant are in balance, or, as suggested by the strong reaction of 35S:CKX plants, that other CKs are at play. At early time points of the interaction, activation of gibberellin 2-oxidase presumably installs a local hormonal setting favorable for meristematic activity that provokes leaf serrations. The results are discussed in the context of symptom development, evasion of plant defense, and the establishment of a specific niche by R. fascians.

Infection of Arabidopsis (Arabidopsis thaliana) with the Gram-positive phytopathogenic actinomycete Rhodococcus fascians provokes an overall bushy and stunted appearance of the aerial plant parts, abnormal flowers, multiple rosettes and inflorescences, and small and serrated leaves (Vereecke et al., 2000;de O. Manes et al., 2004). The shooty phenotype is induced on a wide range of plants and results from the activation of axillary meristems and de novo meristem formation (de O. Manes et al., 2001). The central virulence determinant of R. fascians is encoded by the fas operon located on the linear plasmid pFiD188. The presence of an isopentenyltransferase (IPT) gene indicates that this locus encodes the biosynthetic machinery for cytokinins (CKs) or CK analogs (Crespi et al., 1992), but their identity awaits elucidation. Recently, infection of Arabidopsis and tobacco (Nicotiana tabacum) with R. fascians has been shown to induce genes involved in abscisic acid and GA catabolism, suggesting that modulation of the hormonal levels of the plant might be important for symptom development (Simó n-Mateo et al., 2006).
Transcription factors of the class-I KNOTTED-like homeobox (KNOX) family (SHOOT MERISTEMLESS [STM], BREVIPEDICELLUS [BP]/KNOTTED-LIKE1 [KNAT1], KNAT2, and KNAT6) are crucial for function and maintenance of meristems. The essential role of this gene family in meristem identity is reflected by the partially functional redundancy of BP/KNAT1 and STM, and of the KNAT2 and KNAT6 proteins (Byrne et al., 2002;Gallois et al., 2002;Hay et al., 2004). KNOX genes are required for the undifferentiated state of the cells in the central region of the shoot apical meristem (SAM). KNOX expression is inactivated in daughter cells that moved into the leaf initiation site and allows cell specification (Barton, 2001). The initial downregulation of STM and other KNOX genes upon initiation of simple leaves in the organ founder cells of the SAM is not fully understood, but recent data suggest that auxins might be implicated (Barton, 2001;Byrne, 2005;. KNOX gene expression is also involved in leaf shape determination (Chuck et al., 1996;Reiser et al., 2000;Byrne et al., 2001;Frugis et al., 2001;Hay et al., 2003;Hake et al., 2004;Mü ller et al., 2006). Ectopic expression of BP/KNAT1 in the simple leaves of Arabidopsis results in lobing of the leaf margins and the formation of epiphyllic meristems (Lincoln et al., 1994), whereas KNOX expression occurs naturally in the leaves of a number of species with compound or dissected leaves (Hareven et al., 1996;Bharathan et al., 2002;. Several negative regulators repress KNOX gene expression at the leaf initiation sites of the SAM (Ori et al., 2000). YABBY3 (YAB3) and FILAMENTOUS FLOWER (FIL) down-regulate KNOX genes upon lateral organ development, in addition to their role in establishing abaxial leaf identity and lamina growth (Sawa et al., 1999;Kumaran et al., 2002;Nole-Wilson and Krizek, 2006). ASYMMETRIC1 (AS1), a Myb transcription factor, together with AS2, a lateral organ boundary protein, probably operate as a heterodimer and also negatively regulate BP/KNAT1, KNAT2, and KNAT6 (Byrne et al., 2000), whereas, in turn, STM negatively regulates AS1 and AS2 expression, placing it high in the KNOX hierarchy (Hake et al., 2004).
The alteration of GA and CK levels in plants overexpressing KNOX genes suggests that KNOX proteins operate through interactions with plant hormones (Kusaba et al., 1998). Transcription of GA 20-oxidase (GA20ox) genes, involved in GA biosynthesis, is repressed by KNOX proteins, creating a low GA regime favorable for the meristematic character of the SAM (Jasinski et al., 2005). Constitutive GA signaling or exogenous GA application suppresses KNOX-overexpressing phenotypes, supporting the antagonistic role of GA (Sakamoto et al., 2001;Hay et al., 2002;Rosin et al., 2003;Chen et al., 2004). The strong similarities between the KNOX-overexpressing phenotypes and transgenic plants overexpressing an IPT gene (Kieber, 2002) hint at a function for CKs. Recently, STM and BP/KNAT1 have been shown to activate CK biosyn-thesis by inducing transcription of IPT5 and IPT7 (Jasinski et al., 2005;Yanai et al., 2005;Sakamoto et al., 2006). The resulting higher CK level is necessary and sufficient to stimulate GA catabolism by inducing GA 2-oxidase (GA2ox) gene expression. Whereas the direct control of KNOX proteins over CK biosynthesis is clearly illustrated, there are conflicting data on the role of CKs in triggering STM and BP/KNAT1 expression (Rupp et al., 1999;Rashotte et al., 2003). Hence, the occurrence of a feedback loop between CK and KNOX gene expression remains a matter of debate.
We evaluate the response of Arabidopsis to R. fascians infection by analyzing morphological, molecular, and biochemical alterations that occur in the small and serrated symptomatic leaves generated upon the interaction. Our data demonstrate that the leaf deformations are caused by ectopic KNOX gene expression as a consequence of induced modifications of the hormone balance in the plant.

Rhodococcus fascians Infection Affects Leaf Morphology
To assess the effect of R. fascians on leaf development, Arabidopsis plants (ecotype C24) were infected with the wild-type strain D188 and the plasmid-free nonpathogenic derivative D188-5 by applying a drop of bacterial suspension to the apex of plants with five visible leaves (growth stage 1.05; Boyes et al., 2001). Whereas plants were not responsive to strain D188-5 and developed as noninfected controls, deformations of the rosette leaves and axillary shoot production were observed 2 to 3 weeks after infection with strain D188 (Supplemental Fig. S1, A and B). The first response appeared in leaf 8 and, from then on, the newly formed rosette leaves and the leaves derived from activated axillary shoots were highly affected (Fig. 1, A and B). Typical symptoms were a strongly serrated leaf margin, a narrow and short lamina, the absence of leaf expansion, a loss of leaf symmetry, and an uneven leaf surface (Fig. 1D). Timing of leaf formation and phyllotaxis were not altered, but leaves that developed later after infection had much deeper serrations than the first affected leaves (Fig. 1, A and B). The highly deformed inflorescences did almost not elongate, which contributed to the bushy, stunted appearance of infected plants (data not shown).
These alterations of the leaf shape were reminiscent of the reported leaf phenotype of plants misexpressing KNOX genes (Lincoln et al., 1994;Chuck et al., 1996). However, in the 35S:KNAT1 plants analyzed, the margins were deeply lobed, but not serrated, and neither leaf expansion nor overall plant development were compromised (Supplemental Fig. S1C; Fig. 1E). Also, leaves of dexamethasone (DEX)-induced 35S:STM-GR plants (Gallois et al., 2002) vaguely resembled R. fasciansinfected leaves. In both cases, swelling around the vascular tissue and an often asymmetric, wrinkled, and bulgy lamina were evident (Supplemental Fig.  S1D; Fig. 1F).

Altered Leaf Morphology upon R. fascians Infection Is Correlated with Ectopic KNOX Gene Expression
The putative role of KNOX gene misexpression in leaf deformations induced upon R. fascians infection was assessed by analyzing the transcript levels of the four class-I KNOX genes (STM, BP/KNAT1, KNAT2, and KNAT6) by quantitative reverse transcription (qRT)-PCR in symptomatic leaves 30 d postinfection (dpi). KNOX gene expression was induced by the interaction of Arabidopsis with R. fascians strain D188, albeit to a different level depending on the gene (Fig. 2, A and B). For KNAT2 and KNAT6, an up-regulation was measured. No transcripts of BP/KNAT1 and STM were detected in D188-5-infected controls, confirming that these genes are not expressed in wild-type Arabidopsis leaves. However, high expression values for STM and BP/KNAT1 were obtained in symptomatic leaves upon D188 infection ( Fig. 2A). These observations were confirmed by infection of the STM:YFP and KNAT1:GUS marker lines. Whereas STM or BP/KNAT1 were not expressed in leaves upon D188-5 infection (data not shown), inoculation with strain D188 induced both reporter genes. The expression was localized at the basis and near the vascular tissue of symptomatic rosette leaves (Fig. 2, C-E), and at similar locations in the newly formed axillary shoots (Fig. 2F). Confocal analysis of the infected STM:YFP line revealed expression in cells in sinuses of serrated leaves (data not shown), while staining was occasionally visible with the KNAT1:GUS line at the leaf margin and in patches on the leaf blade ( The morphological discrepancy between the transgenic 35S:STM-GR and R. fascians-infected plants might result from only a partial overlap of the mod-ulation of the KNOX pathway. To test this hypothesis, the transcript profiles of several KNOX pathwayrelated genes were determined in both experimental systems. 35S:STM-GR plants were induced by DEX, and wild-type C24 plants were infected with R. fascians, 16 d after germination. For the R. fascians-infected plants, no symptoms were apparent at 10 dpi, but deformations became visible at 17 and 21 dpi. The effect of DEX induction was already apparent 2 d posttreatment (dpt). Therefore, leaves of treated and untreated plants were harvested for qRT-PCR analysis at 1, 7, 10, 17, and 21 dpt. Expression was quantified of the four class-I KNOX genes, three negative regulators, and GA-and CK-related genes shown to be controlled by KNOX proteins. An overview of the results is given in Figure 3 as relative data of treated versus control samples. Seventeen days after R. fascians infection, coinciding with the appearance of leaf deformations, all four KNOX genes were switched on (see also Fig. 2A). For the 35S:STM-GR control plants, KNOX up-regulation was evident at 1 dpt, in agreement with reported data (Yanai et al., 2005).
We reasoned that the expansion of the KNOX expression domain (Fig. 2, C-G) could result from a down-regulation of negative regulators, such as AS1, AS2, and YAB3. Although YAB3 was indeed repressed in 35S:STM-GR plants upon induction, only a transient repression was detected in R. fascians-infected plants.
In 35S:STM-GR, AS1 was transiently up-regulated, whereas AS2 was not differentially expressed. Upon R. fascians infection, AS1 was up-regulated at the onset of symptom development, whereas AS2 was upregulated after an initial down-regulation. These observations indicate that the modulation of KNOX gene expression upon R. fascians infection does probably not occur via these central regulators.
Although repression of GA biosynthesis by KNOX proteins (Kusaba et al., 1998) was apparent upon DEX treatment of the 35S:STM-GR plants, the GA20ox1 gene was not down-regulated after R. fascians infection. In contrast, expression of the GA2ox3 gene, involved in GA catabolism, increased early in the interaction and preceded by far R. fascians-induced symptom development and KNOX expression. No consistent differential GA2ox3 gene expression was observed in the 35S:STM-GR plants.
Also the CK-related genes responded differently in both experimental setups. During the R. fascians interaction, expression of IPT5 and IPT7 was switched off at 7 dpi, and that of a CK dehydrogenase gene (CKX3), involved in irreversible CK inactivation, was strongly up-regulated. In contrast, no differential expression of CKX3 was observed in the 35S:STM-GR plants and the CK-biosynthetic genes, especially IPT7, were up-regulated. Interestingly, expression of the ARABIDOPSIS RESPONSE REGULATOR5 (ARR5) gene was upregulated with the same kinetics in both experimental systems, indicating the increased CK signaling in both cases.

Modulation of GA Content upon R. fascians Infection
The transcript data presented above imply that R. fascians infection activates GA degradation. The occurrence of a putatively low GA regime during symptom development was tested by treating infected plants with GA and by measuring the GA content in leaves at different time points after infection. GA 3 was exogenously applied at 2 dpi, either by transferring infected plants to GA-containing medium or by treating the plants with a GA-Silwet solution on a daily basis until symptoms arose. Although occasionally leaves had fewer or no serrations, overall no clear and reproducible reversion of the leaf deformations was obtained. Full expansion of the leaf blade was not restored upon GA application and the GA-treated plants still exhibited the typical bushiness induced by R. fascians infection (data not shown).
Next, the level of 10 GAs was determined in the leaves of Arabidopsis in a time course experiment, in which plants were compared that were mock-inoculated with water, infected with R. fascians strain D188-5, or infected with strain D188. No conclusive data could be obtained for GA 8 and GA 29 (data not shown). Infection with the nonpathogenic strain D188-5 did not alter the kinetics of any of the measured GAs in comparison with that of the mock-inoculated control (Fig. 4). In contrast, inoculation with strain D188 caused a biphasic metabolic pattern. In the early phase, corresponding to 1 to 10 dpi, the concentration of most GAs was similar to that of the controls, with an average concentration ratio for D188/D188-5 of 1.2 at 10 dpi. Only the level of GA 34 , a GA2ox3 reaction product, increased over time. Accumulation of this catabolite started between 1 and 4 dpi to reach a 2-fold higher concentration at 10 dpi compared with D188-5 infections, supporting the transcript data. In the second phase from 10 to 21 dpi, coinciding with the appearance of young leaves and shoots, the GA spectrum shifted quantitatively. Relative to the D188-5 control, the GA precursors GA 53 (3.5-fold), GA 19 (5.4-fold), GA 24 (15.7-fold), and GA 9 (9.8-fold) strongly accumulated; the levels of the bioactive GA 1 and GA 4 and of the catabolite GA 34 increased 3-fold. The concentration of GA 20 followed the trend of the controls, but the final level increased only slightly (1.4-fold).
To evaluate the physiological response in the DEXinduced 35S:STM-GR plants, the GA content of the leaves was determined at 0, 1, and 7 dpt. The GA levels were comparable to those of wild-type plants mockinoculated with water at 7 dpi. Despite the pronounced morphological alterations, the metabolic profiles obtained with or without induction did not differ, except for GA 20 whose concentration decreased 2.5-fold at 7 dpt (data not shown), which is in agreement with the transcriptional down-regulation of the GA20ox1.

Modulation of the Plant CK Metabolism by R. fascians Infection
The transcript profiling data showed that early differential gene expression induced upon R. fascians . GA profiles during the R. fascians-Arabidopsis interaction, indicating an increase of GAs and metabolites at 10 to 14 dpi with strain D188, coinciding with the generation of symptomatic leaves. The GA content in leaves of Arabidopsis C24 plants infected with R. fascians strain D188 (black triangle), strain D188-5 (black circle), and mock-inoculated with water (black square). The respective GA metabolite is indicated above each graph. Error bars represent SD.
infection is related to CK metabolism. A more elaborate transcript analysis for all IPT and CKX genes of Arabidopsis on leaves at 30 dpi confirmed that IPT7 and IPT5 were almost completely switched off and that IPT3 was down-regulated ( Fig. 5A). No differential expression was observed for IPT2 and IPT9, which are the two tRNA IPT genes (Miyawaki et al., 2004). IPT1, IPT4, IPT6, and IPT8 were not expressed in the aerial tissues, confirming the data reported by Miyawaki et al. (2004). Strikingly, expression of all seven CKX genes was strongly induced in symptomatic tissue (Fig. 5B). This observation was supported by measuring CKX enzyme activity in infected plants (Fig. 5C). In Arabidopsis tissue infected with strain D188-5, CKX activity was below the detection limit, but upon infection with the virulent strain, specific enzyme activity values were obtained of 1.18 6 0.22 nkat mg 21 with 2,3-dimethoxy-5-methyl-1,4-benzoquinone-indophenol ( Fig. 5C) and 1.30 6 0.12 nkat mg 21 with 2,6-dichlorophenol-indophenol as electron acceptor at pH 6.5 (data not shown).
These data suggest that the CK levels are reduced upon infection with strain D188; therefore, the CK content was quantified in Arabidopsis leaves at different time points for three treatments: mock inoculation with water, infection with strain D188-5, and infection with D188. The metabolite patterns obtained from the mock inoculations and the D188-5 infections were comparable (Fig. 6), but several CK types were affected upon infection with strain D188. For the biosynthetic intermediates (CK monophosphates, the reaction products of the IPT enzymes), a strong and continuous reduction of the trans-zeatin-monophosphate level started from 4 dpi. Although the effect on cis-zeatin-monophosphate and isopentenyladenine (iP) monophosphate was less pronounced, a lower level than that of the control samples was apparent from 17 dpi (Fig. 6).
From 7 dpi onward, the 9-N conjugates and especially the 7-N conjugates decreased significantly. After an initial accumulation, iP, the only free base that could be measured, decreased continuously over time, reaching final concentrations that were half those obtained upon D188-5 infections. iP, 7-N, and 9-N conjugates are substrates of the CKX enzymes, thus the measured decrease upon D188 infection supports the transcript data. Finally, whereas no significant changes could be measured for the CK ribosides (data not shown), a strong decrease in the zeatin-O-glucoside level was observed upon D188 infection. CK measurements on tissues harvested from DEX-induced 35S: STM-GR plants at 0, 1, and 7 dpt showed an increased concentration of all CK metabolites over time, except for the free bases, confirming the data reported (Yanai et al., 2005; data not shown).
To get further insight into the role of exogenous CKs in leaf deformation and symptom development, the apical part of stage 1.05 Arabidopsis plants was treated once with an agar-CK paste and plant development and expression of KNOX pathway genes were evaluated at 28 dpt. iP had no impact on leaf development (data not shown), but application of 10 25 M 6-benzylaminopurine (BAP) partially mimicked R. fascians-induced symptoms (Fig. 7). Newly formed leaves were deformed, but rather lobed than serrated and leaf expansion was less inhibited. The effect emerged immediately from leaf 6 onward, and was only retained in four to five subsequent leaves (Fig.  7B). In addition, the typical bushiness of plants infected with R. fascians was not observed upon CK treatment (data not shown). Transcript profiling of plants treated with BAP revealed that although the trends in gene expression were similar for independent biological repeats, absolute expression values varied a lot (data not shown). Upon BAP treatment, compared with a nontreated control, ARR5, CKX3, and GA2ox3 were up-regulated, and IPT7 and GA20ox1 were down-regulated (Fig. 7E). These expression profiles resemble those obtained upon R. fascians infection, but the effect of exogenous CK addition was much less pronounced. In contrast to R. fascians infection, upon BAP treatment IPT5 expression was up-regulated ( Fig. 7E) and no expression of the four class-I KNOX genes could be demonstrated (data not shown). Furthermore, no differential expression was observed for AS1 and AS2, whereas YAB3 was upregulated.

Functional Analysis of KNOX Pathway-Related Genes in R. fascians-Induced Symptomatology
To strengthen the correlation between KNOX genes, CK signaling, and metabolism during R. fasciansprovoked disease on Arabidopsis, a functional analysis was carried out. KNOX gene involvement was assessed by monitoring symptom development in stm-2, stm-8, bp-1, bp-9, knat2, knat6, knat2 stm-2, knat6 stm-2, and knat2 knat6 mutant lines. The lines with a defective STM allele fail to maintain a functional SAM after germination, but all except the knat6 stm-2 mutant (Belles-Boix et al., 2006) retain organogenic potential and form vegetative shoots (Clark et al., 1996;Fig. 8A) Whereas the knat6 stm-2 line did not respond to R. fascians infection, the other STM mutant plants were smaller, paler, and had less expanded leaves at 28 dpi compared with the noninfected controls (Fig. 8A).
Interestingly, none of these mutants displayed serrated leaves, demonstrating the necessity of the STM protein in the determination of the leaf shape. The differential expression of CKX3, ARR5, GA2ox3, and IPT7 still occurred in these mutants (Fig. 8B), placing the role for STM downstream of the CK response.
The other tested KNOX gene mutants (bp-1, bp-9, knat2, knat6, and knat2 knat6) had no pronounced developmental defects and exhibited shooty symptoms and serrated leaves after infection (Supplemental Fig. S2A). Nevertheless, when analyzed in more detail (see ''Materials and Methods''), the responsiveness of these plants was reduced when compared with wild type (Supplemental Fig. S2E). To quantify the level of leaf serration, a serration index (SI) was calculated as the number of serrations in a symptomatic leaf, multiplied by the average depth of the serrations in that leaf, and divided by the leaf surface area. At least 20 symptomatic leaves per mutant line were analyzed and only for the bp-9 mutant a significant reduction of the SI could be measured (Fig. 8C). A similar result was obtained for the bp-1 mutant in the Ler background (data not shown). Single and higher order mutants in IPT (ipt3; ipt5; ipt7; ipt3,5; ipt3,7; ipt5,7; and ipt3,5,7) and A-type ARR genes (arr5; arr4,5; arr5,6; arr3,4,5,6; arr5,6,8,9; and arr3,4,5,6,8,9) exhibit no major phenotypic changes compared to wild-type plants, apart from subtle alterations for some ARR mutants like elongated petioles and a reduced rosette size for arr5 Miyawaki et al., 2006). Upon infection, all of these mutants developed the typical R. fascians-induced symptoms with the same kinetics as wild-type controls (Supplemental Fig. S2, C and D).

DISCUSSION
The morphological changes induced by R. fascians on its many hosts implicate a profound perturbation of the hormone balance (Vereecke et al., 2000;de O. Manes et al., 2004). A very characteristic phenotype of R. fascians infection on Arabidopsis is the deformation of the newly formed leaves and their highly serrated margins (Fig. 1). These symptoms are reminiscent of the indeterminate growth characteristics of leaves of transgenic plants with ectopic expression of the KNOX pathway (Hake et al., 2004). KNOX transcription factors are believed to function by reducing the GA levels and activating the CK biosynthesis, thus creating hormone settings that favor meristematic tissue identity (Jasinski et al., 2005). The results obtained upon R. fascians infection of wild-type, mutant, and transgenic Arabidopsis plants and upon DEX induction of 35S:STM-GR plants are summarized in Figure 10.
We assessed the involvement of the KNOX pathway in R. fascians-induced leaf deformation of Arabidopsis by evaluating the expression pattern of the four class-I KNOX genes, three central negative regulators of KNOX gene expression, and GA-and CK-related genes that are implicated in the establishment of the meristem-specific hormone setting. Transcript profiling combined with localization of reporter gene expression (Figs. 2 and 3) showed that STM and BP/ KNAT1 were ectopically expressed and KNAT2 and KNAT6 expression was up-regulated in symptomatic leaves. The expression pattern of the regulators YAB3, AS1, and AS2 did not support a role in the activation of the KNOX pathway upon R. fascians infection. Interestingly, comparison with the expression profiles of this set of genes in DEX-induced 35S:STM-GR plants revealed that the sequence of the modulation in the KNOX pathway differed completely in both experimental systems (Fig. 3). Whereas the reported control of the KNOX proteins on GA20ox and IPT expression Figure 7. Partial mimicking of the R. fascians phenotype by local application of CKs to the Arabidopsis shoot apex. Leaf series of an Arabidopsis C24 control plant (A) and upon local treatment at the apex with a BAP-containing agar suspension at 28 dpt (B); leaf phenotype of control (C) and after 10 25 M BAP treatment (D); changes in expression levels of KNOXrelated genes upon BAP treatment determined by qRT-PCR (E). KNOX genes were not ectopically expressed upon BAP treatment. Asterisks mark the time point of CK treatment. Bars, 1 cm. was confirmed in the 35S:STM-GR experiment, the first response to R. fascians infection at 1 dpi was an upregulation of CKX3 and ARR5, followed by a downregulation of the IPT genes and an up-regulation of GA2ox3 at 7 to 10 dpi. Hence, the ectopic KNOX gene expression starting at 17 dpi is not directly involved in the redirection of the hormone balance of the symptomatic leaf, but is a consequence of this modification induced by R. fascians.
The downstream position of the KNOX pathway in the sequence of events occurring during interaction with R. fascians is supported by the responsivity of knox mutants. Although the stm-2, stm-8, and knat2 stm-2 mutants become symptomatic, they do not exhibit serrated leaves (Fig. 8A). Similarly, bp-1 and bp-9 display fewer and milder serrations upon infection (Fig. 8C). These findings demonstrate that BP/KNAT1 and especially STM are essential for R. fascians-induced leaf deformation and confirm the reported unequal functional redundancy between STM and BP/KNAT1 (Byrne et al., 2002;Douglas et al., 2002). Whereas none of the other tested knox mutants exhibit pronounced symptomatic defects, knat6 stm-2 mutants do not respond at all ( Fig. 8A; Supplemental Fig. S2A). The latter observation is not unexpected because these plants have lost all meristematic potential and never develop further than the cotyledonary stage (Belles-Boix et al., 2006). The downstream position of the KNOX genes is supported by the maintenance of the differential expres-sion of the IPT7, CKX3, ARR5, and GA2ox3 genes in the weak stm mutants (Fig. 8B).
The CK-related data indicate that the plant responds to R. fascians infection by activating processes that would result in a reduction of the overall CK concentration (Fig. 5). The levels of three CK monophosphates decreased over time, in correlation with the down-regulation of the IPT genes (Fig. 6). Moreover, the concentration of the CKX substrates, N-glucosides, and iP (Werner et al., 2001) decreased drastically

Control (left) and infection (right) phenotypes of weak STM mutants (A). Transcriptional response upon infection (B). Symptomatic leaf series of knox mutants after infection with strain D188 (C). The SI is indicated in parentheses.
during the experiment, in accordance with the increased CKX transcription and activity (Figs. 5 and 6). The net result of the interaction at the end of the experiment was a 3.5-fold decrease in the overall CK content when compared with the controls. Although these results are in agreement with previous reports Vereecke et al., 2000;de O. Manes et al., 2001;Gális et al., 2005aGális et al., , 2005b, they seem to contradict the pronounced CK-related phenotypes induced on infected plants. R. fascians-induced symptoms are anticipated to be triggered by bacterium-secreted CKs. Indeed, no fewer than 11 different CKs have been identified in the supernatant of several isolates (among them iP; Armstrong et al., 1976;Murai et al., 1980;Eason et al., 1996). Despite the lack of a strong link between any of these CKs and virulence , the central role of bacterial CK production in the symptomatology is emphasized by the nonpathogenic phenotype of an ipt mutant of the R. fascians strain D188 (Crespi et al., 1992) and the strict correlation between the presence of the IPT gene and virulence in other R. fascians isolates (Stange et al., 1996). The implementation of homeostatic mechanisms to keep the hormone concentrations below a lethal level has been reported in transgenic plants that ectopically express a bacterial IPT gene and upon exogenous addition of CKs (Motyka et al., 1996). Therefore, up-regulation of the CKX genes is probably a compensatory mechanism activated by the host in an attempt to restore CK homeostasis in response to bacterial CK secretion. The transient peak of iP measured at 1 dpi could reflect these events (Fig. 6). Moreover, the enhanced response of some of the 35S:CKX transgenic plants upon infection (Fig. 9) implies that the bacterial signal is not a substrate for the CKX enzymes.
Similarly, the inhibition of CK biosynthesis by blocking IPT expression might be a negative feedback mechanism imposed by bacterial CKs, the occurrence of which is supported by the activation of ARR5 expression (Fig. 3). In contrast to the situation in the DEX-induced 35S:STM-GR plants ( Fig. 3; Jasinski et al., 2005;Yanai et al., 2005), IPT expression is not triggered during R. fascians infection, despite the ectopic KNOX gene expression. Hence, the negative control of bacterial CKs on IPT expression is dominant over the stimulatory effect of KNOX proteins (Fig. 10). We had anticipated that infection of ipt mutants might result in an earlier or enhanced infection phenotype, but this was not the case. Functional redundancy between members of the ipt family is a possible explanation.
The deduction that bacterial CK production might cause the observed expression profiles was further explored by simulating R. fascians infection by exogenous addition of BAP at the apex of Arabidopsis. The shape of the newly formed leaves was similar to those infected with R. fascians, although the serrations were much less pronounced and the effect was only transient (Fig. 7, B and C). Moreover, whereas ARR5, IPT7, GA2ox3, and GA20ox1 had comparable expression profiles, albeit less prominent to those obtained upon R. fascians infections, no ectopic KNOX gene expression could be demonstrated (Fig. 7D). The single application of BAP instead of the continuous delivery of CKs by R. fascians might be the reason for these differences. Interestingly, in CK-overproducing plants, the transcript levels of STM and BP/KNAT1 were elevated (Rupp et al., 1999), suggesting that CKs might trigger KNOX expression. However, activation of KNOX transcription was not confirmed by microarray analysis of tissues treated with exogenous CKs (Rashotte et al., 2003). This discrepancy might well be accounted for by the endogenous versus exogenous administration of CKs in both experimental setups. We observe a similar discrepancy for KNOX gene expression upon R. fascians infection and upon exogenous CK application, suggesting that the constant delivery of CKs by R. fascians mimics endogenous changes. The R. fascians-induced effects could result either from the constant and very local delivery of standard CKs concomitant with a high turnover by the plant CKX enzymes, or from the activity of specialized CKs. Although this issue remains to be clarified by bacterial CK analyses that are currently ongoing, the strong response of the 35S:CKX plants supports the latter hypothesis.
Transcript profiling showed that simultaneously with ARR5, GA2ox3 expression was activated very early during the interaction with strain D188 (Fig. 3), coinciding with a buildup of the GA 4 catabolite GA 34 (Thomas et al., 1999;Pimenta Lange and Lange, 2006) from 2 to 4 dpi onward (Figs. 4 and 10). However, this increased deactivation was not accompanied by a reduced level of bioactive GAs that might be the consequence of a very localized occurrence of the GA modulation. Similarly, although it is generally accepted that KNOX transcription factors down-regulate GA biosynthesis, thus decreasing the GA content (Jasinski et al., 2005), in our experimental setup, the amount of GA in leaves of DEX-induced versus control 35S:STM-GR plants did not differ markedly (data not shown). In the second phase of the interaction, from 10 dpi onward, at symptom onset, both the nonhydroxylation and the hydroxylation biosynthetic pathways were highly active (Fig. 4). Although the precursors GA 19 and GA 24 were the most abundant GAs upon D188 infection, their conversion to GA 20 and GA 9 by GA20ox was rate limiting (Talon et al., 1990;Coles et al., 1999;Hedden and Phillips, 2000). The subsequent conversion of GA 20 and GA 9 to the bioactive GA 1 and GA 4 by GA3ox was not (Coles et al., 1999). The same kinetics occur during regular GA biosynthesis. Moreover, the resulting 3-fold increase in GA 1 and GA 4 levels did not lead to GA-overproduc-tion phenotypes, such as accelerated stem elongation and longer internodes, as observed in transgenic or mutant plants with an altered GA metabolism (Huang et al., 1998;Coles et al., 1999). Altogether, the activated GA biosynthesis measured at the later time points of the interaction with strain D188 might reflect the overrepresentation of young GA-producing tissues in these samples generated upon infection. Preliminary data from a microarray experiment in which Arabidopsis is infected with strains D188 and D188-5 support this hypothesis, because a 2-fold increased expression of ent-kaurene synthase, involved in the first part of GA biosynthesis in the plastids, is observed between 14 and 24 dpi, implying an increased flow through the GA pathway (our unpublished data).
Activation of the KNOX pathway has been described for few other plant-microbe interactions, such as the formation of giant cells during root-knot nematode infection and of nodules during rhizobial endosym- At4g00180 qRT TCACGGTCACCGACAAAAGGT GTCCTTGCTGTGAGTGTTCCT Cnops et al. (2004) bioses (Koltai et al., 2001). Recently, ectopic expression of class-I KNOX genes has been demonstrated in the severely deformed leaves of transgenic tobacco and Arabidopsis plants expressing the oncogene 6b from Agrobacterium tumefaciens (Terakura et al., 2006), and in leaves of peach (Prunus persica) infested with the fungus Taphrina deformans (Bruno et al., 2005). Interestingly, ectopic BP/KNAT1 gene expression in Arabidopsis results in a decreased lignification (Mele et al., 2003). Modification of cell walls by regulation of lignin deposition and quality are coordinated by BP/KNAT1 to achieve proper cell differentiation. In the context of plant-microbe interactions, the induction of ectopic KNOX gene expression and the concomitant downregulation of lignification might facilitate cell proliferation and, more importantly, microbial entry and spread. Hence, KNOX gene manipulation by microbes could be a way to decrease plant defense-related lignification (Vance et al., 1980). Thus, the activation of the KNOX pathway in response to R. fascians infection might contribute to the circumvention of plant defense and the formation of local lesions in the epidermis. Indeed, accumulation of defense-related secondary metabolites (Vereecke et al., 1997;Lin et al., 2003) or extensive cell wall thickenings (de O. Manes et al., 2001(de O. Manes et al., , 2004, typical signs of strong defense responses, have not been observed upon R. fascians infection. Our data indicate that CK secretion by R. fascians activates compensatory plant mechanisms in an attempt to restore the overall hormone homeostasis. The continuous bacterial CK delivery probably leads to a very locally increased GA catabolism and a hormone balance in favor of meristematic tissue identity. The activation of KNOX gene expression beyond the SAM domain imposes indeterminate growth characteristics on leaves generating serrations and keeping the leaves in a meristematic or juvenile state. Young tissue is highly susceptible to the signals produced by R. fascians (Lacey, 1939;Faivre-Amiot, 1967;Vereecke et al., 2000); therefore, the fixation of symptomatic tissue in a meristematic state would ensure the amplification of the symptoms. Moreover, endophytic R. fascians has been hypothesized to feed on metabolites typical for meristematic tissues (Vereecke et al., 2002). Hence, the proliferation of tissues that do not mature would warrant the supply of specialized nutrients. The high GA content in symptomatic tissue may positively affect the photosynthetic rate and the development of the vasculature (Biemelt et al., 2004), further contributing to the nutritional value of and nutrient transport in infected tissue.
In conclusion, the ectopic activation of the KNOX pathway by R. fascians demonstrates a strategy to redirect developmental host processes to stimulate the creation of a selective niche. Nevertheless, neither ectopic KNOX expression in transgenic lines nor exogenous addition of CKs result in the typically bushy plant phenotype induced by R. fascians, suggesting that additional pathways are modulated by the bacterial signals.

Infection and Sampling
The Rhodococcus fascians strains used were the pathogenic strain D188, containing the linear virulence plasmid pFiD188, and the plasmid-free nonpathogenic strain D188-5 (Desomer et al., 1988). These strains were grown in liquid yeast extract broth for 2 d at 28°C under gentle agitation until the late exponential phase. Prior to infection, these cultures were washed and concentrated four times by resuspending the bacterial pellets in sterile deionized water.
At 16 d postgermination, Arabidopsis plants were infected by local application of a drop of bacterial culture to the SAM. To extract total RNA or for hormone measurements, leaves were sampled, but without petiole, as closely as possible to the leaf blade to avoid contamination with SAM or axillary meristem tissue at 0, 1, 4, 7, 10, 14, 17, and 21 dpi. At 14 to 17 dpi, symptoms became clearly visible.

Phenotype and Image Analyses
To follow the response of plants infected with R. fascians strain D188 over time, several parameters were taken into account. Before the appearance of shooty symptoms and small serrated leaves, plants were considered responsive based on anthocyan production, trichome induction, curling of existing leaves, reduced chlorophyll content, and/or a higher compactness of the plant, when compared with noninfected controls or controls infected with strain D188-5. The phenotype was scored on minimal 20 plants in three independent biological repeats.
To determine the leaf SI, a symptomatic leaf was dissected, and photographed with a stereomicroscope (Leica). The leaf surface area and the average depth of the serrations for that leaf were determined by analyzing the image in ImageJ. The SI was calculated by multiplying the number of serrations with the average depth of the serrations, and divided by the leaf surface area: the higher the resulting value, the more symptomatic the leaves. The SI was calculated in two independent biological repeats, for at least 20 leaves per line.

RNA Isolation, cDNA Synthesis, and qRT-PCR
During sample collection, the tissue was snap-frozen in liquid nitrogen and stored at 280°C until RNA isolation. RNA was isolated with the RNeasy Plant Mini Kit (Qiagen) according to the manufacturer's instructions. These RNA preparations were DNAse treated and purified through NH 4 Ac (5 M) precipitation. Quality and quantification were controlled with a NanoDrop spectrophotomer (Isogen). Samples of 2 mg of RNA were used for cDNA synthesis with the SuperScript reverse transcriptase kit (Invitrogen), subsequently diluted 50 times, and stored at 220°C until further use.
Each qRT-PRC was carried out with a LightCycler 480 (Roche Diagnostics) with SYBR Green for detection, in a 10-mL volume (5 mL master, 0.5 mL 5 mM of each forward and reverse primer, and 4 mL cDNA) in triplicate in a 384multiwell plate to allow determination of mean and SD. Cycle threshold (C T ) values were obtained with the accompanying software and data were ana-lyzed with the 2 2DDCT method (Livak and Schmittgen, 2001). Obtained values were normalized against those of actin that was used as an internal standard. An overview of all primers is given in Table I. The mean expression level of each gene was calculated from three biological repeats, obtained from three independent experiments. The expression data are presented as relative values for the R. fascians interaction (strain D188 infected versus strain D188-5 infected plants) and for the 35S:STM-GR plants (DEX-induced versus noninduced). The data were converted to heat maps with the TMEV-3D software (Multiexperiment Viewer; The Institute for Genome Research).
For reverse transcription (RT)-PCR, cDNA was prepared following the same protocol, but the cDNA was diluted only five times. In each semiquantitative RT-PCR (Miyawaki et al., 2004), 0.1 mg RNA was used in a reaction volume of 20 mL. After the PCR, 10 mL was loaded on 1% agarose gels containing SybrSafe for visualization.

GUS Staining
GUS-marked plants were submerged in 90% (v/v) acetone at 4°C for 1 h and transferred to a GUS-staining solution of 2 mM 5-bromo-4-chloro-3-indolyl-b-D-glucuronide and 0.5 mM K 3 Fe(CN) 6 in buffer containing 100 mM Tris and 50 mM NaCl (pH 7.0). After 19 h of incubation at 37°C in the dark, the tissue was cleared in 96% ethanol (v/v). Samples were stored in lactic acid (Acros Organics) at 4°C until analysis with a binocular stereomicroscope (Zeiss).

CKX Enzyme Activity Tests
The CK oxidase/CK dehydrogenase enzyme activity was tested as described by Frébort et al. (2002). In short, this spectrophotometric method is based on Schiff base formation of the CKX reaction product with 4-aminophenol under acidic conditions. Total proteins were extracted by grinding 3 g of frozen plant material in CKX buffer (0.2 M Tris-HCl, 1 mM phenylmethylsulfonyl fluoride, 0.3% Triton X-100) and concentrating them over columns (with a cutoff value of 10,000 D; Millipore). The extract was added to tubes containing 300 mL 0.2 M imidazole, 60 mL 10 mM electron acceptor (2,6dichlorophenol-indophenol or 2,3-dimethoxy-5-methyl-1,4-benzoquinoneindophenol), and 30 mL of CKs (10 mM iP dissolved in dimethylsulfoxide). Reaction mixtures were incubated at 37°C overnight, after which reactions were stopped by adding 300 mL 40% (v/v) TCA and 200 mL 2% (v/v) p-aminophenol in 6% TCA. After centrifugation, the supernatants were transferred to quartz cuvettes and wavelength scanning was carried out (between 200 and 800 nm) with a DU-640B spectrophotometer (Beckman). The activity was calculated as described (Frébort et al., 2002). The data are the average of two independent experiments.

CK Analysis
CKs were extracted, purified, and analyzed as described previously (Å stot et al., 1998) ]iPMP. CKs were purified by passing the plant extract through an SPE-C18 cartridge (Bond Elut; Varian) followed by a cation exchanger (Oasis MCX; Waters). Samples were dried and propionylated with acetonitrile/N-methylimidazole/propionic anhydride (5:3:1 v/v/v), evaporated under vacuum, and stored at 220°C until further analysis. The propionylated samples were dissolved in 1 mL acetonitrile with 3% (v/v) formic acid, supplemented with 12 mL water with 3% (v/v) formic acid, and 10 mL injected for HPLC/mass spectrometry as described (Nordströ m et al., 2004). The data were processed by the MassLynx software (MicroMass). Three repeats were used to calculate the average CK levels.

GA Analysis
Frozen plant material (500 mg) was homogenized in liquid nitrogen with a mortar and pestle and extracted in 1 mL of 80% MeOH (0.02% [w/v] versity). After centrifugation (approximately 18,000g for 3 min), the supernatant was evaporated to dryness under vacuum. The residue was dissolved in 1% (v/v) acetic acid, loaded onto preequilibrated Isolute C8-EC cartridges (International Sorbent Technology), and eluted with 80% (v/v) MeOH. After evaporation to dryness, the samples were further purified on an Oasis MCX cartridge (Waters), methylated with ethereal diazomethane, and purified by HPLC. The fractions containing the GAs of interest were dried, trimethylsilylated in 10 mL of pyridine and 10 mL of bis-(trimethylsilyl)-trifluoroacetamide (11% trimethylchlorosilane), and quantified by gas chromatography-mass spectrometry in a selected reaction-monitoring mode with a JMS MStation (JEOL) as described previously (Moritz and Olsen, 1995). Two independent biological repeats were used to determine the average GA levels for the three tested conditions.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Whole-plant phenotype of Arabidopsis C24 plants infected with R. fascians and comparison with 35S:KNAT1 and 35S:STM-GR plants.