The use of FLP-mediated recombination for the functional analysis of an effector gene family in the biotrophic smut fungus Ustilago maydis


  • Yuliya Khrunyk,

    1. Department of Organismic Interactions, Max Planck Institute for Terrestrial Microbiology, Karl-von-Frisch-Strasse 10, D-35043 Marburg, Germany
    Search for more papers by this author
  • Karin Münch,

    1. Department of Organismic Interactions, Max Planck Institute for Terrestrial Microbiology, Karl-von-Frisch-Strasse 10, D-35043 Marburg, Germany
    Search for more papers by this author
  • Kerstin Schipper,

    1. Department of Organismic Interactions, Max Planck Institute for Terrestrial Microbiology, Karl-von-Frisch-Strasse 10, D-35043 Marburg, Germany
    Search for more papers by this author
  • Andrei N. Lupas,

    1. Department of Protein Evolution, Max Planck Institute for Developmental Biology, Spemannstr. 35, D-72076 Tübingen, Germany
    Search for more papers by this author
  • Regine Kahmann

    1. Department of Organismic Interactions, Max Planck Institute for Terrestrial Microbiology, Karl-von-Frisch-Strasse 10, D-35043 Marburg, Germany
    Search for more papers by this author

Author for correspondence:
R. Kahmann
Tel: +49 6421 178 501


  • In the Ustilago maydis genome, several novel secreted effector proteins are encoded by gene families. Because of the limited number of selectable markers, the ability to carry out sequential gene deletions has limited the analysis of effector gene families that may have redundant functions.
  • Here, we established an inducible FLP-mediated recombination system in U. maydis that allows repeated rounds of gene deletion using a single selectable marker (HygR). To avoid genome rearrangements via FRT sites remaining in the genome after excision, different mutated FRT sites were introduced.
  • The FLP-mediated selectable marker-removal technique was successfully applied to delete a family of 11 effector genes (eff1) using five sequential rounds of recombination. We showed that expression of all 11 genes is up-regulated during the biotrophic phase. Strains carrying deletions of 9 or all 11 genes showed a significant reduction in virulence, and this phenotype could be partially complemented by the introduction of different members from the gene family, demonstrating redundancy.
  • The establishment of the FLP/FRT system in a plant pathogenic fungus paves the way for analyzing multigene families with redundant functions.


Plant–pathogen interactions involve secreted pathogen effectors. Such effectors can trigger virulence by suppressing pathogen-associated molecular pattern (PAMP)-triggered immunity. Alternatively, effectors can induce effector-triggered immunity (ETI) in plants that host cognate resistance proteins (Jones & Dangl, 2006). As a result of gene duplications and functional redundancy caused by an arms race between PAMP and ETI receptors in plants and effectors in pathogens, many effectors are individually dispensable (Angot et al., 2006; Birch et al., 2009; Kvitko et al., 2009; Stergiopoulos & de Wit, 2009). Some effectors might also possess overlapping activities, which may be the case for the Avr4 and Ecp6 effectors of Cladosporium fulvum, both of which bind to chitin and thus could potentially protect the fungus against plant chitinases and/or suppress plant defense responses by scavenging chitin fragments released from fungal cell walls in the apoplast during infection (Bolton et al., 2008).

Ustilago maydis is the causative agent of corn smut disease and has become one of the models used to study biotrophic interactions. In U. maydis, 426 genes code for putatively secreted proteins, and, of these, 272 encode either U. maydis-specific proteins or conserved proteins without recognized InterPro domains (Müller et al., 2008). Many of the novel secreted proteins are encoded by gene clusters, and the respective genes are induced in infected tissue during biotrophic growth. Deletions of several effector clusters, but also of a single effector gene, have dramatic effects on virulence (Kämper et al., 2006; Doehlemann et al., 2009). Among the novel secreted U. maydis proteins, some are encoded by gene families (Kämper et al., 2006). One of these U. maydis-specific gene families is family 9, which consists of four genes, and another, designated family 17, contains three genes (Kämper et al., 2006). Neither of these gene families has been functionally characterized because this would require the generation of multiple gene-deletion mutants. Although tools (such as a PCR-based system) for the generation of gene-replacement mutants in U. maydis is in place, only a limited number of dominant drug-resistance markers, such as hygromycin (Wang et al., 1988), carboxin (Keon et al., 1991), phleomycin and nourseothricin (Gold et al., 1994) have been developed. As deletions are usually made in the engineered solopathogenic strain SG200 (Kämper et al., 2006), which is already phleomycin resistant, and the link between phenotype and deletion of a particular gene needs to be established by complementation, this reduces the number of available dominant markers further. Such obstacles can be overcome by recycling the resistance marker, which can be achieved via site-specific recombination (Bucholtz, 2008; Birling et al., 2009).

Site-specific recombinases, such as Cre and FLP, catalyze efficient recombination between two directly oriented recombination sites, termed LoxP and FRT, respectively. This leads to excision of the intervening DNA segment, leaving one recombination site behind. If the selectable marker is placed on the intervening DNA segment, this allows the generation of unmarked gene disruptions and the use of the same selectable marker in subsequent rounds (Wirth et al., 2007). To date, such selectable marker-removal systems have been successfully established in different organisms, such as Saccharomyces cerevisiae (Storici et al., 1999), Candida albicans (Morschhauser et al., 1999), Aspergillus nidulans (Forment et al., 2006), Cryptococcus neoformans (Patel et al., 2010), maize (Kerbach et al., 2005) and mouse (Wu et al., 2008). The 45 kDa FLP recombinase from S. cerevisiae is the best characterized eukaryotic member of tyrosine recombinases. A minimal, fully functional FRT site is 34 bp long and consists of an 8 bp asymmetric spacer that defines the orientation of the FRT site, which is flanked by two, 13-bp-palindromic sequences that constitute the binding sites for FLP (Chen & Rice, 2003). As FLP-mediated recombination does not require any accessory host proteins, the FLP/FRT system has been employed in a wide range of species, ranging from prokaryotes to mammals (Luo & Kausch, 2002; Schweizer, 2003). Another advantage of the FLP/FRT system is that FLP exhibits optimal activity at 30°C (Buchholz et al., 1996), a temperature at which U. maydis can be cultivated under laboratory conditions. As a result of the possible recombination between the FRT sites left in the genome after each round of FLP-mediated marker removal there is a potential problem of undesired chromosome rearrangements. However, this can be circumvented by introducing FRT sites with different point mutations in the core region, which restricts recombination to identical FRT sites (Storici et al., 1999; Barrett et al., 2008).

Here we describe establishment of the FLP/FRT system in U. maydis. As proof of principle we have successively disrupted 11 genes encoding related secreted effectors. We designated this group of 11 genes as the eff1 gene family. This family combines the previously defined families 9 and 17 (Kämper et al., 2006), plus four additional genes, all encoding U. maydis-specific secreted proteins. We show that, depending on the genes deleted, virulence can be significantly compromised when two or more genes are deleted. The virulence phenotype of the nine-gene-deletion mutant could be complemented by introducing different members of the gene family, suggesting redundancy.

Materials and Methods

Strains, plasmids, growth conditions and molecular techniques

For cloning purposes the Escherichia coli strains Top10 (Invitrogen) and DH5α (Gibco BRL, Eggenstein, Germany) were used and grown in yeast extract, tryptone (YT) medium. The solopathogenic U. maydis strain SG200 (a1mfa2bE1bW2) has been described previously (Kämper et al., 2006). Strains were grown on potato dextrose (PD) (Difco, Becton Dickinson, Heidelberg, Germany), complete medium (CM)-glu or yeast extract peptone sucrose light (YEPSL) liquid medium (Holliday, 1974; Molina & Kahmann, 2007) or the respective solid media containing 2% agar in addition. To select transformants, hygromycin B (Duchefa, Haarlem, the Netherlands) and carboxin (Riedel de Haen, Seelze, Germany) were added to a final concentration of 200 μg ml−1 and 2 μg ml−1, respectively. For induction of the crg1 promoter, cells were grown in CM-glu medium to an OD600 of 0.7, washed twice with distilled water, resuspended in CM medium containing 1% arabinose as the sole carbon source (CM-ara) and incubated for the times indicated. To delete the hygromycin-resistance marker flanked by FRT sites, strains were transformed with the FLP-expressing plasmid pFLPexpC. A carboxin-resistant colony was grown overnight in CM-ara medium and plated on PD to obtain single colonies. Colonies were replica-plated on PD plates and PD plates supplemented with hygromycin (PD-hyg). Colonies that had lost resistance to hygromycin were subsequently tested on PD plates supplemented with carboxin to test for the loss of pFLPexpC and were analyzed by PCR to verify deletion events.

To generate deletion mutants, the single-step method developed by Kamper (2004) was adopted. The strategy for generating the codon-optimized FLP-recombinase gene was similar to that used by Stemmer et al. (1995) and Hale & Thompson (1998). For integrating genes into the ip locus, the strategy of Loubradou et al. (2001) was used. Plasmids pBScbx, pRU11, pNEBUH and p123 have been described previously (Keon et al., 1991; Brachmann et al., 2001; Weinzierl et al., 2002; Aichinger et al., 2003). Details for all plasmids and strain constructions can be found in the Supporting Information Methods S1.

Cloning procedures followed standard molecular techniques (Sambrook et al., 1989). All primers used in this study are listed in Table S1. Transformation of U. maydis followed the protocol of Schulz et al. (1990). DNA was isolated as described previously (Hoffman & Winston, 1987). Details for additional molecular techniques are described in Methods S2.

Pathogenicity assays

Pathogenicity assays were performed as described previously (Kämper et al., 2006). For maize (Zea mays) infections, cultures of U. maydis strains indicated in the respective experiments were grown to an OD600 of 0.7–0.8 in YEPSL, pelleted, resuspended in distilled water to an OD600 of 1 and injected into 7-d-old seedlings of the variety Early Golden Bantam (Olds Seeds, Madison, WI, USA). Plants were kept in the glasshouse with a light–dark cycle of 16 h (28°C) and 8 h (20°C). Disease symptoms were scored, according to severity, 14 d after inoculation (Kämper et al., 2006).

FLP activity and recombination assay

To assay the efficiency of FLP-mediated recombination of wild-type and mutated FRT pairs, strains were generated where the um11377.2 gene was deleted in SG200FLP using the hph resistance cassette from pHwtFRT, pHFRTm1, pHFRTm2, pHFRTm3 and pHFRTm4 plasmids, respectively. In each case, one transformant was recovered in which the um11377.2 gene was replaced by a hygromycin resistance cassette flanked by a pair of wild-type or mutated FRT sequences. These strains were grown in CM-ara medium for 24 h and then plated onto PD plates to obtain single colonies. Colonies were replica-plated to PD and PD-hyg plates. The recombination efficiency was calculated by determining the percentage of cells that had lost hygromycin resistance.


Multiple alignments of eff1 genes (see Table S2) were made in macaw (Schuler et al., 1991) and signal sequences were predicted with signalp at (Bendtsen et al., 2004). All other bioinformatic analyses were performed within the MPI Bioinformatics Toolkit at (Biegert et al., 2006). Searches for members of the Eff1 protein family were made by comparison of the profile Hidden Markov Models (HMMs) in hhpred (Söding et al., 2005). To this end, HMMs were generated for all open reading frames (ORFs) of the U. maydis genome and have been made publicly available for searching at The cluster analysis was made with clans (Frickey & Lupas, 2004).


Establishment of the FLP-mediated recombination system in U. maydis

To increase the probability of expression of FLP in U. maydis, a codon-optimized FLP recombinase gene was assembled from oligonucleotides (Methods S1). In total, 384 silent mutations were introduced (Fig. S1). To test FLP expression we generated strain SG200FLP, in which the FLP recombinase is under the control of the arabinose-inducible crg1 promoter. By northern blot analysis, FLP recombinase gene expression was visualized 1 h after the shift to CM-ara (Fig. S2). To assess FLP activity, the self-replicating recombination reporter plasmid, pIF1, was introduced into SG200FLP and into SG200 as a control. In pIF1, lacZ′ is disrupted by a cassette in which the FRT sites flank a constitutively expressed egfp gene. FLP-mediated recombination should excise the egfp cassette and leave a plasmid in which the lacZ ′ gene is restored (Fig. 1a). After growth in CM-glu medium, SG200pIF1 and SG200FLPpIF1 strains were shifted to CM-ara medium for up to 16 h. Starting at 4 h, a decrease of relative fluorescence units was observed in the SG200FLPpIF1 strain, indicating excision of the egfp gene (Fig. 1b). Neither in the control strain SG200pIF1, nor in SG200FLPpIF1 grown in CM-Glu medium was such a decrease in relative fluorescence observed (Fig. 1b). After an induction period of 16 h, DNA was prepared and introduced into DH5α by electroporation. All transformants with DNA isolated from SG200pIF1 were white on 5-bromo-4-chloroindol-3-yl β-d-galactoside (Xgal) plates, whereas 67.8 ± 15.5% (calculated from three independent experiments) of the transformants with DNA isolated from SG200FLPpIF1 were light blue (Fig. 1c), indicating partial restoration of LacZ activity. When FLP activity was not induced in SG200FLPpIF1, 1.8% of the E. coli transformants showed lacZ′ expression, indicating some leakiness of the crg1 promoter. Compared to plasmids isolated from white colonies, plasmids from light-blue colonies were reduced in size by 2.0 kb, indicative of FLP-mediated excision (Fig. S3). In five of these plasmids the expected new junction was verified by sequencing. This illustrated efficient FLP-mediated excision, when FRT sites are located on an autonomously replicating plasmid.

Figure 1.

 Establishment of the FLP/FRT system in Ustilago maydis. (a) Scheme of the recombination reporter plasmid pIF1. egfp is inserted in the lacZ′ gene and is flanked by directly repeated FRT sites (red bars with arrows). egfp expression is controlled by the constitutive Po2tef promoter and the nos terminator (upper panel). FLP-mediated recombination excises egfp, leaving one FRT site. This site is in-frame with the open reading frame of lacZ′ (lower panel). (b) SG200FLPpIF1 and SG200pIF1 were either grown in CM-glu or shifted to CM-ara for the indicated time. Relative eGFP fluorescence/OD600 was determined in triplicate experiments. Average values are shown and SEs are indicated for SG200FLPpIF1 shifted to CM-ara (red) and CM-glu (green) and for SG200pIF1 shifted to CM-ara (blue) and CM-glu (dark red). (c) Sixteen hours after the shift of SG200pIF1 and SG200FLPpIF1 to CM-ara medium, DNA was prepared, transformed into DH5α and transformants were plated on YT/5-bromo-4-chloroindol-3-yl β-d-galactoside (Xgal) plates. All transformants with DNA recovered from SG200pIF1 yielded white colonies (left) while about two-thirds of the transformants with DNA from SG200FLPpIF1 yielded blue colonies (right).

Next, we tested the efficiency of FLP-mediated excision in strain SG200FLP when FRT sites are present in the genome. To achieve this, um01796 was disrupted in SG200FLP by the hph cassette flanked by wild-type FRT sites (Fig. 2a) to yield strain SG200FPLΔ01796FRT/FRT. Genomic DNA was isolated at different time-points after the induction of FLP. By PCR analysis using primers flanking um01796 the appearance of the 2 kb post-excision product was demonstrated after an induction period of only 2 h and the intensity of this band increased with prolonged induction time (Fig. 2b).

Figure 2.

 Time course and efficiency of FLP-mediated excision of the hph cassette from a genomic location. (a) Scheme of the events leading to deletion of um01796 and subsequent FLP-mediated excision. FRT sites are indicated as red boxes and their orientation is given by arrows. In the first step the deletion construct for gene um01796 is integrated in the genomic locus of SG200FLP by homologous recombination (dotted crosses), yielding a hygromycin-resistant transformant. After inducing FLP expression by shifting to CM-ara medium the hygromycin resistance cassette is excised leaving one FRT site in the genome. PCR primers used for demonstrating excision of the hph cassette are indicated together with the length of the amplified fragments. (b) SG200FLPΔ01796FRT/FRT was grown in CM-glu medium and then shifted to CM-ara medium for 24 h. Samples were taken at the indicated times and DNA was prepared. Excision was monitored by PCR using primers P1 and P2. The 4.8 kb fragment corresponds to the nonexcised cassette, the 2.1 kb fragment indicates the post-excision product. (c) SG200Δ01796FRT/FRTpFlpexpC was grown in CM-ara medium for 16 h and then plated on CM. DNA was prepared from nine single colonies (lanes 1–9) and analyzed by PCR using primers P1 and P2. Hygromycin-resistant colonies contained the nonexcised hph cassette (4.8 kb) while all hygromycin-sensitive colonies had excised the cassette (2.1 kb). Using the P1 and P2 primer pair a 3.3 kb PCR product is amplified from SG200 (lane 10). M, 1 kb ladder.

As the low-level basal expression of FLP observed in SG200FLP could cause premature excision of the resistance cassette and complicate the identification of desired transformants, a self-replicating FLP-expressing plasmid (pFLPexpC) was generated (Fig. S4). Such plasmids containing an autonomously replicating sequence from U. maydis have been shown to be mitotically unstable (Tsukuda et al., 1988). pFLPexpC was introduced into SG200Δ01796FRT/FRT, and, after inducing FLP expression, colonies that had lost the hygromycin cassette, as well as the FLP donor plasmid, were identified (see the Materials and Methods section for details). In three independent experiments an excision frequency of 65.71 ± 4.34% was determined and of these cells 21.1 ± 2.6% were also carboxin sensitive (i.e. they had lost pFLPexpC). Excision of the hph cassette was verified by PCR using primers flanking gene um01796 (Fig. 2c). There was a perfect correlation between the presence of the 2 kb PCR product, indicative for deletion of the hygromycin cassette from the genome, and hygromycin sensitivity. For the subsequent experiments, selection for the simultaneous loss of hygromycin and carboxin resistance was applied.

Because the generation of successive deletion mutants was very time consuming, we modified the system in such a way that in future experiments the FLP gene could be introduced into the genome together with the hph cassette. To demonstrate that this is feasible we disrupted um01796 in SG200 by introducing pYUIF-FRTm2. In the resulting strain, a cassette containing FLP and hph is flanked by FRTm2 sites (Fig. S5a). After inducing FLP, 34 ± 9% of the resulting single colonies were hygromycin sensitive. PCR analysis revealed that in these colonies the hph gene, as well as the FLP gene, had been lost (Fig. S5b). This illustrates that the experimental speed in future studies can be significantly increased.

Recombination efficiency of core-mutated FRT sequences

To minimize the likelihood of chromosome rearrangements occurring through intramolecular and intermolecular recombination between identical FRT sites left in the genome after several rounds of FLP-mediated excision, four mutated FRT sequences were designed, each with a different point mutation in the core region (Fig. S6a). For the FRTm1 sequence, functionality had previously been demonstrated (Storici et al., 1999). Plasmids were generated in which the hph marker gene is flanked by two direct copies of these mutated FRT sequences and the left and right flanks of the um11377.2 gene. The hph cassettes and flanking regions from pHwtFRT (Fig. S4), pHFRTm1, pHFRTm2, pHFRTm3 and pHFRTm4 were amplified and introduced into SG200FLP individually to disrupt gene um11377.2. The efficiency of FLP-mediated recombination of mutated FRT sites was assayed (see the Materials and Methods section). Relative to recombination in the strain carrying wild-type FRT sequences, the mutated FRT sequences recombined two to five times less efficiently (Fig. S6b).

To demonstrate that there is no recombination between wild-type FRT and FRT sites carrying mutations in the core region we used SG200Δ01796FRTΔ11377.2FRT/FRT and several derivatives. In SG200Δ01796FRTΔ11377.2FRT/FRT, um01796 on chromosome 3 has been deleted, leaving one wild-type FRT site, and um11377.2 (also residing on chromosome 3 at a distance of 0.43 Mb from um01796) has been replaced with a hygromycin cassette flanked by two wild-type FRT sites. In a parallel experiment, the four strains SG200Δ01796FRTΔ11377.2FRTm1/FRTm1, SG200Δ01796FRTΔ11377.2FRTm2/FRTm2, SG200Δ01796FRTΔ11377.2FRTm3/FRTm3 and SG200Δ01796FRTΔ11377.2FRTm4/FRTm4 (Table S3), which differ from SG200Δ01796FRTΔ11377.2FRT/FRT only by the m1, m2, m3 or m4 mutation in the FRT sites (Fig. S6a) residing in the um11377.2 locus (before excision of hph), were generated. After introducing pFLPexpC and inducing FLP expression (see the Materials and Methods section) DNA was isolated and analyzed by PCR to identify deletion events that had occurred between FRT sites in the um01796 and the um11377.2 loci using the primer combinations shown in Fig. S7a. While recombination could be visualized between a wild-type FRT site in the um01796 locus and wild-type FRT sites in the um11377.2 locus in DNA isolated from SG200Δ01796FRTΔ11377.2FRT/FRT, such a 3.1 kb product was not amplified when strains harbored a wild-type FRT site and any of the mutated FRT sites (Fig. S7b). This illustrates that the FRT mutations introduced greatly reduce, or abolish, recombination with wild-type FRT sites, which should consequently help to maintain strain integrity.

An 11-gene family in U. maydis that codes for novel secreted proteins

U. maydis families 9 and 17 have been described to consist of four and three genes, respectively, encoding novel secreted proteins that are U. maydis-specific (Kämper et al., 2006). Our initial interest in these groups of secreted proteins was based on the finding that three members of these two gene families contain putative nuclear localization signal (NLS) motifs (Müller et al., 2008). However, these putative NLS sequences occur at nonconserved locations and are therefore not considered to be functionally relevant. Additionally, a more rigorous search for related genes using profile HMMs revealed that these seven genes in families 9 and 17 are related and four additional paralogs exist in the genome (Figs 3, S8). We designated this enlarged gene family as the eff1 family (Table S2). The eff1 family comprises um01796 and um11377 on chromosome 3, the adjacent genes um03313 and um03314 on chromosome 8, and the seven genes, um02135, um02136, um02137, um02138, um02139, um02140 and um02141 clustered on chromosome 5. All encoded proteins, except for Um11377, contain putative N-terminal secretion signals, and an analysis of the um11377 gene region showed that sequence similarity with other Eff1 proteins extends well upstream of the predicted Um11377 start methionine and includes a putative signal sequence. This larger frame is, however, disrupted by a stop codon at position 64. We resequenced this gene from the sequenced U. maydis strain 521 and detected a sequencing error. The gene model was corrected and the gene is now designated um11377.2. Additionally, based on sequence similarity between Um02139 and Um02140 (Fig. S9) Met37 is strongly implied to be the true start codon of Um02140. The respective gene is now designated um02140.2. Sequence comparisons between Eff1 proteins showed that the family forms three subgroups (Fig. 3): group I, comprising Um01796, Um11377.2, Um02137 and Um02138; group II, comprising Um02139, Um02140.2, Um02141, Um03313 and Um03314; and group III, comprising Um02135 and Um02136. Group III sequences are highly divergent, yet should be counted as true Eff1 homologs based on the following observations: they make multiple, statistically significant connections to other Eff1 proteins in HMM comparisons (Fig. 3); they are bidirectional best-hits to the other Eff1 proteins in sequence searches by HMM; they have the same domain structure as other Eff1 proteins (Fig. S9); and they are located directly adjacent to the main cluster of Eff1 proteins on chromosome 5.

Figure 3.

 Cluster map of Eff1 proteins. Sequences were compared using blast, seeded randomly in space and connected by forces proportional to the significance of their pairwise blast matches. The map was then equilibrated to convergence in a force-directed manner. blast matches at a P-value of 1.0 or better are shown as grey lines. Three groups of sequences are apparent, colored red, blue and green on the map. Group III is the most distant and loses connection with the other groups at a P-value cut-off of 1e-03. Groups I and II lose connection with each other at a cut-off of 1e-09. Eff1 proteins are labelled by their respective um numbers, as well as by their eff1 gene numbers.

All Eff1 proteins have the same architecture, consisting of an N-terminal signal sequence, a central region predicted to be natively unstructured and a conserved C-terminal domain, which presumably represents the only folded part of these proteins (Fig. S9). We also noted that in group II sequences, the central region predicted to be natively unstructured contains a conserved segment with an area of elevated helical propensity, which is duplicated in Um02139 and Um03314 (Fig. S9b).

Expression patterns of the eff1 effector genes

The expression patterns of all members of the eff1 family were analyzed by quantitative real-time PCR during different stages of fungal development using uninfected plant material as a negative control. The statistical analysis of the real-time PCR data is presented in Table S4. Gene-expression levels were quantified in reference to the constitutively expressed peptidyl-prolyl cis-trans isomerase gene ppi1 (accession number EAK84904). During axenic growth of SG200 in YEPSL, expression of the 11 eff1 genes could not be detected. Twenty-four hours after plant infection, a time-point when U. maydis has developed appressoria and has begun to invade the host tissue (Mendoza-Mendoza et al., 2009), expression of seven eff1 genes was evident (Fig. 4). At this time-point the um02141 transcript was up-regulated c. 1000-fold compared with the ppi1 expression level. Over the next 7 d of biotrophic growth, the expression levels of um02141 decreased by two- to threefold (Fig. 4). The maximum transcript levels of um01796, um03313 and um02139 occurred 5 d postinfection, while um11377.2 and um02140.2 demonstrated the highest expression levels at 3 d postinfection. For um02137 and um02138 there was a continuous increase in expression over the period from 1 to 8 d postinfection. Except for um02139, which was, at most, up-regulated fivefold, all other genes were up-regulated at least 50-fold at one of the chosen time-points after infection (Fig. 4). This illustrates that all members of this gene family are specifically expressed during the biotrophic phase and thus qualify to be called effectors. Accordingly, they were renamed eff1-1 to eff1-11 (Fig. 5a, Table S2).

Figure 4.

 Expression analysis of the eff1 family genes by quantitative real-time PCR. RNA was prepared from SG200 grown saprophytically in YEPSL liquid medium, as well as from maize seedlings, 1, 3, 5 and 8 d postinfection (dpi). Expression levels of the eff1 family genes was standardized relative to the constitutively transcribed ppi gene. Three biological replicates, with two technical replicates each, were obtained. Standard deviations are indicated. P-values calculated using the Student’s t-test are shown in the Supporting Information Table S4. Asterisks indicate significant differences relative to expression values in axenic culture. P-values: *, < 0.05; **, < 0.01; ***, < 0.001. Expression of the eff1 genes was also tested in uninfected plant material (depicted on the right) and in this case P-values were calculated relative to expression values in infected plant material 5 d postinfection (see Methods S2). The color code for individual genes is shown below the figure.

Figure 5.

 Strategy for the construction of eff1 deletion strains. (a) The arrangement of eff1 family genes on chromosomes 3, 8 and 5 is shown in the upper part of each panel using the color code for individual groups shown in Fig. 3. The lower part of the panels shows the intermediate, as well as the final, structure of the respective chromosomes. Wild-type and mutant FRT sites remaining after excision are indicated in color. The binding sites of diagnostic primers and the sizes of the PCR fragments are indicated. (b) Verification of all deletions in SG200eff1Δ1-11. DNA was prepared from SG200 and SG200eff1Δ1-11, and diagnostic fragments were amplified from SG200 DNA (lanes 1) and SG200eff1Δ1-11 DNA (lanes 2) using the primer pairs indicated in panel (a). M, 1 kb ladder.

The generation of mutants lacking members of the eff1 gene family

To generate mutants lacking either all or different combinations of the genes constituting the eff1 gene family, we followed the scheme depicted in Fig. 5(a) using five successive rounds of FLP-mediated recombination (see Methods S1 for details). This allowed us to generate strains SG200eff1Δ1, SG200eff1Δ1,2, SG200eff1Δ1,2,3,4, SG200eff1Δ1,2,3,4,7,8,9,10 and SG200eff1Δ1-11 (Table S3). In addition, another five strains carrying different combinations of eff1 gene deletions were generated. These are strains SG200eff1Δ3,4, SG200eff1Δ1,8, SG200eff1Δ1,3,4,8, SG200eff1Δ1,2,7,8,9,10 and SG200eff1Δ1,2,3,4,7,8,9,10,11 (Table S3, Methods S1). In SG200eff1Δ1-11, disruption of all 11 genes was verified by PCR using primers that bind to the left and right borders of the segments that were deleted (Fig. 5a,b).

Phenotypic analysis of eff1 mutants

To test the phenotype of mutants in which different combinations of eff1 genes were deleted, the mutant strains were grown on various stress media (sorbitol, sodium chloride, calcofluor, Congo red and H2O2). In no case could significant differences in growth between SG200 and mutants be observed (Fig. S10). In addition, all mutants developed vigorous filaments on CM-charcoal plates, comparable to those developed by SG200 (Fig. S10). Next, the 10 strains carrying different deletion combinations of family eff1 genes (Table S3) were tested for pathogenicity. It had been shown previously that virulence can be lost completely when a single effector gene, in this case pep1, is deleted in SG200 (Doehlemann et al., 2009). Compared with SG200 infections, significantly reduced virulence was observed when 9 or all 11 eff1 genes were deleted (Fig. 6, see Table S5 for a statistical analysis of the data). In comparison to symptoms caused by SG200, two effects were seen with SG200eff1Δ1-11: the number of plants showing symptoms decreased from > 95% to c. 50%; and the symptom severity was reduced. The symptoms (ligula swellings and small tumors) observed after infection with SG200eff1Δ1-11 and SG200eff1Δ1,2,3,4,7,8,9,10,11 were also significantly different, suggesting that the two distantly related genes from group III –eff1-5 and eff1-6 (Fig. 3) – also contribute weakly to the phenotype of the 11-gene-deletion strain. Compared with SG200eff1Δ1,2,3,4,7,8,9,10,11, SG200eff1Δ1,2,3,4,7,8,9,10 showed more large tumors and dead plants, suggesting that eff1-11 is an important gene for tumor formation. SG200eff1Δ1,2 induced more chlorosis and fewer tumors, but no reduction in the number of plants showing symptoms (still almost 100%). Thus, eff1-1 and eff1-2 are important for tumor formation and they explain the elevated amount of chlorosis in all strains from which they are deleted. SG200eff1Δ1,2,3,4 and SG200eff1Δ3,4 were similar in type of symptoms, but they showed differences in the number of plants with symptoms, which was lower in SG200eff1Δ1,2,3,4. This shows again that eff1-1 and eff1-2 contribute positively to virulence. As SG200eff1Δ1,2,7,8,9,10 induced symptoms comparable to those of SG200eff1Δ1,2, there is no clear contribution of eff1-7, eff1-8, eff1-9 and eff1-10 to virulence. SG200eff1Δ1,2,3,4,7,8,9,10 showed significantly decreased tumor symptoms compared with SG200eff1Δ1,2,7,8,9,10 (Fig. 6), suggesting a contribution of eff1-3 and eff1-4 to virulence. This conclusion was reinforced by the attenuated virulence seen after infection with SG200eff1Δ3,4 (Fig. 6).

Figure 6.

 Disease symptoms caused by SG200 and SG200 carrying different combinations of eff1 gene deletions. For each strain three independent infections were performed and the total number of infected plants is given above each column. Symptoms were scored 14 d after infection and average scores from the three experiments are shown. To facilitate comparisons, the 11 eff1 genes are listed on the right and their absence in individual strains is indicated by Δ. The color code for disease rating is given below the figure. Standard deviations and P-values are shown in the Supporting Information Table S5. P-values relative to symptoms caused by SG200: *, < 0.05; **, < 0.01; ***, < 0.001.

To assess the contribution of individual eff1 genes to the virulence phenotype, several complementation strains were generated in which individual eff1 effector genes were re-introduced into SG200eff1Δ1,2,3,4,7,8,9,10,11. Two strains carrying single-copy integrations of eff1-1 or eff1-8 were generated. Both strains showed statistically relevant elevated virulence compared with the strain carrying nine eff1 gene deletions, with eff1-1 showing stronger complementation than eff1-8 (Fig. 7). Additionally, an SG200eff1Δ1,2,3,4,7,8,9,10,11 strain was generated in which the four genes eff1-7, eff1-8, eff1-9 and eff1-10 were introduced in single copy, and this strain also showed statistically relevant increased virulence compared with the nine-gene-deletion strain (Fig. 7). As the severely reduced virulence of the nine-gene-deletion mutant could be partially restored by complementation with all genes tested, we assume functional redundancy within the eff1 gene family.

Figure 7.

 Complementation of SG200eff1Δ1,2,3,4,7,8,9,10,11 with different eff1 genes. Disease symptoms caused by SG200, SG200eff1Δ1,2,3,4,7,8,9,10,11 and three different complementation strains are shown. Genes re-introduced into SG200eff1Δ1,2,3,4,7,8,9,10,11 are listed below. The color code for disease rating is given below the figure. Standard deviations and P-values are shown in the Supporting Information Table S5. P-values relative to symptoms caused by SG200eff1Δ1,2,3,4,7,8,9,10,11: *, < 0.05; **, < 0.01; ***, < 0.001.


In this study we successfully implemented the FLP/FRT system to generate multiple gene deletions in U. maydis. To achieve this, a codon-optimized FLP gene was designed and recombination was assayed using a genetic screen as well as by PCR. The subsequent scheme for the FLP-mediated marker deletion and re-use consisted of three main steps: the generation of a deletion mutant in which the selectable marker introduced is flanked by direct repeats of FRT sites; the introduction of an inducible FLP gene on an autonomously replicating plasmid; and the induction of FLP expression and the subsequent screening for the loss of the selectable marker as well as the FLP donor plasmid. The system was highly efficient and c. 21 ± 8% of the single colonies cultured after the induction of FLP had lost the marker as well as the FLP donor plasmid. For future use we have developed the system in such a way that the inducible FLP gene can become an integral part of the cassette used to generate the deletion.

The successful deletion of the 11-gene eff1 effector family in SG200, using five consecutive rounds of gene replacements and subsequent marker excision, showed the utility of the FLP/FRT system in U. maydis. To eliminate possible intermolecular and intramolecular recombination events between identical FRT sites left in the genome after excision, FRT sequences with different point mutations in the core region were employed. None of these sites was found to recombine with wild-type FRT sites, but recombination could be detected in a strain where two identical FRT sites were present on the same chromosome, 0.43 Mb apart. None of the strains generated by several rounds of FLP-mediated recombination showed morphological defects and all of these strains developed vigorous filaments and were comparable to the progenitor strain with respect to growth under oxidative stress, cell wall stress and osmotic stress. This makes it likely that such strains are stable and are unlikely to have acquired additional mutations, a prerequisite for assessing small virulence phenotypes of redundant genes. We expect that the establishment of the FLP system in U. maydis will pave the way towards functional analysis of effector gene families as well as serving as a tool to improve genetic manipulations of plant pathogenic fungi in general.

Expression of all members of the eff1 gene family was strongly induced during biotrophic development. This is one of the few unifying features of filamentous pathogen effectors in which the respective genes are up-regulated during host colonization, either during penetration, in haustoria, or during later stages of pathogen development inside the plant tissue (Hahn et al., 1997; Catanzariti et al., 2006; Kämper et al., 2006; Haas et al., 2009; Oh et al., 2009; Skibbe et al., 2010). However, information on how this regulation is connected to the growth stages inside the host is scarce. In U. maydis the zinc-finger protein, Mzr1, has been defined as a transcriptional activator of mig2 effector genes during host colonization (Zheng et al., 2008). However, Mrz1 was shown to regulate only two of the five mig2 genes and thus is not a general regulator for effector genes. In addition, given the recent finding that several U. maydis effectors are expressed in a maize tissue-specific manner (Skibbe et al., 2010), it is unlikely that such global effector regulators exist at all. In Fusarium oxysporum, a regulatory gene, SGE1, has recently been identified that was crucial for pathogenesis and affected the expression of the four tested six effector genes. U. maydis has two genes related to SGE1 (um06496 and um05853) (Michielse et al., 2009), which have not yet been functionally characterized. However, because SGE1 orthologs exist in all fungi (Michielse et al., 2009), it is unlikely that this transcription factor will be identified as a dedicated regulator of effector genes in phytopathogenic fungi.

The fact that 9 of the 11 eff1 genes reside in two clusters in the genome supports the assertion that they have originally arisen by a local gene-duplication mechanism followed by rapid diversification and dispersion to other chromosomes. The large eff1 gene cluster on chromosome 5 is heterogeneous and contains genes from three different eff1 groups (Fig. 5a) with two, two and three direct copies from each group, respectively. Intriguingly, the two dispersed copies on chromosome 3 are closely related to each other as well as to the two adjacent genes –eff1-7 and eff1-8– in the large cluster on chromosome 5. In addition, the two adjacent copies on chromosome 8 are closely related to each other and to the adjacent genes –eff1-9 and eff1-10– in the large cluster on chromosome 5. It will be very interesting to analyze the number, distribution and groups of eff1 effector genes in geographically distinct isolates of U. maydis as this might provide insights into the evolutionary fate of the clusters as well as the dispersed copies.

Database searches with the conserved C-terminal domain identified in this protein family (Fig. S9) did not produce any statistically significant matches. However, orthologs of the eff1 genes were found in Sporisorium reilianum, the cause of head smut in maize (J. Schirawski, G. Mannhaupt & R. Kahmann, unpublished data) and in Ustilago scitaminea, the cause of sugarcane smut (G. Mannhaupt & R. Kahmann, unpublished data), which makes it likely that this effector family is smut specific and is not involved in determining host range. The combined deletion analysis and complementation studies conducted for members of the eff1 family has revealed that genes eff1-11, eff1-3 and eff1-4 contribute highly significantly to virulence, whereas all other members of this gene family contribute to virulence only weakly. As we have always deleted genes eff1-3 and eff1-4 simultaneously, we are presently unable to assess whether they both contribute to virulence. Interestingly, the eff1 genes with the strongest effect on virulence all belong to group II (Fig. 3). By contrast, Eff1-4 is very similar to Eff1-9, but the effects of eff1-9 on virulence were marginal. This could indicate that additional, currently unrecognized, features determine the strength of a given eff1 gene. Alternatively, because eff1-9 is the gene that shows the smallest increase in expression values during biotrophic growth, the differences in contribution to virulence could be attributed to these differences in expression. This could be investigated by promoter-swap experiments.

Because the observed complementation was, in all cases, partial, it remains possible that some subfunctualization in eff1 genes (Conant & Wolfe, 2008) has occurred. In this respect it would be interesting to determine whether the separate deletion of the three different groups of eff1 genes produces distinct phenotypic effects and whether these can then be complemented only by eff1 genes from the same group or also by members of different groups. Such differences in virulence might become apparent when using different maize cultivars or when infecting different maize tissues such as tassel, the cob or mature leaves (Skibbe et al., 2010), rather than seedlings, as used in this study.

At present, the function of Eff1 proteins after secretion remains highly speculative because the Eff1 effectors do not show any similarity to database entries. The members of this family are highly divergent but show conserved features. The central segment could represent a eukaryotic linear motif, which binds specifically to a protein of the plant host and becomes partly ordered in the process. Although the C-terminal domain also lacks similarity to known domains, as judged by HMM comparisons to the NCBI conserved domain database (CDD), the fact that it is predicted to assume a folded conformation makes it the most likely domain to confer Eff1 protein activity. This is in line with results from oomycete effectors, which revealed that functional domains reside in the C-terminus (Ellis et al., 2009; Tyler, 2009). Preliminary experiments indicate that the 11-gene-deletion mutant does not elicit a hypersensitive response and is able to establish biotrophic growth (K. Schipper, Y. Khrunyk & R. Kahmann, unpublished data). This is different from the phenotype of pep1 effector mutants that are arrested during penetration and elicit strong defense responses (Doehlemann et al., 2009). The Eff1 proteins lack ‘classical’ RXLR motifs found in the N-terminal domains of effector proteins from oomycetes that are translocated to plant cells (Kamoun & Goodwin, 2007; Ellis et al., 2009; Tyler, 2009). Further studies will reveal if the Eff1 effectors remain in the host–pathogen interface or enter plant cells by alternative, RXLR-independent, routes. Moreover, identification of interacting proteins and the results of localization studies may give valuable hints concerning the function of the family members. Given the fact that the deletion mutant lacking all 11 eff1 genes has a strong virulence phenotype, functional studies can now be conducted. It will be of particular interest to analyze the effects of deletion of the conserved C-terminal domain as well as deletions in the central domain. This could reveal whether these domains confer functions that can be separated (i.e. on the one hand are involved in targeting, and on the other hand interact with host proteins).


Y.K. is grateful to the members of her thesis committee and in particular M. Bölker for constructive comments. We thank G. Mannhaupt for genome comparisons. Our work was supported by the DFG-funded Collaborative Research Center SFB593. Y.K. received a fellowship from the IMPRS.