In a previous study we have described two Pex5-like proteins, Pex5a (Um02528) and Pex5b (Um10172) from U. maydis required for optimal growth of the fungus in different nutrient conditions and for pathogenic development (Figure 1A; Freitag et al., 2012). To discriminate the specific contributions of both proteins for growth on different carbon sources (Freitag et al., 2012; Camões et al., 2015), we tested pex5a and pex5b single and double deletion mutants on media containing either glucose or oleic acid as sole carbon source (Figure 1B). We found that ∆pex5a cells did not exhibit a severe growth defect on glucose plates but on oleic acid-containing medium (Figure 1B). Growth of mutants lacking pex5b or both genes was reduced on glucose-containing medium indicated by smaller colonies on solid medium and lower growth rates in liquid medium (Figure 1B and Supplementary Figure S1). On plates containing oleic acid as sole carbon source proliferation of these mutants was almost abolished (Figure 1B). This indicates that the presence of both Pex5 proteins is required for peroxisome function in U. maydis.

Next, we followed localization of the peroxisomal reporter protein mCherry-SKL in ∆pex5a and ∆pex5b cells (Figure 1C). Peroxisomal localization of mCherry-SKL was observed in ∆pex5a mutants but not in ∆pex5b mutants suggesting that Pex5b is necessary for peroxisomal import of PTS1 proteins in U. maydis. To test whether impaired binding of PTS1 cargo was responsible for the phenotype of ∆pex5b cells we deleted the TPR domains of Pex5b. Although the partial deletion of pex5b affected growth on oleic acid-containing medium (Figure 1B), pex5b∆TPR mutants still were able to import mCherry-SKL into peroxisomes (Figure 1C). Pex5b can therefore import PTS1 proteins into peroxisomes in the absence of Pex5a, while Pex5a-mediated protein import depends on the Pex5b NTD.

To address whether the observed growth phenotype of ∆pex5a cells (Figure 1B) results from different specificities of the Pex5a and Pex5b receptors for distinct subsets of peroxisomal matrix proteins or is caused by the reduced overall amount of Pex5 receptors, we overexpressed Pex5 derivatives. Overexpression of Pex5b rescued the growth phenotype of ∆pex5b cells but did not restore the growth phenotype of ∆pex5a cells (Supplementary Figures S2A,B). This indicates that it is not the reduced dosage of TPR proteins but rather the specificity of the Pex5a receptor, which explains the phenotype of ∆pex5a mutants.

We addressed this hypothesis by construction of strains expressing Pex5a and Pex5b chimeric variants (Supplementary Figure S3). A mutant expressing Pex5 proteins containing only the TPR domain of Pex5a (pex5b:TPR _pex5a ) exhibited a much stronger growth defect on oleic acid-containing medium compared to a mutant expressing only the TPR domain of Pex5b (pex5a:TPR _pex5b ), which only showed a slightly reduced colony size (Figure 1B). The TPR domain of Pex5b, therefore, recognizes specific PTS1 proteins required for peroxisome function in these conditions. This is further supported by the finding that, although growth was abolished on oleic acid-containing medium (Figure 1B), mCherry-SKL is localized to peroxisomes in ∆pex5a pex5b:TPR _pex5a cells indicating that this chimeric protein is not generally affected in peroxisomal import of PTS1 proteins (Figure 1C).

In addition, our experiments reveal that the absence of the NTD of Pex5a is primarily responsible for the growth defect of the ∆pex5a strain on oleic acid-containing medium (Figure 1B, compare mutant ∆pex5a with pex5a:TPR _pex5b ). Indeed, overexpression of Pex5a∆TPR was able to rescue the phenotype of a ∆pex5a strain (Figure 1D). Critical peroxisomal matrix proteins are likely to exist, which specifically require the NTD for import. Moreover, the data suggest that Pex5a can act as a targeting factor even in the absence of its TPR domains.

Several proteins from other fungi are known, which are imported into peroxisomes via binding to the NTD of Pex5 although they lack a canonical PTS or the PTS1 has been removed (van der Klei and Veenhuis, 2006). We reasoned that PTS1-containing substrates that require the NTD of Pex5a may as well display specificity for the TPR domain of this cargo receptor. To identify such proteins, we generated a library of GFP proteins with C-terminal dodecameric PTS1 motifs derived from U. maydis enzymes presumably involved in peroxisomal ß-oxidation (Table 1; Figure 2A and Supplementary Figure S4; Brocard and Hartig, 2006; Camões et al., 2015). Constructs were expressed in WT and in ∆pex5a cells (Supplementary Figure S4). GFP fused to PTS1 motifs of the candidate proteins Um01966, Um10665 and Um11001 showed peroxisomal localization in WT cells but accumulated in the cytosol in the absence of Pex5a (Figure 2A and Supplementary Figure S4).

Um01966 is a putative acyl-CoA oxidase with high similarity to yeast Pox1, Um10665 a putative acyl-CoA dehydrogenase and Um11001 a putative enoyl-CoA hydratase (Table 1; Camões et al., 2015). N-terminally GFP-tagged full-length versions of all three proteins co-localized with mCherry-SKL in the presence of Pex5a, but showed pronounced cytosolic localization in ∆pex5a cells (Figure 2B). Cytosolic accumulation of GFP-tagged full-length proteins was even enhanced compared to the C-terminal dodecamers fused to GFP. Both experiments show that the three identified proteins are cargo with a preference for Pex5a.

To discriminate between the function of the C-terminal TPR domain and the NTD of Pex5a for peroxisomal import of Um01966, Um10665 and Um11001 we added canonical SKL containing motifs at the C-terminus of each. Targeting to peroxisomes in ∆pex5a cells was drastically improved for GFP-Um10665-SKL and GFP-Um11001-SKL. Um01966-SKL predominantly co-localized with mCherry-SKL positive foci in wild type cells but substantial cytosolic mistargeting was observed in a strain deleted for pex5a (Figure 3A). This suggests a critical function of Pex5a for import of the acyl-CoA oxidase Um01966, which cannot be entirely bypassed by addition of a C-terminal canonical targeting signal.

Upon masking the PTS1 by C-terminal addition of three alanine residues (AAA), GFP-Um10665-AAA and GFP-Um11001-AAA remained largely cytosolic, while GFP-Um01966-AAA partially co-localized with mCherry-SKL (Figure 3B). A GFP-tagged and AAA-masked C-terminal dodecamer of Um01966 remained cytosolic revealing an additional targeting signal within the full-length protein (Supplementary Figure S5). Microscopic analysis of strains expressing chimeric variants of Pex5a and Pex5b demonstrated that the Pex5a TPRs are not required for efficient targeting of GFP-Um01966 to peroxisomes (Figure 3C). However, in the absence of the NTD of Pex5a (∆pex5a pex5b:TPR _pex5a ) residual peroxisomal targeting of GFP-Um01966 was observed (Figure 3C), while GFP-Um01966 localized almost entirely in the cytosol upon depletion of pex5a (Figure 2B). This is probably due to the presence of the Pex5a TPRs, which might partially compensate for the absence of the Pex5a NTD in ∆pex5a pex5b:TPR _pex5a cells (Figure 3C). These results are in accordance with our initial observation that the isolated PTS1 containing sequence requires Pex5a for efficient import (Figure 2A). Targeting of GFP-Um10665 and to a minor extent GFP-Um11001 was reduced in the absence of each part of Pex5a (Figure 3C). Together, these experiments suggest that it is the combination of affinities to the NTD and the CTD of Pex5a, which facilitates recognition and import of Pex5a specific cargo. The impact of each of the two domains can be different depending on individual substrates.

To test the binding affinities of TPR domains of Pex5a and Pex5b (Figure 2), we set up a yeast two-hybrid experiment (Chien et al., 1991). We attached the TPR domains of either protein to the Gal4-DNA-binding domain (BD), and GFP with C-terminal dodecamers of different candidate proteins to the Gal4-activation domain (AD) (Figure 4A). As a control we used a C-terminal dodecamer, which efficiently mediates import of GFP into peroxisomes in the absence of Pex5a (Supplementary Figure S3; Um03158). Constructs were transformed in respective combinations into AH109∆pex5 (Stehlik et al., 2020). Interactions were monitored for three independent transformants per combination by growth assays on high stringency medium (Figure 4B). The assay revealed a stronger interaction of PTS1 motifs of Um01966, Um10665 and Um11001 with the TPR domain of Pex5a compared to the TPR domain of Pex5b. Among the tested candidates the PTS1 motif of Um03158 had the highest affinity to the TPRs of Pex5b and may thus be efficiently imported via both Pex5 proteins (Figure 4B). Interaction data from the two-hybrid experiment are in agreement with the data on import efficiency obtained by fluorescence microscopy (Figure 2). The weak interaction of the Um10665 PTS1 with the Pex5a TPRs may explain why efficient import of GFP-Um10665 is only observed when a native Pex5a protein containing the original NTD and CTD is expressed (Figure 3C). The strong interaction of the PTS1 of Um01966 with the Pex5a TPRs explains residual peroxisomal import of GFP-Um01966 in ∆pex5a pex5b:TPR _pex5a cells (Figure 3C). Together, the results of the two-hybrid assay confirm that TPR domains of Pex5a and Pex5b show distinct preferences for variations of the C-terminal targeting signal and indicate that both cargo receptors have specific as well as shared substrates.

Import of the putative Pox1 ortholog Um01966 shows similarity to Pox1 in S. cerevisiae since both can be imported via the NTD of a Pex5 protein (Klein et al., 2002). We wondered whether more similarities exist and tested import specificity of the glyoxylate cycle enzyme Mls1 – a Pex9 substrate in S. cerevisiae (Effelsberg et al., 2016; Yifrach et al., 2016) – in cells containing Pex5 derivatives with only one type of TPRs. To this end we fused the C-terminal dodecamers of the U. maydis malate synthase ortholog Mls1 (Um15004) to GFP. Indeed, we detected efficient peroxisomal import in the presence of the TPR domain of Pex5b but pronounced cytosolic localization when only the TPR domain of Pex5a was present (Figure 5A). mCherry-SKL predominantly localized in peroxisomes of both strains (Figure 5A). Mls1 from U. maydis contains the C-terminal tripeptide ARI. Remarkably, also a stop codon readthrough derived isoform of the glycolytic/gluconeogenetic enzyme triosephosphate isomerase (Tpi1; Um03299) harbors this C-terminal tripeptide (Freitag et al., 2012) and efficient import into peroxisomes also depends on the TPRs of Pex5b (Figure 5B). Thus, the highly similar PTS1 motifs of both enzymes are preferentially recognized by Pex5b.

Of the cytosolic receptors, only Pex5b can mediate peroxisomal import as a single protein, while Pex5a requires the NTD of Pex5b (Figure 1). Pex5b might act as a co-receptor for Pex5a similar to Pex5-large for Pex7 in mammals or Pex18 and Pex21 for Pex7 in S. cerevisiae (Braverman et al., 1998; Otera et al., 1998; Purdue et al., 1998; Woodward and Bartel, 2005). Alternatively, Pex5b could be independently required for the activity of the peroxisomal import machinery.

In U. maydis PTS2-dependent transport may also involve Pex5b as a co-receptor, since the NTD of Pex5b contains a putative binding site for Pex7 (Figure 1; Kiel et al., 2006). In addition, the U. maydis genome lacks any ortholog of the yeast co-receptors Pex18 and Pex21 (Kämper et al., 2006). To test Pex5b for targeting of PTS2 proteins, we engineered a reporter protein by fusing an N-terminal fragment of the PTS2 protein Um01090 to GFP (PTS2-GFP; Figure 5A). The N-terminus of this predicted 3-keto-acyl-CoA thiolase related to yeast Pot1 contains a sequence matching the PTS2 consensus [R/K]-[L/V/I]-[X]5-[H/Q]-[L/A] (Lazarow, 2006; Camões et al., 2015; Kunze, 2020). We observed co-localization of PTS2-GFP with mCherry-SKL in peroxisomes and found that PTS2-GFP was retained in the cytosol upon deletion of the pex7 ortholog (um03596) (Figure 6A). To address whether Pex5b acts as co-receptor for Pex7, we expressed PTS2-GFP in pex5a and in pex5b deletion mutants. While peroxisomal localization was not affected in ∆pex5a cells, absence of pex5b completely abolished peroxisomal import of PTS2-GFP (Figure 6B). Reintroduction of the NTD of Pex5b into ∆pex5b mutants partially restored targeting of PTS2-GFP to peroxisomes suggesting that the NTD of Pex5b acts as co-receptor for Pex7 in U. maydis (Figure 6B).

Finally, we focused on peroxisomal matrix proteins without a canonical peroxisomal targeting signal. Several proteins have been described that neither contain a PTS1 nor a PTS2 but, nevertheless, localize inside of peroxisomes (van der Klei and Veenhuis, 2006). In U. maydis, the multifunctional enzyme involved in peroxisomal fatty acid oxidation (Mfe2; Um10038) was characterized earlier (Klose and Kronstad, 2006). Although we could not detect any PTS motif in Mfe2, both N-terminally and C-terminally tagged GFP fusion proteins co-localized with mCherry-SKL suggesting that Mfe2 is imported into peroxisomes independently of recognition by Pex7 or the Pex5 TPRs (Figure 6C). We probed the mode of transport of Mfe2 by localization studies in a set of U. maydis mutants and found that peroxisomal targeting of Mfe2 requires Pex5b. The NTD of Pex5b was sufficient for partial peroxisomal localization of GFP-Mfe2 but import of Mfe2 was more efficient when the full-length protein was expressed (Figures 6C,D). This shows that the NTD of Pex5b also can act as receptor for peroxisomal matrix proteins in U. maydis.

Constructs were generated using standard cloning procedures (Sambrook et al., 1989) or Gibson assembly (Gibson et al., 2009). All plasmids were verified by sequencing. Escherichia coli strain Top Ten (Invitrogen) was used for transformation according to a standard protocol (Hanahan et al., 1991) and propagation of plasmid DNA. Transformation of U. maydis cells was achieved as described previously (Schulz et al., 1990). Deletion strains and chimeric variants were created using an Sfi1 based cloning system (Brachmann et al., 2004; Kämper, 2004; Supplementary Figure S3). Genomic DNA was extracted according to an established protocol (Hoffman and Winston, 1987). Mutant strains were verified by Southern blot analysis (Sambrook et al., 1989). Proteins were expressed under control of the constitutive otef-promoter either integrated into the cbx-locus (Broomfield and Hargreaves, 1992; Spellig et al., 1996) or randomly integrated into the genome of U. maydis (mCherry-SKL). Similar expression levels were confirmed by measuring fluorescence using Synergy Mx multimode microplate reader (BioTek). All plasmids, strains and oligonucleotides used or generated during this study are listed in Supplementary Table S1. Progenitor plasmids used for this study were described (Spellig et al., 1996; Brachmann et al., 2004; Sandrock et al., 2006; Freitag et al., 2012; Stehlik et al., 2020). Genes can be accessed on NCBI.

U. maydis strains were grown at 28°C in liquid YEPSL (1% yeast extract, 0.4% peptone, 0.4% sucrose) or on solid potato dextrose broth containing 1.5% Bacto agar at 28°C. For selection solid media were supplemented with antibiotics (Brachmann et al., 2004). For growth assays 4 µl of serial tenfold dilutions of logarithmically growing cells starting with an OD₆₀₀ of 1 (Figure 1B and Supplementary Figure S2) or 0.1 (Figure 1D) were spotted on solid minimal yeast nitrogen based media (Difco) with a pH of 5.7 supplemented with 0.5% ammonium sulfate. The plates contained 1.5% Bacto agar and either 2% glucose or a mixture of 0.2% oleic acid and 0.05% Tween-40. Plates were incubated for two (Figure 1B and Supplementary Figure S2) to 3 days (Figure 1D) at 23°C. All assays were at least repeated three times with similar results.

Cells of an OD₆₀₀ of 1 were diluted to a starting OD₆₀₀ of 0.1 and inoculated into fresh yeast nitrogen based media (Difco) with a pH of 5.7 supplemented with 0.5% ammonium sulfate and 2% glucose in a volume of 100 µl in flat bottom 96 well plates. Growth assays were performed in a Synergy Mx multimode microplate reader (BioTek) at 23°C with high shaking setting. OD₆₀₀ was determined in 30 min intervals. Each strain was measured in five technical replicates and in three independent experiments. Mean values were plotted and original data and standard deviations are accessible in Supplementary Table S2.

A total of 200 µl of hot 1.5% agarose melted in water was used to create a thin agarose cushion on a 76 × 26 mm microscope slide (Roth). Cells from logarithmic growth phase incubated in yeast nitrogen based media (Difco) with a pH of 5.7 supplemented with 0.5% ammonium sulfate and 2% glucose were washed with water, concentrated fivefold, and 3 µl were spotted onto the middle of the agarose pad and covered with an 18 × 18 mm coverslip (Roth). Microscopy was performed on an Axiovert 200 M inverse microscope (Zeiss) equipped with a 1394 ORCA-ERA-CCD camera (Hamamatsu Photonics), filter sets for enhanced GFP (EGFP) and rhodamine (Chroma Technology), and a Zeiss 63×Plan Apochromat oil lens (NA 1.4). Single-plane bright field or phase contrast images and z-stacks of the cells (0.5 µm z-spacing) in the appropriate fluorescence channels were recorded, using the image acquisition software Volocity 5.3 (Perkin-Elmer). Images were processed and evaluated in ImageJ (Schneider et al., 2012). For protein localization analysis, single plain images or z-projections of deconvolved image stacks were used. Deconvolution was performed on the z-stacks by the ImageJ plugin DeconvolutionLab with 25 iterations of the Richardson – Lucy algorithm (Sage et al., 2017).

Sequences encoding the TPR domains of Pex5a and Pex5b were inserted into pGBKT7 (Matchmaker GAL4 Two-Hybrid System 3; Clontech) between the EcoRI and BamHI restriction sites via Gibson assembly. The ORFs for GFP or GFP modified with C-terminal dodecamers of Um01966, Um10665, Um11001, and Um03158 including PTS1 motifs were cloned into pGADT7 (Matchmaker GAL4 Two-Hybrid System 3; Clontech) between the EcoRI and BamHI restriction sites. Either pGBKT7-Pex5aTPR or pGBKT7-Pex5bTPR were co-transformed with one of the pGADT7 plasmids into YTS398, a derivative AH109 of deleted for pex5 (Stehlik et al., 2020). Three independent transformants of each of the 10 combinations were grown in liquid synthetic defined (SD) medium lacking leucine and tryptophan to an OD₆₀₀ of approx. 1. Cells were washed once with sterile water and 4 µl of fivefold or fiftyfold dilutions (OD₆₀₀ = 0.2 or 0.02) were spotted on solid SD medium lacking leucine and tryptophan as growth control, and on SD medium lacking leucine, tryptophan and histidine, and containing 1.5 mM 3-amino-1,2,4-triazole to test for protein – protein interaction. Plates were incubated for 3 days at 30°C.