Partially Denatured rrnas of Escherichia coli and Bacillus stearothermophilus

Transcription

1 JOURNAL OF BACTERIOLOGY, Mar. 1985, p /85/ $02.00/0 Copyright D 1985, American Society for Microbiology Vol. 161, No. 3 Electron Microscopy of the Secondary Structure in Partially Denatured rrnas of Escherichia coli and Bacillus stearothermophilus BARBARA K. KLEIN, JORGE ROMERO, AND DAVID SCHLESSINGER* Division of Biology and Biomedical Sciences, Washington University School of Medicine, St. Louis, Missouri Received 11 September 1984/Accepted 10 December 1984 Partially denatured 16S and 23S rrnas from the thermophile Bacillus stearothermophilus show characteristic loop patterns when observed by electron microscopy. The patterns are very similar to those seen in rrnas from Escherichia coli. At least 2 of 4 most stable interactions in 16S rrna and 8 of 12 interactions in 23S rrna are in common for the two species. These interactions correspond well to features of secondary structure in models inferred for rrna from phylogenetic sequence comparisons and chemical modification studies. However, two additional large loops, enclosing large portions of the 23S rrna, have been detected in B. stearothermophilus for the first time, and even though other loops are similar, their relative frequencies vary in the two species. Much of the variation is consistent with relative AG' values for putative base-paired stems at the base of different loops; but the 5'-terminal loops in 23S rrna, for example, are unaccountably far less stable in B. stearothermophilus. Also, in general, structural features are not differentially stabilized in B. stearothermophilus; the relative stability of secondary structure in its ribosomes at elevated growth temperatures must involve interactions with ribosomal proteins or other cellular components. Secondary structures have been predicted for Escherichia coli 16S and 23S rrna based in part on comparative analysis of rdna sequences from different species (1, 6, 10, 11, 14, 16; J. L. W. Kop, Ph.D. thesis, University of California, Santa Cruz, 1982). Many of these predicted structural features have been observed directly by electron microscopy of E. coli rrnas (7, 8). In addition, electron microscopy can specify which base-paired stems are among the most stable to denaturation; in E. coli these include some of the stems which define the major structural domains in 16S and 23S rrna (7, 8). Sequence data from E. coli and Bacillus stearothermophilus 23S rrna have been compared to help infer a secondary structure model for this rrna (10, 11). We have compared the qualitative structures observed in E. coli rrna by electron microscopy with those of the evolutionarily distant B. stearothermophilus. Unlike E. coli, which is facultatively anaerobic, gram-negative, and mesothermic, B. stearothermophilus is obligately aerobic, gram-positive, and thermophilic. The results show that the basic domain structure and major loops are very similar in the two organisms, although some differences are observed in the stability of corresponding loops. The consistency of the data with published secondary structure models reinforces the models and further validates microscopy of partially denatured RNA as a tool for investigating secondary structure. MATERIALS AND METHODS Isolation of RNA. E. coli D10 (5) was grown in L broth at 37 C; the standard strain of B. stearothermophilus (from M. Nomura) was grown at 65 C. 30S and 50S ribosomal subunits were prepared from frozen cells, and the 16S and 23S rrnas were isolated, characterized, and stored as described previously for 23S rrna of the 50S ribosomes (7, 8). * Corresponding author. 981 Electron microscopy. The samples were prepared for electron microscopy as described previously (7, 8) with a hyperphase of 50% formamide-10 mm Tris (ph 8.0) and various concentrations of NaCl and either MgCl2 or EDTA. In this analysis, two ionic conditions were investigated for the E. coli and B. stearothermophilus 16S rrna samples: 10 mm NaCl, 1 mm M9Cl2 and 50 mm NaCl, 1 mm MgCl2. A total of 212 B. stearothermophilus molecules were analyzed, with a mean length of ,um, which is in agreement with earlier determinations for E. coli (7). For 23S rrna, 235 B. stearothermophilus molecules had a mean length of 0.99 ± 0.10,um, which is again similar to results with E. coli (cf. reference 8). As in earlier analyses (7, 8), all molecules with clear contours were traced and digitized in the samples of B. stearothermophilus rrna. Base pairing in rrna creates extensive secondary structure, but in the partially denaturing conditions used, only the most stable features remain. Of these, electron microscopy discerns the long-range interactions. Two kinds of structure are seen (7, 8) (Fig. 1): single-stranded open loops stabilized by short base-paired regions at their base (as in domain I of 16S rrna; see below) and smaller, closed loops which look like hairpins (as in domain II of 16S rrna). In the presence of added Mg2+ ions, the loop patterns of some molecules (<10%) were too complex to be traced readily and thus were not included in the analysis. Data analysis. Because the rrna loop patterns of E. coli and B. stearothermophilus are similar, the orientation of the loop patterns in B. stearothermophilus RNA are based on the schemes described previously for E. coli. The terminal loop at one end (5') and the large nonterminal loop near the opposite end (3') of the 16S rrna were used. For the 23S rrna, the two closely spaced loops, the large non-terminal loop near the 3' end, or both were used to orient the molecules. Loop location domains were identified from histograms; a minimum in the histogram defined the border

2 982 KLEIN, ROMERO, AND SCHLESSINGER FIG. 1. Electron micrographs of 16S and 23S rrna from B. stearothermophilus. (A) 16S rrna in 50 mm NaCl-1 mm MgCl2. (B) 23S rrna in 10 mm NaCl-0.3 mm MgCl2. Bar, 0.1,um; 108,225 x. between two domains. The sizes of composite loops were then determined by fitting the loop size distribution for each domain with a set of Gaussian peaks to estimate significance and standard deviations (4). The location of each loop is given by the position of its midpoint (Tables 1 and 2). Loop midpoints were determined as the mean location (nucleotide number) of a given loop size within the domain. Thermodynamic comparisons. To compare the relative stability of individual base-paired regions in the predicted secondary structures of E. coli (1, 6, 10, 11, 15, 17) and B. stearothermophilus (10, 17), the free energies of these regions were estimated by using the Tinoco rules (15) as modified by Salser (14) and Tinoco and others (2) on available sequence data. For the B. stearothermophilus 23S rrna, the free energies of base-paired segments were calculated by using the predicted secondary structure (10, 15, 17; Kop, Ph.D. thesis) (see Table 2). The predicted relative stabilities of corresponding loops in B. stearothermophilus and E. coli were then compared, based on the differences in AG' values (see Table 3). The calculated differences are listed for individual basepaired regions. Since the loops resulting from several of these interactions cannot be resolved by electron microscopy they are bracketed, and a single entry is listed in the last column of Table 3. For the difference in AG' to be significant, the two values had to differ by an arbitrary value of >2 kcal. This cut-off value was chosen because small changes in the parameters used for the free energy calculations could result in variations of that magnitude. Limited sequence information is available for the 3' end of 16S rrna from B. stearothermophilus, as determined in a recent review by Woese et al. (17). Five discrete interactions can be predicted for domain III of B. stearothermophilus 16S rrna and compared with E. coli. Of these interactions, three have identical sequences in the two organisms: 926 to 933/1384 to 1391, 938 to 943/1340 to 1345, and 946 to 955/1225 to 1235 (see Tables 23, 24 and 25 in reference 17). The other two interactions (984 to 990/1215 to 1224 and 1046 to 1067/1189 to 1211) differ by one or two base pairs, respectively, in the two species and differ in estimated free energy by less than 2 kcal for the two species, (see Tables 25 and 28 J. BACTERIOL. in reference 17). Thus, the loops in this domain of 16S rrna are expected to have a very similar stability in the two species. RESULTS AND DISCUSSION The ribosome cycle is similar in all bacteria studied; and rrnas and total r-protein from B. stearothermophilus and E. coli are interchangeable in the formation of chimeric active ribosomes (12). Thus, profound structural similarities between E. coli and B. stearothermophilus ribosomes can be expected. On the other hand, ribosomes from B. stearothermophilus often do not translate mrnas from gram-negative bacteria (3, 9), and their levels of antibiotic sensitivity are rather different. Also, the B. stearothermophilus ribosomes must function at 65 C (far above heat shock temperatures for E. coli). The rrnas may therefore differ considerably in structure or stability, and one might expect the structure in B. stearothermophilus rrna to be generally more stable. Basic loop patterns: domains and orientation. The pattern of loops observed by electron microscopy in partially denatured rrna isolated from B. stearothermophilus was analyzed in detail and compared with the loop pattern observed in E. coli rrna. The mean lengths of rrnas were comparable for the two species (see below). Examples of the prominent features of the loop patterns are shown in the electron micrographs in Fig. 1. The loop patterns are almost identical for the 16S rrna of these two species (cf. Fig. 1A with Fig. 1 of reference 7). In both cases the loops tend to occur in three location domains, with loops more prominent in the two outside domains. A result for 23S rrna, again similar to that for E. coli (8), is shown in Fig. 1B. The most striking feature is a pair of closely spaced loops near one end of the molecule. As a result of the similarity in rrna loops between E. coli and B. stearothermophilus, the orientation of the loop patterns observed in B. stearothermophilus is based on the schemes described previously for E. coli rrna (see above). By using the scaled, oriented molecules, three-dimensional histograms were constructed (Fig. 2A and B), showing the frequency of occurrence of loops as a function of their midpoint position and size. The three loop location domains described previously from electron micrographs of E. coli 16S are clear (cf. in Fig. 8 in reference 7). The 5'-terminal domain is centered at approximately 200 nucleotides, the center domain at 700 nucleotides, and the 3' domain near 1100 nucleotides. Loop patterns within domains of 16S rrna. The predominant feature in domain I is a terminal loop of about 380 ± 50 nucleotides in length, centered at 200 ± 30 nucleotides. In addition, there is a smaller, less-well-defined loop 200 ± 70 nucleotides long, centered at 200 ± 60 nucleotides. Domain II, in the center of the molecule, is dominated by a loop 180 ± 80 nucleotides long, centered at 700 ± 90 nucleotides. In domain III, near the 3' end, the loop size distribution is too complex to fit easily with a unique set of loop sizes. Therefore, the data for this region are presented in the form of a loop size histogram in Fig. 3A (cf. Fig. 3B for E. coli). The loop sizes within this distribution range from 150 to 600 nucleotides in length; the most prominent feature is a loop about 250 nucleotides long. Comparison with E. coli 16S rrna. Most aspects of the loop pattern of 16S rrna from B. stearothermophilus observed in the electron microscope and analyzed as described above are much the same as in E. coli rrna (1, 10,

3 VOL. 161, 1985 rrna STRUCTURE IN E. COLI AND B. STEAROTHERMOPHILUS 983 A B 16S 500ooo LOOP LOCATiON (nuc) L(OuOSIZ) LOOP LOCATION(nuc) LO(OPSIZE FIG. 2. Frequency, location, and size of loops observed in E. coli and B. stearothermophilus 16S rrna. (A) B. stearothermophilus; (B) E. coli. Combined data sets for all ionic conditions used are represented. (A) and (B) are scaled independently to the most frequently occurring loop. Thus, only relative peak heights are compared. 17). The average number of loops per molecule is about three for both species at the two ionic concentrations investigated (1 mm Mg2' and 10 or 50 mm NaCi). Another striking similarity is the three-domain structure, which appears to be a hallmark of the bacterial small subunit rrna (17). Similarities also exist in the loop pattern within each of the domains. These similarities are illustrated in Fig. 4, in which the loops identified in both species are indicated on the widely accepted secondary structure of 16S rrna from E. coli (1, 6, 10, 15, 17). The most obvious similarity is in domain I, in which the most prominent feature in both samples is the large terminal 370- to 380-nucleotide loop. Also in both samples, a smaller, less-frequent loop exists in domain I. However, these smaller loops differ substantially in size, approximately 100 nucleotides in the E. coli structure and closer to 200 nucleotides in the B. stearothermophilus structure. The small loop identified in the B. stearothermophilus structure does not correspond well to any of the features predicted in the E. coli structure; therefore, it is not indicated in Fig. 4. This loop may result from a combination of two different loops that are not well resolved or from a difference between the secondary structures of E. coli 0 0, [ A B.stearothermophilus B E. coli and B. stearothermophilus rrna. Further discussion of this point must await determination of the complete primary sequence and auxiliary evidence about the secondary structure of 16S rrna from B. stearothermophilus. The prominent loop observed in domain II also appears to correspond to a loop found in domain II of the E. coli structure. In domain II of the E. coli sample, three loops were identified at high Mg2+ concentrations (1 to 2 mm), but only one rather broad feature is evident in B. stearother- z U0.05- Lu LU U_ LOOP SIZE (Nuc) FIG. 3. Frequency of loops of various size in the 3' domain of B. stearothermophilus (A) and E. coli (B) 16S rrnas. Combined data sets for molecules prepared for microscopy in 1 mm MgCl2 with 10 mm or 50 mm NaCl. Absolute frequency of loops in the domains is shown. Nuc, nucleotides. FIG. 4. Frequent loops observed in E. coli and B. stearothermophilus 16S rrna. The secondary structure proposed by Woese et al. (17) for E. coli 16S rrna is shown. The filled squares indicate interactions (putative stems) identified in rrna of both species; the filled triangles are interactions observed frequently only in E. coli. Individual loops have not yet been determined for domain III (see text).

4 984 KLEIN, ROMERO, AND SCHLESSINGER TABLE 1. Loop sizes and locations in B. stearothermophilus 23S rrnaa Domain Size + SD (nucleotides)b Midpoint ± SD (nucleotides)c I A I B II ± ± 80 II-2 A ± 110 II-2 B III A III B IV V (A and B) V C V D ± 90 Large loop 1050 ± ± 90 Large loop 1690 ± ± 70 a All the loops inferred from Gaussian fits of the data in Fig. 2B are given. b Note that for loops IA, IB, and IV and the two large loops, the size distribution is much broader than for the other features. c The location of the loop is given for its midpoint along the contour of the rrna. mophilus under the ionic conditions investigated. This feature is likely a composite of unresolved peaks that cover the similar range of loop sizes resolved in E. coli. Another quantitative difference occurs in the shape of the Comparison of observed and predicted loop sizes and TABLE 2. locations in B. stearothermophilus 23S rrnaa Loop sizes Domain Observed Structural models AG' (kcal)b Size Midpoint Size Midpoint I A 200C I B 500C II-1 170c A 180d II-2 B 380C III A 190d III B III B 370C IV 260e V (A and BY 170e V C 380e V D 610e VI Large loop 1050c 2160 Large loop 1690C 2050 a Combined data sets (Fig. 2) were used to determine the sizes and locations of the midpoints of loops (in nucleotides). The predicted loops were obtained from the model for secondary structure of B. stearothermophilus (H. Noller, personal communication). b AG' values were estimated as described in the text. c Loops observed at significant levels. d Loops observed frequently (more than at significant level). ' Loops observed most frequently. f Interactions in parentheses are difficult to resolve in the B. stearothermophilus data (see text). domain III loop size histogram (Fig. 3). In E. coli (Fig. 3B), the loop size distribution is very broad, extending from 150 to 600 nucleotides with few if any individual features evident. Smaller-sized loops tend to be more frequent in the B. stearothermophilus samples. There exists partial sequence data for the 3' end of 16S rrna from B. stearothermophilus. The stabilities of basepaired sequences in E. coli and B. stearothermophilus can thus be compared. The stabilities of predicted base-paired regions are very similar (see above), which leads to a prediction of very similar 3' loop patterns for the 16S rrna molecules in the two species. The small differences actually observed may reflect variations in response to the ionic conditions investigated or could result from variations of relative loop stability in the 3' region which can even occur in different E. coli 16S rrna preparations (7). Loop patterns within domains of 23S rrna. A total of 1462 loops were analyzed in 235 molecules at three ionic conditions: lomm NaCl, 0.3 mm MgCl2; 50 mm NaCl, 0.5 mm MgCl2; and 80 mm NaCl, 5 mm EDTA. The number of loops per molecule decreased from the first to the third of these conditions, from 5.4 to 4.0 to 3.8 (±1.5) for B. stearothermophilus, and from 4.0 to 3.2 to 2.9 (±1.5) for E. coli. This trend is again consistent with earlier findings with E. coli (8). Six discrete loop location domains were identified (Fig. 5, top; two of these are designated as subdomains II-1 and II-2; the others are I and III through V). Of these six domains, domain IV (centered at 1900 nucleotides) and domain V (centered at 2400 nucleotides) are most prominent. In addition, there is a small 5'-terminal domain at 300 nucleotides. This is followed by two closely spaced subdomains: II-1 at 800 nucleotides and II-2 at 1100 nucleotides. Domain III is visible as a shoulder near domain IV, at 1500 nucleotides. c a I 23S LOOP II-2 r LOCATION (nuc) b 1-2 mii _ LOOP SIZE (nuc) ' NUCLEOTIDES 3 FIG. 5. Loops in B. stearothermophilus 23S rrna. (A) Frequency, location, and size of loops. Combined data sets for all ionic conditions used are represented. The domains in 23S rrna are labelled. (B) Schematic drawing of loops identified. Solid bars represent individual loops along the contour of 23S rrna; domain locations are indicated with roman numerals (see Table 1). Lines A and C show the two large loops identified in this analysis. Line B shows all frequent loops determined (see text). nuc, nucleotides. y J. BACTERIOL.

5 VOL. 161, 1985 rrna STRUCTURE IN E. COLI AND B. STEAROTHERMOPHIILUS 985 Table 1 summarizes the loop sizes and loop locations determined. In addition, superscripts c to e in Table 2 indicate the more frequent and most frequent loops (cf. Fig. 5). Domain I, at the 5' end, does not contain any prominent structure, but two loop sizes were identified: a large terminal loop of approximately 500 nucleotides and a broad distribution of small loops averaging about 200 nucleotides in length. Neither of these features was completely resolved because individual structures occur only infrequently. Domain II is best fit as a composite of two subdomains. Subdomain II-1 contains a single small loop approximately 170 nucleotides in length. Subdomain II-2 includes two loops, one approximately 400 nucleotides and the other, more frequent, about 180 nucleotides in length. Two distinct loops were identified in domain III. The smaller one (190 nucleotides) occurs more frequently than the larger one (370 nucleotides). Domain IV is one of the two most structured regions in the 23S rrna pattern (Fig. SA). It is dominated by a loop of 260 nucleotides, but the distribution of loop sizes is broad, indicating the presence of less prominent loops of different sizes that are not resolved in this analysis. The other prominent region in 23S rrna is domain V. In this case, all three loops identified occur frequently: 170, 380, and 610 nucleotides in length. In addition to the 11 loops described thus far, two very large additional loops occur at low but appreciable frequencies. They are 1050 and 1700 nucleotides in length and appear to enclose the features of the individual domains near the 3' end. Their locations and sizes are indicated in Fig. SB. A linear representation of the sequences in major loops is inferred from the data in Fig. 5A. In Fig. SB, the sequence enclosed by the smaller loop is shown as a thick bar along line A; it encompasses domains IV and V within its termini (dashed lines). The larger loop covers the sequence along the bar on line C; thus, the loop encompasses domains III, IV, and V. For comparison, the thick bars along line B indicate the sequences enclosed within the termini of the most frequent loops in individual domains: one loop each in domains 11-2, III, and IV and three loops in domain V. Comparison with predicted secondary structure. Because the 23S rrna from B. stearothermophilus has been sequenced and a proposed secondary structure constructed, the 13 loops (Table 2) can be compared with the predictions of the model for size and positions, and their relative frequencies can be compared with the calculated stabilities of the base-paired stems that define compared loops in the model of Noller et al. (10, 11; Kop, Ph.D. thesis, 1982; see Fig. 7) Because the sequence is known, comparisons can be taken a step further to examine relative stabilities. One can ask whether the relative AG0 values of the stems at the base of loops is in accord with the relative observed frequencies of the loops; i.e., do more frequently observed loops have more stable stems? (The use of putative AG' values to predict structures in RNA has had limited success, partially because of the large number of possible structures. However, when other techniques infer homologous loops and stems in an RNA from two bacterial species, AG0 values permit an assessment of their relative stabilities.) In a previous analysis of E. coli 23S rrna (8), four of the five most stable predicted structures correspond to stems of the prominent loops observed in the composite pattern; for B. stearothermophilus, five of the seven most stable predicted structures correspond to five of the six most promi- TABLE 3. Comparison of predicted stabilities of interactions in B. stearothermophilus and E. coli 23S rrnaa B. stearother- E. coli A(AG0) Domain mophilus thermophi- (B. stearo- variation' Expected Size Midpoint Size Midpoint lus - E. coll) I II III IV V VI a Loop sizes and midpoint are from the models for the two species (see Table 2) (10). The AGO values were calculated as described in the text, and the A(AG0) values were derived from them. b Expected relative variation in B. stearothermophilus loops larger than 100 nucleotides. Symbols: -, predicted to be relatively more stable in B. stearothermophilus; 0, no difference predicted; +, predicted to be more stable in E. coli. nent loops (Table 2). However, there are discrepancies between observed and expected frequencies in several domains, as follows. Domain I shows two broad loops which do not correspond well to the predicted structure. Each loop may include two similar loops that are not well resolved in the analysis. The resolution is poor in part because the frequencies are unexpectedly low. From previous studies, loops with base-paired stems of AG' < -10 kcal may be observed under these conditions of electron microscopy (7, 8). However, for the loop in domain IB, the stabilities of the predicted loops, calculated from AG' values of the two stems (Table 2), are large (negative AG0 values) and comparable to those for loops in domains II, III, and IV; but the feature is observed under the electron microscope much less frequently. In general, the agreement between observation and prediction is much better in other domains, but some additional exceptions are seen. For example, in domain II, there is a good correspondence of the observed and predicted loop patterns; the loop in dotnain II-2A is the most frequently observed, which is in accord with its more negative free energy. However, the large 680-nucleotide loop predicted to encompass domain II has a stability comparable to that of the clearly observed loop in domain II-2B, but it is not seen; in contrast, the observed small loop in domain II-1 corresponds well to a loop in the model but is predicted to be far less stable than the missing 680-nucleotide loop. In domain III, both predicted loops were observed, and the order of their stability is in line with AG' values.

6 986 KLEIN, ROMERO, AND SCHLESSINGER A B J. BACTERIOL. Downloaded from FIG. 6. Frequent loops observed in E. coli and B. stearothermophilus 23S rrnas. The secondary structure proposed by Noller et al. for E. coli 23S rrna (11) is shown. The filled squares indicate interactions identified in rrna of both species; the filled triangles are interactions observed frequently on in E. coli; the filled circle in domain IV is the 220 nucleotide interaction, prominent only in B. stearothermophilus samples. In domain IV, a single loop is observed frequently, rather than the two loops which are both predicted to be frequent from the AG' values (Table 2). The observed loop is also slightly larger than expected and may include an overlapping and incompletely resolved combination of the two predicted loops. The loops in domain V match the predicted results very well (except for the location of the smallest loop of this domain [cf. the similar problem in studies of E. coli 23S rrna] [8]). Consistent with the very negative AG' values, all three loops were among the most frequently observed. The additional larger loops have no defined counterpart in the model structure (see Fig. 6), although they often occur as "superloops" in molecules along with the standard structures of domains III to V (Fig. 5). Detailed comparison with E. coli i3s rrna. The average number of loops per molecule is significantly greater for 23S rrna from B. stearothermophilus (see above). The comparison between the loop structures and frequencies in B. stearothermophilus and E. coli can be based qualitatively on Fig. 5 and Fig. 8 in reference 7. (Data were included for a number of ionic conditions, but the same features are seen when a single ionic condition is used for the RNA of both species, for example in 10 mm NaCl-0.3 mm MgCI2.) In addition, Table 3 lists all the loops larger than 100 nucleotides predicted for the two species and the difference between the AG' values that we calculated for each of the corresponding loops in the two species. Because the predicted and observed loops for the two species are largely in agreement, observed loops can be indicated on the secondary structure map for the E. coli 23S rrna (Fig. 6). The individual features identified in domains II-1, II-2, III, and V are similar in both species (Fig. 6). The primary difference between these samples is in the relative and absolute frequencies of the various composite loops. These differences are most obvious in domains I, IV, and VI. The 5'-terminal loops in domain I occurred much less frequently in B. stearothermophilus than in the corresponding E. coli samples, even though they are predicted to be much more stable (Table 3). Also, in this domain the two large terminal loops are not resolved in the B. stearothermophilus samples, whereas the larger loop clearly dominates the pattern in E. coli. Because of this difference, the interaction is not indicated for B. stearothermophilus rrna in Fig. 6. on September 6, 2018 by guest

7 VOL. 161, 1985 rrna STRUCTURE IN E. COLI AND B. STEAROTHERMOPHILUS 987 In domain II, in contrast, of the twoloops in Fig. 6, the smallii-1 loop is predicted to be less stable than in E. coli, but the loop is appreciably more frequent. The predicted large 680-nucleotide loop is not seen in either species (see above). In domain III, both species show two loops (Fig. 6), but the 190-nucleotide loop should be relativelymore frequent in B. stearothermophilus as was observed. In domain IV, the most prominent features shift from the 340-nucleotide loop in E. coli (Fig. 6) to the 260-nucleotide loop in B. stearothermophilus. As mentioned above, this feature may include an unresolved combination of two loops, favoring the smaller 220-nucleotide loop. The AG' values (Table 3) are consistent with the greater relative frequency of the 220-nucleotide loop in B. stearothermophilus. In domain V, all the loops in B. stearothermophilus were expected to be more stable, and their frequency is indeed higher (Fig. 5 and Table 2). The small domain VI near the 3' end in E. coli samples (Fig. 6) was predicted to be relatively weak in B. stearothermophilus and is not seen under any of the ionic conditions investigated. Thus, the greatest discrepancy from predictions is seen in the 5'-terminal domain. The terminal loops re predicted to have mnore stable AG' values in B. stearoth&mophilus, but they appear significantly less often in B. stearothermophilus rrna and show a poorly resolved, broad distribution of loops. DomainsII and IV show some additional, less marked, deviations from expected relative frequencies. There are at least two explanations for the discrepancies between observed relative stabilities and the predicted AG' values calculated for certain stems. First, artifacts due to differential cytochrome c binding to particular rrna regions or errors in the analysis of complex loop patterns are difficult to exclude entirely. A more interesting possibility is that these discrepancies reflect the differential stabilization of certain loops by other physiological variables, such as divalent ions. For instance, the loops in domain I and the 340- and 220-nucleotide loops in domain IV tend to be much more frequent in the presence of Mg2+ in both species (8; unpublished data). Differential stabilization of some loops compared with others with greater AG' values of their stems is clear both within the data for B. stearothermophilus and in comparisons with E. coli. Within the B. stearothermophilus data, for example, structures are consistently less frequent throughout domain I compared with the other domains. Similarly, in domain IV, the observed relative frequencies of two loops within the same domain are reversed compared with the frequencies expected from AG' values (Table 2). In this case the smaller, 220-nucleotide loop is significantly stabilized in B. stearothermophilus as opposed to E. coli (Table 3), but although it is now observed to be more frequent than the larger 340-nucleotide loop, it is still predicted to be somewhat less stable. In summary, the loop patterns and stabilities observed for 16S rrna from these two species are generally very similar and are in good agreement with the limited sequence data available. The conclusions are somewhat different for the 23S rrna. In this case, the loop patterns are again very similar, but the relative frequencies of specific loops vary significantly. The 5'-terminal loops are seen unaccountably less frequently in B. stearothermophilus rrna, but the relative frequencies of other loops vary from those in E. coli in a manner generally similar to that predicted on the basis of sequence comparison. In particular, whereas most loops are more stable in B. stearothermophilus, a few, notably in domain VI, are less stable both experimentally and in their predicted secondary structure. Thus, particularly for the 16S rrna, loops do not appear to be differentially stabilized in the thermophilic organism. These results are in accord with the earlier findings of Pace and Campbell (13), who studied the UV hypochromicity of free rrna and ribosomes of the two species as a function of temperature. B. stearothermophilus rrna was slightly more stable, but its ribosomes were far more stable than were its E. coli counterparts. Most likely, the additional stability required for growth at elevated temperatures is thus conferred by tertiary structure interactions and bound r-proteins. Because microscopy directly determines both features of secondary structure and their relative stabilities, it can provide a way to extend analyses through evolutionary comparisons. For example, two very large loops which may not have been stable enough to see in E. coli were seen in B. stearothermophilus 23S rrna, suggesting that the domains defined earlier may be organized into "superdomains" in the intact ribosome. As another example, the predicted secondary structures of 16S and 23S rrna have been suggested to have universal features present in eucaryotic rrna as well (17), and it will be of interest to see whether comparable trends in loop patterns and stabilities occur at those evolutionary extremes. ACKNOWLEDGMENTS We thank H. Noller for communicating B. stearothermophilus sequence data in advance of their publication. Grady Phillips and Tom Rucinsky gave invaluable help with the electron microscopy, and Monty Brandenberg aided us in programming three-dimensional histograms on the VAX 11/780, as well as interfacing the digitizer. The work was supported by grant PCM from the National Science Foundation, postdoctoral fellowship 5-F32-GM to B.K.K. from the National Institutes of Health, and auxiliary funds from a Monsanto Biomedical Research Agreement. LITERATURE CITED 1. Brimacombe, R., P. Maly, and C. Zwieb The structure of ribosomal RNA and its organization relative to ribosomal protein. Prog. Nucleic Acids Res. 28: Cech, T. R., N. K. Tanner,I. Tinoco, Jr., B. R. Weir, M. Zuker, and P. S. Perlman Secondary structure of the Tetrahymena ribosomal RNA intervening sequence: structural homology with fungal mitochondrial intervening sequences. Proc. Natl. Acad. Sci. U.S.A. 80: Ehrlich, S. D DNA cloning in Bacillus subtilis. Proc. Natl. Acad. Sci. U.S.A. 75: Fraser, R. D. B., and E. Suzuki In J. A. Blackburn (ed.), Spectral analysis, p Marcel Dekker, Inc., New York. 5. Gesteland, R. F Unfolding of Escherichia coli ribosomes by removal of magnesium. J. Mol. Biol. 18: Glotz, C., and R. Brimacombe An experimentally-derived model for the secondary structure of the 16S ribosomal RNA from Escherichia coli. Nucl. Acids. Res. 8: Klein, B. K., P. Forman, Y. Shiomi, and D. Schiessinger Electron microscopy of secondary structure in partially denatured Escherichia coli 16S rrna and 30S subunits. Biochemistry 23: Klein, B. K., T. C. King, and D. Schlessinger Structure of partially denatured Escherichia coli 23S ribosomal RNA determined by electron microscopy. J. Mol. Biol. 168: Kreft, J., K. Bernhard, and W. Goebel Recombinant plasmids capable of replication in B. subtilis and E. coli. Mol. Gen. Genet. 152: Maly, P., and R. Brimacombe Refined secondary struc-

8 988 KLEIN, ROMERO, AND SCHLESSINGER ture models for the 16S and 23S ribosomal RNA of Escherichia coli. Nucleic Acids Res. 11: Noller, H. F., J. Kop, V. Wheaton, J. Brosius, R. R. Gutell, A. M. Kopylov, F. Dohme, W. Herr, D. A. Stahl, R. Gupta and C. Woese Secondary structure model for 23S ribosomal RNA. Nucleic Acids Res. 9: Nomura, M Assembly of bacterial ribosomes: in vitro reconstitution system facilitates study of ribosome structure, function, and assembly. Science 179: Pace, B., and L. L. Campbell. 1%7. Correlation of maximal growth temperature and ribosome heat stability. Proc. Natl. Acad. Sci. U.S.A. 57: Salser, W Globin mrna sequences: analysis of base J. BACTERIOL. pairing and evolutionary implications. Cold Spring Harbor Symp. Quant. Biol. 42: Stiegler, P., P. Carbon, M. Zuker, J. P. Ebel, and C. Ehresmann Structural organization of the 16S ribosomal RNA from E. coli. Topography and secondary structure. Nucleic Acids Res. 9: Tinoco, I., P. Borer, B. Dengler, M. Levine, 0. Uhlenbeck, D. Crothers, and J. GralHa Improved estimation of secondary structure in ribonucleic acids. Nature (London) New Biol. 246: Woese, C. F., R. Guteli, R. Gupta, and H. F. Noller Detailed analysis of the higher-order structure of 16S-like ribosomal ribonucleic acids. Mkrobiol. Rev. 47: