In Vivo Analysis of the Major Exocytosis-sensitive Phosphoprotein in Tetrahymena

Restriction enzymes were from New England Biolabs (Beverly, MA). All other reagents were from Sigma Chem. Co. (St. Louis, MO), unless otherwise noted.

Cells were grown at 30°C with agitation in 1% proteose peptone, 0.1% yeast extract (both from Difco, Detroit, MI), 2% glucose, and 0.003% ferric EDTA. Tetrahymena thermophila strains derived from inbred strain B were used in this work, and are listed in Table I. Strain B2086 is considered to be wild-type, and all initial strains are wild-type with respect to regulated exocytosis. Strains CU428.1 and B2086 were obtained from Peter Bruns (Cornell University, Ithaca, NY).

Based on known codon usage in Tetrahymena thermophila (Martindale, 1989), degenerate primers (Integrated DNA Technologies, Coralville, IA) were designed to amplify parafusin homologues via the polymerase chain reaction. Primers were designed to correspond to regions conserved between GAL5 (Saccharomyces cerevisiae), PGM1 (Homo sapiens), and parafusin (P. tetraurelia). A total of four forward and four reverse primers were designed and used in all 16 possible combinations. The primers correspond to residues 126-133, 142-149, 321-327, 339-344, 388-384, 369-362, 410-403, and 427-421 of the deduced Pgm1p protein sequence (see Fig. 4). PCR was performed on a Tetrahymena thermophila λgt10 cDNA library generously provided by Tohru Takemasa (University of Tsukuba, Japan), or on Tetrahymena thermophila strain CU428.1 genomic DNA. The 50-μl reactions contained 10⁸ plaque-forming units or 10 ng genomic DNA, 2.5 U Taq polymerase (Promega Corp., Madison, WI), 10 mM Tris-HCl pH 8.3, 50 mM KCl, 2.5 mM MgCl₂, 250 μM each dNTP, and 1 μM each primer. 35 cycles of 92°C, 0.5 min; 40, 45, 50, or 55°C, 1 min; 72°C, 0.5 min, were performed. A total of 12 bands of the expected sizes were cloned into pCRII (Invitrogen, San Diego, CA) and sequenced; 5 products were unrelated to PGM, the other 6 (3 amplified from genomic DNA, 3 from cDNA, using 6 of the 8 primers in all) were related to various PGMs. These were determined to be identical in overlapping regions and derived from a single gene.

The largest PCR product, 1,265 bp (corresponding to bases 559-1824 of Fig. 2) was used to screen a cDNA phage library using the Genius system (Boehringer Mannheim, Indianapolis, IN), which resulted in the isolation of a single clone containing a 1.7-kB insert. After subcloning into pBluescript SK (+) (Stratagene, La Jolla, CA), sequencing revealed that the 5′ end was truncated. On the assumption that some full-length clones were present in the library, we conducted PCR on the cDNA library using a λgt10 primer and a primer complementary to the portion of the gene we had already sequenced. Cloning and sequencing of the resultant 900-bp product revealed that it contained a putative start codon.

Genomic sequences 5′ of the cDNA were obtained by inverse PCR (Ochman et al., 1988) on genomic DNA digested with EcoRI. The forward primer corresponded to bases 306-275, and the reverse primer to bases 1250-1273 of Fig. 2. PCR was performed on 200 ng of circularized genomic DNA using the same reagents and cycling times as above, but with a 1.5-min extension step. These reactions produced a single product which was the expected size (3 kB) based on Southern blotting. This was cloned into pCRII and partially sequenced. The PGM1 sequence reported represents the consensus obtained from at least three independent templates, each sequenced on both strands.

Probe templates were prepared by PCR amplification of cloned sequences followed by purification on QIAprep columns (Qiagen, Chatsworth, CA). Probes labeled with either digoxigenin (Boehringer Mannheim; for Southern analysis) or ³²P (DuPont NEN, Wilmington, DE; for Northern analysis) were then prepared by randomly primed DNA synthesis.

Genomic DNA was prepared as described (Gaertig et al., 1993) except that after isopropanol precipitation, DNA was further purified by dissolving in 50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 100 μg/ml RNaseA for 10 min at 65°C, and then separated from contaminants on a Q-100 column (Qiagen, Chatsworth, CA) according to the manufacturer's directions. DNA was quantified by spectroscopic analysis (A₂₆₀/A₂₈₀ ratios were typically 1.75–1.85). Southern blotting was according to manufacturer's instructions (Genius, Boehringer Mannheim) loading 20 μg of DNA per lane. For low-stringency blots, hybridization was conducted in 50% formamide, 5× SSC, 2.0% DIG blocking reagent (Boehringer Mannheim), 0.1% N-lauroylsarcosine, 0.2% sodium dodecyl sulfate, at 20°C for 16 h. This was followed by four 15-min washes in 2× SSC containing 0.1% SDS, also at 20°C. RNA purification and Northern blotting were performed as previously described (Chilcoat et al., 1996).

PGM1 was disrupted by replacement of the genomic sequences −76 to +2182 relative to the start codon (resulting in removal of the first 486 residues) with a construct containing the neo2 resistance cassette (Chilcoat et al., 1996; Gaertig et al., 1994), which confers paromomycin resistance in Tetrahymena. Electroporation (ECM 600; BTX Corporation, San Diego, CA) was used to transform mating pairs of B2086 and CU428.1, as previously described (Gaertig et al., 1995). Initial drug selection was with 120 μg/ml paromomycin, added 6 h after transformation. 3–6 transformants were obtained per transformation (n = 3). Because only some of the macronuclear PGM1 copies were replaced initially, cells were grown in increasing concentrations of paromomycin for stringent selection of the pgm1-1:: neo2 allele. Several transformants were cloned and grown in paromomycin (up to 4 mg/ml) for 400 generations.

To measure PGM activity, 500-ml cultures were grown to 1–2 × 10⁵ cells/ ml, then rapidly chilled and pelleted at 200 g for 5 min in conical bottles. All subsequent steps were performed cold. Cells were washed once in 10 mM Hepes-KOH, pH 6.9, and the pellet volume measured. The pellet was then washed and suspended in 2 vol of Buffer A (0.3 M sucrose, 10 mM Hepes-KOH, pH 6.9, 36 mM KCl, 2 mM MgCl₂, 2 mM EGTA; Fluka, Ronkonkoma, NY) with the protease inhibitors leupeptin (0.5 μg/ml; Boehringer Mannheim), antipain (12.5 μg/ml), E-64 (10 μg/ml), and chymostatin (10 μg/ml). The resuspended pellet was homogenized by passage through a ball-bearing cell cracker (Hans Issel, Palo Alto, CA) with a nominal clearance of 0.0005 in. In some cases this crude lysate was fractionated by centrifuging in a Beckman Optima TLX ultracentrifuge with a TLA120.2 rotor at 890 g for 10 min (to generate S1 and P1 fractions), 16,000 g for 15 min (S2 and P2) and 300,000 g for 15 min (S3 and P3). Pellets were washed and suspended in the starting volume of Buffer A. Paramecium tetraurelia was grown as a monoxenic culture, and the bacteria were depleted by a brief period of cell starvation before cell homogenization (Sonneborn, 1970). Paramecium homogenates were prepared identically to those of Tetrahymena, and S. cerevisiae (strain GPY60) lysates were prepared as described (Baker et al., 1990).

PGM activity was determined using an assay based on (Joshi, 1982). 200-μl reaction volumes contained a final concentration of 2 mM Glc-1-P, 7.9 nM Glc-1-6-biP, 0.5 mM NADP, 5 mM MgSO₄, 40 mM imidazole-HCl, pH 7.4, and 0.2 U Glc-6-P dehydrogenase. Ribose-5-phosphate was added to some reactions to a final concentration of 2 mM. After all reaction components were assembled they were incubated at 4°C for 5 min to ensure complete activation of PGM. Samples were then warmed to 30°C; after a 2-min equilibration the absorbance at 340 nm was recorded every 10 s for 12 min on a microplate reader (Bio-Tek 340, Highland Park, VT). This resulted in linear plots of absorbance vs. time for all measurements used. This assay was found to be linear for solutions of 4–400 mU/ml PGM. Each assay was calibrated with purified rabbit PGM equilibrated at 4°C for 30–120 min in 40 mM imidazole, pH 7.4, 2 mg/ml BSA (Calbiochem-Novabiochem, La Jolla, CA). Protein concentrations were determined using bicinchoninic acid (Pierce, Rockford, IL).

An overnight culture of cells was pelleted for 5 min at 200 g, then washed and suspended in Buffer H (5 mM Hepes-KOH, pH 7.5, 0.65 mM CaCl₂, 0.1 mM MgCl₂) for overnight starvation at 30°C without shaking. Starved cells are optimal for labeling and secretion. Cells were concentrated by centrifugation to a final concentration of 2–9 × 10⁶ cells/ml and labeled by addition of 1/9 vol [³²P]orthophosphoric acid (DuPont-NEN, 8500-9120 Ci/mmol), to a final concentration of 1 mCi/ml, and incubated at room temperature for 45 min. To induce secretion, 1/5 vol of saturated picric acid was added, followed within ∼2 s by the addition of 2 vol of 95°C SDS-PAGE sample buffer (190 mM Tris-HCl, pH 6.95, 15% glycerol, 6% SDS, 15 mM EDTA, 0.002% bromophenol blue). Samples were heated to 95°C for 10 min, separated by SDS-PAGE, and electrophoretically transferred to nitrocellulose. Membranes were then washed (3 × 1 min) with 1× TBS (20 mM Tris-HCl, pH 7.6, 137 mM NaCl), dried, and exposed to film (Kodak XOMAT AR) for 12–60 h or to PhosphorImager screens (Molecular Dynamics, La Jolla, CA) overnight. After exposure, membranes were stained for total protein using Ponceau S. All experiments were performed at least three times, with similar results.

Cells were grown to 3–6 × 10⁵/ml, pelleted for 5 min at 200 g, washed and suspended in Buffer B (50 mM Bis-Tris-HCl, pH 6.8, 105 mM NaCl, 5 mM CaCl₂, and 1 mM NaH₂PO₄), and homogenized as for PGM assays. 25 μg of this lysate was incubated in Buffer B with 2.8 μCi [β-³⁵S]UDP-Glc (2.8 mCi/ml, 345 Ci/mmol; DuPont NEN, Wilmington, DE) and 0 or 200 μM glucose-1-phosphate, plus 0 or 200 μM UDP-glucose in a total volume of 50 μl for 30 min at room temperature (Marchase et al., 1993). 2 vol of 95°C SDS-PAGE sample buffer were added, and samples were visualized as described above.

Cells were grown to 3–6 × 10⁵/ml, washed and suspended in DMC (0.2 mM sodium phosphate, pH 6.9, 0.17 mM sodium citrate, 0.1 mM MgCl₂, 0.65 mM CaCl₂) at a concentration of 10⁶/ml, and incubated for 6–9 h at room temperature. Aliquots of cells were stimulated by the addition of 1/4 vol Alcian Blue 8 GX in water (concentrations given in text) for 15 s. Cells were then fixed and processed for immunofluorescence using mAb 4D11, which recognizes the granule core protein p80, as previously described (Turkewitz and Kelly, 1992). Cells in randomly chosen fields were scored as either unstimulated or stimulated based on the presence or absence of docked granules.

In both Paramecium tetraurelia and Tetrahymena thermophila, synchronous exocytosis of dense-core secretory granules can be triggered by a transient increase in intracellular calcium. In Paramecium, exocytosis has been correlated with dephosphorylation of parafusin, an abundant phosphoprotein. To see whether a similar protein exists in Tetrahymena, we labeled cells with [³²P]orthophosphate. One major phosphoprotein of the same apparent molecular mass as parafusin (∼60 kD) was observed in unstimulated cells, and was massively dephosphorylated in stimulated cells (Fig. 1). This suggested that a parafusin homologue with a conserved function could be studied in Tetrahymena.

To clone the gene encoding the putative homologue as well as any related proteins, we identified protein sequences conserved between Paramecium parafusin, human PGM, and yeast PGM, and designed eight degenerate PCR primers corresponding to those regions. PCR amplification using combinations of these primers, in conjunction with Tetrahymena cDNA or genomic DNA, generated six products encoding a PGM-like protein. Each overlapped with all of the others, and overlapping sequences were identical, indicating that they were derived from a single gene. The largest product hybridized to a 1.7-kB clone from a cDNA library, whose sequence was identical to that of the gene identified by PCR. We extended the search for potential homologues with low stringency Southern blots of genomic DNA, using probes made both from S. cerevisiae PGM (GAL5) as well as the isolated Tetrahymena cDNA clone. Both probes detected a single macronuclear gene (see Fig. 3). Based on these lines of evidence, it appears that Tetrahymena has only a single gene homologous to PGM and parafusin, which we named PGM1 for reasons described below. Northern blots (see Fig. 6 C) were also consistent with a single gene.

Both cDNA and genomic DNA sequences were determined. The PGM1 sequence displayed characteristics of a Tetrahymena thermophila gene: several short introns and the 5′ and 3′ UTRs were rich in AT-rich sequence, and codon usage showed the expected bias (Fig. 2) (Horowitz et al., 1987; Martindale, 1989). Southern blotting of total Tetrahymena DNA indicated two bands hybridizing with PGM1, which differed in intensity 20-fold (Fig 3). This is the expected difference between macronuclear and micronuclear alleles, since the macronucleus contains 45 copies of each gene, while the micronucleus is diploid. The micronuclear allele appeared to contain a ∼2.0 kB insertion, known as an IES. This was confirmed by direct cloning of the IES, which was determined to lie within intron 4 (Fig. 2). As expected, this micronuclear copy persisted in clones whose macronuclear alleles were disrupted (not shown).

The longest ORF in PGM1 is predicted to encode a 587-residue protein with a calculated mass of 65.5 kD. This product, Pgm1p, is closely related to P. tetraurelia parafusin (60% identity) and known eukaryotic PGMs (40– 50% identity). We note that T. thermophila and P. tetraurelia, while both ciliates, diverged at least several hundred million years ago (Wright and Lynn, 1997). To evaluate the significance of the observed differences between these proteins, we used the known crystal structure of rabbit muscle PGM (Dai et al., 1992) as a framework upon which to compare a group of PGM-related proteins. These included Paramecium parafusin and H. sapiens PGM5, both of which have been speculated to lack enzymatic activity in part on the basis of sequence analysis (Andersen et al., 1994; Subramanian et al., 1994; Moiseeva et al., 1996) and O. cuniculus PGM2, a sarcoplasmic reticulum-associated phosphoprotein (Lee et al., 1992). Strikingly, all six residues predicted to be involved in substrate binding and enzymatic catalysis in conventional PGM (Dai et al., 1992) are completely conserved in these genes (Figs. 4 and 5), arguing that all are enzymatically active. As has been previously noted, parafusin contains several insertions and deletions relative to mammalian PGMs (overlined in Fig. 4); it has been hypothesized that these differences account for a non-PGM activity in parafusin (Subramanian et al., 1994). Tetrahymena Pgm1p contains insertions and deletions at the same positions. However, neither the size nor the sequence of the insertions is conserved between Tetrahymena and Paramecium. Moreover, molecular modeling predicts that all insertions and deletions in the ciliate proteins relative to the mammalian enzyme occur at the end of helices or in regions of random coil, and are located at the periphery of the molecule (Fig. 5). Thus, residues predicted to be involved in enzyme catalysis are maintained in both parafusin and Pgm1p, and variable regions lie in parts of the molecule that are not likely to be important for either folding or enzymatic activity.

To analyze the in vivo role of PGM1, a disruption vector (Gaertig et al., 1994) was used to target the PGM1 gene. The vector was designed so that its recombination with the PGM1 locus eliminated the first 80% of PGM1, including the start codon (Fig. 6,A). Disruption occurs by replacement of macronuclear PGM1 sequences with neo2, which confers resistance to paromomycin. Tetrahymena have both a micro- and a macronucleus, but only the latter is transcribed during vegetative growth. The macronucleus contains ∼45 identical copies of most genes (Bruns, 1986). Immediately after transformation, vector integration results in replacement of a small number of the macronuclear copies. Transformed clones are then grown in increasing concentrations of paromomycin, to select for cells containing a progressively higher fraction of disrupted alleles, through a process called phenotypic assortment (Bruns, 1986). After phenotypic assortment, we verified the complete replacement of macronuclear PGM1 through Southern and Northern blotting (Fig. 6, B and C). A second disruption strategy was also used, in which neo2 was inserted in the 5′ NspV site of PGM1 (not shown). Transformants created using either construct showed identical phenotypes. Furthermore, Southern blotting using a probe derived from neo2 demonstrated that the targeting construct had not integrated elsewhere in the genome (not shown). After complete gene disruption, cells were transferred to drug-free media for subsequent analysis. ΔPGM1 cells displayed no noticeable defects in growth or morphology.

A well-characterized biochemical property of parafusin in Paramecium lysates is that it serves as the acceptor of a glucose phosphotransferase (Satir et al., 1990). To determine whether this property was shared by Tetrahymena Pgm1p, we prepared cell lysates and incubated them with the donor molecule for this reaction, [β-³⁵S]UDP-glucose. A single band with the mobility predicted for Pgm1p was labeled in lysates prepared from wild-type cells (Fig. 7 A). In contrast, no band was labeled in parallel experiments done with ΔPGM1 lysates.

Labeling of PGM in this reaction mix could also be due to the presence of a breakdown product, [α-³⁵S]Glc-1-P, which can label the active site of PGM (Marchase et al., 1987). To eliminate this possibility, cell lysates were incubated with [β-³⁵S]UDP-glucose in the presence of a 200-fold excess of either unlabeled UDP-Glc or Glc-1-P (Fig. 7 B). UDP-Glc, but not Glc-1-P, nearly abolished the labeling, indicating that glucose phosphotransferase activity and not active site labeling was primarily being observed.

Our initial observation was that a ∼60-kD phosphoprotein underwent dephosphorylation in vivo upon exocytosis induction in Tetrahymena. To establish this parafusin homologue as Pgm1p, the experiment was repeated with ΔPGM1 cells. As anticipated, the parafusin homologue at that position was no longer present (Fig. 7 C). In summary, similarity between parafusin and Pgm1p is seen both in sequence and in two kinds of characteristic phosphorylation. No other protein in T. thermophila had this combination of qualities.

To determine whether Pgm1p was enzymatically active, we measured PGM activity in crude cell lysates prepared from wild-type or ΔPGM1 cells (Fig. 8). PGM activity in ΔPGM1 cells was reduced 92% (±1.8%) relative to wild-type cells, indicating that Pgm1p accounts for most of the PGM activity in T. thermophila. Since data presented above argue that T. thermophila does not contain other PGM-like genes, the residual PGM activity observed in ΔPGM1 cells might be accounted for by unrelated phosphomutase(s) with broad specificity. Similarly, S. cerevisiae contains proteins other than PGMs that can catalyze interconversion of glucose-1-P and glucose-6-P (Boles et al., 1994; Hofmann et al., 1994). We asked whether various monosaccharide phosphates, known to act as competitive inhibitors for unrelated phosphomutases, would affect the apparent PGM activity in wild-type or ΔPGM1 lysates. Supporting our hypothesis, ribose-5-phosphate was found to be a more potent inhibitor of the residual activity present in ΔPGM1 cell lysates (82% inhibition) than of the activity in wild-type lysates (30% inhibition; Fig. 8) or purified rabbit PGM (35% inhibition; not shown). Similar differential inhibition was seen with mannose-1-P (not shown). This argues that the residual PGM activity in ΔPGM1 cells comes from enzymes with substantially different properties from classically defined PGMs. Candidates include phosphopentomutase, phosphomannomutase, and N-acetylglucosamine phosphomutase.

To examine the localization of Pgm1p in Tetrahymena, cell homogenates were fractionated by centrifugation, and PGM activity was determined. Less than 0.1% of the total PGM activity could be pelleted (Table II). It remains possible that posttranslational modification of Pgm1p induces a weak or transient membrane attachment. A second potential complication is that a membrane-localized population of PGM might have a latent activity, as has been reported for a sarcoplasmic reticulum-associated PGM (Lee et al., 1992). However, a variety of treatments which had been shown to activate latent PGM activity in that system, including nonionic detergents, high salt, and denaturants, failed to uncover latent activity in Tetrahymena membranes. Therefore, Pgm1p appears to be a soluble protein with no detectable membrane association.

Tetrahymena contain a large number of dense-core granules, called mucocysts, whose exocytosis can be easily triggered and monitored. Virtually all granules in resting cells are docked in positions from which they can fuse and release their contents within seconds of stimulation. This dramatic release can be detected by immunofluorescence visualization of docked granules using a mAb against a secreted granule protein. Before stimulation, both wild-type and ΔPGM1 cells displayed an identical pattern of docked granules, indicating that granule synthesis, targeting, and docking had taken place (Fig. 9). Cells were then visualized after 15 s of stimulation, and individual cells in randomly chosen fields were scored as either unstimulated or having undergone extensive exocytosis. Under optimal stimulation conditions, ΔPGM1 cells displayed the same response as wild-type cells, both showing virtually complete exocytosis in >97% of the cells (Fig. 10). A number of suboptimal stimulation conditions (lower concentrations of secretagogue) were tested to see whether the sensitivity of the ΔPGM1 cells might be altered. Again, the responses of ΔPGM1 and wild-type cells were indistinguishable. In summary, ΔPGM1 cells appeared identical to wild-type in both the extent of exocytosis and the stimulation threshold.

Several other tests were performed to assess the ΔPGM1 exocytotic response. At higher concentrations of secretagogue, >90% of both wild-type and ΔPGM1 cells were seen to become encapsulated, an assay that has previously been used to indicate extensive and synchronous secretion (Tiedtke, 1976). We tested the response of ΔPGM1 cells to other secretagogues including dibucaine, the calcium ionophore A23187, electrical shock, and picric acid. Secretion was evaluated by several methods: immunofluorescence as previously described; observing cell behavior (secreted protein causes cells to clump in a distinctive fashion); isolating secreted proteins (Turkewitz et al., 1991); and electron microscopy (Chilcoat et al., 1996). In all cases, the responses of ΔPGM1 cells were indistinguishable from those of wild-type (data not shown).

Parafusin has been the subject of a large number of papers, and its potential role in regulated exocytosis has been widely cited. The most significant evidence supporting such a role includes the demonstration of dramatic dephosphorylation on a <80-ms time scale, indicating that this modification occurs rapidly enough to be playing a regulatory role in membrane fusion, and that the dephosphorylation does not occur in exocytosis-deficient mutant cells (Zieseniss and Plattner, 1985). The potential significance of these findings was underscored by the finding that a related protein undergoes posttranslational modification upon depolarization of rat brain synaptosomes, suggesting that this step represented an evolutionarily maintained aspect of regulated exocytosis for both dense-core granules and synaptic vesicles (Veyna et al., 1994). This model also fits with data obtained in numerous systems that indicate a role for protein phosphatases during exocytotic triggering (Verhage et al., 1995; Raufman et al., 1997). Nonetheless, these data can be broadly described as circumstantial, and no direct evidence has established a functional role for parafusin or its dephosphorylation.

We approached these questions by taking advantage of an organism whose physiological and experimental features make it highly advantageous for studies of regulated exocytosis. Like Paramecium, Tetrahymena can undergo rapid synchronous exocytosis of dense-core secretory granules under controlled conditions. In addition, recent methodological advances facilitate targeted gene disruption, thereby allowing in vivo analysis of protein function. We have identified and cloned the parafusin homologue in T. thermophila. Results of gene disruption demonstrate that the PGM1 gene encodes enzymatically active PGM. The data do not support a role for Pgm1p in regulated exocytosis.

A parafusin homologue was identified using a variety of approaches, including degenerate PCR, library screening, low stringency Southern blotting, Northern blotting, PGM activity assays, in vivo stimulus-dependent dephosphorylation, and in vitro glucosylphosphotransferase assays. All of these approaches identified PGM1, but no other related genes. This experimental redundancy implies that we have identified the parafusin homologue, demonstrates that it encodes active PGM, and indicates that there are no other PGM-like proteins in Tetrahymena. The gene product displays several highly characteristic features of parafusin.

While Tetrahymena contains a single PGM, multiple isoforms of the enzyme are found in other eukaryotes. In mammals, isoforms of metabolic enzymes with apparently identical functions are commonly observed, sometimes with differences in tissue distribution or gene regulation. Such is the case with human PGM (reviewed in Whitehouse et al., 1992), and a similar situation may account for the number of isoforms in rabbit. The two isoforms of PGM in the generally compact genome of S. cerevisiae result from an ancient large-scale chromosomal duplication (Boles et al., 1994). In P. tetraurelia, where at least two isoforms of parafusin exist, multiple isoforms are the norm rather than an exception (Madeddu et al., 1995). Our results strengthen the impression that single-copy genes are common in T. thermophila.

The identification of a novel PGM from the highly divergent ciliate kingdom led to some interesting sequence-based observations. PGM is an ancient enzyme, highly conserved even between prokaryotes and eukaryotes (Lu and Kleckner, 1994), and we asked if intron/exon boundary sites are conserved between T. thermophila PGM1, H. sapiens PGM1, and P. tetraurelia parafusin. Except for the first intron of Paramecium and the second of Tetrahymena, no boundary sites were conserved (Putt et al., 1993; Hauser et al., 1997). In none of these genes do intron/exon junctions coincide with polypeptide domain boundaries (as determined by Dai et al., 1992).

A number of recently sequenced PGM-like genes have been predicted to lack enzymatic activity because of the perceived divergence of important amino acid residues. The suspected absence of PGM activity was indeed one of the strongest arguments that these proteins have novel roles. As a divergent PGM, the Tetrahymena protein is especially useful for evaluating these predictions. When the deduced product of PGM1 was aligned with these proteins, it was apparent that active site residues were universally conserved. This is in contrast to what was originally stated for parafusin (Subramanian et al., 1994), but is in agreement with the recently observed activity of parafusin expressed in vitro (Hauser et al., 1997). The short insertions and deletions present in both parafusin and Pgm1p, relative to mammalian PGMs, are predicted by molecular modeling to lie in regions where small changes will have little or no effect on enzyme activity. The lack of sequence conservation between the Tetrahymena and Paramecium proteins in these regions does not support the proposal that they endow parafusin with a specialized role in exocytosis, as previously proposed (Subramanian et al., 1994). H. sapiens PGM5 (also known as aciculin), is a dystrophin- and utrophin-associated protein that was suggested, based in part on sequence analysis, to lack enzymatic activity (Belkin et al., 1994; Moiseeva et al., 1996). Nonetheless, it shows complete conservation of active site residues. Similarly, while a Gly-to-Ala substitution in the Mg²⁺-binding hairpin of aciculin was perceived as incompatible with enzymatic activity, both the ciliate sequences predict an Ala-to-Gly substitution in a symmetric position of this hairpin. Given these observations, the reported existence of PGM-like proteins lacking PGM activity may deserve reconsideration.

If phosphorylation plays a regulatory function, the sites of phosphorylation might be conserved among divergent family members. In vivo data suggest that (Ser/Thr) kinases/phosphatases may be involved (Momayezi et al., 1987; Verhage et al., 1995; Raufman et al., 1997), and parafusin can be phosphorylated in vitro by casein kinase and dephosphorylated by calcineurin (Kissmehl et al., 1996, 1997). Disallowing the active site, only 4 of 11 Ser/Thr residues conserved throughout eukaryotes are expected from the crystal structure to be on or near the surface of the protein: residues 30, 286, 584, and 587 (Tetrahymena Pgm1p numbering). Only Thr30 and Thr286 are on the same face of the enzyme as the active site, and molecular modeling suggests that phosphorylation of Thr30 would block entry of substrate.

We evaluated the function of PGM1 by generating three independent cell lines in which all macronuclear copies of PGM1 were disrupted. Lysates prepared from ΔPGM1 cells contain a dramatically reduced level of PGM activity, and the residual activity appears to be the result of an unrelated enzyme(s). The specific activity of PGM in wild-type Tetrahymena was two- to fivefold higher than that usually reported in S. cerevisiae, plants, or liver (Hanson and McHale, 1988; Oh and Hopper, 1990; Mithieux et al., 1995), but was identical to that we observed in Paramecium.

To test if PGM1 was required for regulated secretion, we induced exocytosis using a variety of stimulatory conditions. Wild-type and ΔPGM1 cells displayed indistinguishable sensitivity to stimulation, and extent of release. Since Pgm1p became dephosphorylated under stimulation conditions, our results appear to uncouple these two previously correlated phenomena and indicate that this protein is not essential for exocytosis. Two alternative interpretations deserve mention, although we consider both to be unlikely. First, the Tetrahymena protein may have lost a function conserved in other eukaryotes; however, it is then necessary to explain why rapid dephosphorylation has been maintained. Second, although we stimulated cells using a variety of different conditions, our quantitative measure was based on cells fixed 15 s after stimulation. Because this period is long relative to the minimum time required for membrane fusion, it is possible that ΔPGM1 and wild-type cells have different kinetics of release, which cannot be detected by our assay. In this case, one would have to postulate that Pgm1p acts as a kinetic effector of exocytosis that has no effect on either the threshold or extent of release.

The lack of an observable phenotype in a gene disruption experiment may be due to suppression by an offsetting second site mutation to maintain an advantageous trait. For the results reported here, this could theoretically involve selection for PGM activity or regulated exocytosis. Both seem highly unlikely for the following reasons. In a variety of organisms, PGM mutants with only mild phenotypes have been obtained (Hanson and McHale, 1988; Boles et al., 1994; Lu and Kleckner, 1994; Zhou et al., 1994), indicating that PGM is not an essential enzyme under many growth conditions. Similarly, regulated exocytosis in Tetrahymena appears completely dispensable for growth in the laboratory: a variety of exocytosis-deficient mutants have been obtained that have no growth defects and are nonreverting (Maihle and Satir, 1985; Turkewitz et al., 1991; Gutierrez and Orias, 1992). Finally, indistinguishable results were obtained with three independent transformants.

In summary, our results indicate, in contradiction to the central hypothesis of previous work on parafusin, that Pgm1p is not required upstream of membrane fusion during regulated exocytosis. Some of the earlier evidence can easily be reinterpreted in light of these findings. The absence of parafusin dephosphorylation in exocytosis mutants or in the presence of exocytosis inhibitors (Gilligan and Satir, 1982; Zieseniss and Plattner, 1985) was originally interpreted as evidence that the genetic or pharmaceutical blocks are epistatic to dephosphorylation of parafusin, which was postulated to be epistatic to exocytosis. However, the data are equally consistent with a model in which exocytosis is epistatic to dephosphorylation, leaving open the possibility that parafusin may function in recovery from exocytosis. The time course of dephosphorylation established in Paramecium may suggest that it is not membrane fusion per se that triggers dephosphorylation, but some earlier event.

The family of PGM-like proteins, some of which have been shown to undergo changes in phosphorylation, includes a calmodulin-dependent kinase substrate in sarcoplasmic reticulum (Lee et al., 1992), a yeast PGM which undergoes phosphoglucosylation upon heat shock (Dey et al., 1994), and a dystrophin/utrophin-associated PGM found in adherens junctions (Moiseeva et al., 1996). A possible link between these proteins may be their sensitivity to fluctuations in intracellular calcium. Modification appears to occur under conditions of increased free cytosolic calcium associated with secretion in regulated secretory cells and contraction in muscle (Clapham, 1995; Erxleben et al., 1997). In addition, localization of calcium-sensitive proteins to the adherens junction suggests that it is also a site of calcium flux (Burridge et al., 1988). Therefore, PGM-like proteins may respond (via kinases/phosphatases or phosphoglucose-transferases/esterases) to calcium levels. What is the nature and function of this response? Given the rapidity and extent of the modifications, we considered that the dephosphorylated enzyme may play a role in demobilization of the high calcium required to trigger exocytosis, for example by increasing the rate of calcium uptake into the vesicular stores that underlie the plasma membrane of ciliates, or into the ER or SR in mammalian cells. Recent experiments in mammalian cells suggest that levels of glucose can influence calcium transport, which could mean that modulation of PGM activity affects free calcium levels (Darbha and Marchase, 1996). Therefore, one might expect to see effects from perturbation of calcium homeostasis in ΔPGM cells, but these have not been reported in either S. cerevisiae (Boles et al., 1994) or in Tetrahymena (Chilcoat, N.D., unpublished observation). Further development of these or other in vivo systems may help to address this issue.

We thank Pam Bounelis and Richard Marchase (University of Alabama Birmingham) for advice and for the generous gift of [β-³⁵S]UDP-Glc, Paul Gardner for help with DNA sequencing, Denis Nyberg (University of Illinois Chicago) for Paramecium cultures, and Celeste Richardson for S. cerevisiae cultures and lysates. Our thanks also go to Alex Haddad, Phoebe Rice, Nava Segev, Ted Steck, and Ben Glick for discussion and comments.