Purification and properties of initiation factor IF-3 from Caulobacter crescentus.

Abstract A five-step procedure is described for the purification to homogeneity of initiation factor IF-3 from ribosomal washes of Caulobacter crescentus. The protein, C-IF-3, has a molecular weight of 24,000 to 25,000 and seems to consist of a single polypeptide chain. It has three in vitro activities. (a) It promotes ribosomal binding of N-formyl methionyl transfer RNA directed by phage messengers; (b) it dissociates 70 S ribosomes into 30 S and 50 S subunits; (c) it binds single-stranded natural and synthetic RNAs. The factor is analogous to, and exchangeable with Escherichia coli IF-3. However, E. coli ribosomes do not bind N-formyl methionyl transfer RNA with C. crescentus phage Cb5 RNA as messenger and, conversely, C. crescentus ribosomes do not form an initiation complex with coliphage MS2 RNA or late T4 RNA. C-IF-3, as well as E. coli IF-3, have a high affinity for single-stranded stacked polymers, e.g. polyadenylate. Deviation from this type of structure, whether toward a more organized one, e.g. polyadenylate·polyuridylate, transfer RNA, or toward unstacking, e.g. polyuridylate, has an adverse effect on binding. DNA, native and denatured, is not bound at C-IF-3 concentrations sufficient to bind polyribonucleotides. Binding of C-IF-3 to coliphage MS2 RNA provides some protection from degradation by RNase.

A five-step procedure is described for the purification to homogeneity of initiation factor IF-3 from ribosomal washes of CauZobacter crescentus.
The protein, C-IF-J, has a molecular weight of 24,000 to 25,000 and seems to consist of a single polypeptide chain.
It has three in vitro activities. (a) It promotes ribosomal binding of N-formyl methionyl transfer RNA directed by phage messengers ; (b) it dissociates 70 S ribosomes into 30 S and 50 S subunits; (c) it binds single-stranded natural and synthetic RNAs. The factor is analogous to, and exchangeable with Escherichia coli IF-3.
However, E. coli ribosomes do not bind N-formyl methionyl transfer RNA with C. crescentus phage Cb5 RNA as messenger and, conversely, C. crescentus ribosomes do not form an initiation complex with coliphage MS2 RNA or late T4 RNA.
C-IF-3, as well as E. co2i IF-3, have a high affinity for single-stranded stacked polymers, e.g. polyadenylate.
Deviation from this type of structure, whether toward a more organized one, e.g. polyadenylate-polyuridylate, transfer RNA, or toward unstacking, e.g. poly uridylate, has an adverse effect on binding. DNA, native and denatured, is not bound at C-IF-3 concentrations sufficient to bind polyribonucleotides.
Binding of C-IF-3 to coliphage MS2 RNA provides some protection from degradation by RNase.
Studies with initiation factor IF-3* from Escherichia coli have demonstrated that this protein is required for ribosomal binding of natural mRNAs (1, 2; for review see Ref. 3). However, the mechanism by which ribosomes select proper sites on the messenger for correct initiation is not understood; neither is the mechanism by which IF-3 causes the binding of the messenger to ribosomes.
In addition to this activity, IF-3 dissociates 70 S ribosomes into 30 S and 50 S subunits (dissociation factor * This research was aided by Grants PR.A-61 and NP-25B from the American Cancer Societv and Grant AI 11517 from the National Institutes of Health. " $ Supported by Training Grant GM 01234 from the National Institutes of Health. 1 The abbreviations used are: IF-l, IF-2, IF-3, polypeptide chain initiation factors from Escherichia coli (the prefix C denotes the corresponding initiation factor from Caulobacter crescenlus); DF, ribosomal dissociation factor; SDS, sodium dodecyl sulfate. activity) (4) and remains bound to the 30 S subunits (5-7). This provides a pool of 30 S subunits competent to accept mRNA.
Investigations on initiation of protein synthesis in prokaryotes have been carried out largely on the E. coli system programmed with RNAs from RNA coliphages or T4 RNA.
In vitro studies with systems from bacteria other than E. coli (8-10) have already indicated some species-specific selectivity in the ability of ribosomes to bind and translate coliphage RNAs.
There is, however, a limitation in studies using only one class of hostspecific, i.e. coliphage, natural messengers. This prompted us to develop an amino acid-incorporating system from Cuulobocter crescentus programmed with a C. crescentus phage RNA (11). This system was selected because of the availability of the C. crescentus RNA phage, +C!b5, which is similar in physical properties (12, 13) and analogous in genetic content (12) to RNA coliphages.
RNA phages, other than coliphages, are known only for Pseudomonas aeruginosa (14) and for the Caulobacter group (15). The two parallel in vitro systems with speciesspecific mRNAs (E. co&coliphage RNA and C. crescentus-Cb5 RNA) provide a model for investigations on the specificity of components-ribosomes, initiation and interference (16, 17) factorsresponsible for the correct binding and translation of the messenger, and may provide new insights into the mechanism of ribosome-messenger recognition.
We have previously shown (11) that the ability to translate phage messengers by these two systems depends upon the presence of the corresponding host 30 S subunit; the source of the initiation factors and the 50 S subunits is immaterial.
In this communication we focus on the properties of initiation factor IF-3 from C. crescentus (C-IF-3). First, we describe a relatively simple procedure for the preparation of homogeneous C-IF-3.
The pure protein displays three in vitro activities. (a) It promotes ribosomal binding of fMet-tRNA directed by phage messengers; (b) it has DF activity; and (c) it has RNAbinding activity.
The factor is analogous to, and exchangeable with E. coli IF-3.
However, E. coli ribosomes do not bind fMet-tRNA if Cb5 RNA is used as messenger and, conversely, C. crescentus ribosomes do not bind fMet-tRNA with coliphage MS2 RNA or T4 RNA as messengers.
The binding of C-IF-3 to a number of natural and synthetic polynucleotides was examined.
Efficient binding does not appear to be sequencespecific but requires a rather unique polynucleotide conformation. C-IF-3, as well as E. coli IF-3, have a high affinity for single-stranded stacked polymers, e.g. poly (A). Any deviation from this type of structure, whether in the direction of higher for 10 min at 37" without C-IF-3; this procedure increases the ratio of 70 S to 30 S + 50 S ribosomes.
The factor was added and samples were incubated for 15 min at 37". Control samples without factor were processed identically; 0.1 ml was layered on 5 ml of a 5 to 20%. sucrose gradient in the same buffer and centrifuged at 4" for 90 min at 42,000 rpm in the Spinco SW 50.1 rotor.
Gradients were analyzed in an ISCO analyzer with a lomm light path flow cell.
The proportion of 70 S, 50 S, and 30 S ribosomes was computed from the areas under the peaks; 3.0 pg of freshly prepared Step 5 C-IF-3 produced 85 to 95% dissociation under these conditions and this was taken as a relative DF activity of 1.0 in expressing the results of Fig. 3 and Table 4. and ionic con&tions are given in the legends. Samples were incubated for 15 min at O", unless otherwise specified, diluted to 2 ml with the reaction buffer, and passed through Millipore filters.
Radioactivity retained on the filters was measured. In assays with labeled poly(U), the filters were pretreated with 0.5 M KOH according to Smolarsky and Tal (26) to reduce blanks without C-IF-3. Determination of Molecular Weight-Determination of molecular weight of C-IF-3 &as by SDS gel electrophoresis and Sephadex gel filtration.
Gel electrophoresis was carried out as described by Weber and Osborn (27)  Purification was also followed by SDS gel electrophoresis of steps 3 to 5 protein (Fig. 1). Origin at top, anode at bottom. Samples (0.08 ml) were layered on 6.5ye gels and run for about 6 hours at 5 ma per gel until the trailing edge of the tracking dye (bromphenol blue) reached the bottom of the column (Gels A, B, and C, column length 7.2 cm); in Gels D and E, the column length was 9.0 cm, and samples were run until the tracking dye moved either three-quarters (D) or halfway (E) through the column. Gels were stained with Coomassie brilliant blue (0.25%) and destained in 7.5y0 acetic acid (27) at 30,000 X g in a Sorvall GSA rotor until a clear supernatant (S-30) was obtained. This was spun for 12 hours at 40,000 rpm in Spinco Ti 60 rotors. The ribosomal pellet was washed for 12 hours in standard buffer containing 0.5 M NH&l and pelleted at high speed as above. The 0.5 M NH&l wash was used as the source of C-IF-2 (23). The ribosomal pellet was washed again for 12 hours in standard buffer containing 1.0 M NH&l and pelleted. The resulting ribosomal wash was used as the source of C-IF-3.
Ammonium Sulfate Fractionation (Step 6)-The 1.0 M NH&l ribosomal wash was fractionated with solid ammonium sulfate. The precipitate between 45 and 80% saturation contained over 90% of C-IF-3 activity.
The protein pellet obtairied after a 15-min centrifugation at 10,000 rpm in the Sorvall centrifuge was suspended in and dialyzed against two changes of a buffer containing 0.05 M NH&I, 0.05 M Tris-HCI (pH 7.4), 5 X 10m4 M EDTA, 5 X low4 M dithiothreitol, and 10% glycerol (W/V) (Buffer A&$ for 12 hours. Phosphocellulose Chromatography (Step S)-The entire Step 2 dialysate (250 mg) was layered on a column (0.9 x 25 cm) previously equilibrated in buffer ASO. The column was washed with 50 ml of Buffer AS0 followed by Buffer Aloo until Azso of the effluent was near zero. Over 90% of the protein layered was removed in the wash. In a narallel assav with MS2 13HlRNA. in&Ybation with Step 3 C-IFBdid not yield ethanol-precipitable material.
This result is in agreement with the sucrose gradient profile of degradation and demonstrates the endonucleolytic nature of the cleaving enzyme(s).
tions (20 ml) were pooled and concentrated to 1 ml by dialysis against Buffer Aso containing 50% polyethylene glycol (w/v).

DEAE-Sephadex Chromatography
( Step $)-The material from Step 3 (6 mg in 1 ml) was layered onto a column (0.9 X 10 cm) previously equilibrated in Buffer Aso. The column was washed with 25 ml of Buffer ASO followed by 25 ml of Buffer A loo. C-IF-3 eluted in the second wash. The active fractions (6 ml) were pooled and concentrated in Buffer AbO containing 6 M urea and 50% polyethylene glycol (w/v). SDS gel electrophoresis of Step 4 protein demonstrates that a single contaminant is present in addition to the major band of C-IF-3 (Fig. 1).
Step 4 C-IF-3 is free of an endonuclease activity which copurifies with the factor up to Step 3 (Fig. 2). The endonuclease is removed in the ASO column wash.
Sephadex G-100 Chromatography in 6 M Urea (Step 5)-Step 4 protein (3.6 mg in 0.9 ml) was layered on a column (0.9 X 65 cm) previously equilibrated with Buffer Aso containing 6 M urea and eluted with the same buffer. Exposure to this concentration of urea for up to 24 hours, does not inactivate C-IF-3. As shown in Fig. 3, C-IF-3 appears at a volume greater than the void volume and the incorporating activity with MS2 RNA and Cb5 RNA coincides with DF activity and RNA-binding activity. Note that C-IF-3 promotes the translation of both phage messengers by the corresponding host ribosomes; none of the phage messengers is translated by non-host ribosomes. SDS gel electrophoresis of Step 5 C-IF-3 reveals a single band (Fig.  ID). The active fractions were pooled, dialyzed against Buffer Azoo (containing 50% glycerol, v/v), and stored in small aliquots at -70". Under these conditions, Step 5 C-IF-3 has a biological half-life of approximately 2 months, whereas Step 2 and Step 3 protein can be stored for many months without loss of activity. In all experiments reported in this paper, Step 5 C-IF-3 of specific activity not below 30% of the original was used.   A summary of the purification procedure is given in Table I. Three preparations of C-IF-3 by the method described gave essentially identical results except that the specific activities of homogeneous Step 5 C-IF-3 in the standard incorporation assay (with MS2 RNA and E. coli ribosomes) were 700,850, and 1200. It should be noted that the large amounts of endonuclease present in protein of Steps 1 to 3 of the purification procedure precluded any accurate determination of the specific activity of C-IF-3 at these steps. For this reason it was impossible to assess precisely the degree of purification accomplished by them. The absolute values for the specific activity of Step 5 C-IF-3 are similar to those obtained for E. coli IF-3 in an analogous amino acid incorporation assay with MS2 RNA and E. coli ribosomes (4).
The purification procedure for C-IF-3 is relatively simple and offers certain advantages over previous methods (e.g. Refs.  28). Washing ribosomes with 0.5 M NH&l before applying the 1.0 M NH&l buffer for the extraction of C-IF-3 results in very small losses of activity and removes significant amounts of contaminating proteins. The use of phosphocellulose as the first ion exchange step in the procedure avoids the time-consuming separation of large amounts of protein on an anionic exchanger since it quickly removes over 90% of contaminants. Moreover, in contrast to the homogeneous factor, Step 3 protein (post-phosphocellulose)

1461
is fairly stable on storage and can be purified through Steps 4 and 5 in a short time.
The results of Fig. 2 point to the importance of screening for nuclease activity during the purification of components to be used in protein synthesizing systems. In fact, crude extracts and Step 3 C-IF-3 promote a low level of [*4C]lysine incorporation with either Cb5 RNA and E. coli ribosomes or MS2 RNA and C. crescentus ribosomes.
This most probably reflects some unspecific translation of RNA fragments for it occurs also when partially degraded phage RNAs are used as messengers.
This is not the case with Step 4 C-IF-3.

Properties of C-IF-S illoleculur
Weight-The molecular weight of C-IF-3 determined by SDS gel electrophoresis and gel filtration (see "Methods") was 25,000 and 24,000, respectively.
This indicates that C-IF-3 is a single polypeptide chain. C-IF-3 appears to be somewhat larger than E. coli 28).
[14C]jMet-tRNA Binding Directed by MS2 RNA, Cb5 RNA, and Late T4 RNA-Experiments on the ribosomal binding of [14C]fMET-tRNA directed by three different types of mRNA are presented in Table II. With either E. coli or C. crescentus IF-3, MS2 RNA and late T4 RNA stimulate the binding to E. coli but not to C. crescentus ribosomes, whereas, Cb5 RNA stimulates ['"C]fMet-tRNA binding to C. crescentus but not to E. coli ribosomes.
These results show the same messenger specificity and IF-3 interchangeability as in the amino acid incorporation assay ( Fig. 3 and Ref. 11). We have shown elsewhere that ribosomal specificity for the translation of MS2 RNA and Cb5 RNA resides in the corresponding host 30 S subunit (11); the source of the 50 S subunits is immaterial.
We have also shown that C-IF-3 promotes the correct translat.ion of MS2 RNA by E. coli ribosomes as judged by SDS gel electrophoresis of phage specific proteins synthesized in vitro (11).
DF Activity-DF activity assayed in column fractions at Step 5 of purification overlaps with the amino acid incorporation activity (Fig. 3). Pooled Step 3 and Step 4 fractions were also assayed, showing qualitatively a substantial enrichment of DF activity with respect to Step 1 and Step 2 proteins.
This indicates that, as with E. eoli IF-3, C-IF-3 has probably a dual function in protein synthesis.
The DF activity of C-IF-3 is not species-specific (Fig. 3).

RNA-binding
Activity-E. coli IF-3 is known to bind phage RNA and synthetic oligonucleotides of the type ApUpG(pA), (4,29). With labeled RNAs this binding can be conveniently measured by the retention of label on Millipore filters. Fig. 3 shows that peaks of amino acid incorporation activity, DF activity and RNA-binding activity of Step 5 C-IF-3 overlap. Furthermore, all three activities decrease upon storage of homogeneous C-IF-3 or after repeated freezing and thawing (Table  III).
Inactivation is even more pronounced in the DF-and RNA-binding assays for in these two, in contrast to the amino acid incorporation assay, C-IF-3 acts stoichiometrically rather than catalytically.
There is no change in the SDS gel electrophoresis pattern of aged Step 5 preparations (specific activity in the amino acid incorporation assay decreased about 90% from the original 1200). These observations suggest that the RNA-binding activity of C-IF-3 is not an unspecific property of C-IF-3 as a basic protein but is closely related to its two other activities-DF and ribosomal binding of mRNA. This prompted us to examine more closely the binding of C-IF-3 to selected polynucleotides, natural and synthetic, at various ionic conditions and temperatures.
It was previously noted that E. coli IF-I, although more strongly basic t,han IF-3, has no RNA-binding properties (4). The binding at 0" of C-IF-3 to poly(A), poly(U), poly(A . U), and poly (A . 2U) in sodium phosphate buffer and in Tris-HCl buffer containing Mg2+ is shown in Fig. 4, A and B. In the phosphate buffer there is good binding to the single-stranded stacked poly(A) helix (30) and almost no binding to poly(U) which, under these conditions, exhibits a nearly unstacked, random coil structure (31). The situation is reversed in the Mg2+-containing Tris-HCl buffer (Plate B) where poly(A) binding is almost nil and poly(U) binding increases considerably. These results suggest that binding of IF-3 is not base-specific but depends rather on certain structural features of the polynucleotides.
The increase in poly(U) binding apparently reflects an increased affinity of IF-3 to poly(U) when the polynucleotide is in an ordered state as is the case in Mg2+-containing Tris-HCl buffer (32,33). The drop in poly(A) binding is less clear. Mg2+ may compete with IF-3 for binding sites or poly(A) may attain a more compact structure which prevents binding; the latter alternative is more in line with subsequent experiments. The binding of IF-3 to the poly(A . U) double-stranded complex in sodium phosphate buffer (Plate A) is significantly less Not detectable than that of poly (A), and the binding of the triple-stranded poly(A . 2U) is still less. The change in ionic conditions has little effect on the structure of the double-stranded and triplestranded complexes, and, accordingly, there is little change in IF-3 binding.
Some decrease in binding in the presence of Mg2+ is seen for the double-stranded complex (A and B, Fig. 4); this may reflect a partial transition to a triple-stranded complex or to a more compact structure.
The experiments with synthetic polymers show that secondary and tertiary structure play a critical role in the IF-3 binding reaction.
IF-3 has little affinity for highly ordered helices and even less for unstacked polynucleotides but binds efficiently, albeit in competition with Mg2+, to stacked single-stranded helices. These conclusions are partially borne out by experiments with natural polynucleotides carried out under the same conditions, although a more complex pattern of interaction between IF-3 and RNA emerges. As seen in Fig. 5, A and B, IF-3 has little affinity for tRNA and still less in the presence of Mg2+. This is in agreement with the observation that ordered structure has an adverse effect on binding.
On the other hand, there is good binding to MS2 RNA in the presence of MgZf which is drastically reduced in sodium phosphate buffer. 16 S and 23 S ribosomal RNAs behave similarly, but binding is less efficient in the Mg2+ -containing buffer and the decrease in binding in sodium phosphate buffer is less pronounced.
There is virtually no binding to DNA and, significantly, to denatured DNA under the conditions of the assay. However, at very high concentrations of IF-3 (37.5 pg/O.l ml of assay sample), considerable binding of DNA was obtained (not shown) indicating that unspecific effects occur under these conditions.
The experiments with single-stranded polyribonucleotides shown in Fig. 5 indicate that efficient IF-3 binding requires a unique type of structure.
Certain binding sites on MS2 RNA become available for interaction when a specific Mg2+-induced conformation of the RNA molecule is attained.
To some extent this holds for rRNAs as well. The results with synthetic and natural polymers suggest that efficient binding takes place at unpaired, stacked regions of the RNA. In further experiments we examined the MgZt dependence of IF-3 binding to MS2 RN:\, formaldehyde-treated MS2 RNA and 23 S ribosomal RNA (Fig. 6). A high level of bindiug to MS2.RNA is seen even at 1 mM Mg2f. The extended binding plateau (up to about 10 mM Mg2+) indicates that at moderate concentrations MgH dots not cornpet<> effectively for C-IF-3 binding sit'es; however, at high Rig*+ concentrations (about' 20 mM) a more compact structure of the RNA may inhibit the formation of the complex with IF-3. The binding curves for the modified MS2 RNA and for 23 S RNA have similar shapes except that efficient binding starts at higher Mg2+ levels. Note that in 20 mM Tris-HCl buffer used in this experiment, and in the absence of metal ions, there is much less binding than in 20 mM sodium phosphate buffer (cf. Figs. 5 and 6). Clearly, the presence of a metal counterion (or NH4+, not shown) is a necessary condition for C-IF-3 binding.
This holds also for poly (A) binding which is greatly reduced in 20 mM Tris-HCl (not shown) as compared to the 20 nlM sodium phosphate buffer (Fig. 4).
The temperature stability of C-IF-3-RNA complexes is shown in Fig. 7. The assays were carried out under optimal ionic conditions of binding for each polymer.
Plateau levels of complexes formed at each temperature were obtained from timecourse curves and plotted as a fraction of binding at 0". With poly(U) (20 mM Tris (pH 7.3), 10 mM Mg*) the complex is  virtually dissociated at 10". Under these conditions the ordered state of poly(U) has a T, of about 6" and, as in the experiments of Fig. 4, although at different ionic conditions, there is little binding to the random coil form of the polymer.
The decreased complex formation with poly(A) (20 mM sodium phosphate, pH 7.3) with increasing temperature is gradual and again closely reflects structural changes of the polymer, namely the gradual extension of the stacked helix. With MS2 RNA (20 mM Tris (pH 7.4)) 5 mM Mg2+) binding is unchanged up to 25" and drops considerably at 37". With 23 S RNA (20 mM Tris (pH 7.4), 5 mM Mg*) temperature-induced changes are similar to those with MS2 RNA except that the complex is less stable.
Binding of MS2 [3H]RNA to IF-3 affords partial protection from digestion with pancreatic ribonuclease (Table IV). This effect depends on the enzyme concentration and temperature. In Experiments 1 and 3 of Table IV, digestion by RNase is complete, i.e. it produces fragments which do not bind IF-3; addition of IF-3 prior to RNase affords about 3 to 4% protection.
The protective effect is abolished by increasing the temperature (cf .  Table IV Experiments 2 and 4). These observations may be helpful in finding some relationship between the RNA-binding activity of IF-3 and its role in ribosomal binding to natural mRNA.
Sequence analysis of factor-protected fragments would allow a comparison with ribosome-protected sites (34). Many of the RNA-binding experiments reported in this section were repeated with homogeneous E. co% IF-3. The results were virtually identical in all cases.

CONCLUSIONS
The isolation, in a homogeneous state, of a bacterial IF-3 other than E. coli made possible experiments which demonstrate the exchangeability of the two factors with respect to (a) phage RNA-directed ribosomal binding of the initiator aminoacyl-tRNA (b) phage RNA-directed protein synthesis (see also Ref. ll), (c) DF activity, and (d) RNA-binding activity.
The phage RNAs used in this study were derived from two unrelated but physically and genetically analogous, E. co& and C. crescentus RNA phages. It follows that the two IF-3s do not discriminate between the phage messengers employed, nor are they selective toward different prokaryotic ribosomes. Specificity of messenger recognition in the two systems investigated resides in the 30 S ribosomal subunit (11). The RNA-binding experiments indicate that IF-3, whether E. coli or C. crescentus, has high affinity for certain structurally unique unpaired regions of the RNA which may correspond to initiation sites. These observations provide some clues to the mechanism by which IF-3 assists in the ribosomal binding of phage RNA, but it remains to be established what the relationship is between this activity and the RNA-binding activity, and what additional restrictions on RNA binding are imposed by the 30 S subunit itself (8, 11).
Aclcnowledgments-We thank Dr. S. Ochoa for his interest and comments on the manuscript and Drs. R. Mazumder and S.
Lee-Huang for gifts of purified E. coli IF-l, IF-2, and IF-3. We thank Mr. W. Frazier for large scale growth of C. crescentus and E. coli Ql3, Mr. G. Melders for preparation of phage MS2, and Miss M. DiPiazza for excellent technical assistance.