Genetic, Genomic, and Functional Analysis of the Granule Lattice Proteins in Tetrahymena Secretory Granules
In some cells, the polypeptides stored in dense core secretory granules condense as ordered arrays. In ciliates such as Tetrahymena thermophila, the resulting crystals function as projectiles, expanding upon exocytosis. Isolation of granule contents previously defined five Granule lattice (Grl) proteins as abundant core constituents, whereas a functional screen identified a sixth family member. We have now expanded this screen to identify the nonredundant components required for projectile assembly. The results, further supported by gene disruption experiments, indicate that six Grl proteins define the core structure. Both in vivo and in vitro data indicate that core assembly begins in the endoplasmic reticulum with formation of specific hetero-oligomeric Grl proprotein complexes. Four additional GRL-like genes were found in the T. thermophila genome. Grl2p and Grl6p are targeted to granules, but the transcripts are present at low levels and neither is essential for core assembly. The ΔGRL6 cells nonetheless showed a subtle change in granule morphology and a marked reduction in granule accumulation. Epistasis analysis suggests this results from accelerated loss of ΔGRL6 granules, rather than from decreased synthesis. Our results not only provide insight into the organization of Grl-based granule cores but also imply that the functions of Grl proteins extend beyond core assembly.
Dense core granules (DCGs) are vesicles specialized for storage of highly concentrated proteins, which can be secreted via exocytic fusion with the plasma membrane in response to extracellular stimuli (Arvan and Castle, 1992; Arvan and Castle, 1998). The cargo itself, rather than functioning as a mere passenger in this pathway, seems to play a central role in driving granule formation (Arvan et al., 2002). The mechanisms involved may involve biophysical properties of DCG proteins, including reversible aggregation in response to changes in the ionic composition between successive compartments of the secretory pathway (Chanat and Huttner, 1991). In addition to such conditional solubility, some DCG cargo proteins may interact with the lipid bilayer in noncanonical ways that also may be important for sorting (Dhanvantari and Loh, 2000; Tooze et al., 2001). For these reasons, simply identifying the DCG cargo is an important first step in analyzing granule biogenesis in any specific cell type.
An additional motivation for defining granule cargo exists in the ciliates, single-celled protists that as a group synthesize a variety of DCGs (Rosati and Modeo, 2003). DCG formation in ciliates involves the organization of lumenal cargo proteins as large crystals, which have the remarkable property of undergoing spring-like expansion upon exocytosis in a way that ensures rapid content release (Hausmann, 1978). The assembly of such reproducible crystalline structures within the secretory pathway poses interesting regulatory questions, for example, with regard to nucleation and size control. Molecular analysis of DCGs in ciliates has been chiefly limited to two organisms, Tetrahymena thermophila and Paramecium tetraurelia (Vayssie et al., 2000; Turkewitz, 2004). In the latter, biochemical isolation of granules and protein microsequencing led to cloning of the trichocyst matrix proteins (tmps). These fell into three families that together may include >100 members and that may all adopt a similar overall structure (Madeddu et al., 1995; Gautier et al., 1996). Localization and gene silencing studies provided evidence that Paramecium granule cores were multilayered constructs in which distinct tmp families independently assembled into concentric zones (Vayssie et al., 2001). For example, cells in which the T1 tmp family members were silenced still formed an ordered, central core plug, but they lacked the overlying structure found in wild-type granules.
DCGs in Tetrahymena seem morphologically and chemically less complex than those in Paramecium. The tmp homologues in T. thermophila were named granule lattice (Grl) proteins, and five members were initially identified starting with isolation of the most abundant DCG contents (Chilcoat et al., 1996). Biochemical analysis of Grl1p indicated that conformational changes could underlie the crystalline expansion noted above, whereas gene disruption suggested that Grl1p itself might be present throughout the entire core, because no visibly ordered structure was formed in its absence (Verbsky and Turkewitz, 1998). Granules formed in such ΔGRL1 cells did not undergo rapid expansion upon exocytosis (Chilcoat et al., 1996). This defect, due to the absence of a single granule protein, facilitated an entirely different approach to identifying granule contents based on screening for exocytosis-defective mutants among cells transformed with an antisense library, a technique allowing phenotype-based gene cloning in this organism (Sweeney et al., 1996). This identified a sixth family member, initially called NDC1 (for nondischarge) but here renamed GRL8 for simplicity (Chilcoat et al., 2001). Granules lacking Grl8p, like those lacking Grl1p, had no discernibly organized structure. The tmp/Grl family of proteins have no identified homologues outside of ciliates, but they are nonetheless similar to abundant dense core granule proteins in animal cells, called chromogranins, in having a preponderance of acidic amino acids, which endow them with the capacity to bind calcium (Chanat et al., 1991).
In the work presented here, we have extended the initial genetic screening to ask whether other genes were required for core formation. The results provided strong evidence that the family of six Grl proteins represents the essential structural core components in this organism. At the same time, searching the T. thermophila macronuclear genome revealed four additional paralogs, named GRL2, 6, 9, and 10. We have undertaken functional analysis of a large subset of the GRL family. The results confirm that six GRL genes (GRL1, 3, 4, 5, 7, and 8) each play a nonredundant role in core formation. The assembly defects associated with individual disruption of GRL1, 3, 4, 7, and 8 are similar, suggesting that each may contribute to a common structural unit. In support of this, we present evidence that some Grl proteins assemble into hetero-oligomeric complexes in the endoplasmic reticulum (ER) and that assembly may be required for ER exit. The products of GRL2 and 6 are much less abundant and not essential for lattice formation. Nonetheless, disruption of GRL6 had effects on both granule morphology and number, indicating a novel role for this gene product.
MATERIALS AND METHODS
Cell Strains and Culture
Strain CU428.1, MPRR/MPRR (6-methylpurine-sensitive, VII) bears a dominant allele conferring 6-methylpurine resistance in the micronucleus and the corresponding sensitive allele in the macronucleus; strain B2086, mprS/mprS (6-methylpurine-sensitive, II) has the drug-sensitive allele in both nuclei. Both were provided by Peter Bruns (Cornell University, Ithaca, NY). Cells were grown at 30°C with agitation in 2% proteose peptone, 0.2% yeast extract (both from Difco, Detroit, MI), with 0.009% ferric EDTA. For viewing of green fluorescent protein (GFP)-tagged proteins, cells were grown in 0.2% yeast extract with 0.009% ferric EDTA, to reduce autofluorescence. Unless stated otherwise, reagents were obtained from Sigma-Aldrich (St. Louis, MO).
Genes were disrupted by first using a selectable marker to replace a large region of a cloned copy. The marker consisted of the NEO gene on an autonomous expression cassette (Gaertig et al., 1994). The cloned interrupted copy was then introduced into Tetrahymena, to integrate via homologous recombination. GRL3 was disrupted by replacement by the NEO2 cassette of the sequences –154 to 1526 (these junctions determined by convenient restriction sites) relative to the start codon. Similarly, GRL4 was disrupted by replacement of the sequences –310 to 727 with NEO2. GRL7 was disrupted by replacement of the sequences 1–2402 with a similar cassette, MTT-NEO. The difference between NEO2 and MTT-NEO is that the former uses a histone H4 promoter, whereas the latter uses an inducible metallothionein (MTT1) promoter, to drive expression of the gene responsible for drug resistance (Shang et al., 2002). Disruptions were accomplished by biolistic transformation of CU428.1, as described previously (Chilcoat et al., 1996). Initial drug selection was with 120 μg/ml paromomycin, added 6 h after transformation. Cells transformed with the GRL7 disruption construct were grown in 0.5 μg/ml CdCl2 immediately after transformation and during the entire period of drug selection to induce the MTT1 promoter. Because only some of the macronuclear copies of the target gene are replaced initially, cells were cultured with increasing concentrations of paromomycin (up to 800 μg/ml) and allowed to grow for >60 generations for stringent selection of the disrupted allele.
To make whole cell lysates, ∼8 × 105 cells were pelleted for 30 s in a clinical centrifuge and reduced to a volume of 250 μl before adding an equal volume of 100°C 2 × SDS-PAGE sample buffer (final concentration 100 mM sucrose, 3% SDS, 2 mM Na2EDTA, and 62.5 mM Tris, pH 6.9) and boiled for 3 min. For Western blots, 12 μl of lysate was electrophoresed using SDS-PAGE and transferred to 0.45-μm nitrocellulose (Osmonics, Westborough, MA). Immunoblots with polyclonal antisera were blocked, probed, and washed with 5% milk in tris-buffered saline (TBS). The anti-Grl1p, anti-Grl3p, anti-Grl4p, and anti-Grl8p antibodies were used at 1:1000, 1:400, 1:250, and 1:1500, respectively. Detection of primary antibodies was with horseradish peroxidase-conjugated goat anti-rabbit antibody at 1:2000 (Jackson ImmunoResearch Laboratories, West Grove, PA); blots were developed with Pierce Supersignal (Pierce Chemical, Rockford, IL) and exposed to film.
Immunofluorescence and Microscopy
Fixation and immunolabeling using monoclonal antibody (mAb) 5E9 were as in Bowman and Turkewitz (2001) with two exceptions: the entire procedure was done at room temperature, and the primary incubation was with a 20% (vol/vol) solution of 5E9 hybridoma supernatant. Immunolabeling with 4D11 was as described previously (Bowman and Turkewitz, 2001). 4D11 and 5E9 were the gift of Marlo Nelson and Joseph Frankel (University of Iowa) and recognize p80 and Grl3p, respectively (Turkewitz and Kelly, 1992; Bowman et al., 2005). For live cell microscopy, cells were immobilized in 1% (vol/vol) methyl cellulose (Carolina Biological, Burlington, NC). Samples were viewed under a Zeiss Axiovert microscope interfaced with a Zeiss LSM 510 confocal laser system and software.
Granule content of wild-type and ΔGRL6 lines were analyzed by flow cytometry. For these experiments, cells (∼1.5 × 105/sample) from stationary phase (grown overnight, to 106/ml) or overnight starved cultures were fixed with paraformaldehyde, permeabilized with Triton X-100, and immunostained in 1% bovine serum albumin in TBS, as described above. Cells were incubated with saturating concentrations (1:3 dilution of in vitro hybridoma culture supernatant, to a total volume of 0.4 ml) of either mAb 4D11 or 5E9. After washing, cells were suspended in 200 μl with 1% fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG (Molecular Probes, Eugene, OR), an amount that was also found empirically to be saturating. Negative controls had no primary antibody. Cells were analyzed on an LSR II flow cytometer (BD Biosciences, San Jose CA) using FACSDiVa software. Excitation was done using a 488-nm solid state Sapphire Laser (Coherent, Santa Clara, CA), and the collection filter was a 530/30 band pass filter.
RNA Isolation, In Vitro Transcription, and Northern Blotting
CU428.1 RNA was isolated from exponentially growing cultures or from cells starved in a one-tenth dilution of Dryl's (1.7 mM sodium citrate, 1 mM NaH2PO4, 1.5 mM CaCl2) supplemented with an additional 0.1 mM MgCl2 and 0.5 mM CaCl2 (DMC) for 6 h by using the RNeasy kit (QIAGEN, Valencia, CA). For in vitro transcription reactions, template open reading frames (ORFs) were each amplified from a cDNA library using the following primer pairs: GRL1, 5′-ATGAATAAGAAATTATTAGTTGTCCTTTT-3′ and 5′-TCAGTTAATGAAGTCAATATTGGG-3′; GRL2, 5′-ATGCGCTTAACTGTTATTGTCAC-3′ and 5′-TCAGAATCCACCAGAACTAGCA-3′; GRL6, 5′-ATGAAAATAATTATTATTTTTTTAGCCTC-3′ and 5′-TCATTGGGATTATTCGTCTGAT-3′; and GRL7, 5′-ATGAGAAAAGTCTTCGTTGCTT-3′ and 5′-TCAAGCGCTTTCAGCACTT-3′. The PCR products were directionally cloned into pCRII (Invitrogen, Carlsbad, CA), and the resulting plasmids were linearized for run-off transcription with SP6 (sense) or T7 (antisense) RNA polymerase (Fermentas, Hanover, MD). These transcription products were ethanol precipitated and quantified for use as standards (sense) or as probe templates (antisense). RNA electrophoresis and Northern blotting were performed as per Farrell (1993). After electrophoresis on 1% agarose gels with formaldehyde, RNA was transferred and UV cross-linked to Magna (MSI, Westboro, MA) nylon membranes. Blots were visualized using a PhosphorImager (Amersham Biosciences, Sunnyvale, CA).
Construction of Antisense Libraries
Three antisense libraries were used in this screen. They were constructed using a PCR-based protocol, with one of two cDNA libraries as template. The first antisense library was constructed from the cDNA library described in Chilcoat et al. (2001). After digestion of the cDNA library with SacI and XhoI (to excise it from its pBluescriptvector backbone), we executed a single cycle of PCR with the primer 5′-TGCTAGCCACGGTCCGAGCGGGTACCNNNNNN-3′. The six random nucleotides at the 3′ end of this primer allowed it to anneal randomly within the library inserts, producing truncated copies with an added KpnI site. PCR conditions for this step were 95°C, 3:00; 15°C, 0:30; 70°C, 0:30; and 95°C, 2:00. The product of this reaction was diluted 10-fold and used as template with primers 5′-TGCTAGCCACGGTCCGAGCG-3′ (contained within the above-mentioned primer) and 5′-biotin/ACCGCGGTGGCGGCCGCTCTA-3′ (containing a NotI site, and complementary to the vector backbone adjacent to the 5′ ends of the cDNA inserts). PCR conditions for this step were 95°C, 2:00; 25 cycles of (95°C, 0:45; 53°C, 0:30; 70°C, 0:30); and 70°C, 10:00. After addition of fresh reagents and another cycle of PCR (95°C, 3:00; 50°C, 1:00; and 72°C, 3:00), the final PCR product was digested with KpnI before binding on a streptavidin plate (Pierce Chemical). After washing off unbound products, antisense inserts were eluted by digestion with NotI. After ethanol precipitation, the inserts were ligated into the ribosomal antisense vector 5318DNmod (Chilcoat et al., 2001). The cDNA library used as template for the second and third antisense libraries was constructed using the Creator SMART cDNA Library Construction Kit (BD Biosciences Clontech, Palo Alto, CA). Predigestion was with SmaI and XhoIor EcoRI, XhoI, NotI, ClaI, and AscI to excise the inserts from their pDNRlib vector backbone. PCR reactions were as described above, except that the biotinylated primer used was 5′-biotin/ATGCTCAGATCTATTGCATAGCGGCCGCAATTCGGCCATTACGGCCGGG-3′.
Transformation with Antisense Libraries and Screening
Mating pairs of B2086 and CU428.1 were transformed by electroporation using the ECM 600 (BTX, San Diego, CA) as described previously (Gaertig and Gorovsky, 1995). Initial drug selection was with 100 μg/ml paromomycin, added 16–24 h after transformation. After 3 d at 30°C, the transformants were screened to enrich for those with defects in exocytosis, as follows. After starvation overnight in DMC at 22°C, 50 ml of cells (at a density of 1 × 106cells/ml) was stimulated with one-fourth volume of 0.1% Alcian Blue 8GX, added forcefully via syringe; after 30 s, an equal volume of 2% proteose peptone was added, and the cells were transferred to two 50-ml clear conical centrifuge tubes (Nalgene, Rochester, NY). The cells were pelleted in a clinical centrifuge for 1 min. Most of the supernatant was aspirated to leave 6–8 ml of fluid, in which the pelleted cells were gently resuspended. Over the next 5–10 min, encapsulated cells settled by gravity to the tube bottom, whereas free-swimming (exocytosis-deficient) cells migrated toward the air–water interface and concentrated at the meniscus. The latter were collected and transferred to 50 ml of fresh medium and grown for 8 h at 22°C, at which time the enrichment procedure was repeated. The free-swimming cells collected after this second step were diluted into fresh medium for distribution into 96-well plates (25 μl/well) at a density of ∼30 single cells per plate. After 4 d at 30°C, the cells were starved by addition of 125 μl/well DMC and incubation at 30°C for an additional 4 d. Wells were then screened for the absence of capsules after stimulation with 25 μl of 0.1% Alcian Blue, followed by rescue with 25 μl of 2% proteose peptone. Initially, all encapsulation-deficient wells were confirmed by retesting, but this step was later omitted because of the negligible number of encapsulation-positive wells identified during rescreening, i.e., the assay results were highly reproducible.
Whole Cell PCR
Wells bearing encapsulation-deficient cells were replicated to master plates and grown to stationary phase. The antisense inserts harbored by the cells in each well were identified by whole cell PCR. Briefly, 50 μl of cells were mixed with 12.5 μl of K buffer (50 mM Tris-HCl, pH 8.8, 250 mM KCl, and 0.4% Nonidet P-40) and incubated for 1 h at 55°C and 10 min at 100°C. Then, 25 μl of this cell lysate was mixed with 25 μl of PCR mix (10 mM Tris-HCl, pH 8.8, 50 mM KCl, 0.08% Nonidet P-40, 2 mM MgCl2, 2 mM of each nucleotide, 1 μM of the primers 5318f1 [5′-AAAAGGTCGATGAGTAAGGAAATG-3′) and 5318r2 (5′-CAATCTCAGGGTACGCGG-3′], and 3 U of TaqDNA polymerase [Fermentas]) and subjected to the following: 95°C, 3:00; 30 cycles of (94°C, 0:30; 55°C, 0:30; and 72°C, 1:00); and 72°C, 10:00. Products were visualized on 4% agarose gels, and those showing a single band were purified for sequencing. Products yielding antisense sequences of interest were digested with KpnI and NotI and ligated into the ribosomal antisense vector 5318DNmod for retransformation of Tetrahymena by electroporation, as described above.
Expression of GFP-tagged Proteins
Grl2p-GFP and Grl6p-GFP were expressed in Tetrahymena from the vector pVGF.MTT, a modified version of pVGF.1 (Haddad et al., 2002). pVGF.MTT was created by replacing the ribosomal RPL29 promoter in pVGF.1 (flanked by NotI and PmeI restriction sites) with the MTT1 promoter (amplified from cDNA using primers 5′-CCGGTCACCTTTTTTATTTAGTAAAAATT-3′ and 5′-CCGTTTAAACTATTTTAAGTTTAGTATTAT-3′) after replacing the NotI site with one for BstEII. The GRL2, GRL6, and GRL7 coding sequences were amplified from cDNA using primer pairs 5′-AAGTTTAAACATGCGCTTAACTGTTATTGTCAC-3′ and 5′-TTGTCAGCTGAAGAATCCACCAGAACTAGCACTT-3′, 5′-ATGTTTAAACATGAAAATAATTATTATTTTTTTAG-3′ and 5′-TTAACAGCTGAATTGGGATTATTCGTCTGATC-3′, and 5′-ACGTTTAAACATGAGAAAAGTCTTCGTTGCTT-3′ and 5′-TATCAGCTGAAAGCGCTTTCAGCACTTCC-3′, respectively; these pairs provide restriction sites for blunt ligation into pVGF.MTT at its PmeI site. The second primer of each pair removes the stop codon from the end of the coding sequence, allowing for insertion in frame with an enhanced GFP tag (BD Biosciences Clontech) in the vector. Correct ligation was verified by direct sequencing. After transformation into Tetrahymena, protein expression was induced by growth in 0.2% yeast extract, 0.009% ferric EDTA with 0.1 μg/ml CdCl2 for 24 h at 30°C. Grl1p-GFP was expressed from the ncvB vector (Bowman et al., 2005), which results in the expression of protein from the endogenous chromosomal MTT1 locus. The Grl1p-GFP coding sequence was produced by creating the GFP fusion protein in pVGF.MTT (as above, using primers 5′-GTTTAAACATGAATAAGAAATTATTA-3′ and 5′-TTGTCAGCTGAAGTTAATGAAGTCAATATTGG-3′), and shuttling this fusion into ncvB using the flanking PmeI and ApaI restriction sites. The Igr1p-GFP coding sequence was produced by shuttling the IGR1 ORF from the pVGF.1 vector (Haddad et al., 2002) into the pVGF.MTT vector (using PmeI and XhoI restriction sites), followed by the insertion of the GFP tag and transfer into ncvB, as described above. Grl1p-GFP expression was induced by growth in SPP with 2 μg/ml CdCl2 for 6–8 h at 30°C.
High Level Expression of GRL1
The GRL1 coding sequence was amplified from cDNA using the primers 5′-GTTTAAACATGAATAAGAAATTATTA-3′ and 5′-CTCGAGTCAGTTAATGAAGTCAAT-3′, which provide restriction sites for insertion into pVGF.MTT between the PmeI and XhoI sites. This ligation step removes the GFP sequence from the vector, resulting in the expression of full-length untagged GRL1. Protein expression was induced by growth in SPP with up to 3 μg/ml CdCl2 for 4–5 h at 30°C.
Expression of GRL1–6xHis
The endogenous GRL1 coding sequence was replaced with GRL1-6xHis by modifying a vector used by Bradshaw et al. (2003) to replace the endogenous GRL1 gene with genes that contain altered GRL1 coding sequences. Starting with a template vector that encodes the wild-type Grl1p sequence, the primers 5′-CACCACCACCACCACCACTGAAAAATGATGTGATTT-3′ and 5′-GTTAATGAAGTCAATATTGG-3′ (MWG, High Point, NC) were used to amplify the vector by inverse PCR. The amplified product was blunted with T4 DNA polymerase and then digested with DpnI to destroy the template before self-ligation (New England Biolabs, Beverly, MA). The modified coding sequence (full length GRL1 plus six histidine codons placed immediately upstream of the stop codon) was confirmed by sequencing, and the construct was used for biolistic transformation of CU428.1.
Purification of proGrl1p–His Complexes
Purification followed Bowman et al. (2005) except that the lysis, wash, and elution buffers did not contain urea or guanidine.
Production of Polyclonal Anti-Grl4p Antibodies
To obtain immunogen for the production of anti-Grl4p antibodies, secretory granule contents were purified from wild-type cells (Turkewitz et al., 2000). Approximately 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 (Verbsky and Turkewitz, 1998), Grl4p was excised with a scalpel. Rabbit antiserum against this band was then produced commercially (Zymed Laboratories, South San Francisco, CA).
Exocytosis Testing of ΔGRL Strains
Strains were tested for Alcian Blue-induced capsule formation by using cultures starved overnight in DMC and for dibucaine-induced flocculent release by using stationary cultures (0.8–1 × 106 cells/ml in SPP medium) as described previously (Turkewitz and Kelly, 1992; Melia et al., 1998). To quantify dibucaine-stimulated release from ΔGRL6 cells relative to wild-type, 6 × 106 cells were suspended in 1 ml of 10 mM Na-HEPES, pH 7.0, 0.5 mM CaCl2, and stimulated for 15 s by addition of 110 μl of 25 mM dibucaine. Cells were then diluted with buffer to 15 ml and pelleted for 5 min at 300 × g. The released granule contents, which form a distinct flocculent layer above the cell pellet, were removed with a transfer pipette and transferred to a graduated conical tube for repelleting and volume measurement. Protein concentrations in the flocculent layer were measured using the Pierce protein assay.
Analysis of Functionality through Evolutionary Constraint
The translations of the sequences were first aligned using ClustalW using default parameters. Afterward the codons were mapped directly onto the peptide alignments to produce a multiple alignment with gaps in multiples of three. For estimating rates of evolution, an unrooted phylogeny for the three genes was assumed. The substitution rates at amino acid changing sites (KA) and at synonymous sites (KS) were calculated using the codeml software (Yang 1997). KA is the ratio of number of nucleotide substitutions that result in an amino acid change to the total possible number of amino acid changing nucleotide substitutions. KS is a similar ratio corresponding to synonymous changes within codons. The ratio KA/KS (also called ω) reflects the nature evolutionary constraint between sequences. Because silent substitutions do not affect the peptide, the KS ratio is used as a neutral standard. If KA = KS (ω = 1), this indicates a lack of constraint at the amino acid level. If, however, KA < KS (ω < 1), then the rate of protein evolution is constrained, indicating selection against amino acid substitution and implying protein functionality. Finally, if KA > KS (ω > 1), then protein evolution is accelerated, indicating natural selection for amino acid substitution. Evolutionary constraint was evaluated by testing four separate models of protein evolution. In the first model (H0), the rates of all branches were estimated from the data. In the remaining three models, ω in the branch leading to each sequence was fixed at a value of 1 (neutrality), whereas the other branches were estimated. The models are described as follows: H0) Estimate ω for all branches; H1) Set ω for GRL5 = 1, and estimate the others; H2) Set ω for GRL9 = 1, and estimate the others; and H3) Set ω for GRL10 = 1, and estimate the others.
To evaluate the probability that a given sequence differed from the neutral rate of ω = 1, a likelihood ratio test (Yang, 1998) between H0 and each model was conducted as follows. The X2 statistic is calculated as
Screening for Antisense Ribosomes That Confer Exocytosis Defects
To identify genes required for regulated exocytosis of secretory granules in T. thermophila, we transformed cells with both existing as well as new libraries of antisense ribosomes. We screened transformants to identify the subset with defects in exocytosis, as judged by the inability to form capsules when stimulated with the secretagogue Alcian Blue. Such cells were obtained first by physical enrichment and then by visual screening of individual clones in 96-well plates. Clones that failed to undergo visible exocytosis when challenged with Alcian Blue were expanded, and the antisense inserts were then amplified by whole cell PCR using primers directed against the insert flanking sites at the rDNA locus. As can be seen in a representative agarose gel, multiple PCR products were obtained for many transformant lines (Figure 1). These were due to multiple antisense inserts within individual cells, rather than to nonclonality among the transformant lines, because we obtained the same result when single cells were again isolated and clonally expanded (our unpublished data). For practical purposes, we focused on the subset of cells seeming to have a single major insert.
We obtained sequences of 304 antisense inserts associated with strong defects in exocytosis in the primary screen. From these sequences, we then identified the corresponding genes by querying the expressed sequence tag (EST) database, in which a large fraction of entries include sequence from the 5′ untranslated region. Of these 304, 237 had previously been identified in T. thermophila, and 67 were novel. Among the former were all six of the known GRL family (GRL1, 3, 4, 5, and 7, and NDC1/GRL8, and multiple independent inserts were obtained corresponding to each family member; Figure 2). For the sake of simplicity, we henceforth refer to NDC1 as GRL8. In fact, inserts corresponding to GRL genes comprised 181 of these inserts (Figure 2). We confirmed the significance of a subset of these inserts by reinserting them into the empty rDNA vector and generating new transformants for testing. In all cases, inserts corresponding to GRL genes regenerated exocytosis-deficient transformants. The remaining 56 previously identified genes all fell into the class of “housekeeping” genes, particularly corresponding to ribosomal proteins, whose suppression would be predicted to impair cell viability. Because housekeeping genes were unlikely to have roles specific for granule synthesis, they were not further examined.
To begin analyzing the set of novel genes, we first tested the ability of the inserts to regenerate the exocytosis deficiencies, as described above. However, in contrast to the GRL gene inserts, in no case did retransformation regenerate the exocytosis deficiency. In such cases, the primary transformants may have harbored one or more additional rDNA inserts, not detected by whole cell PCR, that were responsible for the deficiencies. Previous work has demonstrated that an antisense ribosome constituting just a small minority of the total ribosome population in a given cell can produce a null or hypomorphic phenotype (Sweeney et al., 1996). Alternatively, the deficiencies may have been due to mutations unrelated to the rDNA locus, induced as a result of the transformation process. The genes corresponding to these inserts were not analyzed further.
Testing GRL Function by Individual Gene Disruption
Although this functional screen did not succeed in identifying new genes involved in regulated secretion, the results strongly suggested that each of the GRL gene products, previously identified by isolation of DCG contents, plays an essential, nonredundant role in regulated exocytosis. To test this directly and to evaluate the function of individual Grl proteins, we disrupted each of three GRL genes (GRL3, GRL4, and GRL7) in independent cell lines (Figure 3). We did not target GRL5 for disruption because Southern blotting indicated that multiple gene copies might be present. This issue was clarified when the T. thermophila genomic sequence became available and is discussed in a later section. As expected, the ΔGRL3, ΔGRL4, and ΔGRL7 lines were each completely defective in Alcian Blue-stimulated capsule formation (our unpublished data; Figure 11). The deficiencies were due to defects in DCG synthesis and specifically in the assembly of the core structure. As revealed by electron microscopy, each of the ΔGRL lines accumulated aberrant DCGs that differed from the wild type in shape, being spherical rather than elongated (Figure 3F). In addition, the spherical granules lacked any visibly ordered core structure. There were no discernible differences between the granules formed in these different strains; moreover, we also found similar granules in a double-disruption ΔGRL3, ΔGRL4 cell line (our unpublished data). At this level of analysis, the granules in strains lacking GRL3, GRL4, or GRL7 seemed structurally and functionally identical to those previously characterized in ΔGRL1 or ΔGRL8 cells (Chilcoat et al., 1996; Chilcoat et al., 2001). These results differed markedly from those in Paramecium, in which distinct granule assembly defects were induced by silencing of specific tmp families (Vayssie et al., 2001). Instead, our results suggested that the Tetrahymena Grl proteins all play similar roles during core formation. To account for these results, we hypothesized that each protein is present throughout the elongated lattice found in wild-type cells. One possibility is that each contributes to a Grl hetero-oligomeric complex, which could function as an assembly subunit.
In Vitro Evidence for proGrl Hetero-oligomeric Complexes
To determine whether such hetero-oligomers exist, we began with a biochemical approach. Attempts to dissociate mature granule cores into putative subunits were foiled by the striking stability of the assembled lattice under all but strongly denaturing conditions (our unpublished data). On the other hand, before proteolytic processing and core assembly, the Grl proteins are fully soluble (Turkewitz et al., 1991). To ask whether Grl proproteins assemble as hetero-oligomers, we engineered GRL1 with a carboxy-terminal hexahistidine extension. This was recombined at the native GRL1 locus as a gene replacement, to achieve a wild-type pattern and level of expression. ProGrl1p-6His was isolated from nondenaturing detergent cell lysates by nickel affinity chromatography, and antibodies against other Grl proteins were used to ask whether these coeluted with proGrl1p-6His. Mature processed Grl1p is entirely insoluble under these conditions, and was therefore discarded before the chromatography step. Western blotting, after SDS-PAGE and transfer to nitrocellulose, revealed that at least two other proGrl proteins were bound to the nickel-Sepharose, in a manner that depended on the presence of proGrl1p-6His (Figure 4). Although these data do not establish whether one or more of these proteins directly contacts proGrl1p or whether such contacts are specific, they are consistent with assembly of a set of Grl proproteins as an early step in core formation. Because we do not possess antibodies against the remaining Grl proteins, we could not determine whether these were also present.
In Vivo Evidence for Hetero-oligomer Formation
To understand the in vivo significance of such oligomers, we analyzed Grl proteins in a subset of the GRL null lines described previously. The logic of these experiments was that specific oligomer formation, if it occurred at the level of the ER, could be a prerequisite for ER exit. This has been observed for multimeric complexes in many other systems, including at least one complex targeted to secretory granules (Chanat et al., 1999). We therefore asked whether the absence of other Grl proteins blocks ER exit of proGrl1p. As a first measure, we assessed the extent of proGrl1p processing, a step that occurs in a post-trans-Golgi network (TGN) compartment and which therefore requires exit from the ER. The processing of proGrl1p was significantly inhibited in ΔGRL4 cells (Figure 5A). That this reflected a specific interaction between proGrl1p and proGrl4p was supported by the observation that only a partial inhibition was seen in ΔGRL3 and ΔGRL8 cells. The results suggested that proGrl1p, when expressed in the absence of proGrl4p, failed to reach the post-TGN processing compartment due to its retention in the ER. To assess this more directly, we expressed a GFP-tagged variant of GRL1 in wild-type, ΔGRL3, ΔGRL4, and ΔGRL8 cells. In wild-type cells, the protein localized as expected to the docked DCGs (Figure 5B). In ΔGRL4 cells, on the other hand, Grl1p-GFP was found primarily in a subcortical reticular pattern, distinct from the punctate pattern of docked DCGs. This reticular pattern is likely to reflect the cortical endoplasmic reticulum but also may overlap with alveoli, which are flattened cisternae underlying the plasma membrane (Frankel, 2000). Alveoli are involved in calcium-based signaling and may represent a specialized ER subcompartment. We observed identical patterns in cells transformed to express GFP (with an N-terminal signal sequence), fused at the C terminus to carboxy-terminal tetrapeptides that confer ER localization in other species (KDEL or HDEL) (our unpublished data) (Andres et al., 1991). This localization depended upon the ER retention signal, because signal sequence GFP, without any ER-retention signal, localizes to small cytoplasmic vesicles in these cells or is found in the medium (Haddad et al., 2002; Bowman et al., 2005). Importantly, mislocalization of proGrl1p-GFP to the cortical reticulum in ΔGRL4 cells was specific, because Grl1p-GFP expressed in ΔGRL3 and ΔGRL8 cells accumulated primarily in the DCGs. Unlike in wild-type cells, ΔGRL3 and ΔGRL8 cells also showed a low level of proGrl1p-GFP labeling of non-DCG structures, consistent with the partial inhibition of processing (Figure 5B).
GRL1 Overexpression Leads to ER Accumulation
These results suggested that proGrl1p and proGr4p associate at the level of the ER, as a prerequisite to ER exit. This may regulate the stoichiometry of subunit assembly in the multicomponent lattice. A prediction was that overexpression of GRL1 would lead to its partial retention in the ER, because the amount of proGrl4p or other potential binding partners would be limiting under such circumstances. To test this, we expressed GRL1 from a multicopy rDNA-based vector, under the control of the MTT1 promoter. Grl1p accumulated in the cells in the proprotein form, at a level so high that it was visible as a prominent Coomassie Blue-staining band after SDS-PAGE of whole cell lysates (Figure 6A). Analysis of these cells by thin sectioning and electron microscopy revealed that the cytoplasm contained abundant ER filled with a visible aggregate, organized in some regions as a geometric lattice (Figure 6B). This crystallization was probably widespread, because many ER lumena were distended in distinctly angular shapes. Such distended ER is not seen in wild-type cells. Interestingly, cells overexpressing GRL1 also had a normal complement of docked DCGs with wild-type morphology, suggesting that the endogenous level of proGrl1p still exits the ER even while most of the proGrl1p is retained (our unpublished data). Consistent with this, Western blotting of whole cell lysates, derived from cultures induced over a range of cadmium concentrations, showed that a constant amount of mature Grl1p was generated, irrespective of the GRL1 expression level (Figure 6C).
It is also noteworthy that the proGrl1p in such overexpressing cells did not seem aggregated by biochemical criteria. When cells were solubilized in nondenaturing detergent (1% Triton X-100), essentially all of proGrl1p was soluble (350,000 × g; 20 min) (our unpublished data). These results suggest that ER retention of proGrl1p is not due to misfolding.
Genome Analysis Reveals Four Novel GRL Genes
Previous biochemical analysis as well as the functional screen reported here converged on a set of six GRL genes. In particular, the results of the latter suggest that no other genes play essential roles in granule assembly. However, genes with redundant functions would have been invisible. The same may be true for genes with nonredundant functions that are transcribed at a low level, because representation in the antisense libraries depends in part on relative mRNA abundance. We took advantage of the recent completion of the sequencing of the T. thermophila macronuclear genome, in addition to establishment of an EST database, to ask whether these six genes represent the entire GRL family. Homology searches with each of the six genes revealed the likely existence of four novel related genes in the macronuclear genome, which we have named GRL2, 6, 9, and 10 (Figure 7A). The relationship between the GRLs is shown in Figure 7B. The predicted new family members all bear characteristic Grl protein features, including their overall length, highly acidic character, and a wealth of predicted short α-helical coiled coils. The extent of amino acid identity between the six previously identified GRL genes is low, on the order of 20% (Verbsky and Turkewitz, 1998). This is also true for GRL6, but the other three newly detected GRL genes show a pattern indicating a different evolutionary history. GRL9 and 10 are very closely related (≥85% identity) to GRL5, and their hybridization with a GRL5 probe can account for the Southern blotting results mentioned previously. GRL2 is most closely related (>40% identity) to GRL1. The genes seem to represent relatively recent duplications in an otherwise deeply branched phylogeny. Grl2p is unusual in having a nonconservative substitution within a four-amino acid motif that is the most highly conserved 1° structural feature in the Grl/tmp family (Figure 7C) (Verbsky and Turkewitz, 1998).
The fact that these four hypothetical genes eluded previous detection suggested that some of their products might have roles distinct from the six established members. Functional analysis was confined to GRL2 and 6, because the similarity of GRLs 5, 9, and 10 made it challenging to target those genes individually. GRL9 and 10 are likely to be transcribed at relatively low levels, because many independent GRL5 clones, but no GRL9 or 10 clones, were identified in the EST library (all EST sequences available through standard public databases). However, neither GRL9 nor 10 seems to be a pseudogene, based on direct and indirect evidence. The GRL9 product was unambiguously detected in mass spectrometric analysis of isolated granules (our unpublished data). The argument for GRL10 is based on analysis presented below, indicating that this gene is evolving under natural selection for amino acid substitution. Because pseudogenes are expected to evolve neutrally, this is evidence against GRL10 being a pseudogene.
To test for the presence or absence of functional constraint during evolution of GRL5, GRL9, and GRL10, we measured the divergence between the sequences. Some nucleotide substitutions in a codon do not change the peptide sequence (synonymous substitutions), whereas others result in the substitution of one amino acid residue for another (replacement substitutions). The ratio of synonymous substitutions to total number of synonymous sites is called KS, whereas the ratio for replacement substitutions is called KA. Under the assumption that synonymous substitutions are invisible to natural selection and thus represent the mutation rate in the absence of selection, the ratio of KA to KS, or ω, measures the selective constraint on amino acid substitutions in a sequence. Table 1 shows the results of testing the rates of evolution on the unrooted phylogeny between GRL5, GRL9, and GRL10. These results indicate that the rates of nonsynonymous substitution for GRL5 and GRL9 do not differ significantly from the rates of synonymous substitution, because the ratio ω does not differ significantly from 1. However, for GRL10 in particular the rate of replacement substitution is significantly elevated with respect to synonymous substitution, because its value of ω is significantly greater than 1. This suggests that GRL10 is not evolving as would be expected for a pseudogene, i.e., neutrally.
|GRL10|| 0.13429 || 0.0485 || 2.77 || 9.9041 ||0.00165|
In Vivo Analysis of GRL2 and GRL6
GRL2 was present as a single clone in the EST library, and both GRL2 and 6 could be amplified by PCR as full-length copies from cDNA libraries, indicating that these genes are transcribed. As shown by Northern blotting, however, the mRNAs are present at levels much lower than those of the previously described GRL genes (Figure 8). The regulation of expression also seems to be different, because GRL1 and 7, but not 2 and 6, show greater accumulation (as a fraction of total mRNA) in starvation versus growth conditions. To characterize GRL2 and 6, we first asked whether the gene products localize to DCGs by engineering them as GFP-tagged copies. As shown in Figure 9, expression of either Grl2p-GFP or Grl6p-GFP from the MTT1 promoter produced a fluorescent array of cortical puncta, indicating that these proteins accumulate in docked granules. Nonetheless, neither GRL2 nor 6 is essential for DCG assembly and exocytic capsule formation, as shown via gene disruption (Figure 10). Both ΔGRL2 and ΔGRL6 cells were indistinguishable from wild type in Alcian Blue-stimulated capsule formation, which is a qualitative measure of exocytosis (Figure 11A). Exocytosis seemed to be as efficient as wild type, because ΔGRL2 and ΔGRL6 cells underwent complete degranulation after exposure to Alcian Blue as judged by immunofluorescence (our unpublished data). Consistent with this, electron microscopy (EM) analysis of fixed thin sections revealed docked DCGs with crystalline cores (Figure 11B). Based on these observations, we tentatively concluded that neither Grl2p nor Grl6p plays an important role in granulogenesis.
We were therefore surprised to observe that ΔGRL6 cells showed a large quantitative deficiency in exocytosis. When Tetrahymena are treated with the local anesthetic dibucaine, granule contents are released via exocytosis but do not form a capsule, as they do in the presence of Alcian Blue. Instead, the expanded granule lattices dissociate from the cells and form a sedimentable flocculent whose volume is many times greater than that of the cells themselves (Satir, 1977). Cell lines disrupted in GRL1, 3, 4, 7, or 8, or bearing a GRL5 antisense insert, do not produce any visible flocculent when stimulated, because the granule cores in such cells do not expand upon exocytosis (Chilcoat et al., 1996; our unpublished data). When ΔGRL6 cells were stimulated with dibucaine, the volume of the flocculent was reduced 40–70% relative to wild type (n = 10) (our unpublished data). This smaller volume could result from the release of a smaller number of DCGs, or from DCGs containing cores that undergo more limited expansion than those in wild type. The latter was ruled out by measuring the protein content of the flocculent: ΔGRL6 cells not only released a smaller volume of flocculent than wild type, but it contained a correspondingly smaller amount of protein (n = 2) (our unpublished data). In fact, visualization of four independent ΔGRL6 lines by immunofluorescence of fixed cells demonstrated that they contain significantly fewer docked DCGs than in wild-type cells or other ΔGRL lines (Figure 12A).
The DCGs also seemed wider, when viewed en face, than those in wild type. To be certain that this was not a fixation artifact, we transformed wild-type, ΔGRL1, ΔGRL2, and ΔGRL6 cells with Igr1p-GFP, a fluorescent marker for DCGs (Haddad et al., 2002). Analysis of docked DCGs in living cells confirmed that the ΔGRL6 DCGs are both shorter and wider than those in wild type, although clearly different from the spherical DCGs in ΔGRL1 (Figure 12B). ΔGRL2 DCGs were not distinguishable from wild type.
The difference in DCG accumulation also was supported by immunodetection of granule contents in whole cell lysates, because the amounts of several Grl proteins, in their processed forms, seemed reduced in ΔGRL6 cells relative to wild type or ΔGRL2 (Figure 12C). Unexpectedly, the ΔGRL6 cells also showed increased accumulation of the corresponding Grl proproteins. As discussed above, this could reflect either a transport or an assembly defect. Immunolocalization of Grl3p in these cells, using mAb 5E9, demonstrated that all detectible protein was present in the docked secretory granules, suggesting the latter (our unpublished data). It is important to note that these Western blots cannot be used to judge the relative levels of pro- and processed versions of each species, because some antibodies seem to be far more reactive on Western blots against the proprotein forms (Bowman and Turkewitz, unpublished). These results indicate that Grl6p plays an important role in DCG biogenesis. Given the fact that its absence alters core dimensions without eliminating lattice formation, and results in decreased granule accumulation, we conclude that this role is different from that of the six major Grl proteins. Table 2 summarizes the phenotypes associated with the set of GRL disruptions.
Rapid release of granule contents
|None (wild type)||Elongated, crystalline core||Wild type||Yes|
|ΔGRL1, 3, 4, 7, or 8 (macronuclear)||Spherical, amorphous core||∼Wild type||No|
|ΔGRL2 (macronuclear)||Elongated, crystalline core||∼Wild type||Yes|
|ΔGRL6 (macronuclear)||Shorter, wider than WT, crystalline core||Reduced||Yes|
|GRL5 antisense || Spherical, amorphous core || ∼Wild type || No |
Epistasis Analysis of GRL6 Function
The reduction in granule accumulation in ΔGRL6 cells could reflect a decreased rate of synthesis, if this relatively low abundance core protein normally acts as a master regulator. Alternatively, ΔGRL6 granules could be synthesized at the same rate as wild type, but lost at an accelerated pace. Because we saw no indications of cytoplasmic granule degradation in light or electron micrographs of ΔGRL6 cells, we hypothesized that such accelerated loss could occur by an increase in nonstimulated exocytosis. To explore this idea, we disrupted GRL6 in a mutant cell line, MN173, in which granule transport to the plasma membrane is impaired (Figure 13A) (Melia et al., 1998). As a result, MN173 cells accumulate secretory granules in the cytoplasm, rather than docking them at the plasma membrane. If GRL6 is required for granule synthesis per se, MN173 ΔGRL6 double mutants would be expected to accumulate fewer granules than MN173 itself. However, if ΔGRL6 granules are lost from cells by premature exocytosis, this defect might be suppressed in the MN173 background, because such exocytosis requires that the granules first be docked at the plasma membrane.
Neither microscopy nor dibucaine-stimulated protein release could be readily used to assess granule abundance in the MN173 background, because granules are not docked at the cell surface. Instead, we used flow cytometry of fixed permeabilized cells that were immunostained with antibodies against either of two granule contents proteins, Grt1p and Grl3p. Using this approach, we first confirmed that ΔGRL6 cells contain fewer granules than wild type (Figure 13B). Similar results were obtained using either antibody, and with cells from either stationary or starved cultures. For the double mutant analysis, we selected two lines derived from a single transformation in which we introduced a NEO-disrupted copy of GRL6. Drug-resistant clones derived from this transformation are expected to contain a variable number of the wild-type versus disrupted allele, during the several week interval during which the latter is driven to fixation (Turkewitz et al., 2002). Southern analysis of these two lines shows that, in clone 7, the wild-type allele is still present at wild-type levels, whereas complete replacement by the null allele has occurred in clone 10 (Figure 13A). We measured granule content in these two clones. In striking contrast to the results of GRL6 disruption in a wild-type background, the ΔGRL6 MN173 cells (clone 10) showed no decrease in the accumulation of Grl3p, relative to clone 7 (Figure 13B). Interestingly, on average, MN173 cells lacking GRL6 accumulated a somewhat higher level of both granule markers than MN173 cells with wild-type levels of GRL6 (Figure 13C). Immunofluorescence visualization of Grl3p in these cells showed the expected pattern of cytoplasmic granules, and identical flow cytometric results were obtained using either mAb 4D11 or 5E9, which recognize two different, unrelated granule markers (our unpublished data). These results do not support the idea that GRL6 is required for efficient granule synthesis, because MN173ΔGRL6 cells show no deficiency in granule accumulation. Instead, they suggest that absence of Grl6p in an otherwise wild-type cell causes accelerated granule loss. This loss is likely to be via a process requiring transport to the plasma membrane, because the reduction in granule accumulation is suppressed in the MN173 background.
Ciliate DCG formation is driven by protein self-assembly into ordered multisubunit structures (Adoutte et al., 1984). Our results indicate that GRL1, 3, 4, 5, 7, and 8 are essential for such assembly in T. thermophila. Because Tetrahymena granules clearly contain other proteins, their absence from the antisense screen output suggests they do not play essential nonredundant roles in core formation. At the same time, the results indicate limitations to this approach. An unbiased genetic screen would have been expected to yield mutations affecting multiple steps upstream of granule content release (Orias et al., 1983; Melia et al., 1998). The fact that only granule assembly mutants were recovered, linked to GRL deficiencies, reflects a bias for highly expressed genes, due to the fact that antisense libraries were constructed from pooled mRNAs. Thus, a conservative conclusion is that no other nonessential genes, expressed at levels comparable with the six GRLs, are essential for granulogenesis.
Our results reinforce reports that antisense sequences overlapping the 5′ UTR are most effective, because all inserts we obtained met this criterion (Sweeney et al., 1996; Jacobs et al., 2004). The results also suggest useful modifications. First, library normalization could reduce representational biases due to transcript abundance. Second, modifying the transformation protocol to limit the uptake of antisense rDNAs could enhance this approach, because most transformants maintained multiple vectors. We focused on cells from which we recovered single antisense rDNAs. Our results hinted that many of these may have harbored additional inserts, however, because many of the recovered inserts did not regenerate exocytosis phenotypes. Because antisense ribosomes can suppress translation, even when present as a minority, we hypothesize that unrecovered inserts accounted for the exocytosis defects in such transformants. The simplest hypothesis is that these also corresponded to GRL genes.
In ΔGRL1 or ΔGRL8 cells, DCGs are spherical rather than elongated (Chilcoat et al., 1996; Chilcoat et al., 2001). The same was found here following disruption of GRL3, 4, or 7. These results suggest that each of the Grl proteins is distributed throughout a relatively uniform lattice. The possibility that six Grl proteins coassemble to form a repeating lattice subunit is consistent with the comparable abundance of five of the GRL products (Verbsky and Turkewitz, 1998). We infer that the sixth, Grl8p, is also abundant, because the gene is well represented in EST libraries.
Assembly of the DCG lattice begins early in the secretory pathway, because proGrl1p fails to exit the ER in a ΔGRL4 strain, but not ΔGRL3 or 8. This may serve as a mechanism to enhance the efficiency of lattice formation, by imposing a compartmentally controlled order of assembly, and may establish the stoichiometry of Grl1p and Grl4p products. This system is highly stringent, so that the accumulation of processed Grl1p is independent of the amount of proGrl1p available. The disruption of GRL3 or GRL8 partially inhibited proGrl1p processing. If we assume that all proGrl1p that exits the ER undergoes subsequent processing, this inhibition could reflect partial retention of proGrl1p in the ER. This would suggest that other proGrl proteins contribute to transport of proGrl1p, without being absolutely required. However, inhibition also can be explained if processing depends upon subsequent steps that are blocked in the mutants.
An attractive model of lattice assembly posits that soluble Grl (or tmp) precursors are delivered to immature granules, where processing generates assembly-competent products (Adoutte et al., 1984; Vayssie et al., 2001). However, in ΔGRL3 and ΔGRL8 cells, Grl1p-GFP seems overwhelmingly to accumulate in granules, and yet Western blots indicate significant accumulation of proGrl1p. Second, proGrl1p evidently does not require processing to crystallize, because it does so in the ER when overexpressed. This recalls reports in which mammalian DCG proteins assembled in the ER or Golgi under stress conditions, reflecting the tendency of these proteins to crystallize (Tooze et al., 1989). Although proGrl1p crystallization in the ER may not be physiological, assembly of proproteins is also suggested by images of immature granule-like vesicles in two mutants, UC620 and 623. These show very limited processing of proGrl proteins, but the cores nonetheless show regions organized as lattices (Bowman et al., 2005). These results may be reconciled if processing, rather than acting before assembly, instead acts selectively upon proproteins that have recently assembled, thereby rendering the process irreversible. This model could account for the partial inhibition of proGrl1p processing in all ΔGRL strains, because the absence of any Grl protein would preclude formation of a subset of lattice contacts.
Although six GRL genes encode the basic lattice architecture, four additional related genes exist in the genome. Two (GRL9 and 10) are closely related to GRL5. Neither seems to be transcribed at a high level based on their absence from the EST database. Interestingly, analysis of amino acid substitution rates indicated that GRL10 is evolving under elevated rates of amino acid substitution, strongly implicating the role of natural selection in its evolution. The significant deviation from ω = 1 for GRL10 provides strong evidence of nonneutral evolution (either purifying selection or directional selection), suggesting both that the gene is transcribed and that it confers a phenotype. In contrast, the substitution rates of GRL5 and GRL9 cannot be distinguished from the neutral evolution rate. For those genes, the observation of ω near 1 indicates only that on average the sequences in question cannot be distinguished from the neutral rate. This is compatible with either all codons evolving at a rate such that ω is near 1, or a subset of codons evolving under constraint with others evolving under natural selection and/or neutrality such that the average ω is near 1.
GRL2 and 6 were analyzed by several criteria. The GFP-tagged proteins were targeted to DCGs, but the endogenous genes are transcribed at <5% the level of the major GRLs. Consistent with this, gene disruption revealed that neither is required to form exocytosis-competent DCGs. Nonetheless, disruption of GRL6 had a remarkable effect on the size of the exocytic response, reducing the amount of material released, as well as changing granule morphology. The former seems primarily due to a reduction in DCG number. For some mammalian cells, specific cargo proteins, namely, members of the chromogranin family, may play important roles in determining granule number (Kim et al., 2001). The mechanisms are unknown, but chromogranins are major structural components of dense cores (Chanat et al., 1991). The ΔGRL6 phenotype is consistent with this idea, but with the important difference that Grl6p is apparently present at relatively low levels. This estimate of protein scarcity is inferred from transcript abundance, and is consistent with the fact that no GRL6 products have been detected as major granule components. Several different phenomena might contribute to DCG reduction in ΔGRL6 cells. Grl6p could be a master regulator of GRL gene expression, or might be required for correct targeting of the major core proteins. A third possibility is that ΔGRL6 DCGs are synthesized at wild-type levels but are unstable. An elevated level of DCG loss, due to a high level of unstimulated exocytosis, would result in low steady-state accumulation. Consistent with the last model, and inconsistent with the first two, we found that the decrease in granule accumulation in ΔGRL6 cells was suppressed in the background of a second mutation that inhibits efficient transport of granules to the plasma membrane. We propose that unstimulated exocytosis of ΔGRL6 granules reflects a change, relative to wild-type, in the distribution of proteins in the secretory granule membrane. Grl6p may be distributed at the periphery of the wild-type granule lattice, in a position where its absence might affect, directly or indirectly, the recruitment or retention of transmembrane proteins. A change in membrane protein composition also may account for the superaccumulation of granules in MN173ΔGRL6 cells, but the extent and mechanism of granule turnover in MN173 cells are as yet unknown. The limited data on membrane proteins of dense core granules suggest that interactions with lumenal proteins may provide one important mechanism for sorting, but the specificity of those interactions is unclear (Wasmeier et al., 2002). Our in vivo data on defects in ΔGRL granules may suggest that, within this protein family, a subset of members may be adapted less for structural roles within the core lattice than for interactions with nonlattice proteins and may provide an interesting new perspective on secretory granule formation.
This article was published online ahead of print in MBC in Press ( http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E05–01–0028) on June 15, 2005.
† These authors contributed equally to this work.
Monitoring Editor: Randy Schekman
We thank members of this laboratory, including N. D. Chilcoat and J. Verbsky for contributions to analysis of ΔGRL strains; A. Sitikov for construction of antisense libraries; and N. Elde, N. Bradshaw, and M. Nasone for discussion and support. James Marvin and Ryan Duggan (Cancer Research Center Flow Cytometry Facility, University of Chicago) provided invaluable technical support and advice. J.J.E acknowledges helpful discussion with Ying Chen and Kevin Thornton. Lawrence Klobutcher (University of Connecticut, Storrs, CT) shared valuable ideas on antisense library construction. The NEO3 construct and the MTT1 promoter construct were generously provided by M. Gorovsky (University of Rochester, Rochester, NY), and the 4D11 and 5E9 hybridomas were a gift of Marlo Nelson and Joseph Frankel (University of Iowa). We thank Yimei Chen and The University of Chicago EM facility, and the Cancer Research Center, for electron microscopy. A.T.C and G.R.B were supported, respectively, by Training Grants GM-07281 and GM-07183. J.J.E. was supported by a National Science Foundation Graduate research fellowship. This work was supported by National Institutes of Health Grants GM-50946 and GM-59268 (to A.P.T.).