Temporal Modulation in Fluorescence Spectroscopy and

Temporal Modulation in Fluorescence Spectroscopy and Imaging for Biological

Applications

GUSTAV PERSSON

Doctoral Thesis in PhysicsStockholm, Sweden 2009

KTH Applied PhysicsAlbaNova University CenterExperimental Biomolecular PhysicsSE-106 91 StockholmSWEDEN

TRITA-FYS 2009:13ISSN 0280-316XISRN KTH/FYS/--09:13--SE ISBN 978-91-7415-299-9

Gustav Persson. Temporal Modulation in Fluorescence Spectroscopy and Imaging for Biological Applications

Akademisk avhandling som med tillstånd av Kungliga Tekniska Högskolan framlägges till offentlig granskning för avläggande av teknologie doktorsexamen i fysik onsdagen den 20 maj 2009 klockan 13.00 i sal FB42, AlbaNova Universitetscentrum, Roslagstullsbacken 21, Stockholm. Avhandlingen kommer att försvaras på engelska.

Copyright © 2009 Gustav Persson

The following parts are reprinted with permission:Paper I. Copyright © 2007 American Chemical SocietyPaper II. Copyright © 2008 American Chemical SocietyPaper III. Copyright © 2008 Biophysical SocietyPaper V. Copyright © 2009 the PCCP Owner Societies

Print: Universitetsservice US-AB, Stockholm

iii

AbstractThis thesis explores the benefits of intensity modulation for the pur-pose of extending the range of applications of fluorescence spectros-copy and imaging in cellular and molecular biology and medicine.

Long-lived transient states of fluorescent molecules can, because of their long lifetimes, be used to detect subtle changes in the microen-vironment of the molecule. A method for determining the kinetic rates for transitions to and from such states by registration of changes in the average fluorescence intensity related to different modulation of the excitation source is introduced. It combines the detection sensitivity of fluorescence with the environmental sensitivity of the long-lived transient states and allows the use of slow detectors such as CCD cameras, making parallelization and wide-field imaging pos-sible developments. An extension of this method, generating image contrast based on triplet state population using a standard laser scan-ning microscope, is also shown.

A strategy to combine fluorescence correlation spectroscopy (FCS) with modulated excitation, in a way that allows extraction of cor-relation data for all correlation times, is presented. This enables the use of modulation to optimize measurement conditions with respect to photophysical properties of the dyes used. FCS with modulated excitation will probably prove useful in future studies involving mul-tiple kinetic processes occurring in overlapping time ranges. One of the ideas from this project also constitutes a powerful method for generating artifact free correlation curves from data sets where sec-tions have been removed. This is potentially very useful in biological studies where spikes in the measurements often cause problems.

In the final project, cross-correlation and alternating excitation are combined in measurements on a pH-sensitive ratiometric dye to clearly distinguish the protonation–deprotonation dynamics from other processes. The presented approach makes the protonation relat-ed fluctuations manifest themselves as a very distinct anti-correlating component in the correlation curve. This enables robust data analysis using a simple model.

iv

Sammanfattning (summary in Swedish)Det finns många metoder som är användbara för biologiska och medicinska studier, men fluorescensdetektion är ett av få verktyg som har den känslighet som behövs för detektion och karaktärisering av enstaka molekyler. Detta beror till stor del på att det ljus som sänds ut (emitteras) av en fluorescerande molekyl får en annan färg (våglängd) än det ljus som används för belysning (excitation) och därmed kan separeras från spritt ljus med relativt enkla medel. Den exceptionella känsligheten har gjort att fluorescens på bred front sla-git igenom som utläsningsmetod och numera används dagligen för diagnostiska syften på sjukhus, för gallring av möjliga läkemedel och analyser inom läkemedelsindustrin samt för grundforskning inom biologi och medicin.

Genom att belysa ett prov i ett mikroskop med en fokuserad laser-stråle kan fluorescens detekteras från en väldigt liten volym. Fluo-rescensintensiteten från denna volym kommer, om koncentrationen av fluorescerande molekyler är tillräckligt låg, fluktuera p.g.a. att molekyler rör sig in i eller ut ur fokus eller till följd av kemiska eller fotofysikaliska processer. Dessa fluktuationer kan analyseras med s.k. fluorescenskorrelationsspektroskopi (FCS). Därigenom fås informa-tion om genomsnittligt antal molekyler i fokus utifrån fluktuationer-nas relativa storlek, molekylernas storlek utifrån deras genomsnittliga passagetid genom fokus eller hastigheter och jämvikter hos kemiska reaktioner.

Fluktuationer beroende på att de flesta fluorescerande molekyler un-der belysning spontant växlar mellan mörka och ljusa tillstånd spelar en central roll i denna avhandling. Sådana mörka (icke fluorescerande) tillstånd är ofta väldigt långlivade jämfört med det tillstånd från vilket emissionen av fluorescens äger rum. Detta innebär att en molekyl i ett sådant tillstånd har relativt lång tid på sig att interagera med sin omgivning och därigenom tvingas tillbaka till det emitterande tillståndet. Mätning av t.ex. andelen molekyler som uppehåller sig i det mörka tillståndet är alltså ett mycket känsligt verktyg för att skaffa sig information om molekylernas omedelbara omgivning. FCS har den stora fördelen att genom detektion av fluorescens kunna

v

karaktärisera dessa mycket omgivningskänsliga tillstånd som i sig inte ger upphov till någon användbar signal. En nackdel med att an-vända FCS till detta är dock att det krävs mycket känsliga detektorer med hög tidsupplösning och lågkoncentrerade prover med mycket ljusstarka molekyler.

I artikel I presenteras en metod för karaktärisering av ett mörkt tillstånd, det s.k. triplett-tillståndet, genom att excitationen sker puls-vis (modulerat) och genomsnittlig fluorescensintensitet mäts under en relativt lång tid. Förändring av pulslängder och tid mellan pulser påverkar fluorescensintensiteten och detta möjliggör beräkning av triplett-tillståndets egenskaper. På detta vis kodas tidsinformationen i belysningen och detektionen kan ske med billigare detektorer utan krav på tidsupplösning. Metoden som sådan är direkt tillämpbar även på andra fotoinducerade tillstånd. En vidareutveckling av detta koncept demonstreras i artikel II, där ett vanligt fluorescensmik-roskop med svepande laser (LSM) används för att generera bilder där kontrasten ges av andelen molekyler i triplett-tillståndet. Detta är av intresse för biologiska studier bl.a. för att triplett-tillståndet är mycket känsligt för syrekoncentrationen i omgivningen.

Eftersom modulerad excitation kan användas för att påverka andelen molekyler i mörka tillstånd är det intressant även för FCS-mätningar av flera anledningar. Olika processer påverkas i olika utsträckning av modulationen beroende på hur snabba de är i förhållande till pulslängder och pulsseparationer. Detta ger möjlighet att selektivt undertrycka en viss process. Om man t.ex. inte är intresserad av triplett-tillståndet kan populationen av detta undertryckas, vilket ger starkare signal för att färre molekyler befinner sig i det mörka tillstån-det. Mindre triplett-population minskar också fotoblekningen som är ett allvarligt problem i många sammanhang, särskilt vid mätningar i celler eftersom den begränsade volymen innebär att det finns få fluorescerande molekyler som kan ersätta en som blekts. Eftersom FCS bygger på fluktuationsanalys kommer dock även de fluktua-tioner som uppkommer p.g.a. att belysningen varierar att ha en stark inverkan på mätningarna. I artikel III presenteras två metoder för att ta bort denna inverkan. Den ena fungerar tillsammans med traditio-

vi

nell apparatur men förutsätter att den genomsnittliga fluorescensin-tensiteten inte varierar inom en excitationspuls. Den andra metoden fungerar mer generellt men ställer också större krav på datainsamlin-gen. I artikel IV demonstreras vidare att den enklare metoden utgör ett väl fungerande verktyg för att återskapa FCS-kurvor från mätdata där delar tagits bort, exempelvis p.g.a. intensitetsspikar i mätningen. Metoden visas ge bättre resultat och vara mer generellt användbar än den idag vanligaste metoden.

I artikel V demonstreras ytterligare en metod för att tydligt urskilja en specifik process i en FCS-mätning. I detta fall handlar det om protonering och deprotonering av en färgämnesmolekyl, d.v.s. upptagande och avgivande av en proton. Här används ett färgämne vars excitations- och emissionsspektra båda förskjuts mot den blå än-den av det synliga spektret vid protonering. Vid separat detektion av ljus från det protonerade respektive det deprotonerade tillståndet och efterföljande korskorrelation av dessa båda signaler, ger de protoner-ingsrelaterade fluktuationerna upphov till en antikorrelation som lätt kan urskiljas från övriga processer och möjliggör analys med hjälp av en förhållandevis enkel och robust modell.

Det gemensamma målet för de presenterade projekten har varit att utvidga området av möjliga tillämpningar för fluorescensspektroskopi inom biologi och biomedicin. Flera nya metoder har utvecklats, som alla löser specifika relevanta problem. De kommer därför sannolikt komma till användning bland nästa generations fluorescensmetoder för biomolekylära studier.

vii

List of publicationsThis thesis is based on the following papers, which will be referred to by their roman numerals:

I Monitoring Kinetics of Highly Environment Sensitive States of Fluorescent Molecules by Modulated Excitation and Time-Averaged Fluorescence Intensity RecordingTor Sandén, Gustav Persson, Per Thyberg, Hans Blom and Jerker WidengrenReproduced with permission from Analytical Chemistry, 2007, 79(9): p. 3330-3341doi: 10.1021/ac0622680Copyright © 2007 American Chemical Society

II Transient State Imaging for Microenvironmental Monitoring by Laser Scanning MicroscopyTor Sandén, Gustav Persson and Jerker WidengrenReproduced with permission from Analytical Chemistry, 2008, 80(24): p. 9589-9596 doi: 10.1021/ac8018735Copyright © 2008 American Chemical Society

III Modulated Fluorescence Correlation Spectroscopy with Complete Time Range InformationGustav Persson, Per Thyberg and Jerker WidengrenReproduced with permission from Biophysical Journal, 2008, 94(3): p. 977-985doi: 10.1529/biophysj.107.113332Copyright © 2008 Biophysical Society

IV Modulation Filtering Enables Removal of Spikes in Fluorescence Correlation Spectroscopy Measurements without Affecting the Temporal InformationGustav Persson, Per Thyberg, Tor Sandén and Jerker WidengrenSubmitted for publication

http://dx.doi.org/10.1021/ac0622680

http://dx.doi.org/10.1021/ac8018735

http://dx.doi.org/10.1529/biophysj.107.113332

viii

V Fluorescence Cross-Correlation Spectroscopy of a pH-Sensitive Ratiometric Dye for Molecular Proton Exchange StudiesGustav Persson, Tor Sandén, AnnSofi Sandberg and Jerker WidengrenReproduced with permission from Physical Chemistry Chemical Physics, 2009, doi: 10.1039/b822494cCopyright © 2009 the PCCP Owner Societies

Other publicationsThe respondent has also contributed to the following related publica-tions, which are not included in this thesis.

A Modulated or Alternating Excitation in Fluorescence Correlation SpectroscopyGustav Persson, Tor Sandén, Per Thyberg and Jerker WidengrenProc. of SPIE, 2009, 7185, 718509doi: 10.1117/12.808912

B Transient State Microscopy: A New Tool for Biomolecular Imaging Tor Sandén, Gustav Persson and Jerker Widengren Proc. of SPIE, 2009, 7183, 71832Rdoi: 10.1117/12.814908

C Time-Varying Excitation in Fluorescence Spectroscopy for Biological ApplicationsGustav PerssonLicentiate Thesis in Physics, 2007, Royal Institute of Technology, Stockholm, SwedenISBN: 978-91-7178-713-2

http://dx.doi.org/10.1039/b822494c

http://dx.doi.org/10.1117/12.808912

http://dx.doi.org/10.1117/12.814908

http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-4399

ix

List of abbreviationsALEX alternating excitationAOM acousto-optic modulatorAPD avalanche photodiodeCCD charge-coupled deviceCLSM confocal laser scanning microscopeCMOS complementary metal oxide semiconductorCW continuous waveDLS dynamic light scatteringDMA direct memory accessFCCS fluorescence cross-correlation spectroscopyFCS fluorescence correlation spectroscopyFIFO first in first outFRAP fluorescence recovery after photobleachingFWHM full width at half maximumHEPES 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acidHOMO highest occupied molecular orbitalI/O input/outputKI potassium iodideLSM laser scanning microscopeLUMO lowest unoccupied molecular orbitalNA numerical aperturePIE pulsed interleaved excitationPSF point-spread functionQELS quasi-elastic light scatteringRD rhodol derivativeRh110 rhodamine 110Rh6G rhodamine 6GRTSI real-time system integrationTEMPO 2,2,6,6-tetramethylpiperidine-1-oxyl

x

ContentsAbstract iiiSammanfattning (summary in Swedish) ivList of publications vii

Other publications viiiList of abbreviations ixTable of contents x1 Introduction 12 Background 4

2.1 Fluorescence and photophysics 42.1.1 Electronic states 52.1.2 Absorption and emission spectra 72.1.3 Fluorescence lifetime and quantum yield 92.1.4 Fluorescence polarization and anisotropy 102.1.5 Triplet kinetics 12

2.2 Confocal laser scanning microscopy 132.3 Fluorescence spectroscopy 16

2.3.1 Fluorescence correlation spectroscopy 172.3.2 Dual-color fluorescence cross-correlation

spectroscopy 223 Transient state monitoring 23

3.1 Traditional techniques 243.2 Utilizing the effect of modulation on the fluorescence

intensity 263.3 Point measurements 273.4 Imaging 29

4 FCS and modulation 344.1 Modulation filters 364.2 Statistical properties of modulation filtered FCS curves 384.3 Pulse trains 414.4 Measurements with modulated excitation 424.5 Modulation filtering of data where spikes have been

removed 465 Ratiometric cross-correlation 50

5.1 The fluorophore 515.2 FCCS with alternating excitation 52

xi

6 Data acquisition and modulation control 597 Conclusions and future outlook 638 Introduction to the included papers 659 Acknowledgements 69References 72

1

Chapter 1

INTRODUCTION

Fluorescence detection provides several benefits that have col-lectively contributed to the very wide spread of presently available fluorescence techniques. The excellent detection sensitivity is one important contributor. In the last decades this has, by a combination of properties inherent to the fluorescence emission and advances in optics and electronics, reached a level enabling detection and char-acterization of single fluorescently labeled molecules in their native environment. In 1976 Hirschfeld demonstrated with total internal reflection microscopy that fluorescence offers the sensitivity needed for observation of single antibodies in solution labeled with between eighty and hundred fluorescein molecules each[1]. A decade later fluo-rescence detection of single B-Phycoerythrin molecules, containing 34 bilin chromophores, in aqueous solution using hydrodynamically focused flow and confocal microscopy was reported[2, 3]. In 1989 the first single fluorophores were detected, but this time in a cryogenic solid. The samples studied were pentacene molecules in p-terphenyl crystals at liquid helium temperatures[4, 5]. Shera et al. reported the first detection of single fluorophores in aqueous solution in a flow cell using pulsed excitation and time-gated detection in 1990[6]. The

1 Introduction

2

same year Rigler et al. demonstrated detection sensitivity sufficient for freely diffusing single fluorophores with continuous wave (CW) excitation and an epi-illuminated confocal microscope setup[7].

In many cases specificity is even more important than sensitivity. Fluorescence detection provides molecular specificity, allowing ob-servation of labeled molecules in the presence of a vast number of unlabeled molecules. Additionally, the spectral signatures of differ-ent fluorophores enable simultaneous monitoring of several species. Emission intensity, wavelength, polarization and fluorescence lifetime are measurement parameters that can be recorded simultaneously to provide a diverse set of information about the immediate environ-ment of a fluorophore or a about molecule tethered to it. This kind of multiplexing has proven useful in high-throughput screening[8] and fundamental biomolecular studies based on single-molecule detec-tion[9, 10]. It has been applied in flow cytometry[11] and fluorescence microscopy[12], as well as for readout of DNA and protein microar-rays[13].

The field has expanded rapidly, and today fluorescence is used on a daily basis for diagnostic purposes in hospitals, for screening and analysis in the pharmaceutical industry and for fundamental research in life sciences.

The focus of this thesis is the application of intensity modulation, mainly on the excitation side, in fluorescence spectroscopy and imag-ing. The aim of the work has been to further expand the range of possible applications. Five papers are included, four of which present new experimental approaches. The fifth paper (paper IV) is concerned with post-processing of measurement data in fluorescence correlation spectroscopy (FCS)[14] and makes use of the findings of the earlier papers regarding the effects of modulation.

The intention of the thesis is to summarize my work, performed in cooperation with my coworkers in the group for Experimental Biomolecular Physics at the Royal Institute of Technology (KTH) in Stockholm, and put it into context. The thesis begins with a background chapter on fluorescence in general and the techniques on which the work was based. The following three chapters present

3

Temporal Modulation in Fluorescence Spectroscopy and Imaging for Biological Applications

the work behind the five papers. A separate chapter provides a short presentation of the strategy for data collection and modulation con-trol employed in the studies. This is followed by the conclusions and future outlook. Papers I-V constitute the final part of this thesis.

4

Chapter 2

BACKGROUND

2.1 Fluorescence and photophysicsLight emitted when electrons decay from an excited state is referred to as luminescence. The term luminescence was first introduced by Eilhardt Wiedemann in 1888 and comes from the Latin word lu-men, which means light. Photoluminescence is a form of lumines-cence where the decaying electron was initially excited through the absorption of a photon, i.e. light induced luminescence. This in turn comprises two distinct phenomena: fluorescence and phosphores-cence[15, 16].

The first reported observation of fluorescence was made by the Span-ish physician Nicolas Monardes. The year was 1565 and he described blue light from an infusion of the tropical hardwood called narra or Pterocarpus indicus, which was then known as Lignum nephriti-cum. The word fluorescence was first used by Sir George Gabriel Stokes, physicist and professor of mathematics at Cambridge, in his description of an experiment conducted in 1852 where he used the ultraviolet part of a spectrum of sunlight to excite fluorescence of quinine sulphate[17]. He derived the word from the mineral fluorspar

5


(fluor- from fluere, which in Latin means to flow, because the mineral was easily melted), which contains calcium fluoride and commonly exhibits fluorescent behaviour. In this paper he also stated that the light emitted as fluorescence is always of longer wavelength than the exciting light. This statement is now called Stokes’ law.

Ten years earlier the French physicist Edmond Becquerel had studied light emitted from calcium sulphide deposited on paper and also concluded that this was red-shifted in respect to the exciting light[18]. In this case, however, the studied phenomenon was phosphorescence. This term comes from the Greek words φως (“fos”) and φορειν (“for-ein”), which mean light and to bear, respectively. The term phosphor has been used since the Middle Ages for materials glowing in the dark after exposure to light. In the nineteenth century this property was still used to distinguish between phosphorescence and fluorescence: in phosphorescence the light emission persisted after the excitation was stopped, whereas in fluorescence it stopped instantly. Now the distinction is made based on the different molecular electronic states involved. The first theoretical distinction approaching the current view was made by Francis Perrin in 1929[19].

2.1.1 Electronic statesExcitation of a fluorescent molecule normally means promotion of an electron from the highest occupied molecular orbital (HOMO) to the lowest unoccupied molecular orbital (LUMO), referring to the electron configuration in the ground state of the molecule (figure 2.1A). In most fluorescent dyes the ground state is a singlet state, i.e., each electron is paired with another one with opposite spin. Thus all spins cancel out, yielding a total spin quantum number S = 0, and multiplicity M = 2S + 1 = 1 (hence singlet). The ground state is denoted S0. When excited, the electron keeps its spin, meaning that the excited state is also a singlet state, S1. For each electronic state, a molecule has a large number of possible vibrational energy levels. Electronic transitions are much faster than molecular vibrations, and hence the atomic nuclei in a molecule do not have time to move significantly during a transition. This is referred to as the Franck-Condon principle and implies that an electronic transition is most

2 Background

6

likely to occur in conjunction with a vibrational transition between levels with overlapping vibrational wavefunctions. When excited, the molecule most often ends up on a higher vibrational level of S1 but quickly relaxes to the lowest vibrational level of S1 through internal conversion, meaning that the molecule dissipates vibrational energy through collisions with solvent molecules, which occurs within a few picoseconds. From S1, there are several possibilities for the molecule to relax back to its ground state. Relaxation by internal conversion is less likely than to S1 from higher states, because of the larger energy gap between S1 and S0. Nevertheless, for most molecules, including many fluorophores, this is the dominant relaxation process. If the molecule relaxes from S1 to S0 by emitting a photon, the emission is called fluorescence. In some cases, fluorescence may also occur from higher excited singlet states like S2. A third possibility is what is called intersystem crossing, which refers to the molecule crossing from the system of singlet electronic states to the system of triplet states. The spin of the promoted electron is reversed, yielding a total spin quan-tum number of 1 and a multiplicity of 3.

Figure 2.1: A. Population of the highest energy orbitals of a typical fluorescent molecule in different states. The arrows represent electrons with spins in the in-dicated directions. B. Jablonski diagram of a fluorophore, schematically show-ing three electronic states and some of the corresponding vibrational levels.

A B

S0

S1

T1

kISC

kT

k10 k

01

ELUMO π*

πHOMO

S0

S1

T1

In many cases, a three state model including the states S0, S1 and the first excited triplet state T1 is both physically adequate and mathematically convenient to describe the photophysics behind a fluorescence experiment, as long as the excitation intensity is kept sufficiently low[20, 21]. Such a model is often illustrated by an energy level diagram called a Jablonski diagram (figure 2.1B). Each transi-tion between two states is characterized by a transition rate. In this thesis k01, k10, kISC and kT will be used to denote the rate constants for

7


excitation from S0 to S1, relaxation from S1 to S0, intersystem crossing from S1 to T1 and relaxation from T1 to S0, respectively.

In a pure spin system, intersystem crossing is quantum mechanically forbidden. However, due to interaction between orbital and spin angular momenta, so called spin-orbit coupling, the wave functions of the singlet states contain small fractions of triplet state wave func-tions and vice versa. This makes the singlet–triplet transition pos-sible, although normally with very low probability. Upon intersystem crossing from S1 the molecule enters the triplet state T1. A triplet state always has a lower energy than the corresponding singlet state. Emission of a photon due to relaxation from this state is referred to as phosphorescence. Because relaxation requires another intersystem crossing, which is a rather improbable event, the triplet state is very long-lived compared to the S1 state. This explains why phosphores-cence may persist for minutes or even hours at low temperatures or in solid phase, where collisions with surrounding molecules are infre-quent. In solution however, the long life-time of the triplet state leads to a higher probability of a collision with another molecule while in this state and a correspondingly higher probability of vibrational relaxation. The relaxation from the triplet state is efficiently enhanced by molecular oxygen, O2, which is normally ubiquitous in aqueous solutions. Finally, the molecule may also return to the excited singlet state via electronic or vibrational excitation. A subsequent emission of a photon is referred to as delayed fluorescence.

2.1.2 Absorption and emission spectraAt thermodynamic equilibrium the vibrational levels of the ground state are populated according to the Boltzmann distribution. Hence, at room temperature most molecules are in the lowest vibrational energy level of S0 and excitation normally takes place from there. Depending on the energy of the exciting photon the molecule may end up on different vibrational levels of S1. Absorption of a higher energy photon, i.e. a photon with shorter wavelength, will put the molecule on a higher vibrational level. The lowest energy, or longest wavelength, of a photon that will lead to electronic excitation of the molecule is given by the energy difference of the lowest vibrational

2 Background

8

level of S1 and the highest populated level of S0 (cf. figure 2.1B).

The absorption spectrum is a curve showing how efficiently a sub-stance absorbs photons of different wavelengths. At low temperatures or in solid phase the spectrum is typically highly structured with dis-tinct absorption lines corresponding to transitions between different vibrational levels. In solution at room temperature, however, most dyes exhibit a smooth absorption spectrum due to broadening of the absorption lines. The broadening is of two types: homogeneous broadening due to tightly spaced vibrational levels, giving rise to a near-continuous spectrum, and inhomogeneous broadening, which in liquid samples is caused mainly by collisions between molecules. Mostly, the inhomogeneous broadening is dominant. A typical ab-sorption spectrum of a fluorescent dye in solution is shown in figure 2.2A. As stated above, fluorescence emission predominantly occurs from the lowest vibrational level of S1, due to the fast internal conver-sion from higher levels. Since the spacing of the vibrational levels in the different electronic states is similar, the emission spectrum often resembles a mirror image of the absorption spectrum (figure 2.2A).

Figure 2.2: A. Normalized absorption (solid curve) and emission (dashed curve) spectra of rhodamine 6G. B. Structure formulae for two of the fluorescent dyes used in the work presented in this thesis.

A B

H3C

COOC2H5

CH3

C2H5 C2H5

HN O NH+

COO-

-O O O

rhodamine 6G

dianionfluorescein

400 500 600 7000.0

0.2

0.4

0.6

0.8

1.0

λ / nm

Most fluorescent molecules are organic compounds containing chains or rings of carbon with alternating single and double bonds, so called conjugated systems. Figure 2.2B shows the structure of two of the dyes used in the work presented in this thesis: rhodamine 6G (Rh6G) and fluorescein. A double bond consists of one pair of electrons in a σ-orbital and one pair in a π-orbital. A σ-orbital can be formed either

9


from two atomic s-orbitals, one s- and one p-orbital, or from two atomic p-orbitals with symmetry axes along the bond. A σ-orbital is rotationally symmetric around the bond. A π-orbital is formed by two atomic p-orbitals overlapping laterally. In fluorescent dyes the highest occupied molecular orbital (HOMO) is often a bonding π-orbital and the lowest unoccupied molecular orbital (LUMO) is often an antibonding π*-orbital (see figure 2.1A). Absorption of a photon with appropriate energy can induce a π → π* transition. In a conjugated system the π-orbitals of the double bonds overlap and allow the π-electrons to be delocalized over the whole system. As a rule of thumb, the larger the conjugated system, the lower the energy needed for the π → π* transition. Consequently, a molecule contain-ing a long chain of carbons with alternating single and double bonds has its absorption maximum at a longer wavelength than a molecule with a shorter carbon chain. Absorption spectra of the common cya-nine dyes Cy5 and Cy3 may serve as an example here (figure 2.3).

Figure 2.3: A. Structure formulae for the cyanine dyes Cy5 and Cy3 show-ing the difference in extent of the conjugated systems, i.e. the carbon chains with alternating single and double bonds. B. Normalized absorption spectra of the same dyes illustrating how the extent of the conjugated system affects the energy needed for excitation.

2.1.3 Fluorescence lifetime and quantum yieldThe fluorescence lifetime is the time constant of the fluorescence decay after excitation is instantly turned off. Since the fluorescence intensity is proportional to the population of the S1 state this is equivalent to defining the fluorescence lifetime as the average time a molecule spends in the excited state following excitation. The life-

A B-O3S

N+

N SO3-

-O3S

N+

N SO3-

Cy5

Cy3

400 500 600 700

0.0

0.2

0.4

0.6

0.8

1.0

λ / nm

Cy3 Cy5

2 Background

10

time of a fluorophore in the absence of any nonradiative relaxation processes is called the intrinsic (or natural) lifetime and is equal to the inverse of the radiative rate. In principle, the intrinsic lifetime can be calculated from the absorption and emission spectra of a fluorophore using the Strickler-Berg relation[22]:

1

n

= kr =4 n2c ln10

5N A

F ( )d3F ( )d

( )d . (1)

Here F(λ) is the fluorescence emission spectrum in arbitrary units and ε(λ) is the absorption spectrum, n is the refractive index of the medium, c the vacuum speed of light and NA Avogadro’s number. This works well in many cases but assumes no interaction with the solvent, no difference in refractive index between the absorption and emission wavelengths and no change in the excited state geometry. If the fluorescence emission competes with nonradiative processes the observed lifetime becomes shorter than the intrinsic one.

The fluorescence quantum yield of a fluorophore is defined as the ratio of the number of emitted photons and the number of absorbed photons. The fraction of molecules that decay through fluorescence emission, i.e. the quantum yield, is given by:

Φ=

kr

kr + knr

, (2)

where kr is the radiative rate and knr is the sum of the rates for all competing nonradiative processes. The observed lifetime becomes:

τ =Φτn. (3)

2.1.4 Fluorescence polarization and anisotropyThe transition of an electron from one orbital to another induces a transient displacement of charges in the molecule. The transition moment represents the resulting transient dipole. For a π → π* transition in an aromatic hydrocarbon the transition moment lies in the plane of the conjugated molecule. A transition can only be induced by an electromagnetic field with a non-zero electric com-ponent along the transition moment. The absorption efficiency is

11


proportional to the square of the scalar product of the electric field vector and the transition moment, i.e. proportional to the square of the cosine of the angle between the two vectors. The absorption is most efficient for photons polarized parallel to the transition mo-ment. Photons polarized perpendicular to the transition moment cannot be absorbed. The selective absorption of photons of differ-ent polarization is known as photoselection. A photon emitted as fluorescence is polarized along the transition moment. It should be noted that transition moments of a molecule generally have different directions for different electronic transitions. If excitation and emis-sion occur through different transitions, for instance if the molecule is excited to a higher electronic state than the one from which the emission occurs, the excitation and emission transition moments are not necessarily parallel.

For measurement of anisotropy the fluorescence is split into two polarization components, one parallel to the polarization of the ex-citation and one perpendicular. The anisotropy r is defined as the difference between the two detected intensities, normalized by the total intensity:

r =

I ||− I⊥

I || + 2I⊥

. (4)

The factor two in the denominator takes into account that there are two equivalent polarization components, perpendicular to the excita-tion and to each other, contributing to the total emission.

Assuming molecules with random orientation the photoselection gives a theoretically maximal anisotropy of 0.4. This is called the fundamental anisotropy r0, and is what would be expected from molecules in a dilute solution immediately following excitation by a Dirac pulse. Fluorescence depolarizes and the anisotropy decays with time due to rotational diffusion or torsional motion of the molecules. Assuming depolarization only due to rotational diffusion the anisot-ropy decays according to the Perrin equation:

r τ( ) =

r0

1+ τ θ( ), (5)

2 Background

12

where θ is the rotational correlation time and τ the observed life-time.

2.1.5 Triplet kineticsIt has already been mentioned that the triplet state is long-lived due to the low probability of the spin-forbidden intersystem crossing. The transitions are possible thanks to spin–orbit coupling. This coupling gets stronger with increasing atomic number. In flurophores consist-ing of aromatic hydrocarbons the spin–orbit coupling is normally weak due to the relatively light atoms they consist of. The strength of the spin–orbit coupling may be increased by substitution of some constituents with heavier atoms, and thus the intersystem crossing rates may be increased. This is known as the internal heavy atom effect and increases triplet yield and makes triplet relaxation faster[23]. Dyes containing halogens, such as Eosin and Erythrosin have proven useful in phosphorescence spectroscopy thanks to their high triplet yields[24]. Heavy atoms in the surrounding solution have a similar effect, the external heavy atom effect[25].

The long lifetime of the triplet state gives the molecule a lot of time to interact with its surrounding while in this state. This makes the triplet state an excellent probe for changes in the local environment.

Given the three-state model illustrated in figure 2.1B the probabili-ties S0, S1 and T1 of a fluorophore, located at a point r at time t, occupying the corresponding states can be calculated by solving the differential equation system[26]:

ddt

S0 r,t( )

S1 r,t( )

T1 r,t( )

⎛

⎝

⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜

⎞

⎠

⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟

=

−k01 r,t( ) k10 kT

k01 r,t( ) - kISC + k10( ) 0

0 kISC −kT

⎡

⎣

⎢⎢⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥⎥⎥

S0 r,t( )

S1 r,t( )

T1 r,t( )

⎛

⎝

⎜⎜⎜⎜⎜⎜⎜⎜⎜⎜

⎞

⎠

⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟

. (6)

The solution will have one zero eigenvalue meaning that for constant illumination the system will converge towards a steady-state solution. The inverses of the two non-zero eigenvalues are the fluorescence anti-bunching time[27] and the time constant for the triplet popula-tion build-up.. For a detailed description see paper I.

13


2.2 Confocal laser scanning microscopyIn confocal microscopy focused illumination of the object is combined with some mechanism, usually a small aperture in a plane conjugate to the illumination focus called a confocal pinhole, that suppresses detection of light not originating from the focal plane. Marvin Min-sky, then a post-doctoral fellow at Harvard University, applied for a patent for a stage-scanning confocal microscope in 1957[28]. A similar construction had also been suggested for microspectrophotometry by Hiroto Naora six years earlier[29]. It would, however, take another three decades, until 3-dimensional images of fluorescently labeled biological samples were demonstrated by use of laser scanning confo-cal microscopes[30-32], before the technique really started gaining in popularity[33].

The resolution of an optical microscope detecting transmission or scattering is limited by diffraction. Imaging of a point using finite sized optics results in a diffraction pattern, i.e. a spread out intensity distribution. The larger the optical element used for imaging, the less blurring of the image. The property of a microscope objective determining the resolving power is its numerical aperture (NA):

NA = n sinθ, (7)where n is the refractive index of the medium and θ is the the half angle of the cone from which the objective is able to collect light (figure 2.4A).

The intensity pattern generated when imaging a point source into the object plane of the objective is called the point-spread function (PSF) and determines the resolution of the microscope. Laterally in the focal plane, the PSF takes the form of an Airy disc (figure 2.4B), which can be expressed mathematically by:

J12ρ( )

ρ2

, where ρ=

2πλ

rNA, (8)

r is the radial space coordinate, and J1 is the first-order Bessel-func-tion of the first kind. According to the Rayleigh-criterion two equally bright point sources can just be resolved if the intensity maximum in the diffraction pattern of one coincides with the first dark fringe

2 Background

14

of the other (figure 2.4C-D). This gives an intensity dip of 26% be-tween the two intensity maxima. The radius of the first dark fringe of the Airy-disc, and consequently the lateral resolution is given by:

Δrwide-field =

0.61λNA

. (9)

The above expression for the resolution is valid for a microscope with wide-field illumination. In a confocal microscope, the PSF of the illumination source, i.e. the intensity distribution of the illumination focus, combines with the image of the confocal pinhole in object space to form the detection volume (figure 2.4E). In the limit of an infinitely small pinhole, its image will be identical to the PSF and the confocal resolution limit is defined by the squared PSF (figure 2.4F). Defining the resolution, analogously to the wide-field case, as the distance between two points that generates a 26% intensity dip between the maxima of their squared PSFs gives[34]:

Δrconfocal =

0.44λNA

. (10)

A proper calculation of the resolution for a pinhole of finite size requires a convolution of the geometric image of the pinhole in the object plane with the detection PSF. However, for commonly used pinhole sizes, there is only a minor difference in the result.

Along the optical axis the PSF has the form:

sin ζ4ζ

4

⎛

⎝

⎜⎜⎜⎜⎜⎜⎜⎜

⎞

⎠

⎟⎟⎟⎟⎟⎟⎟⎟⎟⎟

2

, where ζ =

2πnλ

NA2z , (11)

and z is the coordinate in the axial direction with its origin in the focus (figure 2.4A). With a definition analogous to the one above the axial resolution of a confocal microscope, based on the square of equation 11, becomes:

Δzconfocal =

1.5nλNA2 . (12)

Thus, in aqueous solution (n = 1.5), the axial resolution is about four times worse than the lateral, even with an objective with high NA

15


(~1.2). It is also common to use the full width at half maximum of the reflected intensity, when moving a surface along the optical axis through the focus, as a measure of the axial resolution:

FWHM = 0.84Δzconfocal =

1.26nλNA2 . (13)

Apparently, the radial resolution of a confocal microscope is slightly

Figure 2.4: A. Illustration of the definition of numerical aperture and the co-ordinate system used in the discussion regarding resolution. B. The airy disk describing the point-spread function (PSF) of an optical system with circular apertures. C. The Rayleigh criterion for resolution. D. The detected intensity distribution from two point-sources just resolved according to the Rayleigh cri-terion. E. The principle of confocal microscopy. The images of the light source and the detection pinhole combine in the object plane to define the observation volume. The confocal pinhole blocks most light emanating from out-of-focus sources. F. The squared PSF defining the radial extent of the observation vol-ume in a confocal microscope.

C D

E F

A B

z

r

θ

focalplane

objective dichroicmirror

light source

confocalpinhole

2 Background

16

better than that of a wide-field microscope, but the most important feature is the axial confinement of the detection (figure 2.4E)[35, 36]. This enables the acquisition of highly resolved 3-dimensional images, if the detection volume is scanned over the sample or the sample is moved through the focus. In combination with fluorescent labeling it also provides the sensitivity needed for single-molecule detection. The Stokes-shift makes it possible to spectrally separate emitted fluorescence from excitation light and thereby contributes to the excellent sensitivity. For fluorescence microscopy the sample is often illuminated from the same side as the detection takes place (as shown in figure 2.4E) to eliminate the contribution to the background of transmitted excitation light entering the detection pathway. This is called epitaxial illumination or epi-illumination and allows the use of the same objective for illumination and detection, which facilitates instrument alignment.

One implementation of a confocal microscope for 3-dimensional imaging is the confocal laser scanning microscope (CLSM or just LSM), where most often mirrors mounted on galvanometers are inserted in pupil planes in the common illumination and detection beam path to simultaneously scan the focus over the sample and descan the emitted light. This allows the latter to be imaged on a fixed detection pinhole (figure 2.5). Scanning in the axial direction is normally performed by refocusing, by moving either the sample stage or the objective. For a thorough exposition of the subject, see the Handbook of Biological Confocal Microscopy[33].

2.3 Fluorescence spectroscopyFluorescence based read-out typically offers high specificity and out-standing sensitivity. Thanks to the sensitivity fluorescence spectros-copy has developed into one of the major means for single molecule measurements and fundamental biomolecular studies. Much of the work presented in this thesis concerns method development in the field of fluorescence correlation spectroscopy (FCS)[14] or makes use of FCS as a reference in the development of related techniques.

17


2.3.1 Fluorescence correlation spectroscopyFCS is a technique that makes use of intrinsic fluctuations to gain information of the dynamics of a system at thermodynamic equilib-rium (see [37-42] for reviews). Two of its greatest virtues are its non-invasiveness, i.e. the possibility to study the fluctuations of a system around its true unperturbed equilibrium, and its high sensitivity. Vir-tually any dynamic process affecting the fluorescence intensity can be studied by FCS. Fluctuations in the detected fluorescence intensity may arise from the diffusion of molecules in and out of the detection volume or by changes in fluorescence characteristics due to reaction kinetics or conformational changes.

Other techniques for studies of relaxation processes in molecular systems, such as the temperature-jump technique[43] and FRAP (fluorescence recovery after photobleaching)[44] rely on perturbation of the system. Quasi-elastic or dynamic light scattering (QELS or DLS)[45, 46] is another non-invasive technique for studying fluctua-tions manifested by a change in refractive index. This, however, lacks the sensitivity of fluorescence to changes in the molecular structure or local environment.

Figure 2.5: Sketch of the most important optical and opto-mechanical components of a confocal laser scanning microscope. An angular change in a pupil plane, introduced by tilting a scanning mirror, translates into a change in position in an image plane or specifically in the object plane.

detector

galvanometer

galvanometer

excitationlight

objective

confocalpinhole

imaginglens

tubelens

scanninglenses

dichroicmirror

slowscanning mirror

fastscanning mirror

2 Background

18

FCS was introduced in 1972 by Magde et al.[14] who used it to extract the binding and dissociation rates of ethidium bromide binding to double stranded DNA. Ehrenberg and Rigler presented the theory for FCS applied to Brownian rotational motion in 1974[47]. An epi-illuminated confocal microscope setup for FCS was first introduced by Koppel et al. in 1976[48] to suppress background noise. However, the real breakthough did not come until Rigler et al. combined this with modern optics to reduce the detection volume and a high quan-tum yield single photon capable avalanche photodiode (APD) for de-tection in the beginning of the 1990’s[7, 49]. Since then the technique has found numerous applications in life sciences and many further technical developments have been presented, e.g. for improved time resolution[50], further confinement of the excitation[51], reduction of the observation volume[52] and parallelization[53, 54]. Dual-color cross-correlation has been introduced for studies of interaction between two differently labeled species[55, 56] (see next section). Approaches have been presented for calibration free absolute determination of diffusion coefficients based on cross-correlation of the signal from two detection volumes[57, 58] and for extraction of correlation curves from subpopulations in the sample based on fluorescence lifetime[59, 60]. FCS has also been implemented with scanning excita-tion beams[61], e.g., for studies of slowly diffusing entities such as aggregates in membranes[62], and the concept has been extended to imaging applications[63-66].

A typical modern setup for a conventional FCS experiment consists of an epi-illuminated confocal microscope and a pair of detectors in a so called Hanbury Brown and Twiss arrangement[67] to circumvent effects of detector dead-time and after-pulsing (figure 2.6). The laser beam is expanded to fill the back aperture of the objective to create a diffraction-limited focus. A dichroic mirror reflects the excitation light into the objective but transmits the collected fluorescence, which is focused onto the confocal pinhole in the image plane of the microscope. The fluorescence is then divided onto the two detec-tors by a beam splitter. In front of each detector there is a band-pass interference filter to block reflected as well as Rayleigh and Raman scattered excitation light. The signals from the two detectors are then

19


cross-correlated, either in a hardware correlator or by software in a computer, for generation of the FCS curve.

The FCS curve is calculated as the normalized auto-covariance of the detected fluorescence intensity:

G τ( ) =

I t( )I t + τ( )

I t( ) I t + τ( ), (14)

or as the cross-correlation of two signals:

G τ( ) =

I1 t( )I2 t + τ( )

I1 t( ) I2 t + τ( ). (15)

Here, the brackets indicate temporal averages over the time t with τ being the lag time or correlation time.

Assuming that the fluorescence intensity is proportional to the ex-

Figure 2.6: Sketch of a typical confocal FCS setup and illustration of how molecules diffusing into and out of the focus introduce fluctuations in the fluorescence signal. An example of an FCS curve exhibiting correlation com-ponents due to diffusion and singlet–triplet transitions is shown in the lower right corner.

LASER

Objective

Beam expander

Dichroic mirror

Tube lens

Confocal pinhole

Beam splitter

Emission filters

Excitation filter

Cover slideSample

APD

APD10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1

1.0

1.5 A

BτT

2.0

2.5

G

τ / s

τDT =A

BN1

2 Background

20

citation intensity, the fluorescence detection efficiency function of the described setup is defined by the excitation illumination profile, the PSF of the objective and the transmission function of the confo-cal pinhole[49]. This function describes the spatial distribution of the average detected fluorescence intensity and determines the effective sample volume. It can be approximated by a 3-dimensional Gaussian distribution, which in a Cartesian coordinate system with the origin in the center of the focus and the z-direction along the excitation beam can be written:

I x, y,z( ) = I0 exp −

2 x 2 + y 2( )ω0

2 −2z 2

ωz2

⎛

⎝

⎜⎜⎜⎜

⎞

⎠

⎟⎟⎟⎟⎟, (16)

where ω0 and ωz define the lateral and axial extent of the volume, respectively. The implications of this assumption have been studied extensively[68, 69]. For a single species of fluorescent molecules diffus-ing through such a volume the analytical expression for the auto-correlation function was derived by Aragón and Pecora[70]:

G( )=1N

1+ 4D

02

1

1+ 4D

z2

12+1=

=1N

1+D

1

1+ 2D

12+1=GD( )+1. (17)

Here N is the average number of fluorescent molecules in the sample volume and D is their diffusion coefficient, τD the lateral diffusion time and = z 0, which is commonly referred to as the structure parameter, is the ratio of the axial and the lateral extent of the detec-tion volume.

For studies of chemical or photophysical processes leading to fluo-rescence fluctuations on a time scale much faster than the diffusion Palmer and Thompson[71] showed that the correlation function can be written separated into two factors, one for the diffusion and one for the faster process:

G τ( ) = GD τ( )R τ( )+1. (18)

21


This expression is also valid for processes occurring in the same time range as the diffusion, as long as all entities contributing to the fluo-rescence signal have the same diffusion properties. For fluctuations due to singlet–triplet transitions[26] the expression becomes:

G( )=GD( )GT ( )+1=GD( )1 T +Te T

1 T+1, (19)

where T is the average fraction of molecules in the dark triplet state and τT is the relaxation time of the singlet–triplet fluctuations, i.e., the inverse of the lowest non-zero eigenvalue of equation 6. An ex-ample curve is shown in figure 2.6. Triplet state build-up increases with higher excitation intensity and brings with it fluorescence inten-sity saturation. One should keep in mind that this saturation makes the spatial fluorescence intensity distribution flatter and increases the apparent detection volume[26, 72].

Protonation–deprotonation dynamics can be treated in the same way as the triplet transitions, assuming that the protonated (or depro-tonated) state is non-fluorescent[73]. Taking into consideration both singlet–triplet and protonation–deprotonation dynamics the auto-correlation function becomes:

G( )=GD( )GT ( )GP ( )+1=

=GD( )1 T +Te T( ) 1 P + Pe P( )

1 T( ) 1 P( )+1, (20)

provided that the two processes take place independently of each other. Here P denotes the average fraction of molecules in the dark protonated state, and τP is the relaxation time of the corresponding fluctuations. Another model that has been used, and which is equiva-lent to the one above if the characteristic times of the singlet–triplet and protonation–deprotonation are very different, is[21]:

G( )=GD( )1 T P +Te T + P e P( )

1 T P( )+1, (21)

where ′P = P 1−T( ).

2 Background

22

2.3.2 Dual-colorfluorescencecross-correlationspectroscopy

Dual-color fluorescence cross-correlation spectroscopy (FCCS)[55, 56] is an extension of FCS where two lasers are used to illuminate the same detection volume in order to excite two different fluorophores, the emission of which is split up by a dichroic mirror and detected by separate detectors. Cross-correlation of the fluorescence signal from two fluorophores has been demonstrated to be a powerful method to study bimolecular interactions. In addition to mere co-localization, FCCS can give information about the extent of concerted movement of two interaction partners, labeled with dyes emitting in different spectral ranges. FCCS is commonly used for monitoring of molecu-lar interactions and dynamic co-localization in solution and artificial membranes as well as in live cells[74]. A major concern in FCCS is the crosstalk related to the degree of overlap of the emission spectrum of one dye with the spectral range of the emission filter of a channel dedicated to detection of another dye. Together with the signal-to-noise levels, this crosstalk eventually sets the sensitivity limit of the technique.

A few years ago, alternating excitation from two or more different excitation sources was introduced for fluorescence-based single mol-ecule spectroscopy under the acronyms ALEX (Alternating Laser EXcitation)[75] and PIE (Pulsed Interleaved Excitation)[76]. Alternat-ing excitation can provide information about which excitation source has generated each detected fluorescence photon. Thereby, spectral crosstalk can be strongly suppressed and the sensitivity by which the interaction between two spectrally distinct interaction part-ners can be characterized by FCCS significantly improved. At the single-molecule level this approach provides information on labeling stoichiometry and allows separation of single- and double-labeled sample molecules.

23

Chapter 3

TRANSIENT STATE MONITORING

Fluorescent dyes exhibiting photoinduced long lived transient states can be used as excellent probes for microenvironmental changes. Such states can be generated by, e.g., trans–cis isomerization, inter-system crossing or photoinduced charge transfer. While the fluores-cence lifetime of a fluorophore normally is in the nanosecond range, these transient states have lifetimes on the order of microseconds or even milliseconds. Consequently, there is much more time for environmental interactions to affect the relaxation of a molecule in such a state than one in the first excited singlet state. In a first project we developed a technique for characterizing the kinetics of long-lived nonfluorescent or weakly fluorescent transient states by measurements with modulated excitation and time-averaged fluo-rescence intensity recording (paper I). This technique benefits from the single-molecule sensitivity of fluorescence without imposing any constraints on the concentration or fluorescence brightness of the molecules under study. Additionally, it requires no time resolution on the detection side, which makes it ideal for use with relatively slow detectors such as CCD-cameras and opens for massively parallel detection. In a second project, we adapted this technique for using

3 Transient state monitoring

24

the population buildup in transient states as a contrast mechanism in biomolecular and microenvironmental imaging with commercial LSMs (paper II).

3.1 Traditional techniquesThe major traditional methods for monitoring triplet-state kinetics in biomolecular studies are transient absorption spectroscopy, phos-phorescence spectroscopy and FCS.

Transient absorption spectroscopy is a well-established tech-nique[77, 78], where various states are monitored via their absorption of a separate probe beam, following an excitation pulse. However, for many compounds, the absorption spectra of these transient states overlap with those of other photoinduced states, making it difficult to separate them from each other. Moreover, this technique is relatively technically complicated and lacks the sensitivity for measurements at submicromolar concentrations[79].

Phosphorescence, i.e. the emission originating directly from the long-lived first excited triplet state can also be used for biomolecular studies[80]. By virtue of its long decay time, phosphorescence is well suited for monitoring slow rotational motions and to probe subtle changes in environmental conditions (viscosities, accessibilities of quenchers, polarities, etc.), revealing both structural and dynamic information of biological macromolecules[81]. However, coupled to the long-lived emission is also the susceptibility of the triplet state to dynamic quenching by oxygen and trace impurities, which can be circumvented only after elaborate and careful sample preparation. This quenching not only shortens the triplet lifetime, but practically makes the phosphorescence undetectable. Therefore, biomolecular monitoring by phosphorescence is largely restricted to deoxygenized, carefully prepared samples and has only been exploited to a rather minor extent.

The population and kinetics of a triplet state of a fluorescence marker can also be followed by fluorescence fluctuation analysis with FCS[26, 82]. Here fluorescence is detected, rather than the faint, easily quenched, phosphorescence signal from the triplet state itself.

25


Hence FCS provides a combination of high signal level and the out-standing environmental sensitivity given by the long lifetimes of the transient states. Furthermore, the experimental realization is rather simple. However, since the technique relies on the relative intensity fluctuations generated by each single molecule it is limited to very low concentration samples and fluorophores with high molecular brightness[83]. The time resolution needed on the detector side for triplet state characterization currently requires point detectors like APDs or photomultipliers, which makes parallelization complex. FCS with high time resolution has indeed been demonstrated with four laser foci and a two-by-two detector array[54], but making large arrays this way would be complex from an alignment point of view and expensive with currently available detection technology. Large arrays of single photon counting detectors, CMOS-APDs, are under development. Nonetheless, the presented approach circumvents the need for time resolved detection, making the transient state dynam-ics available for massively parallel readout or imaging applications by use of simple and relatively cheap detectors such as CCD-cameras.

Figure 3.1: Schematic representation of the modulation approach. The light from a laser is modulated by an acousto-optic modulator, controlled by a PCI-6602 card (National Instruments), and then used to excite the sample. The fluorescence is detected either by two APDs or a CCD camera connected to the PCI-6602 or a PCI-1410 card (National Instruments, not shown), respec-tively.

APD detectors/CCD

Ar+ laser

AOM

NI PCI-6602

SampleO

+

O

NH

NH

CH3

CH3

CH3

H3C

H3C

O


26

3.2 Utilizing the effect of modulation on the fluorescenceintensity

The experimental setup for the first project was essentially a traditional confocal FCS setup with the addition of an acousto-optic modulator (AOM) in the excitation beam path (see figure 3.1). The approach is general and does not depend on objective based fluorescence col-lection or diffraction limited excitation beam focusing. However, high enough excitation intensity to drive the fluorophores into the transient states of interest to a significant extent under continuous illumination is a prerequisite. An estimate of the spatial distribution of the excitation intensity and detection efficiency is also needed for quantitative results.

0

200

400

600

800

200

400

600

800

Fluo

resc

ence

Int

ensi

ty

Time

0

Tim

e Ave

rage

dFl

uore

scen

ce

Pulse Width

Pulse Width

w1 w1w2 w2

T TF1

F2

Fexc,2

Fexc,cw

Fexc,cw

Fexc,1

F1

F2

Fexc,cw

Fexc,1

Fexc,cw

Fexc,2

A B

C

1.0

1.1

1.2

1.3

1.4

1.5

1.6

Nor

mal

ized

Fluo

resc

ence

Rel

ativ

e tr

iple

tpo

pula

tion

0.0

0.1

0.2

0.3

0.4Figure 3.2: A and B show the relationship between pulse duration, w, and time-averaged fluorescence, F. The fluorescence normalized with respect to the duty cycle of the pulse train and the relative triplet population are shown in C.

By modulating the excitation source in the range of the relaxation time of the triplet state of a fluorescent dye, the triplet state can be populated to significantly different degrees depending on the repeti-tion rate and duration of the pulses. For pulse trains with short ex-citation pulse durations, the probability for the fluorescent molecule to enter into the triplet state within the pulses is smaller than for longer pulses with the same repetition rate. This in turn means that the fluorescence intensity is higher during the shorter pulses. This is illustrated in figure 3.2. Also, for pulse trains with a larger temporal

27


separation between the pulses, the probability for a fluorophore to relax into the ground state before the onset of the next excitation pulse increases, with a higher fluorescence intensity during the pulses as a result. The kinetic parameters of the triplet-state (kISC and kT) can be determined by measuring the average fluorescence intensity for a set of different modulation parameters and applying a model deriving the fluorescence intensity variations from the electronic state populations. The model used in this study is based on the one presented in the chapter about fluorescence in general (see figure 2.1B and equation 6).

Figure 3.3: A. Rh6G intersystem crossing rates for different KI concentrations. Values measured by FCS (black squares) compared to the presented approach (grey circles). B. Comparison of measurements of 10 nM (open symbols) and 10 µM (filled symbols) Rh6G with 2 mM KI. Lines are only connecting data points as guides for the eye. Inset shows intersystem crossing and triplet relax-ation rates from a concentration series ranging from 10 nM to 10 µM.

A B

0 1 2 3 4 5

0

5

10

15

20

25

k ISC /

µs-

1

KI concentration / mM0 1 2 3 4 5

1.00

1.05

1.10

1.15

1.20

1.25

1.30

1.35

1.40N

orm

aliz

ed flu

ores

cenc

e

Pulse width / µs

kT

10-8 10-7 10-6 10-50.00.20.40.60.8

468

10

Rat

e /

µs-1

Concentration / M

kISC @ 100 µW kT kISC @ 200 µW

3.3 Point measurementsTo demonstrate and evaluate the approach we measured triplet state kinetic rates of the very well characterized laser dye Rh6G in aqueous solutions with different potassium iodide (KI) concentrations. The intersystem crossing and triplet relaxation rates increase linearly with KI concentration due to the external heavy atom effect. The results were compared with traditional FCS-measurements[26]. Figure 3.3A shows the intersystem crossing rate, kISC, measured for KI concentra-tions from 0 to 5 mM. The linear dependence is evident and the differences between the FCS results and the values achieved with modulated excitation are within experimental errors.


28

In contrast to FCS experiments, which normally require dye con-centrations well below the micromolar range, the new modulation approach has no principal concentration limitation. To confirm this and to demonstrate that the method does not require any sample concentration calibration, measurements were performed at dye concentrations ranging between 10 nM and 100 µM. In the mea-surements of the samples in the upper concentration range, the gen-erated fluorescence intensity tended to saturate the sensitive APDs used. By inserting optical neutral density filters in the fluorescence beam path, this saturation was avoided. Since the background signal from ambient and scattered excitation light was reduced to the same extent as the fluorescence signal by the filters, the increase in signal-to-background ratio due to the higher sample concentration was not compromised.

In figure 3.3B, data from the 10 nM measurement and the 10 µM measurement are plotted in the same graph. The two data sets over-lap, indicating that the triplet-state kinetics was indeed found to be the same for the two concentrations, spanning a range of 3 orders of magnitude. In the inset, the triplet-state rate parameters, measured at two different excitation powers, are plotted versus dye concentration. The kinetic rates display almost no concentration dependence. Thus, the proposed method does not reveal any obvious intrinsic upper limit for the concentrations of fluorescent molecules. The limita-tions are instead set by the influence a high concentration may have on the fluorescence properties of the molecules themselves. In our measurements, a distinct drop in the triplet-state kinetic rates could be observed for a sample containing 100 µM Rh6G. We attribute this drop to quenching of the fluorophore molecules, effectuated by dye-dye interaction. This is well in agreement with dimerization of the dye molecules, which has been reported to take place for Rh6G at similar concentrations in water[84]. Further, insertion of attenua-tion filters (with optical densities up to 4) to avoid APD saturation demonstrates that the approach is compatible with fluorescence brightness attenuation of the Rh6G molecules by up to 4 orders of magnitude. Thus, unlike FCS and other fluctuation spectroscopy ap-proaches, it does not depend on a high fluorescence brightness of the

29


sample molecules.

FCS needs the high time resolution of the APDs, but only the time-averaged fluorescence intensity is used in our modulation approach. To demonstrate that detector time resolution is indeed not an issue, and as a first step toward parallelization, we replaced the APDs in our instrumentation with a CCD camera. Figure 3.4 shows the result of a measurement with a CCD camera, together with a corresponding measurement with APDs. The data sets obtained were nearly identi-cal.

Figure 3.4: Overlaid data sets from measurements on Rh6G with 5 mM KI using APDs (10 nM dye concentration, filled symbols) and a CCD camera (10 µM, open symbols). Lines show the model fit to the APD measurement data. Above the curves are three fluorescence intensity images from the CCD camera.

0 1 2 3 4 5

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

Pulse width / µs

Nor

mal

ized

flu

ores

cenc

e

3.4 ImagingIn the second project we worked with a Carl Zeiss LSM 510 Meta laser scanning microscope equipped with a ConfoCor 3 unit for FCS (Carl Zeiss MicroImaging GmbH, Jena, Germany). Here we exploit-ed the fact that, for a single fluorophore, the scanning of the excita-tion beam over the sample is equivalent to a temporal modulation of the excitation intensity (figure 3.5). Even at the fastest scanning rates, the time between subsequent lines was much longer than the triplet relaxation time and hence, in terms of triplet population buildup, the only effect of changing the scan speed was a change in effective illumination time for each pass of the beam. For imaging, interleaved


30

acquisition of several images at the fastest available scan speed and at a lower speed was used. The sets of images acquired at different scan speeds were separately averaged. To further reduce noise in the result-ing images each pixel was averaged with its four nearest neighbors. The images acquired with fast scanning served as a reference with only a small amount of triplet buildup due to short illumination times and therefore higher fluorescence intensities compared to the slowly scanned images. After averaging the image sets were corrected for distortions due to non-uniform scan speeds over the field of view. Finally, the averaged intensities of the quickly scanned image set were divided by those of the slowly scanned one, on a pixel-by-pixel basis, to achieve image contrast reflecting the triplet population of the fluo-rophores contributing to the signal in each pixel.

To demonstrate that the method can be used to generate images with a commercially available LSM reflecting local fluorophore triplet population, we performed measurements on a model system consist-ing of immobilized liposomes. Liposomes of approximately 5 µm diameter were formed by extrusion. One batch was made using a buffer containing the fluorophores rhodamine 110 (Rh110) and ATTO 655, and another batch was made with Rh110 and TEMPO-choline. TEMPO-choline promotes relaxation of the triplet state of

Exci

tation

inte

nsity

Time

Figure 3.5: Scanning of a laser beam over a sample exposes fluorophores in the sample to an excitation pulse, the duration of which is proportional to the width of the beam and inversely proportional to the scan speed. A high scan speed (dashed line) yields a shorter effective excitation time than a slow scan speed (solid line).

31


Rh110 and thereby reduces the triplet population buildup. ATTO 655 was used as an independent control for identification of lipo-somes without TEMPO-choline. We chose to work with Rh110 instead of Rh6G for the imaging because it is less hydrophobic, and hence exhibits a nice distribution throughout the liposome instead of getting stuck in the lipid bilayer. Figure 3.6 shows a set of images of a mixture of the two liposome batches immobilized in agarose

Figure 3.6: Liposome images demonstrating the ability of the presented method to reveal differences in local microenvironment. A. Image acquired with 12.8 µs pixel dwell time of a sample containing a mixture of liposomes with and without TEMPO, all containing Rhodamine 110. The false color scale is adapted to the intensities of the image. B. Image of the same sample as in panel A but acquired with 0.8 µs dwell time. The fluorescence intensities in the image are somewhat higher than in panel A. C. Image acquired with red ex-citation showing the emission of ATTO 655 present in the liposomes without TEMPO. This shows that out of the four brightest liposomes in the images the two leftmost contain TEMPO and the two rightmost do not. An inspection of the images in panels A and B shows that there is no obvious correlation between absolute fluorescence intensity and the presence of TEMPO. D. Ratio image formed by dividing the intensities in each pixel of the image in panel B with the corresponding intensities of the image in panel A. Here the concentra-tion dependence of the intensities vanishes and the values reflect the triplet population. Liposomes with and without TEMPO can easily be distinguished. E. Composite image where colors represent the values as in panel D and the brightness is a measure of reliability of the values based on the intensities of the underlying images.

0.9

1.0

1.1

1.2

1.3

1.4

1.5

0.9

1.0

1.1

1.2

1.3

1.4

1.5

A

D

CB

E

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

1400

0

100

200

300

400

500

600

700


32

Figure 3.7: Oxygen monitoring. Triplet contrast (intensity ratio) images in the left column and their corresponding composite images on the right. From top to bottom: Sample of liposomes containing rhodamine 110 surrounded by air and after 10, 30 and 60 minutes in argon atmosphere. The triplet population clearly increases with decreasing oxygen concentration.

1.0 1.2 1.4 1.6 1.8 2.0 1.0 1.2 1.4 1.6 1.8 2.0

33


gel demonstrating the ability of the method to make the distinction between the two based on triplet population.

An application of high potential importance for biological and medi-cal studies is imaging and monitoring of local oxygen concentration. Figure 3.7 shows an image series of a sample of liposomes with Rh110. The series starts with the sample oxygen concentration determined by the equilibrium with the surrounding air and proceeds as the sample is gradually deoxygenized by a flow of argon. This shows both the ability of the method to resolve differences in oxygen concentration and the possibility to perform a series of measurements on the same sample.

34

Chapter 4

FCS AND MODULATION

A common complication in FCS is that several processes take place in the same time regime, making them difficult to separate. Many of the best fluorescent markers, synthetic organic dyes as well as fluo-rescent proteins, exhibit fluorescence flickering due to, e.g., triplet formation[26], trans–cis isomerization[85], electron transfer[86] or proto-nation[73]. Whereas each of these processes may be exploited to probe the environment, sometimes they complicate the measurements or the data analysis by obscuring a process of interest falling within the same time range. We have previously shown that the amount of triplet formation may be controlled by modulating the excitation with pulse widths and periods in the range of the time constants of the involved transitions (see paper I). This should also be true for other photo-induced processes, i.e. all of the processes mentioned above except the concentration dependent protonation. Suppressing triplet state population buildup is also useful because it is a way of decreasing photobleaching[21, 87]. It has been shown that modulated excitation may allow immobilized fluorophores to emit 10-20 times more photons before they bleach[88]. However, modulating the ex-citation periodically in FCS measurements induces ringing in the

35


correlation curve, making a direct interpretation difficult. In most cases the modulation even destroys the correlation information for certain time regimes. In paper III we introduced and experimentally verified a method to retrieve the full correlation curves from FCS measurements with modulated excitation and arbitrarily low frac-tion of active excitation. We demonstrated, for Rh6G in water, that modulated excitation can be applied to FCS experiments to suppress the triplet build-up more efficiently than by reducing excitation power with CW excitation. We also showed data from measurements of fluorescein at different pH, where suppression of the triplet sig-nificantly facilitated the analysis of the protonation kinetics generat-ing fluorescence blinking in the same time range as the triplet state kinetics.

Another common problem, particularly when FCS is used in biologi-cal studies—in live cells or with samples containing large biomolecules or vesicles—is the appearance of spikes in the measurements. Spikes are sudden large transient increases of the measured fluorescence intensity that may appear due to formation of aggregates or natural presence of or contamination by strongly fluorescent or reflecting particles. All fluorescent entities contribute to the correlation func-tion used in FCS proportionally to the square of their fluorescence brightness[89]. A few short intense spikes, or even a single one, may thus render a long measurement useless. Sometimes the spikes can be eliminated by very careful sample preparation, but in many cases this is not possible. Therefore, the option to remove the effects of spikes after the measurement is very attractive.

If a time resolved fluorescence trace is acquired, sections affected by spikes may be removed before calculation of the FCS curve. The removal of sections of a fluorescence trace may be seen as an intensity modulation of the signal (figure 4.1). We have demonstrated that one of the filtering methods introduced in paper III for FCS mea-surements with modulated excitation is generally applicable and very powerful for reconstruction of FCS curves from fluorescence traces, from which parts containing spikes have been removed (paper IV).

4 FCS and modulation

36

4.1 ModulationfiltersWhen first approaching the problem of generating correlation curves from measurements with modulated excitation we divided the task into two parts: correlation between excitation pulses for long cor-relation times, and correlation within pulses for short times. The solution to the first task was rather simple. Using multi-tau correla-tion[90] with the shortest bin time corresponding to the modulation period gives a smooth correlation curve, but only provides informa-tion for correlation times longer than the period. After several tries to achieve correct normalization for the “intrapulse” correlations we realized that renormalization, according to the relative number of fluorescence photon pairs provided for each lag time interval by the particular modulation settings, should enable the use of any common correlation algorithm for the whole time span. Renormalization can be achieved by multiplication or division of the modulation distorted correlation curve by a suitable filter function.

Two approaches for generating filter functions for renormalization and retrieval of a correlation curve from a modulated measurement were introduced and evaluated. The first has the advantage of be-ing useful with conventional correlators, whereas the second is more

Figure 4.1: Spike removal and square wave intensity modulation. The top curve is an example of a fluorescence intensity trace containing a spike and the curve below it shows how the same trace could look without the spike. The intensity modulation defined by the third curve yields the same filtered trace (bottom curve) when applied to the traces with and without the spike.

0 20 40 60 80 100

t / ms

37


generally applicable. Here the basic assumptions and the final expres-sions are presented. For the full derivations, see paper III.

If the detected intensity I is assumed linearly dependent on the modulation of the excitation (a situation analogous to modulation of the light emitted from the sample), it can be expressed as a product of two uncorrelated components, the modulation M and the fluores-cence F:

I t( ) = M t( )F t( ). (22)

In this case, for all τ where M t( )M t + τ( ) ≠ 0 the autocorrelation of F can be calculated by dividing the autocorrelation of the detected intensity with the one of the modulation:

GF τ( ) =F t( )F t + τ( )

F 2 =

=I t( )I t + τ( )

I 2

M t( )M t + τ( )

M 2 =GI τ( )GM τ( )

. (23)

This method can easily be used with a hardware correlator, e.g. by measurement on reflected excitation light to get the correlation func-tion for the modulation.

In a more general case the intensity trace collected during a measure-ment with modulated excitation can be considered a product of three components:

I t( ) = M t( )V t( )F t( ). (24)

Here, M is the excitation modulation and F corresponds to the fluo-rescence of the observed molecules, as if they where continuously illuminated, but with a molecular brightness determined by the av-erage state population distribution while actually illuminated. V is a factor taking care of the variations in signal intensity during the pulses due to the modulated excitation. Hence, V is to a first approxi-mation periodic, with the same period as the modulation. The peri-odic nature of the modulation allows generation of a filter function to remove the modulation artifacts by introducing a lag time shift which is a multiple of the modulation period Tp and much larger


38

than the correlation time of any process other than the modulation contributing to the correlation. In the case of sample molecules freely diffusing through the detection volume this means that the shift has to be much larger than the diffusion time τD ( NTp >> τD). If this is fulfilled, the desired correlation function is obtained:

I t( )I t + τ( )

I 2

I t( )I t + NTp + τ( )I 2

=F t( )F t + τ( )

F 2 = GF τ( ). (25)

This operation can only be carried out for lag-times where the filter function has a non-zero value. It also deserves to be noted that an explicit normalization of the correlation curves is not necessary, since the filtering will by itself normalize the curve.

4.2 StatisticalpropertiesofmodulationfilteredFCS curves

Modulation of the excitation, and also the modulation filtering, will affect the statistical properties of the generated FCS curve. Knowing how is important for the design of an experiment, to determine the measurement duration needed to meet a certain precision require-ment and of course to determine whether modulated excitation is the proper choice for the study at hand. Here follows an estimate of the standard deviation of the correlation function, for the case where the filter function has been extracted from the measurement itself by time-shifted correlation according to equation 25. This was presented in paper A based on the weighting function introduced in paper III.

The correlation function G(τ) may be regarded as a function of the number of photon pairs Np(τ) separated by a time within an interval around τ, and the total number of photons in each detection chan-nel, N1 and N2, contributing to the correlation function:

G(τ) = A

N p(τ)N1N 2

. (26)

39


A is here a proportionality constant depending on the binning of the correlation function. Adding indices D and S in equation 26 to denote the direct and the shifted correlation functions respectively we can express the modulation filtered correlation function as a ratio of the two:

Gmod (τ) =

GD(τ)GS(τ)

=N p,D(τ)

N1,DN 2,D

⋅N1,SN 2,S

N p,S(τ). (27)

The proportionality constants cancel out, since both correlation functions must be calculated in the same way. Assuming that all Ni,j are generated by independent Poissonian processes[91] we can ap-proximate their standard deviations:

σ Ni , j⎡⎣⎢⎤⎦⎥= Ni , j ≈ Ni , j . (28)

Here the expectation value

Ni , j is approximated by the measured Ni,j and propagation of errors gives:

σ Gmod (τ)[ ]= Gmod (τ)σ

2 Ni , j⎡⎣⎢⎤⎦⎥

Ni , j2

i , j∑ ≈

≈Gmod (τ) 1Ni , ji , j∑ . (29)

This expression may be used as an empirical weighting function when fitting a model equation to the measured correlation data. For comparison of the statistical properties of FCS measurements it would be more useful to express this using macroscopic properties of the measurements in analogy with the work by Koppel[83]. The number of photon pairs available for a certain lag time τ can for a CW measurement be estimated by:

N p τ( ) = Nw τ( )

2 I 2 = N =T

w τ( )

⎧

⎨⎪⎪

⎩⎪⎪

⎫

⎬⎪⎪

⎭⎪⎪

=Tw τ( )I 2, (30)

where w(τ) is the bin width for the given lag time, N the number of such bins in the measurement, T the total measurement time and I the average count rate per channel. With similar count rates in both detection channels the total numbers of photons are given by:


40

N1 = N 2 =TI . (31)

Consequently, assuming that the measurement is completely back-ground free, an error estimate for an auto- or cross-correlation curve from a CW measurement would be:

σ GCW τ( )( )≈GCW τ( )1

Tw τ( )I 2 +2

TI=

= GCW τ( )1

TI1

w τ( )I+ 2. (32)

For a measurement with modulated excitation the characteristics of the excitation pulse train will affect the statistics of the correlation curve. For some lag times the statistics will be better than for others. Two useful error estimates may however be given without knowledge about the specific pulse train as long as the excitation duty cycle η, i.e. the fraction of time for which the excitation is active, is known:

Gmod ( )( ) Gmod ( )2

Tw( )Ip2 k+1 +

4TIp

=

=Gmod ( )2

TIp

1w( )Ip

k + 2. (33)

Here, Ip is the average fluorescence intensity within the pulses to facilitate comparison with CW measurements. Accordingly, the in-tensity averaged over the whole measurement is given by multiplica-tion by the modulation duty cycle η. The average errors for a certain duty cycle are given by equation 33 if k = 1. Error estimates for the lag times giving the best determined correlation with the excitation pulse train at hand, e.g. integer multiples of the modulation period, are given by k = 0.

This leads to the conclusion that on average an FCS curve from a measurement with modulated excitation filtered by its time-shifted correlation function will exhibit a standard deviation that is a factor 2 larger than that of a measurement of the same duration and the same average fluorescence intensity but with CW excitation. Conse-quently, a measurement with modulated excitation must be twice as

41


long as the corresponding CW measurement, to yield an FCS curve with the same average uncertainty. However, for certain lag times, the random errors may be smaller for the modulated measurement than for the CW measurement.

For the case where the filtering function contains no uncertainty, as in the case where modulation filtering is used to correct the FCS curve of a measurement where spikes have been removed, the fac-tor 2 does not apply. Here, the change in measurement quality compared to an unfiltered measurement is determined by the duty cycle η or, equivalently, the fraction of the measurement left after spike removal.

4.3 Pulse trainsAn important issue when making FCS measurements with rectangular shaped modulated excitation is to get useful correlation information for all time regimes. The excitation must be such that fluorescence photons can occur in pairs separated by any time shorter than the measurement duration. For a regular pulse train, with one pulse per modulation period, the pulse width must be at least half a period to fulfill this requirement. Such a pulse train, with a pulse width that is exactly half the period, has a duty cycle of 50%. The duty cycle is normally defined as the ratio of the pulse width and the period, whereas here, for the more complex pulse trains, we have chosen to

Figure 4.2: A. Design of lower duty cycle pulse trains, allowing extraction of complete correlation curves, by starting with a 50% duty cycle square wave and recursively removing the middle third of each pulse. Ribbon a represents CW excitation, b is a 50% duty cycle square wave, c and d show the two first recursion steps. B. Illustration of the more general approach, removing the n even 1/(2n+1)-parts. Here n = 2.

A

a

b

c

d B

a

b

c

d


42

define it analogously as the fraction of the total time the excitation is actually active. To achieve duty cycles smaller than 50% we designed pulse trains with multiple pulses per modulation period. Based on a rectangular pulse train with 50% duty cycle, a pulse train with an arbitrarily small duty cycle can be generated by recursively removing the middle third of each pulse. Accordingly, the number of pulses per modulation period is doubled for each recursion. This pulse train would still contain information on all time scales, limited only by the resolution of the detection system. A more general realization of the same idea would be to recursively remove the n even 1/(2n+1)-parts (n positive integer) of each pulse, decreasing the duty cycle by a factor of (n+1)/(2n+1) for each recursion (figure 4.2). For all measurements with modulated excitation presented in this chapter, pulse trains were generated according to this scheme with n = 1.

4.4 Measurements with modulated excitationMeasurements were made on Rh6G in water, with collection of time-tagged photon traces, as well as with a hardware multi-tau correlator, to demonstrate the described approaches for generating correlation curves from measurements with modulated excitation.

Figure 4.3A shows two FCS curves from measurements with modu-lated excitation and different pulse widths. The data was acquired using the ALV-5000 hardware correlator (ALV-GmbH, Germany) and the correlation curves were filtered by division by the correla-tion curves from corresponding measurements on reflected light. The method works well for excitation with short pulses, because the fluo-rescence intensity does not change significantly during each pulse. For longer excitation pulses the decrease in fluorescence intensity due to triplet population build-up during a pulse gives rise to peaks in the correlation curve. These peaks appear around lag times correspond-ing to odd multiples of the pulse width. In figure 4.3B FCS curves from similar measurements are shown. In this case data was collected as time-tagged photon traces and software-correlated. Filter func-tions were generated by time-shifted correlation of the same mea-surements, as described above (equation 25). Both of these curves are well-behaved, and the fluorescence decay during the pulses is no

43


longer a problem. An additional major advantage of this method is that it only requires one measurement to generate a filtered correla-tion curve. This makes it more robust and less sensitive to changes in the experimental setup. This method was used for all the following measurements presented here.

A series of measurements were made on Rh6G with different modula-tion parameters and compared to a power series with CW excitation. A relatively high peak excitation intensity was chosen to clearly illus-trate the effect of the modulation on the average triplet population. Figure 4.4A shows a CW measurement at the chosen peak excitation power compared with two modulated measurements. The reduction in triplet fraction is obvious. Figure 4.4B shows that, compared to decreasing excitation power, modulated excitation can be used as a more efficient way to suppress triplet build-up. The triplet fractions were determined by nonlinear least-squares fits of the model shown in the introduction (equation 19). It can be seen that, as expected, the choice of modulation parameters is critical for optimal triplet

Figure 4.3: A. Measurements on Rh6G, acquired by an ALV-5000 correlator. Curves generated by dividing correlation curves for the Rh6G measurements with modulated excitation by curves from measurements on reflected excita-tion light with corresponding modulation. Pulse widths in the excitation pulse trains were 0.9 (solid curve) and 8.1 µs (dashed curve). The modulation period was 16.2 µs for both measurements. B. Measurements corresponding to those in panel A but software correlated after collection of time-tagged photon traces. The filter functions used to calculate these curves were generated by introduc-ing a time shift of 0.3 s between the data channels before correlating. Insets in panels A and B show the unfiltered correlation functions (solid curves) and the corresponding filter functions (dashed curves). The upper insets correspond to the measurements with the longer pulses.

10-6 10-4 0.01

0

1

2

3

G

τ / s

0

2

4

6

G

10-7 10-6 10-5 10-4 10-3 0.01 0.1

1.0

1.2

1.4

1.6

G

τ / s

A

10-6 10-4 0.01

0

1

2

3

G

τ / s

0

2

4

6

G

10-7 10-6 10-5 10-4 10-3 0.01 0.1

1.0

1.2

1.4

1.6

G

τ / s

B


44

suppression. The pulses must be short enough, that fluorophores do not have the time to accumulate in the triplet state, and the times between pulses must be long enough to allow sufficient relaxation back to the singlet ground state (see paper I).

To show that modulated FCS can be used to separate processes overlapping in time, measurements were made on fluorescein at dif-ferent pH. At the relevant pH values, fluorescein has two forms: one nonprotonated dianionic form, which is fluorescent; and one pro-tonated monoanionic form, which is very weakly fluorescent under the conditions of this study. Here we regard the latter form nonfluo-rescent, the only implication of which is a slight underestimation of the fraction of protonated dye in the sample. With this assumption, the protonation kinetics affects the correlation curve in exactly the same way as the triplet transitions. We used the model introduced in equation 20 for curve fitting.

Figure 4.4: A. Three measurements on the same Rh6G sample and their cor-responding fits. All correlation data are shown as solid thin curves. The solid thick curve is a fit to the CW measurement, the thick dashed curve is a fit to the modulated measurement with 5.4-µs period and 900-ns pulse width, and the dotted curve is a fit to the modulated measurement with the same period and 100-ns pulse width. The average triplet population clearly decreases with shorter pulses. B. Average fraction of molecules in the triplet state as a function of average excitation power for a sample of Rh6G. The solid curve with filled squares shows data from a CW power series. The dashed curves with open symbols show measurement data from measurements with modulated excita-tion and constant peak power of 0.84 mW. The curve with circles represents a modulation series with 16.2-µs period and varying pulse widths, and hence duty cycles. Upward-pointing triangles signify 5.4-µs, downward-pointing triangles 1.8-µs and diamonds 600-ns periods.

A B

10-7 10-6 10-5 10-4 10-3 0.01 0.11.0

1.5

2.0

2.5

3.0

3.5G

τ / s

0 200 400 600 800

Average excitation power / µW

0

5

10

15

20

25

30

35

Trip

let

frac

tion

(%

)

45


Lower pH means a higher concentration of protons available for reaction, which increases the fraction of protonated molecules and the rate of protonation. This leads to a shortening of the measured protonation relaxation time. Adding buffer speeds up the exchange and also shortens the protonation time, but leaves the protonation fraction unaffected. As a reference, measurements with both CW and modulated excitation were made on fluorescein at pH 9.4, where essentially all molecules are in the unprotonated state. The triplet parameters were extracted by fitting equation 19 to the data (figure 4.5A).

Figure 4.5: A. Measurements on flu-orescein in 10 mM phosphate buffer at pH 9.4 with 84-µW excitation power (average power for CW, peak power for the modulated measure-ment). No protonation assumed due to the high pH. A global fit with the triplet relaxation time as common fitting parameter is shown as a solid line for the CW measurement and as a dashed line for the modulated measurement with 5.4-µs period and 100-ns pulse width. The fit gives a triplet relaxation time of 0.94 µs and triplet fractions of 48% and 18%, respectively. B. Measurements on fluorescein in 1 mM phosphate buffer at pH 6.0 performed in the same manner as presented in panel A. A global fit, with protonation time and fraction as common fitting parameters and triplet parameters fixed to the results obtained in panel A, indicates 66% protonated dye with protonation time 5.0 µs. C. The same modulated measurement as in panel B but fit with a model assuming no triplet. This gives a protonation fraction of 64% and a protonation time of 4.2 µs.

B

C

A

10-7 10-6 10-5 10-4 10-3 0.01

1.0

1.1

1.2

1.3

1.4

τ / s

G

10-7 10-6 10-5 10-4 10-3 0.01

1.0

1.1

1.2

1.3

1.4

τ / s

G

10-7 10-6 10-5 10-4 10-3 0.01

1.0

1.1

1.2

1.3

1.4

τ / s

G


46

Figure 4.5B shows corresponding measurements on a fluorescein sample with 1 mM phosphate buffer at pH 6.0. This pH gives promi-nent protonation, and the buffer concentration was chosen such that the protonation time is very close to the triplet time, making the two processes hard to separate. A global fit of equation 20, with triplet parameters fixed to the values achieved for the measurements at pH 9.4 and protonation parameters as common fitting parameters, gave a fraction of protonated species of 66% and a protonation time of 5.0 µs.

In figure 4.5C the same measurement with modulated excitation as in figure 4.5B is depicted again, this time with a fit assuming zero triplet population. This yields protonation fraction and time of 64% and 4.2 µs, which compares reasonably well with the values from above. This indicates that modulated excitation can suppress the triplet population to the extent that, for some applications, triplet kinetics need not be taken into account in the analysis anymore.

The introduction of excitation modulation in FCS provides a new set of parameters for optimizing the photophysical conditions for a mea-surement. Properly used it will improve the signal to noise ratio by increasing the fluorescence intensity and reducing photobleaching. It adds the option of applying time gating to improve the signal to background ratio, which may be especially useful for measurements with low excitation duty cycle under conditions with stray light or high dark current. Here, the time resolution is not high enough to allow gating to remove scattered excitation light[9], but with a combi-nation of modulation and a laser generating short pulses this would also be possible. Moreover, modulated FCS offers more robust data analysis and interpretation thanks to the added possibilities to sepa-rate processes from each other.

4.5 Modulationfilteringofdatawherespikeshave been removed

A model system enabling reproducible generation of fluorescence intensity spikes with different brightness and average frequency was devised. This comprised free dye or single-labeled liposomes in aque-

47


ous solution representing clean samples and multi-labelled liposomes with different degree of labeling to be added in different amounts to generate spikes (see paper IV for details).

For spike identification we used a simplified version of an algorithm commonly used for burst selection in single-molecule measure-ments[9, 92]. A sliding average was calculated over a trace of times be-tween consecutive photons, for each data point averaging 50 entries. This average was compared to a threshold value, which henceforth will be expressed as a percentage of the average time between photons in the measurement. If below the threshold value, all photons con-tributing to the average were discarded as being part of a spike. The beginning and end of each spike was marked in the trace to enable fast semi-analytical calculation of the modulation filter function.

Figure 4.6 shows an example of a measurement on free fluorophores only (solid curve) in comparison with the same sample with addition of liposomes with on average ten fluorophores each, generating on average 18 spikes per second (dashed curve). The dotted curve shows the result of spike removal with a 25% threshold and subsequent modulation filtering. The addition of liposomes had a strong effect on the correlation curve, and evidently the filtering was capable of removing this effect. The model of equation 19 fits both the curve

Figure 4.6: Filtering of a spiky measurement. FCS curves from measurements on free dye (solid curve) and the same sample with addition of multi-labeled li-posomes generating intensity spikes (dashed). After spike removal and modula-tion filtering of the data from the second measurement a curve (dotted) similar to the one of free dye only was obtained.

10-6 10-5 10-4 10-3 0.01 0.1 1

1.0

1.2

1.4

G

τ / s


48

from the pure sample and the filtered curve from the spiky sample well and generates nearly identical fit parameters for the two.

We wanted to compare the modulation filtering approach to other methods for generation of FCS curves from data where spikes have been removed, and preferably do it independently of the spike iden-tification. Therefore, traces from measurements without spikes were used, and sections of these were removed randomly to simulate per-fect spike removal. The durations of the removed sections, as well as the separations between them, were generated randomly from expo-nential distributions with given mean values. Figure 4.7 shows FCS curves from three such traces, generated by two different methods. Here, the average duration of the removed sections was 300 µs and the average separation was varied from 1 to 100 ms. For comparison, the FCS curve from the measurement of the spike free sample, in this case single-labeled liposomes, is included in each panel. Panel A shows that the original curve is perfectly regenerated by the modula-tion filtering. In panel B the curves were generated by an established and commonly used algorithm, where a correlation function is cal-culated separately for each part of the fluorescence trace between two consecutive spikes and all of these correlation functions are subse-quently averaged[93]. Using this approach limits the available lag time

Figure 4.7: Comparison of different methods for generation of correlation curves after simulated ideal spike removal. A. Modulation filtering. B. Sepa-rate correlation of the individual remaining parts of the intensity trace and subsequent averaging. Photons were removed, for on average 300 µs long time intervals separated by on average 100 ms (dashed curve), 10 ms (dotted) and 1 ms (dash-dotted), from a spike free measurement on single-labeled liposomes (solid curve).

A B

10-7 10-6 10-5 10-4 10-3 0.01 0.1 1

G

τ / s

1.0

1.1

1.2

G

τ / s

1.0

1.1

1.2

10-7 10-6 10-5 10-4 10-3 0.01 0.1 1

49


range to the time between spikes and if spikes appear frequently each partial trace becomes very short and a significant decrease in both correlation amplitude and decay time is observed. This is consistent with a previously observed bias of the correlation function for short traces[94]. The cases simulated here are rather extreme and this estab-lished algorithm works well in most cases. The spike frequency of the measurement presented in figure 4.6 is however almost twice as high as the lowest frequency in these examples. This indicates that the effect of the bias may be substantial in some practical situations.

The modulation filtering is a generally applicable approach for gen-eration of FCS curves from data where spikes have been removed. It makes efficient use of all available data and keeps the time axis intact. Furthermore, it preserves correlations between data before and after removed spikes, which is particularly important when spikes are ex-tremely short or when they occur very frequently. The calculation of the correlation function for the square wave modulation introduced by the spike removal can be made very quickly in a semi-analytical fashion. Consequently, the processing time added by the modulation filtering after spike removal is minimal.

50

Chapter 5

RATIOMETRIC CROSS-CORRELATION

In this chapter another method for separation of processes occurring in the same time range is presented. We introduced spectral cross-correlation as a tool to selectively highlight a process giving rise to a spectral shift in both the absorption and the emission spectrum of a dye. We used a pH-sensitive so called ratiometric fluorophore and showed that, by the application of spectral cross-correlation, the protonation–deprotonation process manifests itself as an anti-correlation. This makes the distinction from, e.g., diffusion or triplet kinetics substantially more robust.

To improve the selectivity of detection of the two different protona-tion species an alternating excitation scheme similar to the ALEX[75]

and PIE[76] approaches was applied in combination with temporal gating of the detected signal. Application of the previously pre-sented modulation filtering by time-shifted correlation (paper III) eliminates the requirement that the excitation alternation must be faster than the processes of interest. We showed that the results were equal regardless of the excitation alternation period, except for the

51


statistical properties and hence the measurement times required for a certain precision.

5.1 ThefluorophoreA ratiometric dye is a molecule for which the absorption or emis-sion spectrum shifts in response to the environment. In contrast to a reporter molecule that responds to an environmental change by a change merely in fluorescence quantum yield, this enables internal calibration by recording the ratio of absorbance or emission intensity for two different wavelengths[95]. Thereby, differences in dye concen-trations, optical path lengths or other parameters that affect the fluo-rescence output vanish. This is especially important in measurements in live cells, where the variations in local dye concentration may be substantial.

The fluorophore used in this project was a pH-sensitive rhodol de-rivative (a test compound supplied by ATTO-TEC GmbH, Siegen, Germany, henceforth referred to as RD). To determine the spectral properties of the protonated and deprotonated species of this dye absorption and emission spectra were acquired at pH 4 and 10, re-spectively. These are shown in panels A and B of figure 5.1. At low pH most of the fluorophore molecules are in their protonated state, excitable in the blue-green region of the electromagnetic spectrum. At higher pH the molecules are deprotonated and the absorption and emission spectra are red-shifted. The typical shape of the absorption and emission spectra makes properly applied excitation very selective for the red-shifted species, whereas the emission separation can be made with high selectivity for the blue-shifted species. This inher-ent complementarity makes the combination, along with temporal gating of the signal to suppress spectral cross-talk, highly effective as a means to improve the contrast in the cross-correlation. The 488-nm argon ion laser line was used for predominant excitation of the protonated species and the 594-nm light from the helium-neon laser for the deprotonated species (vertical lines in panel A). The emission was separated using optical band-pass filters, the transmission spectra of which are shown as shaded areas in panel B. Obviously, neither the excitation wavelength nor the emission filters used here were optimal

5 Ratiometric cross-correlation

52

for the protonated species. This reduced the achievable signal to noise ratio, but did not affect the general outcome or the conclusions of this proof of principle study. Panel C of figure 5.1 shows the results of a pH titration in a spectrofluorometer, where the two species were separately excited and detected. Apparently, the signals from the two species were complementary and the fluorophore exhibited only one pKa in the pH range covered by the titration. This pKa was deter-mined to approximately 7.6.

Figure 5.1: Bulk dye characteriza-tion. A. Absorption spectra of the protonated (solid curve) and deprotonated (dashed curve) spe-cies of RD. Vertical lines mark the wavelengths corresponding to the laser lines used for excitation in this project. B. Emission spectra of the protonated (solid curve) and depro-tonated (dashed curve) species of RD. The shaded areas represent the transmission of the emission filters. C. Spectrofluorometer bulk pH titration of RD in 150 mM NaCl. Open squares represent measurements with excitation and detection at 594 and 653 nm, respectively, whereas solid circles show measurements made with 488-nm exci-tation and 587-nm detection. All data have been normalized with respect to the lowest and highest values of each data series.

A B

C

450 500 550 600 6500.0

0.2

0.4

0.6

0.8

1.0

Nor

mal

ized

abs

orpt

ion

λ / nm

Nor

mal

ized

inte

nsity/

tran

smis

sion

λ / nm

5 6 7 8 9 10 110.0

0.2

0.4

0.6

0.8

1.0

Nor

mal

ized

inte

nsity

pH

500 550 600 650 700 750 8000.0

0.2

0.4

0.6

0.8

1.0

5.2 FCCS with alternating excitationSeparately detecting the fluorescence from the protonated and the deprotonated species of a pH-sensitive fluorophore and cross-corre-lating the acquired intensity traces makes the protonation kinetics appear as an anti-correlation in the fluorescence correlation curve. However, as opposed to in an ordinary dual-color FCCS measure-

53


ment where the two labels can be selected for good spectral separa-tion, the contrast is here, under continuous illumination, limited by the inherent spectral separation of emission from the two species of the single probe. To minimize spectral cross-talk modulation of the laser beams was introduced for individual and temporally separated

Figure 5.2: A sketch of the setup used for FCCS with alternating excitation. Two lasers serve as excitation sources and are modulated by acousto-optical modulators. The setup has two entirely separate detection pathways and in total four detection channels. The upper inset shows how the blue laser preferably excites the protonated species of the fluorophore, which is mainly detected in the green detection channels, and the yellow laser almost exclusively excites the deprotonated species, which is detected in the red detection channels. The lower inset shows examples of how the switching between the protonated and depro-tonated states contributes to the correlation in the case of autocorrelation (red and green curves) and gives rise to an anticorrelation when cross-correlating the signals from the two spectral detection ranges (black curve). The anticorrelation arises because a molecule cannot reside in both states at once—after emitting a photon in one state it must switch to the other state before emitting a photon contributing to the cross-correlation.

Ar+ LASER

Objective

Beam expander

Dichroic mirrors

Tube lens

Confocal pinhole

Polarizingbeam splitter

Emission filters

Excitation filter

AOM

AOM

Sample container

APD

APD

HeNe LASER

APD

APD

Excitation

Fluorophoreprotonation

Detectedemission

H+

H+ H+

H+

Crosstalk

k-k+[H+]

10-8 10-6 10-4 10-2 11.00

1.05

1.10

τ / s

G(τ)


54

excitation of the two species. The alternating excitation was combined with time-gating of the intensity traces collected from the spectral ranges of the protonated and deprotonated species, respectively. A schematic drawing of the experimental setup for this approach is presented in figure 5.2, along with a sketch of the underlying process and a comparison of the auto- and cross-correlation curves.

The effect of dual color cross-correlation and alternating excitation with gating is illustrated in figure 5.3. Auto-correlation functions for the two different excitation wavelengths and spectral detection ranges from measurements on RD in 500 µM HEPES buffer are shown in panel A of figure 5.3. A cross-correlation function from a measurement with alternating excitation and gating applied to remove crosstalk is shown in panel B. A comparison with panel C, where a cross-correlation function from the same measurement but without gating is shown, clearly demonstrates that the contrast in the protonation part of the cross-correlation is strongly enhanced by the gating procedure that essentially removes the spectral cross-talk.

No evident contribution from singlet–triplet transitions could be observed in our cross-correlation measurements. A simple model without triplet component proved sufficient for good fits to the cross-correlation data and provided results consistent with those extracted from conventional FCS measurements:

G τ( ) =

1N

1+τ

τD

⎛

⎝⎜⎜⎜⎜

⎞

⎠⎟⎟⎟⎟

−1

1+τ

β2τD

⎛

⎝⎜⎜⎜⎜

⎞

⎠⎟⎟⎟⎟

−1 2

1−Ce−τ τP( )+1. (34)

Here, C is a contrast parameter in the range 0 to 1 depending on the achieved discrimination in detection of the two protonation species. Crosstalk reduces the contrast and is reflected by a decrease in C.

A global fit of the auto-correlations and the gated cross-correlation with the protonation relaxation time as common parameter enables direct extraction of the relaxation times and amplitudes pertaining to the triplet and protonation kinetics. Such a fit to the data is shown in figures 5.3A and B. Fits of the model in equation 34, with all param-eters free, to the gated measurement in panel B and the ungated one in panel C, both yielded values of τP within experimental errors of

55


Figure 5.3: The effects of cross-correlation and gating. The panels in this figure all show measurements on RD in 500 µM HEPES buffer at a pH of 7.7. A. Autocorrelation curves from measurements with CW excitation and detection of either the deprotonated (open squares, 594-nm excitation) or the protonated species (solid circles, 488-nm excitation). The solid lines represent a global fit of the models in equations 21 and 34 to these two data sets and the one in panel B, respectively, with τP as common parameter. B. Cross-correlation of the signals from the two spectral ranges in a measurement with alternating excita-tion. Gating was used to remove crosstalk. The solid line represents a model fit as described above. C. Cross-correlation without gating.

1.0

1.5

2.0

2.5

3.0

3.5

10-6 10-5 10-4 10-3 0.01 0.1

10-6 10-5 10-4 10-3 0.01 0.1

10-6 10-5 10-4 10-3 0.01 0.1

10-6 10-5 10-4 10-3 0.01 0.1

10-6 10-5 10-4 10-3 0.01 0.1

10-6 10-5 10-4 10-3 0.01 0.1

τ / s

G

1.0

1.5

2.0

-0.1

0.0

0.1

Res

idua

ls

Res

idua

lsRes

idua

ls

-0.1

0.0

0.1

1.0

1.5

2.0

-0.2-0.10.00.10.2

A B

C

GG

τ / s

τ / s

that extracted from the global fit. In this case, however, the contrast improvement from the gating made the curve fitting more robust and hence the reported parameter values substantially more precise.

For retrieval of the correlation curves from the data collected with modulated excitation, the strategy of division by a time-shifted cor-relation function to remove the periodic variations described in paper III was applied. This makes it possible to use modulation periods longer than the characteristic time of the studied processes, and thereby to design pulse trains adapted to different demands, e.g., to improve the photophysical measurement conditions by suppressing triplet population buildup. Figure 5.4 shows cross-correlation curves from three measurements acquired with different modulation peri-ods ranging from 250 ns to 20 µs. The result was the same regardless


56

of whether the period was shorter or longer than the protonation relaxation time. However, for measurements of the same duration, a shorter modulation period generates more photon pairs available for correlation in the short lag time range yielding a higher signal-to-noise ratio for studies of fast processes.

To ascertain that the anti-correlating part of the correlation curve was due to protonation of the fluorophore, measurements were conducted on RD in HEPES buffer with a fixed pH of approximately 7.7 but with different buffer concentrations (250 µM – 4 mM). Figure 5.5A shows the cross-correlation curves measured with alternating excita-tion and a global fit to these curves based on the model in equation 34 with the contrast parameter C as a common fitting parameter. This yielded a value of C of 0.91, which meant that there was essentially no crosstalk left after gating. Each curve shows a prominent anticor-relating component attributable to the protonation. The protonation of the fluorophore got faster with increasing buffer concentration and the relaxation time of the anti-correlating component changed accordingly. The protonation relaxation rate in the presence of buffer is given by[73]:

kP = k + k+ H+ + B[ ]k+B H+ K aB + k B

H+ K aB +1, (35)

and is hence linearly dependent on the buffer concentration [B]. Here,

Figure 5.4: Cross-correlation curves from measurements on RD in 1 mM HEPES buffer with different excitation alternation periods: 250 ns (open squares), 2 µs (solid circles) and 20 µs (open triangles).

10-6 10-5 10-4 10-3 0.01 0.10.99

1.00

1.01

1.02

G

τ / s

57


k+B and k-B are the rates of proton transfer to the fluorophore from the buffer and v.v., and KaB is the equilibrium constant of the buffer. Already a visual inspection of the curves shows that the protonation relaxation time was shortened by about half for each doubling of the buffer concentration. In figure 5.5B the protonation relaxation rates extracted from the cross-correlations are plotted against the buffer concentration. Values from CW autocorrelation measurements on the same samples, fit globally with the triplet parameters as common fitting parameters, are included for comparison. As can be seen in the graph, the results from the different measurements are very similar; indicating that the protonation time was well determined by a fit of the cross-correlation only. The solid line represents a linear fit to the values retrieved from the cross-correlations.

This study suffered from two main limitations regarding signal to noise ratio, none of which is inherent to the concept itself. The first was the lack of suitable filters, which could be remedied by use of a custom-made filter set. The second was the shallow data buffer of the data acquisition card, which severely limited the useful signal level. An optimized filter set in combination with a data acquisition system

Figure 5.5: Buffer concentration series. A. Cross-correlation curves from a series of measurements on RD in HEPES buffer with concentrations rang-ing from 250 µM to 4 mM. The solid lines show a global fit of the model in equation 34 to all five data sets with the contrast parameter C as common fitting parameter. B. The total protonation relaxation rate, kP, as a function of buffer concentration. The open squares represent the values extracted from the fit shown in panel A. The solid line is a linear fit to these data. Solid circles show values from FCS measurements with CW excitation on the same samples extracted by a global fit of the model in equation 21 with the triplet parameters as common fitting parameters.

A B

10-6 10-5 10-4 10-3 0.01 0.1

τ / s

G

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.50

2

4

6

8

k P /

105

× s

-1

[Buffer] / mM

1.0

1.5

2.0

250 µM500 µM1 mM2 mM4 mM


58

allowing higher count rates should reduce the measurement times for the cross-correlation measurements from hours to minutes, which would make the approach more useful for routine measurements.

Proton exchange across and along biological membranes play a central role in the energy metabolism of virtually all living organisms. Still, in spite of extensive studies, many mechanisms in this context are far from understood. FCS provides a unique and powerful tool for mon-itoring of local ion concentrations and buffering properties without the need to introduce perturbations to the observed system[73, 96, 97]. The low dye concentrations used also minimize the buffering effects of the fluorophores themselves. Turning the protonation component of the correlation curve into an anti-correlation makes it substantially easier to separate from the diffusion component or photophysical processes in the same time range. This in turn enables measurements over a larger range of protonation rates than with ordinary FCS un-der similar conditions. The prominent feature of the cross-correlation curve caused by the anti-correlating protonation process makes the curve fitting more robust.

59

Chapter 6

DATA ACQUISITION AND MODULATION CONTROL

Three of the projects introduced in the previous chapters relied on data acquisition synchronized with the modulation of the excitation laser light. This was achieved using two PCI-6602 counter/timer cards from National Instruments Corporation, USA (NI), equipped with BNC-2121 connector blocks and interconnected over the built-in RTSI (real-time system integration) bus. A PCI-6602 card provides eight 32-bit counters, up to 32 digital I/O lines (shared with the counters) and 3 DMA (direct memory access) channels for fast data transfer with a 16-sample FIFO (first in first out) buffer each. The first acousto-optic modulator (AOM) (AA.MT.200/A0,5-VIS, AA Opto-Electronic, France) and the avalanche photodiodes (APDs) (SPCM-AQR-14/16, PerkinElmer Optoelectronics, USA) for fluo-rescence detection were all connected to the connector block of the same counter/timer card. The internal 80-MHz clock of the same card was used both for data collection timing and to determine pulse widths and periods to eliminate problems with timing drift in the system. The second AOM, which was used when alternating excita-

6 Data acquisition and modulation control

60

tion was needed, was connected to the connector block of the second card. The software controlling the cards was written in LabView 8.20 (NI).

In the first project (paper I), the main strategy was to measure average fluorescence intensity during a few seconds while illuminating the sample with excitation light modulated by a square wave pulse train with a certain pulse width w and period T. This was repeated for a set of different pulse train parameters. The implementation was rather straightforward. One counter per detection channel was set up to count pulses from an APD for a certain amount of time and one counter was set up to generate the pulse train. A program was written to automatize a complete measurement series, correct for detector nonlinearity, subtract background and normalize the data.

Some reference measurements with time tagged photon traces were also needed. Here the goal was to time stamp each detected photon with the 12.5-ns resolution provided by the 80-MHz clock of the card. This was realized by letting one counter per detection channel continuously count clock pulses, and upon photon detection send its current value to the FIFO buffer. The buffers were then read by the computer as often as possible. It turned out, however, that the shal-low buffers of the card presented a problem. Even though the average fluorescence count rates were well below the cards highest sustainable data transfer rates, a fluorescence burst might easily produce 16 pho-tons fast enough to induce buffer overflow. This situation was further aggravated by detector afterpulsing. To circumvent this problem a second counter was introduced as a filter for each channel, config-ured to generate a pulse of certain length when trigged by a pulse from the APD (figure 6.1A). Once back to idle, this counter could be retriggered, but for a well-defined time after detection of a photon it would be insensitive to consecutive pulses. The pulse generated by this counter was in turn used as input for the counter providing the time stamps. Thus detection of photons arriving too close together was suppressed and the stability of the data collection greatly im-proved. When implementing such a filter, care has to be taken not do introduce any unwanted artifacts in the data. If too many detection

61


events are suppressed this will for instance lead to pulse distortions in fluorescence histograms or decreased correlation amplitudes if the data is used for FCS. These artifacts are well known, since most detec-tion systems used in the field exhibit a so called dead-time[98, 99]. The same strategy was used to acquire the time-stamped photon traces needed for the other projects.

For the FCS measurements with modulated excitation presented in paper III the modulation pulse trains needed were more com-plex. Generation of the pulse trains was achieved by letting several counters work together. To define the period of the pulse train, one counter was configured to generate a simple square wave pulse train with 50% duty cycle (50% on, 50% off). This pulse train, in turn,

Figure 6.1: A. Suppression of photon detection pulses arriving within a too short time span after the last registered one was achieved by letting the signal from the APD trigger a counter set up to generate a pulse of certain length. This generated pulse was then used to gate the readout of the counter providing the time stamps. The detection dead time is the sum of the defined high and low times of the pulse generating counter. After finishing the generation of one pulse, this counter is again idle and retriggerable. B. Strategy for generating complex excitation pulse trains. To achieve pulse trains with sufficient time resolution and stability they must be predefined on the hardware level. One counter was set up to generate a simple pulse train defining the period. A number of counter pairs were set up to generate a sequence of a finite number (in this case two) of pulses when triggered. The first pair used the signal from the period defining counter as trigger and in turn triggered the next counter pair. The pulse widths and the times between pulses for each counter pair were chosen such that the time from the onset of the first pulse of a sequence until the end of the last pulse corresponded to the pulse width of the triggering counter pair.

A

B

APD

Filtercounter

dead time

Periodcounter

1st counterpair

2nd counterpair

Tp

wp

6 Data acquisition and modulation control

62

was used to trigger another counter to generate a number of pulses. In this configuration an additional counter is occupied by keeping track of the number of pulses generated. Now, this pulse train could either be sent to the modulator or used to trigger another counter pair. This was done with the period-defining counter together with up to five other counter pairs, i.e., in total up to eleven counters were used, just for the pulse generation (figure 6.1B). Both cards had to be used and to make sure no timing drift problems would arise, the period-defining counter and the counter pair providing the pulses for the AOM should reside on the same card as the counters used for data acquisition. Hence, the trigger pulses had to be sent over the RTSI bus.

For alternating excitation (paper V), one counter on the first card was set up to generate single pulses triggered by the control pulses for the first AOM. The length of these pulses was controlled from the user interface and defined the delay between the pulse train for the first AOM and that for the second one. The falling flanks of the pulses were in turn used to, via the RTSI bus, trigger the generation of single pulses by the counter on the second card providing the control signal for the second AOM. To ensure proper timing, the delays were calibrated by FCS measurements on reflected light.

Correlation of the acquired fluorescence traces was performed us-ing a C-program, written in-house, based on an algorithm presented by Laurence et al.[100] This program also contained functionality for introducing the large time offsets needed for calculating the modulation filters (equation 25). For correlation of the square wave pulse trains defining the modulation introduced by spike removal a modified version of the same program was used. This worked on markers defining the beginning and the end of each pulse, instead of individual photons, and calculated the correlation contribution of each pulse pair analytically. The contributions to each lag time chan-nel were summed up for all pulse pairs and finally the correlation function was normalized.

Built-in Linux command line tools were used for division of correla-tion functions and other minor operations.

63

Chapter 7

CONCLUSIONS AND FUTURE OUTLOOK

In this thesis several new techniques exploiting the benefits of modulated excitation in fluorescence spectroscopy and imaging were introduced and evaluated.

The modulation approach for determining kinetic parameters com-bines the sensitivity of fluorescence with the environmental sensitiv-ity of long-lived transient states without the need for time-resolved detection. This indicates an opportunity to parallelize the approach. Generalization to transient states other than the first excited triplet state should be straight forward. Transient state imaging, which is an extension of the method in paper I was realized using a commercial LSM in paper II. The use of a widely available instrument should make this approach both accessible and attractive to many research-ers in, e.g., cellular biology.

A strategy to circumvent the problems introduced by modulated ex-citation in FCS measurements was presented. Modulation provides a new set of parameters for optimization of the measurement condi-

7 Conclusions and future outlook

64

tions with respect to photophysical properties of the dyes used. This will probably prove very useful in many future studies. One possible development that would make the technique more accessible is the introduction of a real-time correlator for measurements with modu-lated excitation.

The modulation filtering also proved useful for generation of FCS curves from data where certain sections have been removed, e.g., be-cause of spikes in the measurements. Compared to established meth-ods, this approach has several advantages. It preserves correlations over the full time range of the measurement and the generated curves are not affected by high spike frequencies. In combination with a good spike identification algorithm this approach has the potential of facilitating many future FCS studies of difficult samples.

Cross-correlating the fluorescence signals originating from the protonated and deprotonated species of a pH-sensitive fluorophore makes the protonation–deprotonation process manifest itself as an anti-correlating component in the correlation curve. The shift of both absorption and emission spectra of a ratiometric dye upon protonation makes it possible to both excite and detect the species se-lectively. Because it is generally not possible to either excite or detect one species exclusively, the combination of alternating excitation and two-color detection provides a significant improvement in contrast. Application of the approach for combination of FCS and modulated excitation presented in paper III allows the use of alternation periods longer than the characteristic time of a process under study. The pre-sented approach thereby enables the use of modulated excitation also for manipulation of photoinduced processes to further improve the measurement conditions. The concept is not limited to protonation studies but could also be applied to ratiometric dyes sensitive to other ambient conditions and ions.

The common aim of the projects presented here was to extend the range of possible applications of biomolecular fluorescence spec-troscopy. Several new methods were developed, all of which address specific common problems. Hence, they will most likely prove useful in future fluorescence-based biomolecular studies.

65

Chapter 8

INTRODUCTION TO THE INCLUDED PAPERS

This chapter contains short summaries of the papers that are in-cluded in this thesis and the contribution of the respondent to each of them.

Paper I: Monitoring Kinetics of Highly Environment Sensitive States of Fluorescent Molecules by Modulated Excitation and Time-Averaged Fluorescence Intensity RecordingThis paper presents a method for extracting the kinetic parameters governing the population of long-lived transient states from the dependence of the average fluorescence intensity on the mode of excitation modulation. The proposed approach benefits from both the high sensitivity of fluorescence detection and the susceptibility of the long-lived states to effects of the surrounding environment. This has previously been a combination unique for fluorescence cor-relation spectroscopy, which however requires high detection time resolution, fluorescent labels with high molecular brightness, and

8 Introduction to the included papers

66

very low sample concentrations. Encoding the time information in the excitation relaxes these requirements, making the proposed ap-proach more suitable for parallelization and imaging.

The respondent took active part in the experimental work and devel-oped the software for modulation control and data acquisition. The respondent was also actively involved in data processing and analysis, as well as in the writing of the paper.

Paper II: Transient State Imaging for Microenvironmental Monitoring by Laser Scanning MicroscopyHere, the method from the previous paper has been adapted to generate image contrast in a commercially available laser scanning microscope. Triplet state images of liposomes with different internal environments were acquired for demonstration. Images showing the effect of deoxygenation of the sample on the triplet population showed the potential of the technique for oxygen concentration im-aging.

The respondent participated actively in all parts of this project.

Paper III: Modulated Fluorescence Correlation Spectroscopy with Complete Time Range InformationThis paper presents two methods to combine FCS with modulated excitation along with their experimental verification. These methods allow extraction of correlation data for all lag times. One method works with standard hardware correlators but is limited to cases where the average fluorescence intensity does not change significantly dur-ing an excitation pulse. This restriction does not apply to the second method, which however requires acquisition of the full fluorescence trace. That the mode of excitation affects the populations of photoin-duced states of fluorophores was the basis of papers I and II. Here it was also shown that this could be used to optimize the measurement conditions in FCS.

The respondent worked out the major part of the theory, carried out

67


all the experiments of this project and wrote the paper. PT made important contributions in the form of theoretical discussions, simu-lations and programming.

Paper IV: Modulation Filtering Enables Removal of Spikes in Fluorescence Correlation Spectroscopy Measurements without Affecting the Temporal InformationIn FCS studies of biological samples, fluorescence intensity spikes due to aggregation or contamination present a common and serious problem. The first of the methods for modulation filtering intro-duced in paper III is proven useful for the generation of artifact free correlation curves from fluorescence traces where spikes have been removed. This method makes efficient use of all available data and preserves the temporal information throughout the measurement. It properly represents correlation between events separated by removed spikes, which is particularly important in the presence of very short intense spike or if spikes occur very frequently.

The respondent worked out the major part of the theory, carried out all the experiments of this project and wrote the paper. PT assisted with data analysis, programming and simulations. TS assisted with sample preparation.

Paper V: Fluorescence Cross-Correlation Spectroscopy of a pH-Sensitive Ratiometric Dye for Molecular Proton Exchange StudiesTo improve the robustness in the determination of exchange rates in ion exchange studies, we introduced a concept where cross-correla-tion is applied to the emission of a ratiometric dye. We demonstrated the concept with a pH-sensitive dye and showed that separate detec-tion of emission from the two protonation states followed by cross-correlation makes the protonation–deprotonation dynamics appear as an anti-correlation. Alternating excitation and temporal gating of the detected fluorescence were applied to improve the detection selectivity.

8 Introduction to the included papers

68

The respondent built the experimental setup, worked out the data acquisition strategy and developed the software. He took active part in most of the experimental work and data analysis and wrote a large part of the paper.

69

Chapter 9

ACKNOWLEDGEMENTS

Much can be learnt from books, but a lot of the knowledge and al-most all the energy that has carried me to this point emanates from people in my vicinity. Here I want to thank the ones who had the most direct impact on this work.

First, I would like to thank my supervisor, Jerker Widengren, for giving me the opportunity to work in this field and for sharing his knowledge and enthusiasm.

I also want to thank all the past and present members of the Experi-mental Biomolecular Physics group at KTH, in no particular order:Tor, for fruitful collaboration, many laughs and healthy exercise ses-sions.Per, for always taking the time to answer my questions and for invalu-able support regarding statistics, data analysis and programming.Anders, for the introduction to and company in the lab, for guidance in the world of wine, nice bike excursions and much more.Per-Åke, for support in the RXR-project (which we unfortunately had to abandon in the end), for chemistry advice in general and for being a good room-mate in the office.

9 Acknowledgements

70

Hans, for general support, for always knowing where to look for information and for proof-reading the thesis. Evangelos, for the enthusiasm with which you tried over and over again with RXR and for great company for long hours in and outside the lab.Kai, for many interesting discussions and the introduction to climb-ing.Håkan, for proof-reading the thesis.Andriy, AnnSofi, Daniel, Heike, Johan, Lei, Lina, Magnus, Sarah, Shadi, Sofia, Stefan, Thiemo and Yu for always being nice and help-ful.I want you all to know that I consider you not only coworkers but also very good friends.

All the members of the Cell Physics group—thank you for pleasant company during the lunch breaks and for all interesting discussions. You left behind a large void when you moved out of our corridor.

During my four months at the Heinrich-Heine-Universität in Düs-seldorf and the subsequent visits I have got to know a lot of nice people. I want to thank Professor Claus Seidel and all the people at the Institute for Molecular Physical Chemistry for making me feel like I belonged there. Especially, I would like to express my gratitude to:Marcelle, for making me feel at home in your office.Alessandro and Suren, for all the help in the lab.Bärbel Hofmann and Veronica Mendorf, for the help with all practical and administrative details.

Thanks to Jens, for keeping the spirit up during long sessions of mea-surements on polythiophenes. I hope we will be able to squeeze some results out of those data soon.

To carry out these projects would not have been possible without the able support of Rolf Helg and Kjell Hammarström in the mechanical workshop at AlbaNova.

I want to take the opportunity to thank my childhood friends Claes and Daniel for always being there, and all my friends in the F96(95)

71


bunch for making the last thirteen years so much more fun.

Finally and most importantly, I want to thank my father, my mother, Mats-Erik, my sister and Mikaela for all the patience, love and sup-port.

72

REFERENCES

1. Hirschfeld, T., Optical microscopic observation of single small molecules. Applied Optics, 1976. 15(12): p. 2965-2966.

2. Nguyen, D.C., et al., Detection of Single Molecules of Phycoeryth-rin in Hydrodynamically Focused Flows by Laser-Induced Fluores-cence. Analytical Chemistry, 1987. 59(17): p. 2158-2161.

3. Peck, K., et al., Single-Molecule Fluorescence Detection—Auto-Correlation Criterion and Experimental Realization with Phyco-erythrin. Proceedings of the National Academy of Sciences of the United States of America, 1989. 86(11): p. 4087-4091.

4. Moerner, W.E. and L. Kador, Optical-Detection and Spectroscopy of Single Molecules in a Solid. Physical Review Letters, 1989. 62(21): p. 2535-2538.

5. Orrit, M. and J. Bernard, Single Pentacene Molecules Detected by Fluorescence Excitation in a Para-Terphenyl Crystal. Physical Review Letters, 1990. 65(21): p. 2716-2719.

6. Shera, E.B., et al., Detection of Single Fluorescent Molecules. Chemical Physics Letters, 1990. 174(6): p. 553-557.

73


7. Rigler, R. and J. Widengren, Ultrasensitive detection of single molecules by fluorescence correlation spectroscopy. BioScience, 1990. 3: p. 180-183.

8. de Jong, L.A., et al., Receptor-ligand binding assays: technologies and applications. Journal of Chromatography B, 2005. 829(1-2): p. 1-25.

9. Eggeling, C., et al., Data registration and selective single-molecule analysis using multi-parameter fluorescence detection. Journal of Biotechnology, 2001. 86(3): p. 163-180.

10. Widengren, J., et al., Single-molecule detection and identification of multiple species by multiparameter fluorescence detection. Ana-lytical Chemistry, 2006. 78(6): p. 2039-2050.

11. Edwards, B.S., et al., Flow cytometry for high-throughput, high-content screening. Current Opinion in Chemical Biology, 2004. 8(4): p. 392-398.

12. Suhling, K., P. M.W. French, and D. Phillips, Time-resolved fluorescence microscopy. Photochemical & Photobiological Sci-ences, 2005. 4(1): p. 13-22.

13. Schäferling, M. and S. Nagl, Optical technologies for the read out and quality control of DNA and protein microarrays. Analytical and Bioanalytical Chemistry, 2006. 385(3): p. 500-517.

14. Magde, D., E. Elson, and W.W. Webb, Thermodynamic Fluctua-tions in a Reacting System—Measurement by Fluorescence Cor-relation Spectroscopy. Physical Review Letter, 1972. 29(11): p. 705-708.

15. Valeur, B., Molecular fluorescence: principles and applications. 2002, Weinheim: Wiley-VCH.

16. Lakowicz, J.R., Principles of fluorescence spectroscopy. 2nd ed. 1999, New York: Kluwer Academic/Plenum.

17. Stokes, G.G., On the Change of Refrangibility of Light. Philo-sophical Transactions of the Royal Society of London, 1852. (142): p. 463-562.

18. Becquerel, A.E., Les effects produits sur les corps par les rayons so-laires. Annales de Chimie et Physique, 1843. 3(9): p. 257-322.

References

74

19. Perrin, F., The fluorescence of solutions: Molecular induction, po-larization and duration of emission and photochemistry. Annales de Physique, 1929. 10(12): p. 169-275.

20. Widengren, J. and R. Rigler, Mechanisms of photobleaching investigated by fluorescence correlation spectroscopy. Bioimaging, 1996. 4(3): p. 149-157.

21. Eggeling, C., et al., Photobleaching of Fluorescent Dyes under Conditions Used for Single-Molecule Detection: Evidence of Two-Step Photolysis. Analytical Chemistry, 1998. 70(13): p. 2651-2659.

22. Strickler, S.J. and R.A. Berg, Relationship between Absorption Intensity and Fluorescence Lifetime of Molecules. The Journal of Chemical Physics, 1962. 37(4): p. 814-822.

23. McClure, D.S., Triplet-Singlet Transitions in Organic Molecules Lifetime Measurements of the Triplet State. The Journal of Chemi-cal Physics, 1949. 17(10): p. 905-915.

24. Garland, P.B. and C.H. Moore, Phosphorescence of Protein-Bound Eosin and Erythrosin—Possible Probe for Measurements of Slow Rotational Mobility. Biochemical Journal, 1979. 183(3): p. 561-572.

25. Kasha, M., Collisional Perturbation of Spin-Orbital Coupling and the Mechanism of Fluorescence Quenching. A Visual Demonstra-tion of the Perturbation. The Journal of Chemical Physics, 1952. 20(1): p. 71-74.

26. Widengren, J., Ü. Mets, and R. Rigler, Fluorescence Correlation Spectroscopy of Triplet-States in Solution—a Theoretical and Ex-perimental Study. Journal of Physical Chemistry, 1995. 99(36): p. 13368-13379.

27. Mets, Ü., J. Widengren, and R. Rigler, Application of the an-tibunching in dye fluorescence: Measuring the excitation rates in solution. Chemical Physics, 1997. 218(1-2): p. 191-198.

28. Minsky, M., U.S. Patent 3,013,467, Microscopy Apparatus (De-cember 19, 1961; filed November 7, 1957).

29. Naora, H., Microspectrophotometry and Cytochemical Analysis of Nucleic Acids. Science, 1951. 114(2959): p. 279-280.

75


30. Carlsson, K., et al., Three-dimensional microscopy using a confocal laser scanning microscope. Optics Letters, 1985. 10(2): p. 53-55.

31. Carlsson, K. and N. Åslund, Confocal imaging for 3-D digital microscopy. Applied Optics, 1987. 26(16): p. 3232-3238.

32. White, J.G., W.B. Amos, and M. Fordham, An evaluation of confocal versus conventional imaging of biological structures by fluorescence light microscopy. Journal of Cell Biology, 1987. 105(1): p. 41-48.

33. Pawley, J.B., ed. Handbook of Biological Confocal Microscopy. 2nd ed. 1995, Plenum Press: New York.

34. Webb, R.H., Confocal optical microscopy. Reports on Progress in Physics, 1996. 59(3): p. 427-471.

35. Sheppard, C.J.R. and T. Wilson, Depth of field in the scanning microscope. Optics Letters, 1978. 3(3): p. 115.

36. Hamilton, D.K., T. Wilson, and C.J.R. Sheppard, Experimen-tal observations of the depth-discrimination properties of scanning microscopes. Optics Letters, 1981. 6(12): p. 625-626.

37. Rigler, R. and E.S. Elson, Fluorescence Correlation Spectroscopy—Theory and Applications. Springer Series in Chemical Physics, F.P. Schafer, J.P. Toennies, and W. Zinth, eds. Vol. 65. 2001, Berlin: Springer-Verlag. 487.

38. Hess, S.T., et al., Biological and chemical applications of fluo-rescence correlation spectroscopy: A review. Biochemistry, 2002. 41(3): p. 697-705.

39. Widengren, J. and Ü. Mets, Conceptual Basis of Fluorescence Correlation Spectroscopy and Related Techniques as Tools In Biosci-ence, in Single Molecule Detection in Solution, Methods and Ap-plications, C. Zander, J. Enderlein, and R.A. Keller, eds. 2002, WILEY-VCH Verlag Berlin Gmbh: Berlin. p. 69-120.

40. Schwille, P. and E. Haustein, Fluorescence Correlation Spectros-copy, An Introduction to its Concepts and Applications, in Single Molecule Techniques, P. Schwille, Editor. 2002, Biophysical Society, Bethesda, MD, USA.

References

76

41. Thompson, N.L., A.M. Lieto, and N.W. Allen, Recent advances in fluorescence correlation spectroscopy. Current Opinion in Structural Biology, 2002. 12(5): p. 634-641.

42. Krichevsky, O. and G. Bonnet, Fluorescence correlation spectros-copy: the technique and its applications. Reports on Progress in Physics, 2002. 65(2): p. 251-297.

43. Eigen, M. and K. Kustin, The Kinetics of Halogen Hydrolysis. Journal of the America Chemical Society, 1962. 84(8): p. 1355-1361.

44. Axelrod, D., et al., Mobility measurement by analysis of fluores-cence photobleaching recovery kinetics. Biophysical Journal, 1976. 16(9): p. 1055-1069.

45. Cummins, H.Z. and H.L. Swinney, Light Beating Spectroscopy. Progress in Optics, 1970. 8: p. 135-200.

46. Berne, B.J. and R. Pecora, Dynamic Light Scattering. 1976, New York: Wiley-Interscience.

47. Ehrenberg, M. and R. Rigler, Rotational brownian motion and fluorescence intensify fluctuations. Chemical Physics, 1974. 4(3): p. 390-401.

48. Koppel, D.E., et al., Dynamics of fluorescence marker concentra-tion as a probe of mobility. Biophysical Journal, 1976. 16(11): p. 1315-1329.

49. Rigler, R., et al., Fluorescence Correlation Spectroscopy with High Count Rate and Low-Background—Analysis of Translational Dif-fusion. European Biophysics Journal with Biophysics Letters, 1993. 22(3): p. 169-175.

50. Felekyan, S., et al., Full correlation from picoseconds to seconds by time-resolved and time-correlated single photon detection. Review of Scientific Instruments, 2005. 76(8): p. 083104.

51. Berland, K.M., P.T. So, and E. Gratton, Two-photon fluorescence correlation spectroscopy: method and application to the intracellu-lar environment. Biophysical Journal, 1995. 68(2): p. 694-701.

52. Blom, H., L. Kastrup, and C. Eggeling, Fluorescence fluctuation spectroscopy in reduced detection volumes. Current Pharmaceuti-cal Biotechnology, 2006. 7(1): p. 51-66.

77


53. Kannan, B., et al., Electron multiplying charge-coupled device camera based fluorescence correlation spectroscopy. Analytical Chemistry, 2006. 78(10): p. 3444-3451.

54. Blom, H. and M. Gösch, Parallel confocal detection of single biomolecules using diffractive optics and integrated detector units. Current Pharmaceutical Biotechnology, 2004. 5(2): p. 231-241.

55. Schwille, P., F.J. Meyer-Almes, and R. Rigler, Dual-color fluores-cence cross-correlation spectroscopy for multicomponent diffusional analysis in solution. Biophysical Journal, 1997. 72(4): p. 1878-1886.

56. Bacia, K. and P. Schwille, Practical guidelines for dual-color fluo-rescence cross-correlation spectroscopy. Nature Protocols, 2007. 2(11): p. 2842-2856.

57. Dertinger, T., et al., Two-Focus Fluorescence Correlation Spectros-copy: A New Tool for Accurate and Absolute Diffusion Measure-ments. ChemPhysChem, 2007. 8(3): p. 433-443.

58. Korlann, Y., et al., Measuring diffusion with polarization-mod-ulation dual-focus fluorescence correlation spectroscopy. Optics Express, 2008. 16(19): p. 14609-14616.

59. Böhmer, M., et al., Time-resolved fluorescence correlation spectros-copy. Chemical Physics Letters, 2002. 353(5-6): p. 439-445.

60. Kapusta, P., et al., Fluorescence Lifetime Correlation Spectroscopy. Journal of Fluorescence, 2007. 17(1): p. 43-48.

61. Petersen, N.O., Scanning fluorescence correlation spectroscopy. I. Theory and simulation of aggregation measurements. Biophysical Journal, 1986. 49(4): p. 809-815.

62. Petersen, N.O., D.C. Johnson, and M.J. Schlesinger, Scan-ning fluorescence correlation spectroscopy. II. Application to virus glycoprotein aggregation. Biophysical Journal, 1986. 49(4): p. 817-820.

63. Petersen, N.O., et al., Quantitation of membrane receptor distri-butions by image correlation spectroscopy: concept and application. Biophysical Journal, 1993. 65(3): p. 1135-1146.

References

78

64. Wiseman, P.W., et al., Two-photon image correlation spectroscopy and image cross-correlation spectroscopy. Journal of Microscopy, 2000. 200(1): p. 14-25.

65. Digman, M.A., et al., Fluctuation Correlation Spectroscopy with a Laser-Scanning Microscope: Exploiting the Hidden Time Struc-ture. Biophysical Journal, 2005. 88(5): p. L33-36.

66. Digman, M.A., et al., Measuring Fast Dynamics in Solutions and Cells with a Laser Scanning Microscope. Biophysical Journal, 2005. 89(2): p. 1317-1327.

67. Hanbury Brown, R. and R.Q. Twiss, Correlation between Pho-tons in two Coherent Beams of Light. Nature, 1956. 177(4497): p. 27-29.

68. Hess, S.T. and W.W. Webb, Focal volume optics and experimental artifacts in confocal fluorescence correlation spectroscopy. Biophysi-cal Journal, 2002. 83(4): p. 2300-2317.

69. Enderlein, J., et al., Performance of fluorescence correlation spec-troscopy for measuring diffusion and concentration. ChemPhy-sChem, 2005. 6(11): p. 2324-2336.

70. Aragon, S.R. and R. Pecora, Fluorescence Correlation Spectroscopy as a Probe of Molecular-Dynamics. Journal of Chemical Physics, 1976. 64(4): p. 1791-1803.

71. Palmer, A.G. and N.L. Thompson, Theory of Sample Translation in Fluorescence Correlation Spectroscopy. Biophysical Journal, 1987. 51(2): p. 339-343.

72. Gregor, I., D. Patra, and J. Enderlein, Optical Saturation in Fluorescence Correlation Spectroscopy under Continuous-Wave and Pulsed Excitation. ChemPhysChem, 2005. 6(1): p. 164-170.

73. Widengren, J., B. Terry, and R. Rigler, Protonation kinetics of GFP and FITC investigated by FCS—aspects of the use of fluorescent indicators for measuring pH. Chemical Physics, 1999. 249(2-3): p. 259-271.

74. Bacia, K., S.A. Kim, and P. Schwille, Fluorescence cross-correla-tion spectroscopy in living cells. Nature Methods, 2006. 3(2): p. 83-89.

79


75. Kapanidis, A.N., et al., Fluorescence-aided molecule sorting: Analysis of structure and interactions by alternating-laser excita-tion of single molecules. Proceedings of the National Academy of Sciences of the United States of America, 2004. 101(24): p. 8936-8941.

76. Müller, B.K., et al., Pulsed Interleaved Excitation. Biophysical Journal, 2005. 89(5): p. 3508-3522.

77. van Amerongen, H. and R. van Grondelle, Transient absorption spectroscopy in study of processes and dynamics in biology. Methods in Enzymology, 1995. 246: p. 201-26.

78. Chen, E. and M.R. Chance, Nanosecond transient absorption spectroscopy. Methods in Enzymology, 1993. 226: p. 119-47.

79. Korobov, V.E. and A.K. Chibisov, Primary processes in the pho-tochemistry of rhodamine dyes. Journal of Photochemistry, 1978. 9(4): p. 411-424.

80. Jovin, T.M. and W.L. Vaz, Rotational and translation diffusion in membranes measured by fluorescence and phosphorescence methods. Methods in Enzymology, 1989. 172: p. 471-513.

81. Cioni, P. and G.B. Strambini, Tryptophan phosphorescence and pressure effects on protein structure. Biochimica Et Biophysica Acta—Protein Structure and Molecular Enzymology, 2002. 1595(1-2): p. 116-130.

82. Widengren, J., R. Rigler, and Ü. Mets, Triplet-state monitoring by fluorescence correlation spectroscopy. Journal of Fluorescence, 1994. 4(3): p. 255-258.

83. Koppel, D.E., Statistical accuracy in fluorescence correlation spec-troscopy. Physical Review A, 1974. 10(6): p. 1938-1945.

84. Penzkofer, A. and W. Leupacher, Fluorescence Behavior of Highly Concentrated Rhodamine 6G Solutions. Journal of Luminescence, 1987. 37(2): p. 61-72.

85. Widengren, J. and P. Schwille, Characterization of photoinduced isomerization and back-isomerization of the cyanine dye Cy5 by fluorescence correlation spectroscopy. Journal of Physical Chemis-try A, 2000. 104(27): p. 6416-6428.

References

80

86. Widengren, J., J. Dapprich, and R. Rigler, Fast interactions between Rh6G and dGTP in water studied by fluorescence correla-tion spectroscopy. Chemical Physics, 1997. 216(3): p. 417-426.

87. Song, L., et al., Influence of the triplet excited state on the photo-bleaching kinetics of fluorescein in microscopy. Biophysical Jour-nal, 1996. 70(6): p. 2959-2968.

88. Donnert, G., C. Eggeling, and S.W. Hell, Major signal increase in fluorescence microscopy through dark-state relaxation. Nature Methods, 2007. 4(1): p. 81-86.

89. Elson, E. and D. Magde, Fluorescence correlation spectroscopy. I. Conceptual basis and theory. Biopolymers, 1974. 13(1): p. 1-27.

90. Schätzel, K., Correlation techniques in dynamic light scattering. Applied Physics B: Lasers and Optics, 1987. 42(4): p. 193-213.

91. Saleh, B., Photoelectron statistics: with applications to spectroscopy and optical communication. Springer series in optical sciences, 6. 1978, Berlin; New York: Springer-Vlg.

92. Enderlein, J., et al., The statistics of single molecule detection: An overview. Bioimaging, 1997. 5(3): p. 88-98.

93. Operating Manual - Confocor 3. 2007, Jena, Germany: Carl Zeiss MicroImaging GmbH.

94. Saffarian, S. and E.L. Elson, Statistical Analysis of Fluorescence Correlation Spectroscopy: The Standard Deviation and Bias. Bio-physical Journal, 2003. 84(3): p. 2030-2042.

95. O’Connor, N. and R.B. Silver, Ratio Imaging: Practical Con-siderations for Measuring Intracellular Ca2+ and pH in Living Cells—Methods in Cell Biology, in Digital Microscopy, 3rd ed., G. Sluder and D.E. Wolf, eds. 2007, Academic Press. p. 415-433.

96. Widengren, J. and R. Rigler, An alternative way of monitoring ion concentration and their regulation using fluorescence correla-tion spectroscopy. Journal of Fluorescence, 1997. 7(1): p. 211S-213S.

81


97. Brändén, M., et al., Localized proton microcircuits at the bio-logical membrane-water interface. Proceedings of the National Academy of Sciences of the United States of America, 2006. 103(52): p. 19766-19770.

98. Becker, W., Advanced Time-Correlated Single Photon Counting Techniques. Springer Series in Chemical Physics, A. Castle-man, J.P. Toennies, and W. Zinth, eds. Vol. 81. 2005, Berlin: Springer-Verlag.

99. Nishimura, G. and M. Kinjo, Dead-time distortion in fluores-cence correlation measurements. Applied Optics, 2005. 44(17): p. 3458-3467.

100. Laurence, T.A., S. Fore, and T. Huser, Fast, flexible algorithm for calculating photon correlations. Optics Letters, 2006. 31(6): p. 829-31.

Temporal Modulation in Fluorescence Spectroscopy and

Documents

Transcript of Temporal Modulation in Fluorescence Spectroscopy and