The fungal pathogen Ustilago maydis establishes a biotrophic relationship with its host plant maize (Zea mays). Hallmarks of the disease are large plant tumours in which fungal proliferation occurs. Previous studies suggested that classical defence pathways are not activated. Confocal microscopy, global expression profiling and metabolic profiling now shows that U. maydis is recognized early and triggers defence responses. Many of these early response genes are downregulated at later time points, whereas several genes associated with suppression of cell death are induced. The interplay between fungus and host involves changes in hormone signalling, induction of antioxidant and secondary metabolism, as well as the prevention of source leaf establishment. Our data provide novel insights into the complexity of a biotrophic interaction.
Fungal pathogens of plants have developed different strategies to cope with the plant environment. While necrotrophic fungi kill plant cells rapidly after infection to feed on the dead tissue, biotrophic fungi acquire nutrients from living plant tissue. The biotrophic relationship requires a highly specialized adaptation of the pathogen to the host plant. Hyphae of biotrophic fungi can grow intercellularly as well as intracellularly, thereby being ensheathed by the plasma membrane of the host cell. Many biotrophic fungal pathogens like rusts and powdery mildew form specialized feeding structures called haustoria (Hahn and Mendgen, 2001; Voegele et al., 2001). In these structures, a carbohydrate- and protein-containing interface is developed between the hyphal cell wall and the plant plasma membrane that facilitates the exchange of signals and nutrients between fungus and host (Hahn and Mendgen, 2001; Mendgen and Hahn, 2002; Perfect and Green, 2001).
Plants have developed multifaceted defence systems, many of which are induced only upon pathogen attack. These responses include induction of pathogenesis related (PR) genes, production of secondary metabolites as well as the reinforcement of cell walls. Associated with these responses are the production of reactive oxygen species (ROS) and the induction of localized cell death (the hypersensitive response, HR). Induction of the basal plant defence machinery occurs upon the recognition of conserved molecules which are commonly found in a variety of microbial species, but which are absent in the host. These pathogen-associated molecular patterns (PAMPs) include, for example, fungal chitin, β-glucans and ergosterol. Specific virulence factors of the pathogen can be recognized by corresponding R (resistance) genes of the host plant. Resistance-gene-mediated resistance is associated with the activation of a salicylic acid (SA)-dependent signalling pathway that leads to expression of defence-related genes like PR1, the production of ROS and programmed cell death. Other phytohormones involved in pathogen responses are ethylene (ET) and jasmonates (JA). For biotrophs, R-gene-mediated defence responses and SA signalling are thought to result in resistance by restricting fungal growth in infected cells that have undergone hypersensitive cell death. Conversely, programmed cell death supports the growth of necrotrophic pathogens. Thus, plant defence responses appear specifically adapted to the attacking pathogen, with SA-dependent defences acting against biotrophs, and JA- and ET-dependent responses acting against necrotrophs (Glazebrook, 2005; Greenberg and Yao, 2004; Jones and Dangl, 2006; O’Connell and Panstruga, 2006).
One of the best studied biotrophic pathogens is the smut fungus Ustilago maydis, that induces plant tumours on all aerial parts of its host plants maize (Zea mays) and teosinte (Euchlaena mexicana). The biotrophic stage of U. maydis is initiated after the fusion of two haploid sporidia that form the infectious dikaryotic hyphae. Upon formation of a specialized infection structure, the appressorium, penetration of the host cell is most likely facilitated by enzymes that degrade the plant cell wall. During the early infection stage invading hyphae are surrounded by the host plasma membrane. At later stages, hyphae grow both intra- and intercellularly. Tumour development is associated with both plant cell enlargement and increased cell divisions (Banuett and Herskowitz, 1996; Callow and Ling, 1973; Doehlemann et al., 2008); however, it is currently unknown how this is triggered by the fungus. Finally, within the tumour tissue the diploid teliospores are formed (Banuett, 1995; Feldbrügge et al., 2004; Martinez-Espinoza et al., 2002).
In a previous study, a set of 12 maize genes differentially regulated upon U. maydis infection were identified. From the nature of these genes it was inferred that U. maydis triggers a discrete plant defence programme and interferes with the differentiation of plant tissue (Basse, 2005). We have studied the responses of maize to U. maydis infection by extensive transcriptional and metabolic profiling. Using the MapMan tool (Thimm et al., 2004) specifically adapted to maize, we visualized expressional changes of distinct biological pathways. These studies have revealed a complex interplay of U. maydis with its host plant, resulting in massive changes in primary and secondary metabolism.
Ustilago maydis is recognized by its host during the early infection phase
To determine the most appropriate time points to perform the array analysis, we used confocal imaging to follow the course of infection using the U. maydis strain SG200 which is able to infect plants without a mating partner (Kämper et al., 2006). The different stages of disease progression chosen for our study are depicted schematically in Figure 1(a). Twelve hours post-infection (hpi) the majority of SG200 cells had formed filaments and appressoria. During this initial stage, we observed a small fraction of epidermal cells undergoing cell death, indicating that not all hyphae were able to establish a biotrophic interaction (Figure 1b). By 24 hpi hyphae were found in the epidermal cell layer. This intracellular stage persisted at 2 days post infection (dpi) but was now associated with hyphal branching. Fungal cell-to-cell passage was associated with swelling of the hyphal tip (Figure 1c) reminiscent of appressoria. Except for the growing tips, hyphae were surrounded by autofluorescent material accumulating at sites of cell-to-cell passage and around older parts of the hyphae (Figure 1c), suggesting the elicitation of plant defences. By 4 dpi tumour development had commenced and hyphae had colonized meristematic tissue, growing both intra- and intercellularly (Figure 1a, Doehlemann et al., 2008). In colonized areas, strong autofluorescence and occasional small areas with clusters of dead cells were detected (Figure 1d,e). By 8 dpi tumours had increased in size and contained clusters of sporogenic hyphae. At this time point, accumulation of anthocyanin was observed and autofluorescence persisted. Our analysis demonstrates that the infection by U. maydis elicits visible plant defence reactions throughout the various stages of biotrophic development. As the time points used in this microscopic study provided a comprehensive view of distinct steps in fungal development as well as the plant response, we have chosen the same stages for our subsequent studies.
Changes in host gene expression after U. maydis infection
For a comprehensive analysis of host cell responses, we have performed transcript profiling using the Affymetrix maize genome array. On this array, 13 339 genes are represented by 17 555 probe sets. Under our experimental conditions, significant signals were obtained for 70.8% of the probe sets in at least one of the experiments; 53.0% were present in all experiments. Plants were infected with SG200 and leaf samples were taken 12 and 24 hpi, as well as 2, 4 and 8 dpi in three independent experiments. Changes in gene expression were calculated relative to control plants inoculated with water. Statistical analysis revealed that 2891 genes were differentially regulated in at least one of the time points analysed. By 12 hpi, 208 genes were differentially regulated in infected leaves, of which 138 were functionally annotated. Of these, genes with a presumed function in stress response and redox regulation, including defence-related genes, were significantly enriched (Figure 2, Table S2). By 24 hpi, the number of genes with significantly altered transcript levels in infected leaves decreased to 116 (Figure 2a), which mostly reflects the variation around the threshold level of two-fold used as cut-off. Nevertheless, 37 of the genes that were induced 12 hpi were significantly downregulated at 24 hpi, and this downregulation was in most cases maintained at 2 dpi (Figure 3, and see below). The number of stress-associated genes remained significantly high at 24 hpi, and genes involved in redox regulation became less prominent. Furthermore, an increased number of genes related to secondary metabolism were differentially regulated in infected leaves at 24 hpi compared with 12 hpi (Figure 2b). At 2 dpi, the number of differentially regulated genes increased to 575, of which 358 could be functionally annotated (Figure 2a, Table S2). Besides genes of the functional categories ‘stress’ and ‘secondary metabolism’ that were already enriched during the earlier time points, genes associated with ‘cell wall metabolism’ were over-represented (Figure 2b). Furthermore, genes of the categories ‘protein metabolism’, ‘transcription and RNA processing’ and ‘transport’ were significantly enriched 2 dpi (Figure 2b), indicating the onset of a broad metabolic reprogramming in infected tissue. We also observed a slight, but not yet significant, enrichment of genes involved in photosynthesis. By 4 and 8 dpi, the number of differentially regulated genes increased to 1582 and 2420, respectively (Table S2). The number of defence-related genes as well as genes involved in redox regulation increased further (Figure 2b). The major changes concerned the functional categories ‘PS and C4’, ‘primary carbon metabolism’, ‘protein metabolism’ and ‘transcription and RNA-processing’. Transcriptional changes were visualized by the MapMan tool adapted to maize, which allows an assignment to cellular processes (Figure 4). In the following we will discuss the most prominent processes that are subject to differential regulation.
Plant defence responses to U. maydis infection
By 12 hpi, stress-related genes were significantly over-represented in infected plant tissue (Figure 2b). A rather unspecific defence reaction was elicited as several genes known to be induced by abiotic stresses like temperature, osmolarity and wounding were upregulated. However, the majority of induced genes encode PR-like proteins (van Loon et al., 2006; Tables S2 and S3) at a time point when fungal hyphae had just started to penetrate the epidermis. This shows that U. maydis cells were recognized. From a total of 184 PR-like genes identified on the array, 34 genes representing most PR gene classes were found to be induced (Table S3). Furthermore two leucine-rich repeat (LRR) genes were upregulated (Zm10830.1 and Zm8200.1) as well as Zm12900.1, encoding a protein similar to a receptor-like kinase induced after rust infection of wheat (Feuillet et al., 2003). The four most induced genes (200- to 700-fold) encode a terpene synthase (involved in isoprenoid synthesis; Zm14496.1), endochitnase B (PR3-like; Zm1595.1), a barwin-like protein (PR4-like; Zm2227.1) and a 1,3-beta-glucanase (PR2-like; Zm791.1) (Table S2). The Zm-mfs1 gene (major facilitator superfamily; Zm18344.1), which is induced upon infection with the necrotrophic pathogen Cochliobolus carbonum (Simmons et al., 2003) was 26-fold induced by 12 hpi as well. Zm-mfs2 (Zm12717.1) expression was not altered upon C. carbonum infection or upon infection by U. maydis. The an2 gene (ZmAffx.12.1), which is upregulated after infection with Fusarium graminearum (Harris et al., 2005), was also induced by U. maydis 12 hpi, while an1 (Zm228.1) expression was not altered. The corresponding rice orthologues encode ent-copalyl diphosphate synthases. an1 is involved in Gibberellic acid (GA) synthesis and an2 is required for phytoalexin synthesis (Prisic et al., 2004). Both Zm-mfs1 and an2 expression decreased significantly by 24 hpi. Further inspection of the genes induced at 12 hpi revealed that 37 of these genes were downregulated at least two-fold in infected tissue at the 24-h time point. Most of these were defence-related genes, including three of the four most highly induced transcripts (Figure 3). This indicates that the primary plant response is attenuated when U. maydis starts colonizing epidermal cells.
Interestingly, by 12 hpi genes for Bax-inhibitor 1 (Zm12293.1) and a cystatin (Zm8113.1), both involved in cell death suppression, were induced (Table S4). This induction persisted at 2–8 dpi, and additionally two more cystatin genes (Zm14795.1; Zm14272.5) were induced in infected tissue. Conversely, one of the two metacaspases present on the array (Zm18453) was significantly downregulated at these time points (Table S4). Taken together, these data suggest that U. maydis infection is accompanied by an inhibition of the plant cell death programme.
Ustilago maydis induced changes in hormone signalling
Of 25 genes annotated as being involved in JA biosynthesis and responses, nine were significantly induced already by 12 hpi, and expression levels were maintained further for most of the genes (Figure 5). Upregulation of JA signalling is usually associated with an induction of plant defence genes like defensins, hevein-like proteins and chitinases (Glazebrook et al., 2003; Penninckx et al., 1998; Thomma et al., 1998). Consistently, the maize orthologues of such genes were upregulated after U. maydis infection (Table S3). In contrast, PR1 (Zm15280.1), one of the prime marker genes in SA-signalling and described to be induced by both necrotrophic and biotrophic pathogens (Morris et al., 1998), was not induced during the early U. maydis infection stages. Furthermore, a germin-like protein (Zm12518.1), which in Arabidopsis thaliana is SA-induced but repressed in response to methyl-jasmonate (Schenk et al., 2000), was transiently induced 12 hpi and subsequently repressed (Figure 3). Interestingly, three genes involved in auxin biosynthesis and 19 auxin responsive genes were induced up to 44-fold 4 and 8 dpi (Figure 5, Table S5). At these time points, which coincide with the onset of massive cell division and enlargement, genes coding for gibberellin biosynthesis enzymes as well as GA-responsive genes were also induced (Figure 5).
Changes in antioxidant levels and secondary metabolite synthesis during U. maydis infection
Abiotic and biotic stress coincides with changes in oxidation state and content of soluble antioxidants, with glutathione (GSH/GSSG) being the most sensitive component (Ogawa, 2005). Seven glutathione S-transferase (GST) genes were already induced by 12 hpi (Table S2), including gst15 (Zm545.1), a homologue to wheat gst1a that is induced upon pathogen attack (Dudler et al., 1991). To substantiate the transcriptome data, the contents of soluble (glutathione and ascorbic acid) and membrane-bound antioxidants (tocopherols) were determined from the same material used for array analysis. Contents of tocopherol and ascorbic acid did not change during the infection process (data not shown). Elevated GSH levels were observed 24 hpi and increased further during the infection process, while a high reduction state of the glutathione pool was maintained (Figure 6a,b). Increased levels of GSH have also been shown to coincide with the induction of PR genes in A. thaliana (Senda and Ogawa, 2004). Additionally, GSH plays a major role in secondary metabolite synthesis, mainly by regulating the key enzymes phenylalanin ammonium lyase (PAL) and chalcone synthase (CHS) (Gomez et al., 2004; Loyall et al., 2000). Consistently, both PAL enzyme activity and transcript level were strongly increased in tumour tissue 8 dpi (Figure 7a,e). The substrates for PAL, phenylalanine and tyrosine, accumulated about four- and five-fold, respectively, 8 dpi in infected tissue (Figure 7b,c). Accumulation of these two amino acids was significantly higher than the average increase of most other amino acids, which was in the range of two- to three-fold (data not shown). Consistently, many genes from the shikimate pathway were found to be induced in infected tissue starting at 2 dpi (Figure 7a). Shikimate, an abundant key metabolite upstream of phenylalanine and tyrosine, increased about eight-fold in tumour tissue compared with uninfected leaves 4 and 8 dpi (Figure 7d).
The major phenylpropanoid products downstream of PAL are hydroxycinnamic acid (HCA) derivatives predominantly serving as building blocks of lignin and flavonoids, which represent potential phytoalexins, anthocyanins and UV protectants. Anthocyanins accumulated strongly in infected maize tissue (Figure 7f), which was also evident by visual inspection of infected tissue. Similarly, the overall content of HCA derivatives was increased in tumour tissue (Figure 7g), while the total content of flavonoids was not changed significantly 8 dpi [infected tissue 114.45 ± 9.81 A310 g−1 fresh weight (FW); control tissue 121.2 ± 6.9 A310 g−1 FW) 8 dpi. The gene for leucoanthocyanidin dioxygenase (Zm62.1), a key enzyme for anthocyanin synthesis, was induced about 15-fold at late infection stages (Table S2). Induction of genes involved in lignin biosynthesis was observed already 12 hpi and increased further during the infection process. In tumour tissue (8 dpi) transcript levels for genes involved in almost all steps of lignin biosynthesis were significantly induced (Figure S1). The enhanced synthesis of phenolic compounds, such as lignin, flavonoids and phenylpropanoids (Figure 7a), is reflected by the enhanced cell wall autofluorescence observed in U. maydis-infected tissue (Figure 1c–e).
Changes in plant primary metabolism
The experimental conditions used in our experiments resulted predominantly in the infection of the third leaf which was initially still contained within the leaf whorl. One to two days later, when exposed to light, the onset of photosynthesis was expected to lead to a decrease in free hexose content, which in sink tissues is known to originate from cleavage of imported sucrose (Horst et al., 2008).
To determine the influence of U. maydis on the sink-to-source transition, we measured free hexose and sucrose contents in infected and non-infected leaves (Table 1). Hexose content in infected and non-infected leaves decreased 4 dpi, reflecting the onset of photosynthetic activity. Whereas hexose levels remained low in control leaves after 8 days, they increased >20-fold in infected leaves at this time point (Table 1); sucrose contents were not altered by U. maydis infection. Consequently, this resulted in an increased hexose/sucrose ratio in tumour tissue (Table 1).
Table 1. Carbohydrate contents in Ustilago maydis infected and uninfected leaves
dpi, days post-infection.
*Concentrations are given in μmol g−1 fresh weight (FW) for sugars and μmol Glc units g−1 FW for starch. Standard errors are indicated.
25.1 ± 0.7
23.2 ± 0.5
1.8 ± 0.3
5.0 ± 0.4
1.1 ± 0.5
26.6 ± 3.0
27.7 ± 1.5
23.2 ± 0.7
24.5 ± 2.0
30.4 ± 2.0
22.6 ± 1.1
29.6 ± 2.7
0.9 ± 0.04
1.01 ± 0.04
0.07 ± 0.01
0.16 ± 0.01
0.05 ± 0.02
0.94 ± 0.11
3.3 ± 0.4
6.95 ± 0.9
75.8 ± 4.9
113.6 ± 7.6
63.3 ± 6.5
69.2 ± 6.1
To follow the influence of U. maydis on normal leaf development, we determined transcriptional alterations in infected and non-infected leaves at five consecutive time points. In non-infected leaves the most dramatic changes (1678 genes) occurred at the onset of photosynthesis (equivalent to 1 and 2 dpi), affecting mostly genes involved in protein and RNA synthesis, primary metabolism and photosynthesis (Figures 2 and 8, Table S6, Figure S2a). In comparison, expression of only 376 genes was changed in infected leaves (Table S6, Figure S2b). Notably, induction of photosynthesis-associated genes was strongly reduced in infected compared with uninfected leaves (Figure S2a,b), indicating that normal development from sink to source tissue is impaired. At the same time, we observed significant changes in energy metabolism; both glycolysis and lipid metabolism were induced in tumours (Figure 4).
To elucidate which of the observed effects resulted from the absence of photosynthesis and which were caused by an active metabolic reprogramming induced by U. maydis, we compared transcript profiles of infected, uninfected and masked leaves (shielded from light for 6 days) of the same age (8 dpi).
Induction of photosynthetic genes was absent in masked leaves when compared with uninfected leaves. Similarly, primary metabolism was downregulated, with the exception of sucrose degradation that was found to be induced (Table S7, Figure 9). A comparable induction of sucrose degradation was observed in infected leaves (Figure 9), which in addition showed transcriptional induction of glycolysis and the tricarboxylic acid cycle (TCA; Figure 9, Table S2). Induction of genes for hexose degradation was significantly less pronounced in masked leaves compared with tumour tissue, indicating higher energy consumption in the infected cells (Figure 9). Comparison of genes differentially regulated in masked leaves with genes differentially regulated in tumour tissue identified 958 genes as exclusively regulated upon U. maydis infection (Table S7). Although the functional distribution within this group did not change dramatically when compared with all differentially expressed genes (Figure 2a), a significant enrichment of defence-related genes was observed among the genes induced more than 100-fold at 8 dpi (34 of 67; Table S7).
We have analysed changes in the maize transcriptome and metabolic changes in response to infection with the biotrophic pathogen U. maydis. We have focused on changes in distinct cellular processes, which were visualized by a novel version of the MapMan tool adapted to maize. We infer from our data that U. maydis is initially recognized and elicits plant defence reactions. With establishment of the biotrophic interaction, these initial responses are attenuated. Additionally, our data indicate that U. maydis interferes with normal leaf development and prevents the transition from sink to source leaves.
Ustilago maydis induced defence responses and cell death suppression in maize
The transient upregulation of defence-associated genes in infected tissue suggests that U. maydis is recognized by the plant via conserved molecular patterns. The currently known PAMP receptors are LRR receptor kinases and receptor-like proteins with an extracellular LRR domain lacking a kinase domain. It has recently been shown that these PAMP receptors are transcriptionally upregulated after elicitation (Zipfel et al., 2004, 2006). During the early phase of host colonization by U. maydis, we observed the upregulation of two putative membrane-bound LRR-like receptor kinases. The orthologue of one of them (Zm10830.1) is induced in Sorghum bicolor after infection with the hemibiotrophic fungus Colletotrichum graminicola (Hipskind et al., 1996). The second gene (Zm8200.1) encodes a SER kinase and thus belongs to a group that includes BAK1, which has recently been shown to act as a positive regulator in infection-induced cell death signalling (Chinchilla et al., 2007; Kemmerling et al., 2007). We also observed an induction of Zm12900.1, encoding a protein similar to a receptor-like kinase from wheat, which is induced after infection with the rust fungus Puccinia triticina (Feuillet et al., 2003). In analogy to these pathosystems, it is conceivable that the identified maize genes are involved in PAMP perception.
With the onset of biotrophy 24 hpi, defence responses were attenuated, and concomitantly we observed a transcriptional induction of JA signalling components. The observed induction of genes encoding cell death suppressors such as cystatins (Belenghi et al., 2003; Solomon et al., 1999) and Bax-inhibitor 1 (Eichmann et al., 2004) as well as the repression of caspases suggest that U. maydis interferes with the regulation of cell death.
The phytopathogenic bacterium Pseudomonas syringae interferes with programmed cell death by translocation of the effector protein AvrPtoB into host cells (Rosebrock et al., 2007). Similarly, several fungal Avr proteins have been shown to interact with their cognate R gene products in the cytoplasm of host cells (Dodds et al., 2006; Jia et al., 2000). Whether a similar mechanism is employed by U. maydis remains to be shown. In U. maydis novel secreted effectors have recently been shown to be important for virulence (Kämper et al., 2006). However, it is currently unknown whether these effectors target apoplastic or cytoplasmic plant targets.
Ustilago maydis-induced changes in hormone signalling
Plant hormone signalling is dramatically changed in response to pathogen attack. In compatible interactions with necrotrophic pathogens, JA signalling plays a minor role, and instead, SA-dependent cell death responses and the expression of a large set of defence genes including PR1 are observed (Seo et al., 2001). Biotrophic pathogens, on the other hand, induce JA and ethylene responses during compatible interactions. These responses do not lead to cell death and are associated with induction of tryptophan biosynthesis, the accumulation of secondary metabolites and the induction of plant genes encoding defensins (Brader et al., 2001; Glazebrook, 2005; Wasternack, 2007). Consistently, after infection with U. maydis, PR1 expression was undetectable at early time points. At later time points, low expression of PR1 was detected, which likely reflects a mixed response caused by a small fraction of infected plant cells undergoing necrosis. Induction of JA signalling which antagonizes the SA pathway (Glazebrook, 2005) is detected immediately after infection. At the same time, activation of typical JA-responsive defence genes such as defensins, hevein-like proteins and chitinases is observed (Glazebrook et al., 2003; Penninckx et al., 1998; Thomma et al., 1998). Jasmonate synthesis does not depend on the expression level of its biosynthetic genes, but on the substrate availability of stored precursors (Wasternack, 2007). In line with this, we do not observe an induction of Zm13677.1, a homologue to the OPR7 gene from rice that has been shown to be essential for JA synthesis (Tani et al., 2008).
Ustilago maydis-induced tumours contain elevated levels of auxin (Turian and Hamilton, 1960). Recently, it has been shown that auxin produced by U. maydis is unlikely to be important for tumour formation (Reinecke et al., 2008). We now demonstrate transcriptional induction of both auxin synthesis and auxin-responsive genes during tumour development, suggesting that the cell enlargement observed in U. maydis-induced tumours is caused by elevated levels of auxin produced by the plant. Recent studies in A. thaliana demonstrated repression of auxin signalling by SA (Wang et al., 2007). The SA-mediated repression of auxin levels leads to plant resistance, while inhibition of SA signalling allows auxin signalling, which, in turn, would promote fungal growth and host susceptibility. This is in agreement with the minor role of SA signalling in the maize/U. maydis interaction.
Antioxidants and secondary metabolites
The induction of GSTs by pathogens has been shown previously (Greenberg et al., 1994; Hahn and Strittmatter, 1994; Levine et al., 1994; Marrs, 1996) and has been reported to occur very rapidly, preceding the induction of PR genes (Alvarez et al., 1998; Mauch and Dudler, 1993). Seven GST genes were already induced 12 hpi. The antioxidative activity of GSTs has been proposed to reduce damage caused by pathogens, and to restrict cell death during HR (Mauch and Dudler, 1993). It is conceivable that some of the induced GSTs could be involved in scavenging oxygen radicals which also result from respiratory processes of the plant cell. Because it is not possible to implicate function or substrate specificity from the primary protein sequence (Wagner et al., 2002), the specific functions of the GSTs regulated during the infection process has to remain speculative.
Although genes involved in glutathione synthesis [glutathione synthetase, gsh1 (Zm3618.2) and glutamate-cysteine ligase, gsh2 (Zm9043.1)] were not significantly regulated, glutathione content was increased throughout infection. This could reflect the requirement for a higher antioxidative capacity in infected tissue once the integrity of the photosynthetic apparatus is impaired in tumours. In addition, the enhanced glutathione levels could serve as a signal for defence gene induction, as has been described for A. thaliana (Senda and Ogawa, 2004).
Our data revealed that genes involved in secondary metabolism are significantly enriched at all time points analysed. In particular, we observed an induction of genes of the shikimate pathway, as well as a 20-fold increase in PAL activity and an accumulation of the primary pathway products phenylalanine and tyrosine. Phenylalanin ammonium lyase catalyses the committed step in the biosynthesis of phenolic secondary metabolites of the phenylpropanoid class (comprising HCA derivatives, lignans and flavonoids). In accordance, we detected an accumulation of HCA derivatives and anthocyanins. The HCA derivatives can serve as building blocks for lignin biosynthesis, and genes involved in lignin and lignan synthesis were also consistently significantly induced during the late infection stages of U. maydis. This induction is likely to reflect the increased cell wall synthesis resulting from enhanced cell division and cell expansion within tumour tissue. In addition, U. maydis-infected cells show enhanced cell wall epifluorescence, indicative of the deposition of lignin and/or other phenolic compounds. Reinforcement of cell walls by phenolics has been shown to be part of a defence reaction against pathogens (Bruce and West, 1989; Egea et al., 2001; Huang and Hartman, 1998; Lange et al., 1995; Nicholson and Hammerschmidt, 1992). We hypothesize that this late response is also part of the mixed response discussed above where some cells elicit defence reactions while others do not.
In parallel to HCAs, we observe an accumulation of anthocyanins in tissues infected with U. maydis, coinciding with the induction of leucoanthocyanidin dioxygenase gene expression. Anthocyanin accumulation is part of the response towards a variety of biotic and abiotic stress situations such as pathogen attack, waterlogging, high light, salinity or cold stress (Chalker-Scott, 1999). Considering that U. maydis as a biotroph does not get in direct contact with the anthocyanins localized in the vacuole, it is likely that the accumulation is an indirect stress response caused by the fungus.
The induction of phenylpropanoid biosynthesis at late infection stages might reflect the increasing activity of the SA pathway at these time points and it is possible that some not yet described phenolic phytoalexins are produced along with the abundant phenolics. For instance, the hydroxamic acid DIMBOA, a derivative of the aromatic amino acid tryptophan and one of the known defence compounds in maize, has been described to be induced upon U. maydis infection, consistent with an increased expression of the gene for the initial biosynthesis step, Bx1 (Basse, 2005).
Likewise, isoprenoids serve as phytoalexins, which exhibit antimicrobial properties and are synthesized in response to pathogen attack (VanEtten et al., 1994). Maize probably has a much smaller set of phytoalexins compared with other cereals such as rice (Walton, 2001). Mining of public databases did not reveal any evidence for the key enzymes for synthesis of polycyclic diterpenes, which is the major group of phytoalexins in rice. Nevertheless, induction of an ortholog of the an2 gene indicates that U. maydis triggers phytoalexin synthesis throughout the entire infection process.
Photosynthesis and primary metabolism
Ustilago maydis is known to infect young meristematic maize tissue, but is unable to infect differentiated source leaves (Wenzler and Meins, 1987). Between 1 and 2 dpi, we observed a global induction of genes involved in light reaction, Calvin cycle, photorespiration, tetrapyrrole synthesis as well as sucrose and starch synthesis in uninfected leaves, which was not observed in infected leaves of the same age. This indicates that the transition from a juvenile sink tissue to a mature, photosynthetically active source tissue is blocked in infected leaves. This block is consistent with the recently described observation that U. maydis-infected leaves are not able to establish C4 metabolism but continue to perform C3 photosynthesis usually only observed in immature maize leaves (Horst et al., 2008). In addition, U. maydis infection is associated with pronounced chlorosis, and, concomitantly, with a decline in chlorophyll content and reduced rates of CO2 assimilation in infected leaf tissue (Horst et al., 2008).
In infected leaves, about 60% of the differentially expressed genes can be attributed to the downregulation of the photosynthetic apparatus, as shown by comparison with masked leaves. Induction of sucrose degradation and reduction of sucrose synthesis was observed in infected and masked tissues, indicating that both rely on import of sucrose from photosynthetically active source tissues. Alterations specific for infected tissue comprise the induction of glycolysis and the TCA cycle, which might indicate either an elevated flux of carbon skeletons into amino acid biosynthesis or increased respiration of infected maize cells.
The increase in free hexose content found in tumours is typical for a sink tissue. It is likely that the developing tumour itself generates a sink through active proliferation. The free hexoses within tumour cells then could be used by U. maydis as an easily accessible carbon source. This strategy is in strict contrast to powdery mildew infections, where leaves remain as source tissues (Scholes et al., 1994; Walters and McRoberts, 2006). In this case, the fungus aquires nutrients via haustoria from single epidermal cells, leaving the physiological state of the entire leaf largely unaltered. Based on the presented results, the key questions are ‘How are nutrients partitioned between U. maydis and host cells?’ and ‘To what extent the host metabolism is actively reprogrammed by fungal effectors?’. For a more comprehensive understanding of these processes, detailed analysis of single U. maydis-infected cells instead of complex infected tissues will be one of the major challenges for future research.
Ustilago maydis hyphae were stained with WGA-AF 488 (Molecular Probes, Invitrogen, http://www.invitrogen.com/). Plant membranes were visualized using FM4-64 (Invitrogen). Samples were incubated in staining solution (4 μg ml−1 FM4-64, 10 μg ml−1 WGA-AF 488; 0.02% Tween20) for 30 min and washed in 1× PBS (pH 7.4). Confocal images were recorded on a TCS-SP5 confocal microscope (Leica, http://www.leica.com/). FM4-64 excitation was at 561 nm and detection at 600–700 nm; WGA-AF 488 excitation at 488 nm and detection at 500–540 nm. Autofluorescence of cell wall material was excited at 405 nm and detected at 415–460 nm. For RFP fluorescence of hyphae in maize tissue, an excitation of 561 nm and detection at 580–630 nm was used.
Plant material and RNA preparation
For U. maydis infections, maize plants (Early Golden Bantam) were grown in a phytochamber in a 15-h/9-h light/dark cycle; light period started/ended with 1 h ramping of light intensity. Temperature was 28 and 20°C, relative humidity 40% and 60% during light and dark periods, respectively, with 1 h ramping for both values. Plantlets were individually sown in pots with potting soil (Fruhstorfer Pikiererde, http://www.hawita.de) and infected 7 days after sowing 1 h before the end of the light period, as described (Brachmann et al., 2001); plants for 12 hpi samples were infected during the beginning of the light period. For three independently conducted experiments (biological replicates), samples used for both RNA preparation and metabolite measurements were collected 1 h before the end of the light period and directly frozen in liquid nitrogen. For each experiment, 30 plants were sampled and divided for the 12–48 hpi samples into three subsets, and for all other samples into four subsets. Metabolite analysis was carried out independently for all these subsets, i.e. technical replicates. For RNA isolation, material from the subsets was pooled and ground in liquid nitrogen. The RNA was extracted with TRIzol (Invitrogen) and purified using an RNeasy kit (Qiagen, http://www1.qiagen.com/).
DNA microarray and verification by quantitative realtime PCR
Affymetrix Gene chip® maize genome arrays were done in three biological replicates, using standard Affymetrix protocols (Midi_Euk2V3 protocol on GeneChip Fluidics Station 400; scanning on Affymetrix GSC3000). Expression data were submitted to the GeneExpressionOmnibus (http://www.ncbi.nlm.nih.gov/geo/), accession number GSE10023. Data analysis was performed using Affymetrix Micro Array Suite 5.1, bioconductor (http://www.bioconductor.org/) and dChip1.3 (http://biosun1.harvard.edu/complab/dchip/), as described (Eichhorn et al., 2006). We considered changes greater than twofold with a difference between expression values >100 and a corrected P-value <0.001 as significant. For pathway analysis we used the MapMan tool optimized for maize. Expression changes in MapMan pathways were filtered by a P-value <0.001.
To verify microarray results, selected genes were analysed by quantitative (q)RT-PCR (Table S9). For cDNA synthesis, the SuperScript III first-strand synthesis SuperMix assay (Invitrogen) was employed, using 1 μg of total RNA. Quantitative RT-PCR was performed on a Bio-Rad iCycler (http://www.bio-rad.com/) using the Platinum SYBR Green qPCR SuperMix-UDG (Invitrogen). Cycling conditions were 2 min 95°C, followed by 45 cycles of 30 sec 95°C/30 sec 61°C/30 sec 72°C. Primer sequences are listed in Table S10.
Amino acids were extracted in 80% ethanol and determined by HPLC after derivatization with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate as described in Cohen and Michaud (1993). Soluble carbohydrates and starch were determined from the same extract by coupled enzymatic assays as described (Bergmeyer, 1970; Voll et al., 2003). Small thiols were extracted with 0.1 m HCl and quantified after derivatization with monobromobimane by HPLC [Summit Series, Dionex Corp. (http://www.dionex.com/) equipped with a Luna 5u C18(2) column (Phenomenex, http://www.phenomenex.com/) and a Dionex RF2000 fluorescence detector]. Separation of thiols was achieved with an isocratic mixture of 89% 100 mm potassium acetate, pH 5.5 and 11% methanol; detection was performed at 380 nm after excitation at 480 nm. Shikimate was extracted with perchloric acid (Hausler et al., 2000) and isolated by ion exchange chromatography using a Dionex IonPac AG11-HC (2 × 50 mm) pre-column and two Dionex IonPac AS11-HC (2 × 250 mm) columns on a Dionex ICS3000 system using a KOH step gradient (0–100 mm). The eluate was passed on to a mass spectrometer (API3200 Q-trap tandem MS, Applied Biosystems, http://www.appliedbiosystems.com/) and shikimate was detected and quantified by the transition m/z =173 to m/z =93 in the negative ion mode relative to standards.
To determine the total content of phenolics, A534, A310 and A290 were measured spectrophotometrically in perchloric acid extracts (see above) and in 80% methanol extracts of the perchloric acid insoluble material to account for water-soluble and water-insoluble derivatives, respectively. The absorptions of both extracts were normalized to fresh weight and added to yield total content of phenolics.
Leaf material was homogenized in extraction buffer (100 mm borate, pH 8.8, 0.1 mm PefaBloc, 5 mmβ-mercaptoethanol, insoluble polyvinylpyrrolidone) and incubated for 30 min on ice. After sonication for 30 sec, the soluble fraction was desalted using Sephadex G25 columns, and the protein content was determined (Zor and Selinger, 1996) for normalization of activity. The PAL activity was assayed in a buffer containing 240 μl of a 100 mm l-phenylalanine solution, 80 μl extract and 80 μl 0.2 m borate buffer, pH 8.8. The increase in released trans-cinnamic acid was monitored for 1 h at 290 nm and quantified using a standard curve.
We are grateful to the Bioanalytics Group, Department of Biochemistry, FAU Erlangen-Nuremberg for the assistance in ion chromatography-mass spectometry analysis. We thank Dr Olga Levai (Leica Microsystems GmbH, Wetzlar, Germany) for expert technical advice with confocal microscopy. This work was supported by the DFG priority programme FOR666.