Characterization of proteome dynamics during growth in oleate reveals a new peroxisome-targeting receptor

Peroxisomes are found in almost all eukaryotes and participate in central catabolic pathways such as β-oxidation of fatty acids, breakdown of amino acids and nucleotides, and detoxification of reactive oxygen species (Islinger et al., 2010; Smith and Aitchison, 2013). It is therefore obvious why any change in peroxisome functions leads to severe diseases in humans (Waterham et al., 2016). The genetically powerful yeast Saccharomyces cerevisiae has served as an attractive model to study peroxisomes owing to the conservation of pathways required for biogenesis of the organelle, targeting of proteins and metabolic functions.

In S. cerevisiae (from here on termed yeast), peroxisomes are the sole organelles that perform β-oxidation of fatty acids making them essential for growth when oleic acid is the sole carbon source. Given that yeast usually grow while consuming sugar as a main carbon source, they must dramatically alter cellular organization to rely on lipids (Jung et al., 2013; Smith et al., 2002). Among other changes, this includes optimizing peroxisomal functions as well as the functions of other organelles, such as mitochondria, where the metabolic products of the β-oxidation process are consumed (Trotter, 2001). Indeed, it has been shown that during a transition to growth in oleate, yeast actively alter their transcriptional response (Karpichev and Small, 1998; Smith et al., 2002, 2007) as well as the localization of several proteins (Jung et al., 2013). However, little is known about the extent of the organellar and proteomic changes that occur during the transition to oleate-dependent growth.

To further uncover the extent of organelle and protein dynamics during growth in oleate, we followed the localization of nearly all yeast proteins – tagged at either their N or C terminus (N′ or C′ respectively) with GFP (Huh et al., 2003; Yofe et al., 2016). We identified dramatic cellular changes, such as alterations in organelle shapes and sizes as well as numerous changes in protein localizations. Furthermore, we identified several new potential peroxisomal proteins and demonstrated that Ymr018w, that we now name Pex9, is a peroxisomal-targeting receptor that targets a subset of proteins that contain a peroxisomal targeting signal 1 (PTS1) motif under specific conditions. We suggest that such a dedicated targeting receptor also exists in vertebrates. Our findings highlight the complexity of the machinery targeting proteins to peroxisomes and provide an important step towards a deep understanding of peroxisomes, their roles and regulation under various metabolic conditions.

To characterize the proteome dynamics of yeast grown in oleate as a sole carbon source, we performed a high-content microscopic screen of collections of strains containing GFP-tagged proteins (Breker et al., 2014). An imaging-based screen can be easily used to detect the organellar localization of GFP-tagged proteins while avoiding the requirement for subcellular fractionation of each organelle followed by mass spectrometry analysis (Breker and Schuldiner, 2014) which is laborious and might suffer from organelle contaminations. To ensure that we obtained an optimal coverage of all yeast proteins, we used two complementary yeast libraries; the C′ GFP library (Huh et al., 2003) and the N′ SWAT-GFP library (Yofe et al., 2016). The C′ GFP library encompasses all yeast proteins expressed under their native promoter, enabling us to follow changes both in transcriptional and post-transcriptional levels. The N′ SWAT-GFP library contains ∼1800 strains focusing on endomembrane and peroxisomal proteins – all of which are expressed under a constitutive SpNOP1 promoter, which enables us to detect the localization of low abundance or condition-specific proteins. Importantly, tagging proteins at either terminus can mask targeting sequences as well as regulatory sequences that might affect their localization. Therefore, it is essential to utilize both libraries for increased accuracy. One relevant example is the peroxisomal matrix proteins, which contain a PTS1 motif at their C-terminus and are thus only correctly localized in the SWAT library (Yofe et al., 2016).

To identify the proteome changes that occur during growth in oleate compared to in glucose, we imaged the strains after 20 h of growth in medium supplemented with oleate as the sole carbon source (Fig. 1A). The localizations obtained were then compared to the previously described localizations in glucose (Breker et al., 2014; Huh et al., 2003; Yofe et al., 2016). Most of the peroxisomal proteins were observed in the expected punctate pattern only when the SWAT library (Yofe et al., 2016) was used. Hence, in order to differentiate between a peroxisomal localization and other punctate localizations (lipid droplets, Golgi, organelle contact sites etc.) we integrated a peroxisomal marker, Pex3–mCherry, into the SWAT strains (Fig. 1A).

Following our analysis, we observed that growth in oleate led to tremendous changes in the cell including changes in the abundance, size and shape of different organelles. The number of peroxisomes was increased (Fig. 1B, left), as has been previously reported (Veenhuis et al., 1987), the size of lipid droplets was increased (Fig. 1B, middle; see Fig. S1A–C for a quantitative analysis) and mitochondrial morphology was dramatically altered (Fig. 1B, right; see Fig. S1D,E for a quantitative analysis). Interestingly, we found that Fzo1, which is involved in mitochondrial fusion, lost its defined localization during growth in oleate (Fig. S1F). Additionally, we could detect changes in localization that imply downregulation of DNA replication and budding, in line with the general stress that cells undergo when grown in oleate (Fig. 1C).

One striking alteration in localization was the accumulation of proteins in the vacuole when cells were grown in oleate but not in glucose (Fig. 1D), suggesting that many proteins, such as sugar transporters, that have no role during oleate growth, are degraded under this condition.

It is known that dramatic transcriptional changes occur during the transition to growth on oleate. In concordance with this, we found transcription factors that we could detect in the nucleus only during growth in oleate (Fig. 1E). One such factor was Adr1 (Fig. 1E, left), which enhances the expression of peroxisomal genes (Gurvitz et al., 2001; Karpichev et al., 2008). Another factor that changed localization was Opi1 (Fig. 1E, middle), whose activity is known to be regulated by its localization. Interestingly, Opi1 has been previously shown to move from the nuclear endoplasmic reticulum (ER) to the nucleus, as we saw, under conditions in which there is an excess of fatty acids in the yeast cell (Velazquez et al., 2016). However, we could not detect changes in peroxisomal size or abundance upon OPI1 deletion (data not shown); thus, it most probably regulates refined changes in cellular metabolism. Another protein, Usv1, which affects the transcriptional regulation of genes that are involved in growth on non-fermentable carbon sources, was also observed in the nucleus during growth in oleate (Fig. 1E, right).

Overall, we found that 719 proteins changed localization in the N′ library (Table S2; Fig. 1F) and 461 in the C′ library (Table S3 and Fig. 1G) when cells were grown in oleate compared to growth in glucose. Only 25 cellular changes overlapped between both GFP libraries emphasizing the power of using these complementary collections. Interestingly, the most frequent changes in localizations were the loss of a defined localization or movement to the vacuole (Fig. S2). These cellular changes suggest that many proteins are either downregulated or degraded when cells are grown in oleate. The large number of changes in protein localization highlights the dramatic cellular reorganization required to enable yeast to rely on lipids as a sole carbon source.

We next focused on changes in peroxisomes, the sole organelle in yeast that performs β oxidation (Trotter, 2001). It has been previously suggested that the protein content of peroxisomes is highly dynamic (Jung et al., 2013; Smith and Aitchison, 2013); hence, we first examined which known peroxisomal proteins changed their localization during growth in oleate (Table S4). One nice example of the power of using two complementary libraries comes from Pxa2, a subunit of the peroxisomal fatty acid transporter. Whereas Pxa2–GFP that was expressed under its native promoter was only observed during growth in oleate (Fig. 2A, left), the constitutively expressed GFP–Pxa2 could be visualized in peroxisomes in both conditions (Fig. 2A, right), supporting the fact that Pxa2 levels are transcriptionally regulated (Karpichev and Small, 1998).

Having integrated a Pex3–mCherry peroxisomal marker to the N′ GFP library (Fig. 1A), we were able to identify proteins that were colocalized with peroxisomes during growth in oleate, but had not been reported before as peroxisomal proteins. We could detect that Mtc4, a protein with an unknown function, was mainly colocalized with Pex3 in oleate but not in glucose (Fig. 2B, left). Intriguingly, we found that Mtc4 colocalized with lipid droplets in glucose (Fig. 2B, right). Given that we did not see a general overlap between peroxisomes and lipid droplets during growth in oleate (Fig. S3A), the change observed in Mtc4 implies that at least a fraction of Mtc4 is targeted to peroxisomes in oleate, plausibly to improve lipid consumption, as was previously suggested (Binns et al., 2006).

Hem14, a mitochondrial enzyme that catalyzes the seventh step in the heme biosynthetic pathway, was localized to mitochondria in glucose and was observed as puncta in oleate (Fig. 2C). Given that mitochondria were altered during growth in oleate and often appeared more punctate, some of these puncta might simply be mitochondria. However, some of the GFP–Hem14 puncta colocalized with Pex3. The unique localization of Hem14 suggests that either Hem14 is dually localized to mitochondria and peroxisomes in oleate, or that a subset of mitochondria that are enriched in Hem14 are localized in the vicinity of peroxisomes (Cohen et al., 2014; Shai et al., 2016). Currently, we cannot differentiate between these possibilities due to the limitations of fluorescence microscopy.

Finally, Rmd5, a component of the glucose-induced degradation deficient (GID) complex that degrades gluconeogenic enzymes (Braun et al., 2011), colocalized with Pex3 in oleate (Fig. 2D). This suggests that gluconeogenic enzymes might be quality controlled proximally to peroxisomes.

We found several proteins that had not been previously suggested to be peroxisomal proteins colocalized with Pex3 during growth in both glucose and oleate. One such protein, Bet4, is the α subunit of the type II geranylgeranyltransferase (Fig. 2E). Although we did not detect Bet2, the β subunit, in peroxisomes (data not shown), we further found that Bet4 localization to peroxisomes was dependent on the PTS1-targeting receptor Pex5 (Fig. S3B). This was not expected because Bet4 does not contain a canonical PTS1 sequence. Hence, it would be intriguing to further examine whether Bet4 is indeed localized to the matrix of peroxisomes and if so, whether it directly binds to Pex5 in a PTS1-independent manner or whether it is inserted to the matrix by piggybacking on a PTS1-containing protein. Moreover, it now remains to be determined whether Bet4 has a functional role in peroxisomes or whether one subunit is simply sequestered there under certain conditions.

Another newly identified peroxisomal protein that came up from our screen is Ymr018w, which colocalized with Pex3 during growth in glucose and in oleate (Fig. 2F). This protein was suggested to be a unique paralog of Pex5 in S. cerevisiae (Kiel et al., 2006) as it displays 27% identity to the yeast Pex5 and 25% identity to the human Pex5 (Amery et al., 2001). Additionally, transcriptome profiling of yeast has pointed out Ymr018w as having a role in peroxisome biogenesis or function (Smith et al., 2002). Interestingly, we could still detect GFP–Ymr018w colocalizing with peroxisomes when both the targeting receptors PEX5 and PEX7 (the targeting receptor for PTS2-containing proteins), were deleted (Fig. S3C). This suggests that the targeting, and therefore maybe also the function of Ymr018w, is independent of Pex5 and Pex7. Intrigued by its potential role in targeting proteins to peroxisomes, we decided to further investigate Ymr018w.

To examine whether Ymr018w is targeting proteins to the matrix of peroxisomes we first deleted YMR018W, PEX5 or PEX7 and examined whether any of the N′ GFP-tagged peroxisomal proteins (N′ GFP peroxisomal library) lost their punctate localization during growth in glucose or in oleate (Fig. 3A). Although all known targets of Pex5 were affected by the absence of Pex5, and all Pex7 targets lost their peroxisomal localization upon loss of Pex7, we could not detect any obvious change in localization for any protein upon deletion of Ymr018w (data not shown).

Given that both Pex5 and Pex7 were fully functional in the Δymr018w strain, we assumed that their presence could compensate for any loss of Ymr018w activity. By similarity to Pex5 alone it was impossible to predict the function of Ymr018w given that it could either target PTS1 proteins similar to Pex5, or might work in a similar manner to Pex5 Long (Pex5L), the mammalian Pex7 co-receptor (Schliebs and Kunau, 2006), and hence might target PTS2 proteins. Therefore, we created strains in which PEX5 or PEX7 was deleted and mCherry–Ymr018w was overexpressed under a strong TEF2 promoter. After verifying that mCherry–Ymr018w was localized to peroxisomes (Fig. S4A), we examined the localization of all of the GFP-tagged peroxisomal proteins in control strains (no mutation, Δpex5, Δpex7, TEF2-mCherry-Ymr018w) as well as in strains in which Ymr018w was overexpressed in the background of the absence of Pex5 or Pex7 (Fig. 3A). Indeed, under these conditions the substrate range of Ymr018w was exposed.

We could detect that Mls1, a malate synthase that contains a PTS1 signal, completely lost its peroxisomal localization upon deletion of PEX5 but regained peroxisomal targeting upon overexpression of Ymr018w (Fig. 3B) suggesting that Mls1 can use both Pex5 and Ymr018w for its targeting into peroxisomes. Importantly, the effect of Ymr018w was dependent on its levels and independent of the presence of the mCherry tag as it targeted substrates even more robustly when expressed under the GPD1 promoter [which drives an approximately twofold to threefold stronger transcription than the TEF2pr (Janke et al., 2004)] and with no tag (Fig. 3B, “No tag”; Fig. S4B).

Mls2, the second PTS1-containing malate synthase, was targeted to peroxisomes only upon overexpression of Ymr018w (regardless of Pex5 presence). This suggests that Mls2, unlike Mls1, is completely dependent on Ymr018w for its targeting (Fig. 3C).

Another Ymr018w target is Gto1, a glutathione transferase. Unlike the Mls proteins, Gto1 was the only PTS1 protein that we still observed in puncta when PEX5 was deleted and its punctate localization increased upon Ymr018w overexpression (Fig. 3D,E). Interestingly, we did not detect an obvious effect on GFP–Gto1 in the Δymr018w strain (data not shown), most probably due to backup targeting by Pex5.

Importantly, even when Ymr018w was overexpressed under the GPD1 promoter it was not able to target additional PTS1 proteins other than the three described above, to peroxisomes (Fig. S4C). Intrigued by this specificity, we further studied the targeting specificity of Ymr018w and examined whether a GFP protein to which we added the last 10 amino acids of Mls1 would be targeted to peroxisomes by Ymr018w. Interestingly the C-terminus of Mls1 was sufficient to mediate targeting by Ymr018w (Fig. 3F).

Taken together, our results show that Ymr018w targets a subset of PTS1-containing matrix proteins, and that targeting specificity is defined by the context of the PTS1 sequence. In a parallel and independent approach, Ymr018w was found to have the same function in another study (Effelsberg et al., 2016), hence we jointly named Ymr018w as Pex9.

Our work demonstrates that two PTS1 targeting receptors exist in yeast – the well-characterized receptor Pex5, which targets all PTS1 proteins, and Pex9, which targets a specific subset of PTS1 proteins. Why would another targeting receptor be required for only a few proteins? Interestingly, the three proteins that we found to be substrates for Pex9 targeting, Mls1, Mls2 and Gto1, are highly regulated and their expression is induced dramatically in oleate (Barreto et al., 2006; Kunze et al., 2002, 2006). Correspondingly, we could see that just like GFP–Mls1, GFP–Pex9, when driven under its native promoter, was upregulated upon growth in oleate (Fig. 4A,B). We therefore suggest that Pex9 functions under specific conditions, such as oleate growth, to support and specify targeting of essential peroxisomal proteins under such conditions.

To date, two major pathways for targeting proteins to the matrix of peroxisomes have been characterized (Hasan et al., 2013) (Fig. 4C). One pathway is mediated by the Pex5 receptor that mainly targets PTS1-containing proteins. The second pathway is mediated by the Pex7 receptor that targets two PTS2-containing proteins and Pnc1 (Effelsberg et al., 2015; Kumar et al., 2016). Pex7 cannot work independently and hence it uses co-receptors (Purdue et al., 1998). Our results put down a new framework for thinking of peroxisomal targeting – a basal machinery (Pex5 and Pex7 with co-receptor Pex21) that functions under normal conditions, and an environmentally regulated arm (Pex7 with co-receptor Pex18 and Pex9) that gives priority to specific proteins required under these conditions or might be able to work under more adverse conditions.

The requirement for a condition-specific targeting machinery might not be unique to yeast. It has been previously shown that Pex5 has a Pex5-like paralog in humans. Interestingly, this Pex5-like protein is specifically expressed in the brain, which is one of the organs most dependent on peroxisome function (Amery et al., 2001). We found that indeed, most vertebrates contain both a Pex5 protein and a Pex5-like protein (Fig. 4D) implying a duplication of the PEX5 gene early in the vertebrates lineage. The existence of paralogs for Pex5 might suggest that two PTS1-targeting receptors, one which targets all PTS1 proteins, and another, that is expressed under specific conditions or in specific cells and targets preferential proteins, is not restricted to yeast. This highlights the beauty and complexity of targeting proteins to peroxisomes – a fascinating and important organelle.

All strains in this study are based on the BY4741 laboratory strain (Brachmann et al., 1998). A complete list of strains can be found in Table S5. The libraries used were the yeast C′ GFP library (Huh et al., 2003) and the SWAT N′ GFP library (Yofe et al., 2016), which is a collection of ∼1800 strains tagged with GFP at their N-terminus and expressed under the generic constitutive promoter SpNOP1pr. The pFA6a plasmid that was originally used for C-terminal GFP tagging (Huh et al., 2003) was modified to contain the last 10 amino acids of Mls1 at the end of the GFP sequence. Using this plasmid, GFP–PTS1(Mls1) was genomically integrated downstream to MDH3 promoter (Fig. 3F). The strains expressing GFP–Pex9 and GFP–Mls1 under their native promoter (Fig. 4A,B) were picked from the seamless N′ GFP library (Yofe et al., 2016).

To create collections of haploid strains containing GFP-tagged proteins with additional genomic modification such as a peroxisomal marker (Pex3–mCherry) or different deletions (Δpex5, Δpex7, Δpex5Δpex7 and Δymr018w) and overexpression (TEF2-mCherry-Ymr018w), different query strains were constructed on the basis of an SGA compatible strain (for further information see Table S5). Using the SGA method (Cohen and Schuldiner, 2011; Tong and Boone, 2006) the Pex3-mCherry query strain was crossed with the SWAT-GFP library and the other query strains were crossed into a collection of strains from the SWAT-GFP library containing ∼90 strains including known and potential peroxisomal proteins and controls. To perform the SGA in high-density format we used a RoToR bench top 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 Geneticin (200 µg/ml) and/or Neurseothricin (200 µg/ml). 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 Geneticin (200 µg/ml) and/or Neurseothricin (200 µg/ml), alongside the toxic amino acid derivatives Canavanine and Thialysine (Sigma-Aldrich) to select against remaining diploids, and lacking Histidine to select for spores with an A mating type.

The collections were visualized using an automated microscopy setup as described previously (Breker et al., 2013). In short: cells were transferred from agar plates into 384-well polystyrene plates for growth in liquid media using the RoToR arrayer robot. Liquid cultures were grown in LiCONiC incubator, overnight at 30°C in SD-URA medium (for the SWAT-GFP collections) or SD-HIS medium (for the C′-GFP library). A JANUS liquid handler (PerkinElmer) 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) or S-Oleic (6.7 g/l yeast nitrogen base, 0.2% oleic acid and 0.1% Tween-40) supplemented with complete amino acids. Plates were incubated at 30°C for 4 h in SD medium or for 20 h in S-Oleic. 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 screens in oleate) to remove non-adherent cells and to obtain a cell monolayer. The plates were then transferred to the ScanR automated inverted fluorescent microscope system (Olympus) using a robotic swap arm (Hamilton). 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-ER charge-coupled device camera (Hamamatsu). Images were acquired in two channels: GFP (excitation filter 490/20 nm, emission filter 535/50 nm) and mCherry (excitation filter 572/35 nm, emission filter 632/60 nm). All images were taken at a single focal plane.

The localization of the GFP-tagged proteins in oleate was obtained manually and was compared to the published localization databases of cells grown in SD medium that contains glucose. The localizations of the N′-GFP library were compared to the SWAT-GFP localization database (Yofe et al., 2016). The localizations of the C′-GFP library were first compared to the LoQAtE database (Breker et al., 2014). Whenever the localization in oleate was different from the localization observed in LoQAtE, we also compared the localization to the YeastGFP database, which is based on the same library (Huh et al., 2003). Only cases in which the localization of a protein in oleate was different to the localization observed in both of the two databases were further analyzed.

In cases where an obvious GFP signal was viewed but it was impossible to define a specific localization, the localization was described as ‘ambiguous’ and it was not included in the analysis. Cases with technical problems such as contaminations, unfocused images and not enough cells were not included in the analysis as well. Since mitochondria were altered in oleate, we did not consider changes of mitochondrial proteins from mitochondria in glucose to punctate in oleate as different localization. Bud and bud neck localizations were defined as ‘same’ localization.

The proteins that exhibited a different localization in oleate compared to glucose were subdivided into group according to the criteria described in Table S1. The analysis was done using Matlab software version R2015a (http://www.mathworks.com/products/matlab/). All changes in localizations are summarized in Table S2 (N′ GFP library) and Table S3 (C′ GFP library).

In order to get quantitative information about the size of lipid droplets and the shape of mitochondria under growth in glucose and oleate (Fig. S1A–E), an in-house script was written in ImageJ (http://imagej.net/Fiji/Downloads), which includes a rolling ball algorithm for subtracting the background and Otsu's threshold clustering algorithm for adjusting the threshold. After a mask was created, parameters such as area and circularity (the ratio between two perpendicular axes) were measured on the original images (the script is available on request from the corresponding author). The statistical analysis and the relevant plots were generated using R software (https://cran.r-project.org/bin/windows/base/old/3.2.2/). Overall, nine lipid droplets proteins and seven mitochondrial proteins were analyzed including thousands of cells per analysis.

Manual microscopy imaging was performed in strains expressing the following: Mtc4–GFP plus Erg6–mCherry (Fig. 2D), GFP–Mls1 (Fig. 3B), GFP–Mls2 (Fig. 3C), GFP–Gto1 (Fig. 3D), GFP–Pxp1 (Fig. S4C), GFP–Pex9 and GFP–Mls1 under the regulation of their native promoter (Fig. 4A,B). Yeast strains were grown as described above for the high-throughput microscopy with changes in the selection required for each strain. 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; ×60 oil objective; Excitation wavelength of 488 nm for GFP). Images were taken by a connected PCO-Edge sCMOS camera controlled by VisView software. All images were taken at a single focal plane.

The phylogenetic analysis was conducted in MEGA7 software (Kumaret al., 2015; http://www.megasoftware.net/), with the maximum likelihood method (Jones et al., 1992). The tree topology was tested using 500 bootstrap iterations, which confirmed the separation of the two vertebrate branches (bootstrap value 99, Fig. 4D). The scale represents the number of substitutions per site. Positions containing gaps were eliminated. The analysis involved 14 amino acid sequences: PEX5 [human (Ensembl ID ENST00000266563), mouse (Ensembl ID ENSMUSG00000005069), lizard (Ensembl ID ENSACAG00000007658), chicken (Ensembl ID ENSGALG00000014683), frog (Ensembl ID ENSXETG00000023910), fish (Ensembl ID ENSTRUG00000009964)], PEX5L (PEX5-like) [human (Ensembl ID ENST00000465751), mouse (Ensembl ID ENSMUSG00000027674), lizard (Ensembl ID ENSACAG00000004210), chicken (Ensembl ID ENSGALG00000008929), frog (Ensembl ID ENSXETG00000032445)], yeast PEX5 (YDR244W) and yeast PEX9 (YMR018W). Human UDP-N-acetylglucosamine, the closest protein to human PEX5 was used as an outgroup. The sequences were aligned using Clustalx software (http://www.clustal.org/download/current/).

We wish to thank Yoav Peleg for constructing the GFP–PTS1(Mls1) plasmid. We wish to thank Ralf Erdmann and Wolfgang Schliebs for their collegiality.