|Home | About | Journals | Submit | Contact Us | Français|
Giardia lamblia, which is an important parasitic cause of diarrhea, uses activated forms of glucose to make glycogen and activated forms of mannose to make glycophosphosphoinositol anchors. A necessary step for glucose activation is isomerization of glucose-6-phosphate to glucose-1-phosphate by a phosphoglucomutase (PGM). Similarly, a phosphomannomutase (PMM) converts mannose-6-phosphate to mannose-1-phosphate. While whole genome sequences of Giardia predict two PGM candidates, no PMM candidate is present. The hypothesis tested here is that at least one of the two Giardia PGM candidates has both PGM and PMM activity, as has been described for bacterial PGM orthologs. Nondenaturing gels showed that Giardia has two proteins with PGM activity, one of which also has PMM activity. Phylogenetic analyses showed that one of the two Giardia PGM candidates (Gl-PGM1) shares recent common ancestry with other eukaryotic PGMs, while the other Giardia PGM candidate (Gl-PGM2) is deeply divergent. Both Gl-PGM1 and Gl-PGM2 rescue a Saccharomyces cerevisiae pgm1Δ/pgm2Δ double deletion strain, while only Gl-PGM2 rescues a temperature-sensitive PMM mutant of S. cerevisiae (sec53-ts). Recombinant Gl-PGM1 has PGM activity only, whereas Gl-PGM2 has both PGM and PMM activities. We conclude that Gl-PGM1 behaves as a conventional eukaryotic PGM, while Gl-PGM2 is a novel eukaryotic PGM that also has PMM activity.
Giardia lamblia (also known as G. duodenalis and G. interstinalis) is a deeply divergent, binucleate, flagellated protist, which causes two million cases of diarrhea per year in the United States (Steiner et al. 1997; Adam 2001; Simpson et al. 2006). Whole genome sequences of two human isolates of Giardia called WB and GS suggest that they separated from shared common ancestor tens of millions of years ago (Morrison et al. 2007; Franzén et al. 2009). Each Giardia genome is small (~11.7 Mb) and encodes fewer proteins (~4500) than the vast majority of eukaryotes. Giardia has a large number of genes, which have been received from bacteria by lateral gene transfer (LGT) (Nixon et al. 2002; Andersson et al. 2003; Morrison et al. 2007). Giardia genes obtained by LGT encode numerous glycolytic enzymes, most fermentation enzymes, as well as nitroreductases that reduce and activate metronidazole (Pal et al. 2009).
Compared to other eukaryotes, Giardia has many fewer enzymes involved in protein glycosylation:
In this report, we are interested in the Giardia phosphoglucomutase (PGM) that makes the activated form of glucose (Glc-1-P) for glycogen synthesis and the phosphomannomutase (PMM) that makes activated mannose (Man-1-P) for GPI anchor synthesis (Figure 1). While candidate Giardia enzymes have been identified for other steps in the synthesis of glycogen and GPI anchors, whole genome sequences of the organism predict two candidate PGM enzymes: Gl-PGM1 (GiardiaDB 50803_17254) and Gl-PGM2 (GiardiaDB 50803_11448) but no candidates for PMM (Henze et al. 2001; Morrison et al. 2007). This result is surprising because PMM orthologs (defined by common ancestry) of higher eukaryotes may have both PMM and PGM activities (Pirard et al. 1999; Cromphout et al. 2006), but the eukaryotic PGM orthologs have not been shown to have PMM activity (Table I) (Quick et al. 1974). In contrast, all bacteria lack a PMM ortholog, while some bacteria have PGM orthologs that have both PGM and PMM activities (Regni et al. 2004, 2006). The reaction of the bacterial PGM/PMM involves a phosphoryl transfer from a phosphoserine on the enzyme to the bound substrate to form a bisphosphorylated intermediate, which is then reoriented 180° for a second phosphoryl transfer from the intermediate back to the enzyme.
The major hypothesis tested here is that at least one of the Giardia PGM orthologs has both PGM and PMM activities, as has been described for bacterial PGM orthologs. The minor hypothesis tested was that Giardia obtained this bifunctional enzyme by LGT from bacteria. To test this idea, we used nondenaturing gels to separate PGM activities of Giardia and to determine whether any of the Giardia PGMs also have PMM activity. We used phylogenetic methods to determine whether the Giardia PGMs might have derived from bacteria by LGT. In addition, we tested the ability of the Giardia PGMs from the WB isolate to complement a Saccharomyces cerevisiae pgm1Δ/pgm2Δ double deletion strain and a Saccharomyces sec53-ts mutant that lacks PMM activity at the nonpermissive temperature and measured the kinetics of PGM and PMM activities of recombinant Giardia PGMs.
To determine whether Giardia has one or more enzymes with PGM and/or PMM activities, trophozoites of WB were lysed by sonication, and soluble proteins were separated on a nondenaturing polyacrylamide gel. This gel was overlayed with agar containing colorometric reagents to detect PGM and PMM activities. We found that Giardia has two enzymes with PGM activity (Figure 2). The faster migrating Giardia enzyme has much greater PGM activity than the slower migrating enzyme, consistent with either a more active enzyme or a more abundant protein, or both. However, the slower migrating Giardia enzyme also has PMM activity, albeit weaker than the PGM activity. As controls for the PGM assays, wild-type Saccharomyces shows a single wide band of PGM activity, which is absent in a Saccharomyces pgm1Δ/pgm2Δ double deletion strain.
In addition to the activity data shown in Figure 2, we used multiple dimensional protein identification technology (MudPIT) to identify proteins from lysed Giardia. These unpublished data have been deposited at GiardiaDB (Aurrecoechea et al. 2008). MudPIT analyses suggested that one of the candidate Giardia PGMs (Gl-PGM1) with 24 peptides identified is more abundant than the other (Gl-PGM2) with just five peptides identified.
We conclude that Giardia has two enzymes with PGM activity, and one of these PGMs that may be less abundant also has PMM activity. In the following sections, we use bioinformatics methods to compare candidate Giardia PGMs to those of other eukaryotes, and we test the ability of the Giardia PGMs to complement a S. cerevisiae pgm1Δ/pgm2Δ double deletion mutant and a Saccharomyces sec53-ts mutant that lacks PMM activity at the nonpermissive temperature.
The predicted proteins of the WB isolate of Giardia include two PGM orthologs, which we labeled as Gl-PGM1 (GiardiaDB 50803_17254) and Gl-PGM2 (GiardiaDB 50803_11448) (Morrison et al. 2007). Gl-PGM1 is 655-amino acids long and contains throughout its length a PGM domain (Marchler-Bauer et al. 2005). In contrast, the second predicted PGM of Giardia (Gl-PGM2) is shorter than other PGMs (527-amino acids long) and contains only the N-terminal half of the PGM domain. Gl-PGM1 of WB strain and Gl-PGM2 of GS strain have a 37% identity and 54% similarity over a 494 amino acid alignment with 4% gaps (Supplemental Figure 1). Because Gl-PGM2 is much more similar to Gl-PGM1 than to any other PGM (data not shown), it is likely that Gl-PGM2 resulted from gene duplication in Giardia rather than from gene duplication in a common ancestor to Giardia and other eukaryotes.
Gl-PGM1 shows a 31% identity and a 44% similarity over a short (132 amino acid) overlap with the Pseudomonas PGM/PMM that has been extensively characterized (Regni et al. 2004, 2006). While the phosphoryl transfer site including the Ser that is phosphorylated is conserved in Gl-PGM1, phosphate contacts and sugar ring contacts are not conserved (Supplemental Figure 1). Gl-PGM2 cannot be directly aligned with Pseudomonas PGM/PMM or with any of the other PGMs that have been crystallized. However, using Gl-PGM1 as a guide for the alignment, it is clear that the phosphoryl transfer site, including the Ser that is phosphorylated, is not well conserved in Gl-PGM2 (Supplemental Figure 1). Crystallization of the Gl-PGM1 and Gl-PGM2 with various substrates, which is necessary for understanding structure–function relations of the enzymes, is beyond the scope of the present study.
As Gl-PGM2 is so unlike other PGMs and so was unalignable over its C-terminal half with the others, this predicted Giardia PGM was not included in a phylogenetic analysis of PGMs from representative eukaryotes and bacteria (Figure 3). The phylogenetic analyses showed two things. First, all eukaryotes examined (with the exception of Trypanosoma brucei) have at least one PGM, consistent with the importance of this enzyme in the synthesis of activated Glc used to make glycogen, starch and other products (Quick et al. 1974). Second, there is a clear distinction between eukaryotic PGMs, which includes Gl-PGM1, and bacterial PGMs. This result shows that the Giardia PGM was not derived from bacteria by LGT, as we hypothesized (Nixon et al. 2002; Andersson et al. 2003; Morrison et al. 2007).
Giardia is the only eukaryote examined that is missing a candidate PMM (see Supplemental Figure 2 for the PMM tree). As whole genome sequences were examined from two different Giardia isolates (WB and GS), it is highly unlikely that the absence of a Giardia PMM ortholog is due to an artifact in library construction or sequencing (Morrison et al. 2007; Franzén et al. 2009). As PMM orthologs are present in other deeply divergent eukaryotes (e.g. Trichomonas and Trypanosoma), it is likely that secondary loss rather than primary absence best explains why Giardia is missing a PMM gene (Simpson et al. 2006). The absences of Alg enzymes that synthesize N-glycan precursors and of N-glycan-dependent QC proteins in the endoplasmic reticulum have also been best explained by secondary loss of the genes that encode these proteins (Samuelson et al. 2005; Banerjee et al. 2007).
To determine whether the Gl-PGM1 and Gl-PGM2 genes encode functional PGMs, we tested their ability to complement the phenotype of a S. cerevisiae pgm1Δ/pgm2Δ double deletion strain. The complete coding sequences of Gl-PGM1 and Gl-PGM2 were expressed in the Saccharomyces pgm1Δ/pgm2Δ double deletion strain using the pVV214 vector, which contains the URA3 gene that encodes orotidine 5-phosphate decarboxylase. The double deletion strain is incapable of metabolizing galactose (Gal) and presents a lethal phenotype when Gal is used as the sole carbon source (Daran et al. 1997). The deletion strain bearing either Gl-PGM1 or Gl-PGM2 grows normally on Gal (Figure 4), indicating that both Giardia PGM genes encode proteins with PGM activity (Table I).
As a control, we took advantage of URA3 marker in the transformants and grew them on plates with 1% (w/v) 5-fluoroorotic acid (FOA). FOA is converted to a toxic compound 5-fluorouracil by URA3, so the presence of URA3 on a plasmid is lethal in the presence of FOA. In this way, URA3 plasmids were removed from the transformants. As expected, the FOA-treated transformants, which no longer contain the Giardia PGM genes, do not grow on Gal plates (Figure 4).
To determine whether the Gl-PGM1 and Gl-PGM2 genes encode functional PMMs, we tested their ability to complement a S. cerevisiae temperature-sensitive PMM mutant (HMSF33 sec53-6 strain). In addition, as a positive control, we tested the complementation of a functional Saccharomyces sec53 gene. While the Saccharomyces sec53-ts mutant grows normally at 25°C (the permissive temperature) (Figure 5A), the mutant transformed with an empty vector (negative control) does not grow at 37°C (the restrictive or nonpermissive temperature) because the PMM is not functional (Figure 5B). Similarly, the Saccharomyces sec53-ts mutant transformed with Gl-PGM1 does not grow at 37°C. In contrast, the Saccharomyces sec53-ts mutant transformed with either the Gl-PGM2 gene or a functional Saccharomyces sec53 gene (positive control) grows at 37°C. These results, which indicate that Gl-PGM2 has PMM activity while Gl-PGM1 does not (Table I), are supported by kinetic data for the recombinant Gl-PGM1 and Gl-PGM2 that are presented in the next section.
Recombinant Gl-PGM1 that has an N-terminal polyHis tag was expressed in E. coli and purified using Ni-sepharose affinity column chromatography. Recombinant Gl-PGM2 that has a C-terminal polyHis tag was expressed in wild-type Saccharomyces and purified by Ni-sepharose affinity column chromatography. Gl-PGM1 has only PGM activity (Km 8.2 ± 2.0 for Glc-1-P and specific activity is 87 ± 7 μmol/min/mg) (Figure 6A). Gl-PGM2 has both PGM and PMM activities (Km 3.0 ± 0.6 μM, specific activity 76 ± 4 μmol/min/mg for Glc-1-P and Km 5.6 ± 1.7 μM, specific activity 25 ± 2 μmol/min/mg for Man-1-P) (Figure 6B). These results are in agreement with the data from total lysates of Giardia separated on nondenaturing gels (Figure 2), which showed that one Giardia enzyme has PGM activity and no PMM activity, while the other Giardia enzyme has both PGM and PMM activities. These results are also in agreement with the Saccharomyces complementation assays that showed Gl-PGM1 has PGM but not PMM activity, while Gl-PGM2 has both PGM and PMM activities.
The mystery solved here is how Giardia, which lacks a eukaryotic PMM ortholog, is able to make activated mannose that is used to make GPI anchors. As predicted, one of the two Giardia PGM orthologs (Gl-PGM2), which is relatively less abundant in MudPIT analyses, has both PGM and PMM activities (Table I). While this dual PGM/PMM activity is similar to that of bacterial PGM orthologs, it does not appear that the Giardia enzymes were obtained by LGT (Nixon et al. 2002; Andersson et al. 2003; Regni et al. 2006; Morrison et al. 2007). Therefore, the presence of bifunctional PGM/PMM enzymes in Giardia and bacteria likely represents an example of convergent evolution. PMM orthologs of eukaryotes, which are not phylogenetically related to PGMs, have both PGM and PMM activities (Pirard et al. 1999; Cromphout et al. 2006). Broadening of the hexose specificity of these phosphohexosemutases has happened, therefore, on at least three occasions (PGMs acquiring PMM activity in Giardia and in bacteria and a eukaryotic PMM acquiring PGM activity) (Table I). As noted above, a structural understanding of the PGM/PMM activity of Gl-PGM2 will only be possible when the enzyme is crystallized with its various substrates and intermediates (Regni et al. 2004, 2006).
The majority of the “substituted” enzymes in the glycolytic and fermentation pathways of Giardia appear to have derived by LGT (Nixon et al. 2002; Andersson et al. 2003; Regni et al. 2006; Morrison et al. 2007). In addition, the Giardia cis-prenyltransferase that makes dehydrodolichol pyrophosphate was obtained from bacteria by LGT (Grabińska et al. 2010). Here, we infer a plausible historical sequence in which a single Giardia PGM gene was duplicated, one of the PGM paralogs developed PMM activity, and the ancestral eukaryotic PMM ortholog was lost. Why this occurred in Giardia is interesting but likely unanswerable. As T. brucei is missing a PGM ortholog, we expect that its predicted PMM also has PGM activity. Changing substrate specificity is not infrequent in enzymes of protists. For example, Trichomonas has a dolichol-phosphate-glucose synthase ortholog that has dolichol-phosphate-mannose synthase activity, as well as a malate dehydrogenase ortholog that has lactate dehydrogenase activity (Wu et al. 1999; Grabińska et al. 2008).
Trophozoites of the WB isolate of G. lamblia were grown axenically in TYI-S media supplemented with 10% serum and 1 mg/mL bile. S. cerevisiae strains used in this study were as follows: BY4741 (MATα; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0), Δpep4 (as BY4741), pep4::kanMX4, Δprb1 (as BY4741, prb1::kanMX4), PGM1Δ (MATa/MATα; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; met15Δ0/met15Δ0; ura3Δ0/ura3Δ0), PGM2Δ (MATa; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0) and temperature-sensitive PMM mutant (HMSF33 sec53-6) (ATCC). Strains were grown in a synthetic minimal (SD) medium containing 0.67% yeast nitrogen base without amino acids (Difco, Detroit, MI) and 0.5% d-glucose (WAKO, Osaka, Japan), supplemented with an amino acids mixture.
Giardia trophozoites (~106/mL × 50 mL) were chilled on ice for 30 min, concentrated at 2000 rpm for 10 min, lysed by sonication in 10 mM Hepes, pH 7, 10 mM MgCl2 and 25 mM NaCl, supplemented with protease inhibitor cocktail (Sigma, St. Louis, MI) and spun at 13,000 rpm for 30 min to remove insoluble proteins. Saccharomyces (250 mL) were grown in yeast extract peptone dextrose media to O.D. 3 and concentrated by centrifugation at 2000 × g. Yeasts were resuspended in lysis buffer (10 mM Hepes, pH 7, 10 mM MgCl2 and 0.5 mM phenylmethylsulfonyl fluoride) and lysed by sonication in acid-washed glass beads (Sigma). Giardia and yeast cell protein extracts were subjected to electrophoresis toward the anode for about 5 h at 25 mA at 4°C on a 4–12% nondenaturing, discontinuous polyacrylamide gel. The stacking gel buffer consisted of 125 mM Tris (pH 6.8), the resolving gel buffer 375 mM Tris (pH 8.8) and the running buffer Tris–glycine (pH 8.3).
After electrophoresis, the gel was rinsed with deionized water and placed on top of three sheets of Whatman 3MM filter paper saturated with buffer (80 mM imidazole, 5 mM MgSO4) in a Petri dish. An agar gel used as an overlay was prepared, as described (Bevan and Douglas 1969). We melted 0.4 g of Difco Noble agar in 30 mL of 1 mM ethylenediaminetetraacetic acid solution, to which the following ingredients were added and mixed in order: 10 mL of premix (2 mL of 0.8 M imidazole hydrochloride [pH 7.4], 2 mL of 0.1 M MgSO4, 2 mL of 40 mM Glc-1-P, 2 mL of 140 μM Glc-1,6-P, 2 mL of 10 mM NADP+-), 0.4 mL of Glc-6-P dehydrogenase (200 U/mL), 0.4 mL of 60 mM nitroblue tetrazolium and 0.1 mL of 400 mM phenazine methosulfate. This mixture was immediately dispensed evenly over the polyacrylamide gel.
After the overlay agar solidified, the dish was covered with aluminum foil and kept in the dark at 30°C for 2 h to allow development of the purple color indicative of PGM activity. Color development did not occur if Glc-1-P was omitted from the reaction mixture. The stained gel was stored in 7.5% acetic acid until dried or directly photographed with a digital camera. For PMM activity, Glc-1-P was replaced by Man-1-P in the agar premix, and phosphoglucose isomerase (PGI) and phosphomannose isomerase (PMI) were added to the mixture. Again a purple color development indicated PMM activity.
The predicted proteins of G. lamblia of the WB strain (first genome project) and GS strain (second genome project), which have been deposited in the NR data at GenBank or GiardiaDB, were searched with Psi-Blast using the PGM and PMM protein sequences from S. cerevisiae (Altschul et al. 1997; Gao et al. 2001; Morrison et al. 2007; Aurrecoechea et al. 2008; Franzén et al. 2009). Similar methods were used to search the predicted proteins of representative protists (Dictyostelium discoideum, Entamoeba histolytica, Leishmania major, Plasmodium falciparum, Toxoplasma gondii, Trichomonas, T. brucei and Trypanosoma cruzi) and metazoans (e.g. Homo sapiens, Xenopus laevis and Drosophila melanogaster) in the NR database at the NCBI or at specific databases (e.g. PlasmoDB or GeneDB). Predicted PGMs and PMMs were examined for conserved domains using the CD search at the NCBI (Marchler-Bauer et al. 2005). The mutases were aligned using multiple sequence comparison by log-expectation (Edgar 2004). The alignment was manually refined, and gaps were removed using BioEdit. The finished alignment was used to construct the phylogenetic tree using TREE-PUZZLE, a program to reconstruct phylogenetic trees from molecular sequence data by the maximum likelihood method (Schmidt et al. 2002).
PGM1Δ (MATa/MATα; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; met15Δ0/met15Δ0; ura3Δ0/ura3Δ0) diploid strain was sporulated and selected for mating type Matα. The kanamycin cassette in PGM2 was replaced by His5-rt6 cassette. Diploid strain obtained by crossing PGM1Δ (PGM1::kanMX4) with PGM2Δ (PGM2::His-rt6) was sporulated. Screening for appropriate markers and inability to grow on Gal identified spores bearing pgm1 and pgm2 alleles.
WB strain Giardia were chilled on ice for 30 min and concentrated at 2000 rpm for 10 min, and genomic DNA was isolated using Promega kit. Candidate PGM orthologs were amplified from the genomic DNA by polymerase chain reaction using the following primers. For Gl-PGM1 (GiardiDB 50803_17254), the forward primer was ATGGAGGAGAGAGCAAGAGAT, and the reverse primer (that includes a stop codon) was TTAAGGCTCCTTGTCATTGAC. For Gl-PGM2 (GiardiDB 50803_11448), the forward primer was ATGATCAATTTAAAGGACAAAGCAG, and the reverse primer (that includes a stop codon) was CTAAACGCTTTCAACTTGGC. GatewayTM recombinational technology was used to clone the open reading frames into pVV214 vector for expression of Giardia PGMs in Saccharomyces mutants. First, the amplified genes were cloned into entry pCR8/GW/TOPO TA (Invitrogen, Carlsbad, CA) vector. The entry vector bearing Gl-PGM1 or Gl-PGM2 was then recombined with the destination vector pVV214. These plasmids were transformed into S. cerevisiae pgm1Δ/pgm2Δ double deletion strain (described above) or into the S. cerevisiae temperature-sensitive PMM mutant (HMSF33 sec53-6).
Saccharomyces pgm1Δ/pgm2Δ double deletion mutant transformed with vectors containing Gl-PGM1 or Gl-PGM2 were selected in synthetic medium-URA containing 2% glucose and then plated onto synthetic medium-URA supplemented with 2% (w/v) Gal as the only carbon source. In the absence of a functional PGM, the Saccharomyces pgm1Δ/pgm2Δ double deletion mutant does not grow on Gal plates (data not shown). As a control, we took advantage of URA3 marker in the Saccharomyces pgm1Δ/pgm2Δ double deletion mutant transformants and grew them on plates with 1% (w/v) 5-FOA. FOA is converted to a toxic compound 5-fluorouracil by URA3, so the presence of URA3 on a plasmid is lethal in the presence of FOA. In this way, URA3 plasmids were removed from the transformants.
Alternatively, vectors containing Gl-PGM1 or Gl-PGM2 were transformed into the S. cerevisiae temperature-sensitive PMM mutant (HMSF33 sec53-6) in synthetic medium-URA containing 2% glucose at the permissive temperature (25°C). These mutants do not grow at 37°C unless they are transformed with an exogenous PMM-encoding gene. A positive control was a wild-type Saccharomyces sec53 (PMM) gene.
pDESTM17 Gateway vector (Invitrogen) is an adaptive destination vector used to clone and express recombinant protein with N-terminal polyHis tag in E. coli. The entry vector bearing Gl-PGM1 was recombined with pDESTM17 vector and transformed into the BL21-AI strain of E. coli. Bacteria were induced with 0.2% arabinose for 4 h at 37°C according to the protocol provided by Invitrogen. The recombinant Gl-PGM1 was purified using Ni-sepharose resin (Novagen, Madison, WI). Expression and purification of Gl-PGM1 were confirmed by sodium dodecyl sulphate–polyacrylamide gel electrophoresis.
pYES2-Dest52 Gateway vector (Invitrogen) makes recombinant proteins with a C-terminal V5 epitope tag followed by polyHis tag under the Saccharomyces Gal1 promoter. A modified entry vector bearing Gl-PGM2 without the C-terminal stop codon was recombined with pYES2-Dest52 and transformed into S. cerevisiae strain BY4741 (MATα his3Δ1 leu2Δ0 met15Δ0 ura3Δ0). Transformed yeasts were induced with 2% Gal for 16 h at 30°C and then yeasts were lysed with glass beads. Recombinant Gl-PGM2 was purified on the Ni-sepharose resin and confirmed by western blot of purified protein using a horse-radish peroxidase-labeled antibody to the V5 epitope tag.
All enzymatic activities were assayed spectrophotometrically at 340 nm by the reduction of NADP+- to NADPH in a reaction mixture incubated at 30°C and containing, unless otherwise stated, 50 mM Hepes, pH 7.1, 5 mM MgCl2, 0.25 mM NADP+- and 10 μg/mL yeast glucose 6-phosphate dehydrogenase. PGM activity was measured in the presence of 500 μM Glc-1-P and 10 μM Glc-1,6-P, and PMM activity was measured in the presence of 10 μM Glc-1,6-P, 100 μM Man-1-P, 10 μg/mL PGI and 3.5 μg/mL PMI.
Supplementary data for this article is available online at http://glycob.oxfordjournals.org/.
This work was supported by National Institutes of Health grants. We thank Dr. Kariona Grabińska for help with yeast manipulations. National Institutes of Health grants AI44070 (to J.S.) and GM31318 (to P.W.R.).