GeneCluster was used to organize the filtered dataset (2,300 genes) into a two-dimensional 6 × 4 SOM (SOM1). The clusters were then organized into a one-dimensional SOM using Cluster software for display (top). The mean log₁₀ expression ratio for each cluster is represented by a shaded spot. The cluster numbers and genes per cluster of the two-dimensional SOM are at the left. Genes involved in lipid, fatty acid, and sterol metabolism and genes encoding peroxisomal proteins are enriched in cluster 2. A second SOM (SOM2) was generated the same way (bottom) except the log₁₀ expression ratios for each gene were first normalized to have a mean of zero and a standard deviation of one. PEX genes are enriched in cluster 4.

Unsupervised hierarchical clustering. Clustered display of gene expression profiles generated using Cluster and Tree view software with normalized (A) and nonnormalized (B) data. Columns represent the expression ratios for experiments 1–8 (Table I). Rows represent log₁₀ expression ratios of individual genes, with red and green indicating positive and negative ratios, respectively. All genes in the filtered dataset are represented at the left. Candidate genes are those in branches containing PEX genes or genes encoding exclusively peroxisomal proteins, in which all pair-wise correlation coefficients are >0.95. Regions of the tree containing selected branches (red text) are magnified, and genes that encode peroxisomal proteins and/or proteins involved in lipid, fatty acid, and sterol metabolism are marked with an asterisk. Bullets mark YOR084w and YPL112c (renamed PEX25).

Nearest Neighbor analysis. The expression profiles of PEX genes and genes encoding exclusively peroxisomal proteins were compared with the expression profiles of all other genes in the filtered dataset using Nearest Neighbor analysis (GEAP; Galitski et al., 1999). Neighboring genes meeting the correlation coefficient/geometric distance cutoffs outlined in the Materials and methods were selected as candidate genes. Queried genes with at least two neighbors are at the top, and neighboring candidate genes are indicated by shaded boxes. YOR084w and YPL112c (renamed PEX25) are underlined.

Yor084-pA and Pex25-pA are induced by oleate. Yeast strains were grown in oleate, and lysates were prepared at the indicated times. Samples containing equal protein were analyzed by Western blotting to visualize the protein A fusions. Polyclonal antibodies directed against Gsp1p (Leslie et al., 2002) were used to confirm equal protein loads.

Cellular distributions of Pex25p and Yor084p. (A) The DsRed-PTS1 reporter colocalizes with Fox2–GFP chimera to peroxisomes. (B) The subcellular distributions of protein A chimeras were compared with that of DsRed-PTS1 in oleate-induced haploid cells (Pex11-pA and Pex25-pA) and heterozygous diploid cells (Yor084-pA) by double-labeling immunofluorescence microscopy. Bar, 10 μm.

Pex25p-pA is a peroxisomal peripheral membrane protein, and Yor084-pA is a peroxisomal matrix protein. (A) A postnuclear supernatant was separated by centrifugation into a supernatant (20kgS) enriched for cytosol and a pellet (20kgP) enriched for peroxisomes and mitochondria. The 20kgP was treated with Ti8 to lyse peroxisomes releasing matrix proteins (Ti8S). The membrane-containing pellet fraction (Ti8P) was treated with Na₂CO₃ to separate peripherally associated membrane proteins (CO₃S) from integral membrane proteins (CO₃P). Western blotting detected chimeras shown and included Pex11p (peripheral) and Pex13p (integral) as controls. (B) Pex25-pA and Yor084-pA cofractionated with peroxisomes. Organelles in the 20kgP were separated by isopycnic centrifugation. Proteins were identified in each fraction by Western blotting, and peak peroxisomal (P), and mitochondrial (M) fractions (arrows) were identified with antibodies against Fox2p and Sdh2p, respectively.

Subcellular localization of PTS1- and PTS2-targeted proteins in yor084wΔ and pex25Δ cells. DsRed-PTS1 and PTS2–GFP reporters were examined by fluorescence microscopy. DsRed-PTS1 localized to numerous small peroxisomes in pex7Δ and yor084Δ yeast cells. A similar localization was seen with PTS2–GFP in pex5Δ and yor084Δ cells. A diffuse cytoplasmic localization of these reporters was observed in strains lacking components necessary for the targeting of either PTS1 (pex5Δ) or PTS2 (pex7Δ). In pex11Δ cells, both reporters accumulated in a few large peroxisomes as reported previously (Erdmann and Blobel, 1995). In pex25Δ cells, a heterogeneous population was observed; some cells displayed a diffuse localization of the reporters, and some cells showed accumulation of the reporters to a few large peroxisomes. Bar, 10 μm.

Morphological analysis of pex25Δ cells. Ultrastructure of wild-type (A and B) and pex25Δ (C–I) cells. Thin sections of cells prepared after 8 h induction in oleate. Note the heterogeneity in the peroxisome population in pex25Δ cells: large sized, (C–F); normal sized (G and H); and apparently absent (I). P, peroxisome; M, mitochondrion; L, lipid droplet; N, nucleus; V, vacuole. Bar, 1 μm.