super resolution optical imaging – image analysis, multicolor

KTH Engineering Sciences

Super resolution optical imaging – image analysis, multicolor development and biological applications

DANIEL RÖNNLUND

Doctoral Thesis in Physics Stockholm, Sweden 2014

TRITA-FYS 2014:04 ISSN 0280-316X ISRN KTH/FYS/--14:04—SE ISBN 978-91-7595-001-3

KTH Applied Physics Albanova University Center Experimental Biomolecular Physics SE-106 91 Stockholm Sweden

Akademisk avhandling som med tillstånd av Kungliga Tekniska Högskolan framlägges till offentlig granskning för avläggande av teknologie doktorsexamen I fysik den 28 februari 2014 klockan 09:00 i sal FB42, Albanova Universitets-centrum, Roslagstullsbacken 21, KTH, Stockholm.

Copyright © Daniel Rönnlund, 2014

The following articles are reprinted with permission:

Paper I. Copyright © 2012, Fed. of American Societies for Exp. Biology Paper II. Copyright © 2013, Int. Society for Advancement of Cytometry Paper III. Copyright © 2014, National Academy of Sciences Paper IV. Copyright © 2012, WILEY-VCH Verlag GmbH & Co Paper V. Copyright © 2011, BioMed Central Ltd, open access Paper VI. Copyright © 2012, Wiley Periodicals Inc.

Print: Universitetsservice US-AB, Stockholm, 2014

i

Abstract

This thesis focuses on super resolution STED optical imaging. STED provides a wealth of new informational content to the acquired images by using stimulated emission to surpass the diffraction limit in optical fluorescence microscopy. To further increase the informational content, a new method to perform multicolor STED imaging by exploiting differences in the photostability and excitation spectra of dyes is presented. In order to extract information from the images, computational algorithms which handle the new type of high resolution informational content are developed.

We propose that multicolor super resolution imaging in combination with image analysis can reduce the amount of clinical samples required to perform accurate cancer diagnosis. To date, such diagnosis is based mainly on significant amounts of tissue samples extracted from the suspected tumor site. The sample extraction often requires anesthetics and can lead to complications such as hematoma, infections and even cancer cell ceding along the needle track. We show that by applying multicolor STED and image analysis, the information gained from single cells is greatly increased. We therefore propose that accurate diagnosis can be based on significantly less extracted tissue material, allowing for a more patient friendly sampling. This approach can also be applied when studying blood platelets, where we show how the high informational content can be used to identify platelet specific activational states. Since platelets are involved in many different types of diseases, such analysis could provide means of performing truly minimally invasive diagnostics based on a simple blood test.

In addition, our data makes it possible to understand in finer detail the underlying mechanisms rendering cells metastasis competent. We combine the high resolution spatial information provided by STED with information regarding the adhesive forces of cells measured by TFM (Traction Force Microscopy) and the cell stiffness measured by AFM (Atomic Force Microscopy). Such comparisons provide a link between the specific highly resolved protein distributions and different cellular mechanics and functions.

This thesis also includes STED imaging and analysis on the spatial organization of neuronal synaptic regulating proteins, implicating the speed with which neuronal signaling can be regulated.

Keywords: Stimulated emission depletion (STED) microscopy, nanoscopy, multicolor, image analysis, diagnostics, cancer, metastasis

ii

Sammanfattning

Denna avhandling fokuserar på superupplöst STED-mikroskopi. STED ger tillgång till en mängd ny information genom att använda sig av stimulerad emission för att höja upplösningen förbi diffraktionsbarriären inom optisk mikroskopi. En ny flergfärgsmetod presenteras som tillåter fler cellinmärkningar vilket ger möjlighet att ytterligare öka informationen i de resulterande STED bilderna. För att sedan extrahera informationen tillämpas bildanalys, där nya algoritmer beskrivs som kan hantera den högupplösta informationen.

Vi föreslår att flergfärgs-STED-mikroskopi tillsammans med bildanalys kan minska mängden kliniskt prov som behövs för att utföra en korrekt cancer-diagnos. Idag baseras sådan diagnos främst på stora mängder kliniskt prov som extraheras från det suspekta tumörområdet. En sådan provtagning kräver oftast bedövning och kan leda till flertalet komplikationer som blödning, infektionsrisk och risk för spridning av cancerceller. Vi visar att mängden information som kan extraheras från enskilda celler ökar avsevärt genom att applicera flergfärgs-STED. Vi föreslår därför att behovet av extraherade celler minskar vilket innebär att mer patientvänliga metoder skulle kunna tillämpas där mängden extraherat vävnadsprov är mindre. Vi visar också hur en liknande analys kan identifiera specifika aktiveringsstadier hos blodplättar. Då blodplättar är involverade i flertalet sjukdomar kan en sådan analys ge möjlighet att utföra diagnoser baserade på ett enkelt blodprov.

Den extraherade informationen gör det också möjligt att mer i detalj förstå mekanismerna som påverkar cancercellers invasiva förmåga. Genom att kombinera den högupplösta spatiella informationen från STED med information om cellers adhesiva förmåga genom TFM (Traction Force Microscopy) och cellhårdhet genom AFM (Atomic Force Microscopy), kan vi jämföra den specifika proteindistributionen i celler med olika cellmekanismer och funktioner.

Avhandlingen inkluderar också STED studier av synaptiskt-reglerande proteiner i nervceller, där de högupplösta bilderna ihop med bildanalys ger möjlighet att estimera antalet proteiner i nervsynapserna.

iii

List of publications

Paper I

Rho GTPases link cellular contractile force to the density and distribution of nanoscale adhesions Annica K. B. Gad, Daniel Rönnlund*, Alexander Spaar*, Andrii A. Savchenko, Gabor Petranyi, Hans Blom, Laszlo Szekely, Jerker Widengren, Pontus Aspenström FASEB Journal, 2012, Volume 26, Issue 6, pages 2374-2382, DOI: 10.1096/fj.11-195800

Paper II

Spatial Organization of Proteins in Metastasizing Cells Daniel Rönnlund*, Annica K. B. Gad*, Hans Blom, Pontus Aspenström, Jerker Widengren Cytometry Part A, 2013, Volume 83, Issue 9, pages 855–865, DOI: 10.1002/cyto.a.22304

Paper III

Oncogenes induce an HDAC6-mediated vimentin filament collapse that is linked to cell stiffness Li-Sophie Z Rathje*, Niklas Nordgren*, Torbjörn Pettersson*, Daniel Rönnlund*, Jerker Widengren, Pontus Aspenström, Annica K.B. Gad PNAS, early edition, DOI: 10.1073/pnas.1300238111

Paper IV Fluorescence Nanoscopy of Platelets Resolves Platelet-State Specific Storage, Release and Uptake of Proteins, Opening for Future Diagnostic Applications Daniel Rönnlund*, Yang Yang*, Hans Blom, Gert Auer, Jerker Widengren Adv. Healthcare Materials, 2012, Volume 1, Issue 6, pages 707–713, DOI: 10.1002/adhm.201200172

Paper V

Spatial distribution of Na+-K+-ATPase in dendritic spines dissected by nanoscale superresolution STED microscopy Hans Blom, Daniel Rönnlund, Lena Scott, Zuzana Spicarova, Jerker Widengren, Alexander Bondar, Anita Aperia, Hjalmar Brismar BMC Neuroscience, 2011, Volume 12, Issue 16, DOI: 10.1186/1471-2202-12-16

iv

Paper VI

Nearest neighbor analysis of dopamine D1 receptors and Na+-K+-ATPases in dendritic spines dissected by STED microscopy Hans Blom, Daniel Rönnlund, Lena Scott, Zuzana Spicarova, Ville Rantanen, Jerker Widengren, Anita Aperia, Hjalmar Brismar Microscopy Research and Technique, 2012, Volume 75, Issue 2, pages 220–228, DOI: 10.1002/jemt.21046

Paper VII

Spatial distribution of DARPP-32 in Dendritic Spines Hans Blom, Daniel Rönnlund, Lena Scott, Linda Westin, Jerker Widengren, Anita Aperia, Hjalmar Brismar PLoS ONE, 2013, Volume 8, Issue 9, DOI: 10.1371/journal.pone.0075155

Paper VIII

Multicolor fluorescence nanoscopy by photobleaching – concept verification and its application to resolve selective storage of proteins in platelets Daniel Rönnlund, Lei Xu, Anna Perols, Amelie Eriksson Karlström, Gert Auer, Jerker Widengren Submitted Manuscript

Paper IX

Effects of resolution, target density and labeling on co-localization estimates – nanoscopy and modified algorithms to suppress false positives Lei Xu*, Daniel Rönnlund*, Annica Gad, Laura J. Braun, Pontus Aspenström, Jerker Widengren Manuscript in preparation

v

Additional publications not included in this thesis Paper X

LytA, Major Autolysin of Streptococcus pneumoniae, Requires Access to Nascent Peptidoglycan Peter Mellroth, Robert Daniels, Alice Eberhardt, Daniel Rönnlund, Hans Blom, Jerker Widengren, Staffan Normark, Birgitta Henriques-Normark Journal of Biological Chemistry, 2012, Volume 287, Issue 14, pages 11018–11029, DOI: 10.1074/jbc.M111.318584

Paper XI

Analysis of Rho-GTPase-induced localisation of nanoscale adhesions using fluorescence nanoscopy Annica K.B. Gad*, Daniel Rönnlund*, Jerker Widengren, Pontus Aspenström Methods in Molecular Biology, 2014, Vol. 1120, In press

Paper XII

Chromatin In Situ Proximity (ChrISP): Two-parameter, single-cell analysis of chromatin proximities at a high resolution Xingqi Chen*, Chengxi Shi*, Samer Yammine*, Anita Göndör, Daniel Rönnlund, Alejandro Fernandez-Woodbridge, Noriyuki Sumida, Jerker Widengren, Rolf Ohlsson Submitted Manuscript

Paper XIII

SLC10A4, a vesicular aminergic-associated transporter, modulates dopamine homeostasis Martin Larhammar, Kalicharan Patra, Martina Blunder, Lina Emilsson, Christiane Peuckert, Emma Arvidsson, Daniel Rönnlund, Julia Preobraschenski, Carolina Birgner, Christoph Limbach, Jerker Widengren, Hans Blom, Reinhard Jahn, Åsa Wallén-Mackenzie, Klas Kullander Submitted Manuscript * Authors contributed equally

vi

Acknowledgements

I would like to express my gratitude to my group, coworkers and family who made this journey possible.

To start, I would like to thank all past and present members of Experimental Biomolecular Physics.

My supervisor, Jerker Widengren, for opening the doors to science and convincing me to enter, for his constant support, positive spirit and scientific freedom which allowed me to grow as a scientist.

Hans Blom, for his substantial help with the establishment of the STED microscope and all the neuron measurements, for his always appreciated knowledge and advice.

Lei Xu, for all the nice collaborations and scientific discussions, for chats and insights into Chinese culture and politics.

Stefan Wennmalm, for his scientific determination and alternative thinking.

Sofia Johansson, for giving a biologist perspective and showing interest in image analysis.

Volodymyr Chmyrov, for his good taste in music and company in the lab.

Per Thyberg, for his support with computers and insight into bird watching.

Heike Hevekerl, for always being helpful and for nice chats regarding kids.

Johan Tornmalm, for nice company and for bringing new energy to the lab.

Jan Bergstrand, for continuing with STED and the knowledge that I leave the microscope in good hands.

Håkan Lepp, for being a good friend during my first stay in the group, for fun and frustrating FCS measurements.

Evangelos Sisamakis, for his helpfulness and expertise in photophysics, for his laughs and positive spirit.

Thiemo Spielmann, for all the nice chats during lunch and fika, for his help in the lab and support with computer software.

vii

I would also like to thank past and present members who contributed to the good atmosphere in our group: Radek Sachl, Liv Egnell, Bill Söderström, Johan Strömqvist, Andriy Chmyrov, Thor Sandén and Gustav Persson.

I have been privileged to collaborate with several people outside the group and would like to express special gratitude to:

Annica Gad, for all the successful collaborations regarding cell adhesions and mobility, for making the long, repetitive measurements in the lab fun and for the constant collisions between physics and biology.

Gert Auer, for insights into the clinical world of cancer diagnostics, for introducing platelets and their potential diagnostic applications.

Pontus Aspenström, for successful collaborations and good discussions.

I would also like to thank Anna Perols and Amelie Eriksson Karlström for their help and expertise regarding antibody conjugations, especially for the Fluodiamon and multicolor projects. Lena Scott, Anita Aperia and Hjalmar Brismar for the neuron projects and Peter Mellroth, Alice Eberhardt and Birgitta Henriques-Normark for the interesting studies on pneumococci.

During the establishment of the STED microscope, we received a lot of support from people at MPI for Biophysical Chemistry in Göttingen. I would like to express special gratitude to Stefan Hell and Lars Kastrup whose support and expertise has been crucial. I would also like to thank Dominic Wildanger, Benjamin Harke and Sebastian Berning for showing me their setups and explaining the optical components during my short stay in Göttingen.

Finally I would like to thank my family, especially my parents Stellan and Marianne, for their constant love and support. My brothers Kent for the fun times growing up and Johan for being one of my closest and dearest friends. My grandfather Henry, for his love and great spirit.

I would like to express my deepest gratitude to my wife Linda for all her love, patience and support and to my daughter Elin, who is the light of my life.

viii

Contents

Abstract i

Sammanfattning ii

List of publications iii

Acknowledgements vi

Contents viii

List of acronyms x

1: Introduction 1

1.1: Brief history of the microscope 1

1.2: Super resolution imaging 3

1.3: Digital image processing and analysis 8

2: The microscope 11

2.1: Choice of laser 11

2.2: Fluorescent dyes 12

2.3: Optical setup 14

2.4: Microscope handling and alignment 21

2.5: Resolution scaling in STED microscopy 25

3: Image processing and analysis of high resolution images 29

3.1: Image Processing, deconvolution 29

3.2: Size, density and location calculations 31

3.3: Colocalization and nearest neighbor algorithms 33

3.4: Structural analysis, fiber entanglement and orientation 36

4: High resolution as a tool for early cancer diagnostics 41

4.1: Fluodiamon 42

4.2: Metastatic cell adhesion and mobility 44

4.3: Platelets as a minimally invasive diagnostic target 52

ix

5: Dissecting synaptic proteins in dendritic spines 57

6: Multicolor fluorescence nanoscopy by photobleaching 61

6.1: Problems involved in combining multicolor and STED microscopy 61

6.2: A possible solution 62

6.3: Applications and limitations 64

7: Conclusion and outlook 67

Bibliography 69

Summary of papers and author contributions 82

x

List of acronyms

AFM atomic force microscopy APD avalanche photo diode ADP adenosine diphosphate CNB core needle biopsy D1R dopamine receptor 1 Darpp32 dopamine and cyclic adenosine 3´, 5´-monophosphate- regulated phosphoprotein EM electron microscopy FNA fine needle aspiration FWHM full width half maximum ICCS image cross-correlation spectroscopy IGF1R insulin-like growth factor 1 NKA Na+/K+-ATPase or sodium-potassium adenosine triphosphatase PALM photoactivated localization microscopy PMT photomultiplier tube PF-4 platelet factor 4 PSF point spread function rm manders colocalization coefficient rp pearson correlation coefficient SIM structured illumination microscopy STED stimulated emission depletion STORM stochastic optical reconstruction microscopy TFM traction force microscopy VEGF vascular endothelial growth factor

1

Chapter 1

Introduction

A picture is worth more than a thousand words. The ability to capture time in a drawing or a photograph has been a fundamental tool in human life throughout the centuries. From a scientific perspective, microscopy has allowed us to see and image smaller objects than our eyes can normally perceive, greatly enhancing our understanding of the world both in terms of physics, biology and chemistry.

1.1: Brief history of the microscope

The term “microscope” was coined in 1624 from the Greek words “small” and “to view”, however the earliest historical reference to magnification depicting simple glass lenses comes already from ancient Egyptian hieroglyphs dating back to 800 B.C. [1]. The first reliable reference to magnification came 900 years later during the first century by the roman A.D. Seneca who wrote “Letters, however small and indistinct, are seen enlarged and more clearly through a globe of glass filled with water”. This phenomenon was however first thought to be a property of water and it took another 1000 years before Arabian scholar Alhazan built the first lens and described how it focuses light.

The first spectacles for aiding vision were developed in 1280 in Italy and credited Amati but it is not known for certain who the inventor was [2]. The discovery of spectacles aided many persons also outside the field of science and knowledge about optics spread rapidly. Who the first inventors of the microscope are is shrouded in much confusion and probably misinformation. Galileo is often

1.1: Brief History of the microscope

2

credited with the discovery shortly after his invention of the telescope in 1607, however at the time there were already microscopes being sold by instrument makers. Credit is more generally given to the two Dutch spectacle makers Hans Jansen and his son Zacharias in 1595 who developed a two-lens system with sliding tubes, a compound microscope. No early Jansen microscopes have survived but there is a description of one instrument made for the archduke of Austria in the early 1600’s which was said to give a magnification of up to 9 times.

One of the earliest books about microscopy was written by Robert Hook in 1665 called “Micrographia” where he looked at thin slices of cork and found the fundamental units of life, cells. He also reported on the designs of the microscopes he used for his research where he introduced a third lens to further increase the magnification. Robert Hook made several large discoveries in the field of science, where one was to describe the phenomenon of diffraction and the wave theory of light to explain it. Another important contributor to the history of microscopy was Antony van Leeuwenhoek (1632-1723) who was a Dutch tradesman with no wealth or higher education. In 1668 he started making his own lenses and construct simple microscopes which he used to make the discoveries of bacteria, blood cells and sperm cells. During his life he built more than 500 microscopes where the most powerful had a magnification of 300 and a resolution of 1.4 µm [2].

In 1873 Ernst Abbe wrote the famous work where he explicitly states the resolution limit of microscopic images as approximately half the wavelength of light by the formulas:

∆ (1.1)

∆ (1.2)

where ∆r and ∆z is the minimal full width at half maximum (FWHM) resolution in the lateral plane and axial plane respectively, n is the refractive index and α the half aperture angle of the objective [3]. Abbe did not realize that this theory also applies for self-luminous objects (as in fluorescence), this was first discovered by Lord Rayleigh in 1896 who extensively studied the resolution of microscopes for both illuminated objects and self-luminous objects [4]. In his work, Lord Rayleigh also presented his famous criterion, known as the Rayleigh criterion, explaining

Chapter 1: Introduction

3

the minimal distance of two point-like objects where they still can be seen as separated.

1.2: Super resolution imaging

According to Abbe’s formula (eq. 1.1), the best theoretical resolution for visible light (~400-700nm) is about 200 nm in the lateral plane. Biological cells usually have a size ranging from a few to tens of micrometers, which makes this limited resolution more than adequate to study cell populations. However, when researching the more fundamental interactions and mechanics occurring inside cells, where the cellular proteins usually have a size of just a few nanometers, a 200 nm resolution is no longer satisfactory. Therefore many efforts have been made to increase the resolution and thereby the informational contents in the images.

Since the resolution is directly proportional to the wavelength, one approach is to leave the visible spectrum and reduce the wavelength to UV [5] or X-rays [6]. Very high resolutions can be achieved in this way but there are drawbacks in terms of much more complex optics and difficulties in imaging live samples due to the hazardous nature of the high-energy radiation. Electron microscopy is another technique where the very short de Broglie wavelength of electrons can give resolutions in the angstrom regime [7]. However the samples usually need to be sliced in thin sections and imaged in vacuum conditions preventing live cell imaging. Further the sample preparations are more cumbersome and the degree of labeling of proteins is less compared to immunofluorescence based staining [8] [9]. There are also surface scanning techniques operating at very high resolution such as Atomic Force Microscopy (AFM) [10] or Scanning Near field Optical Microcopy (SNOM) [11], however these type of techniques can only image the surface of biological samples.

Despite its relatively low resolution, light microscopy remains the most widely used imaging technique for studying biological samples. Light microscopy together with fluorescence microscopy offers very high specificity and sensitivity with a broad range of markers and fluorescent probes [12]. It offers a high throughput due to relatively easy sample preparation and fast imaging. Also live cell imaging can be achieved, commonly by transfection of the cells to express fluorescent proteins [13]. These benefits in combination with its noninvasiveness and ability to image thick samples and even into tissue, make the technique


4

versatile and applicable in many different research areas. Therefore many efforts have been made to increase the resolution for light microscopy.

Marvin Lee Minsky presented the idea of confocal microscopy in 1957 and received the patent for the invention in 1961 [14]. The idea of a confocal microscope is to scan the sample with a focused beam of light and then detect the emission using a point detector behind a pinhole. The pinhole removes scattered and out of focus light, as shown in Figure 1.1, allowing for axial sectioning of the sample and an increased lateral resolution by a factor of ~1.4 compared to wide field imaging [15]. However when Minsky presented his idea in 1957 the laser was not yet invented (the first laser was made in 1960 [16]) and it was not until the mid to late 1980’s that the first successful confocal microscope images were acquired [17].

The first method to provide a theoretically unlimited resolution for visible light microscopy was stimulated emission depletion (STED) microscopy. The technique was presented in 1994 [18] and experimentally demonstrated in 1999 [19]. STED is based on fluorescence microscopy, where the energy of an excitation photon is absorbed by a fluorescent molecule to excite an electron from the ground state to an excited state, as shown in Figure 1.2a. In most cases the excited electron will lose some of its energy in terms of heat (vibration) before it decays back to the ground state. This decay will release the remaining intensity as a new photon which has less energy and therefore a longer wavelength compared to the excitation photon, allowing for separate detection. STED microscopy uses stimulated emission which is a similar phenomenon but rather than exciting an electron from the ground state to the excited state, it de-excites an already excited electron back to the ground state [20], as shown in Figure 1.2b.

Stimulated emission can occur if the incoming photon lies in the emission spectrum of the excited fluorescent molecule. The emitted photon from such a process will share the same properties as the incoming photon in terms of wavelength, phase, polarization and direction. This makes it possible to easily filter out the stimulated emission from the spontaneous emission. The wavelength of the STED beam is usually selected at the red-tail of the emission spectrum mainly to avoid any excitation caused by the powerful STED beam but also so that the detection can be kept at the emission peak. A schematic excitation and emission


5

Figure 1.1: Confocal microscope schematic. An excitation laser (green lines & arrows) is reflected on a dichroic mirror (A) and focused by an objective lens (B) onto the sample. Fluorescence from the focus point (yellow straight lines and arrows) is collected back through the objective lens and transmitted by the dichroic mirror to the detector lens (C), focusing the beam through the pinhole (D) and onto the detector. Background fluorescence originating from an out of focus plane (yellow slashed line) or scattered from another area of the cell (yellow dotted line) will either not be collimated by the objective lens or focused to the side of the pinhole.

Figure 1.2: Jablonski diagram depicting fluorescence excitation and emission (A) as well as stimulated emission (B).

spectrum with corresponding selected excitation, emission and STED wavelengths is shown in Figure 1.3A.

The trick to increase the resolution by the use of stimulated emission is to cleverly shape the STED focus into a toroidal or “doughnut” shape which is superimposed onto the excitation focus [21], as shown in Figure 1.3B and discussed more in Chapter 2.2. The toroidal shape of the STED beam will have the effect that all excited fluorescent molecules which are outside the center of the


6

STED focus, where the STED intensity is minimal, will be de-excited by stimulated emission. This process will efficiently shrink the volume from which emission occurs and thus increase the resolution, a schematic comparison of confocal and STED resolution is shown in Figure 1.3C. The lateral resolution in STED microscopy can be described by the formula [22]:

∆ (1.3)

which resembles the original formula by Abbe (eq. 1.1) with an additional square root term where I is the intensity of the STED beam and Isat the saturation intensity where half of the fluorophores are stimulated back to the ground state. As can be seen in the formula a higher STED intensity (I) will lead to higher resolution and there is no theoretical limit. The development of the STED technique in combination with development of new more powerful lasers and fluorescent markers keeps pushing the resolution down to reach even molecular sizes, where the current record is 2.4 nm [23].

Localization based super resolution microscopy was introduced in 2006 with the techniques PALM (photoactivation localization microscopy) [24] and STORM (stochastic optical reconstruction microscopy) [25]. These techniques are based on keeping only a small subpopulation of all dye molecules simultaneously in the bright state and imaging them onto a camera. If few enough molecules are emitting and their diffraction pattern does not overlap, they can be mathematically localized with high precision [26]. Single molecules are then switched individually and stochastically and captured in different images so that all molecules can be localized. The final high resolution image will be a computationally calculated image of the combination off all captured images. Usually thousands to tens of thousands of images needs to be acquired but very high resolution can be achieved, down to just a few nanometers [27]. A schematic figure of the image acquisition in localization based super resolution microscopy is shown in Figure 1.4.

Another method which surpasses the diffraction limit is structured illumination microscopy (SIM) [28]. In SIM the sample is scanned with multiple excitation light patterns giving rise to Moiré effects from which high resolution informational content can be extracted. This type of microscopy usually increases the resolution by a factor of two compared to confocal resolution. Apart from these techniques,


7

Figure 1.3: (A): Excitation (black) and emission (red) spectrum of ATTO594 with possible selections of excitation, detection and STED wavelengths. (B): By overlapping the excitation focus with the doughnut shaped STED focus, the latter created by destructive interference from the induced phase shift by the vortex phase plate, the resolution is effectively increased, in this example from 250 nm to 40 nm. (C): Schematic comparison of how a filamentous structure (C1) would be depicted with the above confocal (C2) and STED (C3) resolutions.

Figure 1.4: Schematic representation of localization based super resolution microscopy. Several images are acquired with only a few fluorescent molecules active in each image. The location of each molecule can then be calculated with high precision from each separate image and the combination of all such localizations yields the final super resolution image.


8

there are a multitude of other super resolution techniques where the majority are localization based such as GSDIM [29], and PiMP [30] but there are also correlation based methods such as SOFI [31].

1.3: Digital image processing and analysis

Digital images first appeared in the 1920s when the newspaper industry wanted to send pictures between London and New York [32]. The images were sent by submarine cable where the time required transporting the images could exceed one week and the resulting image quality was poor. However, more efficient ways to code and reconstruct the images were quickly developed and the Bartlane cable transmission system developed in the early 1920s reduced the time to a few hours. The original Bartlane system coded the images in 5 levels of gray which was increased to 15 levels in 1929, thus greatly increasing the image contrast.

However, digital images require so much storage and computational power that progress in the field was intimately tied to the development of the digital computer. The first computers powerful enough to perform meaningful digital image processing appeared in the early 1960s. The first clear records were image processing was used to correct for image distortions was in 1964 when the U.S. spacecraft Ranger 7 took images of the moon surface, were corrections had to be made for inherent aberrations in the camera. In the late 1960s and early 1970s digital image processing started to be used in medical imaging. The development of the first computerized tomography (CT) system in the early 1970s by Godfrey N. Hounsfield and Allan M. Cormack is considered one of the most important events in the application of image processing and was also rewarded the Nobel Prize in Medicine 1979. The introduction of the personal computer by IBM came 1981 and since then there has been a continuous increase in computer performance combined with a decrease in price. With the introduction of the internet in the early 1980s, the applications of digital images has exploded and today they are used in almost all fields of research and industry, as well as in everyday life.

This thesis focuses on the image processing and analysis of optical microscopy images. For conventional optical imaging techniques such as confocal or wide-field microscopy, there are many ways to perform image analysis to extract different types of data. One of the most common image analysis procedures is to


9

look at the intensity overlap or colocalization of separate targets to study their interactions. There are many ways to measure colocalization, but the simplest and most commonly used techniques are the Manders colocalization coefficient (rm) and the Pearson correlation coefficient (rp) defined as [32] [33]:

∑ ·

∑ ·∑ (1.4)

∑ ·

∑ ·∑ (1.5)

where Ri and Gi are the intensity values of pixel i in the “Red” and “Green” color channel and n is the total number of pixels. The equation of rm is very straightforward with resulting values ranging from 0 to 1, where 0 mean no colocalization and 1 mean perfect colocalization, making it seemingly easy to interpret. The equation of rp is very similar to rm with the exception of removing the average intensity for each pixel. Due to this subtraction, rp can have values ranging from -1 to 1 where negative values refer to negative colocalization. Many therefore argue that the rp is more difficult to interpret as compared to rm [34] [35] [36]. However, as will be further discussed in Chapter 3.3 and in Paper IX, the subtraction of the average values makes rp far more robust in terms of less sensitivity to noise, density and resolution effects as compared to rm. Two other commonly used colocalization coefficients are M1 and M2 [34]:

1 ∑ ,∑ , 2 ∑ ,

∑ (1.6)

,

0 , ,

0 (1.7)

where TG and TR are intensity thresholds which should be set above the noise level but include most of the target signal. These coefficients separate the green and red colocalization which can be beneficial when there is uneven labeling of targets.

The set threshold values has major impact on the resulting colocalization values, meaning special care has to be taken as to not artificially alter the degree of colocalization. There are methods to automatically set this threshold, where the Costes method is often considered [37]. Although these methods works well in some specific cases they generally fail to give a satisfying overall applicability,

1.3: Digital image processing and analysis

10

where the Costes method fail at high densities [38] but also at higher resolutions when there is more dark pixels. A more advanced method for measuring colocalization is the ICCS (Image Cross Correlation Spectroscopy) where the fluorescence intensity fluctuations in the pixels are measured by a spatial autocorrelation function as [39] [40]:

, ,, ,

1 (1.8)

where F represents the Fourier transform, F-1 the inverse Fourier transform and F* its complex conjugate. The resulting function is fit to a two dimensional Gaussian where the amplitude corresponds to the number of colocalized particles. This type of colocalization measurement is less straightforward but does provide the best results in most circumstances, see Chapter 3.3 and paper IX for more details.

Colocalization is one of many ways to analyze optical microscopy images. Other analysis methods include long term cell studies where as an example algorithms are made to study and track how the natural killer (NK) cells of our immune system reacts and kills hostile cells [41]. These types of studies can involve thousands of cells studied over long periods of time which is not possible to analyze by hand [42]. A similar type of study is to track individual proteins to study their motility and interactions [43]. There is a continuous development of image processing and analysis tools for optical microscopy imaging, with both commercial and free alternatives covering many different research areas [44] [45]. However, as of yet, few analytical tools have been designed for the new informational content provided by high resolution imaging. Therefore, a significant part of this thesis involves the extraction of such high resolution data, this will be covered in more detail in chapter 3.

11

Chapter 2

The microscope

A cornerstone of this thesis lies in the performance of the dual color STED microscope which makes the super resolution imaging possible. Therefore this chapter will be dedicated to the construction, maintenance and achievable resolution of such a microscope, hopefully providing insights for those interested in performing super resolution STED imaging. Regarding the construction, topics such as suitable lasers, fluorescent dyes and mechanical and optical components will be covered as well as the design of the optical setup. Regarding maintenance, topics concerning alignment and handling of a STED microscope will be addressed. The chapter will end with a short discussion regarding properties affecting the resolution of the microscope.

2.1: Choice of laser

In STED microscopy it is generally desirable to achieve as high resolution as possible. Since the resolution scales with the power of the STED beam (see equation 1.3), powerful lasers are required. Either pulsed lasers or continuous-wave (CW) lasers can be used; however, systems based on pulsed lasers usually provide greater resolution due to the ability to sequentially excite and apply the STED beam together with high, temporally confined pulse energies. On the other hand, CW lasers have the benefit of faster image acquisition, where several images can be acquired per second allowing for video-rate STED [46]. Recent studies have also combined CW STED lasers with pulsed excitation and time-gated

2.1: Choice of laser

12

detection to increases resolution to similar levels as the pulsed STED laser systems at the cost of reduced detection signal [47].

For our microscope, we chose a pulsed supercontinuum laser (SC-450-PP-HE, Fianium Ltd., Southampton, UK) which provided short (~100 ps) high energy pulses over a broad spectrum (400-2000 nm). The broad spectrum of the laser makes it possible to specifically select and extract the desired wavelengths for all excitation and STED beams, providing the additional bonus of having all beams inherently synchronized since they originate from the same laser source.

2.2: Fluorescent dyes

STED microcopy puts a higher demand on the fluorescent dyes both in terms of brightness and stability as compared to conventional light microscopy. The brightness of the dye is important since the STED beam effectively shrinks the excitation volume and therefore fewer dyes will be left to emit, with the exception of single dyes in the excitation volume where the STED image will achieve similar brightness. The stability of the dye is important since higher resolution requires a finer pixel size; where the pixel size has to be equal to or smaller than the resolution divided by two to avoid image artifacts according to the Nyquist criterion [48] [49]. Smaller pixels mean longer acquisition time where the dyes will undergo multiple excitation and de-excitation cycles which could lead to bleaching. Therefore, the selected wavelengths of the microscope are generally set by the fluorescent dyes which will be used for imaging [50].

We choose to construct the setup optimally for imaging the dyes ATTO-647N (Atto-Tec GmbH, Siegen, Germany) and ATTO/Alexa-594 (Life Technologies, Carlsbad, California, US), see Figure 2.1 for spectra. These dyes are both bright and photostable and have previously been successfully imaged with STED microscopy [51] [52]. However, recent advances in the development of new fluorophores have expanded the amount of possible fluorescent markers suitable for STED microscopy [53], providing more flexibility in terms of labeling. The wavelengths for excitation, emission and STED was selected to be 568 ± 5 nm, 615 ± 15 nm and 710 ± 10 nm for ATTO-594 and 647 ± 5 nm, 675 ± 15 nm and 750 ± 10 nm for ATTO-647N, as shown in Figure 2.1.

The two selected dyes are relatively close in spectrum making it possible to keep both STED beams at the far-red part of the visible spectrum (710 ± 10 and

Chapter 2: The microscope

13

Figure 2.1: Excitation and emission spectra of ATTO594 and ATTO647N together with selected excitation, emission and STED wavelengths.

750 ± 10 nm). This red shifted separation of the STED wavelengths comes with the benefit of avoiding excitation of both dyes by the powerful STED beams, making simultaneous acquisition possible. Previous dual color STED microscopes were designed for dyes with more separated spectrum, where the images of the two respective dyes needed to be taken sequentially due to excitation and bleaching of the red shifted dye by the STED beam of the green shifted dye [54]. The simultaneous acquisition provided by our microscope allows for faster imaging with minimal risk of drift between the imaged dyes during capture. It also provides the possibility to take several images of the same area.

Another benefit of having both STED beams at the far-red part of the visible spectrum is less absorption and scattering when imaging tissue samples, where the absorption of hemoglobin is approximately 100 times higher at 600 nm as compared to 700 nm [55]. Further, at the selected excitation and emission wavelengths, there is very little background signal due to autofluorescence originating from the studied cells [56]. A negative effect of having two dyes relatively close in spectrum is the possible overlap of emission wavelengths

2.2: Fluorescent dyes

14

leading to crosstalk, which is the signal of one dye assigned to the wrong color channel. For our selected dyes and emission bands we have approximately 20 % crosstalk from ATTO-594 into ATTO-647N channel and 5 % from ATTO-647N to ATTO-594. This crosstalk can be reduced by applying a time delay between the excitation and STED beams of the two respective dyes before they reach the sample, combined with the same time separation of the detectors [57]. Since the lifetime of organic fluorophores is usually 2-5 ns [58], a time separation of 40 ns makes the spectral crosstalk very low (< 4 % of ATTO594 and < 3 % for ATTO647N).

Other dyes which have successfully been imaged with our setup are ATTO-590, Alexa-594, Dylight594 and Texas-Red for the “green” channel (exc. 568 ± 5 nm, STED 710 ± 10 nm), and Abberior STAR 635 and Dylight650 for the “red” channel (exc. 647 ± 5 nm, STED 750 ± 10 nm). The dyes ATTO-590, Alexa-594, Dylight-594, Texas-Red and AbberiorSTAR635 are all bright and photostable allowing for many repetitive images in the same area. Dylight650 on the other hand is less photostable and only a few images can be acquired before it is completely bleached. However, the reduced photostability of Dylight650 as compared to ATTO-647N and AbberiorSTAR635 can be used as an advantage as discussed further in chapter 6.

2.3: Optical setup

Many optical components are needed to construct a dual color STED microscope based on a supercontinuum laser; a schematic of the setup used for this thesis is shown in Figure 2.2 where the design of the setup is a modification based on the work of Wildanger et al. [51]. The setup can be divided into two major parts. The first part of the setup, including approximately half of the components, is used to select the wavelengths for the two excitation beams and the two STED beams.

First the beam is guided towards two dielectric coated mirrors (G34 0783 000, Linos, Luxembourg) reflecting wavelengths between 400-950 nm and transmitting unwanted higher wavelengths into two separate beamdumps (mechanical workshop, KTH-albanova, Stockholm, Sweden). The beam coming from the laser is randomly polarized and split into two separate linearly polarized beams by a polarization beam splitter cube (PTW 1.15, B. Halle, Berlin, Germany), transmitting parallel (P) polarized light and reflecting perpendicular (S) polarized


15

light. To minimize intensity losses, the polarization of the reflected S-polarized beam is turned to P-polarisation by a half-waveplate (Rac 4.2.15, B. Halle, Berlin, Germany) before being sent to a prism (IB-21.6-60.6-SF10, Melles Griot, Albuquerque, USA) at the Brewster angle [59]. The prism causes a chromatic dispersion of the beam where different wavelengths leave the prism in different angles. The dispersed beam is then collimated and directed onto a slit placed atop a fine tunable translational stage (M-SV-0.5, M-SDS40, Newport, Irvine, USA), offering the possibility to very precisely select the desired wavelength span of the beam by moving the slit and changing its opening width. By reflecting the beam back through the same lens and prism, the dispersed beam is combined again and further guided into a polarization maintaining single mode optical fiber (PMC-630-4,6-NA011-3-OPC-150-P, Schäfter + Kirchhoff, Hamburg, Germany) by a beamcoupler (60SMS-1-0-A15-02, Schäfter + Kirchhoff). This part of the setup effectively acts as a monochromator and is used to select our first STED beam at 710 ± 10 nm.

The other half of the laser intensity, which was transmitted by the polarizing beamsplitter cube, is split into two new beams by a dichroic mirror with high reflection of wavelengths over 660 nm (650dcspxr, Chroma, Bellows Falls, USA). A copy of the components used to extract the first STED beam is used for the reflected beam to select the wavelength of the second STED beam at 750 ± 10 nm. The second STED beam is then coupled into the same type of fiber with the exception of being 8.5 meters longer, effectively adding a ~40 ns delay between the two beams. The remaining intensity is split one final time by a second dichroic mirror (z568rdc, Chroma) reflecting wavelengths below 580 nm and transmitting the remaining wavelengths at 580-660 nm. These two beams are used to select the excitation wavelengths at 568 ± 5 nm and 647 ± 5 nm using filters (z568/10x and 647/10x, Chroma). Thereafter the excitation beams are sent to respective polarization maintaining single mode optical fibers (PM-S460-HP and PM-S630-HP, Thorlabs), where the second excitation at 647 ± 5 nm also has a 8.5 meter longer fiber to add the ~40 ns time delay.

The second part of the setup is used to acquire the images. After the fiber out-coupling all beams are collimated using achromatic lenses (AC254-040-A1-ML for excitation beams and AC254-040-B-ML for STED, Thorlabs). The STED beams pass through two separate vortex phase plates (VPP-1, RPC Photonics, New York, USA), which are crucial elements in the generation

2.3: Optical Setup

16

Figure 2.2: Schematic of the STED microscope used in this thesis together with beam paths for the excitation, STED and emission wavelengths. (a1, a2): Polarizing beamsplitter cubes. (b1, b2): Monochromator part including prism and slit for selecting STED wavelengths. (c1-c6): Dichroic mirrors. (d1, d2): Polarizer and filter to adjust power and select wavelength for each excitation beam. (e1, e2): Vortex phase plates. (f): Objective and piezo-scanning stage with attached sample holder. (g1, g2): APD detectors placed inside a dark box (gray area).


17

of the doughnut-shaped focus of the STED beams. Each phase plate is made of polymer replica on a glass substrate and includes many different vortices, which are small areas of spirally increasing thickness. By matching the vortice thickness to the wavelength of the STED beam, a phase difference of one pi can be achieved. This phase difference, combined with circular polarisation, will create effective destructive interference in the centre of the focus [60].

For our STED wavelengths at 710 ± 10 nm and 750 ± 10 nm, the closest matches on the phase plate are 678.8, 688.9, 727.7, 735.8 and 796.3 nm. We found that for our setup the best matches were when the vortex was slightly below the wavelength of the beam. In our case, both 678.8 nm and 688.9 nm work well for the 710 ± 10 nm beam and 727.7 nm and 735.8 nm for the 750 ± 10 nm beam. New models of phase plates are being developed where a recent modality creates the destructive interference using polarization rather than phase shift [61]. The advantage of using polarization is that it is more tunable in terms of wavelength and can be set to affect only the STED beam. By such a phase plate, both the excitation and STED beam can be coupled together and pass through the same phase plate, making the set-up more robust and easier to align.

In our setup, after passing through the phase plates the two STED beams are combined with their respective excitation beams by two dichroic mirrors (655dcspxr and 690dcspxr, Chroma), reflecting the STED wavelengths while transmitting the excitation wavelengths. Since the fibers maintain the linear polarization of the beams, the fiber outcoupling can be turned so that the first excitation and STED beam pair (568 ± 5 nm and 710 ± 10 nm) have S-polarization and the second pair (647 ± 5 nm and 750 ± 10 nm) have P-polarization. All four beams can then be combined by a polarization beam splitter cube (PTW 1.15, B. Halle).

Good circular polarization is needed in order to achieve the best STED focus with minimal intensity in the center, to achieve this all beam passes through a half- and a quarter-waveplate (RAC 4.2.15 and RAC 4.4.15, B. Halle) before being sent to the objective. As the polarisation is very sensitive, these waveplates were mounted in fine-tuning rotatable holders (PRM1/M, Thorlabs). To check the polarization, we constructed a detector with a rotating polarizer that could be

2.3: Optical Setup

18

coupled to an oscilloscope. If the polarization is not perfectly circular, the polarizer will let through light depending on its rotational position and the rotational frequency will be discernible in the oscilloscope. By fine tuning the position of the wave-plates, the amplitude of the oscillations should get as low as possible and in the best case there will be a flat line indicating perfect circular polarization. The half-waveplate should in principle not be necessary to achieve circular polarization, but we found that the fine tuning of the quarter-waveplate became easier with its addition.

Since a fluorescence event can be very fast, sometimes less than a nanosecond, the timing between the excitation and STED beam reaching the sample is important. If the excitation arrives too early in respect to the STED, some of the excited fluorophores will emit before being stimulated down to the ground state, which will lead to a confocal background signal blurring the image. If on the other hand the STED beam arrives before the excitation, some of the STED intensity will be lost and a lower resolution will be achieved. Therefore, the STED pulse should ideally arrive only a few picoseconds (~10-11 s ) after the excitation, allowing all fluorescent molecules to be excited and release some vibrational energy (~10-15-10-12 s) but not fluoresce (~10-10-10-8 s) [62]. Thus the path length of the STED and excitation beams from the laser to the objective should be similar. To set the path length and thus timing more accurately, we placed either the fiber incoupling or outcoupling of each beam on a railsystem (FLS 40, Linos), allowing manual shift of the positions. This type of coarse alignment is sufficient since one millimeter corresponds to just a few picoseconds due to the high speed of light.

To measure the optimal path lengths, a sample can be prepared with the desired fluorescent marker evenly spread on a cover glass. By illuminating the sample with both excitation and STED beams (without the VPP in the beam path and not at full power to avoid saturation effects), the optimal distance can be found where the emission from the sample is lowest; since this marks where the STED beam is most efficient.

In confocal microscopy, the resolution is limited by the wavelength and the numerical aperture (NA) of the objective (Equation 1.1). Even if this resolution limit is surpassed by applying STED microscopy, the resolution still scales with the wavelength and NA of the objective; since these govern the size of the focal


19

area and therefore the intensity distribution (Equation 1.3). Thus high NA objectives such as oil immersion objectives will provide the best resolutions. In our system, we use a HCX PL APO 100× objective with a NA of 1.4 (Leica, Wetzlar, Germany). To obtain the best focus of the objective, it is important to have immersion oil with the correct refractive index. The company supplying the objective usually also supplies immersion oil. However, since the refractive index is dependent on many parameters, including the temperature in the laboratory, this brand of oil is not necessarily optimal. We routinely use a mixture of two types of oil in a 25/75 ratio (type B and 37; Cargille Laboratories, Cedar Gove, NJ, USA) to obtain the best focus in our lab at ~23.0 °C.

The sample is placed on a piezo stage scanning unit that is coupled to a closed-loop controller (MAX311/M and BPC203; Thorlabs), offering a positional resolution down to 5 nm in three dimensions. The fluorescence coming from the sample is collected back through the objective and then reflected on a dichroic mirror (Laseroptik, Garbsen, Germany). This dichroic needed to be custom made to provide high reflection (> 95%) for the emission wavelengths at 615 ± 15 nm and 675 ± 15 nm while transmitting the excitation wavelengths 568 ± 5 nm, 648± 10 nm (> 70%) and STED wavelengths 710 ± 10 nm and 750 ± 10 nm (> 85%). When selecting filters and mirrors, it is more important to minimize intensity losses of the STED beams rather than the excitation beams. The reason for this is that while the lasers usually provide more than enough laser power for the excitation, as high STED power as possible is required for the best resolution (Equation 1.3).

The fluorescence emission is thereafter separated by a final dichroic mirror (z633rdc, Chroma) and focused into two separate fiber coupled avalanche photo diodes (APD, SPCM-AQRH-13-FC, Becker & Hickl, Berlin, Germany) with emission filters (ET615/30m and HQ675/30, Chroma). Here the fibers (M31L01, Thorlabs) act as pinholes, removing out of focus and stray light, with a core diameter of 62.5 µm corresponding to roughly 1.2-1.4 times the size of an airy disc.

For alignment of the system we use a CCD camera (SPC900NC, Philips, Amsterdam, Netherlands) and a photomultiplier tube (PMT, H8259-02, Hamamatsu, Hamamatsu City, Japan), where the PMT has a larger detection area as compared to the APD which allows for easier detection of misaligned signals. A

2.3: Optical Setup

20

pellicle beamsplitter (BP145B2, Thorlabs) is placed directly after the objective to guide the reflected beams to the CCD or PMT, as shown in figure 2.3. The pellicle is placed on a magnetic mount (KB25/M, Thorlabs) allowing it to be removed when acquiring images using the APDs.

By placing a flippable mirror before the PMT, the reflected light is guided towards the CCD camera to image a large field of view of the sample. A white light source (lamp) is then focused on the back focal plane of the objective to illuminate a larger area of the sample. Also wide-field fluorescence can be acquired by placing a lens focusing the excitation beam to the back focal plane of the objective. By placing a removable mirror before the APD detectors, the emission light can be guided to the CCD rather than APDs. The wide-field images make it easier and faster to locate desired areas and cells in the samples. All of the detection units are placed inside a dark box made of black plastic coated cardboard and floored with textured self-adhesive paper to minimize stray light which may cause background noise.

The detectors together with the piezo scanning stage and shutter are coupled to a computer using two separate NIDAQ cards (NI BNC-2110, NI PCI-6731, National Instruments, Austin-Texas, USA). The software used to control the electronics and capture the images is Imspector (Department of NanoBiophotonics, MPI for biophysical Chemistry, Göttingen, Germany).

Obtaining high resolution comes with the drawback of being more sensitive to drift. A 10 nm drift between two detection channels can be acceptable at confocal or widefield resolutions (i.e., at 200-300 nm), however at super resolution, a 10-nm error can have a major impact on the resulting image and image analysis. Therefore it is crucial to have the optical components mounted in stable and relatively heat insensitive holders (such as M-SN100-F2KN, Newport, Irvine-CA, USA). Further, the microscope setup should be placed in a lab with good temperature control and preferably on a good air dampened optical table.


21

Figure 2.3: Schematic of the CCD and PMT detection pathway. (a): White light source for widefield illumination of the sample to the CCD. (b): Flippable mirror guiding the light to the CCD if up and PMT if down. (c): PMT detector. (d): CCD detector. (e1, e2): Removable lens and mirror for using widefield fluorescence.

2.4: Microscope handling and alignment

Before each imaging session, the setup needs to be aligned. The alignment procedure includes inspecting and correcting size and shape of excitation and STED foci, overlap of these foci, detector pinhole positioning and power optimization.

To inspect the size and shape of the excitation and STED beams individual foci, we detect the reflected light coming from 80 nm colloid gold nanoparticles (henceforth called gold beads, ArrayGC80, BBI Solutions, Cardiff, UK). The gold beads are spin coated onto a cover glass with thickness matched to the objective and sealed in immersion oil to obtain as good refractive index match as possible. We received such samples as a kind gift from Dr. Lars Kastrup (MPI-BPC, Göttingen, Germany).

To start the alignment, the object slide with the gold bead sample is placed on top of the piezo scanning unit. The reflection from the gold beads is usually very


22

strong, so it is advisable to start with the STED and excitation beams at low power. Open the beam path to one of the excitation or STED beams, where we routinely use the first STED beam, and turn on the white light source. By guiding the reflected light to the CCD, using the pellicle and flippable mirror according to Figure 2.3, an image of the gold beads together with the focus of the laser beam can be obtained (the latter reflected by the refractive index mismatch of the cover-slip and the mounting medium). By moving the sample, different gold beads can be placed in the focus of the laser beam which should provide a strong reflection of the laser light. The gold beads are usually not uniform in terms of size and shape and therefore it is beneficial to look at the reflection of several beads before selecting one for alignment. The selected bead should provide a good reflection (too weak will cause problems with noise, too strong could give saturation effects) and a relatively even intensity profile.

After selecting a gold bead by the CCD the beam path should be changed so that the reflective light instead goes to the PMT, for our system this is performed by flipping down the CCD mirror shown in Figure 2.3. The white light source should now be turned off. Start by scanning the bead in one dimension (X, Y or Z) with one of the excitation or STED beams. Fine-tune the position of the bead so that it is centred on the focus of the laser beam and thus provides the strongest reflection. Adjust the beam power so that a good signal-to-noise ratio is achieved without saturating the detector.

Acquire two dimensional images (XY, XZ and YZ) to record the point spread function (PSF) of the laser focus. Look at the XZ and YZ images to see if the PSF is straight or tilted. A tilted PSF indicates that the beam does not go straight into the objective. This can be corrected by beam walking which is performed using the two closest mirrors to the objective which only affects the desired beam. First tilt the mirror furthest away from the objective and then re-align the beam with the second mirror. If the beam profile improves, continue until it is straight; if it worsens, move in the other direction.

The size of the PSF should be close to the diffraction limit (Equation 1.1, 1.2) for the excitation beams and STED beams without the phase plate. If the size is too big, the objective does not focus the laser beam correctly which could be due to a number of reasons. First make sure that the beam profile going in to the detector is large enough to fill the entire objective, if not use a beam expander to


23

increase the size of the beam. The beam also needs to be well collimated, adjust the focusing lens of the laser and see if the profile improves. The focusing of the laser is also affected by the immersion oil, the thickness of the cover-slip and the mounting medium. Always use cover-slips of a thickness specified by the objective. In addition, it is beneficial to experiment with different mounting mediums and immersion oil mixtures to optimise the focus of the system, see chapter 2.2 for the oil mixture we routinely use.

Repeat this procedure for all excitation and STED beams, making sure they are all close to the diffraction limit in size and have straight PSFs. For our setup this type of alignment is rarely needed, usually once every 2-6 months, however it should always be checked before imaging. Overlap the beams by selecting one beam and then moving all of the other beams so that their respective profiles overlap the selected beam; this is performed for each beam separately by adjusting the mirror which only affects the desired beam path positioned closest to the objective. Take two-dimensional XY images of the profiles and make sure that the minima for both STED beams are at the very same position, together with the maxima of the excitation profiles, as shown in Figure 2.4A. Also acquire XZ or YZ images where the STED intensity should be close to zero in the center of the profiles and the excitation intensity should be maximum, according to Figure 2.4B.

After overlapping the excitation and STED foci, the APD detectors should also be aligned to the same foci. Remove the pellicle beam splitter guiding the light to the PMT, turn off the PMT and turn on the APD detectors. Since we have an excitation band (647 ± 5 nm) in between the two emission bands (615 ± 15 nm and 675 ± 15 nm) we can use the reflection of this beam to align the detectors with minimal chromatic errors. Remove the emission filters of the APDs and optimize the pinhole positions by maximizing the reflected signal from the gold bead, the signals should also be aligned for both detectors according to Figure 2.4C. When alignment is done, put back the emission filters. To test for possible chromatic errors, in particular if the wavelength used for alignment is distant from the emission wavelength, it is good to have a sample where fluorescence emission can be used rather than reflection. If possible, use a probe that provides a broad enough emission spectrum so that it can be detected in both of the emission channels, where we use fluorescent crimson beads with emission between ~600-750nm (F-8782, Life Technologies).


24

Figure 2.4: PSFs and alignment of excitation, STED and emission beams. (A): In the XY-plane, the two STED beams should have their minimum at the very same position as the peak of the excitation profiles, as shown by the line profiles measured between the arrows in the overlapped image. (B): In the XZ/YZ plane and with the applied vortex phase plate, the STED beams should have zero intensity along the whole depth direction. (C): The two APD detectors should also be overlapped and focused to the same spot as the excitation and STED beams.

When the detectors and laser beams are all aligned to the same focus the high resolution images can be acquired. In our lab we routinely perform a new alignment procedure every four hours to account for drift during the imaging. During this time we usually have a drift of less than ten nanometers unless there has been a large vibration or temperature difference in the lab.


25

The spectra and polarization of the STED beams should also be measured at least once a month. The spectrum should be measured directly before the vortex phase plate since this component is the most sensitive to a wavelength drift. If there is a shift in spectra, it can be compensated by moving the slit in the monochromator part of the STED beam, shown in Figure 2.2.

The polarization can be measured using a detector with a rotating polarizer coupled to an oscilloscope, as described in Chapter 2.2. If a wave pattern with significant amplitude is seen, due to sub-optimal circular polarization, finely rotate the quarter and half wave plate until the amplitude is minimal. Both a wavelength shift and drift in polarization of the STED beam affects the efficiency of getting destructive interference at the focus of the objective, thus reducing the emission signal by leaving part of the STED intensity to react with the centermost molecules. Therefore, if the acquired high resolution images have less signal than normal, it could be due to a drift in wavelength or polarization of the STED beams.

2.5: Resolution scaling in STED microscopy

As shown in Equation 1.3 the resolution in STED microscopy is proportional to the power of the depletion beam as [22]:

∆ √

(2.1)

where ε is the STED intensity divided by the saturation intensity (I/Isat).

To test the power dependency on the resolution we acquired images on ~40 nm in diameter fluorescent red and crimson beads (F-8793, F-8782, Invitrogen) using different power settings, as shown in Figure 2.5. As seen in both the equation and figure, the resolution quickly improves for relatively low applied STED powers (measured by the power density in the STED focus) and then levels out at higher powers. There is no theoretical lower limit to the resolution, but the required laser intensity to double the resolution quickly expands. In our case for the first STED beam at 710 ± 10 nm, we need a power density of ~0.05 GW/cm2 for the first doubling (240 nm to 120 nm), ~0.3 GW/cm2 for the second (120 nm to 60 nm) and theoretically ~1.2 GW/cm2 for the third (60 nm to 30 nm).

2.5: Resolution scaling in STED microscopy

26

Figure 2.5: Resolution scaling with STED power. (A): 40 nm red and crimson beads were imaged using different STED laser powers. For each set STED power, the size of the beads intensity profiles were measured and then averaged. (B, C): The average size with corresponding error bars given by one standard deviation is plotted against the STED power for the red beads (B) and crimson beads (C) and fitted to the theoretical calculation.

When the highest resolutions are desired, the saturation intensity should also be considered which to a large extent depends on the emission spectra of the dye combined with the selected wavelength of the STED beam. In our case, the dye ATTO647N has almost double the stimulated emission cross section at the selected STED wavelength at 750 ± 10 nm as compared to ATTO594 at 710 ± 10 nm, as shown in Figure 2.1. This greatly increases the efficiency of the STED beam for ATTO647N leading to approximately half the needed intensity to gain the same resolution as for ATTO594. A similar effect can be seen in Figure 2.5 for the crimson- and red-beads, where a resolution of 50 nm is achieved at ~0.47 GW/cm2 at 710 ± 10 nm for the red-beads and at ~0.26 GW/cm2 at 750 ± 10 nm for the crimson-beads.

The current highest achieved STED resolutions have been acquired on fluorescent nanodiamonds having a negatively charged nitrogen vacancy [63]. These nanodiamonds are ideal for STED imaging due to their broad and well separated emission spectrum at high wavelengths (~650-750 nm) and their extreme photostability. The broad and well separated emission spectrum allows for selecting the STED wavelength closer to the emission peak without risking excitation, greatly lowering the saturation intensity. To date, the highest achieved resolution in STED microscopy (2.4 nm resolution with 0.09 nm localization accuracy) was achieved on such nanodiamonds at high STED power combined


27

Figure 2.6: Measured resolution of ATTO594 and ATTO647N. Several intensity profiles were combined of preferably single fluorescent markers attached to the coverglass, some of which are marked with the boxed regions in (A). The resolution can then be approximated by the full width at half maximum (FWHM) value of the combined image of the intensity profiles for ATTO594 (B1) and for ATTO647N (B2), yielding a value of ≤ 40 nm for both dyes (C1, C2).

with a solid immersion objective with a NA = 2.2 [23]. The biggest disadvantage of using nanodiamonds for bioimaging is the difficulty of conjugating these probes to biological proteins. However, research is ongoing in this field and successful bioconjugations have been performed [64] [65], making fluorescent nanodiamonds an interesting future fluorescent marker for STED imaging.

An option to lower the saturation intensity is to intentionally blue shift the STED beam, thus increasing the stimulated emission cross section at the cost of some re-excitation by the STED beam. The excitation effect of the STED beam will compromise the image contrast by adding a confocal “blurring” of the high resolution image. However, this can be compensated by acquiring an additional image with only the blurred profiles originating from the excitation of the STED beam and then subtracting this image from the acquired high resolution image [66].

The achieved resolution of our microscope for the two selected dyes ATTO-594 and ATTO-647N is ≤ 40 nm, as shown in Figure 2.6.

29

Chapter 3

Image processing and analysis of high resolution images

Proteins are the usual objects of study in biological microscopy, having an average size of 200-400 amino acids equaling just a few nanometers [67]. The fine details of protein storage and interaction can therefore not be fully resolved with conventional optical imaging techniques such as wide-field or confocal microscopy, where the resolution is limited to a few hundreds of nanometers. Recent advancements in the development of super resolution imaging techniques such as STED, PALM and STORM, have drastically improved the means to overcome this limitation and are thus providing a wealth of new informational content to the images. Therefore, a substantial part of this thesis involves extracting this type of high resolution informational content. The extraction is performed by applying computational image processing and analysis, where new methods and algorithms need to be designed to cover the spatial mapping and interactions of highly resolved intracellular proteins.

3.1: Image processing, deconvolution

In general, high resolution STED images have a lower signal to noise ratio as compared to their low resolution counterparts. The major reason for this is the effectively smaller excitation volume produced when applying the STED beam, leaving less fluorescent probes active and able to emit. This means that STED

3.1: Image processing, deconvolution

30

images are more susceptible to noise and therefore it is beneficial to perform noise suppressing image processing prior to image analysis.

A common approach to reduce noise in images is to perform low-pass (mean) or median filtering. Median filtering replaces each pixel value by the median of neighboring values while low-pass filtering convolves the image using a smoothing kernel, where each pixel value is averaged by neighboring pixel values. This effectively suppresses high frequency noise but has the drawback of blurring the image, effectively reducing the resolution. For fluorescence microscopy imaging it is generally better to perform deconvolution [68].

In fluorescence microscopy the acquired image can be seen as a blurred version of the true image, convoluted by the PSF of the microscope [69]. In confocal and pulsed based STED microscopy the PSF can be approximated as a Gaussian function [70], [71]:

.

(3.1)

√2 2

where x0 and y0 are the center coordinates of the function and w is the width of the Gaussian where the intensity falls below 1/e2. The FWHM (Full Width at Half Maximum) is the width of the intensity profile located at half the peak intensity and is generally applied to describe the resolution.

By having prior knowledge of the PSF, deconvolution can be performed where the known shape of the PSF is used to calculate a likely model of the real image in an iterative manner [72]. For each iteration, features corresponding to the PSF, likely originating from excited fluorescent molecules, are enhanced while other features not corresponding to the PSF, such as noise, are suppressed. Thus, deconvolution effectively reduces noise while at the same time enhances contrast and slightly improves resolution. However, care has to be taken when selecting the PSF since this type of processing could incur image artefacts; a too small PSF can create artificial features not present in the real sample and a too large PSF can combine separate intensity profiles, effectively blurring the image.

The PSF of the microscope should be measured before any attempt of performing deconvolution. One way to measure the PSF is to prepare a sample containing the desired fluorescent molecules where the labels have a distance

Chapter 3: Image processing and analysis of high resolution images

31

Figure 3.1: (A1): Raw STED image of microtubules stained with ATTO594. (B1): Same image after applied richardsson-lucy deconvolution using a 40 nm PSF and 10 iterations. (A2, B2): Enlargement of boxed regions in the raw and deconvoluted image where the noise is visibly suppressed by the algorithm.

larger- and a size smaller - than the resolution of the microscope. By measuring and averaging several of the resulting emission profiles from such a sample, a good estimate of the microscope PSF can be achieved. An example of deconvolution of STED images is shown in Figure 3.1.

3.2: Size, density and location calculations

A very interesting feature when applying high resolution fluorescence imaging is that individual proteins or protein complexes start to become revealed. For instance, a single cell adhesion particle is measured by electron microscopy to be approximately 25 ± 5 nm in size [73]. Since this size is close to the resolution of our microscope, we can individually reveal all but the most densely packed particles. The densely packed particles will in turn give rise to broader intensity profiles. This provides new possibilities to extract information from the images, where the size, density and location of the labeled targets can be of interest.

To extract such data, the individual intensity profiles originating from the labeled targets first have to be identified and separated. A simple way of doing this is to go pixel by pixel in the image and for each pixel identify whether or not it has neighboring pixels of a higher intensity value. If it has, move to the next pixel, if not, save the pixel position since this pixel is a likely intensity peak. After moving through the entire image, the positions of all intensity peaks should be stored.

3.2: Size, density and location calculations

32

However, the majority of these peaks will be due to noise and therefore a threshold has to be set to exclude these peaks from the analysis.

Thresholding is commonly used in image analysis, to separate interesting features (foreground) from background data. However, the setting of the thresholding can have a large impact on the final analysis, and care should be taken to avoid bias. There are several ways to calculate thresholds automatically, where one possibility is to use the Otsu method which calculates the threshold by minimizing the intra-class variance in the foreground and background data [74]. Although these methods can work for conventional microscopy images [75], they generally fail to work satisfactory with the fine details provided by super resolution images. For this reason, we routinely set the threshold manually. For this, it is important to acquire images using the same parameters (sample preparation, labeling, excitation intensity, pixel size and dwell time), and to set the same threshold value for all compared images.

When the threshold is set, the size of the remaining intensity profiles can be calculated. The size of an intensity profile is commonly set by the profiles FWHM. We therefore performed the size analysis by starting at the peak intensity pixel of each profile and then going one pixel at a time until the pixel intensity value was below half maximum. This was performed in both direction and averaged over 20 different angles to provide a good average for each profile.

The density was measured by summing the amount of identified profiles and then dividing it by the size of the studied cell area. The cell area is preferably acquired by cytoskeletal actin staining of the cell, where filamentous actin to a large extent forms the morphology of the cell [76]. If actin cannot be imaged at the same time as the studied proteins, an approximation of the cell area can be made where the locations of the intensity profiles are connected, much like a “connect the dots” puzzle. This approximation is generally adequate if the studied target is a membrane protein. One way of connecting the profiles is to expand their size using a structure element, for instance an octagon of varying size, until they link with each other. The approximated cell area can then be attained by filling the area spanned by the connected profiles and then shrinking the total area using the same structure element.

The location of the labeled targets in respect to the cell area can now also be acquired. By dividing the cell area into k separate concentric zones of equal size, a


33

Figure 3.2: Analysis of size, density and location of stored proteins in a platelet. (A): Deconvoluted STED image of the pro-angiogenic protein VEGF (Vascular Endothelial Growth Factor) and filamentous actin. (B): Located VEGF intensity profiles for which the number and size is calculated. (C): The F-actin image is used to map the size and shape of the platelet in order to calculate density and location of stored proteins. For this particular platelet the number of VEGF profiles were determined to be 26 or 4/µm2, the average size 60 ± 19 nm and the amount in each zone 35/19/15/23/8% respectively where the first is the innermost and the last the outermost zone.

probability distribution function, P(r), can be made from the number of targets located in each zone where:

∑ 1 (3.2)

This type of analysis will provide information regarding how peripherally located the targets are, which in turn can provide insight into storage and release mechanisms of cells.

A size and density analysis was performed in paper I, II, IV, VIII and XI, and an additional location analysis was performed in IV and VIII. An example of the analysis performed on the image of a blood platelet is shown in Figure 3.2.

3.3: Colocalization and nearest neighbor algorithms

Colocalization algorithms give an indication to the interaction between targets by measuring their intensity overlap, where the intensity overlap is dependent on how spatially close the targets are located. However, since the resolution of the microscope effectively acts as a blurring convolution of the target image, the resolution should also impact the intensity overlap of the targets. Further, other


34

parameters such as density, brightness, labeling size and noise could also influence the intensity overlap. These types of parameters therefore need to be taken into account when comparing colocalization data, both when comparing to others published results but also when comparing between obtained data sets and positive/negative controls.

In paper IX we set out to test how these parameters influence the final colocalization values of the commonly used methods Pearson, Manders, M1 & M2 as well as the more advanced ICCS. Explanations and formulas for the methods can be found in Chapter 1.3, or more detailed in paper IX. The tests were mainly performed by simulating a number of “red” and “green” targets inside a defined space, where the different parameters (resolution, density etc.) could be changed independently of each other. We could then monitor the separate impact of each parameter to the resulting colocalization value, where resolution and density simulations are shown in Figure 3.3A.

The results show clear differences between the studied colocalization methods where Pearson and ICCS are much more robust and insensitive to the resolution and density as compared to Manders and M1 & M2. However, all methods are to some extent affected especially at low resolutions and high densities where also Pearson and ICCS tend to get increased values combined with higher deviations, as shown in Figure 3.3B. The impact of these effects can be lowered by applying an iteratively based thresholding, much like the Costes method [37], where the final threshold is set so that the excluded pixels have a Pearson value lower than 0 (showing no correlation). The modification compared to Costes lies in the iterative steps for setting the threshold, which instead of being set to fixed intensity steps are set as percentile steps based on the number of pixels. This provides a less linear thresholding better suited for intensity differences between the red and green channel, a comparison can be found in supplementary material of paper IX.

A higher resolution generally provides more accurate colocalization values, as seen in Figure 3.3B. However, when the resolution starts to reach the size of the labeling molecules, the intensity overlap will be gone even though the targets are in close proximity. At the achieved resolution of 40 nm of our STED microscope, the rather bulky labeling of primary and secondary antibodies (each ~8.5 nm in length [77]) will affect the overlap, discussed more in paper VI and IX. However,


35

Figure 3.3: (A1): Simulated Images corresponding to a 2.5x2.5 µm2 area with 16 particles/µm2. The particles were randomly placed with no consideration to each other representing no interaction. The resolution of the particles were changed by convolution with a Gaussian PSF of varying size. (A2): Simulation of density effects where the resolution is kept constant at 80 nm. (B): Results from resolution and density effects on the different methods M1&M2, Pearson and ICCS, 10 images were simulated and analyzed for each resolution and density. All methods are affected at worse (> 200 nm) resolutions with slightly increased average values and higher deviations, although the effects are much lower for ICCS and Pearson as compared to M1&M2. Similar effects are seen for increased densities, where the combination of high density and low resolution provides the largest error.

since the individual targets now start to be revealed, the proximity can instead be measured by the actual distance of the targets. This is commonly performed by a nearest neighbor analysis where the position of each target, located by the same procedure as described in Chapter 3.2, is matched to the position of other targets to find the closest match. This will also be affected by the labeling size, but rather than lowering the resulting value as in colocalization, it will broaden the distance distribution while the average distance should remain approximately the same. However, the best option would be to label the targets using smaller probes, such as affibodies [78], or intrinsically by the use of fluorescent proteins.


36

Nearest neighbor analysis is also useful when studying targets which do not have a direct interaction but rather communicate through a signaling pathway where they only need to be in close proximity. A high resolution colocalization analysis would be blind to such interaction since the targets are likely to be separately resolved. On the other hand, nearest neighbor analyses are very dependent on the set threshold and density of the labeled targets. Colocalization methods such as Pearson and ICCS are affected to a much less extent by such effects.

The different methods are complementary to each other and should be separately applied depending on the circumstances. We performed colocalization analysis in paper VI, VIII and IX and nearest neighbor analysis in paper I, II, VI, VII and XII.

3.4: Structural analysis, fiber entanglement and orientation

Common targets for demonstrating super resolution fluorescence imaging are cytoskeletal filaments such as vimentin [79] [80], tubulin [80] [81] and actin [82] [83]. A reason for this is that individual filaments start to become revealed, greatly increasing the visual appreciation of the images. Although actin has been used to study the size and morphological changes of for instance dendritic spines [82], little analysis has been performed on these highly resolved cytoskeletal filaments.

Two features which can be analyzed and extracted from highly resolved cytoskeletal filaments are their relative direction and entanglement. Both of these features can give information regarding the organization of these filaments in cells and provide a deeper understanding of the underlying mechanisms involved in cell mobility, as discussed more in paper III, but it can also serve as a potential diagnostic tool, as discussed more in Chapter 4 and paper II.

The direction of the filaments can be extracted by converting the standard X-Y image into an angle-bin image (θ-r) using radon transformation [84]. Each pixel value in the X-Y image will then be projected for all angles θ and summed into separate bins r. This means that a structure oriented in a certain angle in the X-Y image will be projected to a single bin in the angle-bin image, when the angle corresponds to the structure angle, as shown in Figure 3.4. For other angles, the pixel values will be projected to several bins giving lower and more spread values.


37

Figure 3.4: Example of the Radon transformation. Structures in the X-Y image are projected in different angles (θ), where the angle corresponding to the structure direction will result in a single high value in the corresponding angle-bin (θ-r) image. Angles not corresponding to the structure direction will result in lower and more spread values.

Figure 3.5: Example X-Y images (A1, B1, C1) and corresponding angle-bin (θ-r) images (A2, B2, C2). By taking the cube of the pixel values in the angle-bin image and performing a summation for all pixels belonging to the same angle, a graph displaying the most common structure orientation is obtained (A3, B3, C3). A single value can be extracted by selecting a region of the graph, for instance the peak angle ± 10o, and comparing it to the whole span.


38

This means that the angles having the highest pixel values in the angle-bin image should correspond to the most common structure orientation. By extracting these angles, we can in this case gain information regarding the most common direction of the cytoskeletal filaments, but also to what extent the filaments are oriented in this direction. We found that a simple way of extracting these values is to first take each pixel value in the angle-bin image to the power of three, greatly enhancing the high pixel values corresponding to the structure orientation, and then perform a summation of the pixels for each angle. This will create a one-dimensional graph over the angles and their corresponding summed pixel values, as shown in Figure 3.5, where the highest peak corresponds to the most frequent direction. However, when comparing the extent of fibers oriented in the same direction between cells, it is generally of benefit to compare single values rather than graphs. Therefore, as a second step we performed a simplification where we defined all fibers oriented in the most frequent direction ± 10o to be parallel. The amount of parallel fibers can then be compared to the total amount of fibers in the image, giving a value ranging from 1 if all fibers are oriented within ± 10o to 0.12 where all directions are equally probable (as the peak angle ± 10o will then give a value 21/180 = 0.12).

In order to test this type of analysis we performed it on several simulated images, some of which are shown in Figure 3.5. This form of analysis worked well on the simulated images and could to a large extent provide accurate numbers regarding the direction of the structures. However, this should be seen as an initial test to extract the direction of highly resolved structures and there are likely more optimal methods and algorithms which can be developed.

The width of a single vimentin filament is approximately 10 nm in diameter [85]. Since the resolution of the STED microscope used in this thesis is ~40 nm, tightly entangled filaments will not be separately resolved but instead appears as single, wider, fibers. This provides an opportunity to measure the entanglement of the vimentin fibers by looking at the width of their separate intensity profiles. The same holds true for confocal images where the resolution is ~250 nm. However, the difference in apparent widths between single fibers and entangled fibers is much larger in the high-resolution STED images as compared to with the lower resolution confocal images, as shown in Figure 3.6. Consequently, the high resolution STED images provide a far better sensitivity in terms of determining the fiber entanglement. To measure the width of the fibers, a map of the filament


39

Figure 3.6: Fiber width measurements on deconvoluted confocal and STED images. The width of a single fiber (a) was measured to be 47 nm in the STED image increasing to 94 nm where four fibers come together (b, + 100%), the corresponding increase in the confocal image is 161 nm for the single fiber and 178 nm (+ 11 %) for the combined fibers.

Figure 3.7: Width measurement on single and entangled fibers. Two areas are selected showing single fibers (A1) and entangled fibers (B1) respectively. A skeleton of the fiber structure is created (A2, B2) and the width is calculated at several places along this structure (A3, B3, lines showing selected directions). (C): A histogram of the measured widths showing large separation between the single fibers (black) and the entangled (red).

network can be created for instance in MATLAB (MathWorks, Massachusetts) by the watershed or skeleton and pruning functions. The width of the fibers can then be calculated by moving along this network and measuring the width as the


40

minimum FWHM value calculated for several ( ≥ 20) angles, as shown in Figure 3.7. We generally measured the width at more than 1000 different positions along the filament network, so that each filament is measured on several places yielding a good average.

A structure analysis of the filament direction and entanglement was performed in paper II and III.

41

Chapter 4

High resolution as a tool for early cancer diagnostics

The survival rate of cancer is significantly improved if a reliable diagnosis can be achieved at an early stage when the lesion is still relatively small [86] [87]. Two common modalities to extract tissue samples needed for such diagnostics are fine needle aspiration (FNA) cytology and core needle biopsy (CNB), where both techniques uses needles to extract the sample.

The needle diameter in FNA is usually 22-25 Gauge or 0.5-0.7 mm in diameter [88] while the CNB needle is commonly 14 Gauge or ~2.1 mm in diameter [89]. The larger needle diameter provided by CNB makes it possible to extract tissue samples where a high diagnostic sensitivity and specificity can be reached. However, due to the large needle diameter, CNB also displays higher rates of complications including hematoma and higher risk of infections and even cancer cell seeding along the needle tract. In comparison, FNA offers a patient friendly, minimally invasive, rapid and cost efficient diagnostic procedure. But the smaller needle yields little amount of sample with cells taken out of their tissue context. This makes it more difficult to identify and quantify invasiveness, and the FNA technique relies to a large extent on the experience of the operator and cytopathologist.

It is therefore a trade-off between sampling related side effects and diagnostic reliability between the two modalities where several comparative studies have been

4.1: Fluodiamon

42

performed [90] [91] [92]. To circumvent the need to prioritize either diagnostic reliability or minimized sampling-related side effects, it is strongly motivated to develop new highly sensitive methods by which the necessary information for a specific diagnosis can be extracted from a minimal number of cells.

4.1: Fluodiamon

The objective of the Fluodiamon project (funded by European Community’s 7th Framework Programme FP7, grant agreement 201837) was to develop “Ultra-high-resolution and ultra-sensitive fluorescence methods for objective sub-cellular diagnosis of early disease and disease progression in breast and prostate cancer”. Clinical samples from breast and prostate cancer patients should be extracted with minimally invasive techniques and then labeled with newly developed affibodies and fluorophores enhancing the specificity and sensitivity. The samples were then to be imaged either by super resolution or multiparameter [93] [94] microscopy to greatly increase the informational content in the images. As a last step, computerized algorithms were to be introduced which automatically provides a diagnosis based on the analysis of the multiparameter and high resolution images.

For the minimal invasive extraction of clinical samples, a new design of the FNA needle was developed allowing more efficient sampling [95]. Further, radiofrequency (RF) pulses were added to sterilize the needle track and denature tumor cells [96]. The RF pulses effectively prevent tumor cell ceding and lessen bleeding without additional pain to the patient or degradation of the extracted sample.

Work belonging to this thesis involved the construction of the dual color super resolution STED microscope, described in Chapter 2, and STED imaging of the two membrane bound growth receptors IGF1R (Insulin Growth Factor 1 Receptor) [97] and HER2 (Human Epidermal growth factor Receptor 2) [98]. Initial extensive screening tests were performed on cultured cell lines corresponding to normal (CRL-8798), tumorigenic (CRL-2338) and metastatic (HTB-132) cells to optimize sample preparation and staining procedures as well as optimizing imaging parameters and identifying interesting analysis features.

Thereafter clinical FNA samples were investigated, where the cells were attached to coverglasses by use of cytospin prior to fixation using formaldehyde. From each patient, eight samples could be prepared from the relatively small

Chapter 4: High resolution as a tool for early cancer diagnostics

43

Figure 4.1: Confocal image and corresponding high resolution STED image on a clinical sample from a breast cancer patient. The cells were labeled for the membrane bound growth receptors HER1 and IGF1R.

volume produced by the FNA aspirate, two of which was used for IGF1R-HER2 STED measurements. HER2 was labeled using ATTO594 coupled affibodies (provided by KTH Biotech) [99] and IGF1R was labeled with ATTO647N using primary and secondary antibodies.

For each patient and sample, a minimum of 5 super resolution STED images were acquired on separate cells, one of which is shown in Figure 4.1. In total 63 patients were screened among which 27 were determined by an experienced cytopathologist (Prof. Gert Auer, Karolinska Institutet) to be normal or benign and 36 malignant. The analysis was performed by an image analysis expert (Ville Rantannen, Univ. Helsinki) who created an image classifier based on several different features in the images, including morphological shapes and expression levels for single targets but also interaction analysis between the targets.

4.1: Fluodiamon

44

Unfortunately the final classifier did not provide very satisfactory results since it only had an accuracy of 73%, meaning that more than one out of four patients would get the wrong diagnosis. This can be compared to the immediate on-site evaluation accuracy of an FNA sample by a skilled cytopathologist, which in an American study involving more than 5000 patients was determined to be 85% [100]. There could be several reasons for the relatively low performance of the classifier, where a likely source of uncertainty lies in the choice of targets. It was recently shown that there is a large heterogeneity in the characteristics of HER-family receptors dependent on the choice of host cell line [101]. It is therefore possible that the patient to patient difference of HER2 is larger than the difference between normal and malignant cells. Nonetheless, new modalities for extracting diagnostic information were established and we strongly believe that further optimization regarding especially choice of targets but also labeling, imaging and analysis techniques can provide an accurate diagnostic tool for the future.

4.2: Metastatic cell adhesion and mobility

The metastasis of tumor cells is a multistep process including local tumor cell dissemination or invasion into the blood stream [102], adhesion and extravasation into distant organs [103] and establishment of a vascular network by angiogenesis followed by tumor cell proliferation [104] [105]. A key to this process is the migration of the tumor cells requiring altered adhesive and mechanical properties. Therefore, the proteins that regulate these cellular properties can be expected to have altered spatial distribution patterns in the metastasis competent cells.

Concerning proteins involved in cell adhesive properties, cells in general interact with their surrounding fibrous extracellular matrix via cell adhesion particles, assembled by several different adhesive proteins [106]. On ligation of the cell adhesion particle to the extracellular matrix, a protein signaling cascade is elicited which can be detected as phosphorylation of tyrosine amino acid residues [107].

The cell adhesion particles are determined by electron microscopy to be 25 ± 5 nm in diameter [73], making areas of increased adhesion particle densities appear as larger homogenous adhesion areas, denoted as cell matrix adhesion complexes, at the relatively low resolution of confocal laser scanning microscopes [108].


45

Figure 4.2 STED image of cell adhesion particles and filamentous actin. Areas of increased adhesion particle densities are seen as large almost homogenous intensity areas in confocal microscopy, while the individual particles can be revealed with high resolution STED microscopy.

However, the same areas imaged with super resolution STED microscopy reveals all but the closest adhesion particles, as shown in Figure 4.2. Thus, this resolution improvement provides additional details regarding the underlying structure and organization of cell adhesions but also additional means of discerning cell to cell variations, where the latter could be used as a diagnostic tool separating normal from metastatic cells.

Rho GTPases are known to control the processes of cellular adhesion by regulating actin polymerization, actin bundling and acto-myosin-based contraction [109]. Further, Rho GTPases are believed to play an important role in metastasis,


46

where different Rho GTPases are involved in different modes of tumor cell movement [110] [111].

In paper I, we set out to study the effect of different Rho GTPases in terms of redistribution of adhesion particles in combination with the exerted contractile force of the cells, measured by STED and traction force microscopy [112] [113] respectively. Three different constitutively active variants of Rho GTPases where selected; RhoA (RhoAV14), Rac (Rac1L61) and RhoD (RhoDV26), and separately expressed in mouse NIH3T3 fibroblast cells. Traction force data showed markedly increased contractile force for cells expressing any of the three selected GTPases but particularly for RhoA expressing cells, as shown in Figure 4.3.

When the cell adhesion areas in combination with filamentous actin where studied with super resolution STED microscopy, a clear redistribution of the adhesion particles could be detected, as shown in Figure 4.4A. For the control cells, the adhesion particles were mainly located in cell matrix adhesion complexes at the end of thick actin stress fibers. The RhoA expressing cells showed a clear loss of these thick actin stress fibers and cell matrix adhesion complexes and instead showed a much more homogenous distribution of adhesion particles in combination with many thin actin fibers. A similar more homogenous pattern was seen for Rac expressing cells, however here a higher density of adhesion particles could be seen at the peripheral border of the cell. The RhoD expressing cells showed cell matrix adhesion complexes similar to those of control cells, however these were generally of smaller size and less distinct where instead a higher density of adhesion particles was found outside of these areas.

The adhesion images were analyzed in terms of density and nearest neighbor distances, where the procedure is described in Chapter 3.2 and the results are shown in Figure 4.4B. The density analysis showed similar values for the control, Rac and RhoD cells but a clear increase for RhoA cells. As compared to control cells, the nearest neighbor analysis showed a decreased adhesion particle distance for RhoA cells and increased distances for RhoD and especially Rac cells. The decreased distance for RhoA can be explained by the increased density while the increased distance of RhoD and Rac indicate the more homogenous distribution where the adhesion particles are no longer pact into dense cell matrix adhesion complexes.


47

Figure 4.3: Traction Force measurements on Control, Rac, RhoA and RhoD transfected cells. Stars above error-bars in the histogram indicate significant difference from control cells by the student t-test of unequal variance (* P ≤ 0,05, ** P ≤ 0,01). Scale bar: 10 µm.

Figure 4.4: (A): STED images of adhesion particles and filamentous actin on Control, Rac, RhoA and RhoD cells. (B1): Measured adhesion particle density. (B2): Measured nearest neighbor distances, (** P ≤ 0,01, *** P ≤ 0,001). Scale bar: 1 µm.


48

When compared to the traction force results, the redistribution of adhesion particles from dense actin stress fiber-coupled cell matrix adhesion complexes to several more homogenously spread thin actin fiber-coupled adhesion areas, seem to increase the contractile force of cells. Thus our data supports the notion that the contraction of thin acto-myosin fibers connected to small adhesion-linked particles can function in a relatively autonomous manner [114].

The cell cytoskeleton, composed of actin filaments, microtubules and intermediate filaments (IFs), is believed to govern the mechanical properties of cells and is thus highly involved in cell mobility [115]. In particular, the actin filaments have been studied and shown to control cell mechanics [116] [117]. However, recent studies have also revealed novel functions for the intermediate filament protein vimentin related to the migration of cells [118]. Linking vimentin to properties such as determination of cellular polarity, regulation of adhesion areas and arrangement and transport of signal proteins involved in cell mobility.

In general, the cytoskeletal arrangement largely determines the biomechanical properties of cancer cells, which from recent biophysical investigations have been found to strongly correlate with invasiveness [119] [120]. The vimentin intermediate filament system has been found to enhance tumor cell metastasis in vivo by controlling cell adhesion and motility [121] [122]. Taken together, this suggests that the metastatic process is linked to altered spatial organization of proteins that regulate the adhesive and mechanical properties of cells and their environment. In particular, the organization of cell adhesions and vimentin filaments seem to play a prominent role in metastatic cell mobility.

In paper II we applied STED microscopy to study potential differences in the cell adhesion particles as well as in the vimentin intermediate filament system of cultured normal human skin fibroblasts cells, BJ, and their genetically modified, invasive and metastasizing counterpart, BJHtertSV40TH-RasV12 [123]. Regarding the adhesion particles, we analyzed the images in terms of nearest neighbor distances, size and density of their respective intensity profiles, by the same procedure as in paper I and as explained in Chapter 3.2.

We further extracted the cell matrix adhesion complex areas, in a similar way as the “connect the dots” cell area extraction explained in Chapter 3.2, in order to detect differences in their size and number. A separate analysis of nearest neighbor distance, size and density of adhesion particles was also performed inside


49

Figure 4.5: Scatterplots of adhesion results in control and metastatic cells. (A): Adhesion particle size and density. (B): Size and number of cell matrix adhesion complexes. (C): Density of adhesion particles inside and outside of cell matrix adhesion complexes. Larger symbols represent average values with standard deviations.

and outside of these cell matrix adhesion areas. This type of analysis showed distinct differences between the normal and metastasizing cells, where an overall increase in density and decrease in size of both the adhesion particles and cell matrix adhesion areas was seen for the metastasizing cells, scatterplots of average values of each separate cell is shown in Figure 4.5. By separately analyzing the density inside and outside the cell matrix adhesion complexes, the metastatic cells showed a significant increase outside these areas and a slight reduction inside. If compared to the results obtained from the Rho GTPase study in paper I, this redistribution of adhesion particles from the cell matrix adhesion complexes to a more homogenous cell distribution, indicate higher cell contractility for the metastatic cells.

A separate analysis of nearest neighbor distance, size and density of adhesion particles was also performed inside and outside of these cell matrix adhesion areas. This type of analysis showed distinct differences between the normal and metastasizing cells, where an overall increase in density and decrease in size of both the adhesion particles and cell matrix adhesion areas was seen for the metastasizing cells, scatterplots of average values of each separate cell is shown in Figure 4.5. By separately analyzing the density inside and outside the cell matrix adhesion complexes, the metastatic cells showed a significant increase outside these areas and a slight reduction inside. If compared to the results obtained from the Rho GTPase study in paper I, this redistribution of adhesion particles from the cell matrix adhesion complexes to a more homogenous cell distribution, indicate higher cell contractility for the metastatic cells.


50

For the vimentin intermediate filaments, the images were analyzed regarding direction and entanglement; measured by the ratio of the peak angle ± 10o and width of the fibers respectively, as explained in Chapter 3.4. By this procedure, the filaments were distinguished to be less parallel and more entangled in the metastatic cells as compared to control cells. By combining the direction and entanglement data for each cell separately in a scatter plot, a clear separation can be seen, as shown in Figure 4.6. Hence, the normal and metastasizing cells could be distinguished with regard to the spatial organization of the vimentin intermediate filamentous system in the cells. Whereas the control cell often displayed a well-structured network where most of the filaments were oriented as single filaments in similar direction, the filaments were oriented in a more disordered fashion in the metastatic cells, with less parallel filaments and higher entanglement.

Recent studies have shown that the recruitment of vimentin filaments to sites of cell adhesion promotes adhesive strength and stability [124]. Further, reduced cell adhesion has been shown to give increased focal adhesion turn-over, which can be linked to increased cell motility [125]. Taken together with our findings of the adhesion particles and vimentin organization, this suggests that the more disorganized vimentin structure combined with the more homogenous adhesive particle distribution provides the metastatic cells with more contractile and flexible mechanical properties, which in turn increases cancer cell motility and invasiveness. Although there is a large heterogeneity between cells belonging to the same cell type, regarding both adhesion and vimentin organization as seen in the scatterplots of Figure 4.5 and 4.6, the difference between metastatic and control cells is still significant when averaging between a relatively low number of cells (where we measured 15-20 for each cell type).

In paper II we therefore conclude that high resolution imaging in combination with image analysis of cell adhesion particles and vimentin intermediate filaments can provide novel means to identify metastasizing cells. Further, this type of study can give new insights into the role of these proteins in the metastatic process.

In paper III we wanted to further study how the vimentin intermediate filaments changed during different oncogene expressions, where oncogenes are known to deregulate cellular functions which can lead to tumor cell invasion and the formation of metastases [126] [127]. The work related to this thesis was to


51

Figure 4.6: Vimentin intermediate filaments in control and metastatic cells. Metastatic cells often showed a less parallel and more entangled vimentin network as compared to control cells, measured by the direction and width of the fibers respectively. Stars above error-bars in the histogram indicate significant difference from control cells (* P ≤ 0,05, ** P ≤ 0,01).

perform a vimentin filament entanglement analysis, in the same way as in paper II, between normal and oncogene expressing cells. This analysis showed a more entangled vimentin filament organization for the oncogene expressing cells, indicating an increased cross-linking of the filaments. Further, by inhibiting histone deacetylase 6 (HDAC6), which increases the deacytelated form of microtubules [128], the oncogene effect on the vimentin reorganization was suppressed. This suggests that HDAC6 and microtubule deacetylation are needed for the oncogene induced reorganization of the vimentin filament network. The


52

vimentin organization was further compared to AFM measurements of the cell stiffness and to cell invasion assays, where an increased cellular stiffness as well as invasiveness was seen for the oncogene expressing cells. This suggests that the oncogene induced reorganization of the vimentin filament network via HDAC6 contributes to cellular stiffness and tumor cell invasion.

4.3: Platelets as a minimally invasive diagnostic target

Platelets, or thrombocytes, are small disc shaped cells with a diameter of 2-4 µm2 which circulate our bloodstream. Unlike most cells in our body, platelets do not have a nucleus and can therefore not reproduce themselves by normal cell division and are instead produced by megakaryocytes located in the bone marrow.

Although platelets are most renowned for their capability to form blood clots at the location of a wound [129], they are involved in many more processes where they have been shown to specifically store, sequester and release more than thousand different molecules and proteins [130] [131]. From a cancer diagnostic perspective, patients with newly diagnosed metastatic diseases have shown both increased and decreased platelet content of specific proteins, in particular angiogenesis regulating proteins involved in neovasculature development [132] [133].

Current diagnostic technique require direct tissue sampling from the suspected tumor area, which could lead to several complications as described in the beginning of this chapter. Imagine instead that platelets are the diagnostic target, since they circulate our bloodstream they provide a truly minimally invasive diagnostic tool where a simple blood test could suffice. Further, medication inhibiting platelet activation, such as aspirin, has been shown to reduce the incidence and metastasis of in particular colorectal cancer [134] [135]. Thus, improved means for characterization of platelets may not only provide a minimally invasive diagnostic tool but also provide a deeper understanding of their underlying mechanisms which could provide a basis for various treatment regimes.

To date, research involving platelet imaging has mainly been conducted using confocal laser scanning microscopy (CLSM) or electron microscopy (EM). The fluorescence based readouts of CLSM can provide high specificity in respect to target labeling combined with a high imaging sensitivity and throughput.


53

Figure 4.7: STED images of VEGF, PF-4 and Fibrinogen (green) together with filamentous actin (red) in control platelets as well as platelets activated either by Thrombin or ADP. The number in the bottom left corner represents the amount of platelets which were imaged and analyzed for each category. Scale bars 1 µm.

However, due to the small size of platelets in combination with dense storage of proteins, the 200-300 nm resolution provided by CLSM is not adequate to resolve the distribution patterns of the studied proteins [136]. For this reason, EM is often applied which provides more than necessary resolution, usually ≤ 1 nm. However, the sample preparation in EM is more cumbersome, the examined sections generally do not encompass whole organelles and the degree of labeling is lower compared to fluorescence based microscopy [137] [138], see paper IV introduction for more details.

In paper IV, we sought to overcome the limitations of CLSM and EM by applying super resolution STED microscopy to study platelets. We further wanted to test the diagnostic possibility by imaging the pro- and anti-angiogenic proteins VEGF and PF-4 together with fibrinogen, a protein highly involved in blood


54

clotting, and filamentous actin in un-activated (control) platelets compared to platelets activated by either thrombin or ADP, shown in Figure 4.7.

The high resolution images were analyzed in terms of size, number and location of the proteins respective fluorescence profiles and clear and specific changes could be monitored for the different proteins under different platelet activations, as shown in Figure 4.8. For ADP treated platelets, the pro-angiogenic protein VEGF showed a clear redistribution to the peripheral areas of the platelet combined with a reduced number and size of intensity profiles, strongly indicating release. For the anti-angiogenic protein PF-4, an opposite effect is seen in terms of number of profiles, where the increased amount rather indicates a sequestering of proteins from the blood plasma. This effect of release of pro-angiogenic VEGF and sequestering of anti-angiogenic PF-4, suggest a clear promotion of blood vessel formation at sites of high ADP concentration.

Fibrinogen showed a very clear redistribution upon ADP activation, where the majority of proteins was reallocated to the peripheral areas. This corresponds well to other studies where the fibrinogen has been found to bind to receptors on the platelet surface upon ADP stimulation, promoting platelet aggregation [139]. For thrombin treatment, VEGF seemed little affected with size, number and location parameters similar to that of control platelets. PF-4 showed a clear redistribution of proteins to the peripheral areas followed by lower number and size values, indicating release. For fibrinogen, the reduced number and size values were even more pronounced, where little remained of the originally abundant amount found in control platelets. Thrombin is known to be released at the location of a wound where the released fibrinogen forms fibrin structures required for blood clotting [129].

In the above study, we have activated the platelets by different agonists and monitored how these activations changes the protein distributions of VEGF, PF-4 and fibrinogen. However, from a diagnostic viewpoint it is more interesting if we can do the opposite; identify the type of activation by looking at the individual protein distributions. In order to do this, we compared the number, size and location values of each individual platelet to the corresponding averaged values of the different activation states (control, ADP and thrombin) and assigned each


55

Figure 4.8: Analysis results of platelet images. (A1, B1, C1): Scatterplots of individual control (black squares), thrombin (blue circles) and ADP (red triangles) platelets measured for size and number of located VEGF (A), PF-4 (B) and Fibrinogen (C) intensity profiles. (A2, B2, C2): Average values for the three different blood donors as well as the total average with error bars. (A3, B3, C3): Spatial mapping of the protein distribution within the platelets, darker areas corresponds to higher probabilities of locating proteins in these zones. (D1): Two-Tailed student T-test of unequal variance showing significant differences for thrombin (blue crosses) or ADP (red crosses) activated platelets as compared to control (x P ≤ 0,05, xx P ≤0,01, xxx P ≤ 0,001).

value a state based on the closest match. For the location distribution, P(r), this was performed by a least square fit of the five separate concentric zones. If two or more of the number, size and location assignments were to the same state, the platelet was assigned to that state. This provided an assignment probability matrix where the probability of distinguishing the type of activation, based on the spatial protein distribution of the separate proteins, is shown Table 4.1. As can be seen in the table, the correct assignments (gray boxes) received the highest probability for


56

Table 4.1 Assignment probability matrix. The probability of assigning a platelet with a state given by the vertical position of the matrix, to a state indicated by the horizontal position. The three gray colored diagonal elements in each matrix represent a correct assignment. The off-diagonal elements represent the fraction of misclassified platelets and the right column shows the fraction of unclassified platelets were the platelets received three separate state assignments.

all targeted proteins and states. Labeled VEGF could distinguish ADP activated platelets with 91% probability, while labeled PF-4 and fibrinogen could distinguish thrombin activation by 73% and 85% respectively.

Although these values are not considered high from a diagnostic perspective, it should be noted that this is the probability to distinguish a state based on the image and analysis of one single platelet and one single target. By investigating several platelets combined with multiple targets, the identification strength should be significantly increased. In paper IV we imaged and analyzed close to 400 platelets, where the imaging was performed by our home built high resolution but relatively slow STED microscope. Today commercial STED instruments start to offer the same resolution capabilities with faster acquisition time, making it possible to get an even higher throughput. Combined with the possibility to image multiple targets in the same platelet, as discussed more in Chapter 6 and paper VIII, we believe that high resolution multicolor imaging of platelets can offer very interesting and minimally invasive diagnostic possibilities in the near future.

57

Chapter 5

Dissecting synaptic proteins in dendritic spines

The human nervous system consists of almost one hundred billion neurons where each neuron communicates through approximately a thousand different specialized cell junctions, called synapses [140]. In order to understand such a vastly complex system it is generally beneficial to start on a smaller scale, where part of this thesis is aimed to elucidate different proteins involved in the signaling pathway of postsynaptic dendritic spines.

The dendritic spines are the receiving end of synapses and can be seen as tiny protrusions usually extending 0.5-2 µm from the main shaft of dendrites [141]. During a signaling cascade, neurotransmitters, such as glutamate or dopamine, are released from the presynaptic neuron and binds to different receptor in the postsynaptic dendritic spine. This opens various kinds of ion channels which can either act excitatory, increasing the local spine membrane potential by opening sodium ion (Na+) channels, or inhibitory, decreasing the potential by opening potassium ion (K+) channels.

Generally several dendritic spines need to provide an excitatory signal in order to provoke an action potential from the neuron, which in turn can be blocked by inhibitory signals. In order to reset the ion gradient after excitatory neural signaling, sodium potassium pumps (or NKA, Na+K+-ATPase) are located in the dendrite membrane, transporting three Na+ ions out of and two K+ ions into the

58

dendrite volume for each pumping cycle [142]. The reset of ion gradients is very important for neural signaling and deficiencies in their function have been linked to many behavioral effects such as altered learning, mobility and anxiety as well as diseases such as Parkinson, Alzheimer and migraine [143] [144].

In order to gain information regarding NKA topological organization in dendritic spines, we applied STED microscopy to resolve the proteins spatial distribution in cultured striatum neurons. In paper V, we show that the high resolution STED images identify discernible pools of NKA in the spine-heads and spine-necks as well as within the connecting dendritic structures, as shown in Figure 5.1. Corresponding confocal images lack the needed resolution to show such compartmentalization and instead give the notion of a more homogenous distribution within the spine.

The resolved discontinuous distribution of NKA in spines can have implications for the structural and functional interaction between NKA and other synaptic proteins as well as for the generation of local sodium gradients within the spine. The discerned NKA pools located in the spine heads were determined to be on average 58 ± 11 nm. Given the physical size of NKA, size 65 × 75 × 150 Å [145], this gives an upper limit of 20-30 NKA proteins per pool. This value is however affected by the antibody labeling both in terms of steric hindrance, which can limits the amount of labeled NKA and thus lead to an underestimate of the number of proteins, and antibody size, which can broaden the intensity profile and thus lead to an overestimate. By the use of STED microscopy, it is possible to estimate that roughly around 100 NKA proteins are located in each dendritic spine. This number has implications regarding how fast the system can be reset under physiological conditions.

The activity of NKA can be regulated by ligand binding and by phosphorylation/dephosphorylation processes. One such process is mediated by the dopamine 1 receptor (D1R) which can inhibit the pump activity of NKA and thus regulate neuronal excitability [146]. In paper VI we set out to investigate how D1R was spatially located in relation to NKA by applying high resolution imaging and image analysis. The resulting STED images show a compartmentalization also for D1R where the receptors often were positioned as single pools located in the spine-head and close to the base of the spine-neck, corresponding NKA pools were more frequently occurring, as shown in Figure 5.2A. As can be seen in the

Chapter 5: Dissecting synaptic proteins in dendritic spines

59

Figure 5.1: Confocal and STED images of NKA in dendritic spines. The dendritic spines can be seen as the short (0.5-2 µm) protrusions from the dendritic structure. STED images reveal discernible pools of NKA in the head and neck of the spine as well as in the dendritic structure.

Figure 5.2: (A): STED images of dendritic spines labeled for NKA (green) and D1R (red). (B): Locations of peak intensity profiles. (C1, C2): Nearest neighbor histograms of D1R-NKA and D1R-D1R distances inside dendritic spines.

60

STED images the NKA and D1R clusters were generally not overlapping in the spines, indicating that they are not physically bound to each other in these areas. However, since the inhibitory regulation of D1R to NKA is not direct, but linked by activation of protein kinases [147], this does not prevent their interaction.

Since there is little overlap of the intensity signals, colocalization analyses of the images (as explained in Chapter 1.3 and 3.3) will not provide much information. Instead we performed a nearest neighbor algorithm to study the distances of the NKA and D1R pools, as shown in Figure 5.2B. This analysis provided histograms of the found distances, shown in Figure 5.2C, were the distance between D1R and NKA was determined to be 80 ± 60 nm on average as compared to D1R-D1R distance of 160 ± 60 nm. Thus the D1R pools are generally located much closer to NKA pools than other D1R pools, giving the notion of micro-regulatory domains where each D1R cluster adjusts the activity of nearby NKA proteins.

In paper VII we performed a similar study on the dopamine- and cAMP-regulated phosphoprotein Mr 32 kDa (DARPP-32) in combination with the synaptic scaffolding protein PSD-95 and D1R, where DARPP-32 is known to be important in the dopaminergic signaling pathway [148] [149]. Nearest neighbor analysis revealed a close proximity between DARPP-32 and D1R with an average distance of 70 ± 40 nm. Further, DARPP-32 was found to be located in several small pools close to the resolution limit of the microscope and thus corresponding to only a few proteins in each pool. The relatively low number of Darpp-32 suggests that only a few copies are necessary to regulate neuronal signaling.

61

Chapter 6

Multicolor fluorescence nanoscopy by photobleaching

More targets mean more information. However, two targets do not necessarily mean twice the information as compared to one target since new options arise where the correlation or interaction of the targets can be measured. Thus increasing the amount of differently labeled targets in the images greatly increases the informational content. As stated earlier, the high resolution images also contain much more information than their lower resolution counterparts. Therefore, the topic of this chapter will be the combination of high resolution imaging with multiple target imaging in order to provide as much informational content in the images as possible.

6.1: Problems involved in combining multicolor and STED microscopy

The STED microscope used in this thesis, described in Chapter 2.1, has separate wavelengths and beam paths for each excitation, emission and STED beam. Since we can image two dyes simultaneously this means we have to be able to handle six different beams. Due to the high resolution of the microscope, these six beams have to be handled with high precision since a drift of only tens of nanometers can distort the resulting image. Further, specialized optical components are needed to separately handle the different wavelengths with minimal intensity

6.1: Problems involved in combining multicolor and STED microscopy

62

losses. This is especially important for the STED and emission beams since the resolution is proportional to the STED power and the resulting high resolution images generally suffer from lower signal intensity. The latter due to the effectively smaller excitation volume where less fluorescent probes are left active and able to emit. Additionally, for simultaneous imaging of all dyes, the high power STED beams cannot be in the same wavelength span as any of the dyes excitation spectrum, also limiting the choice of wavelengths. Thus, increasing the amount of targets by adding additional excitation, emission and STED beams is difficult.

One solution is to separate additional targets by the use of fluorescence lifetime rather than spectra [150]. There are several fluorophores which have a similar excitation and emission spectra but different lifetimes, making it possible to use the same excitation, STED and emission wavelengths for multiple dyes. Also the same detector can be used with the added requirement of high time resolution. However, separation of different dyes by the use of their fluorescent lifetime generally requires a high number of photons in order to reduce crosstalk between the dyes. This can be challenging in STED microscopy since the emission intensity, as stated above, is generally low. Nonetheless, successful three color STED images have been acquired where two of the dyes where separated by their lifetime with <10 % crosstalk [151].

6.2: A possible solution

In paper VIII we propose a new method to separate fluorescent dyes by exploiting their difference in photostability and excitation spectra. The procedure regarding image acquisition and crosstalk is explained in detail in paper VIII.

Briefly, two dyes (A and B) having similar excitation and emission spectra are used to label separate targets in the sample. A high resolution image is acquired using a single excitation, emission and STED beam resulting in a combined image of the two targets (A + B). Thereafter the difference in photostability and excitation spectra is exploited where a tuned laser is used to excite and photobleach one dye (A) while the other dye (B) remains largely unaffected. A new image is acquired where only the unaffected dye (B) remains. The image of the bleached dye (A) can now be reconstructed by subtracting the image of the unaffected dye (B) to that of the combined image (A+B) as A = (A+B) – B, as shown in figure 6.1.

Chapter 6: Multicolor fluorescence nanoscopy by photobleaching

63

Figure 6.1: Schematic of multicolor acquisition by photobleaching. Two dyes with similar spectral properties are used to label the sample. (A): The same excitation, STED and emission wavelength is used to acquire an image showing both dyes. (B): Separate bleaching of one dye is performed and thereafter a new image is acquired, now only showing the remaining dye. (C): The bleached dye can be reconstructed by subtracting the second acquired image from the first. (D): A merge of the two separated dyes can be made.

This procedure can be applied in combination with other means of separating targets, such as spectral or lifetime separation, effectively increasing the number of possible targets by a factor of two.

The requirement is that we have two dyes with similar excitation and emission spectra but different photostability combined with a suitable bleaching laser. For our microscope we can image the dyes Dylight650 and ATTO647N using the same excitation (647 ± 5 nm), emission (675 ± 15 nm) and STED (750 ± 10 nm) wavelengths, where Dylight650 is much less photostable and has a slightly red-shifted excitation spectrum as compared to ATTO647N. At a wavelength of 710 ± 10 nm we are positioned at the end-tail of the two dyes excitation spectra where the slight red-shift of Dylight650 provides an approximately 250 % higher excitational cross-section as compared to ATTO647N. At this wavelength we can therefore selectively bleach Dylight650 while keeping ATTO647N largely unaffected.

6.2: A possible solution

64

Since we have a dual color microscope with the possibility to separate two targets by their spectral differences, we can apply the above procedure in both channels. For the second channel we can use the dye Alexa594 coupled to either antibodies or phalloidin, where a significant loss in photostability was seen for the phalloidin conjugated dye as well as a slight red-shift in excitation spectra. Here we selected a bleaching wavelength at 647 ± 5 nm where Alexa594-phalloidin had ~65 % higher excitation, for spectra and bleaching data of all four dyes and conjugates see paper VIII supplementary information. We can thus separately bleach and reconstruct one additional labeled target for each of our detection channels and by doing so create a four color image, where a high resolution four color image of a platelet is demonstrated in Figure 6.2.

The selected bleaching wavelengths at 710 ± 10 nm and 647 ± 5 nm where in part selected out of convenience since they already were available on our microscope, where 710 ± 10 nm is used to STED Alexa/ATTO594 and 647 ± 5 nm is used to excite ATTO647N/Dylight650. Thus we could apply the method without having to modify our optical setup. However, adding one or two bleaching lasers is generally not difficult since the beam quality and focusing is of less importance. Adding bleaching lasers also provides the possibility to further fine-tune the bleaching wavelength in order to optimize bleaching ratios between the dyes.

6.3: Applications and limitations

As stated in the beginning of this chapter, more targets mean more information and thus there should be a broad range of applications were more targets are desirable. Linking back to cancer diagnostics in Chapter 4, more targets should improve the diagnostic readout capability further limiting the needed amount of extracted tissue sample. This should allow for less invasive sampling, providing a less painful and more cost efficient procedure which at the same time reduces risk of cancer cell dissemination. Regarding platelets, more targets should allow for better determination of their activation state increasing both diagnostic sensitivity and specificity for specific diseases. It may also provide means for early screening of multiple diseases by targeting proteins involved in different forms of diseases such as VEGF for cancer metastasis, P-Selectin for inflammatory diseases such as atherosclerosis [152], CLEC-2 for human immunodeficiency virus (HIV) [153] and von Willebrand factor (vWf) involved in arterial thrombosis [154].

Chapter 6: Multicolor fluorescence nanoscopy by photobleaching

65

Figure 6.2: Four color STED and corresponding confocal image of platelets, images were acquired by spectral separation in combination with photobleaching of one dye in each channel.

Regarding the neuronal measurements performed in Chapter 5, more targets would allow for a simultaneous monitoring of all studied proteins NKA, D1R and Darpp32. Since these proteins are known to interact it would be interesting to see a detailed spatial mapping of their positions inside the same dendritic spine, also allowing for nearest neighbor analysis with all three proteins present. Since there is room for an additional label one could consider the actin cytoskeleton to gain a specific location of the proteins in terms of the morphological shape of the spine.

Overall, adding several targets to the same sample should reduce the needed amount of samples and thus increase throughput both in terms of sample preparation and imaging. The image analysis performed in this thesis has either handled the targets separately (size, density, fiber entanglement and direction analysis) or two-by-two (colocalization and nearest neighbor analysis). By adding the possibility to label additional targets, an interesting perspective would also be to develop new analytical tools taking benefit of the high resolution multi-target information by correlating data from more than two color channels. Such algorithms could give better insights into the organization and function of more complex cellular mechanisms involving multiple proteins, for instance the interaction of proteins involved in the regulation of neuronal signaling or in the

6.3: Applications and limitations

66

maturation of a cell focal adhesion, overall giving more detailed information in system biology elucidations [155].

Regarding limitations to the technique, live-cell imaging can be problematic since two or more images need to be acquired where the targeted cell should preferentially not move between capture. This can to some extent be remedied by using the stable dye image to compensate for the movement since its emission intensity profiles should be similar before and after bleaching, this however requires a relatively fast image acquisition and a homogenous movement of targets. Also video-rate imaging will in general not be possible since only a single image can be acquired of the bleachable dye. The exception would be if one target is mainly stationary, such as motor proteins moving along a stationary fiber, where the stationary target can be targeted with the bleachable dye.

67

Chapter 7

Conclusion and outlook

If I was to conclude this thesis with a single word, it would be information. If I was to use three additional words they would be content, extraction and application.

A reoccurring statement throughout this thesis is that “super resolution images provide a much higher informational content compared to their lower resolution counterparts”. This informational content is further increased by a new method to perform multicolor STED by exploiting differences in photostability and excitation spectra of fluorescent dyes, described in paper VIII.

The high informational content also needs to be extracted. This can be performed by applying algorithm based image analysis where different types of targets and interactions call for different types of analysis. The extracted information can then be applied to gain new insights into different biological questions. In paper V the size of NKA pools is used to determine the possible number of NKA-pumps located in dendritic spines, providing insights into the regulatory mechanisms of neuronal signaling. This procedure is continued in paper VI and VII with the addition of nearest neighbor distance measurements of regulating proteins D1R and Darpp32, where the intensity distributions of the proteins were located close but not necessarily overlapping with each other, indicating their indirect interactions. In paper I and II, size, nearest neighbor and density analysis is used on adhesion particles to link their spatial distribution to the adhesive properties of the cells, but also to compare and successfully separate

68

between control and metastatic cells. This type of diagnostic applicability is further tested in paper IV, where we describe how high resolution images in combination with size, density and location analysis can to a high extent correctly identify different activational states of blood platelets. Structural analysis of highly resolved cytoskeletal vimentin intermediate filaments is performed in paper II and III where the extracted information regarding the entanglement and direction of the filaments is linked to cancer cell mobility and invasiveness.

However, care also needs to be taken when extracting information in order to make sure all relevant data is included and background effects excluded. In paper IX, some of the pitfalls involved in applying colocalization algorithms are discussed where parameters such as resolution or density can have a major impact on the extracted data.

As an outlook, it would be of high interest to apply multicolor STED microscopy and image analysis on platelets from healthy donors compared to those of patients suffering from different diseases, such as cancer. If sufficient differences are detected, this type of analysis could in the future be implemented in clinical environments and provide a truly minimally invasive diagnostic procedure. This of course requires direct access to a super resolution microscope, which is not unrealistic given the rapid development of super resolution techniques, where many such microscopes are now commercially available.

Another very interesting possibility would be to further link the high resolution spatial information provided by STED microscopy to information provided by other techniques. In paper I, the highly resolved adhesion particle distribution in cells was linked to the cells adhesive force measured by TFM. In paper III, the nanoscale structure of the vimentin filament system were compared to cellular stiffness measured by AFM and metastatic capabilities measured by cell invasion assays. These types of comparisons help to correlate changes monitored in the highly resolved nanoscale protein distributions to that of cellular mechanics and functions.

With this thesis, I hope that I have convinced you that multicolor high resolution imaging in combination with image analysis provides a powerful tool for biological studies.

69

Bibliography

[1] S. Bradbury, The evolution of the microsope. London: Pergamon Press, 1965.

[2] W.J. Croft, Under the Microscope: A Brief History of Microscopy. London: World Scientific, 2006.

[3] E.C. Abbe, "Beiträge zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung," Archiv für mikroskopische Anatomie, vol. 9, pp. 413–468, 1873.

[4] J.W. Rayleigh, "On the theory of optical images, with special reference to the microscope," Philosophical Magazine, vol. 5, no. 42, pp. 167–195, 1896.

[5] A. Anand et al., "High-resolution quantitative phase microscopic imaging in deep UV with phase retrieval," Optics Express, vol. 36, no. 22, pp. 4362-4364, 2011.

[6] J. Kirz, C. Jacobsen, and M. Howells, "Soft X-ray microscopes and their biological applications," Q. Rev. Biophys, vol. 28, pp. 33-130, 1995.

[7] E. Ruska, "Nobel Lecture: The Development of the Electron Microscope and of Electron Microscopy," , 1986.

[8] J.M. Robinson, T. Takizawa, and D.D. Vandre, "Applications of gold cluster compounds in immunocytochemistry and correlative microscopy: comparison with colloidal gold," Journal of Microscopy, vol. 199, no. 3, pp. 163-179, 2000.

[9] D.A. Meyer, J. Heintz, and R.M. Albrecht, "Detection of Nanoparticles of Differing Composition for High Resolution," Microsc. Microanal., vol. 16, no. 2, pp. 992-993, 2010.

[10] G. Binnig, C.F. Quate, and C. Gerber, "Atomic force microscope," Physical Review Letters, vol. 56, no. 9, pp. 930-933, 1986.

[11] E.A Ash and G. Nicholls, "Super-resolution aperture scanning microscope," Nature, vol. 237, no. 5357, pp. 510-512, 1972.

[12] R.Y. Tsien, L. Ernst, and A. Waggoner, "Fluorophores for confocal microscopy: Photophysics and photochemistry," in Handbook of biological confocal microscopy, 3, Ed. New York: Springer, 2006, pp. 338-352.

[13] R.Y. Tsien, "The green fluorescent protein," Annual Review of Biochemistry, vol. 67, no. 5, pp. 509-544, 1998.

[14] M. Minsky, "US Patent: Microscopy apparatus," 3013467, November 7, 1957.

70

[15] T. Wilson, "The role of the pinhole in confocal imaging system," in In Handbook of Biological Confocal Microscopy. New York: Plenum Press, 1995, pp. 167-182.

[16] T.H. Maiman, "Stimulated Optical Radiation in Ruby," Nature, vol. 187, no. 4736, pp. 493-494, 1960.

[17] K. Carlsson et al., "Three-dimensional microscopy using a confocal laser scanning microscope," Optics Letters, vol. 10, no. 2, pp. 53-55, 1985.

[18] S.W. Hell and J. Wichmann, "Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy," Optics Letters, vol. 19, no. 11, pp. 780-782, 1994.

[19] T.A. Klar and S.W. Hell, "Subdiffraction resolution in far-field fluorescence microscopy," Optics Letters, vol. 24, no. 14, pp. 954-956, 1999.

[20] A. Einstein, "Strahlungs-emission und -absorption nach der Quantentheorie," Verhandlungen der Deutschen Physikalischen Gesellschaft, vol. 18, pp. 318-323, 1916.

[21] K.I. Willig, S.O. Rizzoli, V. Westphal, R. Jahn, and S.W. Hell, "STED microscopy reveals that synaptotagmin remains clustered after synaptic vesicle exocytosis," Nature, vol. 440, no. 7086, pp. 935-939, 2006.

[22] B. Harke et al., "Resolution scaling in STED microscopy," Optics Express, vol. 16, no. 6, pp. 4154-4162, 2008.

[23] D. Wildanger et al., "Solid Immersion Facilitates Fluorescence Microscopy with Nanometer Resolution and Sub-Ångström Emitter Localization," Advanced Materials, vol. 24, pp. 309-313, 2012.

[24] E. Betzig et al., "Imaging intracellular fluorescent proteins at nanometer resolution," Science, vol. 313, no. 5793, pp. 1642–1645, 2006.

[25] M.J. Rust, M. Bates, and X.W. and Zhuang, "Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM)," Nature Methods, vol. 3, no. 10, pp. 793–795, 2006.

[26] R.E. Thompson, D.R. Larson, and W.W. Webb, "Precise nanometer localization analysis for individual fluorescent probes," Biophysical Journal, vol. 82, no. 5, pp. 2775–2783, 2002.

[27] A. Pertsinidis, Y. Zhang, and S. Chu, "Subnanometre single-molecule localization, registration and distance measurements," Nature, vol. 466, pp. 647-651, 2010.

Bibliography

71

[28] M.G.L. Gustafsson, "Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy," Journal of Microscopy, vol. 198, pp. 82-87, 2000.

[29] J. Fölling et al., "Fluorescence nanoscopy by ground-state depletion and single-molecule return," Nature Methods, vol. 5, pp. 943-945, 2008.

[30] S. Munck et al., "Sub-diffraction imaging on standard microscopes through photobleaching microscopy with non-linear processing.," Journal of Cell Science, vol. 125, no. 9, pp. 2257-2266, 2012.

[31] T. Dertinger et al., "Superresolution optical fluctuation imaging (SOFI)," Advances in Experimental Medicine and Biology, vol. 733, pp. 17-21, 2012.

[32] R.C. Gonzales and R.E. Woods, Digital Image Processing, 3rd ed., M.J. Horton, Ed. New Yersey, US: Pearson Education, Inc., 2008.

[33] E.M. Manders, J. Stap, G.J. Brakenhoff, R. van Driel, and J.A. Aten, "Dynamics of three-dimensional replication patterns during the S-phase, analysed by double labelling of DNA and confocal microscopy.," J Cell Science, vol. 103, no. 3, pp. 857-862, 1992.

[34] E.M.M. Manders, F.J. Verbeek, and J.A. Aten, "Measurement of co-localization of objects in dual-colour confocal images," Journal of Microscopy, vol. 169, no. 3, pp. 375-382, 1993.

[35] V. Zinchuk, O. Zinchuk, and T. Okada, "Quantitative colocalization analysis of multicolor confocal immunofluorescence microscopy images: pushing pixels to explore biological phenomena.," Acta Histochem Cytochem., vol. 40, no. 4, pp. 101-111, 2007.

[36] D. Neumann, J. Bückers, L. Kastrup, S.W. Hell, and S. Jakobs, "Two-color STED microscopy reveals different degrees of colocalization between hexokinase-I and the three human VDAC isoforms.," PMC Biophysics, vol. 3, no. 4, 2010.

[37] S.V. Costes et al., "Automatic and quantitative measurement of protein-protein colocalization in live cells.," Biophysical Journal, vol. 86, no. 6, pp. 3993-4003, 2004.

[38] J.W.D. Comeau, S. Costantino, and P.W. Wiseman, "A Guide to Accurate Fluorescence Microscopy Colocalization Measurements," Biophysical Journal, vol. 91, no. 12, pp. 4611–4622, 2006.

72

[39] S. Costantino, J.W. Comeau, D.L. Kolin, and P.W. Wiseman, "Accuracy and Dynamic Range of Spatial Image Correlation and Cross-Correlation Spectroscopy," Biophysical Journal, vol. 89, no. 2, pp. 1251-1260, 2005.

[40] J.W. Comeau, D.L. Kolin, and P.W. Wiseman, "Accurate measurements of protein interactions in cells via improved spatial image cross-correlation spectroscopy.," Molecular Biosystems, vol. 4, no. 6, pp. 672-685, 2008.

[41] M.A. Khorshidi et al., "Analysis of transient migration behavior of natural killer cells imaged in situ and in vitro," Integrative Biology, vol. 3, no. 7, pp. 770-778, 2011.

[42] N.V. Kirienko et al., "Pseudomonas aeruginosa Disrupts Caenorhabditis elegans Iron Homeostasis, Causing a Hypoxic Response and Death," Cell Host & Microbe, vol. 13, no. 4, pp. 406–416, 2013.

[43] D.M. Chudakov, S. Lukyanov, and Lukyanov K.A., "Tracking intracellular protein movements using photoswitchable fluorescent proteins PS-CFP2 and Dendra2," Nature Protocols, vol. 2, pp. 2024 - 2032, 2007.

[44] C. Wählby et al., "An image analysis toolbox for high-throughput C. elegans assays.," Nature Methods, vol. 9, no. 7, pp. 714-716, 2012.

[45] C. Wählby et al., "Finding Cells, Finding Molecules, Finding Patterns," Lecture Notes in Computer Science, vol. 486, pp. 104-114, 2007.

[46] G. Moneron et al., "Fast STED microscopy with continuous wave fiber lasers," Optics Express, vol. 18, pp. 1302-1309, 2010.

[47] G. Vicidomini et al., "Sharper low-power STED nanoscopy by time gating," Nature Methods, vol. 8, pp. 571-573, 2011.

[48] H. Nyquist, "Certain factors affecting telegraph speed," Bell System Technical Journal, vol. 3, pp. 324-346, 1924.

[49] C.E. Shannon, "Communication in the presence of noise," Proc. Institute of Radio Engineers, vol. 37, no. 1, pp. 10-21, 1949.

[50] L. Meyer et al., "Dual-Color STED Microscopy at 30-nm Focal-Plane Resolution.," Small, vol. 4, pp. 1095-1100, 2008.

[51] D. Wildanger, E. Rittweger, L. Kastrup, and S.W. Hell, "STED microscopy with a supercontinuum laser source," Optics Express, vol. 16, pp. 9614-9621, 2008.

Bibliography

73

[52] D. Wildanger, R. Medda, L. Kastrup, and S.W. Hell, "A compact STED microscope providing 3D nanoscale resolution," Journal of Microscopy, vol. 236, no. 1, pp. 35-43, 2009.

[53] T. Müller, C. Schumann, and Kraegeloh A., "STED Microscopy and its Applications: New Insights into Cellular Processes on the Nanoscale," ChemPhysChem, vol. 13, no. 8, pp. 1986-2000, 2012.

[54] G. Donnert et al., "Two-Color Far-Field Fluorescence Nanoscopy," Biophysical Journal, vol. 92, no. 8, pp. 67-69, 2007.

[55] B.J. Tromberg et al., "Non-Invasive In Vivo Characterization of Breast Tumors Using Photon Migration Spectroscopy," Neoplasia, vol. 2, no. 1-2, pp. 26-40, 2000.

[56] M. Monici, "Cell and tissue autofluorescence research and diagnostic applications," Biotechnology Annual Review, vol. 11, pp. 227-256, 2005.

[57] B.K. Müller, E. Zaychikov, C. Bräuchle, and D.C. Lamb, "Pulsed Interleaved Excitation," Biophysical Journal, vol. 89, no. 5, pp. 3508-3522, 2005.

[58] J. Bückers, D. Wildanger, G. Vicidomini, L. Kastrup, and S.W. Hell, "Simultaneous multi-lifetime multi-color STED imaging for colocalization analyses," Optics Express, vol. 19, no. 4, pp. 3130-3143, 2011.

[59] D. Brewster, "On the laws which regulate the polarisation of light by reflection from transparent bodies," Philosophical Transactions of the Royal Society of London, vol. 105, pp. 125-159., 1815.

[60] J. Keller, A. Schönle, and S.W. Hell, "Efficient fluorescence inhibition patterns for RESOLFT microscopy," Optics Express, vol. 15, no. 6, pp. 3361-3371, 2007.

[61] M. Reuss, J. Engelhardt, and S.W. Hell, "Birefringent device converts a standard scanning microscope into a STED microscope that also maps molecular orientation," Optics Express, vol. 18, no. 2, pp. 1049-1058, 2010.

[62] H.H. Jaffe and A.L. Miller, "The fates of electronic excitation energy," J. Chem. Educ., 1966, 43 (9), p 469, vol. 43, no. 9, p. 469, 1966.

[63] F. Jelezko and J. Wrachtrup, "Single defect centres in diamond: A review," physica status solidi (a), vol. 203, no. 13, pp. 3207–3225, 2006.

74

[64] B.M. Chang et al., "Highly Fluorescent Nanodiamonds Protein-Functionalized for Cell Labeling and Targeting," Advanced Functional Materials, vol. Early View, 2013.

[65] M. Mkandawire et al., "Selective targeting of green fluorescent nanodiamond conjugates to mitochondria in HeLa cells.," Journal of Biophotonics, vol. 2, no. 10, pp. 596-606, 2009.

[66] G. Vicidomini, G. Moneron, C. Eggeling, E. Rittweger, and S.W. Hell, "STED with wavelengths closer to the emission maximum," Optics Express, vol. 20, no. 5, pp. 5225-5236, 2012.

[67] L. Brocchieri and S. Karlin, "Protein length in eukaryotic and prokaryotic proteomes," Nucleic Acids Research, vol. 33, no. 10, pp. 3390–3400, 2005.

[68] L. Landmann and P. Marbet, "Colocalization Analysis Yields Superior Results After Image Restoration," Microscopy research and technique, vol. 64, pp. 103–112, 2004.

[69] P.J. Shaw and D.J. Rawlins, "The point spread function of a confocal microscope: its measurement and use in deconvolution," Journal of Microscopy, vol. 163, no. 2, pp. 151-165, 1991.

[70] G. Vicidomini et al., "STED Nanoscopy with Time-Gated Detection: Theoretical and Experimental Aspects," PLOS ONE, vol. 8, no. 1, 2013.

[71] M. Leutenegger, C. Eggeling, and S.W. Hell, "Analytical description of STED microscopy performance," Optics Express, vol. 18, no. 25, pp. 26417-26429, 2010.

[72] W.H. Richardson, "Bayesian-Based Iterative Method of Image Restoration," JOSA, vol. 62, no. 1, pp. 55-59, 1972.

[73] I. Patla et al., "Dissecting the molecular architecture of integrin adhesion sites by cryo-electron tomography," Nature Cell Biology, vol. 12, pp. 909-915, 2010.

[74] N. Otsu, "Threshold selection method from gray-level histograms," IEEE Transactions on Systems, Man and Cybernetics, vol. 9, no. 1, pp. 62-66, 1979.

[75] C. Wählby, J. Lindblad, M. Vondrus, E. Bengtsson, and Björkesten L., "Algorithms for cytoplasm segmentation of fluorescence labelled cells," Analytical Cellular Pathology 24 (2002) 101–111, vol. 24, no. 2-3, pp. 101-111, 2002.

[76] T.D. Pollard and J.A. Cooper, "Actin, a Central Player in Cell Shape and Movement," Science, vol. 326, no. 5957, pp. 1208-1212, 2009.

Bibliography

75

[77] V.R. Sarma, E.W. Silverton, D.R. Davies, and Terry W.D., "The three-dimensional structure at 6 A resolution of a human gamma Gl immunoglobulin molecule," Journal of Biological Chemistry, vol. 246, no. 11, pp. 3753-3759, 1971.

[78] Nygren P.A., "Alternative binding proteins: affibody binding proteins developed from a small three-helix bundle scaffold.," FEBS, vol. 275, no. 11, pp. 2668-2676, 2008.

[79] T. Grotjohann et al., "rsEGFP2 enables fast RESOLFT nanoscopy of living cells," eLife Sciences, vol. 1, pp. 1-14, 2012.

[80] P. Bingen, M. Reuss, J. Engelhardt, and S.W. Hell, "Parallelized STED fluorescence nanoscopy," Optics Express, vol. 19, no. 24, pp. 23716 - 23726, 2011.

[81] B. Hein et al., "Stimulated Emission Depletion Nanoscopy of Living Cells Using SNAP-Tag Fusion Proteins," Biophysical Journal, vol. 98, pp. 158–163, 2010.

[82] N.T. Urban, K.I. Willig, S.W. Hell, and U.V. Nägerl, "STED Nanoscopy of Actin Dynamics in Synapses Deep Inside Living Brain Slices," Biophysical Journal, vol. 101, pp. 1277–1284, 2011.

[83] K. Xu, H.P. Babcock, and X. Zhuang, "Dual-objective STORM reveals three-dimensional filament organization in the actin cytoskeleton," Nature Methods, vol. 9, no. 2, pp. 185-190, 2012.

[84] J. Radon, "On the determination of functions from their integral values along certain manifolds," Ieee Transactions on Medical Imaging, vol. 5, pp. 170–176, 1986.

[85] J.E. Eriksson et al., "Introducing intermediate filaments: from discovery to disease," Journal of Clinical Investments, vol. 119, no. 7, pp. 1763–1771, 2009.

[86] M. Parkin, F. Bray, J. Ferlay, and Pisani P., "Global Cancer Statistics, 2002," CA: A Cancer Journal for Clinicians, vol. 55, no. 2, pp. 74–108, 2005.

[87] O.W. Brawley, "Prostate carcinoma incidence and patient mortality. The effects of screening and early detection," Cancer, vol. 80, pp. 1857–1863, 1997.

[88] H. Sakamoto et al., "Prospective comparative study of the EUS guided 25-gauge FNA needle with the 19-gauge Trucut needle and 22-gauge FNA needle in patients with solid pancreatic masses," Journal of Gastroenterology and Hepatology, vol. 24, no. 3, pp. 384-390, 2009.

76

[89] G. Schueller et al., "US-guided 14-gauge core-needle breast biopsy: results of a validation study in 1352 cases," Radiology, vol. 248, no. 2, pp. 406-413, 2008.

[90] A. Berner, B. Davidson, E. Sigstad, and B. Risberg, "Fine-needle aspiration cytology vs. core biopsy in the diagnosis of breast lesions.," Diagnostic Cytopathology, vol. 29, no. 6, pp. 344-348, 2003.

[91] P.H. Tan and G.M. Tse, "Diagnosing breast lesions by fine needle aspiration cytology or core biopsy: which is better?," Breast Cancer Research and Treatment , vol. 123, no. 1, pp. 1-8, 2010.

[92] A. Nassar, "Core needle biopsy versus fine needle aspiration biopsy in breast--a historical perspective and opportunities in the modern era.," Diagnostic Cytopathology, vol. 39, no. 5, pp. 380-388, 2011.

[93] E. Sisamakis, A. Valeri, S. Kalinin, P.J. Rothwell, and C.A. Seidel, "Accurate single-molecule FRET studies using multiparameter fluorescence detection.," Methods in Enzymology, vol. 475, pp. 455-514, 2010.

[94] M. Cristóvão et al., "Single-molecule multiparameter fluorescence spectroscopy reveals directional MutS binding to mismatched bases in DNA," Nucleic Acids Research, vol. 40, no. 12, pp. 5448-5464, 2012.

[95] H. Wiksell, V. Ekstrand, C. Wadstrom, and G. Auer, "A new specially designed needle significantly increases sample yield during," Physica Medica, vol. 25, no. 1, pp. 47-50, 2009.

[96] H. Wiksell et al., "Prevention of tumour cell dissemination in diagnostic needle procedures," British Journal of Cancer, vol. 103, no. 11, pp. 1706-1709, 2010.

[97] J.I. Jones and D.R. Clemmons, "Insulin-like growth factors and their binding proteins: biological actions.," Endocrine Reviews, vol. 16, no. 1, pp. 3-34, 1995.

[98] M.A. Olayioye, "Update on HER-2 as a target for cancer therapy: Intracellular signaling pathways of ErbB2/HER-2 and family members," Breast Cancer Research, vol. 3, no. 6, pp. 385-389, 2001.

[99] T. Ekblad et al., "Synthesis and chemoselective intramolecular crosslinking of a HER2-binding affibody," Peptide Science, vol. 92, no. 2, pp. 116-123, 2009.

[100] J.F. Nasuti, P.K. Gupta, and Z.W. Baloch, "Diagnostic value and cost-effectiveness of on-site evaluation of fine-needle aspiration specimens: Review of 5,688 cases," Diagnostic Cytopathology, vol. 27, no. 1, pp. 1-4, 2002.

Bibliography

77

[101] P. Barta et al., "Protein interactions with HER-family receptors can have different characteristics depending on the hosting cell line," International Journal of Oncology, vol. 40, no. 5, pp. 1677-1682, 2012.

[102] S.Y. Wong and R.O. Hynes, "Lymphatic or Hematogenous Dissemination: How Does a Metastatic Tumor Cell Decide?," Cell Cycle, vol. 5, no. 8, pp. 812–817, 2006.

[103] G.L. Nicolson, "Organ specificity of tumor metastasis: role of preferential adhesion, invasion and growth of malignant cells at specific secondary sites.," Cancer Metastasis Reviews, vol. 7, no. 2, pp. 143-188, 1988.

[104] J. Folkman, "How is blood vessel growth regulated in normal and neoplastic tissue? G.H.A. Clowes memorial Award lecture.," Cancer Research, vol. 46, no. 2, pp. 467-473, 1986.

[105] I.J. Fidler, "The pathogenesis of cancer metastasis: the ’seed and soil’ hypothesis revisited," Nature Review Cancer, vol. 3, pp. 453–458, 2003.

[106] M. Vicente-Manzanares, C.K. Choi, and A.R. Horwitz, "Integrins in cell migration – the actin connection," Journal of Cell Science, vol. 122, pp. 199-206, 2009.

[107] B. Geiger and K.M. Yamada, "Molecular architecture and function of matrix adhesions," Cold Spring Harbor Perspectives in Biology, vol. 3, 2011.

[108] J.G. Lock, B. Wehrle-Haller, and S. Strömblad, "Cell-matrix adhesion complexes: master control machinery of cell migration," Seminars in Cancer Biology, vol. 18, no. 1, pp. 65-76, 2008.

[109] J.T. Parsons, A.R. Horwitz, and M.A. Schwartz, "Cell adhesion: integrating cytoskeletal dynamics and cellular tension," Nature Reviews Molecular Cell Biology, vol. 11, pp. 633-643, 2010.

[110] V. Sanz-Moreno et al., "Rac activation and inactivation control plasticity of tumor cell movement.," Cell, vol. 135, no. 3, pp. 510-523, 2008.

[111] F.M. Vega and A.J. Ridley, "Rho GTPases in cancer cell biology.," FEBS Letters, vol. 582, no. 14, pp. 2093-2101, 2008.

[112] P. Roy, Z. Rajfur, P. Pomorski, and K. Jacobson, "Microscope-based techniques to study cell adhesion and migration," Nature Cell Biology, vol. 4, pp. 91-96, 2002.

78

[113] W.R. Legant et al., "Multidimensional traction force microscopy reveals out-of-plane rotational moments about focal adhesions," PNAS, vol. 110, no. 3, pp. 881-886, 2013.

[114] K.A. Beningo, M. Dembo, I. Kaverina, J.V. Small, and Y.L. Wang, "Nascent focal adhesions are responsible for the generation of strong propulsive forces in migrating fibroblasts," Journal of Cell Biology, vol. 153, no. 4, pp. 881-888, 2001.

[115] D.A. Fletcher and R.D. Mullins, "Cell mechanics and the cytoskeleton," Nature, vol. 463, pp. 485-492, 2010.

[116] C. Le Clainche and M.F. Carlier, "Regulation of actin assembly associated with protrusion and adhesion in cell migration," Physiological Reviews, vol. 88, no. 2, pp. 489-513, 2008.

[117] M.L. Gardel, I.C. Schneider, Y. Aratyn-Schaus, and C.M. Waterman, "Mechanical integration of actin and adhesion dynamics in cell migration.," Annual Review of Cell and Developmental Biology, vol. 26, pp. 315-333, 2010.

[118] I.S. Chernoivanenko, An.A. Minin, and A.A. Minin, "Role of vimentin in cell migration," Russian Journal of Developmental Biology, vol. 44, no. 3, pp. 144-157, 2013.

[119] S.E. Cross, Y.S. Jin, J. Rao, and J.K. Gimzewski, "Nanomechanical analysis of cells from cancer patients," Nature Nanotechnology, vol. 2, pp. 780-783, 2007.

[120] M. Yilmaz and G. Christofori, "EMT, the cytoskeleton, and cancer cell invasion," Cancer and Metastasis Reviews, vol. 28, no. 1-2, pp. 15-33, 2009.

[121] M.J.C. Hendrix, E.A. Seftor, Y.W. Chu, K.T. Trevor, and R.E.B. Seftor, "Role of intermediate filaments in migration, invasion and metastasis," Cancer and Metastasis Reviews, vol. 15, pp. 507-525, 1996.

[122] M.G. Mendez, S.I. Kojima, and R.D. Goldman, "Vimentin induces changes in cell shape, motility, and adhesion during the epithelial to mesenchymal transition," FASEB, vol. 24, no. 6, pp. 1838-1851, 2010.

[123] W.C. Hahn et al., "Creation of human tumour cells with defined genetic elements," Nature, vol. 400, no. 6743, pp. 464-468, 1999.

[124] R. Bhattacharya et al., "Recruitment of vimentin to the cell surface by β3 integrin and plectin mediates adhesion strength," Journal of Cell Science, vol. 122, pp. 1390-1400, 2009.

Bibliography

79

[125] T. Volk, B. Geiger, and A. Raz, "Motility and adhesive properties of high- and low-metastatic murine neoplastic cells," Cancer Research, vol. 44, no. 2, pp. 811-824, 1984.

[126] E. Makrodouli et al., "BRAF and RAS oncogenes regulate Rho GTPase pathways to mediate migration and invasion properties in human colon cancer cells: a comparative study," Molecular Cancer, vol. 10, no. 118, 2011.

[127] P.M. Campbell et al., "K-Ras promotes growth transformation and invasion of immortalized human pancreatic cells by Raf and phosphatidylinositol 3-kinase signaling.," Cancer Research, vol. 67, no. 5, pp. 2098-2106, 2007.

[128] Y. Zhang et al., "HDAC-6 interacts with and deacetylates tubulin and microtubules in vivo," EMBO, vol. 22, no. 5, pp. 1168-1179, 2003.

[129] A.S. Wolberg, "Thrombin generation and fibrin clot structure," Blood Reviews, vol. 21, pp. 131-142, 2007.

[130] A.D. Michelson, "The Platelet Proteome," in Platelets.: Academic Press, 2012, p. 103.

[131] S.W. Whiteheart, "Platelet granules: surprise packages," Blood, vol. 118, no. 5, p. 1370, 2011.

[132] D. Cervi et al., "Platelet-associated PF-4 as a biomarker of early tumor growth," Blood, vol. 111, no. 3, pp. 1201-1207, 2008.

[133] G.L. Klement et al., "Platelets actively sequester angiogenesis regulators," Blood, vol. 113, no. 12, pp. 2835-2842, 2009.

[134] A.M. Algra and P.M. Rothwell, "Effects of regular aspirin on long-term cancer incidence and metastasis: a systematic comparison of evidence from observational studies versus randomised trials," The Lancet Oncology, vol. 13, no. 5, pp. 518-527, 2012.

[135] C. Medina et al., "Differential inhibition of tumour cell-induced platelet aggregation by the nicotinate aspirin prodrug (ST0702) and aspirin," British Journal of Pharmacology, vol. 166, no. 3, pp. 938-949, 2012.

[136] H.N. Pannerden et al., "The platelet interior revisited: electron tomography reveals tubular α-granule subtypes," Blood, vol. 116, no. 7, pp. 1147-1156, 2010.

80

[137] D.A. Meyer, J.A. Oliver, and R.M. Albrecht, "Colloidal palladium particles of different shapes for electron microscopy labeling," Microscopy and Microanalysis, vol. 16, no. 1, pp. 33-42, 2010.

[138] J.M. Robinson, T. Takizawa, and D.D. Vandré, "Enhanced Labeling Efficiency Using Ultrasmall Immunogold Probes: Immunocytochemistry," The Journal of Histochemistry & Cytochemistry, vol. 48, no. 4, pp. 487-492, 2000.

[139] J.S. Bennett and G. Vilaire, "Exposure of platelet fibrinogen receptors by ADP and epinephrine.," Journal of Clinical Investigation, vol. 64, no. 5, pp. 1393–1401, 1979.

[140] L.A. Kirkby, G.S. Sack, A. Firl, and M.B. Feller, "A Role for Correlated Spontaneous Activity in the Assembly of Neural Circuits," Neuron, vol. 80, pp. 1129-1144, 2013.

[141] H. Hering and M. Sheng, "Dendritic spines: structure, dynamics and regulation," Nature Reviews Neuroscience, vol. 2, no. 12, pp. 880-888, 2001.

[142] J.C. Skou and M. Esmann, "The Na,K-ATPase," Journal of Bioenergy and Biomembranes, vol. 24, no. 3, pp. 249-261, 1992.

[143] A.E. Moseley et al., "Deficiency in Na,K-ATPase alpha isoform genes alters spatial learning, motor activity, and anxiety in mice.," Journal of Neuroscience, vol. 27, no. 3, pp. 616-626, 2007.

[144] J.B. Lingrel, M.T. Williams, C.V. Vorhees, and A.E. Moseley, "Na,K-ATPase and the role of alpha isoforms in behavior.," Journal of Bioenergetics and Biomembranes, vol. 39, no. 5-6, pp. 385-389, 2007.

[145] H. Hebert, P. Purhonen, H. Vorum, K. Thomsen, and A.B. Maunsbach, "Three-dimensional structure of renal Na,K-ATPase from cryo-electron microscopy of two-dimensional crystals," Journal of Molecular Biology, vol. 314, no. 3, pp. 479-494, 2001.

[146] A.M. Bertorello, J.F. Hopfield, A. Aperia, and P. Greengard, "Inhibition by dopamine of (Na+ + K+)ATPase activity in neostriatal neurons through D1 and D2 dopamine receptor synergism," Nature, vol. 347, pp. 386-388, 1990.

[147] X.J. Cheng, J.O. Höög, A.C. Nairn, P. Greengard, and A. Aperia, "Regulation of rat Na(+)-K(+)-ATPase activity by PKC is modulated by state of phosphorylation of Ser-943 by PKA," American Physiological Society, vol. 273, pp. 1981-1986, 1997.

[148] P. Greengard, P.B. Allen, and A.C. Nairn, "Beyond the dopamine receptor: the DARPP-32/Protein phosphatase-1 cascade.," Neuron, vol. 23, pp. 435-447, 1999.

Bibliography

81

[149] P. Svenningsson et al., "DARPP-32: an integrator of neurotransmission.," Annual Reviews Pharmacology and Toxicology, vol. 44, pp. 269-296, 2004.

[150] C.W. Chang, D. Sud, and M.A. Mycek, "Fluorescence Lifetime Imaging Microscopy," Methods in Cell Biology, vol. 81, pp. 495–524, 2007.

[151] J. Bückers, D. Wildanger, G. Vicidomini, L. Kastrup, and S.W. Hell, "Simultaneous multi-lifetime multi-color STED imaging for colocalization analyses," Optics Express, vol. 19, no. 4, pp. 3130-3143, 2011.

[152] K. Suzuki-Inoue, O. Inoue, and Y. Ozaki, "Novel platelet activation receptor CLEC-2: from discovery to prospects.," Journal of Thrombosis and Haemostasis, vol. 9, pp. 44-55, 2011.

[153] K.J. Woollard and J. Chin-Dusting, "P-selectin antagonism in inflammatory disease," Current Pharmaceutical Design, vol. 16, no. 37, pp. 4113-4118, 2010.

[154] Z.M. Ruggeri, "Von Willebrand factor: looking back and looking forward.," Thrombosis and Haemostasis, vol. 98, no. 1, pp. 55-62, 2007.

[155] A.L. Barabási and Z.N. Oltvai, "Network biology: understanding the cell's functional organization," Nature Reviews Genetics, vol. 5, pp. 101-113, 2004.

82

Summary of papers and author contributions

Paper I

Rho GTPases link cellular contractile force to the density and distribution of nanoscale adhesions Annica K. B. Gad, Daniel Rönnlund*, Alexander Spaar*, Andrii A. Savchenko, Gabor Petranyi, Hans Blom, Laszlo Szekely, Jerker Widengren, Pontus Aspenström FASEB Journal, 2012, Volume 26, Issue 6, pages 2374-2382, DOI: 10.1096/fj.11-195800 Summary: In this work, the effects of different constitutive active variants of Rho GTPases are studied regarding cell adhesive properties, by use of TFM and STED microscopy. We conclude that all studied variants, RhoAV14, Rac1L61 and RhoDV26, increase the contractile force of cells as compared to control cells, where RhoAV14 induced the largest increase. All of the studied variants also reported less thick actin stress fibers coupled to dense cell matrix adhesion complexes, and instead a more homogenous distribution of adhesion particles coupled to thinner fibers. Since only RhoAV14 showed increased densities of adhesion particles, the combination of TFM and STED data indicates that a more homogenous distribution of adhesion particles increases the contractile force of cells. Author contribution: The author was responsible for the STED measurements and corresponding image analysis, involving development of analysis algorithms. The author also contributed to the manuscript writing and discussions regarding the findings in these areas.

Paper II

Spatial Organization of Proteins in Metastasizing Cells Daniel Rönnlund*, Annica K. B. Gad*, Hans Blom, Pontus Aspenström, Jerker Widengren Cytometry Part A, 2013, Volume 83, Issue 9, pages 855–865, DOI: 10.1002/cyto.a.22304 Summary: The cytoskeletal and adhesive properties of tumor cells controls their capability of invading surrounding tissues. In this work we aimed to elucidate the subcellular distribution patterns of adhesion particles and vimentin intermediate filaments to identify possible differences between control and metastatic cells.


83

Image analysis procedures were established allowing an objective detection and characterization of these differences, where a distinct separation was obtained from a sparse number of cells. Author contribution: The author took main responsibility for all imaging and analysis, where new algorithms were developed for the high resolution structural information of the vimentin filaments. The author also took the major responsibility in the manuscript writing regarding methods and results.

Paper III

Oncogenes induce an HDAC6-mediated vimentin filament collapse that is linked to cell stiffness Li-Sophie Z Rathje*, Niklas Nordgren*, Torbjörn Pettersson*, Daniel Rönnlund*, Jerker Widengren, Pontus Aspenström, Annica K.B. Gad PNAS, early edition, DOI: 10.1073/pnas.1300238111 Summary: In this work we show how different oncogenes change the vimentin intermediate filament structure in cells as well as induce cellular stiffness and invasiveness. Expression of these oncogenes resulted in up regulation of tubulin deacetylase histone deacetylase 6 (HDAC6), which was required for the vimentin reorganization. We conclude that the oncogene induced reorganization of the vimentin filament network via HDAC6, contributes to cellular stiffness and tumor cell invasion. Author contribution: The author was responsible for the STED measurements and corresponding image analysis. The author also contributed to the manuscript writing and discussions regarding the findings in these areas.

Paper IV

Fluorescence Nanoscopy of Platelets Resolves Platelet-State Specific Storage, Release and Uptake of Proteins, Opening for Future Diagnostic Applications Daniel Rönnlund*, Yang Yang*, Hans Blom, Gert Auer, Jerker Widengren Adv. Healthcare Materials, 2012, Volume 1, Issue 6, pages 707–713, DOI: 10.1002/adhm.201200172 Summary: In this work, high resolution imaging and analysis is applied to map the spatial distributions of VEGF, PF-4 and fibrinogen in platelets. Upon platelet activation by thrombin or ADP, the studied proteins undergo specific spatial alterations regarding the size, number and location of the protein clusters. We performed a simple assignment procedure where the activation state of the

84

platelet (control, Thrombin or ADP) was identified with good accuracy based on these properties, which could have interesting diagnostic implications. Author contribution: The author took main responsibility for the STED measurements and developed and applied new algorithms for image analysis. The author also took the major responsibility in the manuscript writing regarding methods and results.

Paper V

Spatial distribution of Na+-K+-ATPase in dendritic spines dissected by nanoscale superresolution STED microscopy Hans Blom, Daniel Rönnlund, Lena Scott, Zuzana Spicarova, Jerker Widengren, Alexander Bondar, Anita Aperia, Hjalmar Brismar BMC Neuroscience, 2011, Volume 12, Issue 16, DOI: 10.1186/1471-2202-12-16 Summary: In this work, the spatial distribution of NA+K+-ATPase or NKA in dendritic spines is imaged by STED microscopy. The NKA was identified as separate pools located in the spine head and spine shaft as well as in the connecting dendritic structure. This compartmentalization could have implications in the generation of local sodium and potassium ion gradients within the spine and thus affect neuronal signaling. Author contribution: The author together with HB took responsibility for the STED measurements and analysis.

Paper VI

Nearest neighbor analysis of dopamine D1 receptors and Na+-K+-ATPases in dendritic spines dissected by STED microscopy Hans Blom, Daniel Rönnlund, Lena Scott, Zuzana Spicarova, Ville Rantanen, Jerker Widengren, Anita Aperia, Hjalmar Brismar Microscopy Research and Technique, 2012, Volume 75, Issue 2, pages 220–228, DOI: 10.1002/jemt.21046 Summary: In this work, nearest neighbor and colocalization analysis is performed to study the spatial correlation of the NKA pump and D1R receptor in dendritic spines. Due to the high resolution provided by STED microscopy and the non-direct interaction of NKA and D1R, standard colocalization algorithms fail to offer a satisfactory analysis of the possible interaction between these proteins. Instead nearest neighbor algorithms are employed which provides


85

statistics regarding the distances between these proteins, where D1R had an average distance of 80 ± 60 nm to NKA and 160 ± 60 nm to other D1R pools. Author contribution: The author together with HB took responsibility for the STED measurements and part of the analysis.

Paper VII

Spatial distribution of DARPP-32 in Dendritic Spines Hans Blom, Daniel Rönnlund, Lena Scott, Linda Westin, Jerker Widengren, Anita Aperia and Hjalmar Brismar PLoS ONE, 2013, Volume 8, Issue 9: DOI: 10.1371/journal.pone.0075155 Summary: In this work, the spatial distribution of Darpp-32 in combination with D1R in dendritic spines is analyzed with STED microscopy. The acquired high resolution images revealed several small discernible pools of Darpp-32 corresponding to just a few proteins in each pool. Similar analysis of D1R revealed fewer and larger pools mainly located in the spine head. The relatively low number of Darpp-32 suggests that only a few copies are necessary to regulate neuronal signaling. Author contribution: The author together with HB took responsibility for the STED measurements and analysis.

Paper VIII

Multicolor fluorescence nanoscopy by photobleaching – concept verification and its application to resolve selective storage of proteins in platelets Daniel Rönnlund, Lei Xu, Anna Perols, Amelie Eriksson Karlström, Gert Auer, Jerker Widengren Submitted Manuscript Summary: In this work, a new method is developed which exploits the photostability and excitation spectrum of dyes to increase the number of targets in super resolution STED microscopy. The procedure is used to demonstrate four color STED imaging at ≤ 40 nm resolution and at low crosstalk. The method is applied to study platelets where specific information is obtained regarding the spatial organisation of the proteins VEGF, PF-4 and fibrinogen together with filamentous actin.

86

Author contribution: The author designed the method and conducted the procedure. The author also analyzed the results and took a major responsibility in the manuscript writing.

Paper IX

Effects of resolution, target density and labeling on co-localization estimates – nanoscopy and modified algorithms to suppress false positives Lei Xu*, Daniel Rönnlund*, Annica Gad, Laura J. Braun, Pontus Aspenström, Jerker Widengren Manuscript in preparation Summary: This manuscript covers how different parameters such as resolution, density and labeling size affect the obtained results of different colocalization algorithms. These parameters needs to be taken into account when comparisons are made to others published results but also when comparing between obtained data sets and positive/negative controls. Author contribution: The author designed and conducted the initial study involving simulations and STED measurements. The study was continued and advanced by LX to include new colocalization algorithms and improved analysis. The author took part in the manuscript writing. * Authors contributed equally

super resolution optical imaging – image analysis, multicolor

Documents