Biochimie (1992) 74. 319-326 © Soci6t6 fran~aisedc bi~chim~eet biologic mol~ulaire / Elsevier, Pads
319
Structure-function correlations (and discrepancies) in the 16S ribosomal RNA from Escherichia coli R Brimacombe Max-Planck-lnstitutfiir Molekulare Genetik, Abteilung Wittmann,Ihnestrasse 73, lO00 Berlin 33, Germany (Received 26 August 1991; accepted 15 November 1991)
Summary - - The published model for the three-dimensional arrangement of E coli 16S RNA is re-examined in the light of new
experimental information, in particular cross-linking data with functional iigands and cross-links between the 16S and 23S RNA molecules. A growing body of evidence suggests that helix 18 (residues 500-545), helix 34 (residues 1046-1067/!i89-1211) and helix 44 (residues 1409-1491) are incorrectly located in the model. It now appears that the functional sites in helices 18 and 34 may be close to the decoding site of the 30S subunit, rather than being on the opposite side of the 'head' of the subunit, as hitherto supposed. Helix 44 is now clearly located at the interface between the 30S and 50S subunits. Furthermore, almost all of the modified bases in both 16S and 23S RNA appear to form a tight cluster near to the upper end of this helix, surroundingthe decoding site. ribosomal RNA / three-dimensional models / functional sites / electron microscopy / modified bases
Introduction
A few years ago a model was published from this laboratory [1, 21 for the three-dimensional organization of the 16S ribosomal RNA from E coli. The model was based on the topographical information available at the time, including RNA-protein and intra-RNA cross-linking data (summarized in [1 ]), the map of the mass centres of the 30S ribosomal proteins as determined by neutron scattering [3], and several immunoelectron microscopic localizations of individual bases of tile 16S RNA on the surface of the 30S subunit (summarized in [4]). These data were combined with the phylogenetically established secondary structure of the 16S RNA molecule [5, 61 to derive a plausible three-dimensional model. A similar - but by no means identical - model was published a little later by Stem et al [71, which incorporated in addition a substantial set of RNA-protein foot-printing data. Both models [ 1, 7] showed the property that whereas the less highly-conserved regions of the 16S RNA secondary structure were located at the upper and lower extremities of the 30S subunit, the more highly-conserved regions were concentrated into a 'belt' around the centre of the subunit. Sites on the ribosomal RNA that had been implicated in one way or another in the ribosomal function (eg sites of binding of tRNA [8] or antibiotics [9]) were also concentrated into this 'central belt' area, and, moreover, these functional
sites appeared to be divided into two groups, the larger group centred in or near to the 'cleft' between the 'platform' and the 'head' of the 30S subunit, and the other (smaller) group on the opposite side of the 'head'; our 16S model [1], showing this bimodal distribution of the functional sites [10] is shown in figure 1. The separation of the functional sites into two groups in the models was to a large extent a consequence of the immunoelectron microscopic localizations of the methylated G-residue at position 527 in the 16S RNA [16] and of the C-residue at position 1400 [4], which can be cross-linked efficiently to the anticodon loop of P-site bound tRNA [14]; these two positions are both centres of functional importance, but they appeared to be on opposite sides of the 30S subunit [4] (fig 1). It was concluded [7, 10] that the smaller group of functional sites near the position of mG-527 must represent some kind of allosteric phenomenon, whereas the larger 'C-1400' group in the cleft constitutes the actual decoding site. However, more recent data from a number of sources have begun to suggest a direct rather than an allosteric involvement of the functional sites near to mG-527 (fig 1), which in turn implies that there must be serious errors in the three-dimensional 16S RNA models [1, 7]. The purpose of this paper is to describe and discuss these new data, and to indicate those parts of the 16S model that are specifically in need of revision. Three regions of the RNA are concerned,
320 namely helix 18 (residues 500-545), helix 34 (residues 1046-1067/1189-1211) and helix 44 (residues 1409-1491). It is not clear to what extent a relocation of these helices will affect other parts of the structure, and it will therefore take some time before it is possible to present a fully revised version of the model. The electron microscopic evidence
It is appropriate to begin by summarizing the status of the electron microscopic evidence concerning the 527 and 1400 regions of the 16S RNA, because this evidence in itself is by no means unequivocal. Sketches of the appropriate electron microscopic models from the various research groups involved are illustrated in figure 2. The most extreme separation of the two sites
(positions 527 and 1400) is represented by the model of Boublik's group [4], in which the mG-527 is located on the small lobe opposite to the cleft (fig 2A). The 16S RNA could readily be folded in such a way as to meet this criterion [1], and the 'figure 2A-type' location is reflected in the model RNA arrangement of figure 1. (Position 527 and neighbouring nncleotides are shown lying 'outside' the main perimeter of the model in figure 1, because they lie in the singlestranded loop at the end of helix 18, this loop-end is now known to be folded back into the helix in a pseudo-knot interaction ([22], and see figure 3, below).) The RNA model of Stem et al [7] also indicates a wide separation, although not quite so extreme as ours, of the 527 and 1400 positions (fig 22 of [7]). In the electron microscopic model of the St~iffler group [17], the mG-527 is located at the base of the
t,::e
A
Fig 1. The three-dimensional model of 16S RNA, including functional sites, taken from [I0]. The cylinders represent the helical regions of the RNA, numbered as in figure 3. The functional sites are denoted by the small black polygons together with the appropriate position in the 16S sequence, t denotes foot-print sites for tRNA in the absence of a poly(U) template [8], subsequently shown to be P-site positions [ 11]; tu correspondingly denotes sites in the presence of poly(U) subsequently shown to be A-site positions. Str, Spc, Neo etc denote foot-print sites of antibiotics [9], and m indicates a modified nucleotide in the 16S RNA [ 12, 13]. tRNA is the cross-link site to the anticodon loop at position C-1400 [ 14], and poly(A) is the site of cross-linking to a poly(A) messenger analogue at positions 1394-1399 ([10] cf [15]). A. View of the model from the interface side. B. View from the solvent side.
321 head of the subunit rather than on the side lobe (fig 2B), but is still well separated from the presumptive C1400 position at the base of the cleft (the latter position is not in fact included in the model of [17]). In contrast, however, Shatsky and Vasiliev [18] interpret the two sites as being virtually coincident (fig 2C); in this model there is no cleft, but rather a 'ledge' around the base of the head of the subunit. Figures 2D--F show the locations of related sites, interpreted in terms of the 30S subunit model from Lake's group [23]. Figure 2D gives the original location of mG-527 from Glitz's group [ 16], which - as in the St6ffler model of figure 2B - is at the base of the head on the opposite side to the cleft. Figure 2E summarizes the results of Iocalizations by Lake's group [19, 20] of three RNA positions, namely C-518, G-1392 and A-1492, using the techLaique of DNA hybridization electron microscopy. C-518 is in the same hairpin loop-end as mG-527 (see figure 3, below), and by virtue of the pseudoknot just mentioned - is very close to the latter position. G-1392 is similarly close to C-1400, and this area is furthermore connected to the 1500-region (see below) both by phylogenetically established tertiary interactions [22] and by intraRNA cross-links [47]. In agreement with the 1400-1500 neighbourhood, figure 2E indicates that positions 1392 and 1492 map at indistinguishable sites [20] which correspond to the location of C-1400 near to the base of the cleft (fig 2A). However, it should be noted that in this case (fig 2E) the sites are indicated as being on the solvent side rather than the interface side of the 30S subunit. Furthermore, in one set of images (fig 5D in [19]) the 1392 probe was observed on the opposite side of the head, effectively in the 'mG-527 position' (as in fig 2D). The location of C-518 in figure 2E is similar to that of mG-527 in figure 2D. although again the site mapped extends towards the solvent side of the subunit rather than the interface side, and indeed overlaps with the G-1392/A-1492 sites. Despite this interchangeability and overlap of the two sites, Oakes et al [20] interpreted the arrangement of the 16S RNA so as to give the maximal separation of the 1392 and 518 regions that was compatible with their observations (as do the models of [1, 71 as already noted). Figure 2F gives the position observed by Montesano and Glitz [21] of N6-monomethyladenosine in wheat-germ ribosomes. This nucleotide clearly maps at the 'mG-527' position (cf fig 2D), but - by comparison with the sequence of Xenopus laevis 18S RNA [24] - is at a position corresponding to A-1500 in the E coli 16S RNA, and should therefore map on the cleft side (as does the A-1492 in fig 2E). In the same publication [21], the authors also mapped the m7G residue from wheat germ at the same location (cf fig 2D), although in eukaryotes [25] this nucleotide is at a
;,.
'*b;,ut~lx~"
B.
"$tofflur"
".
"'.'ns~;at:v'*
-
Fig 2. Sketches of models of the E coli 305 ribosomal subunit, derived from immuno-electron microscopic studies, showing the locations deduced for mG-527, C-1400 and related positions. A, B, C. The models of [4, 17, 18] respectively. I). The location of mG-527, according to [16]. E. The locations of G-1392, A-1492 and C-518, determined by DNA hybridization electron microscopy [19, 20]. F. The location of mono-methyl adenosine on the eukaryotic 40S subunit [21]; this nucleotide is at a position corresponding to A-1500 in E coll. The views in A, B, C, D and F are from the interface side of the subunit, that in E from the solvent side.
position corresponding to G-1338 of the E coli sequence (see fig 3, below). The latter nucleotide has been cross-linked to protein S13 [26], and lies at a position in the 16S RNA (fig 3) which also favours a location on the cleft side of the head of the 30S subunit (cfthe model of fig 1). From these conflicting results it can only be concluded that, although the 527 and 1400 regions do indeed both appear to lie at the base of the head of the 30S subunit, the electron microscopic techniques are not capable of distinguishing between the two positions in this particular instance. In this ~.ontext it is important to remember that the head of the subunit is only joined to the body by a single helix of the 16S RNA (helix 28, fig 3, and see [1, 7]), and that the head can be severed from the body and isolated as a ribonucleoprotein fragment under very mild conditions of ribonuclease digestion [27]. The 'neck' of the 30S subunit may thus be much narrower than the electron microscopic models suggest, which would in turn imply that the two regions (527 and 1400) do in fact
322 Biochemical evidence 527 and 1400 regions
lie close together in the three-dimensional structure o f the 16S R N A . This c o m m o n location will for convenience be referred to simply as the decoding site; whether this site is indeed in the cleft, or further around the subunit (as in fig 2C), remains unclear.
c
for the juxtaposition
of the
There are four lines o f e v i d e n c e which suggest a physical link b e t w e e n the 527 and 1400 regior~:+ o f the
:,
+'+ c - - ~
+++-+'+-°
F
B--
,'% ?~--iiiI
...... ,-,.,,,
."°-°"-']~,
, _ , .•
...........
,-+ w-+
'"... ..
~_-2I I
j
z
:--~
.... .
...... ,
,-+
+ .' . . . .
,,.-.
319
Structure-function correlations (and discrepancies) in the 16S ribosomal RNA from Escherichia coli R Brimacombe Max-Planck-lnstitutfiir Molekulare Genetik, Abteilung Wittmann,Ihnestrasse 73, lO00 Berlin 33, Germany (Received 26 August 1991; accepted 15 November 1991)
Summary - - The published model for the three-dimensional arrangement of E coli 16S RNA is re-examined in the light of new
experimental information, in particular cross-linking data with functional iigands and cross-links between the 16S and 23S RNA molecules. A growing body of evidence suggests that helix 18 (residues 500-545), helix 34 (residues 1046-1067/!i89-1211) and helix 44 (residues 1409-1491) are incorrectly located in the model. It now appears that the functional sites in helices 18 and 34 may be close to the decoding site of the 30S subunit, rather than being on the opposite side of the 'head' of the subunit, as hitherto supposed. Helix 44 is now clearly located at the interface between the 30S and 50S subunits. Furthermore, almost all of the modified bases in both 16S and 23S RNA appear to form a tight cluster near to the upper end of this helix, surroundingthe decoding site. ribosomal RNA / three-dimensional models / functional sites / electron microscopy / modified bases
Introduction
A few years ago a model was published from this laboratory [1, 21 for the three-dimensional organization of the 16S ribosomal RNA from E coli. The model was based on the topographical information available at the time, including RNA-protein and intra-RNA cross-linking data (summarized in [1 ]), the map of the mass centres of the 30S ribosomal proteins as determined by neutron scattering [3], and several immunoelectron microscopic localizations of individual bases of tile 16S RNA on the surface of the 30S subunit (summarized in [4]). These data were combined with the phylogenetically established secondary structure of the 16S RNA molecule [5, 61 to derive a plausible three-dimensional model. A similar - but by no means identical - model was published a little later by Stem et al [71, which incorporated in addition a substantial set of RNA-protein foot-printing data. Both models [ 1, 7] showed the property that whereas the less highly-conserved regions of the 16S RNA secondary structure were located at the upper and lower extremities of the 30S subunit, the more highly-conserved regions were concentrated into a 'belt' around the centre of the subunit. Sites on the ribosomal RNA that had been implicated in one way or another in the ribosomal function (eg sites of binding of tRNA [8] or antibiotics [9]) were also concentrated into this 'central belt' area, and, moreover, these functional
sites appeared to be divided into two groups, the larger group centred in or near to the 'cleft' between the 'platform' and the 'head' of the 30S subunit, and the other (smaller) group on the opposite side of the 'head'; our 16S model [1], showing this bimodal distribution of the functional sites [10] is shown in figure 1. The separation of the functional sites into two groups in the models was to a large extent a consequence of the immunoelectron microscopic localizations of the methylated G-residue at position 527 in the 16S RNA [16] and of the C-residue at position 1400 [4], which can be cross-linked efficiently to the anticodon loop of P-site bound tRNA [14]; these two positions are both centres of functional importance, but they appeared to be on opposite sides of the 30S subunit [4] (fig 1). It was concluded [7, 10] that the smaller group of functional sites near the position of mG-527 must represent some kind of allosteric phenomenon, whereas the larger 'C-1400' group in the cleft constitutes the actual decoding site. However, more recent data from a number of sources have begun to suggest a direct rather than an allosteric involvement of the functional sites near to mG-527 (fig 1), which in turn implies that there must be serious errors in the three-dimensional 16S RNA models [1, 7]. The purpose of this paper is to describe and discuss these new data, and to indicate those parts of the 16S model that are specifically in need of revision. Three regions of the RNA are concerned,
320 namely helix 18 (residues 500-545), helix 34 (residues 1046-1067/1189-1211) and helix 44 (residues 1409-1491). It is not clear to what extent a relocation of these helices will affect other parts of the structure, and it will therefore take some time before it is possible to present a fully revised version of the model. The electron microscopic evidence
It is appropriate to begin by summarizing the status of the electron microscopic evidence concerning the 527 and 1400 regions of the 16S RNA, because this evidence in itself is by no means unequivocal. Sketches of the appropriate electron microscopic models from the various research groups involved are illustrated in figure 2. The most extreme separation of the two sites
(positions 527 and 1400) is represented by the model of Boublik's group [4], in which the mG-527 is located on the small lobe opposite to the cleft (fig 2A). The 16S RNA could readily be folded in such a way as to meet this criterion [1], and the 'figure 2A-type' location is reflected in the model RNA arrangement of figure 1. (Position 527 and neighbouring nncleotides are shown lying 'outside' the main perimeter of the model in figure 1, because they lie in the singlestranded loop at the end of helix 18, this loop-end is now known to be folded back into the helix in a pseudo-knot interaction ([22], and see figure 3, below).) The RNA model of Stem et al [7] also indicates a wide separation, although not quite so extreme as ours, of the 527 and 1400 positions (fig 22 of [7]). In the electron microscopic model of the St~iffler group [17], the mG-527 is located at the base of the
t,::e
A
Fig 1. The three-dimensional model of 16S RNA, including functional sites, taken from [I0]. The cylinders represent the helical regions of the RNA, numbered as in figure 3. The functional sites are denoted by the small black polygons together with the appropriate position in the 16S sequence, t denotes foot-print sites for tRNA in the absence of a poly(U) template [8], subsequently shown to be P-site positions [ 11]; tu correspondingly denotes sites in the presence of poly(U) subsequently shown to be A-site positions. Str, Spc, Neo etc denote foot-print sites of antibiotics [9], and m indicates a modified nucleotide in the 16S RNA [ 12, 13]. tRNA is the cross-link site to the anticodon loop at position C-1400 [ 14], and poly(A) is the site of cross-linking to a poly(A) messenger analogue at positions 1394-1399 ([10] cf [15]). A. View of the model from the interface side. B. View from the solvent side.
321 head of the subunit rather than on the side lobe (fig 2B), but is still well separated from the presumptive C1400 position at the base of the cleft (the latter position is not in fact included in the model of [17]). In contrast, however, Shatsky and Vasiliev [18] interpret the two sites as being virtually coincident (fig 2C); in this model there is no cleft, but rather a 'ledge' around the base of the head of the subunit. Figures 2D--F show the locations of related sites, interpreted in terms of the 30S subunit model from Lake's group [23]. Figure 2D gives the original location of mG-527 from Glitz's group [ 16], which - as in the St6ffler model of figure 2B - is at the base of the head on the opposite side to the cleft. Figure 2E summarizes the results of Iocalizations by Lake's group [19, 20] of three RNA positions, namely C-518, G-1392 and A-1492, using the techLaique of DNA hybridization electron microscopy. C-518 is in the same hairpin loop-end as mG-527 (see figure 3, below), and by virtue of the pseudoknot just mentioned - is very close to the latter position. G-1392 is similarly close to C-1400, and this area is furthermore connected to the 1500-region (see below) both by phylogenetically established tertiary interactions [22] and by intraRNA cross-links [47]. In agreement with the 1400-1500 neighbourhood, figure 2E indicates that positions 1392 and 1492 map at indistinguishable sites [20] which correspond to the location of C-1400 near to the base of the cleft (fig 2A). However, it should be noted that in this case (fig 2E) the sites are indicated as being on the solvent side rather than the interface side of the 30S subunit. Furthermore, in one set of images (fig 5D in [19]) the 1392 probe was observed on the opposite side of the head, effectively in the 'mG-527 position' (as in fig 2D). The location of C-518 in figure 2E is similar to that of mG-527 in figure 2D. although again the site mapped extends towards the solvent side of the subunit rather than the interface side, and indeed overlaps with the G-1392/A-1492 sites. Despite this interchangeability and overlap of the two sites, Oakes et al [20] interpreted the arrangement of the 16S RNA so as to give the maximal separation of the 1392 and 518 regions that was compatible with their observations (as do the models of [1, 71 as already noted). Figure 2F gives the position observed by Montesano and Glitz [21] of N6-monomethyladenosine in wheat-germ ribosomes. This nucleotide clearly maps at the 'mG-527' position (cf fig 2D), but - by comparison with the sequence of Xenopus laevis 18S RNA [24] - is at a position corresponding to A-1500 in the E coli 16S RNA, and should therefore map on the cleft side (as does the A-1492 in fig 2E). In the same publication [21], the authors also mapped the m7G residue from wheat germ at the same location (cf fig 2D), although in eukaryotes [25] this nucleotide is at a
;,.
'*b;,ut~lx~"
B.
"$tofflur"
".
"'.'ns~;at:v'*
-
Fig 2. Sketches of models of the E coli 305 ribosomal subunit, derived from immuno-electron microscopic studies, showing the locations deduced for mG-527, C-1400 and related positions. A, B, C. The models of [4, 17, 18] respectively. I). The location of mG-527, according to [16]. E. The locations of G-1392, A-1492 and C-518, determined by DNA hybridization electron microscopy [19, 20]. F. The location of mono-methyl adenosine on the eukaryotic 40S subunit [21]; this nucleotide is at a position corresponding to A-1500 in E coll. The views in A, B, C, D and F are from the interface side of the subunit, that in E from the solvent side.
position corresponding to G-1338 of the E coli sequence (see fig 3, below). The latter nucleotide has been cross-linked to protein S13 [26], and lies at a position in the 16S RNA (fig 3) which also favours a location on the cleft side of the head of the 30S subunit (cfthe model of fig 1). From these conflicting results it can only be concluded that, although the 527 and 1400 regions do indeed both appear to lie at the base of the head of the 30S subunit, the electron microscopic techniques are not capable of distinguishing between the two positions in this particular instance. In this ~.ontext it is important to remember that the head of the subunit is only joined to the body by a single helix of the 16S RNA (helix 28, fig 3, and see [1, 7]), and that the head can be severed from the body and isolated as a ribonucleoprotein fragment under very mild conditions of ribonuclease digestion [27]. The 'neck' of the 30S subunit may thus be much narrower than the electron microscopic models suggest, which would in turn imply that the two regions (527 and 1400) do in fact
322 Biochemical evidence 527 and 1400 regions
lie close together in the three-dimensional structure o f the 16S R N A . This c o m m o n location will for convenience be referred to simply as the decoding site; whether this site is indeed in the cleft, or further around the subunit (as in fig 2C), remains unclear.
c
for the juxtaposition
of the
There are four lines o f e v i d e n c e which suggest a physical link b e t w e e n the 527 and 1400 regior~:+ o f the
:,
+'+ c - - ~
+++-+'+-°
F
B--
,'% ?~--iiiI
...... ,-,.,,,
."°-°"-']~,
, _ , .•
...........
,-+ w-+
'"... ..
~_-2I I
j
z
:--~
.... .
...... ,
,-+
+ .' . . . .
,,.-.
