Core formation and the acquisition of fusion competence are linked during secretory granule maturation in Tetrahymena

Abstract

The formation of dense core secretory granules is a multi-stage process beginning in the trans-Golgi network and continuing during a period of granule maturation. Direct interactions between proteins in the membrane and those in the forming dense core may be important for sorting during this process, as well as for organizing membrane proteins in mature granules. We have isolated two mutants in dense core granule formation in the ciliate Tetrahymena thermophila, an organism in which this pathway is genetically accessible. The mutants lie in two distinct genes but have similar phenotypes, marked by accumulation of a set of granule cargo markers in intracellular vesicles resembling immature secretory granules. Sorting to these vesicles appears specific, since they do not contain detectable levels of an extraneous secretory marker. The mutants were initially identified on the basis of aberrant proprotein processing, but also showed defects in the docking of the immature granules. These defects, in core assembly and docking, were similarly conditional with respect to growth conditions, and therefore are likely to be tightly linked. In starved cells, the processing defect was less severe, and the immature granules could dock but still did not undergo stimulated exocytosis. We identified a lumenal protein that localizes to the docking-competent end of wildtype granules, but which is delocalized in the mutants. Our results suggest that dense cores have functionally distinct domains that may be important for organizing membrane proteins involved in docking and fusion.

Introduction

In the secretory pathway, each step of vesicle budding involves selection of both lumenal and membrane proteins. The best-understood sorting mechanisms involve direct and indirect links to cytosolic coat proteins and their associated adaptor complexes (1–3). Some vesicles, however, form without the benefit of such coats, by mechanisms that are less well established. Dense core granule (DCG) biogenesis at the trans-Golgi network (TGN) does not appear to involve cytosolic coats, so other mechanisms must allow the cargo to be segregated from proteins destined for constitutive release, lysosomes, or retrograde transport (as discussed in references 4–6). An important aspect of sorting is likely to be the tendency of DCG proteins to self-aggregate under the conditions encountered in the TGN (7, 8). How such aggregates are then sorted to a single vesicle type, together with specific membrane proteins, is not entirely clear. Recently, several lines of indirect evidence have suggested that protein sorting may occur in part via interactions between the lumenal domains of membrane constituents and the proteins that condense to form the dense core (9, 10). By this model, the self-aggregation of lumenal proteins could induce formation of membrane microdomains enriched in specific proteins. Alternatively, pre-existing membrane microdomains, perhaps defined by their lipid composition, could serve as platforms for lumenal protein aggregation (11). Consistent with this idea, several lumenal DCG cargo proteins appear to interact in non-conventional ways with cholesterol-rich bilayers (10, 12, 13). These models are not mutually exclusive, and both posit that interactions between core proteins and membrane constituents are important in determining DCG composition during biogenesis, during budding from the TGN and/or maturation.

A related but non-identical question is whether such interactions persist or develop in mature DCGs, where they might play a role in organizing the membrane. A provocative example comes from the ciliate Paramecium, in which secretory granule membrane proteins involved in docking and fusion may localize based on the underlying dense core structure (14). In this protozoan, the secretory granules contain an elongated tip structure, clearly distinct from a protein lattice that forms the dense core, which is directly involved in docking and fusion (15). Analysis of Mendelian mutants defective in tip formation revealed that tip assembly is essential for efficient docking and fusion (14, 16), and silencing of lattice-forming genes demonstrated that the tip cannot assemble independently of the lattice (17). While the proteins that make up the dense-core lattice have been isolated and the corresponding genes cloned (18), those involved in tip formation have not yet been identified. Whether the Paramecium tip offers a general model for core-driven membrane sub-domain formation is unknown, but is called into question by the fact that distinct tips are not a known feature of secretory granules in other eukaryotes, and in fact are not even seen in some other ciliates. These include Tetrahymena thermophila, another organism in which both genetic and molecular studies of secretory granule formation have been pursued (19). The lattice-ordered core in Tetrahymena granules consists of homologs of the Paramecium lattice proteins (20), but no tip structure is visible.

In this paper we describe two mutants in T. thermophila, generated by random chemical mutagenesis, which are defective in secretory granule formation. A single well-characterized mutant in DCG synthesis per se (as opposed to docking, fusion, etc.) has previously been described. Called SB281, that mutant is grossly defective in the sorting of DCG lumenal cargo proteins, probably at the level of the TGN, and these are instead rapidly and constitutively secreted in the proprotein form (21, 22). We have now extended this approach to intermediates in DCG synthesis. Mutant lines UC620 and UC623 are genotypically distinct though phenotypically similar. DCG core proteins are efficiently sorted, but nonetheless accumulate in vesicles that are deficient in several characteristic DCG maturation activities, including the proteolytic processing of granule cargo precursors, the assembly of lattice cores, and the docking of vesicles to the plasma membrane. These phenotypes suggest that the mutations affect an early stage of post-TGN granule maturation, and the striking observation is that core assembly and docking are tightly linked. In the course of developing markers to characterize the mutants, we identified a granule core protein with a tip-like localization in wildtype granules, which becomes delocalized in the mutant granules. The corresponding gene, unrelated to those involved in formation of the lattice core, has strong homologs in Paramecium, and represents a candidate for a core protein involved in organization of a membrane docking/fusion domain.

Results

Isolation of exocytosis mutants

Exocytosis mutants in T. thermophila can be isolated on the basis of the wildtype cells’ response to the polycyclic cation Alcian Blue, which triggers global synchronous secretion from docked DCGs (23). Alcian Blue also appears to cross-link the DCG proteins as they are being released, immobilizing each cell in a robust capsule. In contrast, exocytosis-defective cells remain free swimming under the same conditions. Following previous work (24), we mutagenized cells with nitrosoguanidine and then exploited a trick of Tetrahymena genetics to derive homozygous progeny, as described in Materials and Methods, thereby uncovering any recessive mutations. After Alcian blue stimulation, we isolated the non-encapsulated fraction, i.e., free-swimming cells which migrate toward the air-water interface. This fraction would be expected to include both bona fide exocytosis mutants, as well as wildtype cells that underwent exocytosis but either failed to form, or rapidly escaped from, capsules. The capsule formation step was repeated to enrich for the desired mutants, and the roughly 600 free-swimming cells from the second round were distributed into 96-well plates at a density estimated to deliver roughly 30 cells per plate, in which about 75% of the wells would have arisen from a single Tetrahymena clone. We then tested the exocytosis competence of the individual clones, using Alcian Blue and evaluating each well by light microscopy. Clones that showed no visible capsules were used for further analysis. Beginning with 10⁶ mutagenized cells, we obtained 69 clones that were completely defective in capsule formation.

Characterizing these 69 cell lines was directed towards identifying the subset that were defective in granule synthesis rather than in subsequent steps, such as fusion with the plasma membrane. Two assays were used. In one, we prepared whole cell lysates, then used SDS-PAGE and Western blotting to determine whether the DCG cargo proteins were proteolytically processed. The second criterion used was morphological, asking whether the cellular location of DCG cargo in the mutant strains differed from wildtype. It was important to monitor multiple proteins, to determine whether a defect was specific to a single DCG protein, or whether it more generally affected the DCG cargo. Previous work has relied largely on a single antiserum directed against Grl1p, one of a family of six lumenal DCG proteins that together constitute most of the secretory cargo. To increase the number of proteins that we could trace, we generated antibodies against Grl3p (20) and Grl8p (called Ndc1p by Chilcoat et al. (25)). Synthesized as proproteins, the Grls are proteolytically processed during granule maturation and the mature products stored in DCGs (20, 26). Rabbit antisera were raised against mature Grl proteins that were released from stimulated wildtype Tetrahymena, and their use in Western blotting is demonstrated in figure 1A and B. We also characterized a monoclonal antibody (mAb) directed against Grl3p, called 5E9, whose specificity and usefulness for localization studies is demonstrated in figure 1C and D. The antibody targets the multitude of DCGs that are docked in a linear array at the plasma membrane. This mAb was initially obtained by Marlo Nelson and Joseph Frankel (U. Iowa), who generously provided it to us, but its molecular target was not previously known.

Figure 1.

Specificity of new antibodies against Grl proteins. (A) Western blots of cell lysates, either from wildtype or cells expressing HIS-tagged variants of Grl3p or Grl8p (+HIS), were probed with anti-Grl3p and anti-Grl8p antisera. HIS-tagged protein expression was induced with 0.2 µg/ml CdCl₂ for 1 h before lysis. Cells expressing a HIS-tagged copy show both the endogenous protein (arrowhead) as well as the tagged species (arrow). The proteins were resolved by SDS-PAGE using 10% polyacryamide. (B) The HIS-tagged variants of proGrl3p (lane 1) and proGrl8p (lane 2) were purified from Tetrahymena lysates, resolved by SDS-PAGE (10% polyacrylamide) and probed after transfer to nitrocellulose with the anti-Grl3p and anti-Grl8p antisera. The appearance of doublets may reflect alternative sites of amino terminal processing. (C) The 5E9 monoclonal antibody recognizes proGrl3p. Lysates of wildtype cells, and cells expressing 6HIS-tagged Grl3p, were resolved by SDS-PAGE (10% gel) and probed on Western blots with the 5E9 antibody. The arrowhead indicates endogenous proGrl3p and arrow indicates the tagged species. (D) Indirect immunofluorescence using mAb 5E9. (left) 5E9 labels docked DCGs in wildtype cells. (right) In cells lacking all macronuclear copies of GRL3 (ΔGRL3), DCGs are still present (not shown), but are not recognized by the 5E9 antibody.

Molecular cloning of a DCG marker unrelated to the GRL family

To extend the range of our analysis to non-Grl proteins, we focused on an 80kD species (p80) also released by regulated exocytosis. Several characteristics of this protein previously suggested that p80 was not closely related to the Grls (27). Among these differences, p80 did not appear to be endoproteolytically processed in the chase period following biosynthetic labeling. To confirm this, we have now isolated the protein and obtained the N-terminal sequence. This was used to design degenerate PCR primers, with which we amplified and cloned the full-length gene. This confirmed the absence of proteolytic processing, since the N-terminal sequence of the ~80kD secreted protein, determined by Edman degradation, corresponded to the sequence immediately following the predicted signal peptide in the translated cloned gene.

As previously demonstrated, p80 is recognized by mAb 4D11 (27), an antibody useful for indirect immunofluorescent visualization of DCGs. Confirming the cloning, 4D11 did not illuminate DCGs of Tetrahymena in which the p80-encoding gene, which we have given the name GRT1, had been disrupted (Figure 2A).

Figure 2.

Cloning and immunolocalization of Grt1p. (A) The mAb 4D11 recognizes Grt1p. Indirect immunofluorescence using 4D11 illuminates cortical rows of docked DCGs in wildtype cells (left panel), but not in cells lacking all macronuclear copies of GRT1 (ΔGRT1) (right panel). Analysis of ΔGRT1 cells with a different antibody (5E9) demonstrates that the absence of signal with 4D11 is not due to absence of DCGs per se (bottom panel) (B) Deduced features of Grt1p and Grl1p. Numbers refer to amino acid primary sequence, and secondary structure was predicted using DNASTAR. (C) Alignment of the C-terminus of Grt1p with the C-termini of the Tetrahymena DCG protein Igr1p and the putative Paramecium homolog Pcm3p. Sequence identity is indicated by an asterisk, and similarity is indicated by one or two dots. Conserved cysteine residues are boxed in grey. (D) The Grt1p epitope is sub-localized within individual granules. Grt1p was localized by indirect immunofluorescence using 4D11 and confocal microscopy, in wildtype (left panel) or MN173 (right panel). The image of wildtype is a tangential optical section through the docked DCGs, with an inset showing a sagittal section at higher magnification. The image of MN173 is an optical slice through the cytoplasm of the cell. Bar = 1 µm (E) Electron micrograph of an individual DCG, shown docked at the plasma membrane. There is no discernable difference between one end of the DCG and the other. Bar = 0.2 µm.

Analyzing the amino acid sequence confirmed that GRT1 is not a member of the GRL family; most obviously, it does not contain the acidic coiled-coil domains characteristic of the Grl proteins (20). Instead, Grt1p is predicted to fold primarily as β-sheets (Figure 2B). Sequence analysis suggests that it contains four tandem copies of a single ~155 amino acid domain, plus a distinct C-terminal domain. The protein is unusually rich in tryptophan residues (16/766), which are mostly conserved between adjacent domains. The C-terminus bears significant homology with the C-terminal domain of the protein Igr1p (Figure 2C), which is the only previously identified Tetrahymena granule cargo protein that is not related to the Grl’s (28). This domain may therefore be used to define a distinct class of granule cargo proteins in Tetrahymena.

Interestingly, we found evidence that distribution of Grt1p in DCGs may be different from that of Grls. While the reactivity of mAb 5E9 suggested that Grl3p is distributed throughout the granule core (Figure 3B, wildtype), mAb 4D11 recognized an epitope that is concentrated at the docking end, producing half-elliptical patterns of immunofluorescence in DCGs viewed from the side (Figure 2D). The epitope also appears concentrated at the periphery of the lumen, judging by the distribution of immunofluorescence in a ring-like pattern in those granules viewed down their long axis. This suggests that Grt1p has a polarized distribution within individual vesicles, although another possibility is that the epitope is generated or exposed only in a localized subset of the Grt1p cohort. The polarized distribution is established prior to docking, rather than being a consequence of docking, since the same half-elliptical shapes were observed in MN173 mutant cells, which produce normal DCGs that fail to dock at the plasma membrane (24)(Figure 2D). We named the protein Grt1p (for Granule tip) based on this apparent localization. The polarized distribution of the Grt1p epitope was a surprise, since EM images of Tetrahymena DCGs gave no hint of a specialized zone at the tip (Figure 2E).

Figure 3.

Incomplete processing and aberrant localization of Grl precursors in UC620 and UC623. (A) Whole cell lysates of growing cells were resolved by SDS-PAGE (17.5% polyacrylamide) and analyzed by Western blotting, using 3 antisera against Grl proteins. In each panel, the arrowhead indicates the position of the Grl proprotein, and the asterisk indicates the corresponding processed product. (B) Grl3p in UC620 and UC623 is contained in cytoplasmic puncta distinct from the wildtype pattern of docked DCGs. Cells in early log phase were fixed and immunostained with mAb 5E9. In wildtype, the increasing intensity of the signal towards the DCG tip is likely to reflect differential accessibility to the antibody, as no such gradation is observed for cytosplasmic DCGs in the MN173 strain (not shown). Each of the upper panels is a compilation of multiple optical slices, to reconstruct the top ~50% of a single cell. The bottom panels show compiled sections collected from the middle third of the same cells. Bar = 10 µm.

Since granule tips are, in contrast, prominent in Paramecium, we asked whether GRT1 homologs were present in that organism. At least 16 matches in the Paramecium tetraurelia macronuclear database gave matches with probability scores of 1e-20 or better, with the best scores corresponding to genes in the PCM family (29). Although initially identified as putative cortex proteins, our previous analysis suggested that the PCM proteins were instead likely to be dense core granule components (discussed in (28)). The Pcm proteins show extensive similarity over almost the entire protein length (27% identity, 43% positive, over 434 amino acids for PCM4). Several of the PCM homologues, including PCM3, contain a C-terminal region that is similar to that found in Grt1p and Igr1p (Figure 2C). A conserved domain search (30) found that this domain is distantly related to the β/γ-crystallin family of metazoan proteins, which are predicted to fold into a “Greek key” motif (31). Notably, two pairs of cysteine residues may contribute to the stability of the structure by forming disulfide bonds between individual strands of the β-sheet (28), and are conserved between the ciliate proteins. A weaker putative GRT1 homolog was also identified in Plasmodium falciparum, an apicomplexan parasite that is distantly related to the ciliates, and which also has a regulated secretory pathway.

Analysis of DCG cargo processing and localization in the exocytosis deficient mutant

To screen the 69 exocytosis mutants for those with defects in proteolytic processing, we used antisera against Grl1p, Grl3p and Grl8p in Western blots of cell lysates. In most lines as in the wildtype, the majority of protein was present in the processed form. This reflects the accumulation of a large amount of material in mature DCGs relative to the amount of newly synthesized material, in transit through the early secretory compartments, that has not yet undergone processing. However, several lines showed the inverse, relatively high levels of proGrl1p, proGrl3p and proGrl8p and a smaller amount of the corresponding processed products. Two such lines, called UC620 and 623, are shown (Figure 3A).

The entire set of exocytosis deficient mutants were also fixed, permeabilized, and screened for aberrant localization of Grl3p using mAb 5E9. UC620 and 623 were again distinguished by a non-wildtype appearance, consisting of undocked cytoplasmic puncta (Figure 3B). This aberrant localization was also true for other DCG markers, namely Grt1p as well as a GFP-tagged copy of Grl1p that we expressed in both wildtype and mutant cells (Figure 4). In UC620 and UC623, each of the proteins was found in numerous cytoplasmic puncta, clearly distinct from the pattern of docked DCGs in wildtype cells. In both mutants as well as wildtype, pairwise analysis indicated extensive co-localization of these markers, demonstrating that Gr11p, Grl3p, and Grt1p were chiefly targeted to the same vesicles in the mutants.

Figure 4.

Accumulation of multiple DCG proteins within cytosolic vesicles in UC620 and UC623. (A) Co-localization of Grl1p–GFP and proGrl3p. Wildtype and mutant cell lines (indicated at left) were transformed to express GRL1-GFP, and the expression in log phase cultures was induced with 2 µg/ml CdCl₂ for 4 h. After fixation, cells were labeled for two color imaging using rabbit anti-GFP and anti-Grl3p (mouse mAb 5E9) antibodies. Individual cells were imaged in both the green (GFP) and the red (Grl3p) channels. The example shown for wildtype is a reconstruction of the top surface of the cell, roughly corresponding to a tangential section. The images of mutant cells are reconstructions of the middle two thirds of each cell. (B) Co-localization of Grl1p–GFP and Grt1p. The analysis was identical to (A) except that mAb 4D11 (anti-Grt1p), rather than 5E9, was used. Bar = 10 µm.

We used the tools available in this system to characterize the mutants genetically. The approaches, detailed in the Materials and Methods, included outcrossing to test for dominance vs. recessivity, self-crossing to ask whether more than one mutation was involved in each mutant, and mapping of mutations by crossing to strains called nullisomics, which lack individual micronuclear chromosomes (32, 33). These manipulations also allowed us to isolate the mutations of interest from the mutagenized background of the primary cells.

The mutant phenotypes were lost in the F1 generation of an outcross to wildtype, indicating that the mutations are recessive. The phenotypes were also lost in outcrosses to SB281, indicating that the mutations are not caused by weaker alleles of this previously isolated granule synthesis mutant. The results of phenotype segregation analysis were consistent with single genetic lesions in each mutant (Table 2), although we cannot rule out the possibility of multiple, tightly-linked mutations. Analysis of the progeny of crosses with a panel of nullisomic strains allowed us to assign the mutations in UC620 and 623 to chromosomes I and III, respectively, as the exocytosis defect in UC620 failed to be complemented by CU743, which lacks micronuclear chromosome I, and UC623 failed to be complemented by CU734, which lacks micronuclear chromosome III. All other nullisomic crosses resulted in phenotypically wildtype progeny. Our results indicate that the two mutants bear non-allelic mutations that result in very similar phenotypes.

Table 2.

Phenotype quantitation of B* crosses and recovery of homozygous mutants.

Strain			% mutant progeny	Isolated clonea
UC613	X	B*VI	49% +/− 8%b	UC620
UC614	X	B*VI	58% +/− 11%b	UC623

As described in Materials and Methods, heterozygous F1 lines UC613 and UC614 were mated with B*VI, and then uniparental cytogamy was induced The progeny of these crosses were distributed in 96-well plates and the exocytosis phenotypes were scored by the ability to form capsules in response to Alcian Blue. The percentage of mutant progeny was calculated from a derivation of the Poisson formula.

^aRefers to the name given to a single mutant clone from each B* cross, obtained by single cell isolation, which was used for subsequent phenotypic analysis in this work.

^bError refers to the standard deviation in the data from ten separate plates

The mutant vesicles are structurally similar to immature secretory granules

The apparently global defect in the processing of proGrl proteins in these mutants, together with the accumulation of precursors, suggests that DCG cargo protein is sorted to a storage compartment that is blocked in maturation. We therefore asked whether these vesicles morphologically resemble immature secretory granules. Cells expressing Grl1p–GFP were fixed by rapid freezing, and the vesicles labeled using anti-GFP antibodies and colloidal gold. In both UC620 and 623, the gold was concentrated over densely-filled oblong vesicles of a roughly constant size, about half the length of wildtype DCGs, though slightly more wide (Figure 5A). In some vesicles the luminal cargo was clearly ordered in a periodic structure, similar to that in wildtype cores (Figure 5B). Unlike mature wildtype granules, however, the UC620 and 623 vesicles that contained lattices also contained other material, dense but amorphous. To ask whether these vesicles were specific for DCG cargo storage or whether they also contained proteins that are constitutively secreted in these cells, we expressed GFP linked to the signal sequence from GRL1 (called ssGFP). This protein was found predominantly in the culture supernatant of non-stimulated cells, indicating that it is not stored in DCGs (Figure 6). Consistent with this, the ssGFP in wildtype cells was seen chiefly in small cytoplasmic vesicles, not overlapping with a marker in docked DCGs. Similarly, in neither UC620 nor 623 was there any significant overlap between ssGFP and DCG markers. In contrast, ssGFP and the DCG markers co-localized extensively in SB281. Taken together, these results suggest that the vesicles in UC620 and 623 resemble immature secretory granules, to which selective sorting of lumenal proteins is largely intact but in which both processing and the associated physical maturation are at least partially inhibited.

Figure 5.

DCG-like features of the cytosolic vesicles in UC620 and UC623. (A) The vesicles in UC620 and UC623 contain partially ordered cores. UC620 and UC623 were transformed to express GRL1-GFP, as in fig. 3, and analyzed by electron microscopy following cryofixation and immunostaining with anti-GFP antibodies and colloidal gold. Bar = 0.4µm. (B) The arrowhead in (A) indicates a vesicle from UC620 that is shown at high magnification (B, left panel), and a vesicle from UC623 is shown at the same magnification (B, right panel). Bar = 0.1 µm. Similar partial cores were present in UC620 and UC623 cells not transformed with GRL1-GFP (not shown). (C) As analyzed using 4D11 indirect immunofluorescence, Grt1p appears to be sub-localized within individual vesicles in wildtype, UC620, and UC623. This contrasts with SB281, in which Grt1p appears uniformly distributed within each vesicle. The wildtype image represents reconstruction from tangential sections of the top surface of a cell, including part of a row of docked DCGs. The mutant cell images are reconstructions from sagittal sections. Bar = 4 µm.

Figure 6.

The DCG-like vesicles in UC620 and UC623 do not contain constitutive secretory cargo. (A) ss-GFP is constitutively released from wildtype cells. Wildtype cells were transformed to express GFP linked to an N-terminal ER translocation signal sequence (ss-GFP). Protein expression was induced with cadmium for 4 h, and after separating the cell pellet from the media, the amount of ss-GFP in each fraction was evaluated by immunoprecipitation with rabbit anti-GFP antibodies, followed by SDS-PAGE and Western blotting with a mouse anti-GFP mAb (left panel). To control for protein release via cell lysis, cells expressing GRL1-GFP were analyzed in parallel. As expected, proGrl1p–GFP was found entirely in the cell pellet (right panel). The processed mature form of Grl1p–GFP is insoluble in wildtype cells, and is therefore not accessible to immunoprecipitation in this experiment. (B) Wildtype and mutant cell lines (UC620, UC623, and SB281) were transformed to express ss-GFP, as in (A). The indicated cell lines (at left) were then fixed and prepared for two color imaging using an anti-GFP antibody (green channel) and the anti-Grl3p antibody (red channel). Each image is a reconstruction from optical slices that transect the mid-section (~50%) of an individual cell. Bar = 10µm, except in the “merge 4x” column, where partial views of the merged images appear at increased magnification (bar 5 µm).

An independent way to evaluate core assembly in the mutants was to ask whether the distribution of the mAb 4D11-reactive Grt1p epitope within the mutant cytoplasmic vesicles resembled that in wildtype DCGs. Immunoreactive Grt1p showed a non-uniform distribution within the individual vesicles of mutant cells: the protein appeared to be concentrated at the vesicle periphery, as observed near the tip region of normal DCGs (Figure 5C, compare to MN173 in figure 2B). However, in contrast to MN173, in which a high percentage of DCGs are stained in a crescent-shaped pattern, the vast majority of immunostained vesicles in UC620/UC623 appear as circles. This suggests that the Grt1p epitope is uniformly present at the vesicle periphery in UC620/623. To ask whether this reflects an inherent association of Grt1p with the vesicle membrane, we analyzed the sub-localization of Grt1p within individual vesicles in the mutant SB281. In SB281, DCG proteins are found in vesicles whose contents appear aggregated but not organized (34). Unlike in UC620 or 623, the 4D11 mAb epitope appeared uniformly distributed within the individual SB281 vesicles. We hypothesize that the non-uniform distribution of this epitope reflects an assembly process that is shared in UC620, UC623 and wildtype, but not SB281.

Functional analysis of the immature granule-like vesicles in UC620 and 623

To gain insight into the regulated secretory pathway in Tetrahymena and to potentially illuminate the nature of the mutations in these cell lines, we developed functional assays to characterize the intermediates that accumulate in these cells. The first question was whether these immature granule-like vesicles were true intermediates; that is, were they capable of undergoing further maturation, as measured either by docking or proprotein processing? This might be the case if the defects were in cytosolic factors required for maturation. To test this, we performed cytosol mixing experiments via cell fusion. Cells were transformed with a construct encoding Grl1p–GFP under the control of the cadmium-inducible MTT1 promoter (35). The vesicles in UC620 or 623 were first labeled by inducing expression of Grl1-GFP, followed by a chase in cadmium-free medium for 1 hour. The cells were then electrofused with MN173. The DCG docking defect in MN173 was previously shown to be rescued when these cells were fused with other lines, so MN173 provided a positive control as well as allowing us to identify the desired heterotypic doublets (36). UC620 (Figure 7A) and UC623 (Figure 7B) cells were separately electrofused to MN173, and cytoplasmic exchange was allowed to occur for 1 h before the cells were fixed, permeabilized, and incubated both with mAb 5E9 (red) and anti-GFP antibodies (green). In this experiment, all vesicles containing Grl3p (derived from both the UC620/623 and the MN173 partners) should stain red, but only vesicles derived from UC620/623 will also be illuminated in the green channel. Importantly, any vesicles synthesized after the time of fusion will only be visible in the red channel, since no cadmium was present to drive expression of Grl1p–GFP.

Figure 7.

Undocked vesicles in UC620 and UC623 cannot be rescued. UC620 and UC623 cells were induced overnight to express GRL1-GFP, followed by an 85 min chase in cadmium-free medium. The cells were then electrofused to unlabeled MN173, and allowed to recover for 1 h, during which cytoplasmic mixing occurs. Cells were then fixed and analyzed by two-color indirect immunofluorescence using anti-Grl3p mAb 5E9 (red channel) and anti-GFP antibodies (green channel) with confocal microscopy. (A and B) heterotypic fusions of UC620xMN173 and UC623xMN173, respectively. (C) non-fused cells in the same samples. Using MN173 as a fusion partner served as a positive control for cytoplasmic mixing, since the docking defect in this mutant can be rescued by cell fusion (22). In A and B, the MN173 partner is oriented towards the top of the image, and the patterned cortical rows of docked granules labeled by 5E9 (red) in these cells indicate that cytoplasmic rescue has occurred. These patterns can be distinguished from the undocked DCGs in non-fused MN173, in panel C. The cytoplasmic DCG-like vesicles in UC620 and 623, labeled green with GFP, do not show a change in distribution in the paired cells (compare A and B to C), indicating that cytoplasmic mixing has not rescued the docking defect in these vesicles. All images are reconstructions of the top halves of the cells. Bar = 10 µm.

As expected, prior to fusion, the vesicles in UC620 and UC623 were visible in both red and green channels (Figure 7C). In contrast, the undocked DCGs produced by MN173 were only visible in the red channel. After fusion, large numbers of red docked vesicles are present, but none of these is also visible in the green channel. We conclude that, while cytoplasmic exchange complements the docking defect in MN173, it does not allow docking of the vesicles produced by UC620 and UC623. We could not determine whether cytoplasmic exchange facilitates further proprotein processing, because only a small fraction of cells undergoes fusion under these conditions.

Both proprotein processing and docking/fusion are conditional with respect to growth conditions

Wildtype Tetrahymena synthesize and accumulate DCGs under both growth and starvation conditions, but previous results nonetheless suggested that aspects of DCG synthesis were sensitive, in at least some mutant backgrounds, to growth conditions (26). We therefore asked if granule maturation in UC620 and 623 were affected by a shift from growth medium into a minimal inorganic medium. Indeed, the phenotypes of starved cells were markedly different from those of growing cells. First, proGrl proteins were more extensively processed in starved cells relative to growing, though still not equivalent to wildtype (compare figure 8A to 3A). Even more strikingly, starved mutant cells immunostained with mAb 5E9 exhibited an aligned array of docked vesicles, rather than the undocked vesicles seen in growth conditions (compare figure 8B to 3B). Thus both defects were significantly rescued under starvation conditions, though rescue of the lumenal activity, i.e., processing, was clearly incomplete. Further analysis showed that the same was true of the activities associated with the vesicle membrane; while the vesicles were docked, this docking was not functionally equivalent to that of wildtype DCGs. In particular, we asked whether the docked vesicles in starved cells were capable of undergoing exocytosis, like wildtype DCGs, in response to secretagogue (Figure 9A). Starved wildtype cells treated with Alcian Blue undergo extensive degranulation within seconds (23, 37). In contrast, the starved UC620 and 623 cells did not undergo visible degranulation upon stimulation. The failure to degranulate was not due to lack of lattice expansion, since other mutants lacking organized lattices (e.g., cells in which the GRL1 gene is disrupted) showed a degree of degranulation similar to wildtype in this assay (not shown). Similarly, Grt1p localization was not restored to the polarized wildtype pattern in these starved cells, and the protein appeared to be distributed around the periphery of individual docked granules (Figure 9B). The partial rescue in starvation of lumenal and membrane functions, in both UC620 and UC623, supports the idea that the processes underlying the defects are linked.

Figure 8.

Multiple maturation defects are conditional with respect to culture conditions. (A) The proGrlp processing phenotype is partially rescued by starvation. Growing wildtype and mutant cultures were pelleted, washed and suspended in nutrient-free starvation buffer (DMC) for 6 h, and the extent of proGrlp processing was assessed by probing cell lysates with anti-Grlp antibodies. Proteins were resolved by SDS-PAGE, with 17.5% polyacrylamide. The positions of pro- and mature Grlp are indicated by the arrowheads and asterisks, respectively. Molecular weight standards (in kD) are indicated at the right of each panel. (B) Under starvation conditions, UC620 and UC623 contain docked vesicles. Growing wildtype and mutant cultures were starved as in A prior to fixation and immuno-localization of Grl3p using mAb 5E9. The panel layout is identical to that of figure 2. Bar = 10 µm.

Figure 9.

Defects in regulated exocytosis and core organization are retained under starvation conditions. (A) The granule cargo vesicles in starved UC620 and UC623 cells are unresponsive to exocytic stimulus. Starved cells (6 h in DMC at 22°C) were either fixed directly (left hand panels) or exposed to the secretagogue Alcian Blue and fixed 15 min later (right hand panels). Vesicles containing Grl3p were visualized by indirect immunofluorescence with mAb 5E9. The fluorescent images are reconstructions made from optical slices that transect the top third of the cells. The corresponding bright field (Nomarski) images are also shown. Bar = 10 µm. (B) Distribution of immunoreactive Grt1p in docked vesicles. UC620 and UC623 cells were starved as in figure 8 prior to fixation and preparation for indirect immunofluorescence using mAb 4D11. Images of sagittal sections that included docked vesicles were obtained; the bright field phase contrast image indicates the position of the cell membrane. The wildtype example, showing polarized distribution of the Grt1p epitope at the granule tip, is an image of a growing cell, as starved wildtype cells accumulate so many docked DCGs that individual vesicles cannot be resolved using this technique. Bar = 2 µm.

A trivial explanation of the rescue seen in starvation would simply be that the starvation period allows more time for pre-existent vesicles to mature. To examine this, we asked whether phenotypic rescue following starvation could act upon the vesicles already present at the time of the shift, or instead was limited to vesicles synthesized after the shift. The latter appeared more likely, in light of the lack-of-complementation by heterologous cytosol, shown above in growing cells. We tested this using the same approach, by transferring Grl1p–GFP labeled cells to starvation conditions in the absence of cadmium. After 4 h, the cells were fixed, permeabilized, and immunostained with 5E9. Under these conditions, wildtype cells retained many of the docked granules after the transition to starvation conditions, and these could be seen with both the 5E9 antibody and GFP fluorescence (Figure 10). UC620 and 623, in contrast, displayed many docked vesicles that were labeled by mAb 5E9, but did not contain GFP signal. The results indicate that docked vesicles observed in starved mutant cells contain DCG cargo that is synthesized exclusively following the change in growth conditions, and that the undocked granules that accumulate before the switch to starvation cannot be subsequently rescued to docking-competence.

Figure 10.

The vesicles that dock in starved mutant cells are synthesized de novo. UC620, UC623, and wildtype were transformed with GRL1-GFP under the inducible MTT1 promoter. To label vesicles produced during exponential growth, Grl1p–GFP expression was induced by growing cells in the presence of cadmium for 12 h, and this was followed by washing into cadmium-free media for 1 h to chase newly-synthesized Grl1p–GFP out of the early secretory pathway. Small aliquots of the cells were fixed and GFP was localized by fluorescence microscopy (left panels). To follow the fate of these vesicles during starvation, the remaining cells were starved for 4 h in the absence of cadmium, and then fixed for GFP localization (center-left panels). Wildtype cells retained most of the fluorescent docked vesicles that had accumulated during growth. In contrast, starved UC620 and UC623 retained only a small fraction of their fluorescent vesicles, and these remained in the cytoplasm. No Grl1p–GFP was observed in docked vesicles in the starved mutant cells. To confirm that the mutants produced docked vesicles during starvation, cells were immunostained with mAb 5E9 to visualize Grl3p (center-right and right panels, focused on sagittal and tangential planes respectively). Bar = 10 µm.

Unregulated protein release in UC620 and 623

One potentially informative feature of UC620 and 623, particularly notable in starvation, was that the cells accumulate only a modest number of granule-like vesicles, compared with the number of DCGs in wildtype cells. This suggested either that fewer granules are synthesized in the mutants, or that they undergo a relatively high rate of unregulated secretion. Studies in mammalian systems have suggested that immature granules can exhibit an elevated rate of non-regulated secretion compared to mature DCGs, perhaps due to the presence in immature granule membranes of proteins that promote vesicle fusion in the constitutive secretory pathway (38). In Tetrahymena, it is difficult to quantify unregulated release by biochemical assay because the core proteins are variably insoluble. We therefore took the strategy of tracking individual GFP-labeled vesicles over time in live cells, both wildtype and mutant. We measured the number of docked vesicles that were lost from the cell surface over a 45 min time period (Figure 11), and found that wildtype cells lost 24 of 318 vesicles (8%), UC620 cells lost 16 of 100 vesicles (16%), and UC623 cells lost 13 of 164 vesicles (8%). These numbers are likely to overestimate the fraction of granules undergoing exocytosis, since DCG loss in this assay could also reflect de-docking and movement into the cytoplasm, a phenomenon which appears to occur in these cells (unpublished observations). The results suggest that the basal rate of secretion of those vesicles that are stably docked is not dramatically higher in the mutants than in wildtype. This does not rule out the possibility that some vesicles in the mutants may fuse rapidly upon initial docking at the plasma membrane.

Figure 11.

The granule cargo vesicles in starved UC620 and UC623 cells are persistently docked to the plasma membrane. Cells expressing Grl1p–GFP were maintained for 4 h at 22°C in DMC with 0.2 µg/ml CdCl₂. Cells were then immobilized with a rotocompressor, and the top surface of individual cells was followed by capturing an image of this part of the cell every 5 min during a 45 min time course. The panels in the left column contain the first images captured in representative time courses, and the last images from the same time courses are shown in the panels in the right column. Bar = 10 µm.

The second possibility, that the mutants synthesize fewer regulated secretory vesicles than the wildtype, would imply either that less protein is synthesized, or that some of the protein normally destined for DCGs meets another fate, such as rapid degradation in lysosomes or unregulated release. To examine the latter, starved cells were biosynthetically pulse-labeled with ³H-lysine, followed by immunoprecipitation of proGrl1p from both cells and the chase medium, and visualization by SDS-PAGE and autoradiography (Figure 12A). We note that one cannot detect mature Grl1p in such experiments, due to its insolubility. Under these conditions, wildtype cells do not release any proGrl1p (26). SB281, in marked contrast, rapidly secretes the entire cohort of newly-synthesized proGrl1p under starvation conditions. We found that both UC620 and UC623 released some proGrl1p into the medium under these conditions. Most interestingly, proGrl1p release from these mutants took place with a longer lag than seen for the corresponding release from SB281. In a separate experiment (Figure 12B), the labeled cells were pelleted and resuspended in fresh medium between time intervals during the chase, allowing us to directly compare the amount of proGrl1p secretion occurring at early versus later times. Both UC620 and UC623 continued to secrete proGrl1p up to 45 min after the pulse, whereas SB281 cells secrete the vast majority of proGrl1p within 15 min after labeling (Figure 12C). The simplest interpretation of this difference is that secretion from SB281 occurs from an early compartment in the exocytic secretory pathway, whereas the other mutants release at least a portion of proGrl1p from a later compartment. In particular, if secretion in SB281 occurs entirely from the TGN, some of the secretion from UC620 and 623 may occur via a post-TGN maturation compartment.

Figure 12.

Non-stimulated secretion of proGrl1p from UC620 and 623. Starved wildtype or mutant cells (UC620, 623, and SB281) were biosynthetically labeled for 15 min with ³H-lysine. (A) At the time points indicated, cells were separated from the chase medium, and proGrl1p was immunoprecipitated from each fraction. Samples were analyzed by SDS-PAGE (12.5% polyacrylamide) and autoradiography. Like SB281 but unlike wildtype cells, UC620 and UC623 release proGrl1p into the medium. However, both the extent and the kinetics of secretion differ in the mutants. (B) In order to confirm the observation that UC620 and UC623 release a larger fraction of proGrl1p at later time points compared to SB281, the chase period was broken into three separate 15 min intervals. At the end of each interval, an aliquot of cells was removed and processed for immunoprecipitation as in (A). The remaining cells were washed and resuspended in fresh buffer. (C) The results were quantified by normalizing the signal intensities of later time points with the first time point (this quantity was arbitrarily set at one) and plotted on a logarithmic scale. Left panel: proGrl1p in cell pellets. Right panel: proGrl1p recovered from media. Black squares = UC620, grey diamonds = UC623, black triangles = SB281.

Discussion

Few mutations, either naturally occurring (i.e., diseases) or laboratory-induced, are known to inhibit DCG biogenesis, except where the cargo proteins themselves fail to be synthesized (39, 40). The possible exception comes from two ciliates, T. thermophila and P. tetraurelia, which are the only organisms in which random mutagenic approaches have been combined with screening for mutants specific to regulated exocytosis via DCGs. In the latter, recent technical advances have made it possible to identify the genetic lesion in such mutants, though no genes affecting biogenesis per se has yet been reported (41). The study of such mutants is complicated in Paramecium by the unusual role that docked DCGs appear to play in stabilizing microtubules during nuclear division, leading to complex and progressive pleiotropic defects in any cells lacking docked DCGs (42). DCGs do not appear to play such a role in Tetrahymena, thereby making it possible to study stable cell lines with defects in DCG biogenesis. However, only one such line has previously been well studied, while a second (UC1) could be analyzed only superficially because it was sterile (24). Here, we report the isolation of two granule synthesis mutants in T. thermophila, called UC620 and UC623, which were generated by nitrosoguanidine mutagenesis, and which display essentially indistinguishable phenotypes. The mutations impair regulated exocytosis, but are unlikely to inhibit all forms of exocytosis, since the constitutive secretory pathway is essential for cell growth (43), and UC620 and UC623 grow equally as well as wildtype. Genetic analysis indicated that the mutations in UC620 and 623 are located in chromosomes I and III, respectively. Each segregated as a single recessive mutation, but our analysis cannot eliminate the possibility of contributions from tightly linked loci.

The most striking features of the mutant phenotypes are, first, the uncoupling of two steps involving lumenal cargo, namely sorting and processing, and, second, the tight linkage between activities in the DCG lumen and membrane. These will be considered in turn. In growing mutant cells, the sorting of lumenal cargo appears to be largely intact. At least three different DCG proteins are sorted to a compartment likely to be specific to the regulated secretory pathway, having hallmark characteristics of immature DCGs: the vesicles contain partially-crystallized electron-dense cores, and furthermore appear to exclude a constitutively-secreted marker, namely GFP linked to a signal sequence. The non-overlapping localization of DCG and constitutive secretory markers in these cells was in marked contrast to the case in SB281, a previously characterized mutant in which DCG cargo can be rapidly and constitutively secreted. The absence of ssGFP from the UC620/623 vesicles is consistent with the idea that soluble proteins are removed during maturation, but could also be accounted for if the initial sorting at the TGN was relatively selective in Tetrahymena, as has been observed in an insulinoma cell line (44).

Although the DCG proteins in UC620/623 appear to be properly sorted, their processing is greatly inhibited, as much of the protein in growing cells appears to be the full-length form. This accumulation of proGrl proteins supports earlier work showing that granule protein sorting at the TGN precedes processing (26). Additionally, the details of secretion in these cells indicate a connection between processing and retention during granule maturation. Like SB281, UC620/623 exhibit non-regulated secretion of proGrl1p, but this occurs with a longer lag and on a smaller scale than is observed in SB281. We infer that the difference reflects secretion from two different compartments. If Grl proproteins are being secreted directly from the TGN in SB281, the corresponding pool in UC620/623 may enter into immature granules before a fraction of it is released. This could occur by direct exocytic fusion of a subset of the immature granules or via clathrin coated vesicles that bud from the immature granule membrane. The latter has been described in mammalian cells and is termed “constitutive-like” (45, 46). Non-regulated secretion of proGrl1p does not appear to occur in wildtype Tetrahymena, however, where proproteins are rapidly processed and form a highly insoluble lattice. The chemical and physical conversion of the Grl proproteins may underlie their highly efficient retention during maturation in wildtype cells, whereas limited processing in UC620/623 permits partial loss of material at the corresponding stage. Similar accumulation and unstimulated secretion of DCG precursors have been noted in the pancreatic β-cells of mice lacking the proinsulin processing enzyme PC-1/3, though the precise route of secretion has not been determined (47). These results suggest a similarity between maturation in Tetrahymena and in mammalian cells that was not apparent from previous studies in ciliates, and support a relationship between macromolecular assembly and efficient retention during this stage of DCG biogenesis. Given the very limited processing in the growing mutant cells, it was surprising to find evidence of ordered granule cores, since previous data had suggested that processing generated assembly-competent polypeptides (26). Our results with UC620/623 may suggest that processing lies downstream of assembly, perhaps acting as a switch that locks assembled proteins into place, thereby making assembly an irreversible process. However, the images could also be explained by core assembly in a subset of vesicles in which a substantial amount of processing has occurred.

The similarity between the UC620/623 phenotypes and that of the PC1/3 null mouse, raises the question of whether the primary genetic lesions in UC620 and 623 might lie in genes encoding processing enzymes. We cannot answer this question directly because complementing such mutated genes has not yet been achieved in this organism. However, current evidence suggests that multiple co-regulated proteases with overlapping specificities may process proGrl proteins (48). In such a system, single mutations in protease genes would be unlikely to produce the global processing defects seen in UC620/623. In contrast, mutation of proteins that influence the lumenal milieu, i.e., an ion channel or pump in the maturing granule membrane, could affect multiple proteases. For that reason, mutations that indirectly inhibit the processing activities, via inactivation of membrane transporters, represent potential candidates to account for these phenotypes.

Such mutations could account for the noted tight correlation between processing and docking defects in UC620/623, by several mechanisms. First, the correlation may reflect a requirement for specific interactions between core proteins and membrane components. In Paramecium, DCG docking occurs via a distinct tip structure that extends at the end of each granule, whose assembly depends upon the organization of the underlying dense core. When the genes encoding Paramecium core proteins (orthologs of the GRLs in Tetrahymena) were silenced, tip assembly was perturbed and the resulting vesicles were defective in docking (49). Similarly, a number of randomly-generated Paramecium mutants with defects in proprotein processing also showed docking deficiencies (42), which appear similar to the UC620/623 phenotype. Although T. thermophila DCGs show no differentiated tip structure, the asymmetric distribution of the Grt1p epitope, reported in this paper, suggests for the first time that some polarized docking-related structure may be present in wildtype DCGs. It is therefore intriguing that immunoreactive Grt1p in UC620/623 still appears to be concentrated at the granule periphery, though it is no longer focused at one end. Nonetheless, Grt1p polarization is not completely correlated with a docking deficiency, as shown by microscopy of starved cells in which the aberrant granules have become competent for docking but not fusion. These results are consistent with the idea that docking and fusion have overlapping but not identical requirements, and that the latter are more stringent. Grt1p itself is unlikely to be directly involved in either, since it is not predicted to traverse the membrane. Instead, it could play a role in organizing, via interaction with proteins or lipids, membrane subdomains specialized for these functions. We propose that the GRT1 homologs in Paramecium, in which this gene family has undergone a large expansion, assemble to form the distinct granule tip in that organism. An open question is whether core proteins in metazoan DCGs play similar functions. Our working model to account for the UC620/623 phenotypes is that incomplete proprotein processing prevents full core assembly, and thus precludes proper organization of a docking/fusion zone in the granule membrane. The primary defect may be in establishing the lumenal conditions required for efficient processing during granule maturation.

Alternatively, the correlation between defects in processing and docking may not reflect a direct connection between the two, but instead the shared dependence on an upstream step. In particular, both processes depend upon the appropriate cohort of granule membrane proteins, as we have argued above, so mutations in proteins involved in membrane protein sorting could account for the UC620/623 phenotype. However, current models of DCG formation suggest that no such proteins may be required; instead, membrane proteins may be sorted by co-aggregation with condensing lumenal components (50, 51). Interestingly, the uncoupling of membrane and lumenal sorting has recently been demonstrated in the formation of secretory lysosomes, organelles which share some characteristics with DCGs including regulated exocytosis, but whose synthesis depends more closely on mechanisms related to those relevant for conventional lysosomes. In that case, genetic disruption of the murine AP-3 adaptor complex led to accumulation of secretory lysosomes that were defective in intracellular transport (52). The similarity between this mouse phenotype, and those in UC620/623, raises the question of whether ciliate DCGs might be more closely related to secretory lysosomes than has previously been suspected, although phylogenetic analysis of AP adaptors in T. thermophila indicates that this organism has AP-1, AP-2 and AP-4 complexes, but not AP-3 (N. Elde and A. Turkewitz, unpublished.) In addition, the failure of UC620/623 to be complemented via cell fusion argues against a defect in a cytosolic adaptor component per se. The relationship of protozoan DCGs to secretory lysosomes has also recently arisen for a class of secretory bodies, called rhoptries, in the Apicomplexan parasite Toxoplasma (53). Apicomplexans, like Ciliates, belong to the Alveolata, and represent an early and deep branch of eukaryotic phylogeny. It is widely recognized that the core mechanisms in cell biology are universally conserved, but in fact the data to support this view are largely derived from organisms that are relatively closely related. For example, animals and fungi are much more closely related than are a large number of “primitive” eukaryotes, including the Alveolates (54). Molecular analysis of the mechanisms involved in secretory granule formation in these organisms may therefore yield insights into the evolution of membrane traffic.

Materials and Methods

Cells and cell culture

T. thermophila strains used in this work are indicated in (Table 1). Unless stated otherwise, cells were grown at 30°C in SPP media (1% proteose peptone, 0.2% dextrose, 0.1% yeast extract, 0.009% ferric EDTA) and starved at 30°C in DMC, a one-tenth dilution of Dryl’s (1.7 mM sodium citrate, 1mM NaH2PO4, 1mM Na2HPO4, 1.5 mM CaCl2) supplemented with an additional 0.1 mM MgCl2 and 0.5 mM CaCl2).

Table 1.

Tetrahymena strains.

Straina	Drug resistanceb	Phenotypec	Source
B2086	mpr1-1/mpr1-1(mp-s)	(exo+)	Jacek Gaertigd
Cu428	mpr1-1/mpr1-1(mp-s)	(exo+)	Peter Brunse
Cu427	chx1-1/chx1-1(cy-s)	(exo+)	Peter Brunse
B*VI	--	(exo+)	S.L. Allenf
SB281	gal1-1/gal1-1(dg-r)	(exo−)	Ed Oriasg
CU743 (Nulli-I)	mpr1-1/mpr1-1(mp-s)	(exo+)	Peter Brunse
CU737 (Nulli-II)	pmr1-1/pmr1-1; chx1-1/chx1-1(pm-r, cy-s)	(exo+)	Peter Brunse
CU734 (Nulli-III)	pmr1-1/pmr1-1; chx1-1/chx1-1(pm-r, cy-s)	(exo+)	Peter Brunse
CU739 (Nulli-IV)	pmr1-1/pmr1-1; chx1-1/chx1-1(pm-r, cy-s)	(exo+)	Peter Brunse
CU741 (Nulli-V)	pmr1-1/pmr1-1; chx1-1/chx1-1(pm-r, cy-s)	(exo+)	Peter Brunse
UC613	MPR/mpr1-1; CHX/chx1-1 (mp-r, cy-s)	(exo−)	This laboratory
UC614	MPR/mpr1-1; CHX/chx1-1 (mp-r, cy-s)	(exo−)	This laboratory
UC620	mpr1-1/mpr1-1; chx1-1/chx1-1(mp-r, cy-r)	(exo−)	This laboratory
UC623	mpr1-1/mpr1-1; chx1-1/chx1-1(mp-r, cy-r)	(exo−)	This laboratory

^afor the nullisomic strains, the deleted germ line chromosome is indicated in parentheses.

^bmpr, mp = 6-methyl purine, chx, cy = cycloheximide, gal1, dg: 2-deoxygalactose, pmr, pm = paromomycin. Mutant alleles confer drug resistance, and the geneotypes are wildtype where not specified. –s and –r indicate sensitive and resistant phenotypes respectively.

^cPhenotype refers to the capacity for regulated secretion in response to Alcian Blue.

^dJacek Gaertig is at the University of Georgia.

^ePeter Bruns is at HHMI.

^fSally Allen is at the University of Michigan.

^gEd Orias is at the University of California, Santa Barbara.

Generation and screening of exocytosis mutants

This proceedure is described by (24). Cells were mutagenized with nitrosoguanidine, brought to homozygosity, and mass screening for exocytosis mutants was performed as described.

Genetic methods

Large-scale matings, as well as single-cell isolations, were performed as described in Hamilton and Orias (55). Selection with cycloheximide and 6-methyl purine was done at 15µg/ml, and paromomycin was used at 120µg/ml. Capsule formation tests, to detect exocytosis-deficient mutants, were done as described by Turkewitz et al. (56).

Genetic analysis of mutant strains

To assign complementation groups, mutant strains of different mating types were starved and mixed for mating overnight, and the culture tested the following day for the ability to form capsules when exposed to Alcian Blue. For genetic characterization, we outcrossed the 6-methyl purine-resistant mutant lines with CU427, a strain that bears a cycloheximide-resistance allele in the micronucleus but the sensitive allele in the macronucleus . Progeny were identified within 4 d by selection with both 6-methyl purine and paromomycin, and 48 clones were obtained by single cell isolation. These F1 lines were tested for Alcian blue-stimulated capsule formation. To ask whether the phenotypes were due to mutations in single genes, we analyzed the F1 progeny by uniparental cytogamy, an approach that is genetically equivalent to performing a self-cross (57). First, we identified a subset of F1 cells which had lost cycloheximide resistance via phenotypic assortment (58), and established clonal lines from single cells. One F1 clone from each line was then mated with B*VI (table 2), and uniparental cytogamy was induced during conjugation. The desired cytogamonts were selected using cycloheximide (55, 59). The genome of each cytogamont is the duplicated product of a single haploid meiotic pronucleus, generated by the non-B*VI partner in the mating. Therefore, for a single recessive mutation, 50% of the cytogamonts will manifest the defect.

We measured the fraction of exocytosis-deficient cytogamonts (table 2) as follows: cytogamonts were selected by distributing mating pairs (at 7 h after mixing) into 96-well plates and growing them in the presence of cycloheximide. Before plating, the concentration of mating pairs was adjusted so that drug selection would result in approximately 40 wells bearing viable cells per plate. As the efficiency of cytogamy differed in the UC613 and UC614 matings, the pair concentration was determined empirically for each cross by plating at several different pair concentrations. Within a week after mating, the cytogamonts were tested for encapsulation, and the fraction of exocytosis deficient wells was used to determine the frequency of the mutant phenotype in the cytogamonts. Since some wells in a plate will contain multiple cytogamont clones, any encapsulation-defective clone that shares a well with an encapsulation-competent clone will not be recognized. However, the actual frequency of mutant cytogamonts (given by the term “x” in the formulae below) can be calculated from the observed number of exocytosis-deficient wells (given by “a”), since the two figures are related in the following expression:

$$\mathrm{a}=[(\#\text{single clone wells})\mathrm{x}+(\#\text{double clone wells}){\mathrm{x}}^2+(\#\text{triple clone wells}){\mathrm{x}}^3+\dots ]/\text{total}\#\text{wells scored}$$

This equation was converted by expressing the numerator and denominator in terms of the Poisson formula and solving for x:

$$\mathrm{x}=[\text{ln}(\mathrm{a}-{\text{ae}}^{-\mathrm{\lambda }}+{\mathrm{e}}^{-\mathrm{\lambda }})+\mathrm{\lambda }]/\mathrm{\lambda }$$

where λ is the average number of clones per well, obtained by applying the Poisson formula to the fraction of empty wells following cycloheximide treatment (this number is equal to e^−λ).

Homozygous lines for each of the mutations were established from single cells isolated from encapsulation-defective wells. These included the UC619, UC620 and UC623 strains, and were used for further genetic and phenotypic analysis.

The mutations in UC620 and UC623 were mapped using complementation matings, described above, between mutants and five nullisomic strains. Each nullisomic strain lacks one of the five chromosomes in the germ line nucleus, while maintaining the full set in the somatic nucleus (32) (33). The progeny of such matings will be wildtype for exocytosis, except when the mutation resides on the chromosome that is absent in the nullisomic partner. Four of the five nullisomic strains bear a paromomycin-resistance allele in the germ line but not the somatic nucleus, allowing progeny to be selected with this drug. That gene is present on chromosome 1, and is therefore absent in the germline of the nulli-1 strain. To select for the progeny of crosses with this strain, conjugating cells were transformed with the vector pVGF.1, which encodes paromomycin resistance via a different mechanism (60). Since only cells that successfully complete conjugation can be transformed by such so-called processing vectors (61) , this approach allowed us to select progeny in these matings.

Production of polyclonal anti-Grlp antibodies

To obtain immunogen for the production of anti-Grl3p and anti- Grl8p antibodies, secretory granule contents were purified from wildtype cells (56). ~ 1 mg of this material was resolved on a 20% polyacrylamide gel in SDS, and then stained with Coomassie Blue. Guided by previous identification of polypeptides in such samples (20), several bands were excised with a scalpel. Rabbit antisera against these bands were then produced commercially (Zymed, S. San Francisco CA).

Cell lysates and western blotting

For whole cell lysates, 10⁶ cells were washed and resuspended 0.3ml DMC before adding an equal volume of 100°C 2X SDS-PAGE sample buffer (final concentration 100mM sucrose, 3%SDS, 2 mM Na₂EDTA, 62.5 mM Tris pH 6.9) and boiled for 3 min. For western blots, 12µl of such lysates was resolved by SDS-PAGE and transferred to 0.45 µM nitrocellulose. Immunoblots with polyclonal antisera were blocked, probed, and washed with 5% milk in TBS as described previously (22). The anti-Grl1p, anti-Grl3p, and anti-Grl8p antibodies were used at 1:2000, 1:800, and 1:3000 respectively. The monoclonal antibodies 5E9 (anti-Grl3p) and F56-BA1.2.3 (anti-GFP, Exalpha Biologicals, Maynard MA) were used at concentrations of 1:200 and 1:100 respectively, and were used in 3% bovine serum albumin in TBS. Primary antibodies were detected with a Horse Radish Peroxidase-conjugated secondary antibody (Jackson Immunoresearch, West Grove PA) and the Pierce (Rockford IL) SuperSignal detection kit, and images were acquired with a digital camera system (Alpha Innotech, San Leandro CA).

Purification of proGrlp complexes

Wildtype cells were transformed with the ncvB expression vector (see below) containing either GRL3HIS or GRL8HIS, and protein expression was induced by adding 2 µg/ml CdCl2 to 100ml of mid-log phase cells for 8 h. Prior to lysis, the cells were washed and resuspended in 2 ml of ice-cold DMC with protease inhibitors (5 µg/ml leupeptin, 10 µg/ml E-64, 10 µg/ml chymostatin, 12.5 µg/ml antipain). Cells were lysed for 30 min by adding 8ml chilled 1.25X guanidine lysis buffer (40mM Na₂HPO₄, 20mM NaH₂PO₄, 200mM NaCl, 1%Triton X-100, 10mM imidazole, 10% glycerol, 6 M guanidinium-HCl, pH 7.0). Lysates were centrifuged at 20k × g for 30 min in an SW41 rotor, and the supernatants incubated with TALON affinity resin (BD Biosciences, Palo Alto CA) for 2 h at 4° C. The resin was washed with guanidine lysis buffer followed by room temperature urea lysis buffer, which is identical to guanidine lysis buffer except that guanidine is replaced by 8M urea. The bound material was eluted with 150 mM imidazole in urea lysis buffer.

Immunofluorescence

Fixation and immunolabeling was as in Bowman et al. (22) except that it was performed at room temperature . In two-color labeling experiments, the natural GFP fluorescence was enhanced by including 0.5% (v/v) rabbit anti-GFP primary antibodies (Molecular Probes, Eugene OR), followed by 1% (v/v) fluorescein-conjugated anti-rabbit antibody (Jackson Immunoresearch, West Grove PA). This also guaranteed detection of the entire pool of GFP-labeled protein. Samples were viewed under a Zeiss (Thornwood NY) Axiovert microscope interfaced with a Zeiss LSM 510 confocal laser system and software. To measure secretion efficiency, ~10⁷ cells from exponentially-growing cultures (1–3 × 10⁵/ml) were washed and resuspended for 6 h in DMC. The cells were then stimulated with Alcian Blue, and fixed after 15 min for immunofluorescence analysis using a Zeiss Axioplan 2 microscope interfaced with a Zeiss Axiocam and Axiovision software(48).

Cloning of GRT1

Secreted Tetrahymena DCG proteins were purified as previously described (26) and visualized by SDS-PAGE and Coomassie blue staining. The single prominent band with an apparent mass near 80kD was excised and prepared for NH2-terminal sequencing by the Protein Sequence Analysis Lab, University of Kentucky. The amino acid sequence DVKLVTSEFSANKFD was deduced from the band and used for the design of degenerate primers in both orientations to amplify fragments of GRT1 from a cDNA library by PCR, followed by amplification of full-length GRT1 from cDNA and genomic DNA. The predicted translation of GRT1 contained the exact sequence determined by NH2-terminal sequence near the amino terminus.

Disruption of GRT1 through homologous recombination

DNA sequences flanking GRT1 were obtained as described (62) using PCR techniques. A knockout construct was generated with the NEO2 drug resistance cassette (63) inserted at a PstI site just downstream of the predicted translational initiation site, thereby creating the null allele grt1-1::neo2. Macronuclear knockouts were confirmed on the basis of the restriction maps, as detected by Southern blotting with a probe corresponding to the GRT1 gene.

Expression of GFP- and 6HIS-tagged proteins

A T. thermophila expression vector, called ncvB (Supplementary figure 1), was constructed within a pCR2.1 (Invitrogen, Carlsbad CA) plasmid backbone. It is designed to integrate via homologous recombination at the macronuclear MTT1 locus (35), genbank accession # AY061892) following linearization with SfiI and transformation via particle bombardment. The truncated, but functional, version of the MTT1 promoter in the middle of the cassette contains the final 930 nucleotides of the promoter region before the start codon. Cells transformed with this construct were selectively resistant to 60µg/ml blasticidin in the presence of 2µg/ml cadmium. For outgrowth following selection, drug resistant clones were passaged every other day for 10–12 days in successively lower concentrations of CdCl2, ending with 60µg/ml blasticidin plus 0.1 µg/ml CdCl2. A variety of GFP- or His-tagged chimeras (denoted ORF in supplementary figure 1) were expressed, in a cadmium-inducible fashion, by being placed downstream of the full length MTT1 promoter. The 3’ untranslated region was derived from the RPL29 gene (denoted RPL29 3’ UTR, ref), and the 3’ UTR for the BLAST message is supplied by the right flank of the construct, which contains the MTT1 3’ UTR.

To construct GRL1-GFP , the 3’ end of the GRL1 cDNA sequence was modified by PCR amplification, using a primer set (forward primer GTTTAAACATGAATAAGAAATTATTA, reverse primer TTGTCAGCTGAAGTTAATGAAGTCAATATTGG,IDT) that replaced the GRL1 stop codon with a Pvu II restriction site. Similarly, an eGFP (BD Biosciences, Palo Alto CA) clone was PCR-modified to add a Pme I site upstream of the start codon (forward primer AATAGTTTAAACAAAATGGTGAGCAAGGGCGAGGA, reverse primer TCACTCGAGGTTCTTGTACAGCTCGTCCATGC). After cutting with the appropriate restriction enzymes, these products were fused by blunt ligation, resulting in the insertion of a short in-frame linker with the translated sequence “FRNK” between the GRL1 and GFP open reading frames. The 6HIS tag was appended to the 3’ ends of GRL3 and GRL8 by PCR amplification of cDNA clones using primers with tails that had an insertion of the tag sequence (translated as GSHHHHHH) immediately upstream of the stop codon (For GRL3: forward primer GTTTAAACATGAAGAATTTAGCTATC, reverse primer AAGGGCCCTCAATGATGATGATGATGATGGGATCCAACATCAACAGTTAATTCATCAGCA For GRL8: forward primer CGGTTTAAACATGAACAAGGCCTTAGTTTTCT, reverse primer AAGGGCCCTCAATGATGATGATGATGATGGGATCCGAAACCGAGCTTCTTTTGAACGAAG). GFP linked to an N-terminal ER translocation signal sequence (ssGFP) was created using the signal sequence (“MNSKVIIALFCLVAVTLAASATPGANAIT” from an immobilization antigen-like protein identified in the T. thermophila EST library database (ID Tt_280). (This signal sequence was amplified from a cDNA library (25) forward primer TGTTTAAACAAAAAAATGAACTCCAAGGT, reverse primer AGCTCCTCGCCCTTGCTCACGGTAATTGCGTTGGCACCAG) and was fused to the N-terminus of GFP (amplified from pEGFP (Clontech), forward primer GTGAGCAAGGGCGAGGAGC, reverse primer TCACTCGAGGTTCTTGTACAGCTCGTCCATGC) using the overlap PCR method. All clones had the Pme I and Apa I or XhoI restriction sites on the 5’ and 3’ ends respectively, allowing for the sequence to be inserted into the ORF site (Supplementary figure 1) in the ncvB expression vector.

Cryo-fixation and immuno-electron microscopy

Cells were grown in SPP with 150mM mannitol, and 2µg/ml CdCl₂. After collecting cells by centrifugation, they were washed with SPP and mannitol, and a few µl of the cell pellet/slurry were transferred to an aluminum planchette (Type A) with a 100 µm deep well (Engineering office M. Wohlwend, Senwald, Switzerland), and sandwiched with the flat side of a Type B aluminum planchette, coated with hexadecene (64). Cells were cryofixed in a BAL-TEC HPM-010 high-pressure freezer, and then freeze-substituted in 0.25 % glutaraldehyde and 0.1% uranyl acetate in acetone. Embedding was a described by Giddings (64) except that isopropyl alcohol was used in place of methanol to maintain the sample at −45° C for polymerization of the HM20.

Embedded cells were serially sectioned (50–60 nm) and put on formvar-coated nickel grids. Some sections were immuno-labeled with a polyclonal anti-GFP antibody (65) diluted 1:200 in a blocking solution of 1% non-fat dry milk in PBST, and 15 nm colloidal gold-conjugated secondary antibodies (Ted Pella, Redding, CA). Samples were stained with 2% uranyl acetate in 70% methanol/30% water for 5 min and lead citrate for 4 min. The sections were viewed on a Phillips CM10 electron microscope operating at 80 kV. Images were captured with a Gatan digital camera and viewed with the Digital Micrograph Software package (Gatan, Pleasanton, CA).

Microscopy of living cells

Cell swimming was first slowed by the addition of methyl cellulose (Carolina Biological, Burlington NC) (0.5% v/v final concentration) and the cells were further physically restrained in a rotocompressor ((66), made by Warren Ringlien, Carleton College, Northfield MN). Cells were viewed under a Zeiss Axioplan microscope interfaced with a FxHQ CCD camera (Photometrics, Huntington Beach CA) and Openlab software (Improvision, Lexington MA).

Electrofusion

Cells were grown overnight to mid-log phase density in SPP in the presence or absence of cadmium. Cell fusion was as in Bowman et al. (22), following a 15 min incubation in fusion buffer.

Pulse-chase and immunoprecipitation

Labeling with ³H-lysine was performed as in Bowman and Turkewitz (2001)(22) except that the chase medium included 1 mg/ml bovine serum albumin. At indicated intervals, cells were washed by pelleting into an underlayered 5% Ficoll pad, then rapidly withdrawing the supra-pad layer and resuspending the cells in fresh medium. The immune complexes were purified with protein A–Sepharose beads (Amersham), after incubation with the primary antibody for 4 h.

Supplementary Material

Suppl fig 1

Inducible expression of stably integrated trans-genes in T. thermophila with the expression vector ncvB. Transformation results in the replacement of the chromosomal MTT1 gene with the ncvB expression cassette by homologous end recombination. The MTT1 promoter responds to the presence of cadmium in the cell culture media by sharply upregulating gene expression, allowing control over the expression of a trans-gene (ORF), in addition to a selectable marker conferring resistance to blasticidin (BLAST) in transformed cells.