Discrimination between mono- and trimethylated cap structures by two isoforms of Caenorhabditis elegans eIF4E

Abstract

Primitive eukaryotes like Caenorhabditis elegans produce mRNAs capped with either m⁷GTP or m₃^2,2,7GTP. Caenorhabditis elegans also expresses five isoforms of the cap-binding protein eIF4E. Some isoforms (e.g. IFE-3) bind to m⁷GTP–Sepharose exclusively, whereas others (e.g. IFE-5) bind to both m⁷GTP– and m₃^2,2,7GTP–Sepharose. To examine specificity differences, we devised molecular models of the tertiary structures of IFE-3 and IFE-5, based on the known structure of mouse eIF4E-1. We then substituted amino acid sequences of IFE-5 with homologous sequences from IFE-3. As few as two changes (N64Y/V65L) converted the cap specificity of IFE-5 to essentially that of IFE-3. Molecular dynamics simulations suggested that the width and depth of the cap-binding cavity were larger in IFE-5 than in IFE-3 or the N64Y/V65L variant, supporting a model in which IFE-3 discriminates against m₃^2,2,7GTP by steric hindrance. Furthermore, the affinity of IFE-5 (but not IFE-3) for m₃^2,2,7GTP was reversibly increased when thiol reagents were removed. This was correlated with the formation of a disulfide bond between Cys-122 and Cys-126. Thus, translation of m₃^2,2,7GTP-capped mRNAs may be regulated by intracellular redox state.

Introduction

Initiation of protein synthesis proceeds by progressive assembly of initiation complexes, each stage catalyzed by a different set of initiation factors (Hershey and Merrick, 2000). Recruitment of mRNA to the 43S initiation complex in eukaryotes requires the eIF4 factors, which include the 25 kDa cap-binding protein eIF4E. Alterations in eIF4E levels and activity have a profound effect on cell growth and phenotype, presumably due to the differential recruitment of mRNAs specifically required for cell growth and cell cycle progression (De Benedetti and Harris, 1999).

The tertiary structure of mouse eIF4E-1 was solved by X-ray crystallography (Marcotrigiano et al., 1997) and that of yeast eIF4E by NMR spectroscopy (Matsuo et al., 1997). The binding of m⁷GDP results from the stacking of the purine ring between Trp-56 and Trp-102 (using the mouse numbering), formation of H-bonds between N1 and N2 of m⁷G and Glu-103, van der Waals interactions between the ribose moiety and Trp-56, and ionic interactions of the α- and β-phosphate oxygen atoms with Arg-157 and Lys-162. Addition of one methyl group at the N2 position of m⁷G has little effect on cap binding to mammalian eIF4E, but addition of a second methyl group, to form 2,2,7-trimethylguanine (m₃^2,2,7G), markedly decreases binding (Darzynkiewicz et al., 1988; Carberry et al., 1990; Cai et al., 1999).

Two types of capped RNAs exist in eukaryotic cells. The primary transcripts for both mRNAs and small nuclear RNAs (snRNAs) are modified to form a 5′–5′ GpppN linkage, which is subsequently methylated in the nucleus to yield m⁷GpppN (Varani, 1997). The cap of snRNAs is further methylated in the cytosol at N2 of m⁷G, forming m₃^2,2,7GpppN (Mattaj, 1986). These methylations are dependent upon the binding of Sm proteins to form U-type small nuclear ribonuclear protein complexes (snRNPs) and are the signal for import of snRNPs into the nucleus (Hamm et al., 1990; Gorlich and Mattaj, 1996). The trimethylated cap structure is recognized by Snurportin1, a receptor for spliceosomal mRNPs that utilizes the importin β pathway for nuclear import (Huber et al., 1998). The strong preference of most eIF4Es for m⁷G- versus m₃^2,2,7G-containing caps ensures that mRNAs rather than snRNAs are recruited to the translational machinery.

An exception to this paradigm is the acquisition of m₃^2,2,7G-containing caps by mRNAs through the process of trans-splicing, in which a 22 nucleotide spliced leader is transferred from the 5′-end of an snRNA to an acceptor site in the 5′ end of the pre-mRNA (Blumenthal, 1998). As a result, the original m⁷G-containing cap is replaced by the m₃^2,2,7G-containing cap (van Doren and Hirsh, 1990). Trans-splicing of mRNA has been most studied in primitive eukaryotes like the nematode Caenorhabditis elegans (Zorio et al., 1994), but it also occurs in more complex chordate species (Vandenberghe et al., 2001). In C.elegans, the majority of mRNAs are trans-spliced. Both m⁷G- and m₃^2,2,7G-capped mRNAs are found in the cytosol of C.elegans and are translated on polyribosomes (Liou and Blumenthal, 1990).

The fact that the well-studied eIF4Es of higher eukaryotes strongly prefer m⁷G- to m₃^2,2,7G-capped mRNAs, yet C.elegans translates both types of mRNAs, led us to examine the eIF4E of C.elegans (Jankowska-Anyszka et al., 1998; Keiper et al., 2000). Surprisingly, five eIF4E isoforms are expressed in C.elegans, named IFE-1 to IFE-5. IFE-3 is the most homologous to mammalian eIF4E-1, is retained on m⁷GTP–Sepharose but not m₃^2,2,7GTP–Sepharose, and is the only IFE that is essential for viability. IFE-4 has the same cap specificity, but is not essential. By contrast, two isoforms with closely related amino acid sequences, IFE-1 and IFE-5, are retained on both m⁷GTP–Sepharose and m₃^2,2,7GTP–Sepharose. IFE-2, while not retained on m₃^2,2,7GTP–Sepharose, is nonetheless prevented from binding to m⁷GTP–Sepharose by m₃^2,2,7GTP. IFE-1, -2 and -5 are partially redundant for viability, but at least one isoform is required. IFE-1 is expressed only in the gonad of maturing worms and plays an essential role for spermatogenesis in both males and hermaphrodites (Amiri et al., 2001).

In the present study, we have examined the structural basis for differences in cap discrimination among the IFE proteins. Such knowledge may give insight into the biological roles of the five IFE proteins of C.elegans, as well as the multiple eIF4E isoforms found in other organisms (Browning et al., 1987; Rychlik et al., 1987; Wakiyama et al., 1995; Gao et al., 1998; Myers et al., 2000). We find that changing just two amino acid residues can alter the cap discrimination of IFE-5 to resemble that of IFE-3. Based on dynamic tertiary structure models, we propose that this change results from a reduction in the average size of the cap-binding cavity. Un expectedly, the cap discrimination of IFE-5 is reversibly changed by the formation of a disulfide bond between two Cys residues.

Results

Rationale and approach for determining the basis for cap discrimination

It is possible that a small subset of amino acid residues is responsible for the difference in cap specificity of the two classes of C.elegans eIF4E isoforms. Our broad strategy was to choose a representative of each eIF4E class, IFE-3 and IFE-5, interchange amino acid residues, and examine the specificity of cap recognition. One way to choose candidates for discriminatory amino acid residues is to examine regions that are likely to be in close proximity to the bound cap structure. Unfortunately, the tertiary structure has not been solved for any of the IFE proteins. However, the co-crystal structure of mouse eIF4E with m⁷GDP has been solved and can serve as a template for model building of the IFE proteins.

Amino acid sequence alignment shows that mouse eIF4E is more similar to IFE-3 than to IFE-5 (Figure 1), in agreement with the calculated sequence identities (47% versus 41%) and the strong preference of both mammalian eIF4E and IFE-3 for m⁷G-containing caps (Jankowska-Anyszka et al., 1998; Cai et al., 1999). Importantly, there is a high degree of sequence identity between mouse eIF4E and both IFE-3 and IFE-5 in the β-sheets identified by X-ray crystallography (s1, s3, etc.). We approximated the tertiary structures for IFE-3 and IFE-5 by homology modeling to the mouse eIF4E structure (Dwyer, 1996, 2001; Figure 2). After energy minimization, the root-mean-square deviations are 2.2 and 1.1 Å, respectively. The residues homologous to Trp-56 and Trp-102 of mouse eIF4E, which ‘sandwich’ the guanine ring, are Trp-51 and Trp-97 in IFE-3, and Trp-28 and Trp-74 in IFE-5.

Fig. 1. Sequence comparisons between C.elegans IFE-3, mouse eIF4E and C.elegans IFE-5. Amino acid sequences were deduced from the cDNA sequences of IFE-3 (Jankowska-Anyszka et al., 1998), mouse eIF4E (Altmann et al., 1989) and IFE-5 (Keiper et al., 2000). Alignment was performed using the CLUSTAL W algorithm (version 1.8; http://www.ddbj.nig.ac.jp). Shading indicates identical residues. Secondary structure elements are designated as follows: S1, β-sheet 1; S2, β-sheet 2; H1, α-helix 1; H2, α-helix 2; etc. Amino acid residues that contact the cap in mouse eIF4E are shown in bold.

Fig. 2. Molecular models of IFE-3, IFE-5 and the NV-YL variant of IFE-5 in comparison with the crystal structure of mouse eIF4E. The figure was produced with the graphics interface of the Insight II software. N-terminal amino acid residues homologous to those truncated from mouse eIF4E prior to X-ray crystallography are not shown. The position of m⁷GDP is shown in the mouse eIF4E structure. The structures of IFE-3, IFE-5 and IFE-5(NV-YL) were obtained from equivalent time points in MD simulations and are intended to show dynamic aspects of the binding site rather than the minimized structures. Selected amino acid side chains are shown in different colors and labeled in upper case letters using the one-letter code. Secondary structure elements (β-sheets in yellow, loops in dark blue, α-helices in light blue) are labeled in lower case letters, e.g. β-sheet 3 is ‘s3’, etc. Loops connecting β-sheets are labeled 1–2 loop, etc. IFE structures are oriented the same as mouse eIF4E.

The similarity of the IFE-3 and IFE-5 models (Figure 2), coupled with a knowledge of where the amino acid sequences deviate (Figure 1), allowed us to choose amino acid sequences that were likely to be involved in cap discrimination. The loop connecting β-sheets 1 and 2, referred to as the 1–2 loop, forms one side of the cap-binding pocket and could contribute side chains that determine specificity. The 3–4 loop, connecting β-sheets 3 and 4, forms the other side of the cap-binding pocket. It also contains a Glu residue homologous to Glu-103 in mouse eIF4E, which forms H-bonds with the N1 and N2 protons of m⁷G; these protons are absent in m₃^2,2,7G. The loops are, therefore, attractive sites for modification.

Replacement of amino acid sequences in IFE-5 with homologous sequences from IFE-3

In variant 1–2 loop, the 1–2 loop in IFE-5 is replaced by the equivalent sequence from IFE-3 (Figure 3). This loop contains one of the two Trp residues sandwiching the cap guanine. In variant 3–4 up, the upstream (N-terminal) half of the 3–4 loop is replaced together with part of β-sheet 3, whereas in variant 3–4 down, the downstream (C-terminal) half of this same loop is replaced. This stretch of amino acids includes the other Trp residue sandwiching the m⁷G. In other variants, the entire 3–4 loop is replaced (3–4 loop), or various types of 1–2 and 3–4 loop substitutions are combined (1–2 & 3–4 up, 1–2 & 3–4 down, and 1–2 & 3–4 loop). Finally, the upstream portion of the 3–4 loop is dissected with more selective substitutions (variants N-Y, V-L, NV-YL, DDIQPK-EGIKPM and QPK-KPM).

Fig. 3. Amino acid substitutions in IFE-5. Variants, named on the left, were constructed by substituting the indicated sequences of IFE-3 into the homologous positions of IFE-5. Letters in bold indicate amino acid residues that are conserved between IFE-3 and IFE-5.

Qualitative testing of variants using affinity chromatography

IFE-5 variants were initially screened qualitatively for cap specificity by measuring their capture from Escherichia coli lysates on columns of either m⁷GTP– or m₃^2,2,7GTP– Sepharose, followed by elution with m⁷GTP or m₃^2,2,7GTP, respectively (Figure 4). In this qualitative assay, it is not possible to distinguish between proteins that are more highly expressed in soluble form and those with enhanced cap-binding activity; this distinction is made by measuring cap affinities of purified IFEs using the fluorescence-quenching assay (see below). Rather, the qualitative assays served to guide the construction of IFE-5 variants that showed a change in the relative binding to m⁷GTP–Sepharose versus m₃^2,2,7GTP–Sepharose compared with wild-type (wt) IFE-5.

Fig. 4. Qualitative testing of IFE-5 sequence variants using affinity chromatography. Escherichia coli was transfected with the indicated expression vectors, and extracts were applied to columns of m⁷GTP–Sepharose (M) or m₃^2,2,7GTP–Sepharose (T). Proteins were eluted with m⁷GTP or m₃^2,2,7GTP, respectively, and analyzed by SDS–PAGE, with Coomassie Blue staining.

As reported previously (Jankowska-Anyszka et al., 1998), IFE-3 is retained on m⁷GTP–Sepharose (Figure 4A, lane M under IFE-3) but not m₃^2,2,7GTP–Sepharose (lane T). This indicates high selectivity for m⁷G-containing caps, i.e. a ‘mono-specific’ IFE. By contrast, IFE-5 is retained both by m⁷GTP–Sepharose (Figure 4A, lane M under IFE-5) and, to a lesser extent, by m₃^2,2,7GTP– Sepharose (lane T). This is the behavior of a ‘dual-specific’ IFE.

Comparing wt IFE-5 to several variants for relative binding to m⁷GTP–Sepharose versus m₃^2,2,7GTP– Sepharose, variant 3–4 up appears to show the highest selectivity for m⁷GTP (Figure 4A). The variant 3–4 down, on the other hand, binds m₃^2,2,7GTP–Sepharose more efficiently than wt IFE-5, despite similar binding to m⁷GTP–Sepharose. Thus, it displays lower selectivity for m⁷GTP than IFE-5. These results suggest that a discriminatory element (favoring m⁷GTP over m₃^2,2,7GTP) exists in the upstream but not the downstream portion of the 3–4 loop.

We next explored combinations of 1–2 loop sequences with 3–4 loop sequences. Variant 1–2 & 3–4 down is more selective for m⁷GTP than the corresponding variant 3–4 down (Figure 4A). Variant 1–2 & 3–4 up (Figure 4B), on the other hand, is less selective than the corresponding variant 3–4 up. Similarly, variant 1–2 & 3–4 loop is less selective than variant 3–4 loop. In all three cases, the changes in cap selectivity caused by substitutions in the 3–4 loop are attenuated by substitutions in the 1–2 loop, causing the proteins to behave more like wt IFE-5.

Based on these findings, we sought the specific amino acid residues in the upper 3–4 loop sequence responsible for increased m⁷GTP selectivity. Of several single and double amino acid substitutions, the variant NV-YL produces the highest selectivity: binding to m₃^2,2,7GTP– Sepharose is nearly undetectable despite robust binding to m⁷GTP–Sepharose (Figure 4C and D). The combination of changing both Asn-64 and Val-65, rather than either alone, confers the maximum change in m⁷GTP selectivity.

We explored the remaining portion of the 3–4 up region with the variant DDIQPK-EGIKPM (Figure 4E). Despite the fact that binding to both affinity resins is lower than for wt IFE-5, the ratio of binding appears similar to that of wt IFE-5, suggesting no change in selectivity. By contrast, the variant QPK-KPM displays increased binding to m₃^2,2,7GTP–Sepharose, but the same or even less binding to m⁷GTP–Sepharose than IFE-5 (Figure 4F). This indicates that QPK-KPM is even more permissive than IFE-5 for m₃^2,2,7GTP binding.

A consistent picture emerges from these observations. The 1–2 loop region alone does not affect m⁷GTP selectivity. By contrast, the 3–4 loop region is responsible for m⁷GTP selectivity, but changes in selectivity are ‘damped’ in variants combining 1–2 loop and 3–4 loop substitutions. Within the 3–4 loop region, the N-terminal portion (3–4 up) increases selectivity, while the C-terminal portion decreases selectivity. Within the 3–4 up region, simultaneously replacing Asn-64 with Tyr and Val-65 with Leu confers the greatest selectivity, indistinguishable from IFE-3 by this assay. The individual substitutions at positions 64 or 65 do not confer this degree of enhanced selectivity, nor do substitutions in the C-terminal portion of the 3–4 up region. In fact, replacement of the C-terminal one-third of the 3–4 up region has the opposite effect, lowering selectivity. This may account for the observation that exchanging ⁶⁴NV⁶⁵ (variant NV-YL) increases selectivity more than exchanging ⁶⁴NVFRDDIQPK⁷³ (variant 3–4 up).

Quantitative testing of 1–2 and 3–4 loop IFE-5 variants by fluorescence quenching

To obtain quantitative binding data, we employed the fluorescence-quenching assay with homogeneous IFE proteins. The dissociation constants of the cap analog· protein complexes, K_D, were calculated for both m⁷GTP and m₃^2,2,7GTP titrations (Table I, Experiment 1). The binding of m⁷GTP to IFE-3 is 1.3-fold stronger than to IFE-5 (K_D = 0.38 versus 0.50 µM). By contrast, the binding of m₃^2,2,7GTP is 2.3-fold stronger to IFE-5 than to IFE-3 (K_D = 1.6 versus 3.5 µM). The combined effects of these nucleotide-binding differences is that IFE-3 discriminates in favor of m⁷GTP versus m₃^2,2,7GTP caps by a factor of 9.2, whereas IFE-5 discriminates by a factor of only 3.2 (Table I, K_D ratio).

Table I. Dissociation constants (K_D) for m⁷GTP and m₃^2,2,7GTP from various IFE proteins.

Proteina	Titration bufferb	KD (µM)c		KD ratiod
		m7GTP	m32,2,7GTP
Experiment 1
IFE-3	A1	0.38 ± 0.03	3.5 ± 0.1	9.2 ± 0.8
IFE-5	A1	0.50 ± 0.06	1.6 ± 0.1	3.2 ± 0.6
1–2 loop	A1	0.35 ± 0.02	1.5 ± 0.1	4.3 ± 0.5
3–4 up	A1	0.43 ± 0.03	1.9 ± 0.1	4.4 ± 0.5
3–4 down	A1	0.41 ± 0.03	0.81 ± 0.02	2.0 ± 0.2
3–4 loop	A1	0.46 ± 0.03	1.9 ± 0.1	4.1 ± 0.5
1–2 & 3–4 up	A1	0.55 ± 0.05	1.5 ± 0.1	2.7 ± 0.4
1–2 & 3–4 down	A1	0.46 ± 0.02	1.4 ± 0.1	3.0 ± 0.3
1–2 & 3–4 loop	A1	0.42 ± 0.03	1.1 ± 0.1	2.6 ± 0.4
N-Y	A1	0.41 ± 0.02	1.4 ± 0.1	3.4 ± 0.4
V-L	A1	0.43 ± 0.04	2.3 ± 0.1	5.3 ± 0.7
NV-YL	A1	0.38 ± 0.03	2.6 ± 0.1	6.8 ± 0.8
DDIQPK-EGIKPM	A1	0.51 ± 0.04	1.6 ± 0.1	3.1 ± 0.4
QPK-KPM	A1	0.45 ± 0.03	0.85 ± 0.04	1.9 ± 0.2
Experiment 2
IFE-3	A0	0.36 ± 0.03	3.5 ± 0.1	9.7 ± 1.1
IFE-5	A0	0.54 ± 0.07	0.92 ± 0.08	1.7 ± 0.4
3–4 up	A0	0.47 ± 0.03	0.94 ± 0.03	2.0 ± 0.2
3–4 loop	A0	0.45 ± 0.04	0.95 ± 0.04	2.1 ± 0.3
NV-YL	A0	0.68 ± 0.03	1.2 ± 0.1	1.8 ± 0.2
QPK-KPM	A0	0.57 ± 0.03	1.0 ± 0.1	1.8 ± 0.3
IFE-3	A10	0.39 ± 0.03	3.5 ± 0.1	9.0 ± 0.9
IFE-5	A10	0.48 ± 0.05	1.9 ± 0.1	4.0 ± 0.6
NV-YL	A10	0.32 ± 0.04	2.7 ± 0.1	8.4 ± 1.6

^aProteins were dialyzed against buffer A₁ in Experiment 1 and against buffer A₀ in Experiment 2 prior to titration. Buffer A₀ is 50 mM HEPES pH 7.6, 1 mM EDTA, 50 mM KCl, 5% glycerol. Buffer A₁ is buffer A₀ containing 1 mM DTT.

^bTitrations were carried out in the indicated buffers. Buffer A₁₀ is buffer A₀ containing 10 mM DTT.

^cK_D values were determined by measuring the quenching of intrinsic Trp fluorescence during titration with the indicated cap analogs.

^dThe K_D obtained with m₃^2,2,7GTP was divided by the K_D obtained with m⁷GTP.

The quantitative assay confirmed that substitutions of IFE-3 sequences into IFE-5 change m⁷GTP selectivity. Variants 1–2 loop, 3–4 loop and 3–4 up are slightly more selective than wt IFE-5 (increase in K_D ratio), whereas 3–4 down is less selective. However, the largest increase is observed with variant NV-YL (K_D ratio = 6.8), which discriminates nearly as well as IFE-3 (K_D ratio = 9.2). The change in selectivity for variant NV-YL occurs because of both an increase in affinity for m⁷GTP (1.3-fold) and a decrease in affinity for m₃^2,2,7GTP (1.7-fold). Interest ingly, neither the N-Y nor V-L variants alone show as much increase in K_D ratio as the combination. Some variants are less selective for m⁷GTP than wt IFE-5, as reflected in a lower K_D ratio. The largest effect is for variant QPK-KPM, which has a K_D ratio of 1.9 compared with 3.2 for wt IFE-5. This change results from an increase in m₃^2,2,7GTP affinity rather than a decrease in m⁷GTP affinity.

Change in selectivity as a function of redox state

A preliminary tertiary structure for IFE-5 revealed a close proximity of Cys-122 and Cys-142, suggesting the possibility of a disulfide bond between them. Also, the two proteins retained on m₃^2,2,7GTP–Sepharose (IFE-1 and IFE-5) contain Cys residues at four homologous positions, Cys-61, Cys-126, Cys-142 and Cys-185 (using the IFE-5 numbering), that are not present in IFE-3. Conceivably, the tertiary structure of IFE-5 is governed, in part, by disulfide bonds.

We therefore repeated the quantitative determination of K_D values in the absence of reducing agent. Titration of IFE-5 with m⁷GTP in buffer A₀, which does not contain dithiothreitol (DTT), reveals a K_D for m⁷GTP of 0.54 µM (Table I, Experiment 2). This is statistically the same as the value obtained in buffer A₁, 0.50 µM (Experiment 1). By contrast, titration with m₃^2,2,7GTP in buffer A₀ yields a K_D of 0.92 µM, compared with 1.6 µM in buffer A₁. Thus, the affinity of IFE-5 for m₃^2,2,7GTP increases in the absence of DTT. The overall effect is a 1.9-fold loss of m⁷GTP selectivity under more oxidizing conditions (K_D ratio decreases from 3.2 in buffer A₁ to 1.7 in buffer A₀).

For variants 3–4 up and 3–4 loop, the affinity for m⁷GTP is statistically the same in the presence or absence of DTT, but the affinity for m₃^2,2,7GTP increases in buffer A₀. This leads to an overall decrease in m⁷GTP selectivity of 2.0- to 2.2-fold for these two variants. The results are even more dramatic for variant NV-YL; the K_D values obtained in buffer A₀ indicate a loss of affinity for m⁷GTP and a gain of affinity for m₃^2,2,7GTP compared with buffer A₁, for a loss of m⁷GTP selectivity of 3.8-fold. Overall, the K_D ratios of IFE-5 and all variants tested decreased to approximately the same value in buffer A₀ (average of 2.2 ± 0.6), whereas they ranged as high as 6.8 ± 0.8 in buffer A₁.

We tested the specificity of this effect with several controls. First, the selectivity of IFE-3 is statistically unchanged in buffer A₀ (K_D ratio = 9.7 ± 0.8; Experiment 2) versus buffer A₁ (K_D ratio = 9.2 ± 1.1; Experiment 1). Second, the proteins that had been dialyzed against buffer A₀ were titrated with cap analogs in buffer A₁₀ (Experiment 2). The K_D values for IFE-3 are all statistically the same in buffer A₀, A₁ or A₁₀. However, the K_D values and ratios for IFE-5 and variant NV-YL revert to their pre-dialysis values. The fact that these proteins recover their high selectivity for m⁷GTP in the presence of DTT shows that subjecting them to air oxidation (dialysis against buffer A₀) does not result in protein denaturation. As a third control, we tested whether redox state influences overall protein stability. IFE-3 and IFE-5, previously dialyzed against buffer A₀, were subjected to progressively higher concentrations of guanidine isothiocyanate (0–2.67 M) in the presence or absence of 10 mM DTT. The proportion of denatured protein was estimated from the ratio of fluorescence emission at 335 nm to that at 355 nm (Hammarström et al., 2001). The denaturation curves for both proteins were the same with or without DTT (data not shown), indicating that air oxidation does not affect gross IFE structure.

Since DTT (Cleland, 1964) is a much stronger reducing agent than glutathione (GSH), the predominant reducing agent in the cell, we considered the possibility that the change in cap specificity induced by 1 mM DTT occurs outside the range of normal intracellular reduction potentials. We therefore repeated the titration of IFE-5 with m⁷GTP and m₃^2,2,7GTP at five physiological combinations (Shan et al., 1990) of reduced and oxidized (GSSG) glutathione. The K_D ratio was then determined for each redox condition (Figure 5). The results indicated that the affinities of IFE-5 for m⁷GTP and m₃^2,2,7GTP are nearly equal (K_D ratio = 1.3 ± 0.3) in 0 mM GSH and 2.5 mM GSSG. Selectivity increases as the GSH/GSSG ratio increases, reaching a maximum K_D ratio of 2.6 ± 0.4.

Fig. 5. Change in cap specificity of IFE-5 as a function of reduction potential. The K_D values for dissociation of m⁷GTP and m₃^2,2,7GTP from IFE-5 were determined by fluorescence quenching in buffer A₀ containing the indicated concentrations of reduced (GSH) and oxidized (GSSG) glutathione. The K_D ratios were computed as described in Table I.

Identification of a disulfide bond between Cys-122 and Cys-126

We sought direct evidence that IFE-5 undergoes reversible formation of disulfide bonds. Since the five Cys residues of IFE-5 occur in four different tryptic peptides, we adopted the following strategy. IFE-5 in either the oxidized or reduced states (i.e. dialyzed against either buffer A₀ or A₁) was alkylated with acrylamide, digested with trypsin and the peptides subjected to matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry. Peptides containing Cys residues in the sulfhydryl state will have masses that are 71.0 Da greater for every acrylamide moiety incorporated. Peptides containing Cys residues in the disulfide state will not react with acrylamide, and their masses will correspond to the peptide containing a sulfhydryl, an intrastrand disulfide, a mixed disulfide (with another peptide), or some combination of these. By comparing IFE-5 subjected to the same reducing or oxidizing conditions that produced the change in cap specificity, we tested whether the oxidation state of Cys residues could be correlated with the functional change.

Figure 6 shows selected regions of the mass spectra for tryptic peptides from IFE-5 that had been subjected to acrylamidation in either the oxidized (Figure 6A and B) or reduced (Figure 6C and D) states. The peak in Figure 6C at 1580.4 Da is within 0.3 Da of the predicted monoisotopic mass for the doubly acrylamidated peptide ¹¹⁷DMESIC GLVCNVR¹²⁹, which contains Cys-122 and Cys 126. (The peaks at 1581.4, 1582.4 and 1583.4 are isomers of the same peptide containing one, two or three ¹³C atoms, respectively.) Additional evidence that the cluster of peaks at m/z = 1580.4–1583.4 represents peptide ¹¹⁷DMESIC GLVCNVR¹²⁹ comes from a second cluster of peaks at 1596.4–1599.4, representing the same peptides containing Met sulfoxide. The m/z = 1580.4–1583.4 peaks are present in much lower amounts for IFE-5 first dialyzed against buffer A₀ (Figure 6A), indicating that Cys-122 and Cys-126 are predominantly in the disulfide form and therefore unable to react with acrylamide. The absence of a peak at m/z = 1509.4 in Figure 6A, which would have been produced by singly acrylamidated ¹¹⁷DMESIC GLVCNVR¹²⁹, indicates that both Cys-122 and Cys-126 were in the disulfide form at the time of alkylation.

Fig. 6. Determination of the redox state of Cys-122 and Cys-126 in IFE-5 by MALDI-TOF mass spectrometry. Cys residues were alkylated with acrylamide in buffer B containing either no DTT (A and B) or 13 mM DTT (C and D). The proteins were separated by SDS–PAGE and digested with trypsin. (A and C) Spectra in the m/z range of doubly acrylamidated peptide ¹¹⁷DMESICGLVCNVR¹²⁹ (calculated monoisotopic mass 1580.7 Da), indicating that both Cys-122 and 126 are modified by acrylamide under reducing (C) but not oxidizing (A) conditions. (B and D) Spectra in the m/z range of peptide ¹¹⁷DMESICGLVCNVR¹²⁹ containing a disulfide bond between Cys-122 and Cys-126 (calculated monoisotopic mass 1436.6 Da), indicating that this peptide is present under oxidizing (B) but not reducing (D) conditions.

The non-acrylamidated peptide ¹¹⁷DMESICGLVC NVR¹²⁹ has a mass of 1438.6 Da, but a form of this peptide with an intrastrand disulfide between Cys-122 and Cys-126 has a mass of 1436.6. A cluster of isotope-resolved peaks within 0.1 Da of this value, at m/z = 1436.5, 1437.5, etc., is observed in the oxidized (Figure 6B) but not the reduced (Figure 6D) IFE-5 sample. Therefore, some or all of the Cys-122 and Cys-126 residues are linked together by an intrastrand disulfide bond in peptide ¹¹⁷DMESICGLVCNVR¹²⁹ in the oxidized form of IFE-5, but this bond is reduced when the protein is dialyzed against buffer A₁.

The protection of Cys-122 and Cys-126 against acrylamidation could also be caused by mixed disulfides with other Cys residues in oxidized IFE-5. We therefore examined the oxidation states of Cys-61 and Cys-142, as determined by acrylamide reactivity. Peaks with m/z of the acrylamidated form of peptide ³⁵VYTFNTVPEFWA FYEAILPPSGLNDLCDYNVFR⁶⁷ were observed in equal abundance in oxidized and reduced IFE-5, yet none was observed for the non-acrylamidated form (data not shown), indicating that Cys-61 occurs only in the sulfhydryl form. Similar results were obtained for peptide ¹⁴¹NCNDDDTNMR¹⁵⁰, indicating that Cys-142 occurs only in the sulfhydryl form. Unfortunately, Cys-183 is present in a tryptic peptide of <500 Da. This region of the mass spectrum contains too much background noise to yield interpretable results. However, based on the tertiary structure model of IFE-5 (Figure 2), this Cys residue is 12–15 Å from either Cys-122 or Cys-126, so it is unlikely to form mixed disulfide bonds with them.

We obtained similar results using a different alkylating reagent, iodoacetamide, for modification of Cys residues (data not shown). We also obtained mass spectra of tryptic peptides, fractionated by reverse-phase HPLC, from oxidized and reduced IFE-5 without alkylation. A peak corresponding to the intrastrand disulfide form of ¹¹⁷DMESICGLVCNVR¹²⁹ was observed, but no mixed disulfides were found (e.g. between ¹¹⁷DMESICGLV CNVR¹²⁹ and ¹⁴¹NCNDDDTNMR¹⁵⁰; data not shown).

The most likely explanation for these results is that Cys-122 and Cys-126 are linked by an intrastrand disulfide bond in oxidized but not reduced IFE-5, whereas Cys-61 and Cys-142 are in the sulfhydryl form in both oxidized and reduced IFE-5. We attempted to obtain further evidence for the involvement of disulfide bonds in cap discrimination by producing IFE-5 variants. However, substituting either Cys-122 or Cys-126 with Ser resulted in proteins that were almost exclusively in the insoluble fraction of the E.coli lysate. Although this does allow us to measure cap specificity, it suggests that disulfide bonds may play a role in IFE-5 tertiary structure. Interestingly, IFE-3, which does not undergo a redox-induced change in cap specificity (Table I, Experiment 2), contains a Cys residue (Cys-148) at the equivalent position of Cys-122 in IFE-5, but no equivalent to Cys-126 (Figure 1).

Hypothesis for cap selectivity based on molecular dynamics (MD) modeling

We performed MD simulations to suggest structural hypotheses for the difference in m⁷GTP selectivity. Distances were measured between homologous amino acid residues in IFE-3 and IFE-5 during the dynamics trajectories to indicate the dimensions of the cap-binding pocket (Figure 7). The trajectories revealed that both the width (Figure 7A) and depth (Figure 7C) are greater in IFE-5 (thick lines) than IFE-3 (thin lines) over the course of the simulation. The volume of the binding pocket is determined, in part, by the degree of coordination among Trp-51, Tyr-87, Trp-97 and Trp-164 in IFE-3 (Trp-28, Asn-64, Trp-74 and Trp-138 in IFE-5). In IFE-3, the cap-binding pocket is more tightly circumscribed as a result of π interactions among the aromatic residues.

Fig. 7. Dimensions of the cap-binding cavity of IFE-3, IFE-5, and IFE-5 variants NV-YL and N-Y as determined by molecular dynamics simulations. The width of the cavity (A and B) was measured between the CE2 atom of Trp-51 and the CE2 atom of Trp-97 in IFE-3 (Trp-28 and Trp-74 in IFE-5). The depth of the cavity (C and D) was measured between the CG atom of Trp-51 and the CZ atom of Tyr-87 in IFE-3 (CG of Trp-28 and ND2 of Asn-64 in IFE-5). Measurements were obtained for each frame (equivalent to 0.3 ps) of the trajectories. In (A) and (C), thin lines represent IFE-3 and thick lines represent IFE-5. In (B) and (D), thick lines represent IFE-5, thin lines represent variant NV-YL and shaded lines represent variant N-Y. The average width and depth dimensions for the simulations were (in Å): 5.61 × 6.53 for IFE-3, 9.39 × 8.84 for IFE-5, 7.34 × 7.23 for NV-YL and 7.56 × 8.87 for N-Y.

Amino acid substitutions in IFE-5 that change cap-binding specificity also change the results of MD simulations (Figure 7B and D). The NV-YL variant (thin lines) has a narrower (Figure 7B) and shallower (Figure 7D) binding pocket than IFE-5 (thick lines). This relationship is maintained even with longer simulations (20 ps). By contrast, the depth of the binding pocket of the N-Y variant (Figure 7D, shaded lines) is more like that of IFE-5 than variant NV-YL.

These results suggest a correlation between m⁷GTP selectivity and the dimensions of the binding pocket. The pocket is smaller for IFE-3 and the NV-YL variant of IFE-5, which mimics the high m⁷GTP selectivity of IFE-3. The pocket is larger for IFE-5 and the N-Y variant of IFE-5, which fails to mimic the m⁷GTP selectivity of IFE-3. The measurements of the binding pocket reflect the overall more ‘open’ structure of the entire IFE-5 protein when compared with IFE-3 (Figure 2). By contrast, variant NV-YL more nearly resembles the ‘closed’ structure of IFE-3 (Figure 2), particularly in the cap-binding pocket, despite the fact that 99% of its amino acid residues are identical to wt IFE-5.

Discussion

Previous fluorescence-quenching studies have shown that human eIF4E binds m⁷G-containing cap analogs 7- to 8-fold more strongly than the equivalent m₃^2,2,7G-containing analogs (Carberry et al., 1990; Wieczorek et al., 1999). β-globin mRNA capped with m⁷G is translated 6-fold more efficiently in rabbit reticulocyte lysate than the same mRNA capped with m₃^2,2,7G (Darzynkiewicz et al., 1988). Finally, m⁷GTP is 17-fold more effective than m₃^2,2,7GTP for inhibiting translation of rabbit globin mRNA in reticulocyte lysate (Cai et al., 1999). We find that IFE-3 binds m⁷GTP 9-fold more strongly than m₃^2,2,7GTP (Table I), indicating that C.elegans IFE-3 is similar to the characterized mammalian eIF4Es in cap specificity. By contrast, IFE-5 has only a 3.2-fold preference for m⁷GTP over m₃^2,2,7GTP (Table I).

Several models can be proposed to explain this difference in selectivity. Model 1 postulates that the cap-binding pocket of IFE-3, but not of IFE-5, contains a binding determinant for some feature of m⁷GTP that is not shared by m₃^2,2,7GTP, e.g. the ability to form H-bonds at N2. This model makes several predictions (Table II): (A) the affinity of IFE-3 for m⁷GTP is greater than for m₃^2,2,7GTP, because of the extra binding determinant; (B) the affinities of IFE-5 for m⁷GTP and m₃^2,2,7GTP are approximately equal, because the extra determinant is not present in IFE-5; (C) the affinity of IFE-3 for m⁷GTP is greater than the affinity of IFE-5 for m⁷GTP, because IFE-5 lacks the extra determinant; and (D) IFE-3 and IFE-5 bind m₃^2,2,7GTP with nearly the same affinity, because the extra determinant does not affect m₃^2,2,7GTP. Model 2, a mirror image of Model 1, postulates that IFE-5, but not IFE-3, has a binding determinant for some feature of m₃^2,2,7GTP that is not shared by m⁷GTP, e.g. van der Waals interactions with the N2 methyl groups. The predictions for each of the four interactions are different from Model 1 (Table II, Model 2), but the reasoning is similar. Model 3 postulates that steric hindrance impedes access of m₃^2,2,7GTP to the cap-binding pocket of IFE-3 but not of IFE-5, whereas m⁷GTP can readily enter the binding pocket of both proteins. The set of predictions is different from either of the two previous models (Table II, Model 3). For example, the affinities of IFE-5 for m⁷GTP and m₃^2,2,7GTP (Table II, Model 3, Comparison B) are more similar than those of IFE-3 (Comparison A), because once m₃^2,2,7GTP is admitted, the binding determinants are the same as for m⁷GTP. When the predictions are compared with the actual data (Table II, Found), it is clear that none of the three models agrees with all of the data. We therefore propose a fourth model that combines elements of Models 1 and 3. In Model 4, m₃^2,2,7GTP is impeded from entering the binding pocket of IFE-3, but not of IFE-5, by steric hindrance. Furthermore, there is a positive determinant for binding m⁷GTP, but not m₃^2,2,7GTP, in both IFE-3 and IFE-5. The prediction of Model 4 for Comparison A is satisfied because of the m⁷GTP-specific binding determinant, e.g. Glu-98 of IFE-3, which would form an H-bond with N2 of m⁷G. Predictions for Comparisons B and C are satisfied because IFE-5 also contains this determinant (Glu-75). The prediction of Comparison D is satisfied because steric hindrance prevents IFE-3 from admitting m₃^2,2,7GTP.

Table II. Predictions of various models for relative binding of cap analogs to IFE-3 and IFE-5.

Comparison of binding	Model 1a	Model 2b	Model 3c	Model 4d	Founde
(A) IFE-3·m7GTP versus IFE-3·m32,2,7GTP	>	≅	>	>	>
(B) IFE-5·m7GTP versus IFE-5·m32,2,7GTP	≅	<	≅	>	>
(C) IFE-3·m7GTP versus IFE-5·m7GTP	>	≅	≅	≅	≅
(D) IFE-3·m32,2,7GTP versus IFE-5·m32,2,7GTP	≅	<	<	<	<

^aIFE-3, but not IFE-5, contains a positive binding determinant for some feature of m⁷GTP that is not shared by m₃^2,2,7GTP.

^bIFE-5, but not IFE-3, contains a positive binding determinant for some feature of m₃^2,2,7GTP that is not shared by m⁷GTP.

^cSteric hindrance impedes the access of m₃^2,2,7GTP to the cap-binding pocket of IFE-3 but not IFE-5.

^dThere are m⁷GTP-specific binding determinants in both IFE-3 and -5, but m₃^2,2,7GTP is sterically hindered from entering the binding pocket of IFE-3.

^eComparison of binding affinities (1/K_D) for the interactions shown.

The more restricted binding pocket of IFE-3 compared with IFE-5 (Figures 2 and 7) might not admit m₃^2,2,7GTP because this cap structure is bulkier and more disordered than m⁷GTP (Stolarski et al., 1996), requiring both a wider Trp spacing to enter the pocket and also a greater depth to accommodate the dimethyl moiety. In addition, the MD simulations indicate that IFE-5 is sufficiently flexible to close the gap between the ‘sandwich’ Trp residues, allowing the base stacking that provides favorable binding energy (Figure 7A). These results are consistent with the steric hindrance feature of Models 3 and 4, and suggest that Trp spacing impedes m₃^2,2,7GTP entry in both the mammalian eIF4Es and IFE-3.

All IFE-5 variants that include substitution of Tyr-64 and Leu-65 with Asn and Val, respectively, gain in their discrimination against m₃^2,2,7GTP, yet maintain a constant affinity for m⁷GTP. MD simulations indicate that the substitutions in variant NV-YL are alone sufficient to narrow the inter-Trp stacking distance [Figures 2, IFE-5(NV-YL) and 7B]. This narrowing may be due to interaction of the phenol ring of Tyr-64 with the indole ring of Trp-74 through perpendicular π interactions, restricting the movement of Trp-74 relative to Trp-28 at the opening of the pocket [Figure 2, IFE-5(NV-YL)]. Interestingly, Asn-64 and Val-65 are conserved in all three dual-specific IFEs, but are absent from the two mono-specific IFEs. We constructed molecular models of the other dual-specific IFEs, IFE-1 and IFE-2, and subjected them to MD simulations (data not shown). The models behaved more like IFE-5 than variant NV-YL or IFE-3 in overall dynamics as well as in the dimensions of the cap-binding pocket. We also aligned sequences of all known eIF4Es, from yeasts, insects, plants and vertebrates, and found that positions 64 and 65 are occupied by Asn and Val, respectively, only for IFE-1, -2 and -5. These results suggest that Asn-64 and Val-65 constitute a unique specificity determinant.

No data have been published showing that any of the IFE isoforms discriminate between m⁷GTP- and m₃^2,2,7GTP-capped mRNAs in vivo. Rather, we have only shown in the present study and in previous work that these proteins differ in their binding to the free nucleotides or Sepharose-linked forms of the nucleotides. It is, nonetheless, tempting to speculate that, in vivo, the more m⁷GTP-selective IFE isoforms recruit canonical m⁷GTP-capped mRNAs to the ribosome, whereas those IFE isoforms that are more permissive for m₃^2,2,7GTP recruit trans-spliced mRNAs. Because at least some of the IFE proteins differ in tissue distribution (Amiri et al., 2001), and because m⁷GTP- and m₃^2,2,7GTP-capped mRNAs are transcribed from different genes (Blumenthal, 1998), the presence of certain combinations of IFE proteins differing in cap specificity may account in part for the particular spectrum of proteins expressed in a given tissue.

Our results also suggest the possibility that mRNA cap-binding selectivity may be regulated by intracellular redox changes in the case of IFE-5 but not IFE-3. In the absence of a thiol reagent, there is both formation of a disulfide bond between Cys-122 and Cys-126 (Figure 6) and a decrease in the m⁷GTP selectivity of IFE-5 (Table I; Figure 5). It is conceivable that the formation of this disulfide bond serves as a reversible switch that responds to the redox state of the cell, as has recently been reported for the molecular chaperone Hsp33 (Jakob et al., 1999). Such a disulfide bond switch could modulate the recognition of m₃^2,2,7GTP-containing mRNAs by IFE-1, -2 and -5 in vivo, altering the relative translation of trans-spliced versus canonical mRNAs.

Materials and methods

Materials

Restriction enzymes and Pfu DNA polymerase were obtained from Promega. A DNA ligation kit (version 2) was purchased from Takara Shuzo Co., Ltd. m⁷GTP was purchased from Sigma, and m⁷GTP–Sepharose from Pharmacia Biotech, Inc. m₃^2,2,7GTP and m₃^2,2,7GTP–Sepharose were synthesized (Jankowska et al., 1993). Buffer A₀ is 50 mM HEPES, pH 7.6, 1 mM EDTA, 50 mM KCl and 5% glycerol; buffer A₁ is buffer A₀ containing 1 mM DTT; and buffer A₁₀ is buffer A₀ containing 10 mM DTT.

Construction of expression vectors for IFE-3, IFE-5 and variant forms of IFE-5

pETife3 and pETife5, encoding IFE-3 and IFE-5, respectively, were prepared as described previously (Keiper et al., 2000). A segment between the AflII and XhoI sites in pETife3, corresponding to a portion of the 3′-UTR of the transcribed mRNA (372 bp), was deleted to create plasmid pETife3-2, improving the yield of IFE-3. Site-directed mutagenesis of pETife5 was performed using the overlap extension method (Horton et al., 1989) to modify sequences between the NdeI and XhoI restriction sites. The resultant fragment was ligated into the same sites of pET21b. The sequences of all constructs were confirmed at the Iowa State University facility.

Expression, initial characterization and purification of recombinant IFE proteins

Escherichia coli BL21(DE3)pLysS cells were transformed with pETife3-2, pETife5 and mutated forms of pETife5. Cells were cultured in Luria–Bertani medium containing 100 µg/ml ampicillin and 20 µg/ml chloramphenicol at 25°C. Cultures of 100 ml and 2 l were used for analytical and preparative experiments, respectively. When the A₆₀₀ reached 0.5, expression of IFE proteins was induced for 6 h with 50 µM IPTG, after which cells were harvested by centrifugation. Subsequent steps were performed at 4°C. The cells were suspended in 2 ml of buffer A₁ containing 100 µM GTP for analytical experiments (20 ml for preparative experiments), disrupted by sonication and the lysate centrifuged.

For analytical characterization of cap-binding properties, 1 ml of the supernatant was applied to 0.3 ml columns of either m⁷GTP– or m₃^2,2,7GTP–Sepharose equilibrated with buffer A₁. The columns were washed three times with 5 ml portions of buffer A₁, and the bound proteins were eluted with 1.5 ml of either 100 µM m⁷GTP or m₃^2,2,7GTP, respectively, in buffer A₁. Proteins were precipitated by addition of trichloroacetic acid to a final concentration of 10% at 0°C for 30 min, collected by centrifugation, washed twice with 1 ml aliquots of –20°C acetone and dissolved in 50 µl of SDS–PAGE sample buffer. Half of each sample was analyzed by SDS–PAGE using a 15% gel with an acrylamide:N,N′-bisacrylamide ratio of 29:1.

For fluorescence studies, the cleared E.coli lysate was subjected to affinity chromatography on 1 ml columns of m⁷GTP–Sepharose as described above. The protein eluted with m⁷GTP was further purified to homogeneity on a MonoQ column (Amersham Pharmacia Biotech) and dialyzed against either buffer A₀ or buffer A₁, as indicated in the text. Protein concentrations were estimated by dye binding (Bradford, 1976).

Fluorescence measurements

The affinity of IFE proteins for cap analogs was measured by quenching of intrinsic Trp fluorescence during titration with either m⁷GTP or m₃^2,2,7GTP (Carberry et al., 1989; Wieczorek et al., 1998). Measurements were performed on a Model 750 Strobe-Master lifetime spectrofluorometer from Photon Technology Inc. (South Brunswick, NJ). The instrument was equipped with an SE-900 steady-state fluorescence option that utilizes a 75 W xenon arc lamp with photon counting detection. All measurements were made in a 5-mm-ID cylindrical quartz cuvette at 25°C using an excitation wavelength of 295 nm, emission wavelength of 348 nm, lamp slit of 2 nm, monochrometer slit of 3 nm and emission slit of 5 nm. Titrations were performed by adding m⁷GTP or m₃^2,2,7GTP to a 300 µl solution containing 1 µM IFE protein in either buffer A₀, A₁ or A₁₀. Fluorescence values were corrected for the dilution of the sample during titration. The contribution of buffer components to overall fluorescence was determined separately and subtracted.

ΔF was calculated from the expression:

ΔF = F₀ – F_a(1)

where F₀ is the fluorescence before addition of cap analog and F_a is the fluorescence after each addition. The inner filter effect (Lakowicz, 1999) was corrected using the relationship:

F_corr = F_observed × antilog₁₀[(A_295nm + A_348nm)/2](2)

Dissociation constants (K_D) were calculated using the program KaleidaGraph (Version 3.06; Synergy Software, Reading, PA) by non-linear least squares fitting of equation (3):

ΔF/ΔF_max = a/(K_D + a)(3)

where a is the cap analog concentration in micromolar and ΔF_max is given by the expression:

ΔF_max = F₀ – F_{40 μM cap analog}(4)

Chemical modification of Cys residues in IFE-5

IFE-5 was dialyzed against buffer A₀, and 5 µl of a 1 mg/ml solution were diluted into either 45 µl of buffer B₀ (0.5 M Tris–HCl, pH 8.5, 7 M guanidine–HCl, 10 mM EDTA) or buffer B₁₃ (buffer B₀ containing 13 mM DTT). The protein was allowed to stand for 2 h at room temperature, and then 2 µl of a 7 M acrylamide solution were added (Sechi and Chait, 1998). After standing for 40 min in the dark at room temperature, the reaction mixture was diluted with 38 µl of deionized H₂O, and protein was precipitated with trichloroacetic acid and separated by SDS–PAGE. The IFE band was visualized by Coomassie Blue staining and excised. In-gel digestion with trypsin and MALDI-TOF mass spectrometric analysis were performed at the LSUHSC-S Research Core Facility as described previously (Bradley et al., 2002). A similar procedure was used for alkylation of Cys residues using iodoacetamide (Hirs, 1967).

Molecular modeling

Computational analyses were performed on a Silicon Graphics Indigo² workstation with the Insight II software package from Molecular Simulations, Inc. (now Accelrys; San Diego, CA). Tertiary structure models of C.elegans IFE-3 and IFE-5 were created by homology modeling (Dwyer, 1996, 2001). The atomic coordinates for mouse eIF4E (Marcotrigiano et al., 1997) were kindly provided by Dr Stephen Burley, Rockefeller University, and served as the template for model building. The amino acid sequence of IFE-5 was automatically aligned with that of mouse eIF4E, and adjustments were made to account for gaps or differences in length. Atomic coordinates for structurally conserved regions were assigned, best-fitting loops were selected from the Protein Data Bank, and steric clash was eliminated. In order to model a disulfide bond between Cys-122 and Cys-126 in IFE-5, the segment containing these residues was constructed as a helical turn with the two Cys residues approaching a disulfide bond. These two residues were then fixed in subsequent minimizations and MD simulations to retain their relative positions. The model of IFE-5 was then subjected to energy minimization to convergence. The model of IFE-3 was similarly built.

MD studies of IFE-3, IFE-5 and variants of IFE-5 were carried out as described previously (Dwyer, 2001). Briefly, the AMBER force field was used with a distance-dependent dielectric constant of 4.0 and a non-bond 1–4 scaling factor of 0.5. The time step was 1 fs and the temperature was 310°K. Each run was initiated with 200 iterations of energy minimization (steepest descents algorithm) to resolve structural alterations introduced by amino acid substitutions in variant proteins. MD simulations were conducted for a total of 10–20 ps with 2000 iterations of equilibration, and frames were saved every 300 steps.