an evaluation of genetically encoded fret-based biosensors for quantitative metabolite analyses in...
TRANSCRIPT
B
Aq
RQ1
MI
a
ARRAA
KGEECFI
1
(atetahwst2G
ge
(((
h0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
ARTICLE IN PRESSG ModelIOTEC 6762 1–10
Journal of Biotechnology xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Journal of Biotechnology
j ourna l ho me page: www.elsev ier .com/ locate / jb io tec
n evaluation of genetically encoded FRET-based biosensors foruantitative metabolite analyses in vivo
oland Moussa, Anna Baierl, Victoria Steffen, Tina Kubitzki, Wolfgang Wiechert,artina Pohl ∗
BG-1: Biotechnology, Forschungszentrum Jülich GmbH, 52425 Jülich, Germany
r t i c l e i n f o
rticle history:eceived 1 May 2014eceived in revised form 25 June 2014ccepted 7 July 2014vailable online xxx
eywords:FP
a b s t r a c t
A broad range of genetically-encoded fluorescence biosensors has been developed, allowing the detectionof signaling intermediates and metabolites in real time. Many of these biosensors are based on FoersterResonance Energy Transfer (FRET). The two biosensors of the well-known “Venus-flytrap” type exemplar-ily studied in this work are composed of a central sugar binding protein flanked by two green fluorescentprotein derivatives, namely ECFP as well as Citrine and EYFP, respectively. In order to evaluate FRET-basedbiosensors as an in vivo tool for quantitative metabolite analyses, we have thoroughly studied the effectsof pH, buffer salts, ionic strength, temperature and several intracellular metabolites on the signal inten-
YFPCFPitrineluorescent proteinntracellular metabolite analysis
sity of both biosensors and both fluorescence proteins. Almost all micro-environmental variations ledto considerably different FRET signals, because either the fluorescent proteins or the metabolite bindingdomains were affected by the tested parameters. This resulted not only in altered FRET ratios betweenthe apo state and the saturated state but also in significant shifts of the apparent binding constant. Thisunderlines the necessity of careful controls in order to allow reliable quantitative measurements Q2in vivo.
© 2014 Elsevier B.V. All rights reserved.
37
38
39
40
41
42
43
44
45
46
47
48
. Introduction
Due to the rapid signal response time FRET-based technologiesFörster, 1948) are highly promising for non-invasive intracellularnalyses. Respective genetically encoded biosensors consist essen-ially of two fluorescent proteins (FPs), which may react on specificnvironmental changes. Most biosensors contain FPs derived fromhe green fluorescent protein (GFP) of the jellyfish Aequorea victoriand related FPs. As GFP and many of its variants were shown to beighly sensitive toward various environmental parameters, amonghich the influences of pH and halide ions were most intensively
tudied (Martinière et al., 2013; Newman et al., 2011), less sensi-
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
ive variants have been generated (Rizzo et al., 2004; Zhang et al.,003). One of the frequently used less pH- and halide sensitiveFP-variant is Citrine, which additionally shows a high quantum
Abbreviations: AxP, AMP ADP ATP; GTP, guanosine triphosphate; cGMP, cyclicuanosine monophosphate; EYFP, enhanced yellow fluorescent protein; ECFP,nhanced cyan fluorescent protein; FP, fluorescent protein.∗ Corresponding author. Tel.: +49 2461 614388; fax: +49 2461 613870.
E-mail addresses: [email protected] (R. Moussa), [email protected]. Baierl), [email protected] (V. Steffen), [email protected]. Kubitzki), [email protected] (W. Wiechert), [email protected]. Pohl).
ttp://dx.doi.org/10.1016/j.jbiotec.2014.07.007168-1656/© 2014 Elsevier B.V. All rights reserved.
49
50
51
52
53
54
55
56
yield (Griesbeck et al., 2001). The broad range of available GFP-based biosensors has recently been extensively reviewed (Newmanet al., 2011; Okumoto et al., 2012). Besides, in recent years furtherFPs, such as the novel class of cyan-green fluorescent flavoproteinswere identified as useful FRET partners for GFP-derived FPs e.g.for the construction of oxygen-independent in vivo bioreporters(Drepper et al., 2013 and references therein) and as oxygen biosen-sors (Potzkei et al., 2012).
For the quantitative analysis of intracellular metabolites, genet-ically encoded biosensors of the Venus-flytrap type are frequentlyused. Such multi-domain biosensors FPD:BD:FPA consist of a ligandbinding domain (BD) between two usually GFP-based fluorophoresacting as FRET donor (FPD) and acceptor (FPA) domain, respec-tively. As a consequence of FRET (energy transfer: FPD→FPA) theemission intensity of the FRET-donor is reduced, whereas the emis-sion of the FRET-acceptor increases. In case of a ligand-bindingevent to the central BD the FRET-efficiency should change sig-nificantly to visualize the presence of bound ligands with a highsignal-to-noise ratio (Okumoto et al., 2012). During the last decadethe application of many genetically encoded biosensors has suc-
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
cessfully been demonstrated following signaling dynamics in livingcells (Deuschle et al., 2005; Palmer et al., 2006; Palmer and Tsien,2006; San Martin et al., 2013, 2014; Spiering et al., 2013; Takanagaet al., 2008).
57
58
59
60
ARTICLE IN PRESSG ModelBIOTEC 6762 1–10
2 R. Moussa et al. / Journal of Biotechnology xxx (2014) xxx–xxx
Fig. 1. Flowchart showing the application of genetically encoded FRET biosensors. Step 1 illustrates a common in vivo measurement with a cell expressing geneticallyencoded biosensors. After excitation of the donor fluorophore a FRET-ratio (YFP/CFP) is calculated from the fluorescence emission of the donor- and acceptor fluorophores.S y recot
mbwtaYsiiid
itdvitroucoSriac
aoa1
bwil2Fa
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
ubsequently, the corresponding ligand concentration is deduced from a previouslhe investigated ligand (metabolite) is deduced (Step 3).
The application of such ratiometric sensors for quantitativeetabolite analysis in living cells often relies on an in vitro cali-
ration by measuring a ligand-titration curve in an aqueous bufferith arbitrary pH, buffer salt, and ionic strength. From such titra-
ion curves the affinity (Kd-value) of the sensor is deduced from semi-logarithmic plot of the fluorescence intensity ratio (e.g.FP/CFP) of both chromophores over a logarithmic concentrationcale (Fig. 1). The central inflection point of the resulting S-curves subsequently used to deduce the Kd-value. In order to analyzen vivo data, the respective YFP/CFP fluorescence intensity ratios compared with the respective value of the calibration curve toetermine the metabolite concentration in vivo.
Although this method is generally accepted, it has some lim-tations. First of all the Kd-value can only be estimated from theitration curve, which resembles typical substrate concentration-ependent hyperbolic Michaelis–Menten curves known fromelocity studies with enzymes. In both cases a strong signal changes observed in a narrow concentration range. In order to deducehe Kd-value from such curves, the data set must be fitted with aespective (ideal) function. However, a usual dataset shows morer less deviations from this ideal function, especially under sat-ration conditions. In the semi logarithmic plot (Fig. 1) Kd is theentral inflection point and strongly dependent on the numberf data points, their errors and the fit quality to the ideal model.econd, and this is the topic of the present manuscript, the fluo-escence intensity ratio and the apparent Kd can dramatically benfluenced by environmental factors, such as pH, temperature, ionsnd cellular metabolites, which can hardly be simulated in in vivoalibrations.
Several factors are known which affect the FRET-efficiency, suchs the distance between the donor and acceptor fluorophores, theverlap of the donor emission and the acceptor absorption spectra,nd the orientation of the dipoles of both fluorophores (Dale et al.,979; Okumoto et al., 2012; Stryer and Haugland, 1967).
The goal of this study was to evaluate the reliability of such FRET-ased sensor systems for an in vivo-application, focusing on factorshich are most likely different between in vitro calibration and
n vivo application. For this purpose, we selected two well estab-
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
ished biosensors for glucose and maltose, respectively (Fehr et al.,002; Takanaga et al., 2008). These biosensors use very commonRET-pairs namely the enhanced cyan fluorescent protein (ECFP)s FRET donor and enhanced yellow fluorescent protein (EYFP) and
rded in vitro-titration curve (Step 2). As a result a putative in vivo-concentration of
Citrine, respectively, as FRET acceptors. Their central ligand bindingdomain consists of well characterized periplasmic binding proteinsfrom E. coli with known crystal structures (PDB: 2FVY, 1ANF). Bothsugar biosensors were broadly used for quantitative metaboliteanalyses in prokaryotes, yeasts, plants, and mammalian cells show-ing an excellent signal-to-noise ratio (Bermejo et al., 2011; Fehret al., 2002; Hou et al., 2011; Liemburg-Apers et al., 2011; Takanagaet al., 2008).
We characterized both sensors thoroughly to explore the influ-ence of micro-environmental effects, such as pH, buffer salts, bufferconcentration, ions, and cellular metabolites on the biosensor andalso on the single FPs. Our data show that the tested biosensors arevery sensitive toward most of the studied parameters, demonstrat-ing that care should be taken during the calibration of such systemsin order to deduce quantitative data.
2. Material and methods
2.1. Sensor preparation
The glucose sensor pRSET FLII12Pglu600� (Deuschle et al., 2005;Fehr et al., 2005) and the maltose sensor pRSET FLIPmal-25� (Fehret al., 2002) were from addgene (http://www.addgene.org). Theprotein sequences of both sensors are shown in Fig. S15. Charac-teristic mutations relative to GFP (UniProt accession no. P42212) inthe respective FPs were:
ECFP: F64L, S65T, Y66W, N146I, M153T, V163A ((Kremers et al.,2006), Clontech)
EYFP: S65G, V68L, S72A, T203Y (variant 10c in (Ormö et al.,1996), Clontech)
Citrine: S65G, V68L, Q69M, S72A, T203Y (Griesbeck et al., 2001)E. coli BL21 (DE3) GOLD cells (Stratagene) were transformed
with those vectors and positive colonies were selected on agarplates through ampicillin resistance. BL21 (DE3) Gold cells weregrown in 2 L shaking flasks with baffles using 400 mL lysogenybroth medium (LB) containing 100 �g mL−1 ampicillin. Cultiva-tion and sensor expression was carried out for 48 h at 28 ◦C in
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
the dark, without IPTG-induction. Cells were harvested by cen-trifugation, resuspended in 20 mM MOPS buffer (pH 7.3), anddisrupted by ultrasonication at 4 ◦C using an UP200s (Hielscher,Teltow, Germany). The sensor proteins carried an N-terminal
137
138
139
140
ING ModelB
iotech
haGGp5ftp2bpi0ai
2
snFsptadUspoer
arwtE2ercoia
R
was
p(
2
pwOt
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
ARTICLEIOTEC 6762 1–10
R. Moussa et al. / Journal of B
exahistidine tag and were purified via immobilized-metal chelateffinity chromatography on Ni-NTA aragose (Qiagen, Hilden,ermany) by fast protein liquid chromatography (Äkta Purifier,E Healthcare Life Sciences). Typically, we obtained about 2.5 mgure glucose sensor per gram wet cell weight (gwcw
−1) and ca. mg gwcw
−1 of the maltose sensor. All purification steps were per-ormed at room temperature with a flow rate of 3 mL min−1. Forhe washing steps we used 20 mM MOPS buffer (pH 7.3). Theurified sensors were eluted with 20 mM MOPS buffer containing50 mM imidazole (pH 7.3). Subsequently, imidazole was removedy ultrafiltration using a Centricon centrifugal filter units (Milli-ore, Darmstadt, Germany) and the sensor solution was stored
n 20 mM MOPS buffer (pH 7.3) with a protein concentration of.5 mg mL−1 at −20 ◦C. Ultrafiltration using Centricon tubes waslso used for buffer exchange and to adjust different pH-values andon concentrations.
.2. Calibration curve and Kd-determination
All measurements were carried out using a microtiter platepectrofluorimeter with monochromator unit (Tecan M-200). Ifot otherwise stated, transparent flat bottom 96 well plates (Nunc, Thermo Fisher Scientific) were used by adding 90 �l biosensorolution (0.5 mg mL−1) and 10 �l of different sugar concentrationser well. In order to record reliable titration curves, 24 stock dilu-ions of glucose and maltose in the range of 0.0001–2000 mM forll measurements were prepared, respectively. The sugars wereissolved in 20 mM MOPS buffer, pH 7.3, if not otherwise stated.pon 1:10 dilution with the respective sensor solution the final
ugar concentration was between 0.00001 and 200 mM. All dataoints represent the arithmetic average of 15 measurement cyclesf the microtiter plate reader. The standard deviation given forach data point denotes the difference between three independenteplicates.
The donor (ECFP: Clontech, Kremers et al., 2006)) was excitedt 428 nm with a bandwidth of 9 nm. The acceptor proteins (Cit-ine (Griesbeck et al., 2001), EYFP: (Ormö et al., 1996), Clontech)ere both excited at 485 nm with the same bandwidth. The accep-
or emission was recorded at 528 nm with a bandwidth of 20 nm.mission of ECFP was measured at 485 nm with a bandwidth of0 nm. Changes of the FRET signal were determined as YFP/CFP-mission intensity ratio, by only exciting the donor protein andecording the emission intensity of both, the CFP and YFP fluores-ence signal. The dissociation constant (Kd) was calculated basedn the change of the FRET-ratio upon ligand binding. Therefore anterative fitting of the substrate titration curves for the binding of
ligand to a protein was computed with the equation:
= (Rmax − Rapo) · [S]Kd + [S]
+ Rapo
here [S] is the sugar concentration, Rapo the YFP/CFP-value inbsence of sugar, and Rmax is the highest YFP/CFP-value at sugaraturation.
The emission spectra of all purified sensor- and fluorescentroteins were recorded using a FluoroLog®-3 spectrophotometerHoriba) with a bandwidths of 2.5 nm each (Figs. S10–S13).
.3. Influence of pH
The purified biosensor proteins were adjusted to five different
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
H-values in the range of pH 6.3–8.3 in 20 mM MOPS buffer, eachith a protein concentration of 0.5 mg mL−1. This resulted in anD428 of 0.086 in microtiter plates with a light path of 3 mm. The
emperature during the measurements was kept at 25 ◦C, also the
PRESSnology xxx (2014) xxx–xxx 3
sensor solution was pre-warmed at this temperature to ensure thecorrect pH during the experiment.
2.4. Influence of temperature
Due to the fact that every buffer has a different tempera-ture coefficient (�pKa/◦C) (MOPS −0.011 ◦C−1), all MOPS bufferswere prepared for the temperature-dependent studies usingan online tool from Liverpool University (http://www.liv.ac.uk/buffers/buffercalc.html) to ensure a constant pH of 7.3 at therespective temperatures. To ensure the correct buffer temperature(26–42 ◦C) in the microtiter plate, we adapted a colorimetric tem-perature assay described by (Schilling et al., 1993) to the microtiterplate format. This method makes use of the high temperature coef-ficient of TRIS buffer using a cresol red indicator system. Briefly,10 mg mL−1 cresol red (Sigma: 114472-25G) were dissolved inultrapure ethanol as a stock solution. This stock solution wasdiluted (1:500) with Tris–HCl buffer (100 mM, pH 7.5) at 37 ◦C,resulting in a final concentration of 20 �g mL−1 cresol red, whichwas pipetted in a 96-well plate (100 �l per well) to analyze the tem-perature distribution over the microtiter plate and to calibrate thesystem. Afterwards the microtiter plate was sealed with an adhe-sive sealing film (e.g. polyester film, VWR Cat. No.: 60941-062)and placed into the microtiter plate reader, which was stepwiseheated to 25 ◦C, 30 ◦C, 35 ◦C, 40 ◦C and 42 ◦C. Each temperature waskept constant for 1 h to ensure thermal equilibrium. Afterwards,the microtiter plate was scanned at 438 nm, 574 nm and 680 nm.Fig. S12 shows the shift of the two maxima 438 nm and 574 nm inrelation to the temperature-caused pH-shift. 680 nm is a baselinevalue in order to compare the shifts. Based thereon the correlationbetween temperature and the change of the photometric signal wascalculated. The results of these studies are shown in Fig. S14 andTable S3. All other measurements were performed at 25 ◦C. There-fore the samples were pre-warmed at 25 ◦C in a water bath and thepeltier element in the microtiter plate reader was always adjustedto 25 ◦C for at least 1 h before the measurements were started. Toensure reliable, only temperature-dependent measurements thesugar stock solutions used for the temperature-dependent studies,were prepared with the respective MOPS buffers to assure constantpH.
2.5. Influence of buffer salts and buffer concentration
The influence of different buffer salts was tested, which are fre-quently used in studies with genetically encoded FRET-biosensors.MOPS, HEPES, sodium-phosphate and Tris–HCl were all appliedwith the same concentration of 20 mM at pH 7.3.
MOPS buffer was prepared with concentrations of 5, 10, 20, 50,75 and 100 mM, respectively. All sugar stock solutions were pre-pared in the respective buffers. As the ionic strength I of a bufferis directly correlated with its pH and thus with temperature, thelatter two parameters were kept constant at pH 7.3 and 25 ◦C.
2.6. Influence of intracellular metabolites and salts
The influences of intracellular metabolites and salts were testedin a broad concentration range in order to cover physiological rel-evant concentrations.
The following 16 compounds were tested in the given concen-tration ranges, each in 12 concentration steps, respectively:
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
- Metabolites: ATP, ADP, AMP, NAD, NADH, NADP, NADPH, cGMP,GTP [10 �M–20 mM]
- Salts: NaCl, CaCl2, MgCl2, KCl, MnCl2 [0.625 �M–625 mM]
253
254
255
ARTICLE IN PRESSG ModelBIOTEC 6762 1–10
4 R. Moussa et al. / Journal of Biotechnology xxx (2014) xxx–xxx
F ensors y ratios f thre
sspifiap(ttifw(s
2
bsfWctr
3
3
Feeaior8ei
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
ig. 2. Ligand titration curves at different pH-values. Measurements of the glucose sensor in 20 mM MOPS at 25 ◦C. The Y-axis represents the YFP/CFP-emission intensitensor: EYFP). For details see Table S1. Each data point represents the mean value o
All tests were performed with and without the respectiveugar-ligand to ensure that the basic signaling properties of theensors were not affected by the added compounds. Here, weresent only data in presence of the respective sugar. The exper-
ments with different metabolites have been performed withnal biosensor concentrations of 0.35 mg mL−1 and were addition-lly carried out with the single donor and acceptor fluorescentroteins (0.35 mg mL−1). For all measurements 70 �l biosensor0.5 mg mL−1), 20 �l metabolite solution and 10 �l sugar solu-ion (1 M) or buffer, respectively, were used. All measurements inhe presence of different salts were carried out with biosensorsn final concentrations of 0.325 mg mL−1 and were also per-ormed with the single donor and acceptor fluorescent proteins asell (0.325 mg mL−1). Each experiment contained 65 �l biosensor
0.5 mg mL−1), 25 �l of the respective salt solution and 10 �l sugarolution (1 M) or buffer, respectively.
.7. Analyses of ligand specificity
In order to test the specificity of the glucose and mannose sensor,oth were tested with different mono- and disaccharides and theugar surrogate IPTG. The studies have been carried out as describedor glucose and maltose (calibration curve and Kd-determination).
ith the maltose sensor glucose, cellobiose, trehalose, lactose, sac-harose, and IPTG were tested, respectively. The glucose sensor wasested with galactose, maltose, mannose, ribose, fructose, and IPTG,espectively.
. Results and discussion
.1. Influence of pH
The sensitivity of the fluorescence properties of GFP-derivedPs toward pH has been broadly described in literature (Bizzarrit al., 2009; Griesbeck et al., 2001; Llopis et al., 1998; Newmant al., 2011; Wachter et al., 1998). The pKa of the fluorophores is
function of the chromophore forming amino acid residues andnteracting side chains. In this study we measured the influencef pH on the apparent Kd of both sensors for glucose and maltose,
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
espectively, in the physiologically relevant range from pH 6.3 to.3. The results are shown in Fig. 2. Details concerning the influ-nce on the YFP/CFP-ratio and the apparent Kd-value are givenn the Supplementary Information (Table S1). The specific terms
(a) and the maltose sensor (b) were performed in a 96-well plate with 0.5 mg mL−1
of the donor (ECFP) and the acceptor fluorophores (glucose sensor: Citrine; maltosee independent experiments.
YFP/CFP-ratioapo, YFP/CFP-ratiomax, and YFP/CFP-�ratioshift, whichare used below to discuss the data are explained in Fig. S1.
Titration curves of both sensors with glucose and maltose,respectively, at the respective fixed pH values (Fig. 2) show anopposed behavior toward pH-changes. It is clearly visible thatpH influences the fluorescence ratio of both FRET-sensors in theabsence of the respective sugar (YFP/CFP-ratioapo) and under satu-rating condition (YFP/CFP-ratiomax) as well as the dynamic range(YFP/CFP-�ratioshift). Whereas the fluorescence ratio (YFP/CFP-ratio) rapidly drops with rising pH in case of the glucose sensor,the fluorescence ratio of the maltose sensor increases with pH.As a consequence, the correlation between the fluorescence ratioand the metabolite concentration is strongly pH-dependent forboth sensors, meaning that the estimation of metabolite concen-trations in vivo based on in vitro calibration curves (Fig. 1) is afunction of the pH chosen for the calibration relative to the pH inthe cell or organelle. In case of the glucose sensor the apparentKd varied in the range of 0.2–4.7 mM glucose and for the maltosesensor Kd-values between 7 and 54 �M maltose were calculated.This behavior cannot be explained with the pH-sensitivity of thedifferent FPs (ECFP, EYFP, Citrine), which were separately studiedin the respective pH range (Fig. S13). Based on the literature datawe expected weak influence on the fluorescence properties of ECFP(pKa 4.8, Kremers et al., 2006) in the tested pH range from 6.3 to 8.3,whereas EYFP (pH >7, Miyawaki et al., 1997) and Citrine (pKa 5.7Griesbeck et al., 2001) were assumed to be more effected. However,as demonstrated in Fig. S13 both YFP-variants showed the samepH-dependent behavior: their fluorescence intensity only slightlydropped with increasing pH and there is no significant differencevisible between EYFP and Citrine. Besides, ECFP showed a slightincrease of the fluorescence intensity with increasing pH.
The inverse behavior of both sugar sensors is thus probably dueto the binding domain, which goes in line with the observed pH-dependent change of their apparent Kd-value.
3.2. Influence of buffer salt and buffer concentration
Since a broad range of different buffer salts and buffer concen-trations was described for the calibration of FRET-based biosensorsin the literature (e.g. (Deuschle et al., 2005; Fehr et al., 2002; Llopis
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
et al., 1998; Miyawaki et al., 1999; Potzkei et al., 2012; Rodriguez-Garcia et al., 2014; San Martin et al., 2013, 2014; Takanaga et al.,2008), we investigated their influence in more detail. Starting withdifferent concentrations of MOPS-buffer (5–100 mM) the typical
332
333
334
335
ARTICLE IN PRESSG ModelBIOTEC 6762 1–10
R. Moussa et al. / Journal of Biotechnology xxx (2014) xxx–xxx 5
F nts of
w s see T
tt
ssYctwe(mifttKbvt
tic
Fi
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
ig. 3. Ligand titration curves at different MOPS-buffer concentrations. Measuremeith 0.5 mg mL−1 sensor at pH 7.3 and 25 ◦C in MOPS buffer (5–100 mM). For detail
itration curves were recorded with glucose and maltose, respec-ively.
As demonstrated in Fig. 3, the buffer concentration has atrong impact on the YFP/CFP-ratio especially for the glucose sen-or regarding the YFP/CFP-ratioapo and the YFP/CFP-ratiomax, withFP/CFP-�ratioshift varying between 0.5 and 1.2 (Table S2). Byontrast, the maltose sensor is less sensitive toward the varia-ion of the buffer concentration in the range of 20–100 mM MOPS,hereas the YFP/CFP-ratiomax intensities measured at the low-
st MOPS concentrations (5–10 mM) were significantly reducedTable S2). Regarding the apparent Kd-values, we observed only
inor influences in case of the glucose sensor, where Kd variedn the range of 1.1–1.5 mM glucose. However, the decreasing dif-erence between the YFP/CFP-ratioapo and the YFP/CFP-ratiomax inhe presence of 5 mM, 75 mM and 100 mM MOPS and the concomi-ant flattening of the titration curve hampered the estimation ofd. Although the titration curves for the maltose sensor seem toe more reproducible on a first glance, the deduced apparent Kd-alues for maltose differ in the range of 26–50 �M depending onhe buffer concentration.
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
It is important to note that there is no clear correlation betweenhe MOPS buffer concentration and the observed effects. Concern-ng the relevance of these results for in vivo studies literature dataoncerning the ion concentration in E. coli were analyzed. The
ig. 4. Ligand titration curves in HEPES, MOPS, sodium phosphate, and Tris–HCl buffer. Mn a 96-well plate with 0.5 mg mL−1 sensor at pH 7.3 and 25 ◦C. Each data point represent
the glucose sensor (a) and the maltose sensor (b) were performed in a 96-well plateable S2.
results were collected in Table S4. From these data it can be deducedthat potassium salts play an important role for the ionic strengthin E. coli.
HEPES, sodium phosphate, and Tris–HCl are further bufferscommonly used for studies with FRET-based biosensors. As demon-strated in Fig. 4 the results are rather comparable for MOPS, HEPES,and Tris–HCl, whereas sodium phosphate buffer gave differentresults. While for the glucose sensor a significantly reduced FRET-shift upon ligand saturation (YFP/CFP-ratiomax) was observed,sodium phosphate buffer caused the opposite effect regardingthe YFP/CFP-ratiomax value for the maltose sensor, resulting in anincreased maximal signal (Fig. 4, highlighted). Influences on theapparent Kd-values were minor for both sensors (<10%) (data notshown).
3.3. Influence of metal chlorides and metabolites
In order to elucidate possible influences of essential cellularmetabolites, a selection of 16 compounds was tested concerningtheir influence on the YFP/CFP-ratio of both biosensors and on the
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
fluorescence intensity of the single FPs. The intracellular concen-tration of different metabolites in E. coli is described in severalpublications (Anjem et al., 2009; Bennett et al., 2009; Hurwitz andRosano, 1967; Lo et al., 2006; Shabala et al., 2009; Watkins et al.,
easurements of the glucose sensor (a) and the maltose sensor (b) were performeds the mean value of two independent experiments.
378
379
380
381
ARTICLE IN PRESSG ModelBIOTEC 6762 1–10
6 R. Moussa et al. / Journal of Biotechnology xxx (2014) xxx–xxx
F (a) anp −1 at p7 n, but
1bebwi
rtocotaotteEees
Fc
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
ig. 5. Influence of different metal chlorides. Measurements of the glucose sensorerformed in a 96-well plate. Each well contained 65 �l sensor solution (0.5 mg mL.3 instead of the sugar ligand. Hence, the X-axis does not represent a sugar dilutio
995) (Table S5). However, many values differ strongly most proba-ly due to differences in the strains, metabolic status, cell cycle, andxperimental setup. Taking this limitation into account we choseroad concentrations ranges for the tested metabolites and saltshich cover the physiologically relevant ranges for many organ-
sms.First of all we tested different mono- and bivalent metal chlo-
ides in different concentrations. The results in Fig. 5 demonstratewo distinct effects of different mono- and divalent metal chloridesn the YFP/CFP ratio of both biosensors: First, with increasing con-entrations a strong decay of the fluorescence intensity ratio wasbserved (>30 mM for the glucose sensor and >10 mM for the mal-ose sensor). Second, among the tested salts calcium-, magnesium-nd manganese chloride are special, since they caused an increasef the YFP/CFP-ratio in a concentration range of 2–5 mM. Whereashis effect seems to be independent of the chloride concentration,here is a clear impact of the bivalent metal ions visible. While theffect observed with MnCl2 could be deduced to the decay of theCFP fluorescence intensity in the presence of >1 mM MnCl2, the
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
ffects observed with CaCl2 and MgCl2 could not be explained byffects on the FPs. As shown in Fig. S2 Citrine did not show anyensitivity toward the metal chlorides in the concentration range
ig. 6. Influence of AMP, ADP, and ATP. Measurements of the glucose sensor (a) and the montaining 70 �l sensor solution (0.5 mg mL−1), 20 �l AxP and 10 �l 1 M glucose (a) or ma
d the maltose sensor (b) in the absence of glucose or maltose, respectively, wereH 7.3 in 20 mM MOPS; 25 ◦C); 25 �l salt solution; 10 �l of 20 mM MOPS buffer, pHthe concentrations of the tested salts. Data points result from a single experiment.
tested, whereas the fluorescence intensity of EYFP decreased onlyabove 10 mM with all tested salts, which could be explained bythe halide sensitivity known for this variant. The increase of theYFP/CFP-ratio with CaCl2 and MgCl2 in the range of 2–5 mM is mostpronounced for the glucose sensor (Fig. 5a). It can most probably beexplained with a calcium binding site, which has been identified inthe crystal structure of the glucose/galactose binding protein (Vyaset al., 1987). It can be assumed that binding of bivalent cationsto the glucose binding domain alters its structure and therewithalso the FRET-effects of the biosensor (Vyas et al., 1987). This effectwas most probably also observed by (Strohhöfer et al., 2011) whostudied the effect of Ca2+ on the fluorescence life-time of a FRETsensor consisting of ECFP and EYFP flanking both sites of the glu-cose/galactose binding domain. So far, no calcium binding site hasbeen described for the maltose binding protein. In order to dissectthe ion-sensitive part in the sensors, the fluorescent properties ofthe fluorescent proteins were additionally analyzed (Fig. S2).
We therefore conclude that the observed effect is due to thesugar-binding domains in both sensors. Also the decrease of the
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
FRET ratio observed with higher salt concentrations cannot directlybe explained by an effect on the FPs, since only EYFP was shown tobe sensitive toward the different chlorides above 10 mM (Fig. S2)
altose sensor (b) were performed in a 96-well plate in 20 mM MOPS, pH 7.3 at 25 ◦Cltose (b) per well, respectively. Data points result from a single experiment.
423
424
425
ARTICLE IN PRESSG ModelBIOTEC 6762 1–10
R. Moussa et al. / Journal of Biotechnology xxx (2014) xxx–xxx 7
F senss ata po
uEa
AtoaEaflpidu
tt
FssApp
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
ig. 7. Influence of cGMP and GTP. Measurements of the glucose (a) and the maltoseensor (0.5 mg mL−1), 20 �l cGMP or GTP, 10 �l 1 M glucose or mannose per well. D
nder the tested conditions. As only the maltose sensor containsYFP the similar effects observed with both sensors (Fig. 5b) mustlso be traced back to the sugar-binding protein.
Subsequently, the influence of cellular metabolites such as ATP,DP, AMP, cGMP, GTP, NAD+, NADH, NADP+, and NADPH was
ested on the biosensors and also on the fluorescence intensitiesf the single fluorescent proteins. Physiological concentrations ofll tested cellular compounds have been recently published for. coli by (Bennett et al., 2009) (Table S5). These data were takens a guideline for the analysis of metabolite influences on theuorescence properties of the sensors and the single fluorescentroteins. However, it should be noted that the quantification of
ntracellular metabolites is a difficult task and the results stronglyepend especially on the sample preparation the analytical method
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
sed.The results in Figs. 6–8 clearly show a strong influence of the
ested compounds on the FRET-ratio of the biosensors, althoughhe glucose sensor appears to be stronger affected at much lower
ig. 8. Influence of NAD(P)(H). Measurements of the glucose sensor (a) and the maltose senensor (0.5 mg mL−1), 20 �l of the respective redox cofactor, 10 �l 1 M glucose or mannopectra of the reduced redox cofactors overlap with those of ECFP, EYFP and Citrine. Thelthough the simple subtraction of the control seems to be valid for the complete tested cresence of NADH, where concentrations >10 mM NADH resulted in a dramatic increase ohoto-physical effects. However, such effects could only be observed with the sensor pro
or (b) were performed in a 96-well plate in 20 mM MOPS, pH 7.3 at 25 ◦C with 70 �lints result from a single experiment.
concentrations compared to the maltose sensor. In many caseswe could deduce the effects observed with both biosensors to therespective FPs (Figs. S3–S8).
The results with AMP, ADP and ATP (Fig. 6) demonstrate anincreasing decay of the FRET ratio with the number of phos-phate ester groups. As demonstrated in Fig. S3, this effect couldbe deduced to the FPs, were the fluorescence intensity decreasedin all cases in the series ATP > ADP > AMP. This decrease was mostpronounced for Citrine, whereas ECFP was only affected by ATPabove 10 mM. The influence of ATP on the fluorescence proper-ties of CFP has already been described earlier (Borst et al., 2010;Willemse et al., 2007). Willemse et al. (2007) studied a FRET basedATP sensor and found effects of ATP on not further specified CFPand YFP variants. Borst et al. (2010) assigned the influence of ATP
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
on the fluorescence intensity of ECFP to possible electrostatic inter-actions of the negative charge of ATP with the positively chargedhistidine 148, which is close to the chromophore (Campanini et al.,2013; Wachter et al., 1998). This assumption was based on the fact
sor (b) were performed in a 96-well plate in 20 mM MOPS, pH 7.3 at 25 ◦C with 70 �lse per well. The results obtained for NAD(P)H are corrected, since the fluorescencerefore controls were measured under identical conditions at 485 nm and 528 nm.oncentration range of NADPH, the situation is different for the YFP/CFP-ratio in thef fluorescence (marked with *). This is probably the consequence of more complexteins but not with the single fluorescence proteins (see Fig. S8).
458
459
460
461
ARTICLE IN PRESSG ModelBIOTEC 6762 1–10
8 R. Moussa et al. / Journal of Biotechnology xxx (2014) xxx–xxx
Fig. 9. Ligand specificity of the glucose sensor (a) and the maltose sensor (b). Measurements were performed in a 96-well plate in 20 mM MOPS, pH 7.3 at 25 ◦C. The respectivea
twiceItiaiitcfl
rocicfl(d
rctoearcbbsnFrNoFocq
the determination of a respective intracellular metabolite concen-
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
pparent Kd-values are given.
hat a respective His-free variant (Cerulean) in a fusion with EYFPas not affected by ATP. Even though the authors did not find an
nfluence of ATP on the single FPs up to 10 mM (J. Fransen, personalommunication), we clearly observed similar effects on EYFP andven stronger on Citrine (Fig. S3a,b) at ATP concentrations >10 mM.n our investigations, the overall effects of AxP on the YFP deriva-ives were comparable. This simultaneous, asymmetric decreasen donor- and acceptor fluorescence of both fluorescent proteinsffects the overall FRET efficiency dramatically. As the physiolog-cal concentration of ATP in E. coli is about 10 mM (Table S5), thentracellular ATP concentration hampers the fluorescence proper-ies of such biosensors significantly. In contrast, the intracellularoncentration of AMP and ADP in E. coli and thus their effect on theuorescence properties can be neglected (Table S5).
As demonstrated in Fig. 7, cGMP and GTP decrease the YFP/CFP-atio as well, with GTP showing the strongest impact especiallyn the glucose sensor (Fig. 7a, highlighted). The physiological con-entration of GTP in E. coli is about 0.5 mM (Table S5), whichnduced a decrease of the FRET-ratio from 5.3 to 3.8 with the glu-ose sensor. Concerning the effect of both metabolites on the singleuorescent proteins no significant difference could be observedFig. S4), suggesting an interaction of GTP with the glucose bindingomain.
The results obtained with ATP motivated us to further studyelated metabolites containing this moiety, like the nucleotideofactors NAD(H) and NADP(H). The results in Fig. 8 demonstratehe enormous influence of the reduced and oxidized cofactorsn the fluorescence properties of both sensors, whereas theffects were most pronounced with the glucose sensor alreadyt considerably lower concentrations. Analysis of the single fluo-escent proteins revealed a reason for this effect (Fig. S5): NAD+
auses a very strong quenching of the fluorescence intensity ofoth acceptors (Citrine and EYFP) with increasing concentration,ut has a much less dramatic effect on ECFP. Since the mea-ured FRET-signal is the YFP/CFP-fluorescence intensity ratio, aon-proportional decrease of the fluorescence intensities of bothRET-partners results in a pronounced change of the YFP/CFP-atio. This conclusion is supported by results obtained with NADH,ADP+ and NADPH, which mostly caused a very strong decreasef the fluorescence of the acceptor molecules (Fig. S6, Fig. S7,ig. S8). So far we cannot explain the molecular reasons for this
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
bservation. The presented results demonstrate that physiologicaloncentrations of NADPH in E. coli (Table S5) could cause similaruenching effects in vivo. For other organisms with higher internal
NAD+-concentrations than E. coli also the NAD+-concentrationcould affect the fluorescence properties.
Both sensors were further tested with related mono- and disac-charides in order to detect cross-binding activities of the bindingdomains. The results indicate that the glucose sensor, which con-tains a glucose/galactose binding domain, shows affinity also forseveral other sugars like galactose, maltose and mannose, althoughthe latter two Kd-values are tenfold higher (Fig. 9a). Additionally,we tested the affinity toward glucose-6-phosphate, since glu-cose is phosphorylated by the phosphotransferase system (PTS)while transported into the cytosol of E. coli. Fig. S9 shows thelack for any affinity of the glucose sensor for phosphorylatedglucose. This result is consistent with our attempts to measure glu-cose in E. coli in vivo, which was not successful with the glucosesensor.
In contrast the selectivity of the maltose sensor is higher andonly glucose was found to bind to the sensor, however with a 10−4-fold lower Kd-value (Fig. 9b).
4. Conclusions
We have exemplarily studied the sensitivity of two fre-quently used genetically encoded FRET-biosensors toward micro-environmental effects, in order to elucidate influencing effects forquantitative in vivo applications. Both sensors were tested underdifferent in vitro conditions and in the presence of a broad range ofadditives; many of them are cellular metabolites or trace elements.Since such FRET-sensors are multi-domain proteins consisting oftwo fluorescent proteins and a binding domain, the single fluo-rescent proteins were also tested in order to retrace the observedeffects to the respective parts of the sensor. Although factors affect-ing the FRET ratio of such sensors have frequently been observed,there is a less pronounced awareness concerning effects on theapparent affinity of such sensors. However, earlier studies e.g. withthe glucose sensors show that the apparent Kd in vivo stronglydiffers from in vivo data (see Table 1 in Takanaga et al., 2008).
Among the tested parameters in this work, especially the pH(Fig. 2) and the concentration of the buffer (Fig. 3) used for thecalibration of the sensors had a strong impact on the apparent affin-ity and the fluorescence properties of the sensors, which makes
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
tration according to the common procedure shown in Fig. 1 inour opinion rather uncertain. Although this general problem hasalready been recognized (Liemburg-Apers et al., 2011; Okumoto
545
546
547
ING ModelB
iotech
es
tb(
lpAappn(tsbcrnttcdloFg
chosv7Kp(laoteKvpspitwtrtbm
twilTals
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
ARTICLEIOTEC 6762 1–10
R. Moussa et al. / Journal of B
t al., 2012; San Martin et al., 2013, 2014), this is the first systematictudy on such effects.
Whereas most of the different tested buffer salts gave consis-ent results, deviations were observed with phosphate buffer foroth sensors, which is frequently used in studies using such sensorsFig. 4).
For the first time we studied the effects of a variety of cellu-ar metabolites in a broad concentration range on the fluorescenceroperties and affinities of both sensors and on the single FPs.
large majority of cytosolic metabolites is negatively chargednd electroneutrality is maintained by a high concentration ofotassium ions and to a lesser degree by magnesium ions andolyamines (Cossins et al., 2011). Whereas monovalent cations didot show significant effects on the sensor properties, Ca2+ and Mg2+
2–5 mM) increased the FRET-ratio of the glucose sensor aboutwofold and showed similar but weaker effects on the maltoseensor (Fig. 6). This could be explained by an interaction of theivalent cations with the sugar binding domain. Toward higher con-entrations (>50 mM) of the different metal chlorides both sensorseact with a strong decrease in FRET-ratio, an effect which couldot completely be deduced to the FPs, as only EYFP was showno be sensitive toward chloride concentrations >10 mM under theested conditions. Thus, the sensitivity toward the different testedhlorides must also result from influences on the sugar-bindingomains. Both sensors were significantly affected by the cellu-
ar metabolites ATP, GTP, NAD+ and NADPH (Figs. 6–8). In casef ATP, NAD+ and NADPH these effects could be deduced to thePs (Figs. S3, S5-S8), whereas GTP most probably influences thelucose/galactose binding domain (Fig. S4).
Besides, our studies demonstrate that even under equivalentonditions in our lab determination of apparent Kd-values areardly precisely reproducible. For example different preparationsf the glucose sensor were measured in different experimentaletups under almost equivalent conditions, which resulted in Kd-alues between 1.3 mM and 2 mM (Tables S1–S3, 20 mM MOPS, pH.3, 25–26 ◦C). Thus, in our studies we observed a mean apparentd value of 1.65 ± 0.35 mM, which is significantly higher than theublished Kd for this sensor (583 ± 8.0 �M, in 20 mM MOPS, pH 7Takanaga et al., 2008). This difference is probably partly due to theower pH used by Takanaga et al. (2008). As demonstrated in Fig. 2nd Table S1 a pronounced decrease of Kd with decreasing pH wasbserved in our study for both biosensors. However, in both studieshe error is >20% (21% in our studies and 27% in reference Takanagat al., 2008), indicating the principle limitation in the accuracy ofd-determination. This relatively large error accumulated from thearious pipetting steps, the error of the spectroscopic device, tem-erature control and the stability of the sensors. Depending on thetorage time, the storage temperature and the concentration of theurified sensor during storage the measurements were influenced
n repetitive experiments. We observed a continuous decrease ofhe fluorescence intensities of both FPs, eCFP and eYFP, where eCFPas particularly sensitive toward photobleaching. As the decay of
he FPs occurred in a non-proportional manner, the starting FRETatio and also the FRET ratio shift were significantly influenced byhis effect. Further, also effects correlated to denaturation of theinding domain cannot be excluded; however they are less easilyeasureable.Although only two Venus-flytrap sensors were exemplarily
ested in this study, it can be assumed that we identified generaleaknesses of sensors of this type. This is supported by recent stud-
es of San Martin et al., who characterized two FRET biosensors foractate and pyruvate, respectively (San Martin et al., 2013, 2014).
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
hese sensors contain mTFP and Venus as FPs. In both cases theuthors studied also the influences of pH and different other cel-ular metabolites on the sensor response. Whereas the pyruvateensor seems to be very robust toward environmental changes,
PRESSnology xxx (2014) xxx–xxx 9
showing only minor influences toward pH and oxalacetate concern-ing Kd and the FRET �ratioshift, the lactate sensor is more sensitive.Here, the apparent affinity for lactate is clearly influenced by thepH, pyruvate, and citrate (although the effects of Kd are not furtherdiscussed).
Based on the complex structure of these sensors and the differ-ent effects giving rise to a FRET signal, the actual binding event atthe central binding domain is not directly measured but deducedfrom the FRET event. Under optimal conditions the fusion of theFPs to the binding domain will not influence the binding affinityof the binding domain. However, depending on linker variationand altered environmental conditions all three parts of the fusionprotein might respond differently to such changes and minor con-formational changes may result in altered apparent affinities ofsensor.
As a result there is an urgent need for appropriate calibrationsystems. A relatively simple solution would be calibration of thesensors in buffers that are as close as possible to the cellular inte-rior. However, many of these buffers termed “intracellular buffer”(e.g. San Martin et al., 2013, 2014) or “artificial cytosol” (Takanagaet al., 2008), do not contain cellular metabolites, but different phys-iological salts.
Our studies demonstrate that different cellular compounds (e.g.AxP, NAD(P)(H)) influence the signal and thus must be includedinto the calibration system. We tested only few metabolites apartfrom the actual metabolite specifically recognized by the bindingdomain, but based on these results and the number of primaryand secondary metabolites present in bacterial and specificallyeukaryotic cells, respective interactions can also be expected withother metabolites. Therefore use of respective crude cell extracts ismore appropriate. However the preparation of crude cell extractsis usually connected with dilution of the cellular components andnon-soluble fractions are not regarded (e.g. Borst et al., 2010;Jones et al., 2014). The best way to cope with such effects at leastfor sensors located in the cytosol is to measure in permeabilizedwhole cells which guarantees a constant cellular environment, ifrespective non-interfering protocols are available. However, mea-surements in organelles would require isolation of the respectiveorganelle (Liemburg-Apers et al., 2011). Additionally, an alternativeanalytical method to determine the target metabolite concentra-tion for the sensor in the cell/organelle would be desirable, whichallows calibration of the sensor via such an independent method.However, such analyses are usually offline and require rapid stopof the metabolic activity in cells, cell lysis, and a respective ana-lytical protocol that allows a reliable analysis of the respectiveanalyte, which is often a difficult task. A method to eliminate theeffects induced by interaction with cellular metabolites on the flu-orescent properties of the FPs in the sensor is the use of controlsensors with very low binding affinity for the metabolite to bequantified (e.g. Fehr et al., 2003). However, such controls wouldnot identify interactions of cellular metabolites with the bindingdomain.
Further, molecular crowding effects must also be regarded. Inthat respect a recent theoretical study described effects of molecu-lar crowding, ionic strength, and calcium binding on the calmodulindynamics, demonstrating that increased levels of molecular crowd-ing in addition to calcium binding and an ionic strength, typical forthe cell interior, can impact the conformation of calmodulin andtherewith the affinity of calcium binding (44).
Author contributions
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
Conceived and designed the experiments: RM, VS, AB, TK. Ana-lyzed the data: RM, VS, AB, TK, MP. Wrote the manuscript: RM, VS,MP, WW.
674
675
676
ING ModelB
1 Biotech
A
oD
A
f2
R
A
B
B
B
B
C
C
D
D
D
F
F
F
F
G
H
H
J
K
L
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
ARTICLEIOTEC 6762 1–10
0 R. Moussa et al. / Journal of
cknowledgements
We thank Thomas Gensch, Institute of Complex Systems (ICS-4)f Forschungszentrum Jülich, and Thomas Drepper, University ofüsseldorf, for helpful discussions.
ppendix A. Supplementary data
Supplementary data associated with this article can beound, in the online version, at http://dx.doi.org/10.1016/j.jbiotec.014.07.007.
eferences
njem, A., Varghese, S., Imlay, J.A., 2009. Manganese import is a key element ofthe OxyR response to hydrogen peroxide in Escherichia coli. Mol. Microbiol. 72,844–858.
ennett, B.D., Kimball, E.H., Gao, M., Osterhout, R., Van Dien, S.J., Rabinowitz, J.D.,2009. Absolute metabolite concentrations and implied enzyme active site occu-pancy in Escherichia coli. Nat. Chem. Biol. 5, 593–599.
ermejo, C., Haerizadeh, F., Takanaga, H., Chermak, D., Frommer, W.B., 2011. Opticalsensors for measuring dynamic changes of cytosolic metabolite levels in yeast.Nat. Protocols 6, 1806–1817.
izzarri, R., Serresi, M., Luin, S., Beltram, F., 2009. Green fluorescent protein basedpH indicators for in vivo use: a review. Anal. Bioanal. Chem. 393, 1107–1122.
orst, J.W., Willemse, M., Slijkhuis, R., van der Krogt, G., Laptenok, S.P., Jalink, K.,Wieringa, B., Fransen, J.A., 2010. ATP changes the fluorescence lifetime of cyanfluorescent protein via an interaction with His148. PLoS One 5, e13862.
ampanini, B., Pioselli, B., Raboni, S., Felici, P., Giordano, I., D’Alfonso, L., Collini, M.,Chirico, G., Bettati, S., 2013. Role of histidine 148 in stability and dynamics of ahighly fluorescent GFP variant. Biochim. Biophys. Acta 1834, 770–779.
ossins, B.P., Jacobson, M.P., Guallar, V., 2011. A new view of the bacterial cytosolenvironment. PLoS Comp. Biol. 7, e1002066.
ale, R.E., Eisinger, J., Blumberg, W.E., 1979. The orientational freedom of molecularprobes. The orientation factor in intramolecular energy transfer. Biophys. J. 26,161–193.
euschle, K., Okumoto, S., Fehr, M., Looger, L.L., Kozhukh, L., Frommer, W.B., 2005.Construction and optimization of a family of genetically encoded metabolitesensors by semirational protein engineering. Protein Sci. 14, 2304–2314.
repper, T., Gensch, T., Pohl, M., 2013. Advanced in vivo applications of blue lightphotoreceptors as alternative fluorescent proteins. Photochem. Photobiol. Sci.12, 1125–1134.
ehr, M., Frommer, W.B., Lalonde, S., 2002. Visualization of maltose uptake in livingyeast cells by fluorescent nanosensors. Proc. Nat. Acad. Sci. 99, 9846–9851.
ehr, M., Lalonde, S., Lager, I., Wolff, M.W., Frommer, W.B., 2003. In vivo imagingof the dynamics of glucose uptake in the cytosol of COS-7 cells by fluorescentnanosensors. J. Biol. Chem. 278, 19127–19133.
ehr, M., Takanaga, H., Ehrhardt, D.W., Frommer, W.B., 2005. Evidence for high-capacity bidirectional glucose transport across the endoplasmic reticulummembrane by genetically encoded fluorescence resonance energy transfernanosensors. Mol. Cell. Biol. 25, 11102–11112.
örster, T., 1948. Zwischenmolekulare Energiewanderung und Fluoreszenz. Annalender Physik 437, 55–75.
riesbeck, O., Baird, G.S., Campbell, R.E., Zacharias, D.A., Tsien, R.Y., 2001. Reducingthe environmental sensitivity of yellow fluorescent protein: mechanism andapplications. J. Biol. Chem. 276, 29188–29194.
ou, B.H., Takanaga, H., Grossmann, G., Chen, L.Q., Qu, X.Q., Jones, A.M., Lalonde, S.,Schweissgut, O., Wiechert, W., Frommer, W.B., 2011. Optical sensors for mon-itoring dynamic changes of intracellular metabolite levels in mammalian cells.Nat. Protocols 6, 1818–1833.
urwitz, C., Rosano, C.L., 1967. The intracellular concentration of bound andunbound magnesium ions in Escherichia coli. J. Biol. Chem. 242, 3719–3722.
ones, A.M., Danielson, J.A.H., ManojKumar, S.N., Lanquar, V., Grossmann, G., From-mer, W.B., 2014. Abscisic acid dynamics in roots detected with geneticallyencoded FRET sensors. Elife 3, e01741.
Please cite this article in press as: Moussa, R., et al., An evaluationmetabolite analyses in vivo. J. Biotechnol. (2014), http://dx.doi.org/10
remers, G.J., Goedhart, J., van Munster, E.B., Gadella, T.W.J., 2006. Cyan and yellowsuper fluorescent proteins with improved brightness, protein folding, and FRETForster radius. Biochemistry 45, 6570–6580.
iemburg-Apers, D.C., Imamura, H., Forkink, M., Nooteboom, M., Swarts, H.G.,Brock, R., Smeitink, J.A., Willems, P.H., Koopman, W.J., 2011. Quantitative
PRESSnology xxx (2014) xxx–xxx
glucose and ATP sensing in mammalian cells. Pharmaceut. Res. 28, 2745–2757.
Llopis, J., McCaffery, J.M., Miyawaki, A., Farquhar, M.G., Tsien, R.Y., 1998. Measure-ment of cytosolic, mitochondrial, and Golgi pH in single living cells with greenfluorescent proteins. Proc. Nat. Acad. Sci. 95, 6803–6808.
Lo, C.J., Leake, M.C., Berry, R.M., 2006. Fluorescence measurement of intracellularsodium concentration in single Escherichia coli cells. Biophys. J. 90, 357–365.
Martinière, A., Debrosses, G., Sentenac, H., Paris, N., 2013. Development and prop-erties of genetically encoded pH sensors in plants. Front. Plant Sci. 4, 523.
Miyawaki, A., Griesbeck, O., Heim, R., Tsien, R.Y., 1999. Dynamic and quantita-tive Ca2+ measurements using improved cameleons. Proc. Nat. Acad. Sci. 96,2135–2140.
Miyawaki, A., Llopis, J., Heim, R., McCaffery, J.M., Adams, J.A., Ikura, M., Tsien, R.Y.,1997. Fluorescent indicators for Ca2+ based on green fluorescent proteins andcalmodulin. Nature 388, 882–887.
Newman, R.H., Fosbrink, M.D., Zhang, J., 2011. Genetically encodable fluorescentbiosensors for tracking signaling dynamics in living cells. Chem. Rev. 111,3614–3666.
Okumoto, S., Jones, A., Frommer, W.B., 2012. Quantitative imaging with fluorescentbiosensors. Ann. Rev. Plant Biol. 63, 663–706.
Ormö, M., Cubitt, A.B., Kallio, K., Gross, L.A., Tsien, R.Y., Remington, S.J., 1996. Crys-tal structure of the Aequorea victoria green fluorescent protein. Science 273,1392–1395.
Palmer, A.E., Giacomello, M., Kortemme, T., Hires, S.A., Lev-Ram, V., Baker, D., Tsien,R.Y., 2006. Ca2+ indicators based on computationally redesigned calmodulin-peptide pairs. Chem. Biol. 13, 521–530.
Palmer, A.E., Tsien, R.Y., 2006. Measuring calcium signaling using genetically tar-getable fluorescent indicators. Nat. Prot. 1, 1057–1065.
Potzkei, J., Kunze, M., Drepper, T., Gensch, T., Jaeger, K.E., Büchs, J., 2012. Real-timedetermination of intracellular oxygen in bacteria using a genetically encodedFRET-based biosensor. BMC Biol. 10, 28.
Rizzo, M.A., Springer, G.H., Granada, B., Piston, D.W., 2004. An improved cyan fluo-rescent protein variant useful for FRET. Nat. Biotechnol. 22, 445–449.
Rodriguez-Garcia, A., Rojo-Ruiz, J., Navas-Navarro, P., Aulestia, F.J., Gallego-Sandin,S., Garcia-Sancho, J., Alonso, M.T., 2014. GAP, an aequorin-based fluorescentindicator for imaging Ca2+ in organelles. Proc. Natl. Acad. Sci. 111, 2584–2589.
San Martin, A., Ceballo, S., Baeza-Lehnert, F., Lerchundi, R., Valdebenito, R., Contreras-Baeza, Y., Alegria, K., Barros, L.F., 2014. Imaging mitochondrial flux in single cellswith a FRET sensor for pyruvate. PLoS One 9, e85780.
San Martin, A., Ceballo, S., Ruminot, I., Lerchundi, R., Frommer, W.B., Barros, L.F., 2013.A genetically encoded FRET lactate sensor and its use to detect the Warburgeffect in single cancer cells. PLoS One 8, e57712.
Schilling, K., Cumme, G.A., Hoffmann-Blume, E., Hoppe, H., Horn, A., 1993. Multi-wavelength photometry of thermochromic indicator solutions for temperaturedetermination in multicuvettes. Clin. Chem. 39, 251–256.
Shabala, L., Bowman, J., Brown, J., Ross, T., McMeekin, T., Shabala, S., 2009. Ion trans-port and osmotic adjustment in Escherichia coli in response to ionic and non-ionicosmotica. Environ. Microbiol. 11, 137–148.
Spiering, D., Bravo-Cordero, J.J., Moshfegh, Y., Miskolci, V., Hogdson, L., 2013. Quan-titative ratiometric imaging of FRET-biosensors in living cells. Methods Cell Biol.114, 593–609.
Strohhöfer, C., Forster, T., Chorvat, D., Kasak, P., Lacik, I., Koukaki, M., Karamanou, S.,Economou, A., 2011. Quantitative analysis of energy transfer between fluores-cent proteins in CFP-GBP-YFP and its response to Ca2+. Phys. Chem. Chem. Phys.13, 17852–17863.
Stryer, L., Haugland, R.P., 1967. Energy transfer: a spectroscopic ruler. Proc. Natl.Acad. Sci. U. S. A. 58, 719–726.
Takanaga, H., Chaudhuri, B., Frommer, W.B., 2008. GLUT1 and GLUT9 as majorcontributors to glucose influx in HepG2 cells identified by a high sensitivityintramolecular FRET glucose sensor. Biochim. Biophys. Acta: Biomembranes1778, 1091–1099.
Vyas, N.K., Vyas, M.N., Quiocho, F.A., 1987. A novel calcium binding site in thegalactose-binding protein of bacterial transport and chemotaxis. Nature 327,635–638.
Wachter, R.M., Elsliger, M.A., Kallio, K., Hanson, G.T., Remington, S.J., 1998. Struc-tural basis of spectral shifts in the yellow-emission variants of green fluorescentprotein. Structure 6, 1267–1277.
Watkins, N.J., Knight, M.R., Trewavas, A.J., Campbell, A.K., 1995. Free calcium tran-sients in chemotactic and non-chemotactic strains of Escherichia coli determined
of genetically encoded FRET-based biosensors for quantitative.1016/j.jbiotec.2014.07.007
by using recombinant aequorin. Biochem. J. 306, 865–869.Willemse, M., Janssen, E., Lange, F.d., Wieringa, B., Fransen, J., 2007. ATP and FRET –
a cautionary note. Nat. Biotech. 25, 170–172.Zhang, J., Campbell, R.E., Ting, A.Y., Tsien, R.Y., 2003. Creating new fluorescent probes
for cell biology. Nat. Rev. Mol. Cell Biol. 3, 906–918.
810
811
812
813
814