Extensive functional overlap between sigma factors in Escherichia coli.

Abstract

Bacterial core RNA polymerase (RNAP) must associate with a sigma factor to recognize promoter sequences. Escherichia coli encodes seven sigma factors, each believed to be specific for a largely distinct subset of promoters. Using microarrays representing the entire E. coli genome, we identify 87 in vivo targets of sigma32, the heat-shock sigma factor, and estimate that there are 120-150 sigma32 promoters in total. Unexpectedly, 25% of these sigma32 targets are located within coding regions, suggesting novel regulatory roles for sigma32. The majority of sigma32 promoter targets overlap with those of sigma70, the housekeeping sigma factor. Furthermore, their DNA sequence motifs are often interdigitated, with RNAPsigma70 and RNAPsigma32 initiating transcription in vitro with similar efficiency and from identical positions. SigmaE-regulated promoters also overlap extensively with those for sigma70. These results suggest that extensive functional overlap between sigma factors is an important phenomenon.

Full text

Page 1

Extensive functional overlap between s factors in
Escherichia coli
Joseph T Wade1, Daniel Castro Roa2, David C Grainger3, Douglas Hurd4, Stephen J W Busby3,
Kevin Struhl1& Evgeny Nudler2
Bacterial core RNA polymerase (RNAP) must associate with a r factor to recognize promoter sequences. Escherichia coli encodes
seven r factors, each believed to be specific for a largely distinct subset of promoters. Using microarrays representing the entire
E. coli genome, we identify 87 in vivo targets of r32, the heat-shock r factor, and estimate that there are 120–150 r32promoters
in total. Unexpectedly, 25% of these r32targets are located within coding regions, suggesting novel regulatory roles for r32. The
majority of r32promoter targets overlap with those of r70, the housekeeping r factor. Furthermore, their DNA sequence motifs
are often interdigitated, with RNAPr70and RNAPr32initiating transcription in vitro with similar efficiency and from identical
positions. rE-regulated promoters also overlap extensively with those for r70. These results suggest that extensive functional
overlap between r factors is an important phenomenon.
E. coli RNAP consists of five subunits, a2bb¢o, that make up the core
enzyme. Core RNAP can synthesize RNA, but it must associate with
an accessory s subunit to form RNAP holoenzyme in order to
associate with specific DNA sequences located at promoters1. RNAP
holoenzyme containing s70(Es70), the predominant s factor in
E. coli, binds the promoters of housekeeping genes. E. coli also encodes
six ‘alternative’ s factors that allow RNAP holoenzyme to associate
with smaller subsets of promoters2. Each alternative s factor is
required for expression of certain genes in response to a specific
environmental stimulus. Most bacterial genomes encode multiple s
factors that are required for complex cellular processes such as stress
response, morphogenesis and virulence2–5. Alternative s factors gen-
erally recognize different promoter sequences from the housekeeping
s factor and from one another.
The vast majority of genes are predicted to be transcribed by a
single RNAP holoenzyme. In E. coli, among genes whose transcription
is known to be driven by an alternative s factor other than s38,
only B10% are known to be transcribed by s70holoenzyme6. s38
represents an unusual case because the DNA sequence specificity of
s38is very similar to that of s70(refs. 7,8). Most instances of multiple
holoenzymes transcribing a single gene involve binding of different
holoenzymes to separate promoter sequences. Hence, the different
holoenzymes transcribe messenger RNAs with 5¢ UTRs of different
lengths. There are very few described examples of different holoen-
zymes binding overlapping promoter sites in any bacterial species8–14.
The E. coli alternative s factor s32, and its homologs from different
bacteria, are master regulators of the heat-shock response3,15. The
general model of the heat-shock response postulates that the increased
level of s32upon heat shock leads to s switching—that is, sub-
stitution of the housekeeping s70subunit of RNAP with heat
shock–specific s32. s32holoenzyme (Es32) recognizes distinct –10
(CTTGAAA) and –35 (CCCCATNT) promoter elements, thus direct-
ing RNAP to heat-shock genes16. Traditionally, members of the s32
regulon have been identified using biochemical techniques comparing
the abundance of individual transcripts or proteins before and after
heat shock17. Recently, s32-regulated genes were identified on a
genome-wide scale by virtue of transcriptional changes in response
to artificial overproduction of s32(ref. 18). This first systematic study
of the heat-shock regulon revealed at least 26 new heat-shock gene
candidates, suggesting that the heat-shock response is far more
complex than previously thought.
Here we use chromatin immunoprecipitation coupled with micro-
arrays (ChIP-chip) to identify 87 s32promoters, many more promo-
ters than have previously been identified. Notably, 25% of the s32
promoters we identify are located within the coding sequences of
known genes, suggesting a previously uncharacterized regulatory
function for s32. We demonstrate that two of these promoters drive
transcription of a novel class of RNA and that a third drives
transcription of an antisense RNA. We also compare the locations of
s32promoters identified in our ChIP-chip study to the locations of
s70promoters identified in a separate study (J.T.W. and K.S.,
unpublished data). We show that the majority of s32promoters can
also be transcribed by Es70both in vivo and in vitro, and that Es32
and Es70initiate transcription in vitro from the same start site at five
© 2006 Nature Publishing Group http://www.nature.com/nsmb
Received 16 April; accepted 13 July; published online 6 August 2006; doi:10.1038/nsmb1130
1Department of Biological Chemistry and Molecular Pharmacology, Harvard University, Boston, Massachusetts 02115, USA.2Department of Biochemistry, New York
University School of Medicine, New York, New York 10016, USA.3School of Biosciences, The University of Birmingham, Edgbaston, Birmingham B15 2TT, UK.
4Oxford Gene Technology, Begbroke Science Park, Sandy Lane, Yarnton, Oxford OX5 1PF, UK. Correspondence should be addressed to E.N.
(evgeny.nudler@med.nyu.edu) or K.S. (kevin@hms.harvard.edu).
NATURE STRUCTURAL & MOLECULAR BIOLOGY
ADVANCE ONLINE PUBLICATION1
ARTICLES

Page 2

of five promoters tested. Furthermore, we show that B40% of sE
(s24) promoters are also transcribed by Es70in vivo, and we
demonstrate transcription of two sEpromoters by s70in vitro.
Thus, we show that the promoters of two alternative s factors with
apparently distinct core promoter regions overlap extensively with the
promoters of s70. We suggest that extensive overlap between alter-
native and housekeeping s factors is a significant phenomenon in
E. coli and perhaps in other prokaryotic organisms.
RESULTS
Identifying r32promoters
We used ChIP-chip to directly assess s32binding across the E. coli
genome in vivo19–22. Sites of s32association must represent s32-
dependent promoters, as s32does not appreciably associate with
elongating RNAP23. To validate our ChIP assay, we first determined
the association of s32with three known target promoters, dnaK, groE
and clpB, before and after heat shock. We also determined the
association of core RNAP (b subunit) in the same cells22,24(Fig. 1).
As expected, there is a large increase in both b and s32association
with all three promoters after heat shock. Notably, s32is detectable at
these promoters before heat shock, demonstrating that a low level of
Es32can transcribe in normally growing cells. Before heat shock, there
is an appreciable level of core RNAP present at each promoter. As the
ratio of s32to b association before heat shock is much lower than that
after heat shock, it is likely that much of the b association before heat
shock is due to transcription initiation by a non-s32-containing
RNAP holoenzyme.
To identify s32targets on a genome-wide level in an unbiased
manner, we performed a ChIP-chip analysis using microarrays with
60-base oligonucleotides that cover the E. coli genome with an average
spacing of 223 base pairs (bp)22. Cells were harvested after heat shock.
Using data from two independent experiments, we identified 87
targets for s32, with an estimated false discovery rate (FDR) of 1%
(see Methods). At a slightly less stringent FDR of 2%, we identified
134 s32targets. For subsequent analysis, we used the set of 87 targets,
although the results are similar with the set of 134 targets. The list of
87 s32targets is summarized in Tables 1 and 2. Raw data are available
online (http://www.ogt.co.uk/cms-chip-publications.htm).
For 34 of the 38 targets previously identified by transcriptional
profiling18, we observed s32association at the same gene, the
promoter of the gene or the promoter of a gene located a short
distance upstream that is likely to be the upstream gene of an operon.
We suspect that the three of the four exceptions (ydaM, htrC and
cpxP) are artifacts of the transcriptional profiling experiment, because
they ranked near the bottom of the list of candidates. The transcrip-
tional profiling analysis did not identify six putative s32targets listed
by EcoCyc6, and our ChIP-chip analysis identified only one of these six
genes (rpoD), suggesting that the remaining five (ppiC, mlc, pphA,
rfaD and metA) are not s32targets, are bound weakly by s32or
are regulated by s32only under specific environmental conditions.
Of the 65 s32targets located in promoter regions, the corresponding
gene is in the top 10% of genes upregulated by heat shock in 42
cases25(Table 1). Promoters corresponding to heat shock–induced
genes typically show higher levels of s32association than do other
s32targets.
Many r32promoters are within known coding sequences
Notably, 22 s32targets (25%) identified by the ChIP-chip analysis
were not located in promoter regions; 20 were assigned to ORFs and
2 to intergenic regions between convergently transcribed genes
(Table 2). To validate these sites, we used ChIP and quantitative
PCR to measure the association of s32with five targets within ORFs
(sbcD, cycA, macB, dhaM and ydeP) and a target in the intergenic
region between the convergently transcribed tdk and ychG genes
(Fig. 2a). After heat shock, s32associated substantially with each of
these regions, whereas little or no s32association was detected before
heat shock (Fig. 2a), indicating that each of these regions is a genuine
s32target. As the ChIP-chip score (a measure of s32association
derived from the microarray) for the ydeP ORF is one of the lowest for
any s32target identified, the number of false positives from our
ChIP-chip analysis must be very low, in accord with our FDR estimate
of B1% (see Methods).
Several lines of evidence suggest that many of the ORF-located s32
targets represent transcriptionally active and heat shock–regulated
promoters. Two ORF-internal s32targets (macB and cycA) are within
genes upregulated by overproduction of s32(ref. 18), and six ORF-
internal s32targets are within the top 10% of genes upregulated after
heat shock, which is highly significant (Fisher exact test P o 0.01)25. It
is important to note that these microarray expression studies would be
unable to distinguish between full-length mRNAs and shorter RNAs
that represent only a fraction of the known gene.
To further investigate the function of Es32bound within ORFs, we
selected three genes that contained s32targets, cycA, macB and sbcD.
By sequence analysis, we were able to predict the likely binding site for
s32within both cycA and macB. In each case, the binding site for s32
was oriented in the same direction as the associated ORF. We
determined the association of s32and core RNAP (b subunit) before
and after heat shock at three locations within cycA, macB and sbcD:
(i) the 5¢ end, (ii) close to the predicted s32binding site (for sbcD, the
location of the predicted binding site was estimated using the ChIP-
chip data) and (iii) downstream of the predicted s32binding site
(Fig. 2b). At all three genes, s32association increased greatly after heat
shock, and the association pattern is consistent with binding of s32to
the predicted site (Fig. 2c). Notably, very little s32was associated with
the 5¢ end of each gene. At cycA and macB, RNAP association also
increased greatly after heat shock, both at the site of s32association
and downstream of this site (Fig. 2d). Very little RNAP was associated
with the 5¢ end of either gene. This strongly suggests that Es32is
required for heat shock–dependent transcription of RNAs that corre-
spond to a downstream portion of known coding transcripts. It is
likely that these shorter transcripts are noncoding, as they lack a
suitable Shine-Dalgarno sequence and initiation codon, suggesting a
more complex regulatory role. At sbcD, RNAP association also
increased greatly after heat shock, both at the site of s32association
and upstream of this site (Fig. 2d). Very little RNAP was associated
with the 3¢ end of the sbcD gene. This suggests that Es32is required
© 2006 Nature Publishing Group http://www.nature.com/nsmb
0
40
80
120
160
200
240
280
320
360
400
440
Occupancy units
a
clpBgroE dnaKclpBgroE dnaK
0
20
40
60
80
100
Occupancy units
Core RNAP association
σ32 association
No heat shock
Heat shock
No heat shock
Heat shock
b
Figure 1 In vivo binding of s32and RNAP to previously identified s32target
promoters. (a,b) Association of s32(a) and the b subunit of RNAP (b) with
known s32targets, before and after heat shock. Occupancy was measured
as a background-subtracted ratio of binding of s32or b to the tested region
over binding to a control region located within the coding sequence of sgrR.
Error bars represent one s.d. from the mean.
ARTICLES
2 ADVANCE ONLINE PUBLICATIONNATURE STRUCTURAL & MOLECULAR BIOLOGY

Page 3

for heat shock–dependent transcription of an antisense transcript that
covers some of the sbcD gene.
Determining the Er32consensus binding site
Using BioProspector26, a we defined a two-part consensus motif
for s32(Fig. 3a) that is a good match to the previously defined
consensus16,18. BioProspector identifies conserved sequence motifs,
including bipartite motifs, such as the consensus binding site for s32.
The motif we identified is more degenerate than previously identified
motifs, perhaps owing to our ability to identify activator-dependent
s32binding sites. At these sites, the affinity of s32for the promoter in
the absence of a transcription activator is likely to be low. The two
motif halves were separated by 13, 14 or 15 bp in most cases. This is
consistent with a previous study that showed optimum spacing of
15 bp at the groE promoter, with 14 bp also being suitable for
transcription and 16 bp being incompatible16.
Determining the overlap of r32and r70promoters
In a separate study, we identified s70targets comprehensively across
the E. coli genome (J.T.W. and K.S., unpublished data; see Supple-
mentary Data online). To compare the positions of s32target sites
with those identified for s70, we first selected only those s32targets for
which the peak microarray probe corresponded to a position within a
promoter that is not adjacent to a divergently transcribed gene
(divergently transcribed genes being those whose 5¢ ends are adjacent
but are transcribed on opposite strands). For each of these 36 s32
targets, we calculated the minimum distance from a s70target. We
also calculated the minimum distance from each of 1,000 randomly
selected probe coordinates to a s70target. The distributions of these
distances are shown in Figure 3b. There is a highly significant (Fisher
exact test P o 3 ? 10–10) overlap between the positions of s32targets
and s70targets. Notably, 56% (20 of 36) of s32targets analyzed are
located within 200 bp of a s70target, demonstrating that the majority
of s32-regulated promoters are also bound by Es70. 56% should be
regarded as a lower bound, as s70targets were only identified under a
single growth condition (J.T.W. and K.S., unpublished data). This
overlap is particularly dramatic when compared to a model where s
factors bind mutually exclusive sets of promoters. In that case, there
would be a substantially lower overlap than that seen with random
genome positions. To model this situation, we took advantage of the
unbiased transcript map identified in our separate study (J.T.W. and
K.S., unpublished data; see Supplementary Data). We determined, for
254 transcript 5¢ ends (not including transcripts of divergently
transcribed genes; transcripts were mapped from cells grown to
© 2006 Nature Publishing Group http://www.nature.com/nsmb
Table 1 r32targets assigned to promoters
Peak positiona
ChIP-chip scoreb
Assigned promoterc
Heat shock mRNAd
12137
1338236
517489
1382108
3325836
692705
1910676
4366611
4120324
455885
661847
1860543
2732219
458043
494360
3643277
4368697
4435901
4397378
3472832
2879078
1329066
1120249
3865622
3527116
3117399
2748770
4570170
4524200
1027984
1894792
3210711
1744318
3437568
2533473
111.8
72.1
59.9
58.0
51.0
49.9
49.0
46.8
43.4
41.2
37.1
34.7
31.1
30.3
29.1
28.3
27.4
27.2
23.5
22.4
22.0
21.0
20.9
20.4
19.7
18.3
17.8
16.3
15.9
15.7
15.3
14.6
13.6
13.2
12.6
dnaK
yciS
ybbN
ycjX
ftsJ
ybeZ
htpX
fxsA
hslV
clpP
ybeD
gapA
clpB
lon
htpG
prlC
groE
ytfI
miaA
rpsL
ygcI
topA
yceP
ibpA
yrfG
yghJ
grpE
yjiT
yjhI
yccV
sdaA
rpoD
ydhQ
yhdN
crr
58.5
5.6
9.9
39.3
9.1
9.6
36.1
50.7
16.2
3.3
ND
–3.1
36.5
20.3
33.8
16.7
77.5
–0.2
11.9
ND
4.8
5.9
25.5
297.4
12.1
ND
24.1
2.2
5.5
34.3
23.6
7.7
3.7
9.5
–2.5
Table 1 Continued
Peak positiona
ChIP-chip scoreb
Assigned promoterc
Heat shock mRNAd
2166182
2925989
1441543
3543459
2735126
3766465
231049
918375
3764283
63622
239047
2209265
1581861
1710454
2217821
4124864
3725552
1063460
3427148
2288414
2798757
4482322
1173268
1189625
4465355
705196
22199
2520600
2771222
4538807
11.7
10.1
9.0
8.7
7.9
7.8
7.6
7.2
6.9
6.8
6.1
5.4
5.4
5.3
4.7
4.7
4.7
4.7
4.5
4.4
4.3
4.3
4.2
3.9
3.9
3.8
3.7
3.7
3.4
3.2
b2084
sdaC
ldhA
yhgH
yfiA
yibG
yafD
ybjX
yibA
hepA
yafU
yehR
ydeO
ydgR
yehZ
rpmE
yiaA
yccE
yrdA
narP
nrdH
holC
mfd
phoP
treR
glnS
ileS
xapR
b2641
fimB
4.6
2.2
25.9
1.9
2.3
ND
6.7
–2.1
ND
6.8
3.8
3.0
3.2
–6.8
0.1
5.7
3.7
5.7
2.0
4.2
0.8
3.8
1.6
0.7
1.5
–0.7
–1.4
3.0
2.6
–1.0
aGenome coordinate corresponding to the center of the peak microarray probe.bNormalized
ratio of s32binding relative to the genomic DNA control; derived from the ChIP-chip analysis.
cFor probes found within nondivergent promoter regions, the gene with the nearest 5¢ end to
the peak position was chosen. For divergent promoter regions, the gene with the greatest
transcriptional induction after heat shock25was chosen. Promoters of previously known s32
target genes18are underlined.dThe level of induction (log2) of mRNA abundance after heat
shock for the gene corresponding to the assigned promoter25. Values in the top or bottom 10%
of this analysis are in bold. ND, not determined.
ARTICLES
NATURE STRUCTURAL & MOLECULAR BIOLOGY
ADVANCE ONLINE PUBLICATION3

Page 4

exponential phase at 30 1C), the minimum distance to another
transcript 5¢ end. This simulates the situation for s factors that
bind mutually exclusive sets of promoters, as all transcript 5¢ ends
are distinct. As a control, we determined, for 254 random genomic
coordinates, the minimum distance to a transcript 5¢ end. The
distributions of these distances are plotted in Figure 3c. This simula-
tion demonstrates that any s32targets located within 300 bp of a s70
target are highly likely (499%) to represent s32promoters that can
also be transcribed by s70. (Note that the resolution of this analysis is
not sufficient to determine the precise positions of the promoter
sequences for each s factor.)
Close examination of heat-shock promoters reveals many good
matches to the –35 and –10 hexamers required for Es70association.
Furthermore, the upstream half of the s32consensus sequence
(CTTGAA) is notably similar to the consensus sequence of the
s70–35 hexamer (TTGACA). We aligned the consensus sequence of
a s32-dependent promoter with that of a s70-dependent promoter so
that the overlap of the consensus sequences for the two s factors was
maximized. As the two binding elements for each s factor can be
spaced in three different ways, there are nine potential ways these
promoter elements can overlap (Fig. 3d). In each case, there is a
mismatch of only 1 bp in the upstream promoter elements for s32and
s70. In five of the nine cases, there is 0 or 1 mismatched bp in the
downstream promoter elements for s32and s70. This suggests that
s32and s70can regulate the same promoters by binding overlapping
DNA sequences.
The sequences of many s32-bound promoters contain excellent
matches to both the –10 and –35 hexamer consensus sequences for
s70. Quantitative analysis of s32and s70association with each of three
such promoters, yciS, rpsL and hepA, revealed a large increase in the
binding of s32after heat shock (Fig. 3e). s70associated with each
promoter before and after heat shock, although binding decreased
about two-fold after heat shock (Fig. 3f). Thus, both s factors are able
to bind these promoters, sometimes under the same growth conditions.
Transcription by Er32and Er70in vitro
To examine whether Es70can indeed recognize heat-shock promoters,
we reconstituted in vitro transcription using highly pure E. coli RNAP
core enzyme and individually purified s70and s32. We compared the
efficiency of initiation by Es32and Es70at five heat-shock promoters,
groE, rpsL, clpB, htpG and lon, that are known to be regulated by Es32.
The s70-specific T7A1 promoter template was used as a negative
control for Es32initiation. Notably, Es70was at least as efficient at
each of the heat-shock templates as Es32at either 37 1C or 45 1C
(Fig. 4a). In contrast, only Es70was able to initiate at the T7A1
promoter, showing the strict specificity of s32. This control result also
demonstrates the high purity of s32and s70(Supplementary Fig. 1
online) and excludes any potential cross-contamination.
To examine whether the two holoenzymes initiate from exactly the
same or different start sites, we modified the transcription reaction to
© 2006 Nature Publishing Group http://www.nature.com/nsmb
987654321
987654321
0
10
20
30
40
50
60
Occupancy units
No heat shock
Heat shock
No heat shock
Heat shock
Core RNAP association
σ32 association
No heat shock
Heat shock
σ32 association
0
56
23
10
20
30
40
50
60
70
80
90
100
Occupancy units
987
4
1
sbcD
ydeP
dhaM
cycA
tdk/ychG
macB
sbcD
macB
cycA
0
10
20
30
40
50
60
70
80
90
100
Occupancy units
c
d
ab
Figure 2 In vivo binding of s32and RNAP to novel s32targets not located in known promoter regions.
(a) Association of s32with novel nonpromoter targets before and after heat shock. (b) Schematic of the
cycA, macB and sbcD genes, indicating positions of PCR products used for ChIP and quantitative PCR.
Gray rectangles, ORFs; gray arrows, s70-dependent promoters; black arrows, putative s32-binding sites.
(c,d) Association of s32(c) and the b subunit of RNAP (d) with three regions of cycA, macB and sbcD,
before and after heat shock. In a, c and d, occupancy was measured as in Figure 1.
Table 2 r32targets assigned to nonpromoter regions
Peak
positiona
ChIP-chip
scoreb
Assigned
genec
Heat shock
mRNAd
Distance
from nearest
5¢ end (bp)e
2780215
2385670
4429093
415417
921091
1293531
1579583
2236301
2319337
2923838
1789863
1247539
2769946
4359431
2764390
1624219
516521
1213931
1609353
1584068
2762841
3878830
17.8
15.5
10.7
10.4
10.2
6.8
6.7
6.5
5.7
5.5
5.4
5.1
5.0
4.7
4.5
4.4
3.8
3.8
3.8
3.5
3.4
3.3
ypjA
yfbM/yfbN*
cycA
sbcD
macB
tdk/ychG*
ydeN
mglA
atoS
yqcD
ydiV
dhaM
yfjU
cadC
yfjN
dcp
ybbM
ycgF
yneF
ydeP
yfjL
recF
2.9 662
714 (yfbM)
913 (ytfE)
440 (sbcC)
722 (cspD)
708 (ychG)
754 (ydeM)
526 (mglC)
551 (atoC)
468
181
801
230
526
450
1,185
741
649 (ycgE)
525
442
334
414
2.5/2.3
1
18.9
5.2
0.5/2.9
3.7
6.7
3.1
0.8
ND
–5.8
1.4
0.8
ND
–1.4
2
3.4
0.4
ND
1.2
5.8
aGenome coordinate corresponding to the center of the peak microarray probe.bNormalized
ratio of s32binding relative to the genomic DNA control; derived from the ChIP-chip analysis.
cGene within which the peak probe position was located. For intergenic regions of convergently
transcribed genes, both genes are listed (marked with asterisk). Previously identified s32target
genes18are underlined.dThe level of induction (log2) of mRNA abundance after heat shock for
the gene corresponding to the assigned promoter25. Values in the top or bottom 10% of this
analysis are in bold. ND, not determined.eDistance from the peak position to the nearest 5¢ end
of a gene. This is not always the assigned gene; when it is not, the gene with the nearest 5¢ end
is listed in parentheses.
ARTICLES
4 ADVANCE ONLINE PUBLICATIONNATURE STRUCTURAL & MOLECULAR BIOLOGY

Page 5

allow the formation of promoter-proximal stalled elongation com-
plexes (‘walking’ transcription assay; see Methods). Three biotin-
labeled heat-shock promoter templates, groE, lon and dnaK, were
prepared and immobilized on streptavidin beads. Transcription of
these immobilized templates was carried out using either Es32or
Es70, but only three of the four NTPs, so that transcription stalled at a
position close to the promoter. In each case, the same predominant
RNA products of predicted length were generated, indicating identical
transcription start sites for both holoenzymes (Fig. 4b). These RNA
products belonged to mature elongation complexes, as they were
retained on beads after high-salt washing and could be extended
upon addition of the missing nucleotide. We repeated the walking
transcription assay with the template for groE, using different salt
conditions (150 mM potassium glutamate) and not including dinu-
cleotides in the transcription reaction (Supplementary Fig. 2 online).
We detected transcripts of the same size as those detected in the
previous walking transcription assay (Fig. 4b), confirming that our
results are not attributable to low-salt conditions or the presence of
initiating dinucleotides.
In contrast to our work, an earlier study27did not detect transcrip-
tion of dnaK by Es70in vitro. We clearly detect transcription of dnaK
by Es32and Es70from the same start point (Fig. 4b). We suggest that
this discrepancy could result from differences in the reaction condi-
tions or the method of RNAP holoenzyme and template preparation.
Determining the overlap of rEand r70promoters
A combination of transcriptional profiling and bioinformatics has
identified a large number of promoters regulated by sE(s24), which is
required for the envelope stress response28. Using the same approach
as described above for s32(Fig. 3b), we compared the locations of
such sE-regulated promoters to the targets of s70. Just as observed for
s32, there is extensive overlap between the positions of sEtargets and
those of s70targets. Indeed, depending on the promoters analyzed
© 2006 Nature Publishing Group http://www.nature.com/nsmb
T7A1
Core
rpsL
Core
clpB
Core
htpG
Core
groE
Core
Ion
Core
EC18EC17 EC14Marker
12
14
18
7070 7070 32 3232
Size (nt)
σ
18
17
14
Size (nt)
groEIon dnaKT7A1 Promoter
σ
Promoter
12345678 9 10 11 12 13 14 15 16 17 18
45 °C
37 °C
–32 70– 32 70– 3270– 32 70–32 70– 32 70
Run-off
Run-off
a
b
yieE
sixA
ycbK
mdoG
σE promoters
rpoH
hepA
rpsL
yciS
σ32 promoters
σ70 association
σ32 spacing
σ32 association
σ70 spacing
0
20
40
60
80
100
120
140
Occupancy units
No heat shock
Heat shock
No heat shock
Heat shock
hepArpsLyciS
0
50
100
150
200
250
300
Occupancy units
N13
N14
N15
N15
N15
N14
N14
N13
N13
N16
N17
N18
N16
N17
N18
N16
N17
N18
16
17
18
1314 15
2,000 1,600 1,200800 400
Distance to nearest
transcript 5′ end (bp)
0
0
1
2
3
4
5
6
7
8
9
Frequency (%)
Distance from transcript 5′ ends
Distance from random
genomic coordinates
Distance from σ32 peaks
Distance from random
probe coordinates
2,0001,6001,200 800 4000
0
5
10
15
20
25
30
35
40
Distance to nearest σ70 peak (bp)
Frequency (%)
3′
5′
5′
87654321
87654321
3′
2
1
0
2
1
0
a
d
e
f
bc
Figure 3 Extensive overlap between s70targets and targets of s32and sE. (a) s32binding motifs identified using BioProspector. (b) Distributions of
minimum distances to a s70target from s32target promoters and from 1,000 random probe positions. (c) Distributions of minimum distances between all
transcript 5¢ ends and transcript 5¢ ends of 254 randomly selected, non–divergently transcribed genes, and between all transcript 5¢ ends and 254 randomly
selected probe positions (see Methods). (d) Potential overlap between s32and s70promoter elements. Matched bases are indicated by vertical lines and
bold, underlined text in the consensus s70promoter elements. (e) Association of s32with promoters shared by s32and s70, before and after heat shock.
(f) Association of s70with s32- and sE-dependent promoters, before and after heat shock. Occupancy was measured as in Figure 1.
Figure 4 Transcription from heat-shock promoters by Es70and Es32in vitro.
(a) Autoradiogram shows full-size [32P]RNA products (run-off) from reconsti-
tuted transcription reactions (see Methods)40–42,44. RNAP core was mixed
with a three-fold molar excess of s70or s32(as indicated) and all four NTP
substrates. The reaction was initiated by adding an excess of the DNA tem-
plate (a PCR fragment with the indicated promoter) and carried out for 5 min
at 37 1C or 45 1C. Increasing the concentration of either s did not result in
a greater yield of RNA (data not shown), indicating that both mixtures con-
tained a saturating amount of active s molecules. The s70-specific promoter
T7A1 (lanes 1–3) was used as negative control for RNAPs32and as a
reference for RNAPs70activity. (b) Formation of the promoter-proximal
elongation complexes (EC) by Es70and Es32on three heat-shock promoter
templates (see Methods). Sizes of [32P]RNA products are indicated. Marker
was prepared by RNAP walking on T7A1 template with Es70(ref. 40).
ARTICLES
NATURE STRUCTURAL & MOLECULAR BIOLOGY
ADVANCE ONLINE PUBLICATION5