Determining the targeting specificity of the selective peroxisomal targeting factor Pex9

Expanding the cargo range of Pex9 using a microscopy screen

To define motifs by aligning the PTS1 of Pex9 cargoes, it is important to have the biggest possible sample size. However, previous to this work only three Pex9 cargos were identified. Therefore, we sought to find additional proteins that can be targeted by Pex9. To this end, we performed a microscopic screen on a collection of ∼40 yeast strains representing a set of recently identified yeast peroxisomal proteins (Yifrach et al. 2021), which were never tested for targeting by Pex9 (Figure 1A). All proteins in this collection harbor a Green Fluorescent Protein (GFP) tag at the amino terminus (N′) for visualization and to allow their C′ to be exposed in case they have a PTS1. To examine the dependence of the peroxisomal proteins on Pex9, we used an automated mating procedure to insert several genetic traits into each strain: I) a peroxisomal marker, Pex3-mCherry, II) a deletion of the main PTS1 targeting factor (Δpex5) and III) a constitutive expression of PEX9 to enable visualization of its function in glucose-containing medium, as was done previously (Yifrach et al. 2016). We imaged the entire collection using a high content screening setup and looked for proteins that co-localize with peroxisomes when only Pex9 is expressed as a PTS1 targeting factor. Following analysis of all strains, we were able to identify two proteins that co-localized with peroxisomes in this condition – the Family of Serine Hydrolases 3 (Fsh3) protein, a newly identified peroxisomal lipase (Yifrach et al., 2021), and the Alpha-Factor Receptor regulator 1 (Afr1) protein, which is required for the formation of pheromone-induced projections in yeast (Konopka 1993) (Figure 1B).

Figure 1:

Expanding the cargo range of Pex9 using a microscopy screen and yeast 2-hybrid (Y2H) assays. (A) A query strain constitutively expressing Pex9 as the sole PTS1 targeting factor (Δpex5 + TDH3pr-PEX9) and a peroxisomal marker (Pex3-mCherry) was used to genomically integrate these traits into a yeast N′ GFP collection of recently identified peroxisomal proteins (Yifrach et al. 2021) by utilizing an automated mating procedure. Then, fluorescence microscopy was applied to identify proteins that co-localize with the peroxisomal marker when the cells grew on media containing either glucose or oleate as the carbon source. (B) Fsh3 and Afr1 both co-localize with peroxisomes when Pex9 is constitutively expressed and PEX5 is deleted, suggesting that they are newly identified cargos of Pex9. For all micrographs, a single focal plane is shown. The scale bar is 5 µm. (C) Y2H assays demonstrate that Fsh3 interacts with Pex9 in a PTS1-dependent manner, but Afr1 does not. (D) All Pex9-directed cargos show similar amino acid properties at positions −4 (hydrophobic residue L or I) and a negatively charged residue D in position -5 or -6 from the C′ of the protein.

To assess the capacity of Fsh3 and Afr1 to bind Pex9, we used a Yeast-2-hybrid (Y2H) assay. We found that Fsh3, but not Afr1, physically interacts with Pex9 (Figure 1C). The specific interaction between Pex9 and Fsh3 is suggested to be comparably weak or that Fsh3 abundance is low. This is demonstrated by colony formation, which takes up to ten days on a medium that contains adenine and 5 mM 3-amino triazole (3-AT) (for more information on this assay see materials and methods). Afr1 does not contain a canonical PTS1 at its C’ (the last six amino acids of the protein are FTHYLI), which altogether suggests that Afr1 may rely indirectly on Pex9. For example it could piggyback on another Pex9 cargo that contains a PTS1 similarly to a previously shown Pex5 cargo (Gabay-Maskit et al. 2020). Despite being weak, the binding of Fsh3 to Pex9 does depend on the PTS1 tripeptide of Fsh3 (Figure 1C). Put together, this suggests that Fsh3 is an additional direct cargo for Pex9. Interestingly, in the Y2H assay, neither Fsh3 nor Afr1 interacted with Pex5. This was unexpected since the microscopy analysis showed a reduction in peroxisomal localization for both proteins following the deletion of PEX5. This discrepancy between the two assays suggests that the Y2H is less sensitive than the microscopy assay.

After we expanded our cargo list, we could better assess unique sequence features. We compared the context of the PTS1 (the amino acids upstream to the PTS1 tripeptide) of the four Pex9 cargos Mls1, Mls2, Gto1, and Fsh3 (Figure 1D). We noticed that all cargos have a hydrophobic residue one amino acid upstream to the tripeptide (position −4 from the C′) and a negatively charged residue in position −5 or −6 from the C’. This similarity suggests that these features are important for recognition by Pex9.

Mutagenesis in the PTS1 context of known cargos uncovers Pex9 determinants

To explore whether the negatively charged residue in the PTS1 context of the Pex9 cargos plays an important role in their recognition, we substituted the charged residues at this position in various cargo proteins and looked for changes in the targeting ability. To visualize the effect of the PTS1 context, we constructed an integration plasmid containing a GFP fused at its C′ to the last 10 amino acids of Fsh3. We transformed this construct into the genome of a strain that constitutively expresses only Pex9, but not Pex5 (Δpex5 + TDH3pr-PEX9). We validated that the PTS1 motif of Fsh3 is sufficient to target the GFP to peroxisomes by Pex9 (Figure 2A, wtPTS1). Then, we mutated the negatively charged residue aspartate (D) in position −6 of the C′ to A, which has no charge. We indeed observed a reduction, but not a complete loss, of the peroxisomal localization of the protein (Figure 2A, D(-6 )A).

Figure 2:

Mutagenesis in the PTS1 context of known cargos uncovers Pex9 determinants. (A) Directed point mutagenesis was applied on an integration plasmid containing a GFP fused at its C′ to the last 10 amino acids of Fsh3 to substitute position −6 or −5 from aspartic acid (D) or serine (S) to alanine (A). While the substitution of D to A at position −6 showed reduced GFP localization to peroxisomes, the S to A substitution at position −5 showed moderately enhanced GFP localization in peroxisomes when the cells express Pex9 as the sole PTS1 targeting factor. For all micrographs, a single focal plane is shown. The scale bar is 5 µm. (B) Y2H assays show that a D to lysine (K) substitution in the PTS1 context of Fsh3 obliterates the interaction with Pex9, but a similar substitution of D to K in Mls1 and Mls2 does not affect the interaction with Pex9, suggesting that D in the context of the PTS1 is not the sole nor definitive determinant for Pex9 cargo recognition.

Furthermore, when we performed Y2H assays and mutated the D to a positively charged residue, K, it obliterated the interaction of Fsh3 with Pex9 (Figure 2B). However, when we mutated the D to K in the other cargos Mls1 and Mls2, the proteins could still interact with Pex9. These data suggest that the negatively charged residue in position −6 or −5 of the PTS1 plays a role in some cargo proteins but is not the sole nor definitive determinant for Pex9 cargo recognition.

Since factor-specific targeting requires that a cargo both binds Pex9 with enhanced affinity and that it binds Pex5 with reduced affinity, we also assayed the effect of the above-mentioned mutations on Pex5 binding. Interestingly, while the wild-type Mls1 and Mls2 (harboring a D at position −5) did not interact with Pex5, the D to K exchange in Mls1 and Mls2 promoted the interaction with Pex5 (Figure S1). These observations are in agreement with previous data showing that positively charged residues in the PTS1 context enhance targeting by Pex5 (DeLoache, Russ and Dueber 2016). We speculate that the D in position −5 or −6 of the Pex9 cargos serves to reduce capture by Pex5, rather than to enhance the binding to Pex9.

Next, we tested whether a mutation closer to the PTS1 tripeptide affects the targeting ability of Pex9. We changed S to A in position −5 of Fsh3 and checked how well the GFP-PTS1 (last 10aa of Fsh3) localizes to peroxisomes. Although the basal targeting of the native PTS1 construct was already quite good, we were able to observe moderately-enhanced peroxisomal localization (which we define as targeting) in the mutated construct compared to the WT PTS1 (Figure 2A, S(-5)A). S is a polar amino acid, while A is non-polar. This suggests that Pex9 prefers a non-polar amino acid at a position closer to the PTS1 tripeptide.

Exploring Pex9 targeting specificity using natural non-binders

Intrigued by the ability to enhance import by Pex9, we decided to mutate the PTS1 motifs of a cargo protein that does not normally use Pex9 to potentially induce import (non-binders). We reasoned that this would facilitate visualization as there will be a sudden gain of targeting from no initial targeting. To start with a Pex9 non-binding PTS1, we first tested the last ten amino acids of several proteins whose full-length version is Pex9 independent (they are not localized to peroxisomes when Pex9 is expressed and Pex5 is absent). We focused on three cargo proteins with a high priority to Pex5, Malate Dehydrogenase 3 (Mdh3), Carnitine Acetyltransferase 2 (Cat2), and Lysine requiring 1 (Lys1) (Rosenthal et al. 2020), and fused their last ten amino acids to the C′ of a GFP. Then, we integrated these constructs into an inert locus in the yeast genome. While the full-length proteins are not Pex9 cargos (Figure S2 and (Yifrach et al. 2016), surprisingly, we found that the fusion of only the last ten amino acids of both Mdh3 and Cat2 was sufficient to co-localize GFP to peroxisomes in a Pex9-dependent manner (Figure 3A and B). We used a Y2H assay to validate the microscopy-based observation and showed that indeed the full-length Mdh3 does not interact with Pex9, while a peptide consisting of only the last ten amino acids of Mdh3 does (Figure 3C). Conversely, we found that the full-length Pex9 cargos, Fsh3, and Mls1, do not interact with Pex5 in Y2H, but peptides consisting of only their last ten amino acids do (Figure 3C). This striking observation suggests that additional domains in the full protein prevent, or reduce, the interaction with both Pex5 and Pex9 and block targeting even when the PTS1 by itself could enable binding.

Figure 3:

Exploring Pex9 targeting specificity using natural non-binders. Integration plasmids containing a GFP fused at its C′ to the last 10 amino acids of (A) Mdh3 or (B) Cat2, show that the last 10 amino acids of both Mdh3 and Cat2 were sufficient to co-localize the GFP to peroxisomes in a Pex9-dependent manner, although the full-length proteins are not cargos of Pex9. The asterisk indicates that GFP-last 10aaCat2 Control (in B) had a higher signal compared to the other images in the panel, hence a different image contrast was used. (C) Y2H assays validate that the last 10 amino acids of Mdh3 (aa334–343) can interact with Pex9 despite the fact that the full-length protein does not. Similarly, the last 10 amino acids of the Pex9 cargos Fsh3 (aa257–266) and Mls1 (aa545–554) interact with Pex5, more strongly than the full-length proteins. This suggests that additional parameters in the full protein prevent, or reduce, the interaction with the inappropriate targeting factor. (D) An integration plasmid containing a GFP fused at its C′ to the last 10 amino acids of Lys1 shows that the native PTS1 of Lys1 was not sufficient to co-localize the GFP to peroxisomes in a Pex9-dependent manner. Directed mutagenesis substitution of positions −5 and −4 to the negatively charged residue D and/or a non-polar residue leucine (L) show that the L mutation enabled a weak GFP peroxisomal localization, demonstrating that a single non-polar residue at position −4 is sufficient to mediate Pex9 specificity. For all micrographs, a single focal plane is shown. The scale bar is 5 µm.

We continued to work on the last ten amino acids of Lys1, which were not sufficient to mediate Pex9-dependent targeting (Figure 3D, upper panel). To enhance Pex9 binding, we replaced the amino acids at positions −5 and −4 with the negatively charged residue D and a non-polar residue L, either alone or in combination. While the D mutation had no visible effect on the localization of the GFP, the peptide with the L mutation showed a weak peroxisomal signal, demonstrating that a single amino acid change to a non-polar residue at position −4 is sufficient to allow targeting by Pex9.

Overall, our data imply that the PTS1 targeting specificity of Pex9 is largely dependent on the presence of a non-polar, hydrophobic residue flanking the PTS1 tripeptide. Moreover, we suggest that the negatively charged residue at positions −5 or −6 of the PTS1 has two roles – to enhance the binding of specific cargo proteins to Pex9 and to prevent binding to Pex5.

Pex9 molecular modeling support hydrophobicity in the PTS1 binding area as a major determinant of Pex9 binding

Our targeted approach highlighted the preference of Pex9 for a hydrophobic residue preceding the PTS1 tripeptide. This distinguishes Pex9 from Pex5, which is known to prefer positively charged residues in this context (DeLoache, Russ and Dueber 2016). To uncover the structural features that differentiate the two targeting factors in their binding preference, we constructed a model structure of the Pex9 PTS1-binding domain in a similar manner to the modeling of Pex5 that we described previously (Gabay-Maskit et al. 2020) (Figure 4A). The models highlight that the Pex5 surface around the PTS1-binding cavity has large negative electrostatic potential patches (red), while the surface of the cognate Pex9 regions appears more hydrophobic (white). This is especially obvious for the shallow cavity that binds the amino acid in position −2 (black arrows) and the bottom of the PTS1 binding cavity (Figure 4A). In addition, Pex9 has several positive electrostatic patches (blue) at the edge of the peptide-binding cavity.

Figure 4:

Pex9 molecular modeling support hydrophobicity in the PTS1 binding area as a major determinant of Pex9 binding. (A) Molecular modeling of the PTS1 binding domains of Pex9 and Pex5 indicates that the Pex5 surface around the PTS1 binding cavity has large negative electrostatic potential patches (red), while the surrounding area of the PTS1 cavity of the cognate Pex9 regions appears mostly hydrophobic (white) with several positive electrostatic patches (blue) at the edge of the peptide-binding cavity. Black arrows point to the shallow cavity that binds peptide residue −2. (B) Molecular dynamics simulations of complexes of Pex9 with peptides consisting of six C′ amino acids of the known Pex9 cargo Mls2 show that the sidechain of the peptide’s −4 residue points toward the bottom of the PTS1 binding cavity and makes tight contacts with hydrophobic Pex9 residues (all other Pex9 cargo behaved in the same manner). (C) A detailed snapshot from the simulations depicts the contacts between Pex9 residues and the Mls1 L-4 sidechain. Legend for atom coloring on the bottom left.

We then performed Molecular Dynamics (MD) simulations of a cargo factor (Pex5 or Pex9) with peptides consisting of six C′ amino acids of four PTS1 proteins that are Pex9 cargo: Mls1, Mls2, Gto1, and Fsh3. We previously showed that the binding stability of a peptide to Pex5 could be estimated from the constancy of hydrogen bonds (H-bonds) between peptide backbone atoms and specific Pex5 residues, particularly H-bonds of peptide residues −1 and −3 (Yifrach et al. 2021). The Pex5 residues that form these H-bonds are conserved between Pex5 and Pex9 supporting the importance of the C′ tripeptide for binding. The context residues, however, behaved differently in the simulations for Pex5 and Pex9 complexes. In the starting structures of all the complexes, the side chain of peptide residue −4 pointed outwards, making either no, or little, contact with the cargo factor, as seen in the experimental structure that was used as the modeling template (Gatto et al. 2000). While in the Pex5 simulations this side chain remained mostly exposed, in the Pex9 simulations the peptide changed conformation within 50–100 ns and the sidechain of its −4 residue pointed toward the bottom of the PTS1 binding cavity (Figure 4B and Figure S3). This hydrophobic sidechain (L or isoleucine (I)) made tight contacts with hydrophobic Pex9 residues that replace more polar residues of Pex5 (e.g. tyrosine (Y) 410, and A437 in Pex9 versus S507 and S534 in Pex5). Together with additional residues, they form a more hydrophobic surface at the entrance to the peptide-binding cavity of Pex9 and form hydrophobic contacts with L or I at position −4 of Pex9 cargos (Figure 4C). These results strongly support the notion that Pex9 cargos are selected by the hydrophobic residue in position −4 of the PTS1.

A systematic screen shows a correlation between Pex9-dependent peroxisomal localization and hydrophobicity of residue −4 of PTS1 sequences

To validate our findings in an unbiased manner and to investigate more variations of the PTS1 context that may favor Pex9, we took advantage of a library of integration plasmids containing randomized sequences preceding the canonical PTS1 tripeptide S-K-L residues (DeLoache, Russ and Dueber 2016). This library was previously used to find the optimal Pex5 cargo sequence–the enhanced PTS1 import sequence. Each integration plasmid in this pooled library contains a Yellow Fluorescent Protein (YFP) linked at its C′ to a variable sequence of six amino acids (XXXXXX) preceding the SKL tripeptide. We genomically integrated this YFP-XXXXXX-SKL library into a strain constitutively expressing Pex9 as the sole PTS1 targeting factor and picked single colonies into an arrayed format (Figure 5A). Then, we imaged each strain and sequenced the contained plasmid to match the ratio of the YFP-XXXXXX-SKL peroxisome/cytosol localization to the identity of the variable amino acid sequence. We found a wide range of peroxisome/cytosol localization ratios, from very high to very low. Despite redundancy in sequences, we could retrieve 15 distinct linkers that correspond to the wide range of phenotypes (Figure 5B). We manually validated two sequences from the screen by fusing them to the C′ of GFP (Figure S4). We looked for potential determinants in the variable sequences that will explain the different peroxisome/cytosol localization ratios. We tested the various positions as well as many size/hydrophobicity/charge aspects. The best correlation was achieved when the peroxisome/cytosol ratio was plotted against the hydrophobicity score of the residues at position −4 (Figure 5C). The hydrophobicity scores were taken from an averaged scale built from 43 different hydrophobicity studies (Trinquier and Sanejouand 1998), and gave the best correlation relative to any singular hydrophobicity scale examined. The analysis validates our hypothesis that Pex9 cargos are selected by the hydrophobic residue in position −4 of the PTS1.

Figure 5:

A systematic screen shows a correlation between Pex9-dependent peroxisomal localization and hydrophobicity of residue −4 of PTS1 sequences. (A) A pooled library of plasmids containing YFP fused to a stretch of six random amino acids followed by the tripeptide SKL was transformed into a strain constitutively expressing Pex9 as the sole PTS1 targeting factor (Δpex5 + TDH3pr-PEX9) and a peroxisomal marker (Pex3-mCherry). Cells from single colonies were imaged and the PTS1 context of each plasmid was sequenced. (B) Fifteen distinct PTS1 context sequences (the random stretches) were identified and matched to the peroxisome/cytosol localization ratio of their cognate YFP construct. The white arrow is pointing at a peroxisome. For all micrographs, a single focal plane is shown. The scale bar is 5 µm. (C) Amino acid analysis shows that the YFP peroxisome/cytosol localization ratio correlates with the hydrophobicity score of the residue at position −4 from the C′ of each construct. This combined hydrophobicity rank (Trinquier and Sanejouand 1998) gives low values for hydrophobic residues (most φ) and high values for hydrophilic residues (least φ). The analysis supports our hypothesis that Pex9 cargos are selected by the hydrophobic residue in position −4 of the PTS1.

Yeast strains and strain construction

All strains in this study are based on the BY4741 laboratory strain (Brachmann et al. 1998) except for the PJ69-4a that were used for the Y2H assays (James et al. 1996). See the complete list of yeast strains and primers in Table S1. Cells were genetically manipulated using a transformation method that includes the usage of lithium-acetate, polyethylene glycol, and single-stranded DNA (Daniel Gietz and Woods 2002). A pYM-based plasmid (Janke et al. 2004) was modified to contain the last 10 aa of different PTS1 proteins at the C′ of the GFP sequence. Point mutations were introduced using restriction-free cloning. Plasmids are described in Table S2. Constructs were genomically integrated into the HO locus in strains containing Pex3-mCherry, with or without pex5 deletion and constitutive PEX9 expression. Primers for validation of correct insertion were designed using the Primers-4-Yeast website (Yofe and Schuldiner 2014).

Yeast growth media

Synthetic media used in this study contains 6.7 g/L yeast nitrogen base with ammonium sulfate (Conda Pronadisa #1545) and 2% glucose, with complete amino acid mix (oMM composition, Hanscho et al. 2012), unless written otherwise. When Hygromycin or Geneticin antibiotics were used, media contained 0.17 g/L yeast nitrogen base without ammonium sulfate (Conda Pronadisa #1553) and 1 g/L of monosodium glutamic acid (Sigma-Aldrich #G1626) instead of yeast nitrogen base with ammonium sulfate. When mentioned, 500 mg/L Hygromycin B (Formedium), 500 mg/L Geneticin (G418) (Formedium), and 200 mg/L Nourseothricin (Silcol Scientific Equipment LTD) were used.

Yeast library preparation using a synthetic genetic array (SGA)

To create collections of haploid strains containing GFP-tagged proteins with additional genomic modification (i.e. Pex3-mCherry (a peroxisomal marker), PEX5 deletion (Δpex5), and PEX9 constitutive-expression (TDH3pr-PEX9)), a query strain was constructed based on an SGA compatible strain (for further information see Table S1). Using the SGA method (Cohen and Schuldiner 2011; Tong and Boone 2006) the query strain was crossed with a collection of strains from the SWAT N′-GFP library (Yofe et al. 2016; Weill et al. 2018) containing ∼40 strains of recently-identified peroxisomal proteins (Yifrach et al. 2021) together with controls. To perform the SGA in an arrayed format, we used a RoToR benchtop colony arrayer (Singer Instruments). In short: mating was performed on rich medium plates, and selection for diploid cells was performed on SD-URA plates containing Nourseothricin, Hygromycin, and Geneticin antibiotics. Sporulation was induced by transferring cells to nitrogen starvation media plates for 7 days. Haploid cells containing the desired mutations were selected by transferring cells to SD-URA plates containing the same antibiotics as for selecting diploid cells, alongside the toxic amino acid derivatives 50 mg/L Canavanine (Sigma-Aldrich) and 50 mg/L Thialysine (Sigma-Aldrich) to select against remaining diploids, and lacking Histidine to select for spores with an A mating type.

Yeast library preparation from a library of pooled plasmids

A library of pooled Venus-PTS1 plasmids (DeLoache, Russ and Dueber 2016) was transformed genomically into a strain containing Pex3-mCherry as a peroxisomal marker, Δpex5, and TDH3pr-PEX9. 288 single colonies were picked to a 384-well plate containing SD-URA liquid media supplemented with Nourseothricin, Hygromycin, and Geneticin for selection. Then, the collected strains were imaged using automated fluorescence microscopy. In parallel, the DNA of each strain was extracted by dissolving in 20 mM NaOH followed by boiling at 95 °C for 20 min in a PCR machine. Then, a 270 bp DNA containing the variable PTS1 region was amplified using a PCR reaction with appropriate primers (F–cgaaaagagagatcacatgg, R–gaaagcaacctgacctacag). The amplified DNA was cleaned using a GenElute 96 Well PCR Clean-up kit (Sigma-Aldrich) and sent for sequencing.

Automated fluorescence microscopy

The collections (∼40 strains of NOP1pr-GFP-recently identified peroxisomal proteins, Figure 1 and 288 strains with YFP-XXXXXX-SKL, Figure 5) were visualized using an automated microscopy setup: cells were transferred from agar plates into 384-well polystyrene plates for growth in liquid media using manual handling. Liquid cultures were grown in a LiCONiC incubator, overnight at 30 °C in an SD-URA medium. An EVO freedom liquid handler (TECAN) connected to the incubator was used to dilute the strains to an OD₆₀₀ of ∼0.2 into plates containing SD medium (6.7 g/L yeast nitrogen base and 2% glucose) supplemented with –URA amino acids. For the ∼40 NOP1pr-GFP strains we performed an additional screen in S-oleate (6.7 g/L yeast nitrogen base, 0.2% oleic acid, and 0.1% Tween-80) supplemented with –URA amino acids. Plates were incubated at 30 °C for 4 h in SD medium or for 20 h in S-oleate. The cultures in the plates were then transferred by the liquid handler into glass-bottom 384-well microscope plates (Matrical Bioscience) coated with Concanavalin A (Sigma-Aldrich). After 20 min, wells were washed twice with SD-Riboflavin complete medium (for screens in glucose) or with double-distilled water (for the screen in oleate) to reduce autofluorescence, remove non-adherent cells, and obtain a cell monolayer. The plates were then transferred to the ScanR automated inverted fluorescence microscope system (Olympus) using a robotic swap arm (Peak Robotics). Images of cells in the 384-well plates were recorded in the same liquid as the washing step at 24 °C using a 60× air lens (NA 0.9) and with an ORCA-flash 4.0 digital camera (Hamamatsu). Images were acquired in two channels: YFP (excitation at 488 nm, emission filter 525/50 nm) and mCherry (excitation at 561 nm, emission filter 617/73 nm). Image analysis was performed manually using ImageJ software.

Manual microscopy

Manual microscopy imaging was performed for strains with NOP1pr-GFP-last 10 aa of Fsh3 (Figure 2) Mdh3, Cat2, and Lys1 (Figure 3). Yeast strains were grown as described above for the high-throughput microscopy with changes in the selection required for each strain (see yeast strain information in Table S1). Imaging was performed using the VisiScope Confocal Cell Explorer system, composed of a Zeiss Yokogawa spinning disk scanning unit (CSU-W1) coupled with an inverted Olympus microscope (IX83; x60 oil objective; Excitation wavelength of 488 nm for GFP). Images were taken by a connected PCO-Edge sCMOS camera controlled by VisView software.

Yeast two-hybrid assay to assess protein-protein interactions

PJ69-4A cells (James, Halladay and Craig 1996) were transformed with plasmids derived from pPC86 (GAL4-activation domain, AD (‘Prey’)) and pPC97 (GAL4-DNA-binding domain, BD (‘Bait’)) (Chevray and Nathans 1992; Kerssen et al. 2006), containing genes encoding proteins of interest. Transformed cells were selected on YNBG (0.17% [w/v] yeast nitrogen base without amino acids, 0.5% [w/v] ammonium sulfate, amino acids according to auxotrophic requirements, pH 6.0) plates lacking leucine (leu) and tryptophan (trp) containing 2% [w/v] glucose (YNBG). Clones were streaked onto YNBG-trp-leu (control), YNBG -trp-leu-his-ade, and/or YNBG -trp-leu-his + 5 mM 3-amino triazole (3-AT) plates and incubated for 3, 7, or 10 days at 30 °C, respectively. The reporter genes HIS3 and ADE2 are under the control of GAL1 or GAL2 promoters, respectively. Thus, they are only expressed when Gal4-AD and Gal4-BD of the bait and prey fusion proteins are in close proximity due to protein-protein interaction. Since the absence of adenine is a very stringent selection condition (James, Halladay and Craig 1996), weak interactions were investigated in the presence of adenine. However, to suppress unspecific leakage of the growth phenotype upon longer incubation on plates lacking histidine, the competitive His3-inhibitor 3-AT was added to avoid false-positive results. Overall, the addition of 3-AT and longer incubation times, when compared to the control plates, were chosen to visualize weak protein-protein interactions by cell growth on plates lacking histidine or adenine.

Molecular modeling and molecular dynamics (MD) simulations

Model structures of Pex9’s TPR (PTS1 binding) domain, residues 287–484, with bound 6 amino-acid peptides were constructed based on the experimental structures of human Pex5 complexes. The sequence identity for the TPR domain was only 29% to the human Pex5 TPR domain structure in Gatto et al. 2000 (PDB entry 1FCH) but it was spread along the whole sequence. The starting model of yeast Pex9 was constructed using Modeller (Šali and Blundell 1993) as implemented in UCSF-Chimera (Pettersen et al. 2004). The position of the cargo peptide in the starting models was based on the structure of human Pex5-peptide complex 1FCH. As a result, the sidechain of residue −4 of the peptide pointed away from and made no direct contacts with Pex9. We explored the stability of the Pex9/peptide complexes using molecular dynamics. Each starting model was immersed in a box of water, neutralized and energy minimized. Two trajectories of 300 ns each were calculated for every model complex, and frames were extracted at 5 ns intervals and inspected manually. MD simulations were executed with the Gromacs package (Van Der Spoel et al. 2005). The minimum distance between peptide and Pex9 residues was calculated using the Gromacs package analysis options. UCSF-chimera was used to visualize frames from the Pex9/peptide trajectories and to produce Figure 4.