Growth and form of secretory granules involves stepwise assembly but not differential sorting of a family of secretory proteins in Paramecium

The regulated secretory pathway allows some differentiated metazoan cells and a variety of unicellular eukaryotes to elaborate secretory granules for intracellular storage of biological molecules and their release by exocytosis in response to signals from the environment. Histamine release, insulin secretion, and exocytosis of apical granules during host cell invasion by malaria parasites are some examples. Despite extensive studies of secretory granule biogenesis, especially in mammalian hormone-producing, neuroendocrine and exocrine cells and tissues, the molecular mechanisms involved in intracellular transport, sorting, processing, and condensation of secretory proteins during granule biogenesis remain in large part undefined (for reviews of secretory granule biogenesis see Halban and Irminger, 1994; Thiele and Huttner, 1998; Tooze 1998; Arvan and Castle, 1998).

Paramecium provides a model system for analysis of secretory granule biogenesis. This ciliated protozoan elaborates ∼1000 secretory granules (trichocysts) per cell and stores them at specialized sites in the plasma membrane, ready for rapid exocytotic release (for review see Vayssié et al., 2000). Mature trichocysts are a few microns in size and have a highly constrained shape necessary for function, consisting of a voluminous spindle-shaped body surmounted by a tip that serves to anchor the granule to cortical exocytotic sites. Mutants with aberrantly shaped trichocysts are deficient in exocytosis because the affected vesicles cannot attach to the cortical exocytotic sites and be secreted (Pollack, 1974; Ruiz et al., 1976; Adoutte et al., 1984; Gautier et al., 1994). Upon exocytotic membrane fusion, the vesicle content, which is a metastable protein crystal, rapidly and irreversibly expands to an insoluble needle-shaped form, thus propelling itself out of the cell. This exocytotic response is probably involved in defense against predators (Harumoto and Miyake, 1991).

Although Paramecium trichocysts like many protozoan extrusomes (for review see Hausmann, 1978) are very elaborate, their properties are shared by secretory granules of higher eukaryotes. Granule contents are in all instances stored in a condensed, osmotically inert state and decondense upon release. The state of the condensed molecules can be an amorphous aggregate but can also be an ordered crystalline array as in the insulin granules of many species. Vesicle shape can be constrained (i.e. non-spherical) and is then a property of the protein contents of the granule. For example, the rod shape of the Weibel-Palade bodies of endothelial cells is conferred by the proteins stored within them, as demonstrated by transfection of von Willebrand factor propolypeptide into pituitary or insulinoma cell lines (Wagner et al., 1991).

Mutations affecting granule shape have been isolated in Paramecium (Pollack, 1974; Ruiz et al., 1976) and in another ciliate, Tetrahymena (Orias et al., 1983), which secretes cylindrically shaped granules with crystalline contents known as mucocysts. Although the corresponding genes have not yet been cloned, studies of the mutant cells indicate that perturbation in proteolytic maturation of the secretory proteins compromises formation of functional secretory granules, indicating a key role for protein processing in the biogenesis of these organelles (Adoutte et al., 1984; Turkewitz et al., 1991; Gautier et al., 1994; for review see Vayssié et al., 2000).

Biochemical characterization in both ciliate models has shown that, as in insulin-producing cells (Orci et al., 1987; for reviews see Halban and Irminger, 1994; Arvan and Castle, 1998), the major protein components of ciliate secretory granules are produced as soluble proproteins that follow the classical biosynthetic pathway and are delivered to immature secretory granules before any processing to mature polypeptides (Turkewitz et al., 1991; Garreau de Loubresse, 1993; Gautier et al., 1994). The products of the proprotein conversion, which occurs in post-Golgi immature and/or maturing granules (Hausmann, 1988; Shih and Nelson, 1992; Gautier et al., 1994; Garreau de Loubresse et al., 1994), finally assemble into the insoluble, metastable crystalline arrays characteristic of these secretory organelles. In Paramecium, the crystalline matrix of the granule is composed of antigenically distinct core and cortex layers (Hausmann et al., 1988), reminiscent of the organization of rat insulin secretory granules. The latter consist of a dense core of crystalline insulin surrounded by a non-crystalline layer containing C-peptide (Michael et al., 1987; Bendayan, 1989). This two-layered structure is the consequence of proinsulin processing given that insulin crystallizes and the crystal excludes more soluble components (proinsulin, C-peptide) which concentrate at the granule periphery.

Whether delivery of cargo proteins to immature granules is a passive, bulk flow process in ciliates or involves some active sorting and recruitment of membrane is not known. In addition to the major cargo proproteins that will condense into the crystalline matrix after processing, Paramecium trichocysts also contain a minor population of soluble, membrane-associated glycoproteins that link the insoluble cyrstalline matrix to the trichocyst membrane via a delicate mesh-like structure (Momayezi et al., 1993). These proteins, which have not yet been characterized at the molecular level, might be involved in membrane recruitment during trichocyst formation as recently proposed for proteoglycans during zymogen granule formation in rat pancreatic acinar cells (Schmidt et al., 2000).

Thus in Paramecium (Adoutte et al., 1984; Gautier et al., 1994; Gautier et al., 1996; Ruiz et al., 1998) as well as in Tetrahymena (Turkewitz et al., 1991; Chilcoat et al., 1996; Verbsky and Turkewitz, 1998), the ensemble of the biochemical and genetic data indicate that the cargo proteins themselves play an important role in the formation of the secretory granules. We have therefore focused our efforts on characterization of the major protein components of trichocysts, the trichocyst matrix proteins (TMPs), in order to investigate how these proteins are delivered to immature granules and then processed and assembled during granule maturation.

The TMPs are the products of a large family of ∼100 genes encoding 40-45 kDa proproteins which are processed to mature 15-20 kDa polypeptides; the latter are the polypeptides that crystallize to form the condensed trichocyst matrix (Gautier et al., 1994). TMP genes, which appear to be co-expressed and co-regulated at the transcriptional level (Galvani and Sperling, 2000), can be classified in subfamilies consisting of 4-8 very similar genes (>85% nucleotide identity) that specify nearly identical proteins (Madeddu et al., 1995). TMPs that belong to different subfamilies share only ∼25% amino acid identity. However, all TMPs are predicted to have the same protein fold (Gautier et al., 1996). Genetic evidence that TMP heterogeneity is of functional significance and does not simply reflect a need for large amounts of proteins that can co-assemble, was obtained by targeted gene silencing experiments. Cells in which expression of either of two TMP^* gene subfamilies was specifically reduced contained aberrantly shaped trichocysts which could not be secreted (Ruiz et al., 1998).

In order to further analyse the role of the TMP multigene family in the biogenesis of a functional secretory organelle, we have prepared sequence-specific antibodies against polypeptides specified by genes belonging to each of three different TMP gene subfamilies. The antibodies allowed us to determine the localization of the corresponding polypeptides in mature and developing wild-type trichocysts and in mutant trichocysts created by gene silencing. Our data show that each of the different TMPs is restricted to one or the other of the concentric core and cortex regions of the trichocyst matrix. Double-label experiments provide evidence that this does not result from differential sorting of the corresponding polypeptides, since they are present as a mixture even in small post-Golgi vesicles. We propose a model in which TMP assembly is controlled by protein processing, each layer of the crystalline edifice corresponding to one phase of trichocyst growth.

The wild-type strain used in this study is Paramecium tetraurelia d4-2, derived from stock 51 (Sonneborn, 1974). Cells were grown at 27°C in grass infusion (Wheat Grass Powder, Pines International, Lawrence, KS; USA), inoculated with Klebsiella pneumoniae the day before use, and supplemented with 0.4 μg/ml β-sitosterol (Sonneborn, 1970). The secretory mutant tam38 (Ruiz et al., 1976) was also used.

Synthetic genes were designed for optimal expression in E. coli of three different polypeptides, the first mature polypeptide of the gene TMP2c (GenBank U47116), the first mature polypeptide of the gene TMP4a (GenBank U47117), and the second mature polypeptide of the gene TMP1b (GenBank U47115), as diagrammed in Fig. 1A. The protocol for construction of synthetic genes is essentially as described (Dillon and Rosen, 1993). Overlapping oligonucleotides of 70-100 nucleotides were synthesized and gel purified. Each set of oligonucleotides (TMP1b, oligonucleotides 2-9; TMP2c, oligonucleotides 12-19; TMP4a, oligonucleotides 22-29) was annealed and extended with Pfu polymerase (Stratagene). The reactions of 100 μl contained 5 mM of each dNTP, 20 pmol of each oligonucleotide, 2.5 units Pfu in the buffer provided by the supplier; the PCR involved 5 minutes denaturation at 94°C, 10 cycles of denaturation for 1 minute at 94°C, hybridization 1 minute at 55°C and elongation 2 minutes 30 seconds at 68°C with an augmentation of 15 seconds at each cycle, followed by a final elongation of 6 minutes at 68°C. All temperature changes were ramped at 1°C/second. The products of the first reaction were then re-amplified (5 minutes denaturation at 94°C, 25 cycles of 1 minute denaturation at 94°C, 1 minute annealing at 55°C and 2 minutes 30 seconds elongation at 68°C plus an additional 10 seconds per cycle, final elongation 10 minutes 68°C) with external oligonucleotides of smaller size (TMP1b, oligonucleotides 1 and 10; TMP2c, oligonucleotides 11 and 20; TMP4a, oligonucleotides 21 and 30) containing at each extremity restriction sites for cloning in the expression vector. PCR products were gel purified, restriction digested, and cloned into the pET21-a expression vector (Novagen), which adds a C-terminal His6 tag to the polypeptide.

TMP1:

Oligo1: 5′cgccatatgTCGCTGACCAGGGTGCTCTGCG3′;

Oligo2: 5′cgccatatgTCGCTGACCAGGGTGCTCTGCGTGAAATC-CTGACCGCTTTCAACAACCTGCGTGTTCAGCTGGTTGACTC-C3′;

Oligo3: 5′GCTTCGAAGTCTTTCTGAGCTTCAGCTTCGTCAGC-GGTCAGCTGGTTCAGGGAGTCAACCAGCTGAAC3′;

Oligo4: 5′CAGAAAGACTTCGAAGCTCGTGTTATCCAGCTGA-ACCAGGAACACGCTGAATTCCGCGTGCTGTTGTTG3′;

Oligo5: 5′CCAGGGTCTGTTCGATTTTGTTAGCGTTAGCTTCG-ATTTCAGCGGTTTTAACAACAACAGCACGCTGG3′;

Oligo6: 5′CGAACAGACCCTGGACCTGATCGACGTTCTGCAC-GCTGACCTGGACACCCTGAACGGTCAGCTGC3′;

Oligo7: 5′CGGTAGCGTTGTAAACGTCGGTAGCGAAAGCGTA-GTCGTCGTTTTCAGCCTGCAGCTGACCGTTCAGG3′;

Oligo8: 5′GTTTACAACGCTACCGTTTCCGAATACAACAAAGA-ACTGAACGCTGCTCACCAGGCTCTGGACCTGCTGAACC3′;

Oligo9: 5′cgcctcgagTCAGAAAGCACCTTTCAGCTGGGATTTAA-CGTAGTCGGTGAAACGCGGCTGGTTCAGCAGGTCCAGAGC3′;

Oligo10: 5′cgcctcgagTCAGAAAGCACCTTTCAGCTGGG3′.

TMP2:

Oligo11: 5′cgccatatgTCCTCCACCC3′;

Oligo12: 5′cgccatatgTCCTCCACCCAGGCTGACGTTATCGCTAC-CATCAAAAAAATCGACCAGTCTCCGTTCGGTC3′;

Oligo13: 5′GTCAACGGTGCAAGCGTCCTGGTATTCGTGGTTA-CGAGCGTCGTCTTCTTTCTGTTCAGCAACGTAACGGTC3′;

Oligo14: 5′GGACGCTTGCACCGTTGACATCAAAGCTTTCGAC-AAAGACCTGGCTGAATCCAACCGTAAAAAAATCGAAC3′;

Oligo15: 5′CCTGCAGGATACCACGCTGTGGATACAGCTGACC-TTCCAGACGAGCTTCCAGTTCGATTTTTTTACGGTTG3′;

Oligo16: 5′CAGCGTGGTATCCTGCAGGGTCTGGTTGCTCAGA-AACAGGCTGAAGTTAAAGGTTACCAGAAAGACCTGGACG3′;

Oligo17: 5′CAGAACTTTTTCTTCGAAGTCAGCTTTTTCTTCAG-CACGCTGAGCGTCCAGTTCGTCCAGGTCTTTCTGGTAA3′;

Oligo18: 5′GACTTCGAAGAAAAAGTTCTGGAACACCAGGAA-GCTACTGCTATCATCGCTGAAGCTCGTCGTCTG3′;

Oligo19: 5′cgcctcgagCTGGATGAAGGATTCGTGTTCGATGTTG-TCAGCGAACAGACGACGAGCTTCAGC3′;

Oligo20: 5′cgcctcgagCTGGATGAAGG3′.

TMP4:

Oligo21: 5′cgccatatgGGTCCAGTTGGTGAAATCC3′;

Oligo22: 5′cgccatatgGGTCCAGTTGGTGAAATCCAGATCCTGC-TGAACAACATTGCTTCCCAGCTGAACGGTGACCAG3′;

Oligo23: 5′CACTGTTGCTTTCGAAAAAATCATTGCTGACCTG-GAACAGGAAATTGCTTACCACCAGACCCAGATCG3′;

Oligo24: 5′GAAGCTCTGGGTGAAGCTGAAGTTGAAGTTCGT-GTTGTTACCTCCGACATTGCTAACAACGAAAAATCC3′;

Oligo25: 5′CAGCACGACACCTGGGTTCGTAAAGATGCTGAAC-ACGTTGACCAGATGGAAGCTATCGACGAAGCTTCC3′;

Oligo26: 5′TTTTTTCGAAAGCAACAGTGTCAGATTCGTCCAG-TTTGTCAGCTTTTTTCTGGTCACCGTTCAGCTGGG3′;

Oligo27: 5′CAGCTTCACCCAGAGCTTCAGTGGTGGAGTCACG-CAGGTTGGACAGAGCAACGATCTGGGTCTGGTGG3′;

Oligo28: 5′CATCTTTACGAACCCAGGTGTCGTGCTGGGACTG-ACGAGTAGCAGATTCGTCAGCAAAGGATTTTTCGTTGTTAGC-

3′;

Oligo29: 5′cgcctcgagCTGAGCAAAAGCAACACCAGCCTGCAG-GTGCTGAACGATTTTGGAAGCTTCGTCGATAGC3′;

Oligo30: 5′cgcctcgagCTGAGCAAAAGCAACACC3′.

For bacterial expression of the different synthetic genes, BL21-DE3 cells were transformed with the appropriate pET21 constructs. Colonies were inoculated with 2% glucose and 100 to 150 μg/ml ampicillin. After 3 hours of non-induced culture, cells were harvested by centrifugation, and resuspended in fresh medium containing 100 to 150 μg/ml ampicillin and 0.4 μg/ml IPTG for 3 hours of induced culture. Cells were harvested and the pellet frozen at −80°C. The pellet was thawed and supplemented with 8 M urea, 0.1 M NaH₂PO₄, 0.01 M Tris-HCl, pH 8, and mixed for one hour using a wheel. The preparation was centrifuged and the clear lysate was loaded directly onto a Ni-NTA column (Qiagen), washed with the same buffer at pH 8 and then pH 6.5 and the tagged protein eluted at pH 4.5.

The proteins were then dialysed in PBS pH 6.45, and 500 μg of protein injected into rabbits in Complete Freund’s adjuvant. Injections were repeated using incomplete Freund’s adjuvant at three week intervals until high affinity antisera were obtained as judged by immunoblots against the different affinity purified polypeptides, purified trichocysts and whole cell extracts (4 or 5 injections).

For morphological observations, whole cell pellets were fixed in 2% glutaraldehyde in 0.05 M cacodylate buffer pH 7.2 for 90 minutes at 4°C. After washing in the same buffer, the samples were post-fixed in 1% osmium tetroxide in 0.05 M cacodylate buffer, for 60 minutes at 4°C. Post-fixed cells were dehydrated by passage through a series of ethanol and propylene oxide baths before embedding in Epon. Thin sections were contrasted with ethanolic uranyl acetate and lead citrate, then examined using a Philips EM410.

For post-embedding immunolocalization, cell pellets were fixed in 2% paraformaldehyde, 0.15% glutaraldehyde in 0.05 M cacodylate buffer, pH 7.4, at room temperature for 2 hours. After washing in the same buffer, cells were dehydrated by passage through a series of ethanol baths before embedding in LR White (London Resin Ltd). Thin sections were collected on nickel grids and saturated and processed with 3% BSA in PBS. This fixation procedure is optimized for antibody recognition at the expense of ultrastructure preservation. In particular, the periodic aspect of the crystalline regions of the trichocyst matrices is poorly preserved. In most of the images presented here, the crystalline regions appear more electron dense than surrounding amorphous regions.

For single-label experiments, primary polyclonal antibodies were diluted 1/250 (anti-TMP1 and anti-TMP2 antisera) or 1/400 (anti-TMP4 antiserum). Used at these dilutions, background level and density of specific staining were roughly equivalent for the anti-TMP1 and anti-TMP4 sera. However both rabbits immunized with TMP2 polypeptide gave specific sera with somewhat lower affinity as judged by the weaker staining density in immunolocalization experiments; the dilution of 1/250 was used for this serum in order to obtain the same low background as for the other sera. After washing, the sections were incubated with a 1/100 dilution of 10 nm colloidal gold-conjugated anti-rabbit immunoglobulins (GAR G10, Amersham-Pharmacia). After extensive washing, the sections were contrasted with ethanolic uranyl acetate.

Double-label experiments were carried out as described (Slot et al., 1991) using Protein A – gold purchased from Dr J. W. Slot, Utrecht University. Briefly, after saturation in 3% BSA – 0.1% fish skin gelatin in PBS, the sections were treated with anti-TMP4 primary antibody at a dilution of 1/600, then washed and incubated with Protein A coupled to 10 nm gold particles at the dilution recommended by the supplier (1/70). The sections were then treated with 1% glutaraldehyde in the saturation buffer for 5 minutes and quenched in the presence of 50 mM glycine in PBS. After washing, the same procedure was repeated with or without the second primary antibody (anti-TMP1 at a dilution of 1/350) and the sections were incubated with Protein A coupled to 15 nm gold particles (1/65 dilution). Finally, the sections were treated with glutaraldehyde as above, washed in water and contrasted with ethanolic uranyl acetate.

2D gel electrophoresis and immunoblots of partially purified extruded trichocysts were performed as previously described (Gautier et al., 1996) with anti-TMP1 and anti-TMP2 sera at 1/400 dilution and the anti-TMP4 serum at 1/1000 dilution. Serum 031 (Gautier et al., 1994) was raised against and recognizes almost all of the TMPs and was used at a dilution of 1/2000.

Wild-type cells were microinjected with the coding region of either the TMP1b or the TMP4a gene cloned in pUC18 and linearized prior to injection, as previously described (Ruiz et al., 1998). Phenotypes of the clonal descendants of microinjected cells were scored by examining the exocytotic capacity of the cells and by observation of trichocyst morphology using immunofluorescence techniques as previously described (Ruiz et al., 1998). The exocytosis-deficient phenotypes of the silenced cells are reproducibly correlated with the reduction of the specific mRNA, as measured by Northern blots (Ruiz et al., 1998). Cell pellets of appropriately expanded cultures of the chosen clones were fixed and processed for electron microscope morphological observation and for post-embedding immunolocalization experiments as described above.

TMPs are highly immunogenic (Fok et al., 1988), and immunization of rabbits with isolated extruded trichocysts or with gel-purified TMPs raises antibodies that recognize essentially the entire mixture of up to 100 distinct polypeptides that compose the crystalline matrix (Adoutte et al., 1984; Gautier et al., 1994). Monoclonal antibodies are more specific, but still recognize TMP subsets containing many polypeptides (Fok et al., 1988; Shih and Nelson, 1991). In order to obtain more specific reagents, we chose to direct bacterial expression of a single mature polypeptide from each of three previously characterized TMP subfamilies. Since Paramecium uses two of the stop codons recognized by bacteria (TAA and TAG) to specify glutamine, we constructed synthetic genes (see Materials and Methods) corresponding to the first mature polypeptides of TMP2c and of TMP4a and the second mature polypeptide of TMP1b (Fig. 1A). These are the polypeptides for which we originally obtained the N-terminal sequences used to clone the genes, by microsequencing 2D gel spots (Madeddu et al., 1994; Madeddu et al., 1995). We thus know with certainty both the N-terminus of these polypeptides and their position on a 2D gel.

As shown in Fig. 1B, each of the antisera recognizes 2 to 4 of the over 30 major spots resolved on a 2D gel of extruded trichocyst matrices. These 2D trichocyst gel profiles are highly reproducible, and in each case, the serum recognizes the spot used to obtain the original microsequence that served to clone the gene. In the case of the anti-TMP4a serum, which recognizes 2 adjacent spots, the microsequences of both spots were previously determined and are very similar, identifying two polypeptides encoded by members of the TMP4 gene subfamily. The antisera were thus judged to be specific for the sequences they were designed to recognize, and to constitute subfamily-specific, if not polypeptide-specific, reagents.

Fig. 2 shows typical images of mature trichocysts labelled by the different TMP antisera. Each of the antisera recognizes every mature trichocyst in the cell. However, the antisera recognize distinct regions of the trichocyst body matrix. The anti-TMP2 and the anti-TMP4 sera recognize the same central or ‘core’ region of the trichocyst in both longitudinal and cross-sections, while the anti-TMP1 serum decorates the outer or ‘cortex’ layer, not recognized by the first two sera. We found, using an HA (influenza haemagglutinin) epitope tag, that the second mature polypeptide encoded by the TMP4a gene also localizes to the core region (L. Vayssié and L. Sperling, unpublished observations).

As a parallel control, we used antiserum 031 (cf. Materials and Methods) that recognizes the whole set of TMPs and found uniform staining of both the core and cortex regions of the trichocyst body (not shown; see Fig. 2 in Garreau de Loubresse et al., 1994). We note that the elaborate trichocyst tip (cf. Fig. 2), which serves to anchor the vesicle at exocytotic sites in the plasma membrane, is not recognized by any of the available antibodies (Fig. 2; Garreau de Loubresse, 1993; Garreau de Loubresse et al., 1994).

Previous immunolocalization studies (Garreau de Loubresse, 1993; Garreau de Loubresse et al., 1994) showed that the TMPs follow the classical secretory pathway. The earliest post-Golgi vesicles have amorphous contents corresponding to TMP proproteins, which are soluble molecules (Garreau de Loubresse et al., 1994; Gautier et al., 1994). These vesicles increase in volume by homotypic fusion events, which have been documented by cryomicroscopy (Hausmann et al., 1988) and by quick freezing-deep etching (Allen and Fok, 2000). Once they are about 1 μm in size, crystallization begins. The vesicles continue to increase in volume while the body matrix crystallizes and attains an elongated shape. The tip assembles last, as the body matrix finishes its crystallization.

In Fig. 3, the localization of the TMP1, TMP2 and TMP4 polypeptides are shown through different post-Golgi steps of trichocyst biogenesis. The smallest vesicles with amorphous contents are decorated uniformly with each of the three sequence-specific antisera. As soon as crystallization begins, the TMP2 and TMP4 antigens are concentrated in the crystal and the excluded TMP1 antigens in the surrounding amorphous material. This pattern is maintained until the trichocysts attain their final size, a stage at which electron dense material becomes visible at one pole of the vesicle, where the tip will assemble (tip junction, Fig. 3C,F and I). At this point, the core assembly is complete, and the outer TMP1 containing layer crystallizes. These images indicate that the two distinct layers of TMPs found in mature trichocysts correspond to two temporally distinct phases of growth.

This organization in two concentric layers of TMPs appears to be quite robust. A number of secretory mutants have been characterized in which trichocyst biogenesis is affected and the resulting aberrantly shaped organelles are unable to attach to the cortical exocytotic sites or be secreted. Fig. 4 shows immunolocalization images for one such mutant, tam38, using the specific antisera. The organization of the TMPs in two concentric layers is conserved, however the inner layer fails to elongate before the other layer assembles (cf. wild type, Fig. 3C,F and I), accounting for the ovoid form of these abortive trichocysts. Amorphous material visible at the periphery of some vesicles is not recognized by any of our antibodies (Fig. 4A and B and data not shown) and could correspond to tip material which is unable to assemble (Fig. 4A and B, arrows).

In Paramecium, the phenomenon of homology-dependent gene silencing can be used to specifically reduce the expression of the cellular homologues of a given gene (Ruiz et al., 1998; for review see Bastin et al., 2001). Since the phenomenon is homology dependent, similar genes can be silenced together, as demonstrated for TMP gene subfamilies by quantification of mRNA (Ruiz et al., 1998). Silencing is achieved by microinjection at high copy number of the coding region of the gene of interest into the somatic macronucleus. The silencing effect is observed in the clonal descendants of the microinjected cells throughout vegetative growth.

We previously reported that microinjection of the coding region of either TMP1b or TMP4a led to specific reduction in TMP1 or TMP4 subfamily mRNA and that the silenced cells contained aberrantly shaped trichocysts that could not be secreted. Immunofluorescence analysis of the silenced cells showed that the form of the mutant trichocysts is different depending upon which TMP subfamily was silenced, suggesting that TMP1 and TMP4 polypeptides have different roles in trichocyst biogenesis (Ruiz et al., 1998).

We have now further characterized the mutant trichocysts in TMP1 and TMP4 silenced cells at the ultrastructural level, using the sequence-specific antibodies. Fig. 5 shows morphological and immunogold images of trichocysts in TMP4 and TMP1 silenced cells. The trichocysts in TMP4 silenced cells are never elongated. Most (Fig. 5A) have round or slightly ovoid cross-sections with a visible but highly abortive tip. In a few cells (Fig. 5B) the trichocysts have irregular contours suggesting multiple foci of crystallization and no visible tip. Immunolocalization with the anti-TMP4 antibody (Fig. 5C) shows that, just as TMP4 mRNA is reduced by ∼75% but not completely abolished (Ruiz et al., 1998), some TMP4 antigen is present in silenced cells. The core assembly does begin as in wild type, but does not progress to the point of elongation, probably because TMP4 polypeptides are in short supply. The TMP1-containing cortex assembles around the misshapen core (Fig. 5D) and most of the trichocysts contain tip material but the tip cannot assemble. These phenotypes are completely consistent with localization of TMP4 in the core region and demonstrate that if the core cannot assemble correctly, cortex and tip are also affected.

Analysis of trichocysts in TMP1 silenced cells completes the picture (Fig. 5E-G). Within each silenced cell, all of the trichocysts are elongated, however, they appear thinner, more irregular and less rigid than wild-type trichocysts. A few trichocysts have a functional tip assembly and are attached to the cortex but most have an abortive tip (Fig. 5E) and remain free in the cytoplasm. These images are consistent with assembly of an elongated core as in wild type, as revealed by the anti-TMP4 antibody (Fig. 5F), but an incomplete cortex layer (Fig. 5F and G), presumably owing to reduced amounts of TMP1 antigen in the cell. In the absence of a complete cortex layer, the tip material present in the vesicles does not usually form a functional tip. The anti-TMP2 antiserum (not shown) decorates the same core region of the mutants as the anti-TMP4 antiserum. Gene silencing thus primarily affects the layer of the trichocyst matrix that contains the products of the silenced genes. These results underscore the fact that the different TMPs have different morphogenetic roles in trichocyst assembly.

The Paramecium Golgi apparatus is fragmented into dictyosomes (Estève, 1972; for review see Allen and Fok, 2000) and the trans compartment buds coated vesicles which may correspond to the ∼75 nm clathrin coated vesicles containing acid phosphatase identified previously (Fok et al., 1984) as primary lysosomes. Other small irregular vesicles, apparently uncoated, are often found in close vicinity to the trans compartment. These vesicles contain amorphous dense material and represent the first step of TMP condensation (Garreau de Loubresse, 1993). These structures are shown in Fig. 7a by conventional Epon embedding. The clathrin coated vesicle budding from the Golgi indicates the polarity of the stack and two uncoated secretory vesicles, recognizable by their relatively dense, amorphous contents, lie in the vicinity of the trans compartment.

Since mature trichocysts several microns in size develop from such small TMP-containing vesicles, it is possible that ‘core’ TMPs are sorted away from ‘cortex’ TMPs as the proteins exit the Golgi complex, and that this sorting followed by specific vesicle fusion controls the order of TMP assembly. To look for differential sorting, we first examined all identifiable small secretory vesicles in immunolocalization experiments with each of the three sequence-specific antibodies. If core and cortex TMPs are differentially sorted, we would expect to see some unlabeled vesicles. Examination of thin sections of hundreds of different cells for each antibody indicated that all morphologically recognizable vesicles were decorated with gold particles (data not shown), strongly arguing that each antigen is found in every small condensing vesicle.

In order to obtain more direct evidence that the small post-Golgi vesicles contain mixtures of core and cortex TMPs, we performed double-label experiments (see Materials and Methods) using the anti-TMP1 and anti-TMP4 antisera. Thin sections were first incubated with anti-TMP4 serum and then Protein A-10 nm gold. After destruction of the epitope recognized by the Protein A using glutaraldehyde, the sections were incubated with anti-TMP1 serum and then Protein A-15 nm gold. As a control for the method, we evaluated images of mature and developing wild-type trichocysts obtained either including or omitting the 2nd antibody directed against TMP1. Typical control images are shown in Fig. 6: each antibody decorates the same region of developing (Fig. 6A) and mature (not shown) trichocysts as in single label experiments. Furthermore, in the absence of the 2nd antibody, there is no indication of ‘clustering’ of the Protein A-15 nm gold near the 1st antibody-gold complexes (Fig. 6B).

Double-labelling in the region of Golgi complexes is shown in Fig. 7. Paramecium Golgi is not preserved by the LR White post-embedding protocol used for immunocytochemistry, in fact these delicate membranes are poorly preserved even in cryosections as discussed (Hausmann et al., 1988). Nonetheless, the presence in some images of a still recognizable Golgi complex (Fig. 7B and C), or of coated vesicles (Fig. 7C,E,H,I,J), indicates that the small vesicles labelled by the anti-TMP antibodies are close to Golgi. All of the small vesicles that we observed contained both sizes of gold label, attesting to the presence of both antigens. The systematic co-localization of core and cortex TMPs, even in small post-Golgi vesicles, provides good indication that TMP proproteins are not differentially sorted but travel as a mixture through the compartments of the biosynthetic pathway to immature secretory granules where the proprotein conversion will occur.

We have designed synthetic genes that direct bacterial expression of three TMP mature polypeptides with different primary structures. The polypeptides were used to obtain sequence-specific rabbit antisera. Immunolocalization experiments with the antisera have allowed us to evaluate the way in which the protein heterogeneity generated by the TMP multigene family is used to build the crystalline matrix of Paramecium secretory granules.

Each antiserum decorates one or the other of two distinct, non-overlapping concentric regions of the spindle-shaped crystalline matrix: the central ‘core’ or the peripheral ‘cortex’. Examination of trichocysts at different stages of biogenesis showed that crystallization of the core, which is first spherical and then elongates, occurs while the vesicles grow to their final size. The end of this phase of growth is marked by appearance of the electron dense tip junction underneath the vesicle membrane, anchored to the wide end of the growing crystal. Crystallization of the cortex then occurs and the trichocyst tip assembles last. These results were independently confirmed with genetic experiments. Examination of mutant trichocysts created by gene silencing, which reduces the expression only of the targeted TMP gene subfamily, showed that the layer of the trichocyst matrix containing the products of the silenced genes was specifically affected.

The organization of the trichocyst matrix in antigenically distinct core and cortex regions was previously reported by Fok and colleagues (Fok et al., 1988; Hausmann et al., 1988) and by Shih and Nelson (Shih and Nelson, 1991) based on immunolocalization experiments with monoclonal antibodies that recognize subsets of TMPs. These studies, carried out before any knowledge of the primary structure of TMPs or of the organization and expression of the TMP multigene family was available, are consistent with the data presented here.

How is assembly of the TMPs in two distinct layers controlled? Since the inner layer assembles while the vesicles are increasing in volume, one possibility is that specific sorting of TMPs delivers them to the growing trichocyst in an ordered fashion.

It is now generally accepted that sorting of newly synthesized lumenal and membrane proteins can occur both in the TGN and in immature secretory granules. The aggregative properties of granule proteins constitute most or all of the information for their sorting to granules (Chanat and Huttner, 1991), be it ‘sorting for entry’ upon exit from the TGN or ‘sorting by retention’ in the maturing granules (Kuliawat and Arvan, 1994; reviewed by Arvan and Castle, 1998). We have previously argued (Garreau de Loubresse et al., 1994) that TMP precursors, which are soluble molecules, are delivered to post-Golgi vesicles with the bulk flow and only remain in the regulated compartment if processed to mature polypeptides, consistent with ‘sorting by retention’. Our evidence was based on genetic as well as morphological data. In particular, cells of the mutant trichless contain no trichocysts at all, but do synthesize TMP precursor molecules, which localize to small post-Golgi vesicles with amorphous contents. The precursors cannot be processed to mature polypeptides owing to the mutation, and they are constitutively secreted into the culture medium (Gautier et al., 1994).

We have presented here two lines of evidence that TMP-containing vesicles enclose both TMP1 and TMP4 polypeptides. First, examination of hundreds of cells in single-label experiments failed to reveal any unlabeled vesicles. Second, double-label experiments directly show colocalization in the same small vesicles of TMP4 (core) and TMP1 (cortex) polypeptides. Our data are thus consistent with bulk flow of soluble lumenal proteins, including all the TMPs, into post-Golgi vesicles upon exit from the Golgi apparatus. We conclude that sequential delivery of core then cortex TMPs to developing trichocysts is extremely unlikely and cannot account for the ordered assembly of the TMPs.

It remains possible that the antigenically distinct and as yet uncharacterized trichocyst tip polypeptides are sorted into distinct vesicles, as in the ciliate Pseudomicrothorax dubious. This ciliate produces trichocysts consisting of a crystalline ‘shaft’ equivalent to the Paramecium body matrix surmounted by four ‘arms’ which have the same anchorage function as the Paramecium trichocyst tip. Peck et al. have shown that the arm material, which is electron dense, and the shaft material for which they obtained specific antibodies, are sorted into distinct post-Golgi vesicles as they exit the TGN (Peck et al., 1993). These vesicles fuse with the developing trichocysts. The electron dense arm material remains at the periphery of the growing trichocysts and does not assemble until the shaft crystallization is complete.

We propose that in Paramecium, TMP processing controls trichocyst matrix assembly. Our model involves two hypotheses, first of all that TMPs crystallize as soon as they are processed, and second that TMPs are in competition for the same processing enzyme(s). The core TMPs compete more successfully and are thus converted at a greater rate than the cortex TMPs.

That TMPs crystallize as soon as they are processed implies that the amorphous regions of developing trichocysts consist of TMP precursors while the crystalline regions consist of mature polypeptides. Shih and Nelson provided direct experimental evidence that this is indeed the case (Shih and Nelson, 1992). Antibodies specific for TMP precursor molecules were prepared. In immunolocalization experiments, these antibodies recognized only the amorphous regions of developing trichocysts, allowing the authors to conclude that proteolytic processing occurs in parallel with crystallization.

The second hypothesis is that core TMPs are better substrates for the processing enzyme(s) than cortex TMPs. This implies that all TMPs are converted by the same enzyme(s) and that core TMP precursors (such as TMP2 and TMP4) are converted more rapidly than cortex TMP precursors (such as TMP1). Support for this hypothesis is provided by examination of the TMP sequences (Fig. 8). Sequence analysis previously indicated that Paramecium TMPs and Tetrahymena Grlps are homologous proteins, and that the as yet uncharacterized processing enzymes may be the same in both ciliates given the good conservation of the cleavage sites (Verbsky and Turkewitz, 1998). It may therefore be significant that the deduced amino acid sequences at the putative cleavage site N-terminal to the second mature polypeptide is nearly identical for the ‘core’ proteins TMP2c and TMP4a and closely resembles cleavage sites of Tetrahymena Grl1p and Grl4p (Fig. 8A). The cleavage site of the ‘cortex’ protein TMP1b is different from these but nearly identical to the cleavage site of Tetrahymena Grl3p (Fig. 8B) as previously noted (Verbsky and Turkewitz, 1998). These sites do not fit the consensus for the kexin/prohormone convertase family of processing enzymes conserved from yeast to mammals (Seidah et al., 1999). Some additional support for the substrate competition hypothesis comes from 2D analysis of TMP pulse-chase experiments. Despite limited temporal resolution owing to the fact that the experiments involved feeding paramecia with metabolically labelled bacteria and chasing with cold bacteria, different precursors are processed at different chase times (Gautier et al., 1996).

We therefore propose that all TMPs are delivered together to post-Golgi vesicles which incorporate into developing trichocysts by homotypic fusion events. Throughout growth to the final volume, the developing trichocysts are thus continually supplied with a stoichiometric mixture of precursors. As long as core TMP precursors are present, they will successfully compete for the processing machinery and be preferentially converted to mature polypeptides, which assemble as soon as they are produced. The core TMPs might have higher affinity constants than the cortex TMPs for an endoprotease with a low K_m, like the prohormone convertases of multicellular organisms (for reviews see Steiner, 1998; Zhou et al., 1999). The phase during which the trichocyst grows in volume thus corresponds to the period during which the core crystallizes.

The appearance of the tip junction (cf. Fig. 2A, Fig. 3C,F,I) marks the transition to a second growth phase during which TMP-bearing vesicles no longer fuse with the developing trichocyst which has attained its final volume. As no more TMPs enter the vesicles, the processing machinery runs out of its preferred substrate, the core TMPs, whose conversion and parallel crystallization go to completion. The cortex TMP precursors, despite the postulated lower affinity for the enzyme, can now be processed and crystallize.

This simple model, involving only substrate competition for processing enzymes, cannot explain the sharp boundary between the core and cortex regions. It would predict a fuzzy boundary with a concentration gradient of TMPs between the two layers. However, the gene silencing experiments dramatically demonstrate that TMPs belonging to different subfamilies are not interchangeable. Without entering into the details of how TMPs might assemble and be packed in the crystal lattice of the trichocyst matrix (Sperling et al., 1987), this is formally equivalent to saying that ‘core’ and ‘cortex’ TMPs cannot co-crystallize. Our model can thus account for the organization of the trichocyst matrix in two concentric crystalline layers. It also illustrates a strategy for the modular construction of a complex edifice within a membrane bound compartment that could be of general relevance. The successful completion of each part of the final structure is only achieved if the previous part, upon which it builds, has assembled correctly.

We are grateful to Claude Antony for useful advice and help with the double-label experiments. We thank Janine Beisson, Jean Cohen and Carl Creutz for critical reading of the manuscript. This work was supported by the CNRS (Cell Biology Program, grant 9064) and the Microbiology Program of the Ministère de l’Education Nationale, de la Recherche et de la Technologie (Programme de Recherche Fondamentale en Microbiologie et Maladies Infectieuses et Parasitaires). LV was supported by a fellowship from the Association pour la Recherche contre le Cancer (ARC) and by the Association Française contre les Myopathies (AFM).