Genetic and Biochemical Interactions Among Yar1, Ltv1 and RpS3 Define Novel Links Between Environmental Stress and Ribosome Biogenesis in Saccharomyces cerevisiae

Construction of plasmids:

We constructed pYAR1cen (Ld25) by amplifying the intergenic region between YAR1 and HSP82 using a forward primer that anneals upstream of HSE1 in the HSP82 promoter and a reverse primer that anneals within the YAR1 coding region. The PCR product was cleaved with BamHI and NdeI and ligated into pRS314 (Sikorski and Hieter 1989) containing the YAR1 ORF (Ld3). Plasmid structure was confirmed by restriction analysis. Transformation of LY103 with Ld25 confirmed that the centromere vector is able to fully complement the cold-sensitive, heat-sensitive, and slow-growth phenotypes of the Δyar1 strain.

To create pLexA-YAR1 (Ld17), the YAR1 ORF was amplified by PCR, ligated into pCRII-TA (Invitrogen, San Diego), and then subloned into SalI/PstI-digested pLEXA (gift of Stan Hollenberg). Ld17 retains the first codon for methionine of the YAR1 ORF and adds two codons (for amino acids Pro and Asp) between the SalI site and the YAR1 ATG. The in-frame LexA-YAR1 fusion junction was confirmed by DNA sequencing. Ld20 was derived from Ld17 by cleaving with AflII and religating the vector to delete 1223 bp and the NdeI site from the ADE2 gene. To create pGEX-YAR1 (Ld19), we subcloned YAR1 from Ld17 as a SalI/NotI fragment into pGEX-6P-1 (Pharmacia). The presence of an in-frame fusion between GST and YAR1 and the absence of any amino acid substitutions in the YAR1 ORF were confirmed by DNA sequencing.

RPS3 and LTV1 ORFs were each cloned into pCITE2a (Novagen) for in vitro transcription/translation as follows. The LTV1 ORF was amplified from genomic W303 DNA using Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA) and PCR primers. The primers create an NcoI site that includes the initiator codon (Met) for the protein at the 5′ end and an AvaI site immediately upstream of the stop codon at the 3′ end. The PCR product was cleaved with NcoI and AvaI and ligated into pCITE2a cleaved with the same enzymes. This creates an S2A substitution at position 2 in the LTV1 sequence. The sequence of Ltv1 is otherwise unchanged except for the addition of LEHHHHHH at the C terminus. RPS3 was cloned into pCITE2a similarly, except pSK1 (gift of M. Finken-Eigen) was used as a template for PCR instead of genomic DNA, and the forward primer creates an NcoI site without altering the amino acid sequence of the protein. Thus the RPS3 sequence is fully wild type except for the addition of LEHHHHHH at the C terminus. The sequence of all constructs was confirmed to be wild type, except for the changes expected, by DNA sequencing.

Site-directed mutagenesis:

We mutated both K91 and K125 to alanine using a QuikChange site-directed mutagenesis kit following the protocol provided by the manufacturer (Stratagene). The mutagenic primer sets used to generate the K91A mutation were 5′-gttaatgaggtgaatgcaacaggcaacacggctttacattgggc-3′ and its complement, and for K125A, 5′-gcagacccctttattagaaacgcgttcggccacgatgc-3′ and its complement. Engineered BsmI and MluI restriction sites are italicized. The presence of the intended mutations and the absence of other mutations were confirmed by DNA sequencing.

Two-hybrid screen:

A two-hybrid screen for Yar1-interacting proteins was performed in yeast strain L40 harboring the bait plasmid pLexA-YAR1 (LD17). L40 pLexA-YAR1 was transformed with a yeast cDNA library fused to the Gal4 activation domain in pGAD24 (gift of S. Elledge). We screened 3.8 × 10⁵ yeast Trp⁺ transformants for lacZ expression using standard protocols (Vojtek and Hollenberg 1995) and picked 175 positives, of which 162 retested positive in a second assay of purified colonies. All of these were identified to be one of three genes by direct sequencing or by screening with RPS3-specific primers. For all three genes, we showed that loss of the bait plasmid correlated with loss of lacZ transcriptional activation and that each of the activating plasmids could again confer the His⁺Lac⁺ phenotype when reintroduced into cells containing the pLexA:YAR1 bait plasmid.

β-Galactosidase activity assays:

β-Galactosidase activity was assayed as described (Breeden and Nasmyth 1985) and normalized to protein levels using the following equation: 1000 × (OD₄₂₀/ml of supernatant)/protein concentration (mg/ml of supernatant) × reaction time (min). Individual single colonies were inoculated into 5 ml selective media and grown for 17 hr at 30° until the OD₆₆₀ of the culture was between 0.7 and 1.3.

In vitro interaction assays:

GST pulldown assays were performed essentially as described (Ausubel et al. 1995). BL21 cells, transformed with the pGEX-6p-1 plasmid (Pharmacia) or with pGEX-YAR1, were induced for 2 hr with 0.4 mm isopropyl-1-thio-b-d-galactopyranoside (Fisher). Harvested cell pellets were lysed by sonication and cleared cell lysates (2 mg/ml total protein) were incubated with glutathione-linked agarose beads (Sigma) at room temperature for 5 min. The ratio of lysate volume to bead volume was between 5:1 and 10:1 in all experiments. After binding, beads were washed five times with five bed volumes of bead binding buffer.

Potential Yar1 partner proteins (cloned in pCITE2a vectors) were synthesized in vitro using an STP3 transcription/translation kit (Novagen) and 40 μCi [³⁵S]methionine (Amersham Pharmacia) according to the manufacturer's specifications. Radiolabeled protein was incubated with GST or GST-YAR1-loaded glutathione-agarose beads in the presence of 1–2 mg of BL21 extract for 1 hr at 4°. GST and GST-YAR1 loaded beads were then washed three times and bound protein was eluted by boiling beads 5 min in 1× electrophoresis sample buffer. Eluted proteins were analyzed on 12% polyacrylamide gels. Proteins were visualized and quantified using phosphorimaging (Molecular Dynamics, Sunnyvale, CA).

Ribosome sedimentation and protein analysis:

Yeast cell lysates were prepared as previously described (Nelson et al. 1992) from strains grown in YPD or minimal media to an OD₆₀₀ of 0.4–0.7 at 30°. For polyribosome sedimentation, 10 OD₂₆₀ units of lysate were loaded onto 10 ml continuous 15–50% sucrose gradients in CB buffer (20 mm HEPES-K, pH 7.5, 1 mm EDTA, 5 mm MgCl₂, 10 mm KCl, 10% (v/v) glycerol, 2 mm β-mercaptoethanol). For ribosome subunit analysis, 10 OD₂₆₀ equivalents of cleared homogenate were loaded onto 15–50% sucrose gradients in CB buffer lacking MgCl₂ and including 10 mm EDTA. All gradients were centrifuged for 10 hr at 200,000 × g in an SW40 rotor (Beckman). The absorbance of gradient material at 254 nm was measured continuously and 0.5-ml fractions were harvested using an automated collector (ISCO). Where indicated, gradient fractions were precipitated with acetone and processed for SDS-PAGE and immunoblotting. After transfer to nitrocellulose membranes, individual proteins were detected using antibodies directed against the Myc and HA epitopes (provided by C. Lingenfelter) or antibodies directed against the yeast ribosomal proteins S2 and L32 (provided by J. Warner) at 1:2000 dilutions. Primary antibodies were detected with HRP-coupled secondary antibodies at a 1:20,000 dilution and chemiluminescence detection reagent (Perkin-Elmer Life Sciences).

The Yar1 ankyrin repeats are required for the two-hybrid interaction between Yar1 and Ltv1:

The co-crystal structures of a number of ankyrin-repeat proteins and their nonankyrin partners have identified variable residues at the tips of the β-hairpin loops of the ankyrin motif as frequent sites of specific protein:protein interaction (Gorina and Pavletich 1996; Brotherton et al. 1998; Russo et al. 1998). We aligned the amino acid sequence of Yar1 with the known structure of the CDK inhibitor, INK4p18. The optimal alignment of Yar1 is with the third and fourth ankyrin repeats of INK4p18 (Figure 2A). This alignment predicts that each β-hairpin loop in Yar1 has a positively charged lysine residue at the tip (highlighted in yellow).

Open in a separate window

Open in a separate window

Figure 2.—

The β-hairpin lysines in the ankyrin domain are critical for the Yar1:Ltv1 interaction. (A) The Cn3D program at NCBI was used to map the aligned (by gapped BLAST) amino acid sequence of the Yar1 ankyrin domain onto the 3D crystal structure of INK4p18. The program optimally aligned the Yar1 repeats with the third and fourth ankyrin repeats of INK4p18. In this representation, red represents identical residues; blue, nonconserved residues; and purple, conservative substitutions. The two positively charged lysines at the tip of each Yar1 loop have been highlighted in yellow. (B and C) Two-hybrid strains containing wild-type (WT Yar1) or mutant forms of Yar1 (K91A, K125A, or K91,125A), fused in frame with the LexA-DB and the Ltv1- or the RpS3-activation domain fusion plasmids, were grown at 30°. Three individual colonies were inoculated in each case and cell pellets were analyzed for β-galactosidase activity and protein content. Values shown represent the mean of three separate assays, with error bars representing 1 SD of the data.

We changed each of these lysines to alanine, both individually and in combination, by site-directed in vitro mutagenesis of YAR1 in the pLexA:YAR1 bait plasmid. These plasmids were then reintroduced into yeast carrying either RPS3 or LTV1 as activation-domain fusion plasmids. The resulting transformants were assayed for β-galactosidase activity as an indirect measure of their interaction. A representative experiment is shown in Figure 2, B and C. Mutation of lysine 91 reduces, and mutation of both lysines, eliminates the interaction between Yar1 and Ltv1 (Figure 2B). These data indicate that the tips of the ankyrin repeats are critical for the Yar1:Ltv1 two-hybrid interaction. Mutation of both lysines at the β-hairpin tips has essentially no effect on the interaction between Yar1 and RpS3 (Figure 2C). This suggests that Yar1 interacts with RpS3 and Ltv1 through different surfaces of the protein.

Stress-sensitive phenotypes of Δyar1 and Δltv1 mutants:

We previously reported that Δyar1 strains are viable but slow growing under either heat-shock (37°) or low temperature (25°) growth conditions (Lycan et al. 1996). As incubation at such temperatures is stressful for yeast, we asked whether Δyar1 cells exhibited any other stress-sensitive phenotypes. Specifically, we tested Δyar1 cells for sensitivity to oxidative stress, high osmolarity, and nutrient and amino acid starvation. Cells lacking YAR1 were sensitive to oxidative stress induced by treatment with diamide, at least as sensitive as cells lacking YAP1, the major oxidative stress-activated transcription factor (Jamieson 1998; see Figure 3A). Diamide acts to deplete glutathione pools and oxidizes thiol groups (Kosower and Kosower 1987). The Δyar1 strain was also sensitive to osmotic stress induced by sorbitol, although not as sensitive as the Δpbs2 mutant. PBS2 encodes the MAPKK that phosphorylates Hog1 in response to high-salt or sorbitol growth conditions (Hohmann 2002). Thus YAR1 is required for optimal long-term growth under both oxidative and osmotic stress conditions. However, the Δyar1 mutant showed no defects in viability after prolonged nitrogen deprivation (9 days). Mutant cells also accumulated storage carbohydrates comparable to wild-type cells, and the sporulation efficiency was essentially the same as in wild-type controls (data not shown). We also tested the phenotype of Δyar1 mutants constructed in the S288c background, in addition to our W303 background. Δyar1 cells exhibit the same phenotypes in both strain backgrounds (Figure 3B).

Open in a separate window

Figure 3.—

Δyar1 strains and Δltv1 cells exhibit stress-sensitive phenotypes. (A) Wild-type parental strain W303 and isogenic deletion derivatives Δyar1::URA3 and Δpbs::LEU2 were grown to midlog phase in YPD media, as were parental strain SEY6210 and its isogenic deletion derivative, Δyap1-Δ2. Cell density was determined by hemocytometer and 5-μl aliquots containing 16,000 (left) and 4000 (right) cells were spotted onto YPD or YPD plus the indicated drugs. In this experiment, 1.3 times more Δyar1 and Δpbs2 cells were spotted at each position compared to the other strains to compensate for the slightly slower growth rate of these mutants under nonstress conditions. Plates were incubated at 30° and photographed after 2 days. The W303 strain background is more sensitive than the SEY6210 strain to diamide. (B) Parental strain LY134 (derived from S288c) and isogenic deletion derivatives LY135 and LY136 were grown as in A. Equal numbers of wild-type and mutant cells were spotted at each position. All plates were photographed after 2 days at 30° except the 18° plate, which was photographed after 3 days of incubation.

If Yar1 and Ltv1 physically interact in vivo, then we would predict that strains lacking either protein might share one or more phenotypes. We therefore asked whether strains lacking LTV1 shared any of the stress-sensitive phenotypes of Δyar1 cells. We found that the Δltv1 strain is hypersensitive to both osmotic and oxidative stress and is especially cold sensitive (Figure 3B). However, while LTV1 stands for low-temperature viability, we find the Δltv1 strain is slow growing but viable at all of the low temperatures that we tested (18°, 15°, 12°, and 10°; doubling time at 18° is ∼7.5 hr). The Δltv1 strain, like the Δyar1 strain, has a slight slow-growth phenotype at 30°. The fact that Δyar1 and Δltv1 cells share multiple genetic phenotypes (as well as antibiotic hypersensitivity; see below) argues that the two-hybrid interaction between these two proteins reflects a common biological function in vivo.

Genetic interaction between YAR1 and RPS3:

Eukaryotic RpS3 is a highly conserved 36-kD protein located on the solvent side of the 40S subunit on the beak of the head region (Spahn et al. 2001). Multiple functions have been proposed for this small protein. Mammalian RpS3 can be crosslinked to eIF3 (Tolan et al. 1983) and eIF2 (Westermann et al. 1979), the basis for its proposed role in translation initiation (Westermann et al. 1981). In yeast, the missense allele of RPS3, suf14-1, is resistant to aminoglycosides and acts as an extragenic suppressor of +1 frameshift mutations, implicating RpS3 in translational decoding (Hendrick et al. 2001). Finally, in vitro assays have implicated mammalian and Drosophila RpS3 in the repair of oxidative/UV DNA damage (Kim et al. 1995; Yacoub et al. 1996; Deutsch et al. 1997; Sandigursky et al. 1997), and yeast RpS3 has been shown to have an endonuclease activity on apurinic DNA in vitro (Jung et al. 2001).

To test the biological function of the detected two-hybrid and in vitro interactions between RpS3 and Yar1, we turned to genetic analysis. One way to uncover a genetic interaction, when one of the two genes is essential, is to reduce the dosage of the essential protein in a background in which the nonessential protein is eliminated. We constructed a diploid strain, homozygous for Δyar1 and heterozygous for Δrps3, and tested it for a synthetic growth defect under various stress conditions. The heterozygous Δrps3/+ strain was indistinguishable from the wild-type parent under both optimal and all stress growth conditions that we tested (Figure 4a). The Δyar1/Δyar1 diploid is slow growing at 23° and at 37°, but is less affected by temperature than the haploid strain. However, the Δyar1/Δyar1, Δrps3/+ strain grew more slowly than either the Δyar1/Δyar1 diploid or the ΔrpS3/+ strain at all temperatures, but especially at 23° and 37° (Figure 4a). In addition, Δyar1/Δyar1, ΔrpS3/+ is more sensitive to osmotic stress and oxidative stress than either the Δyar1/Δyar1 strain or the ΔrpS3/+ strain (Figure 4b). The synthetic phenotype produced when the RPS3 dosage is reduced in the Δyar1/Δyar1 background is evidence of a genetic interaction between YAR1 and RPS3. These data are consistent with the evidence of a physical interaction between the two proteins in the two-hybrid screen and in vitro.

Open in a separate window

Figure 4.—

Genetic interaction between YAR1 and RPS3. (a) Diploid strains carrying the mutations indicated below were streaked for single colonies on YPD plates and incubated at 23°, 30°, or 37°. Homozygous Δyar1 diploids are not as sensitive to temperature stress as are the haploid strains. (A) +/+, Δyar1/Δyar1, (B) Δrps3/+, Δyar1/Δyar1, (C) +/+, +/+, and (D) Δrps3/+, +/+. (b) Diploid strains bearing the indicated mutations were grown as in Figure 1 and 16,000, 4000, and 1000 cells were spotted (left to right) onto YPD, YPD + diamide, and YPD + sorbitol plates. To compensate for the slower growth rate of the two homozygous Δyar1/Δyar1 strains on YPD, 1.3 times more cells were spotted at each position for these two strains relative to the others. Plates were incubated at 30° and photographed after 25 hr of incubation.

YAR1 and LTV1 mutants are hypersensitive to protein synthesis inhibitors:

Identification of RpS3 and Ltv1 as interacting partners for Yar1 led us to look for further evidence that might link Yar1 to ribosome structure or function in vivo. Strains with mutations in genes involved in ribosome function or biogenesis often exhibit an altered sensitivity to drugs that affect protein synthesis (Lee et al. 1992; Nelson et al. 1992; Kressler et al. 1997; Yan et al. 1998; Liu and Thiele 2001). We tested four inhibitors of translational elongation. Aminoglycosides like paromomycin and neomycin bind to the A-site decoding region of the small subunit reducing translational accuracy (Carter et al. 2000; Ryu do et al. 2002; Lynch et al. 2003). Anisomycin binds the 28S rRNA in the 60S subunit and inhibits the peptidyl transferase reaction (Rodriguez-Fonseca et al. 1995). Cycloheximide inhibits translation by abrogating the translocation of the peptidyl tRNA from the A-site to the P-site (Pestka 1971). Cells lacking YAR1 are hypersensitive to anisomycin, but are no more sensitive to cycloheximide than are wild-type cells (Figure 5B). They are also hypersensitive to both of the aminoglycosides tested, neomycin and paromomycin (Figure 5A). Strains lacking LTV1 were sensitive to the same spectrum of protein synthesis inhibitors as are Δyar1 cells (Figure 5). These data further link Yar1 and Ltv1 function and suggest that both proteins have a role in ribosome biogenesis or function.

Open in a separate window

Figure 5.—

Both Δyar1 and Δltv1 strains are hypersensitive to protein synthesis inhibitors. The sensitivity of wild-type (WT) and isogenic Δyar1 and Δltv1 cells to various protein synthesis inhibitors was tested by growing cells as described in Figure 1. (A) A total of 16,000, 4000, or 1000 cells (left to right) were spotted onto YPD plates with or without the denoted drugs. Only the 1000-cell column is shown for the neomycin (neo; 1 mg/ml) and paromomycin (paro; 1 mg/ml) plates. (B) A total of 400 cells of each strain were spotted onto YPD, YPD + cycloheximide (CHX; 1μg/ml), or YPD + anisomycin (Anis; 7.5 μg/ml) plates. Equal numbers of wild-type and mutant cells were spotted at each dilution. Plates were incubated at 30° unless otherwise indicated.

Yar1 and Ltv1 are required for normal ribosome biogenesis:

Both biochemical interactions and mutant phenotypes link Yar1 and Ltv1 to RpS3 and to the ribosome. To test whether ribosome biogenesis is impaired in mutants lacking either of these proteins, we profiled cellular ribosomes from wild-type, Δyar1, and Δltv1 strains on sucrose density gradients. As shown in Figure 6, cells lacking YAR1 or LTV1 have aberrant polysome profiles relative to wild-type cells. Both mutants have a significantly enlarged 60S peak relative to isogenic wild-type cells. In addition, the total number of 40S subunits, determined by dissociating monosomes and polyribosomes into free subunits, is significantly diminished in the mutants; the 40S:60S ratio is reduced by 52% in Δyar1 and by 58% in Δltv1 (see insets in Figure 6). The reduction in total 40S subunits in both mutants is consistent with the excess of free 60S subunits in the polysome profile, and both pieces of data are consistent with a defect in 40S subunit biogenesis/maturation in both Δyar1 and Δltv1 mutants.

Open in a separate window

Figure 6.—

Strains lacking YAR1 and LTV1 have altered polysome profiles. Strains containing chromosomal deletions of YAR1 (LY135) and LTV1 (LY136) and their isogenic wild-type parental strains were grown at 30° in rich medium and then fractionated by mechanical disruption and centrifugation. The postnuclear cytosolic homogenate was separated on continuous sucrose gradients either containing 5 mm MgCl₂ (full-size graphs) or lacking Mg²⁺ and containing 10 mm EDTA (insets). An RNA absorbance profile was determined using a continuous flow UV detector and is represented here in relative units. Only the portion of the gradient containing small subunits (40S), large subunits (60S), monomer ribosomes (80S), and small polysomes is shown. The arrow denotes the shoulder peak. The shaded line in the insets denotes the size of the wild-type small subunit peak.

One other aspect of the polysome profiles of Δyar1 and Δltv1 mutants is worth noting. In both mutant profiles, there is a distinct shoulder on the 80S peak. This shoulder peak disappears if extracts are centrifuged under conditions that dissociate monosomes and polysomes into free 40S and 60S subunits. Similar shoulder peaks, called “half-mer” peaks, have been noted in numerous 60S subunit biogenesis mutants and in mutants with defects in subunit joining (Woolford and Warner 1991; Kressler et al. 1999) and are composed, in these mutants, of 80S monosomes plus a stalled 43S preinitiation complex attached to the same mRNA. However, in 60S mutants, there are also “half-mer” peaks associated with each of the polysome peaks, which we did not observe. In addition, gel analysis of RNA extracted from the shoulder peak fraction reveals that it is enriched for large subunit rRNA (the 25S:18S ratio in the shoulder peak is 2.8 compared to a ratio of 1.7 in the 80S peak). This ratio suggests an excess of 60S subunits, not an excess of small subunits. Since both the Δyar1 and Δltv1 mutants have a deficiency in 40S subunits and an excess of free 60S subunits, the shoulder peak in these mutants is most likely dimers of free 60S subunits that form in the absence of sufficient 40S partners.

Ltv1 comigrates with ribosome subunits:

The observed defects in 40S subunit biogenesis in Δyar1 and Δltv1 mutants suggested that the interaction among Yar1 and Ltv1 and/or RpS3 might occur in the process of ribosome assembly. If so, we might expect Yar1 or Ltv1 to cosediment with ribosomes or preribosomal particles in vivo. To test this hypothesis, we epitope tagged each protein by integration of the epitope sequence at the appropriate chromosomal locus in different strains. Correct integration and synthesis of the fusion protein were confirmed in each strain by PCR and Western blot analysis. To generate a single strain (LY161) expressing both tagged proteins, we crossed the Yar1-HA strain to the Ltv1-Myc strain, sporulated the diploid, and selected for haploids with both fusion genes. We assessed the ability of each fusion protein to replace the function of the wild-type protein by testing these strains for slow growth and for stress sensitivity relative to wild-type and Δyar1 or Δltv1 strains. The Yar1-HA and the Ltv1-Myc strains, as well as the LY161 strain with both tagged proteins, grew as well as wild-type cells at 30° and on plates containing sorbitol, indicating that the tag does not interfere with the function of either protein (data not shown).

The association of Yar1 and/or Ltv1 with ribosomal particles was assessed by sucrose gradient centrifugation and Western blotting (polysome blots) of extracts of the Yar1-HA Ltv1-Myc (LY161) strain. When whole-cell lysates are prepared under low-salt conditions, most of the Yar1-HA is present in the cytosolic fractions, although some Yar1 is also detected in denser fractions of the gradient in the 40S region (Figure 7). The opposite is true of Ltv1; while some Ltv1-Myc is detected at the top of the gradient, the majority comigrates with 40S ribosomal subunits (Figure 7). Intriguingly, Ltv1 does not comigrate with polysomes. This is in contrast to RpS2, which, as expected for a structural component of the mature small subunit, comigrates with both 40S particles and with translating ribosomes. These data suggest that Ltv1 associates with the 40S subunit, but is released some time prior to its incorporation into a translating ribosome.

Open in a separate window

Figure 7.—

Association of Yar1p and Ltv1p with ribosomal subunits. Wild-type cells expressing epitope-tagged Yar1 and Ltv1 fusion proteins (LY161) were cultured at 30° and fractionated to yield a postnuclear cytosolic homogenate. This was separated on a continuous sucrose gradient as in Figure 6. Individual gradient fractions were processed for SDS gel electrophoresis and Western blotting with antibodies to small (S2) and large (L3) ribosomal subunit proteins and the myc and hemagglutinin (HA) epitopes.

We acknowledge the following Lewis and Clark College alumni for their contributions to this project as undergraduates: Kimberley Stafford for the construction of pLEXA::YAR1 and Yuji Mishina for construction of the K91A YAR1 mutant. We thank B. Baxter for construction of the pCITE fusion vectors. We are grateful to Jim Posakony and Greg Hermann for reading the manuscript and for very helpful discussions. This work was supported by National Institutes of Health (NIH) grant GM-061643 to D.E.L. and by grant 97120:JVZ:02/19/98 from the M. J. Murdock Charitable Trust to D.E.L. In addition, R.M.S. was supported by NIH grant GM-068208.