JB Accepted Manuscript Posted Online 2 May 2016 J. Bacteriol. doi:10.1128/JB.00062-16 Copyright © 2016, American Society for Microbiology. All Rights Reserved.
1 1
Rapid Curtailing of the Stringent Response by Toxin-Antitoxin Encoded mRNases
2
Chengzhe Tiana,*, Mohammad Roghanianb,*, Mikkel Girke Jørgensenc, Kim Sneppena, Michael
3
Askvad Sørensenb, Kenn Gerdesb, Namiko Mitaraia,#
4
Niels Bohr Institute, University of Copenhagen, Copenhagen, Denmark a; Department of Biology,
5
University of Copenhagen, Copenhagen, Denmark b; Department of Biochemistry and Molecular
6
biology, University of Southern Denmark, Odense, Denmark c.
7
Running Head: Curtailing of the Stringent Response by TAs
8
* Equal contribution
9
# Address correspondence to Namiko Mitarai, E-mail:
[email protected]
10
Abstract (Max 250 words)
11
Escherichia coli regulates its metabolism to adapt to changes in the environment, in particular to
12
stressful downshifts in nutrient quality. Such shifts elicit the so-called stringent response
13
coordinated by the alarmone guanosine tetra- and pentaphosphate [(p)ppGpp]. At sudden amino-
14
acid (aa) starvation, RelA [(p)ppGpp synthetase I] activity is stimulated by binding of uncharged
15
tRNAs to a vacant ribosomal site; the (p)ppGpp level increases dramatically and peaks within the
16
time scale of a few minutes. The decrease of (p)ppGpp level after the peak is mediated by the
17
decreased production of mRNA by (p)ppGpp associated transcriptional regulations, which reduces
18
the vacant ribosomal A-site and thus constitutes a negative feedback to the RelA dependent
19
(p)ppGpp synthesis. Here we showed that, at a sudden isoleucine starvation, this peak was higher in
20
an E. coli strain that lacks the 10 known mRNase-encoding TA modules present in the wt strain.
21
This observation suggested that toxins are part of the negative feedback to control the (p)ppGpp
22
2 level during early stringent response. We built a ribosome trafficking model to evaluate the fold of
23
increase in the RelA activity just after the onset of aa starvation. Combining this with a feedback
24
model between the (p)ppGpp level and the mRNA level, we obtained reasonable fits to the
25
experimental data for both strains. The analysis revealed that toxins are activated rapidly within a
26
minute after the onset of starvation, reducing the mRNA half-life by ∼30 %.
27
Importance (Max 120 words)
28
The early stringent response elicited by amino-acid starvation is controlled by a sharp increase of
29
the cellular (p)ppGpp level. Toxin-antitoxin encoded mRNases are activated by (p)ppGpp through
30
enhanced degradation of antitoxins. The present work shows that this activation happens at a very
31
short time scale, and the activated mRNases negatively affects the (p)ppGpp level. The proposed
32
mathematical model of (p)ppGpp regulation through mRNA level highlights the importance of
33
several feedback loops in the early (p)ppGpp regulation.
34
Introduction
35
Upon sudden nutritional stress, bacteria cope with the adverse conditions by rapidly producing
36
guanosine tetra- and pentaphosphate ((p)ppGpp), the master regulators of the stringent response (1-
37
4). The ”alarmone” (p)ppGpp coordinates multiple physiological processes, such as reduction and
38
adjustment of transcription and activation of biosynthetic pathways (5-9). One of the central topics
39
in stringent response research is to understand how (p)ppGpp triggers these processes at an
40
appropriate timescale.
41
Numerous measurements revealed that the early (p)ppGpp profile exhibits a sharp peak within a
42
few minutes after the onset of amino-acid (aa) starvation (10-18) (e.g. Fig. 1a). Fig. 1b shows the
43
3 main feedbacks in the (p)ppGpp regulation relevant for the early stringent response. (p)ppGpp is
44
synthesized by the synthetase RelA and degraded by the hydrolase SpoT. In Escherichia coli, SpoT
45
also has a weak synthetase activity, but RelA is the major producer of (p)ppGpp during aa
46
starvation (7). RelA activity is significantly increased under the presence of uncharged tRNAs and
47
ribosomes (1). Probably, the presence of an uncharged tRNA loaded at the ribosomal A-site allows
48
RelA to rapidly sense aa starvation. In addition, there are positive feedbacks acting on the level of
49
(p)ppGpp. The hydrolase activity of SpoT may be inhibited by high concentrations of (p)ppGpp
50
(19). This inhibition provides with a delay in the normal degradation process and facilitates the
51
accumulation of (p)ppGpp. Furthermore, (p)ppGpp has been shown to increase the activity of RelA
52
in vitro (20). These two effects are indicated as positive feedbacks on the (p)ppGpp concentration
53
(Fig. 1b, a gray arrow).
54
The decrease of (p)ppGpp after a few minutes is conjectured to be caused by the reduced
55
mRNA level, known to show a rapid change upon starvation (21, 22). This leads to a less RelA
56
synthetase activity through reducing the number of empty A-sites, forming a negative feedback (Fig.
57
1b, black arrows). The decrease of mRNAs may be caused by the (p)ppGpp mediated reduction of
58
transcription (21-23). On longer time scales, effects such as modulation of ribosome copy number
59
(24, 25) and the induction of amino-acid synthesis (26) will also affect the (p)ppGpp level, but here
60
we focus on the short time scale.
61
relBE, one of the 10 known type II Toxin-Antitoxin (TA) modules encoding mRNases (also
62
called mRNA interferases) in E. coli, is suggested to be activated upon starvation (27). Toxin RelE
63
cleaves mRNAs at the ribosomal A-site (28). In non-stressed, rapidly growing cells, RelE activity is
64
neutralized by the antitoxin RelB that forms a tight complex with RelE (29, 30). During aa
65
starvation, the protein synthesis rate is about 5% of that of non-starved condition in relBE+ strain,
66
while in the relBE− strain it is about 10% (27). This indicates that RelE is released to act as a part of
67
4 the stress response to rapid shutdown of translation. The release is thought to be mediated by
68
(p)ppGpp, as Lon protease that degrades antitoxin RelB, is known to exhibit increased activity
69
under nutritional stress (25, 27, 31). Under such conditions, activated Lon is also suggested to
70
degrade the other antitoxins (32, 33). Therefore, it is likely that all the 10 corresponding toxins are
71
to some degree activated during aa starvation.
72
These observations lead us to the hypothesis that TA modules may work not only as a
73
downstream component of the stringent response, but also as a component to provide negative
74
feedback to the (p)ppGpp level since activated toxins help to reduce mRNAs that are required to
75
activate RelA (Fig. 1b, blue arrows). Here, we experimentally measured the dynamics of the
76
(p)ppGpp level at aa starvation in the wild-type (wt) strain and in the strain lacking the 10 TA loci
77
(∆10TA) (32) to investigate the extent of the feedback as well as changes in the kinetics of the
78
stringent response. We then used quantitative modeling to analyze the observations in a two-step
79
procedure. In the first step, we evaluated the initial starvation signal by analyzing a stochastic
80
model of ribosome translation dynamics with explicit consideration of binding of uncharged tRNAs
81
to empty ribosomal A-sites. In the second step, we used the resulting estimate as an input to a
82
simple feedback model of (p)ppGpp control to study how this initial signal and the regulatory
83
feedbacks mediated the changes in (p)ppGpp and mRNA levels. Fitting the feedback models to the
84
experimental data showed that feedbacks shown in Fig. 1b is indeed enough to reproduce the
85
experimental data, and revealed that the toxins are likely to be activated immediately after onset of
86
starvation, within one minute.
5 87
Materials and Methods
88
In vivo (p)ppGpp measurement
89
To determine the (p)ppGpp content formed by wt (MG1655) and ∆10TA strains upon valine
90
induced isoleucine starvation we modified the method from Ref. (33). In brief, overnight cultures
91
were diluted 100-fold in 10 ml of MOPS glucose minimal medium supplemented with all
92
nucleobases (10 µg ml−1 of each) as previously described (34) and incubated at 37◦C with shaking.
93
MG1655 and ∆10TA strains were previously reported to have an identical growth rates in LB (32)
94
and the doubling time in MOPS glucose minimal medium was just over an hour (∼62 min) for both
95
strains. At OD600 0.5 cells were diluted 10-fold to OD600 0.05 and were left to grow shaking at
96
37◦C with H332PO4 (100 µCi/mL). After 2-3 generations (OD600 0.2-0.3) amino-acid starvation was
97
induced by the addition of valine (0.5 mg.mL−1). 100 µl samples were withdrawn before and 1, 2, 3,
98
5, 10, 20, and 30 minutes after addition of valine. The reactions were stopped by the addition of 20
99
µl of ice-cold 2M formic acid. A 10 µl aliquot of each reaction was loaded on PEI Cellulose TLC
100
plates (purchased from GE Healthcare) and separated by chromathography in 1.5M potassium
101
phosphate at pH 3.4. The TLC plates were revealed by PhosphoImaging (GE Healthcare) and
102
analysed using ImageQuant software (GE Healthcare). The increase in the levels of (p)ppGpp was
103
normalized to the basal levels (time=0) for each strain.
104
Ribosome trafficking model
105
Model description We modeled the ribosome traffic upon starvation by extending the previous
106
models for translation process in exponential growth (35-37). Summary of the model is presented in
107
the result section. In the model we considered one mRNA chain modeled as a one-dimensional
108
6 lattice sites, where each site represents one codon. The total length was chosen to be 300 sites based
109
on the average protein length (38). In every simulation, we randomly generated the mRNA
110
sequence as follows: the first codon is fixed to be the start codon; the remaining codons are
111
randomly generated with probability proportional to the usage of the amino-acids in E. coli (the
112
usage is available in Ref. (39)). We considered a pool of 15 ribosomes available for translation per
113
mRNA (Table S1). We assumed that each ribosome occupied 11 codons and translated the codon in
114
the middle(35). We did not consider degradation of mRNA, therefore focusing on the steady state
115
of the ribosome traffic.
116
The reactions in the ribosome trafficking model are illustrated in Fig. 2a. We considered that a
117
ribosome takes one of the following three states: (i) the ”free” state where a ribosome is not on the
118
mRNA chain (brown in Fig. 2a); (ii) the ”idle” state where a ribosome is on the mRNA but its A-
119
site is empty (yellow in Fig. 2a); and (iii) the ”elongating” state where a charged tRNA is bound to
120
the ribosomal A-site and the ribosome is ready for translocation (green in Fig. 2a). The possible
121
reaction rates are assigned as follows.
122
i When a ribosome is in the ”free” state (the first column in Fig. 2a): If there are no ribosomes
123
occluding the initiation (i.e., the center of all translating ribosomes are located at codon 12 or
124
later), a ”free” ribosome may initiate translation by binding to the start codon with rate kinit and
125
change its state to ”idle”.
126 127
ii When a ribosome is in the ”idle” state (the second column in Fig. 2a): An ”idle” ribosome at the codon coding amino-acid i may experience one of the following reactions: "elongating"
at rate
"idle" → "idle"+(p)ppGpp "free"
at rate
at rate
(1)
+ (1 − )
(2) (3)
128
7 Here, the first reaction (1) represents binding of a charged tRNA to ribosomal A-site, which
129
changes the ribosome state to ”elongating”. A charged cognate tRNA binds at rate kchci, where
130
ci is the charging level of tRNA for the amino-acid i. A non-cognate charged tRNA can also
131
bind at a small rate pmistrl, representing the mistranslation process. We set
132
where ξj is the usage of amino-acid j and the cj is the charging level for the corresponding tRNA.
133
The second reaction (2) represents the binding of an uncharged cognate tRNA at rate kun(1 − ci),
134
that triggers the production of (p)ppGpp by one unit (in arbitrary unit as we consider only the
135
fold change before and after starvation). We assume that uncharged tRNA is immediately
136
ejected from A-site, with ribosome remaining in ”idle” state.
137
The third reaction (3) represents the abortion at rate pabort, where the ribosome dissociates from
138
mRNA and joins the ”free” pool.
=
∑
,
139
iii When a ribosome is in the ”elongating” state (the third column in Fig. 2a): If the codon in front is
140
not occupied, an ”elongating” ribosome may translocate by one codon, eject the now uncharged
141
tRNA and change its state to ”idle” with rate kel. Otherwise, the ribosome will not translocate
142
and keeps its status.
143
Finally, when a ribosome finishes translating the last codon, it exits the mRNA chain and changes
144
its state to ”free” (the fourth column in Fig. 2a).
145
We implemented the model with Gillespie algorithm (40) and the parameter values used in the
146
simulation are listed in Table S1.
147
Data acquisition The RelA activity was calculated as the number of events where an uncharged
148
tRNA bound to the ribosomal A-site in a time window. The number of translating ribosomes was
149
calculated as the number of ribosomes (”idle” and ”elongating” states) on the mRNA chain. The
150
mistranslation fraction was computed by dividing the number of mistranslation by the number of
151
8 translocation. The trajectory and the relative ribosome occupancy in Fig. 2b and Fig. S2 were based
152
on a single simulation. The values for other figures were the average of 100 independent
153
simulations with randomly generated mRNA. We smoothed the data with a window size of 0.05s
154
for the time series plots. The relative RelA activities (η) were computed by dividing the average
155
RelA activity during the last second of simulation by the pre-starved steady state activity.
156
For trafficking models without abortion and mistranslation (Fig. S2), we set pabort and pmistrl to be
157
zero. For trafficking models for general amino-acid starvation (Fig. S6), we artificially set the usage
158
of one amino-acid to be the fraction of starvation and set the charging level for this amino-acid to
159
0.02 after starvation.
160
Parameter Fitting and Error of Fitting
161
We used a deterministic model of (p)ppGpp mediated feedback presented in Fig. 3a to reproduce
162
the experimental data. The parameters were determined by fitting to the experimental data as
163
follows.
164
The error of fitting for one strain was defined as
=
( )−
( ) ( )
165
where mexp is the average and σexp is the standard error of mean of the experimentally measured
166
(p)ppGpp level and msimu is the simulated number of (p)ppGpp. The error of fit was computed by
167
summing over all measured time points except for t = 0. In Fig. 1a and S4b, the overall error was
168
defined as the L2-norm of the errors for the two strains. In Fig. 3d and S4c, the error was the one
169
for fitting to the wt strain. In Fig. S5c, the error was the one for the ∆10TA strain. MatLab
170
(Mathworks) function fmincon was used to obtain the parameter values by minimizing the error of
171
9 fit and 100 independent minimizations with random initial guesses were carried out to search for the
172
globally optimal values.
173
Global mRNA Half Lives
174
The measurement of mRNA half-lives was done essentially as described by (41). Briefly, cells were
175
grown in M9 minimal medium supplemented with 0.2% glucose at 37°C to an OD450 of 0.5. The
176
culture was diluted to an OD450 of 0.1 when the first sample of 200 μl was taken and placed on ice
177
with 10% SDS and 20% TCA. Immediately after, 1 mg ml−1 rifampicin, 50 ug/ml nalidixic acid,
178
and [3H]uridine (100 mCi mmol−1 ) was added to the culture and sampling continued every 30 sec
179
for 20 min. Samples were boiled at 95°C for 5 min and applied to Nitrate cellulose filters
180
(Whatman) and washed with cold 10% cold TCA. The filters were transferred to vials and the
181
incorporated radioactivity was counted in a liquid scintillation counter. When half-lives were
182
measured during aa starvation, the cells were starved with 0.4 mg ml−1 SHX for one hour before
183
sampling.
184
Results
185
Early stringent response to isoleucine starvation
186
“Hyperactivation” of the toxin RelE in the cell was previously shown to be enough to impair the
187
activation of RelA (42). This was suggested to be due to the ribosome-dependent cleavage of
188
mRNA by RelE resulting in inhibition of translation, and thus inhibiting RelA from sensing the
189
charging state of the tRNAs (42). More recently, high levels of (p)ppGpp have been suggested to
190
10 activate Lon protease to degrade all type II antitoxins of E. coli (31). Based on these reports we
191
wanted to measure the (p)ppGpp profile during the early stringent response in the wt strain
192
(MG1655) and the isogenic strain with all 10 Toxin-Antitoxin loci deleted (∆10TA) to quantify the
193
strength of the toxin-mediated negative feedback loop (Fig. 1a). We induced isoleucine starvation
194
by supplementing excessive valine to the medium (43) at time zero. We quantified the levels of
195
(p)ppGpp at 1, 2, 3, 5, 10, 20 and 30 minutes after the addition and we normalized the data with the
196
pre-starved level of (p)ppGpp for each strain (Materials and Methods, Fig. S1). Agreeing with the
197
previous reports (10, 12), the (p)ppGpp profiles of both strains exhibited a typical shape of
198
(p)ppGpp accumulation (Fig. 1a). The (p)ppGpp level in the ∆10TA strain reached the maximal
199
value of around 13-fold increase in 5 minutes, while the level in wt strain reached the maximal
200
value of 8-fold within that time frame. The visible difference in 5-minute samples suggests that
201
toxins are released rapidly after the starvation signal. Interestingly, the same measurement for the
202
∆relBE strain showed the (p)ppGpp profile similar to that of the wt strain, suggesting that this early
203
time difference is not solely determined by the relBE locus (Fig. S8).
204
Quantitative analysis by mathematical modeling
205
We aimed to quantitatively understand the obtained (p)ppGpp profiles (Fig. 1a). The first signal of
206
aa starvation is the activation of RelA. This is promoted by the binding of uncharged tRNAs to the
207
ribosomal A-sites, which results in (p)ppGpp production per mRNA. This leads to the above
208
mentioned feedback loops to take effect and alter the (p)ppGpp levels further. For a quantitative
209
analysis, we needed to separate the initial direct signal of the aa starvation from the effect of the
210
feedbacks. We therefore quantified the early stringent response in two steps. In the first step, we
211
analyzed a ribosome trafficking model (Fig. 2a; Materials and Methods for detail) where the
212
binding of charged and uncharged tRNAs to the translating ribosomes were explicitly considered.
213
11 From the model, we evaluated the initial direct signal of aa starvation (η) defined as the fold change
214
in the (p)ppGpp production rate per mRNA just after the aa starvation. No regulatory feedbacks nor
215
effects of toxins were considered at this step. In the next step, we used this η as an input to the
216
(p)ppGpp regulatory feedback model where we would study the effect of feedbacks.
217 218
A ribosome trafficking model quantified the starvation signal.
219
The ribosome trafficking model simulated the movements of multiple ribosomes on a typical
220
mRNA of 300 codons in length (38). We model the charging levels for tRNA cognate to each
221
amino-acid explicitly, and the codons of the mRNA chain were randomly chosen with the
222
probability proportional to the usage of their cognate amino-acids (the usage of the isoleucine
223
codons being around 6% (39)). We excluded mRNA degradation and cleavage as these processes
224
took place in a slower timescale (see below). Consequently, we cannot distinguish between wt and
225
Δ10TA strains here. Each ribosome stochastically initiates translation with a rate kinit when there is
226
no occlusion by the other ribosomes. The kinetics of the translating ribosome was modeled with
227
several possible reaction steps to address starvation-induced (p)ppGpp production. When
228
translating a codon, a ribosome can be bound by a charged cognate tRNA with a rate kchc with c
229
being the charging level of the corresponding tRNA, or can be bound by an uncharged cognate
230
tRNA with a rate kunu with u = 1 − c denoting the fraction of the uncharged cognate tRNA. If an
231
uncharged tRNA is bound, (p)ppGpp molecules are produced and the tRNA is ejected, while if a
232
charged tRNA is bound, the ribosome moves to the next codon at a rate kel as long as the codon in
233
front is not occupied by another ribosome. When a ribosome reaches the end of the mRNA, it is
234
released to join the free ribosomes ready to initiate another round of translation. We also accounted
235
for mistranslation (44) and the abortion of translating ribosome mediated by tmRNA (45), but with
236
12 realistic parameters they did not have a significant effect on the result (see below). Parameters were
237
chosen to fit to the previously published data (Table S1).
238
Fig. 2b left panel presents a typical spatiotemporal plot of ribosome traffic with isoleucine
239
starvation at time zero. The charging levels of all tRNA in the pre-starved state were set to be c =
240
0.8 (23, 46), and we reduced the charging level of tRNAIle to cIle = 0.02 (Table S1) from time zero
241
and kept the levels of other tRNA unchanged to simulate the isoleucine starvation. The immediate
242
reduction in the charging level can be justified by the rapid turnover of tRNA (47). We see that
243
ribosomes elongate along the mRNA chain rapidly in the pre-starved state (trajectories shown in
244
blue) with few events of (p)ppGpp synthesis (shown in red). Upon starvation, (p)ppGpp synthesis
245
events occur more frequently. The average translation elongation time for ribosomes however, is
246
decreased by 20% due to frequent stalling, a number similar to the previously reported 15%
247
reduction (21). To visualize the ribosome movement further, we plotted the relative occupancy of
248
each mRNA codon by ribosomes in the post-starved state (Fig. 2b, right panel). Ribosomes were
249
shown to occupy isoleucine codons (black dash lines) more often than the unstarved ones, and
250
ribosome jamming was visualized by the stair-like occupancy curve in the upstream of the starved
251
codons. No visible difference was observed between the occupancy at unstarved codons close to the
252
5’ end and the ones close to the 3’ end, meaning that the translation abortion was rare.
253
To quantify the starvation signal, we computed the average statistics of 100 independent
254
simulations. The relative RelA activity (η), defined as the ratio of the frequency of (p)ppGpp
255
synthesis events in the post-starved state to the pre-starved one, increased by around 8 folds almost
256
immediately upon starvation (Fig. 2c). This fast convergence pinpointed that the depletion of
257
amino-acid pool can be rapidly and accurately sensed through the binding of uncharged tRNAs, and
258
suggested that the initial increase of (p)ppGpp synthesis upon aa starvation can be decoupled from
259
slower biological processes including the mRNA cleavage and regulatory feedbacks (Fig. 1b).
260
13 We fitted the parameters so that the fraction of mistranslated isoleucine codons in the pre-
261
starved state to be 0.5% (44). We found that mistranslation did not have a significant impact on the
262
relative RelA activity, as a trafficking model without this process generated a 10-fold increase in
263
the RelA activity after the onset of starvation (Fig. S2, S3d).
264
The fold change of the (p)ppGpp synthesis upon starvation η was sensitive to the charging level of
265
isoleucyl-tRNA at the starvation, as cIle = 0.04 gives a 5-fold increase and cIle = 0 gives a 15-fold
266
increase (Fig. 2d). We here adopted cIle = 0.02 as a typical value, based on the measurement that the
267
charging level of tRNALeu being between 1% and 3% upon leucine starvation (23) and the
268
assumption of a similar value for isoleucine starvation due to the similarity in the amino-acid
269
structures. This gave η = 8, and we used this value as an immediate input to the control of (p)ppGpp
270
at aa starvation as we analyzed a model of the (p)ppGpp regulatory feedbacks in the next section.
271
A feedback model of early stringent response suggests a rapid activation of toxins.
272
In the second step of our quantitative analysis, we looked at how the initial aa starvation signal and
273
the regulatory feedbacks mediated the early dynamics of (p)ppGpp levels by constructing a simple
274
feedback model. The model equations were formulated in Fig. 3a, where we considered the
275
concentration of (p)ppGpp, [P], and the mRNA, [M]. We assumed that the toxins are not active
276
before the aa starvation in wt, therefore no difference in model parameters before the starvation. We
277
modulated the mRNA degradation rate in the wt strain after aa starvation (see below, Fig. 3b right
278
panel), as mRNAs should have shorter half-lives in the wt strain with the toxins’ mRNase activity.
279
The synthesis rate of (p)ppGpp was formulated in two terms: the first term accounted for basal
280
production including the SpoT-mediated synthesis, and the second one for the RelA-mediated
281
production proportional to the mRNA level. The starvation dependence of this term was represented
282
in the parameter η. η equals to one before starvation and increases to another constant at the
283
starvation to represent the fold-change of the (p)ppGpp synthesis rate per mRNA after starvation.
284
14 For isoleucine starvation, we chose the post-starvation value of η to be 8 based on the presented
285
ribosome trafficking model analysis (Fig. 3b left panel).
286
This value of η suggests that the starvation signal alone is not enough to induce a 13-fold
287
increase in the (p)ppGpp profile for the ∆10TA strain (Fig. 1a). We therefore needed to include
288
some positive feedbacks on the (p)ppGpp concentration to get the peak high enough.
289
The degradation of (p)ppGpp was modeled with saturation in the Michaelis-Menten form to
290
capture the starvation-dependent hydrolysis activity (19). It turned out that this saturated
291
degradation could provide enough ”positive feedback” to accumulate (p)ppGpp more than the fold
292
increase of η to fit the data, so for simplicity we focused on this result in the main text. The version
293
of the model with an additional positive feedback of (p)ppGpp synthesis (20) is presented in Fig. S4,
294
which also gave a reasonable fit.
295
Note that we also analyzed a model without any positive feedback but with high value of η,
296
because it is possible to produce a strong starvation signal by modulating tRNA charging levels
297
(Fig. 2d) at the starvation. However, a good fitting to the experimental curves required an
298
unrealistically long mRNA half-life (Fig. S5). Thus, we concluded that the model with lower η with
299
some positive feedbacks is more realistic.
300
For the mRNA level, we modeled the transcription rate with [P]-dependent repression of
301
Michaelis-Menten form. We assumed a constant degradation of mRNA with a half-life of 2 minutes
302
(dm = ln2/(mRNA half-life))(48, 49). For the wt strain, we considered the additional toxin-mediated
303
mRNA cleavage by modulating the half-lives of mRNA. By changing the half-lives of mRNA from
304
2 minutes to 2ν−1 min at time τ min after starvation (Fig. 3b right panel), we accounted for both the
305
timescale of toxin activation (τ) and the strength of the mRNA cleavage activity (ν).
306
We obtained the parameter values from the literature and by fitting to the experimental curves
307
(Materials and Methods, Table S2). We obtained reasonable fits to the data for both strains (Fig. 1a).
308
15 Our model illustrated that the saturated degradation of the (p)ppGpp level can provide a delay to the
309
peak to fit the 5-minute timescale of the (p)ppGpp peak. Without the saturated degradation, even
310
with a sufficiently strong starvation signal, a good fit required a long mRNA half-life during the
311
rise of the peak but a short one during the fall of the peak even for ∆10TA strain, a phenomenon
312
unlikely to occur (data not shown).
313
The best fit to the wt strain gave the estimate that the toxins were released shortly (τ ∼ 10
314
seconds) after isoleucine starvation and the mRNA half-lives decreased by around 30% (ν ≈ 1.4).
315
This gave a faster drop in the mRNA level in the wt strain than in the ∆10TA strain (Fig. 3c),
316
reducing the (p)ppGpp peak height (Fig. 1a). To verify this further, we sampled the values of τ and
317
ν in a wide range and computed the distance between the simulated curves and the experimental
318
(p)ppGpp profile (or ”error-of-fit”, refer to Material and Methods). Fig. 3d shows that a good fit
319
required the effective half-life of mRNA to be reduced by 20%-40% and toxins to be released
320
within one minute after starvation. Therefore, our analysis suggests a rapid and mild mRNA
321
cleavage activity is induced from the Toxin-Antitoxin complexes in the cytoplasm.
322
Our model could generate predictions consistent with experimental measurements. For example,
323
the model predicted that the wt strain shows a four-fold reduction in the transcription rate
324
(computed by the term
325
(Fig. 3). Earlier literatures reported a similar rapid transcription inhibition (21, 22, 50). O’Farrell
326
(50) also reported that the transcription rate of the wt strain dropped to 30%-35% of the unstarved
327
level under isoleucine starvation, and this number was consistent with our model’s prediction.
328
Furthermore, if we assumed that the translation rate (aa incorporation rate) could be estimated by
329
the product of the mRNA level and the translation elongation rate, we may predict translation rates
330
from our model. We showed earlier that ribosome trafficking model predicted that the translation
331
elongation rate dropped by 20% after aa starvation. Combined with the mRNA levels, we predicted
/(
+
)) within 10 minutes after the onset of isoleucine starvation
332
16 that the translation rate dropped to 15% of the pre-starved level. This number is in accordance with
333
earlier reported 10%-13% (50).
334
Discussion
335
In this manuscript, we combined experimental measurements and modeling to study the early
336
dynamics of the (p)ppGpp levels after isoleucine starvation. Our models demonstrates that the
337
feedbacks through the mRNA level are enough to reproduce the (p)ppGpp peak. We also showed
338
that the TA systems not only modulate physiological activities in the post-starved steady state (27),
339
but also serve as a negative feedback on the (p)ppGpp levels during early stringent response. Our
340
analysis predicts that activation of toxins occurs rapidly (within a minute) after the onset of
341
starvation and, on average, induces a ∼30% reduction of the mRNA half-life.
342
Although our modeling framework was developed for valine-induced isoleucine starvation, it
343
can be applied to a wide range of experimental setups by setting the charging levels of tRNAs and
344
the fraction of the starved codons explicitly. The ribosome trafficking model predicts a sublinear
345
increase of relative RelA activity η on the fraction of starved amino-acid compared to total amino-
346
acid usage (e.g., 2% for histidine starvation, 16% for isoleucine and leucine starvation, Fig S6a).
347
The early (p)ppGpp response curve in different situations can be obtained by applying the obtained
348
fold increase of RelA activity η to the proposed model, which naturally predicts the increasing
349
(p)ppGpp peak height with starvation of more frequently used amino-acid (Fig. 4). Following the
350
procedure described in the result section, one may also utilize our modeling framework to estimate
351
the mRNA levels and translation rates shortly after aa starvation. For example, our model predicts a
352
3-fold reduction in the mRNA level and a 70% drop in translation rate within 10 minutes after the
353
17 onset of histidine starvation, while the numbers are 10-fold and 95% for isoleucine/leucine
354
starvation (Fig. S6de).
355
Our findings on the rapid activation of the toxins by (p)ppGpp supports the idea that the TA
356
systems could serve as an immediate protection mechanism against stress. Meanwhile, it is
357
probably uneconomical for cells to enter a complete shutdown of translation regardless of the extent
358
of the adversity. This suggests that rather weak mRNA cleavage activity from the toxins probably
359
facilitates the cells to enter a temporary status (for aa starvation, the ”hunger state” as suggested in
360
(18)), to evaluate the extent of austerity, ensuring that the cell may recover without a high cost in
361
metabolism if the stress is modest or transient.
362
Since we focused only on the early time behavior of aa starvation, our model might be
363
inconsistent with experimental data in the long-term behavior. Our model predicts that the Δ10TA
364
strain has a lower transcription rate (due to higher (p)ppGpp levels) and longer mRNA half-lives
365
(due to the absence of toxins) than the wt strain, resulting in the two strains to have similar mRNA
366
levels and translation rates (Fig. 3c). This prediction does not match the previous report where RelE
367
alone was shown to reduce translation rates by 2-fold in the long term (27). This inconsistency can
368
be explained by long-term behaviors of the transcription rates and mRNA half-life which were not
369
considered in the current model. Firstly, the model parameters related to transcription rates were
370
fitted to early stringent response when the down-regulation of growth-related genes is dominant (26,
371
51). Consequently, the model prediction primarily reflected the differences between the two strains
372
in the expression of these genes, and the (p)ppGpp-dependent up-regulation of stress-response
373
genes in the post-starved steady states was less covered. If we also account for the stress-response
374
genes, the transcription rate may exhibit a smaller difference between the two strains than model
375
prediction. Secondly, our model did not consider many factors modulating the global mRNA
376
stability in the post-starved state. In reality, the change in mRNA species (26) and protection of
377
18 mRNA from degradation by stalled ribosomes (41, 52) may increase the stability. To check this, we
378
measured the average mRNA half-lives in the pre- and post-starved (60 min after the aa starvation)
379
states (Materials and Methods). We found that mRNA half-lives in the Δ10TA strain indeed
380
increased by two-folds from 1.8 min before starvation to 4.4 min in the post-starved steady states.
381
Meanwhile, the wt strain showed less stabilization (from 1.5 min to 2.4 min) (Fig. S7). As a result,
382
the mRNA decay rate in the wt strain was two-fold higher than the Δ10TA strain in the post-starved
383
states. In summary, the aspect that the two strains may have a similar transcription rate but a two-
384
fold difference in mRNA decay rate indicates a two-fold difference in the cellular mRNA levels in
385
the long term, explaining the translation rate differences in the two strains. It would be interesting to
386
extend our present model to include these effects so that it becomes also applicable to long-term
387
behavior after aa starvation.
388
Acknowledgments
389
Authors thank Steen Pedersen for fruitful discussions.
390
Funding Information
391
This work was supported by the Danish National Research Foundation Centre of Excellence
392
BASP (DNRF120) (MR, MAS, KG, NM), CMoL (CT, KS, NM), a Novo Nordisk Foundation
393
Laureate
394
Research Grant (MR,KG) and the ERC Advanced Investigator Grant PERSIST (294517) (MR, KG).
19 395
References
396
1.
Haseltine WA, Block R. 1973. Synthesis of guanosine tetra-and pentaphosphate requires
397
the presence of a codon-specific, uncharged transfer ribonucleic acid in the acceptor site of
398
ribosomes. Proc Natl Acad Sci U S A 70:1564-1568.
399
2.
400 401
11:181-185. 3.
402 403
Gallant J, Palmer L, Pao CC. 1977. Anomalous synthesis of ppGpp in growing cells. Cell
Battesti A, Bouveret E. 2006. Acyl carrier protein/SpoT interaction, the switch linking SpoT-dependent stress response to fatty acid metabolism. Mol Microbiol 62:1048-1063.
4.
404
Vinella D, Albrecht C, Cashel M, D’Ari R. 2005. Iron limitation induces SpoT-dependent accumulation of ppGpp in Escherichia coli. Mol Microbiol 56:958-970.
405
5.
Gallant JA. 1979. Stringent control in E. coli. Annu Rev Genet 13:393-415.
406
6.
Potrykus K, Cashel M. 2008. (p) ppGpp: Still Magical?*. Annu Rev Microbiol 62:35-51.
407
7.
Dalebroux ZD, Swanson MS. 2012. ppGpp: magic beyond RNA polymerase. Nat Rev
408 409
Microbiol 10:203-212. 8.
Ross W, Vrentas CE, Sanchez-Vazquez P, Gaal T, Gourse RL. 2013. The magic spot: a
410
ppGpp binding site on E. coli RNA polymerase responsible for regulation of transcription
411
initiation. Molecular cell 50:420-429.
412
9.
413 414 415
Scott M, Klumpp S, Mateescu EM, Hwa T. 2014. Emergence of robust growth laws from optimal regulation of ribosome synthesis. Molecular systems biology 10:747.
10.
Ryals J, Little R, Bremer H. 1982. Control of rRNA and tRNA syntheses in Escherichia coli by guanosine tetraphosphate. J Bacteriol 151:1261-1268.
416
11.
20 Cashel M. 1969. The control of ribonucleic acid synthesis in Escherichia coli: IV. relevance
417
of unusual phosphorylated compounds from amino acid-starved stringent strains. J Biol
418
Chem 244:3133-3141.
419
12.
Lagosky PA, Chang FN. 1980. Influence of amino acid starvation on guanosine 5'-
420
diphosphate 3'-diphosphate basal-level synthesis in Escherichia coli. J Bacteriol 144:499-
421
508.
422
13.
423 424
levels in stringent and relaxed strains of Escherichia coli. J Biol Chem 246:4381-4385. 14.
425 426
Lazzarini RA, Cashel M, Gallant J. 1971. On the regulation of guanosine tetraphosphate
Hughes J, Mellows G. 1978. Inhibition of isoleucyl-transfer ribonucleic acid synthetase in Echerichia coli by pseudomonic acid. Biochem J 176:305-318.
15.
Metzger S, Schreiber G, Aizenman E, Cashel M, Glaser GJ. 1989. Characterization of
427
the relA1 mutation and a comparison of relA1 with new relA null alleles in Escherichia coli.
428
J Biol Chem 264:21146-21152.
429
16.
430 431
Lund E, Kjeldgaard NO. 1972. Metabolism of guanosine tetraphosphate in Escherichia coli. Eur J Biochem 28:316-326.
17.
Fiil NP, Willumsen BM, Friesen JD, von Meyenburg K. 1977. Interaction of alleles of
432
the relA, relC and spoT genes in Escherichia coli: Analysis of the interconversion of GTP,
433
ppGpp and pppGpp. Mol Gen Genet 150:87-101.
434
18.
Traxler MF, Zacharia VM, Marquardt S, Summers SM, Nguyen H-T, Stark SE,
435
Conway T. 2011. Discretely calibrated regulatory loops controlled by ppGpp partition gene
436
induction across the ‘feast to famine’ gradient in Escherichia coli. Mol Microbiol 79:830-
437
845.
438 439
19.
Murray DK, Bremer H. 1996. Control of spoT-dependent ppGpp synthesis and degradation in Escherichia coli. J Mol Biol 259:41-57.
440
20.
21 Shyp V, Tankov S, Ermakov A, Kudrin P, English BP, Ehrenberg M, Tenson T, Elf J,
441
Hauryliuk V. 2012. Positive allosteric feedback regulation of the stringent response enzyme
442
RelA by its product. EMBO Rep 13:835-839.
443
21.
Sørensen MA, Jensen KF, Pedersen S. 1994. High concentrations of ppGpp decrease the
444
RNA chain growth rate: implications for protein synthesis and translational fidelity during
445
amino acid starvation in Escherichia coli. J Mol Biol 236:441-454.
446
22.
Svitil AL, Cashel M, Zyskind JW. 1993. Guanosine tetraphosphate inhibits protein
447
synthesis in vivo. A possible protective mechanism for starvation stress in Escherichia coli.
448
J Biol Chem 268:2307-2311.
449
23.
Sørensen MA. 2001. Charging levels of four tRNA species in Escherichia coli Rel+ and
450
Rel- strains during amino acid starvation: a simple model for the effect of ppGpp on
451
translational accuracy. J Mol Biol 307:785-798.
452
24.
453 454
Bremer H, Dennis P. 2008. Feedback control of ribosome function in Escherichia coli. Biochimie 90:493-499.
25.
Kuroda A, Nomura K, Ohtomo R, Kato J, Ikeda T, Takiguchi N, Ohtake H, Kornberg
455
A. 2001. Role of inorganic polyphosphate in promoting ribosomal protein degradation by
456
the Lon protease in E. coli. Science 293:705-708.
457
26.
458 459
stringent response in Escherichia coli. J Bacteriol 190:1084-1096. 27.
460 461
Durfee T, Hansen A-M, Zhi H, Blattner FR, Jin DJ. 2008. Transcription profiling of the
Christensen SK, Mikkelsen M, Pedersen K, Gerdes K. 2001. RelE, a global inhibitor of translation, is activated during nutritional stress. Proc Natl Acad Sci U S A 98:14328-14333.
28.
Pedersen K, Zavialov AV, Pavlov MY, Elf J, Gerdes K, Ehrenberg M. 2003. The
462
bacterial toxin RelE displays codon-specific cleavage of mRNAs in the ribosomal A site.
463
Cell 112:131-140.
464
29.
465 466
22 Galvani C, Terry J, Ishiguro EE. 2001. Purification of the RelB and RelE proteins of Escherichia coli: RelE binds to RelB and to ribosomes. J Bacteriol 183:2700-2703.
30.
Pedersen K, Christensen SK, Gerdes K. 2002. Rapid induction and reversal of a
467
bacteriostatic condition by controlled expression of toxins and antitoxins. Mol Microbiol
468
45:501-510.
469
31.
470 471
persistence by stochastic induction of toxin-antitoxin activity. Cell 154:1140-1150. 32.
472 473
Maisonneuve E, Castro-Camargo M, Gerdes K. 2013. (p) ppGpp controls bacterial
Maisonneuve E, Shakespeare LJ, Jorgensen MG, Gerdes K. 2011. Bacterial persistence by RNA endonucleases. Proc Natl Acad Sci U S A 108:13206-13211.
33.
Germain E, Roghanian M, Gerdes K, Maisonneuve E. 2015. Stochastic induction of
474
persister cells by HipA through (p) ppGpp-mediated activation of mRNA endonucleases.
475
Proc Natl Acad Sci U S A 112:5171-5176.
476
34.
477 478
Methods Mol Genet 3:341-356. 35.
479 480
36.
Ciandrini L, Stansfield I, Romano MC. 2010. Role of the particle’s stepping cycle in an asymmetric exclusion process: a model of mRNA translation. Physical Review E 81:051904.
37.
483 484
Mitarai N, Sneppen K, Pedersen S. 2008. Ribosome collisions and translation efficiency: optimization by codon usage and mRNA destabilization. J Mol Biol 382:236-245.
481 482
Cashel M. 1994. Detection of (p) ppGpp accumulation patterns in Escherichia coli mutants.
Mitarai N, Pedersen S. 2013. Control of ribosome traffic by position-dependent choice of synonymous codons. Phys Biol 10:056011.
38.
Sundararaj S, Guo A, Habibi-Nazhad B, Rouani M, Stothard P, Ellison M, Wishart
485
DS. 2004. The CyberCell Database (CCDB): a comprehensive, self-updating, relational
486
database to coordinate and facilitate in silico modeling of Escherichia coli. Nucleic Acids
487
Res 32:D293-D295.
488
39.
23 Pramanik J, Keasling JD. 1997. Stoichiometric model of Escherichia coli metabolism:
489
incorporation of growth-rate dependent biomass composition and mechanistic energy
490
requirements. Biotechnol Bioeng 56:398-421.
491
40.
492 493
Gillespie DT. 1977. Exact stochastic simulation of coupled chemical reactions. J Phys Chem 81:2340-2361.
41.
Pato ML, Bennett PM, von Meyenburg K. 1973. Messenger Ribonucleic Acid Synthesis
494
and Degradation in Escherichia coli During Inhibition of Translation. J Bacteriol 116:710-
495
718.
496
42.
497 498
of RelE. Mol Microbiol 53:587-597. 43.
499 500
44.
45.
46.
509
Dittmar KA, Sørensen MA, Elf J, Ehrenberg M, Pan T. 2005. Selective charging of tRNA isoacceptors induced by amino-acid starvation. EMBO Rep 6:151-157.
47.
507 508
Keiler KC, Waller PRH, Sauer RT. 1996. Role of a peptide tagging system in degradation of proteins synthesized from damaged messenger RNA. Science:990-993.
505 506
Parker J. 1989. Errors and alternatives in reading the universal genetic code. Microbiol Mol Biol Rev 53:273-298.
503 504
Umbarger HE, Brown B. 1955. Isoleucine and valine metabolism in Escherichia coli V: Antagonism between isoleucine and valine. J Bacteriol 70:241-248.
501 502
Christensen SK, Gerdes K. 2004. Delayed-relaxed response explained by hyperactivation
Gouy M, Grantham R. 1980. Polypeptide elongation and tRNA cycling in Escherichia coli: a dynamic approach. FEBS Lett 115:151-155.
48.
Pedersen S, Reeh S, Friesen JD. 1978. Functional mRNA half lives in E. coli. Mol Gen Genet 166:329-336.
510
49.
24 Pedersen M, Nissen S, Mitarai N, Svenningsen SL, Sneppen K, Pedersen S. 2011. The
511
functional half-life of an mRNA depends on the ribosome spacing in an early coding region.
512
J Mol Biol 407:35-44.
513
50.
514 515
O'Farrell PH. 1978. The suppression of defective translation by ppGpp and its role in the stringent response. Cell 14:545-557.
51.
Bremer H, Dennis PP. 1996. Modulation of chemical composition and other parameters of
516
the cell by growth rate. In Neidhardt F (ed), Escherichia coli and Salmonella. ASM Press,
517
Washington D.C.
518
52.
Reimers JM, Schmidt KH, Longacre A, Reschke DK, Wright BE. 2004. Increased
519
transcription rates correlate with increased reversion rates in leuB and argH Escherichia coli
520
auxotrophs. Microbiol 150:1457-1466.
521 522
53.
Goldman E, Jakubowski H. 1990. Uncharged tRNA, protein synthesis, and the bacterial stringent response. Mol Microbiol 4:2035-2040.
523
Figure Legends
524
Figure 1. Early stringent response to isoleucine starvation. a. (p)ppGpp response of wild-type
525
MG1655 strain (wt) and the strain with all 10 known Toxin-Antitoxin loci deleted (∆10TA) to
526
valine-induced isoleucine starvation at t = 0. The levels of (p)ppGpp were normalized to the pre-
527
starved level for each strain. The triangle (wt) and square (∆10TA) symbols represent the average
528
fold-increase in three independent measurements and the error bars represent the standard error.
529
The simulated levels of (p)ppGpp from the model of the early stringent response were plotted as the
530
blue (wt) and red (∆10TA) curves. b. A schematic illustration of the regulatory network. (p)ppGpp
531
synthesis is triggered by sudden aa starvation imposed on rapidly growing cells. The level of
532
(p)ppGpp is self-amplified by RelA and SpoT and (p)ppGpp induces transcriptional inhibition.
533
25 (p)ppGpp also activates Lon protease to release toxins from the Toxin-Antitoxin complexes to
534
deplete mRNA.
535
Figure 2. The ribosome trafficking model. a. A schematic description of the ribosome trafficking
536
model. Free ribosomes may initiate translation with rate kinit if the start codons of mRNA are not
537
occupied. Every translating ribosome may (1) bind a charged cognate tRNA with rate kchc where c
538
is the charging level of the tRNA and elongate one codon with rate kel, (2) bind an uncharged
539
cognate tRNA, eject and produce (p)ppGpp (increase RelA activity by one unit) with rate kunu
540
where u = 1 − c is the uncharging level, or (3) abort translation with pabort. We also allow non-
541
cognate tRNA to bind ribosomal A-sites with a small rate pmistrl to represent mistranslation.
542
Elongation only occurs if the codon in front is free. Translation is terminated when a ribosome
543
finishes translating all the codons and exits the mRNA. A detailed description of the algorithm is
544
available in the Material and Methods section. b. Ribosome occupancy on mRNA. Left panel: A
545
spatiotemporal plot of ribosome traffic on a randomly generated mRNA upon isoleucine starvation
546
at t = 0. The lines represent the movement of ribosomes. The blue lines represent normal translation
547
and the red lines represent (p)ppGpp synthesis. Right panel: The relative mean occupancy of
548
mRNA codons by ribosomes in the post-starved steady state. The dash lines represent the location
549
of isoleucine codons. c. The relative RelA activity (η) in the pre- and post-starved states. The
550
averages of 100 independent simulations were plotted in the red curves. The dash lines represent
551
the induction time of starvation. d. The level of isoleucine starvation was modulated by sampling
552
the uncharging level of tRNAIle (uIle) between 0 and 1. The relative RelA activity (η) in the post-
553
starved steady states were plotted. The figure confirms the previous claim that the stringent
554
response is significant when the more than 85% of tRNA are uncharged (53).
26 555
Figure 3. A model of early stringent response. a. The formulation of the model. b. Left panel:
556
illustration of the relative RelA activity (η) upon isoleucine starvation at time zero. Right panel:
557
illustration of the average mRNA half life (= ln 2/dm) upon isoleucine starvation at time zero for
558
both the wt strain (blue) and the ∆10TA strain (red). c. The simulated concentration of mRNA for wt
559
(blue) and ∆10TA (red) strains. (refer to Fig. 1a) d. The effects of the timescale (τ) and strength (ν)
560
of mRNA cleavage activity by toxins on the fitting error to the wt measurements.
561
Figure 4. The predicted ppGpp level upon general amino acid starvation. The charging level of
562
the starved codons was set to be 0.02 during starvation. The percentages in the legend represents the
563
usage of the starved amino-acids and the values of η indicate the strengths of the corresponding
564
starvation signal. The prediction assumes that the toxins are released immediately upon starvation
565
and the reduction of mRNA half-lives by toxins is not modulated by the number of starved codons.
a wt (simu) wt (exp)
16
(p)ppGpp (Fold)
14
∆10TA (simu) ∆10TA (exp)
12 10 8 6 4 2 0 −5
0
5
10
15
20
25
30
Time (min) b
Aa shock (η)
(p)ppGpp
mRNA
Toxin
35
a
congate tRNA
Charged
Uncharged
(Fraction c)
(Fraction u=1-c)
wrong tRNA
(2) Initiation kinit
kunu
(3)
kchc
i. Free
b
ppGpp
Exit
Abortion pabort
(1)
kel
mistranslation pmistrl
ii. Idle
iii. Elongating
300
mRNA Codons
250
200
150
100
50
0 -0.5
0
0.5
0.5
1.5
2
d
c 10
16
RelA Activity (Fold, η)
RelA Activity (Fold, η)
1
Relative Occupancy
Time (min)
8 6 4 2 0 -0.5
0
Time (min)
0.5
14 12 10 8 6 4 2 0 0
0.2
0.4
0.6
0.8
Uncharged Fraction of tRNA
1
a
b
η
mRNA Half-life (min)
8
2 2ν-1
1 Time
0
c
d 1000 800
τ
2
6
1.8
∆10TA (simu)
Time Error of Fitting
5 1.6
600
4
ν
mRNA (µ m−3)
wt (simu)
0
400
1.4
200
1.2
0
1
3
0
10
20
Time (min)
30
2
0
2
4
6
τ (min)
8
10
25
−His (2%, η=4) −Ile (6%, η=8) −Ala (10%, η=12) −Ile,Leu (16%, η=17.5)
(p)ppGpp (Fold)
20
15
10
5
0 −5
0
5
10
15
Time (min)
20
25
30
35