an evaluation of genetically encoded fret-based biosensors for quantitative metabolite analyses in...

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

dx.doi.org/10.1016/j.jbiotec.2014.07.007


http://www.sciencedirect.com/science/journal/01681656

http://www.elsevier.com/locate/jbiotec

mailto:[email protected]







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-


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


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


http://www.addgene.org/

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


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:


- 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


http://www.liv.ac.uk/buffers/buffercalc.html

http://www.liv.ac.uk/buffers/buffercalc.html



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,


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


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



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.


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


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




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


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


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



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


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


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




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


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


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


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).


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


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


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.


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


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


http://dx.doi.org/10.1016/j.jbiotec.2014.07.007

http://dx.doi.org/10.1016/j.jbiotec.2014.07.007

an evaluation of genetically encoded fret-based biosensors for quantitative metabolite analyses in...

Documents