A method for unbiased analysis of fluorescence microscope images of Alzheimer’s disease related amyloids

(1)

Linköping University | Department of Physics, Chemistry and Biology Master thesis, 30 hp | Master of Science in Engineering Biology

Spring term 2020 | LITH-IFM-A-EX—20/3767--SE

A method for unbiased analysis

of fluorescence microscope

images of Alzheimer’s disease

related amyloids

Samuel Haglund

Examinator, Per Hammarström Tutor, Sofie Nyström

(2)

Avdelning, institution

Division, Department

Department of Physics, Chemistry and Biology Linköping University

URL för elektronisk version

ISBN

ISRN: LITH-IFM-A-EX—20/3767--SE

_________________________________________________________________ Serietitel och serienummer ISSN

Title of series, numbering ______________________________ Språk Language Svenska/Swedish Engelska/English ________________ Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport _____________ Titel Title

A method for unbiased analysis of fluorescence microscope images of Alzheimer’s disease related amyloids

Författare Author

Samuel Haglund

Nyckelord Keyword

Amyloid beta, Luminocent conjugated oligothiophene, Hyperspectral image analysis

Sammanfattning

Abstract

Alzheimer's disease is a widespread disease that has devastating effects on the human brain and mind. Ultimately, it leads to death and there are currently no treatment methods available that can stop the disease progression. The mechanisms involved behind the disease are not fully understood although it is known that amyloid fibrils play an important role in the disease development. These fibrils are able to form plaques that can trigger neuronal death, by interacting with receptors on the cell surface and the synaptic cleft or by entering the cell and disturb important functions such as metabolic pathways. To study the plaque formation of amyloid proteins, both in vitro and in vivo methods are used to investigate the characteristics of the protein.

Luminescent conjugated oligothiophene probes are able to bind in to amyloid beta fibrils and emit light when excited by an external light source. This way fibrillation properties of the protein can be studied. Developing probes that can serve as biomarkers for detection of amyloid fibrils could change the way Alzheimer's is treated. Being able to detect the disease in its early disease course, and start treatments early, is suggested to stop the progression of neural breakdown. In this project a software is developed to analyze fluorescent microscopy images, taken on tissue stained with these probes. The software is able to filter out background noise and capture parts of the picture that are of interest when studying the amyloid plaques. This software generates results similar to if the images were to be analyzed using any software where the regions to analyze are selected manually, suggesting that the software developed produce reliable results unbiased by background noise.

Datum

Date 2020-06-09

(3)

Copyright

The publishers will keep this document online on the Internet – or its possible replacement – from the date of publication barring exceptional circumstances.

The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/her own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility.

According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.

For additional information about the Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.

(4)

Abstract

Alzheimer's disease is a widespread disease that has devastating effects on the human brain and mind. Ultimately, it leads to death and there are currently no treatment methods available that can stop the disease progression. The mechanisms involved behind the disease are not fully understood although it is known that amyloid fibrils play an important role in the disease development. These fibrils are able to form plaques that can trigger neuronal death, by interacting with receptors on the cell surface and the synaptic cleft or by entering the cell and disturb important functions such as metabolic pathways. To study the plaque formation of amyloid proteins, both in vitro and in vivo methods are used to investigate the characteristics of the protein.

Luminescent conjugated oligothiophene probes are able to bind in to amyloid beta fibrils and emit light when excited by an external light source. This way fibrillation properties of the protein can be studied. Developing probes that can serve as biomarkers for detection of amyloid fibrils could change the way Alzheimer's is treated. Being able to detect the disease in its early disease course, and start treatments early, is suggested to stop the progression of neural breakdown. In this project a software is developed to analyze fluorescent microscopy images, taken on tissue stained with these probes. The software is able to filter out background noise and capture parts of the picture that are of interest when studying the amyloid plaques. This software generates results similar to if the images were to be analyzed using any software where the regions to analyze are selected manually, suggesting that the software developed produce reliable results unbiased by background noise.

(5)

Acknowledgements

I would like to thank:

Per Hammarström and Sofie Nyström for allowing me to do my thesis in their group. Thank you for the support and guidance, the flexibility in the working plan, input in the writing of the report and being close when there were questions. You have both been very helpful.

Alexander Sandberg for your patience when introducing me to the laboratory equipment and the discussions of the work in general. You have been a good tutor and always taken time to explain things thoroughly.

Farjana for sharing data and discussing results. It was really nice of you to share the data I needed for analysis during these restricted times.

Max for discussions in general and basic help with getting started coding in R.

Ganesh for also answering questions about general things.

(6)

Abbreviations

AD Alzheimer's disease

Aβ Amyloid Beta

APP/AβPP Amyloid (beta) Precursor Protein

FTIR Transform Infrared Spectroscopy

E. Coli Escherichia Coli

IPTG Isopropyl β-d-1-thiogalactopyranoside MilliQ water Distilled and filtered water

EDTA Ethylenediaminetetraacetic acid

TRIS Tris(hydroxymethyl)aminomethane

DEAE Cellulose Dimethylaminoethyl cellulose

LCO Luminescent conjugated oligotiophene h-FTAA Heptamer-formyl thiophene acetic acid

q-FTAA Quadro-formyl thiophene acetic acid

p-FTAA Pentamer-formyl thiophene acetic acid ThT Thioflavin T

IEC Ion exchange chromatography

IPTG Isopropyl ß-D-1-thiogalactopyranoside

LB Lysogeny broth

MWCO Molecular weight cut-off

OD600 Optical density 600 nm

RCF Relative centrifugal force

SDS-PAGE Sodium dodecyl sulfate-polyacrylamide gel electrophoresis

SEC Size exclusion chromatography

LED Light Emitting Diode

Laser Light Amplification by Stimulated Emission of Radiation

CAA Cerebral amyloid angiopathy

ROI Region of Interest

σ Standard deviation

GUI Graphical User Interface

I500/540 Intensity ratio for 500 nm and 540 nm

ROI Region of Interest

cryoEM Cryogenic electron microscopy

(7)

Introduction

Alzheimer's disease (AD) is a neurodegenerative disease that affects around 1-2 percent of the population [1]. The disease is characterized by loss of cognitive functions that progress over time and lead to dementia. There is a long pre-clinical phase leading up to the stage of dementia, and AD is thought to begin many years before the actual symptoms become apparent which indicates a need of reliable biomarkers for discovery of the disease in its early developmental stages [2].

The implications of AD are devastating on many functional planes compared to other general health conditions such as cardiovascular disease, cancer or metabolic disease. It affects not only cognitive functions as reasoning, memory and language but also psychological stability and emotional control. Therefore, it is one of the most feared diseases to develop among people, but still disease effective modifying treatments remain beyond reach. Preventing AD depends on understanding early steps in the slowly progressive pathogenesis where extracellular plaques and intra neuronal neurofibrillary tangles play an important role. These plaques and tangles mainly consist of amyloid beta (Aβ) proteins and protein tau respectively [3]. Even though there are monoclonal antibodies developed to clear out amyloid beta protein that have shown to improve cognitive function, they do not reverse disease progression [4]. This might be because of the treatment being initiated to late, when the symptoms already have started to develop. Treatment with amyloid protein neutralizing agents is therefore suggested to be started early on in the disease course, before the onset of frank symptoms [5].

During the 1980s the primary sequence of Aβ was first discovered from extracellular amyloid aggregates [6]. These aggregates can cause severe neuronal and synaptic dysfunction and ultimately lead to death of neuronal cells, where the patient sometimes have lost as much as one third of the brain matter during late stages of the disease. Aβ is produced by cleavage of the amyloid precursor protein (AβPP) which results in peptides of different length. The most common Aβ peptide is 40 amino residues in length and thereby named Aβ1-40 whereas a small proportion of the peptides belong to the 42 residue long variant called Aβ1-42. The latter variant is more hydrophobic and therefore more prone to fibrillation and formation of aggregates that can be seen surrounding neural cells. There is also a collection of other amyloid proteins of varying length involved in the disease. Furthermore, the Aβ peptides also cause damage to the brain by interfering with neural functions intracellularly such as decreasing enzymatic activity of respiratory chain complexes in the mitochondria. [7]

(10)

2

Background and aim of the project

The aim of this project is to get familiar with the laboratory methods of producing and analyzing amyloid beta peptides and then develop a software that can analyze fluorescence microscopy images. To do so the R programming language is going to be used to analyze intensities from specific wavelengths in images taken by a fluorescence microscope. The fluorescence microscope deliver information about the intensity of each pixel in the image, for every wavelength it is able capture. This information can be used to calculate ratios between different wavelengths to see how probes bind to the tissue that is being analyzed, which in turn can give information about other things such as how much mature fibrils there are in comparison to pre mature fibrils in the sample.

Process

The project can be divided into three main activities protein expression, protein purification and measurements and analysis of the results, which include the development of the software. During protein expression the peptide was recombinantly expressed in E coli. After cell harvest and cell lysis, the desired protein was purified by ion exchange and size exclusion chromatography. Amyloid fibrils were obtained in a plate reader assay.

Analysis of the results both involve measurements of fibrillation in vitro using fluorescence spectroscopy, and analysis of amyloid aggregates in vivo using fluorescence microscopy to investigate aggregation in tissue. Pictures from the fluorescent microscope are processed using the R programming language, where a method of filtering the images from background noise is developed.

Project plan

As the project involves many different types of tasks that has to be carried out simultaneously the need of establishing a project plan in the beginning of the project is of importance. There is a GANTT scheme illustrated in Figure 1 that display the different activities of this project. The scheme was created with some help from supervisors that could assist in the time planning.

In order to complete the overall objectives of the project, testing fibrillation kinetics with the Aβ-monomer and developing the analytical tool, the main objective in the beginning of the project is to obtain sufficient amounts of protein. Therefore, the focus during initial weeks is on expressing and purifying the protein. This involves a lot of laboratory work and getting familiar with the equipment and methods used during each step of the process which is crucial for maintaining quality and safety. The first week is dedicated to introduction of the laboratory and going through the upcoming laboratory work and reading through some basic literature. During the following week the actual project begins with cell culture and harvesting. Thereafter, two weeks were needed for the purification the first time because of failing to purify the protein during the first week due to contamination. When the whole cycle of protein expression and purification has been run through a first time the planning report is written reflecting on the upcoming project.

(11)

3

The whole process of expression and purification is then repeated with the second Aβ-peptide as the fibrillation kinetic tests on the first Aβ-monomer was successful after a second purification. When both Aβ-monomers has been produced, kinetic experiments of both variants mixed together can be started. Also, by this time it will be suitable getting familiar with the computer analysis software that is going to be used. The writing of the report will be ongoing during the whole project, and there are a couple of weeks in the end of the project dedicated for the report and the final presentation only.

Figure 1. GANTT scheme showing the activities in the project and when they take place on a timeline represented in a chronological order.

Process analysis

In the beginning of the project, the main priority was to obtain purified amyloid protein. This was harder than imagined due to lack of laboratory experience as a result of my educational background. At the second attempt of purification enough protein was obtained, and successfully analyzed. It was demanding to culture cells and purify recombinant protein without any previous experience, since it involves so many steps and require strictness and a lot more experience of actual labwork than I had. Because of this, the decision was made to focus on the computer analysis of microscopy images instead of continuing on with the laboratory work risking the project to be delayed. This is also where the Hammarström group could make best profit of my knowledge, which was suggested before the start of the laboratory work. As they predicted this was also the case in this project, putting me in a more comfortable position even though it resulted in working with a completely unfamiliar programming language. This would also lead to less impact on the project by the restrictions of laboratory work what was put up shortly thereafter due to the corona pandemic 2020.

Moving on with the coding part earlier than planned affected the structure of the project and the way of working. During the laboratory work the procedures were structured in detail, whereas the programming part was more open and allowing for trying out different ways of solving problems. During the first two weeks of coding almost all the time was spent to understand the programming language and testing my new skills using old data, rather than actually working towards the actual goal. When the basics of the coding part was understood, the goal was set to develop a method of filtering out parts of images that does not need to be analyzed. For instance, this could be dark background and exo-skeletal parts from a fly. Most of the redundant information in the images has to be removed in order to make as accurate analyses as possible. To achieve this, the work revolved around finding a method that can handle as many types of images as possible without having to change the settings too much.

(12)

4

The hard part was to figure out where to start experimenting, where the idea of a relative filter pretty soon evolved.

Having a function for filtering the images, large sets of data were now ready to be analyzed and presented in suitable plots. Then a software for loading in the raw data and filtering it before presenting the results in plots had to be coded. As with all types of coding this took longer time than planned even though the programming language was mastered pretty well at this point. After having developed a software that could handle both fly and mouse tissue, data from different genotypes of the two species was analyzed using the method. Finally, a simple graphical interface was coded to make it easy for the rest of the group to use. The final flowchart is presented in Figure 2.

(13)

5

Theory

Amyloid formation

Many medical conditions are caused by proteins forming characteristic extracellular ensembles of misfolded proteins known as amyloids. This is a dynamic process generating insoluble protein aggregates that often are toxic and deposited in tissue as fibrillar protein. There are different proteins recognized to cause amyloid disease and the essence of amyloidosis lies within the capacity of the protein to acquire more than one conformation. The folding process of newly synthesized polypeptide is a rapid sequence where the pathological process is closely linked to the physiological normal protein folding. In the end of the protein folding process the protein can reach a relatively stable misfolded state, with an energy state similar to the native protein, illustrated in Figure 3. [8]

Figure 3. (Jevgenij A. Raskatov 2017) In the schematic plot the energy state of the molecule is illustrated with regards to the structural form of the protein. This shows that the amyloid protein structures are associated with a lower conformational energy state than that of the native protein state, which partly explains the stability of these aggregate formations..

(14)

6

Disease linked to amyloid proteins

Alzheimer's disease is the most well-known amyloid disease. It is linked to Aβ-protein aggregates found in extracellular plaques and to intracellular tangles composed of the Tau protein. Other neurodegenerative disease coupled to amyloid formation are Lewy body dementia and Parkinson's disease [9]. But also diseases like inclusion body mitosis is associated with amyloid proteins. It is an inflammatory disease characterized by slowly progressive weakness of muscles where deposits of amyloid protein is causing toxicity to skeletal muscles, most notably the Aβ1-42 variant [10]. Some forms of cancer have also been suggested being related to amyloid proteins, where increased levels of Aβ-levels is associated with developing these cancer forms [11].

In the case of Alzheimer's disease the proteolytic remodeling of the APP plays an important role in the amyloidosis where only a limited proportion of the amyloid proteins form fibrils. Although the mechanisms behind amyloidosis are not fully understood, the consequences are clear. Accumulation of Aβ in the brain cause oxidative stress and lead to inflammatory response in the cerebral cortex (Figure 4), which activates the apoptotic pathway of neural cells. The fibrils may also interact with local receptors on the cell. [12]

Figure 4. https://upload.wikimedia.org/wikipedia/commons/c/cc/Alzheimers_brain.jpg. The picture illustrates a healthy brain to the left and a brain suffering from Alzheimer's disease to the right. The main differences between the healthy brain and the sick brain that can be observed is that the sick brain has lost much of its size due to neuronal death. Hippocampus has shrunken greatly and the ventricles filled with cerebrospinal fluid grow larger.

(15)

7

Amyloid beta

Amyloid beta peptides are one of the main components of the plaques found in brain tissue from people with Alzheimer's disease. Aβ is formed by proteolytic processing of AβPP by β- and γ-secretase (Figure 5). One hypothesis is that misfolded oligomers of this protein can act as seeds and lead to a chain reaction where other amyloid beta

peptides also begin misfolding [8]. When aggregating they become toxic to nerve cells and the levels of aggregate formation correlate well with cognitive impairment. [13]

Although the normal function of Aβ is not well understood there are some evidence suggesting it could protect against oxidative stress and play a role in antimicrobial activity [14] [15]. The most common isoforms of the protein are Aβ1-40 and Aβ1-42 where the longer form is cleaved in the endoplasmic reticulum and the

shorter form being produced in the trans Golgi network [16]. They are thought to be intrinsically unstructured meaning that the protein lacks a fixed or ordered tertiary structure instead attaining a set of three-dimensional structures [17]. Furthermore, there are differences between the two isomers conformational states with the Aβ1-42 (Figure 6) variant showing a much more diverse formation of conformational states. These states are in general attributed to the topologies of the C-terminal part of the sequence [13].

.

Figure 6. (Anneli Hidalgo, 2018) The chemical structure of the amyloid beta 42 peptide is illustrated in the picture.

Figure 5. Adapted from Li Hongyun 2010 with permission. The amyloid precursor protein (AβPP) is a transmembrane protein that can be cleaved by secretase enzymes into a series of different protein sequences. In the illustration AβPP is first cleaved by β-secretase, and is then further cleaved by γ-secretase which results in amyloid peptides of different length depending on the site of cleavage by γ-secretase.

(16)

8

Amyloid fibril and plaque formation

Amyloid fibrils are formed when monomeric protein precursors aggregate into fibrils through a common nucleation growth mechanism. The monomers may be unfolded or partially unfolded and in rare cases aggregation is initiated by the native protein itself. During the first step of fibril formation, assembly of oligomers that vary in structure occur. They can then associate further to produce essential precursors of amyloid fibrils. Not all oligomers assemblies will form fibrils. They may however still be cytotoxic and relevant to the disease progression. Before the rapid polymerization into amyloid fibrils can happen, a critical stage where a nucleus is formed must have been reached. This stage is kinetically defined as the most unstable point during the fibril formation. The probability of nucleus formation determines the length of the lag time of amyloid assembly (Figure 7). [12]

Later in the self-assembly process, each precursor undergoes a structural transformation to form β-strand rich secondary structure which allows for fragmentation. This generates new fibril ends that can recruit new monomers, resulting in exponential fibril growth. This is known as primary nucleation. Secondary nucleation, oligomer formation on the surface of pre-existing fibrils, also enhance fibril formation. [12]

Figure 7. Adapted from (Xi Wen-Hui 2016) with permission. The fibrillation process happens in three steps. During the lag phase, monomers self-aggregate into disordered oligomers of beta sheets forming a nucleus that can be elongated in the phase of exponential fibril formation. Then during the elongation phase the fibrils grow rapidly by extension of the nucleus where new monomers attach, until a steady state is finally reached.

Structural studies of the amyloid fibrils from brain tissue have been proven difficult due to their insolubility. Therefore, examination of their structure has focused on fibrils formed in

vitro from synthetic homologous peptides. Techniques like FTIR have been used to examine

amyloid structure and that has led to a considerable amount of information about their morphology and internal structural conformation. The β-sheet content of the Aβ-peptide is linked to insolubility and thus related to neurotoxicity. It is thought that the N-terminus of the peptide is responsible for initiating α-β conformational switching where deprotonation of side chains cause destabilization of the α-helix that is accompanied by rearrangement to an

(17)

9

oligomeric β-sheet. Once the Aβ-peptide is fibrillated it is shown to be resistant to protease degradation [18]. Recent developments in cryoEM and ssNMR have captured high resolution structures of several proteins that are coded by different genetic polymorphs. The differences that appear between these high resolution models indicates that the Aβ-fibrils are polymorphic [19].

Understanding the molecular architecture of amyloid fibrils is an important step towards developing therapeutic treatments against the processes leading up to the disease. There is a wide variety of existing amyloid structures with different organizations of α-helices and β-strands making the disease difficult to understand. Despite their heterogeneity they also share common underlying architecture, like cross amyloid folding, where the β-strands in each protofilament align perpendicular to the long axis of the fibril. [12]

One thing amyloids of different sorts seem to have in common is a hollow core around which the proto filaments are organized. The morphology of synthetic fibrils can change depending on the surrounding solution, where they under some conditions form sheets and ribbons rather than discrete fibrils [11]. Different fibril morphologies of the same peptide also have significantly different toxicity in cell culture [20].

Transgenic animal models

Drosophila melanogaster model

Drosophila melanogaster, commonly known as the fruit fly, is used to study toxic aggregation properties of the amyloid beta peptide. Expression of the Aβ protein in neural and glial cells allows for studying mechanisms of the aggregation in the brain of the fly in detail. The model is also used to investigate the effects of different types of drugs that can have potential effects on disease progression. A majority of the disease related genes in human have similarities with the fly in their function. This makes the fly a very valuable tool in the understanding of human disease progression, and in the discovery of new treatments against Alzheimer's disease. In this case most of the human genes have homologs in Drosophila, such as the AβPP that is termed APP-like and the beta and gamma secretases. [21]

There are mainly two transgenic fly models to study Alzheimer's disease in Drosophila which mimics the proteolytic process completely. The first method expresses human AβPP and in the second model the human Aβ gene is fused to the N-terminal secretion signal peptide directly, which is crucial for a controlled and systematic investigation of various isoforms of the molecule. These models result in pathological characteristics similar to what is seen in humans and the progression of AD-like neurodegeneration mimics also mimics the progression seen in human. [22]

Expression of the Aβ-peptide the Drosophila model result in shortened lifespan, amyloid protein accumulation and locomotor impairment that progressively worsen throughout the life time. These are hallmarks of Alzheimer's disease that can be difficult to detect in rodent animal models. With a Gal4/UAS line where a reporter fly is paired with a driver fly, different Aβ-peptides can be expressed in the offspring [23]. The peptides considered in this project are the Aβ1-42, Aβ1-40 as well as the double expressing variants of Aβ1-42/-1-42, Aβ1-42/-1-40, Aβ1-42/-1-38 and Aβ1-42/-1-37, expressed in neuronal and glial cells.

(18)

10

AβPP mouse models

Various mouse models expressing human AβPP have been developed to study the amyloid beta peptide in vitro. The models differ between each other both regarding how the genes are expressed and how it affects tissue in the body. The London type AβPP mutation was the first genotype to be discovered in a family with AD, followed by the India mutation from which the first mini gene containing human AβPP cDNA and introns was designed. Mice containing this mutation show deficits in spatial learning and memory that progress slowly. Another genotype is the Swedish AβPP mutation that exist in several mice models, such as the AβPP 23 model used to study CAA, increasing the Aβ levels drastically. Other models used to study AD pathology include Flemish, Arctic, Dutch and Iowa AβPP mutation within the Aβ domain. These models have contributed to the understanding of molecular pathogenesis but are also used for therapeutic and functional studies. [22]

Probes used for staining the protein

The ThT probe is commonly used in vitro to monitor amyloid fibril formation. It gives a strong fluorescence signal at 482 nm with an excitation wavelength of 450 nm and can be used to quantify and monitor fibril formation [24]. Another common probe for in vitro studies is p-FTAA, with a fluorescence peak at 475 nm. P-FTAA belongs to the probe family luminescent conjugated oligothiophenes, LCOs.

LCOs are fluorescent molecules that can be used for staining amyloid protein aggregates. They display different fluorescent properties depending on the morphology of the target protein. These dyes are a powerful tool in fluorescent imaging of amyloid beta aggregates in tissue with Alzheimer's disease pathology. The ability to detect and assign the protein deposits gives the opportunity to study structural changes involved in aggregate formation [25]. The oligothiophenes porphyrin hybrids have showed excellent specificity toward amyloid beta deposits [23]. Furthermore, LCOs can optically differentiate between protein aggregates due to their flexible backbone that adopts different conformation depending on the polymorphic nature of the target protein. Changing the effective conjugation length if the molecule become twisted will cause different color of the emitted light [26].

(19)

11

Figure 8. Adapted from (Pal G. Ellingsen 2013) with persmission. Intensity emission spectrum of the LCOs q-FTAA h-FTAA used to stain fibrils with and their chemical structures..

The heptameric LCO h-FTAA (Figure 8) binds to mature amyloid fibrils and pre-amyloid fibrils emitting red fluorescence at peak wavelengths of 540 nm and 585 nm. In contrast, q-FTAA (Figure 8) with a maximum intensity of fluorescence at 500 nm, only bind to mature fibril structures displaying blue blue-shifted fluorescence. By combining these two LCOs and multiple types of microscopy methods different stages of the amyloid formation and maturation pathways can be assign. [27]

Certain molecular requirements of the LCOs have been proven crucial for detection of amyloid fibril formation in order to emit light. Preferably, the backbone of the molecule should consist of four to seven thiophene units with carboxyl groups as extension of the conjugated backbone. LCOs that meet these requirements can serve as sensitive diagnostic tools for studying amyloid aggregates. [27]

(20)

12

Protein purification

Ion-exchange chromatography

Ion-exchange chromatography is a method that separate molecules based on their net affinity to the ion exchanger, working on charged molecules including proteins, nucleotides and amino acids. There are two types of methods, anion-exchange and cation exchange that are used depending on if the molecule of interest is negatively or positively charged. The primary advantage with this method is that only one interaction is involved during the separation, and also the predictability of elution patterns where the charged molecules come out last [28].

Figure 9. Adapted from (Hidayat Ullah Khan 2012) with permission. Illustration of the principle of ion exchange chromatography using salt gradient elution. The substances in this case are the negatively charged proteins.

The principle of the chromatography method is that the molecules of interest are being bound to the column through ionic interaction. Because of the column matrix consisting of charged ionizable functional groups, electrostatic interaction between the stationary phase and the opposite charged analyte will form [29]. To separate the analyte from the rest of the solution the column matrix has to be washed with changeable counter ions. The elution conditions are usually altered so that either pH or salt concentration of the mobile phase are increased according to a predetermined schedule. First, molecules that has an opposite charge of the analyte will be cleared out, then the ionizable molecules are eluted based on their charge. Initially, molecules weakly bound to the stationary phase are washed away. Then as concentration or pH increases, also referred to as gradient elution, more strongly bound analyte molecules will release from the stationary phase (Figure 9) [30].

(21)

13

Dialysis

The biochemical process of dialysis uses a semipermeable membrane to separate molecules in a solution by their difference in diffusion rate. It is performed to remove unwanted small molecules such as salts from large macromolecules like proteins and relies on the natural phenomenon of diffusion, where molecules move from higher to lower concentration until an equilibrium is reached. Two solutions separated by a membrane are needed to perform the separation process, one buffer solution and then the sample solution. The buffer solution can then be adjusted, by controlling the pH or salt concentration for instance, so that the sample solution will attain the desired condition. Choosing membrane with regard to the sample molecule is important so that the pore size of the membrane has the right molecular weight cut of (MWCO) and can keep the sample molecules from diffusing through the membrane [31, 32].

Running the process of dialysis is rather uncomplicated and straightforward. Usually, the membrane is in the form of a tube containing the sample inside that is immersed into a larger container with buffer. The buffer might then be changed during the dialysis process to speed it up or allow for further dialysis when an equilibrium has been reached. When the dialysis is finished the tube is simply removed from the container and emptied from solution. If the process is carried out successfully, the molecules wished to dialyze have diffused through the membrane out into the buffer solution [32].

Reverse-flow centrifugal membrane filtration

Reducing the volume of the protein sample, and thereby increasing the concentration of protein, is a crucial step before gel filtration can be performed. This is done through reverse-flow centrifugal filtration where the sample is transferred into a tube and placing a smaller tube inside on top, containing a membrane on the bottom. When the tube containing the sample and the smaller tube is put into a centrifuge, the centrifugal force will push the liquid towards the bottom of the tube, and up through the membrane. This way, the sample can be concentrated down to a smaller volume if the cut off weight of the membrane is adjusted to the protein size. [33]

(22)

14

Size-exclusion chromatography (SEC)

Gel filtration chromatography is a form of SEC typically

applied to large

macromolecules such as proteins and peptides. The separation method involves a column usually packed with agarose or dextran polymers containing porous beads where the molecules can travel [36]. Larger molecules will have access to smaller volumes to travel in and therefore stay in the column for a shorter time, whereas smaller molecules will be able to travel inside pores which extends the travel path through the column [37]. Size dependent exclusion results in fractions where larger molecules are collected first and it is a separation method that works well on larger

molecules (Figure 10). In combination with other purification methods that separate on characteristics such as acidity or charge, it fits well as a final step [38].

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis

SDS PAGE is an analytical separation method used in biochemistry for separation of charged molecules in an electric field by their molecular mass. The matrix is composed of a polyacrylamide discontinuous gel and the anionic detergent, sodium dodecyl sulfate, binds to protein to form negatively charged complexes. When the protein binds to the detergent it is generally becoming denatured and solubilized, which cause the protein complex to form an ellipsoid or rod like shape with a length proportional to is molecular weight [39]. Applying a constant electric field to the gel containing protein will then force the protein to migrate towards the anode. The migration speed depend on the mass of the protein, which allows for separation of the proteins in reference to their mass. Light weight molecules will have the longest travel distance in the gel [39].

Fluorescence spectroscopy

Fluorescence spectroscopy is an analytical method that detects fluorescence from a sample by using a beam of light that excites electrons in the molecules of the sample, causing them to emit light during relaxation. The light can also be absorbed from a fluorophore, in that case the method is called absorption spectroscopy. [40]

Figure 10. Adapted from (Yen-Hao Lin 2009) with permission. The principal of SEC illustrated as a function of time.

(23)

15

Molecules have different energy levels, also referred to as vibrational states. In fluorescence spectroscopy it is the electronic ground state and the excited electronic state that are of interest. When the molecule absorbs a photon from the incident light it becomes excited, entering a higher electronic energy state. The excited molecule will eventually lose vibrational energy from collisions with other surrounding molecules, which in time results in de excitation and the release of a photon which can be detected by an instrument. Since the electron that is being de excited can drop back into different vibrational levels of the electronic ground state, the light emitted from the molecule will have different frequencies. This phenomenon of different vibrational energy states within each electronic state results in excitation and emission spectrums with different intensities of light emitted and absorbed depending on the wavelength which is illustrated in Figure 11. The loss of energy in the exited state will give rise to a shift in the energy level between the emission and absorption peak of the two states, referred to as Stokes shift. [40]

A spectrometer uses a filter or a monochromator to isolate the desired wavelength from the indecent light beam before passing it on to the sample. There are various types of lights that can be used as excitation sources where LASERs and LEDs belong to the most common ones. Using a laser as excitation source makes the filter or monochromator unnecessary since it only emits light in a very narrow wavelength. Then there is a sensor measuring the fluorescence from the sample being hit by the excitation light. Depending on the machine the detector can either be a single channel or a multi-channel variant, meaning that it can detect the intensity of only a

very narrow wavelength or it can detect a spectrum of wavelength simultaneously. During the measurements the excitation light is kept constant at a wavelength where the analyzed molecule has high absorption, and preferably the emission spectrum is scanned for in a span of wavelengths. [41]

Fluorescence microscopy

In order for fluorescence microscopy to work the sample must be fluorescing. Fluorescence occurs when light is being emitted in another wavelength than that of the excitation source, as a result of absorption of light. The difference between absorbed and emitted light is called the Stokes shift and is a critical property of this method. By filtering out everything except the emitted light completely it is possible to see only the objects of interest, the ones that fluoresce.

Figure 11. (D. Guti´errez 2019) Jablonski diagram illustration the different energy levels and vibrational states of an electron that is being excited.

(24)

16 It is a sensitive method where

even single fluorescent molecules are visible if the background has no auto fluorescence. Fluorophores are the compounds that deliver fluorescent properties and their efficiency is determined by the outermost electron orbitals where conjugated systems often go hand in hand with fluorescent properties. As in the case with fluorescence spectroscopy the fluorophore is excited when light excites the molecule, that then release a photon during relaxation. [42]

The main components of the fluorescent microscope are the light source, the excitation filter, the dichroic mirror and the emission filter (Figure 12). Adjusting the dichroic mirror and the filters to match spectral excitation and emission characteristics of the fluorophore is important to be able to image the distribution of a single fluorophore. [43]

R Programming language

R is a programming language mainly used for statistical computing consisting of a language plus a run-time environment with graphics. The purpose of R is to contain functionality for a large number of statistical procedures and provide a flexible graphical environment for creating various kinds of data presentations [44]. R is widely used among statisticians and there are more than 400 contributed packages available for free, making the programming easier, which arguably makes it one of the most extensive resources for statistical computing currently available [45].

Packages used

The code written for this project relies on the functions of several packages. They are a collection of compiled code stored under a directory called “library”. To plot data in different kinds of plots ggplot2 and ggpubr are mainly used whereas pheatmap deliver the heatmap visualizations of intensity ratios. Loading and exporting .xlsx files openxlsx and xlsx are used. R shiny package is used to create the interactive web application for loading microscopy images for filtering and comparison.

Figure 12. Schematic explanation of the principle of the fluorescent microscopy.

(25)

17

Materials and Methods

Transformation and expression of protein

Electro competent Escherichia Coli cells (E.coli) of strain BL21 star (DE3) pLysS stored in -80° C freezer and a plasmid, containing the Aβ-gene of interest and a gene for antibiotics resistance, was thawed. Then 50 μl of the E.coli was mixed with 1 μl of the plasmid before pipetting the solution into an electroporation cuvette. The cells were transformed using electroporation with the equipment MicroPulser Bio-Rad at 2.5 mV for about 5-6 ms. Immediately after the procedure was performed 200 μl SOC medium is added to the cuvette and gently mixed with the pipette before the cells in the medium were transferred into an Eppendorf tube. Agar plates containing antibiotics, in this case 50 μg/ml ampicillin and 30 μg/ml chloramphenicol which the E.coli are resistant to, were used to grow colonies on by pouring the bacteria on the plate and incubating it over a night. One colony was then picked and transferred into 50 ml LB broth pre culture tube containing 15 ml LB medium. This tube was also incubated over a night. The incubations were done at 37° C .

Protein expression started after the incubation of the pre-culture is performed by adding it to 1.5 L of LB medium containing the same concentrations of antibiotics as above. The culture was then incubated at 37°C in a shake incubator running at 100 rpm. Optical density was controlled continuously with a spectrophotometer from MIDSCI at 600 nm by taking samples from the culture. When the OD has reached a value between 0.4 and 0.6, induction with IPTG was performed to a final concentration of 0.5 mM, following an incubation for four more hours. After the expression was done, the cells were harvested by centrifugation for 30 min at 4000 g at 4° C. The pellet was resuspended in 30 ml Milli Q water and stored in -80° C until further use.

Extraction of the protein was done when the cells have been harvested. Lysis of the cells occurs when distilled water was added after harvest to free inclusion bodies containing the protein. The cells were then thawed and centrifuged before they were resuspended in Buffer A (10 mM Tris-HCl, 1 mM EDTA, pH 8), the solution in which sonication is performed during the first two steps. All sonication were done at 30% amplitude for a total of two minutes of time with 30 second intervals of alternating on time and off time. Centrifugation was performed at 18000 g at 4°C for 10 min. After two steps of sonication and centrifugation with resuspension in Buffer A the final sonication was performed and the inclusion body pellet was resuspended in Urea buffer (8M Urea, 10mM TRIS-HCl, 1mM EDTA, pH 8).

Purification of the protein

Ion-exchange chromatography

Before Ion exchange chromatography is performed the DEAE-cellulose has to be equilibrated in two steps. First it was washed with 100 mL 1M NaOH, then two times with 100 mL 1M NaCl and finally two times with 100 mL milli-Q water. Then the DEAE-cellulose was resuspended in Buffer A2 (1M TRIS-HCl, 10M EDTA, pH 8) and when the cellulose was sedimented the supernatant is removed and 40 mL of Buffer A2 is added where the pH is adjusted to pH 8.

(26)

18

The resuspention in Buffer A2 was repeated two times, and the cellulose was kept in a fridge until further use.

After washing and equilibrating the Urea containing protein was added to the DEAE-cellulose and placed under gentle agitation for 20 min. The cellulose was then poured into a Buchner funnel with a filter on and drained until the surface of the buffer has reached the surface of the sediment. Then the cellulose was washed according to the following steps:

- Two times with 20 mL Wash buffer (10mM TRIS-HCl, 1mM EDTA, 25mM NaCl, pH8) - Four times with 10 mL Elution buffer “Low Salt” (10mM TRIS-HCl, 1mM EDTA,

150mM NaCl, pH8)

- Four times with 10 mL Elution buffer “High Salt” (10mM TRIS-HCl, 1mM EDTA, 500mM NaCl, pH8)

The factions are collected between elution with each solution. They were collected in Falcon tubes and stored in a fridge until Dialysis was performed. The DEAE-cellulose is washed with distilled water and then 20% EtOH, before storage in 20% EtOH.

Dialysis

The low salt and high salt elution fractions collected from the ion-exchange chromatography were transferred into dialysis tubes with a MWCO at 3 kD. They were placed in a 8 L bucket containing distilled water and 2 mM NaOH pre cooled to 4°C. After approximately 4 hours the dialysis tubes were moved into another bucket where they are stored over a night in 4°C. The next day the solution from the tubes are moved to Falcon tubes using an injection needle.

Concentration using centrifugal membrane filtration

Membrane filtration was performed in a centrifuge at 3000 g for 30 min per run until approximately 3 mL of the protein solution is left and the rest of the solution not containing protein has diffused out through the membrane. The concentrated fractions are then frozen in liquid nitrogen and lyophilized before they are stored in -80°C.

Size-exclusion chromatography with Superdex 75

SEC was performed in the GE Healthcare system ÄKTA. Pellet containing protein from the “low salt” fraction is dissolved in 0.55 mL 6M guanidine hydrochloride and kept on ice. The column was washed with water and equilibrated with PBS before injecting the sample with a syringe into the system through a 1 mL loop. Fractions are collected at a speed of 0.5 mL/min and the column is run at a maximum pressure of 0.8 MPa (Figure 13). The chromatogram is shown in Fig X. All solutions used during the SEC were degassed to keep air bubbles away from the system. SDS-PAGE was performed on the fractions containing protein as a verification method and the absorbance is measured to determine the concentration.

(27)

19

Figure 13. Chromatogram for the SEC showing the absorbance at 280 nm during purification of Aβ1-42. Fractions C3-C7 is likely to contain protein.

Analysis and fibrillation

Concentration determination

By measuring the absorbance at 300 nm and 280 nm using a nanodrop spectrophotometer with a wave scan function the protein concentration can be determined. It is necessary to perform the protein concentration corrections at 300 nm since there are no tryptophan amino acids giving strong signal to noise at 280 nm. The absorbance value at 300 nm is used to remove light scattering which may affect the read out. PBS was used as a reference medium and the spectrum was set to a range of 250-450 nm. According to Beer-Lambert’s law and the extinction coefficient of 1490 𝑀−1 _𝑐𝑚−1 _{at 280 nm for the protein the following equation can} be used calculate the concentration:

𝐶 =

𝐴

280 − 𝐴

300 1490

Table 1. Concentrations of amyloid beta fibrils in the columns delivered by the SEC.

Column C4 C5 C6 C7

(28)

20

SDS-PAGE

Gel-electrophoresis was performed as a verification step of the protein purity. The gel was produced by mixing a separation gel and a concentration gel, first letting the separation gel polymerize before adding the concentration gel on top. Polymerization takes approximately 20 min and the gels were stored in a fridge in moisturized paper until further use.

Running one gel, 12 µL of the sample is mixed with 4 µL of 4x SDS cocktail containing DTT loading dye. Also, 1 µL of β-mercaptoethanol is added to each sample to prevent the formation of disulfide bonds. Then the samples were heated to 95° C for 5 minutes, before loading them into the wells. A protein ladder from Bio-rad was used as a molecular weight marker reference. Electrophoresis separation is then run at 100 V for 75 min to obtain optimal separation. When the electrophoresis is finished, the acrylamide gel was washed with distilled water and stained with Coomassie blue G-250 for 20 min. Destaining is then performed by washing the gel with vinegar, until the color is removed.

Fluorescence spectroscopy measurements

Fluorescence measurements are done in a microplate reader, M1000 PRO from Tecan, to investigate the binding of LCO probes to the fibrillated amyloid protein. The experiment consist of p-FTAA titrated with Aβ1-42 at different concentrations, running a fluorescence emission scan in a wide wavelength spectrum. The software Magellan is used to control the measurements running the bottom up mode during the process and the wells were loaded according to Figure 14.

(29)

21

Fluorescent microscopy image analysis

To be able to compare the data from the images taken by the microscope, each picture has to be processed first. There are images from two different systems to be analyzed. A) Drosophila which have a highly autofluorescent exoskeleton and rather weak amyloid signal and B) mouse brain samples with distinct amyloid plaque.

Only the parts of the picture that are relevant to the analysis of the amyloid beta protein plaque formation are to be included in the analysis. In this matter, the parts of the tissue that luminescent probes bind into are interesting to study. Other parts of the picture such as dark background, eyes from the fly and other exoskeletal parts have to be removed in a suitable way without affecting important parts of the image negatively. This is done using a filter that only include pixels that has an intensity corresponding to an interval close to the mean value of the whole picture.

Previous data from the lab have used Regions of Interest (ROI) analysis of portions of amyloid structures. My goal is to perform full image analyses.

Processing the image

Text files containing information about the intensity of the light for pixel in the image are the output from the microscope. The data is structured so that there is one file containing information about the intensities of the pixels corresponding to each wavelength in the range that the detector in the microscope is set to scan for. This way, luminescence of different probes used to stain the tissue in one image can be studied, since the intensity of each probe will be highest in the file corresponding to its peak emission wavelength. For instance, q-FTAA and h-FTAA have maximum emission at 500 nm and 540 nm which make the files containing the intensities for these wavelengths most interesting. Studying the ratio between two different emission wavelengths allows for comparing how probes bind to aggregates in the tissue of different genotypes.

First the files have to be loaded into R and stored as matrices. For analysis of the fluorescence of q-FTAA and h-FTAA, the 500 nm and 540 nm files are loaded in and stored as separate matrices. Then these files have to be filtered so that unnecessary information is not included, such as dark background and exoskeleton that reflect very high intensity light. This is done by applying a filter to the files. For the low wavelength matrix high intensity noise is filtered out, and for the high wavelength matrix low intensity information is filtered out. This appears to be the most effective way of filtering since the high intensities in the low wavelength matrix most likely correspond to the bright noise that is considered to be redundant information. Filtering the high wavelength matrix, for high intensities, would most likely result in filtering out important information too. Vice versa, filtering out the dark background noise, the high wavelength matrix is where the filter is applied to reduce risk of filtering out important parts of the picture. The dark parts in the high wavelength matrix are more likely to correspond to those really dark parts in the picture that is just background noise or parts of the tissue that is not desired to be included.

(30)

22

Figure 15. Schematic picture of the filtration function. First the two files are loaded into the software before they are filtered using a normal distribution function, where the mean value of each matrix is calculated and using standard deviations to add or subtract to the mean to get suitable cut-off values. The output is an intensity ratio matrix where values outside the desired range of the normal distribution are removed.

Removing pixels that do not belong to a predetermined interval of intensities is simply the function of the filter. The question is how to generalize this interval of intensities so that it fits in to an as broad range of image types as possible without having to modify the filter range too much between images. At least, filter range adjustments should be easy to do in a way that allow for qualified guesses of the settings. There has to be a robust reference interval on what may work, otherwise the reproducibility will be affected negatively. It turns out that using the Gaussian normal distribution as a reference interval is a suitable method (Figure 15). By calculating the mean value of the matrices and using standard deviations as upper and lower limits, an interval that include the information of interest can be created. Through experience it seems that the upper limit usually is the mean value plus three standard deviation, unless there are no exoskeleton parts or other things that reflect bright light. In that case the upper limit can be set to infinity. As a lower limit the mean value minus one or one half standard deviation is usually good, unless the picture is very dark, then it might be better to use the mean or plus one standard deviation as the lower limit cut-off. Where to put these interval limit cut-off values become quite easy to see after some filtering iterations displaying the results in a heat map, and when using a relative filter it is easy lower and higher the cut-off limits in a way that deliver predictable results (Figure 16).

(31)

23

Figure 16. The general results of the filter function displayed on fly brain tissue (A) and mouse brain tissue (B). A represents the fly image that has been filtered, A1 a heat map of the intensity ratio matrix and A2 the filtered intensity ratio matrix. The same order is applied to B, B1 and B2.

Signal to noise ratio

To obtain information about how the signal in the picture is related to the noise, signal to ratio analysis is performed. The method of this analysis is to add all the matrices for each wavelength together and divide every pixel by the total number of samples, and then simply divide the new matrix, containing information from the individual matrices, with a matrix containing same information from a healthy genotype. The result will be a new matrix containing information about the relation of matrices containing fluorescent signal and matrices that does not contain such information. From this matrix, a mean value is calculated that represent the signal to noise ratio.

Plotting results

Illustration of results is done in different types of plots. How the probes bind to the tissue is best illustrated using a heat map where a palette of colors correspond to the individual value of each pixel. When the ratio between the different wavelength intensity matrices has been calculated, the color palette is adjusted so that lower values corresponds to one specific colors while higher values correspond to another specific color. High ratios indicate that the binding of one type of probe is dominant in that area, while low ratios indicate that binding of another of probe is dominant. In between there is an interval that correspond to a third color indicating neither of the probes investigated is more likely than the other to be found in that area. In this work the colors red, green and blue have been used to represent the range of values in the ratio matrix calculated from the image wavelength intensities. It is suitable since the fluorescence is commonly referred to as either being red shifted or blue shifted depending on if q-FTAA (blue) or h-FTAA (red) is prominent in the tissue.

Individual comparisons between images are usually done using simple histograms where the number of counts within a ratio interval is plotted. These plots allow for distinguishing what probe is more prominent in what picture. Analyzing populations as a whole and how different

(32)

24

genotypes compare, the violin plot is a better options. Since the analysis of complete images usually generate large amount of data, it has to be narrowed down in a suitable way before the plotting package can handle it. This is done by adding the data together in a vector and then picking values evenly from the vector so that it reflects the characteristics of the original data as good as possible. Every n:th value is included, choosing an n so that the new data set is in the range of thousands of points instead of millions of points.

Graphical User Interface

R studio offers a package named Shiny that is used to create an interactive web application containing the filtering code. The main advantage using a GUI in this case is that the process of loading files into the software becomes easier. Instead of typing the name of every file into the programming code, multiple files can simply be selected at the same time using the mouse, and loaded into the software. The code that is doing the filtering is the same as in the original code, but the rest of the code is different in the Shiny version compared to the original version. This is due to the layout of the programming script in the Shiny version. In the original R script, the software executes the commands from the top of the script to the bottom of the script. When the script has been run, the software ends and results are delivered. However, in the Shiny version this process works differently.

Shiny applications consist of two key components, a server and a user interface. The user interface contains a layout that can handle input fields and output fields. Input fields are things as menus, textboxes and slots for uploading files that are then sent to the server for processing. Calculations and visualizations performed in the server are then sent back to the output section of the user interface, where it can be displayed. Shiny applications always need a server running R to work. R can launch a mini web server locally on the computer to run the application. In contrast to running the original R Script where the software ends when it hits the final line of the script, the server continues running until it is shut down. This means that if the inputs change, new calculations will be performed and new outputs will be generated.

In the Shiny version the inputs will be the intensity files for each wavelength that are loaded in for each genotype. Then the additional information about the filter settings and genotype names are also loaded in as text documents. The filter settings document has to be structured so that there is one setting per genotype on each row. First the upper cut-off limit is typed in, and then a space is made using TAB, followed by the lower cut-off limit and dot is used as comma. The settings for Genotype 1 has to be on row 1 and so on. When the last row is modified, enter is pressed so that the text indicator hits an empty row before saving the document. The same goes for the genotype names, where the first genotype has to be on the first line and so on. Outputs generated from the software are a simple boxplot for the genotypes and then some data about the plots. Also a file containing the plot data vector can be stored, in case the data is wished to be plotted using a different script or software. The difference between the original version and the Shiny interface is illustrated in Figure 17 and Figure 18, where Figure 17 illustrates the interface without Shiny and Figure 18 illustrates the interface using Shiny.

(33)

25

Figure 17. Running the large scale analysis code in R studio.

(34)

26

Results

Fibrillation kinetics

Amyloid fibril formation of Aβ1-42 produced in E.coli were monitored in the fluorescence plate reader as previously described. The native protein was stained with the LCO p-FTAA and fluorescence was measured in a plate reader using different protein concentrations. Scanning over a spectrum of wavelengths, the intensity is building up with increasing concentration of protein which is a positive sign that experiment is correctly set up. When comparing time and intensity, a characteristic fibrillation kinetics curve is produced.

Figure 19. Fluorescence emission plot of Aβ1-42 stained with p-FTAA at different concentrations.

In Figure 19, where Aβ1-42 at different concentrations is stained with p-FTAA, a distinct double peak of fluorescence is building up with higher concentration. This confirms the existence of fibrils in the sample. The fibrillation starts as the protein exits the SEC column, but the speed is greatly reduced when it is kept on ice.

0 2000 4000 6000 8000 10000 12000 14000 460 485 510 535 560 585 610 635 Int en sity Wavelength (nm)

p-FTAA

75 μM 50 μM 10 μM 5 μM 1 μM PBS

(35)

27

Figure 20. Fibrillation kinetics plot of Aβ1-42 at different protein concentrations. The emission wavelength is 540 nm.

The kinetics curve of the fibrillation process illustrated in Figure 20 is very similar looking to the sigmoid shapes portrayed in the literature. It clearly shows a lag phase, exponential phase and a stationary phase that is slowly decreasing which is characteristic for amyloid fibril formation.

Image filtration

Filtration of fly and mouse tissue images show that it is possible to greatly reduce the amount of background noise by applying the relative filter constructed in this project. Performing minor adjustments to the settings of the relative filter suggest that it is possible the filter can fit the purpose of filtrating background noise from any type of tissue. There seem to be similarities in the intensity profile within the specific tissue type, deciding the general settings of the filter, where minor adjustments have to be made for individual images. With some experience from the method one can with certainty in most cases decide which settings to use for a specific image by only looking at it, since the settings follow a predictable pattern depending on the contrast between dark and bright parts in the picture. The tissues analyzed in this project have two different intensity profiles, where the difference within the genotype is greatest in the case with the flies due to the great contrast between bright and dark parts in the images that is common for fly images. Mouse tissue is mostly disturbed by dark background noise, and therefore the settings for this tissue does not vary as much between pictures as for fly images. The heat map images presented in this section show the ratio matrix of 500 nm and 540 nm.

Fly tissue filter

In the fly brain tissue, the intensity range of interest is dependent on dark background areas and exo-skeletal parts that such as parts of the eyes that usually reflect bright light. If there are regions in the image that contain parts that reflect very high intensity noise the upper cut-off value of the filter range is set three standard deviations. The lower cut off value is usually somewhere between minus one standard deviation to the mean value.

0 2000 4000 6000 8000 10000 12000 14000 0 200 400 600 800 1000 1200

Fibrillation kinetics

(36)

28

Figure 21. Images representing brain tissue from Nsyb flies expressing Aβ1-42, 1-42 (double expressed Aβ1-42). The settings are the same for each image, where the lower cut-off value is -1σ, -0.5σ and μ in heat map 1,2 and 3, and the upper cut-off value is set to infinity. As seen in the heat map there are similarities between the different pictures for each setting. When the cut-off limit is closer to the mean, more pixels are removed from the noise but also from the tissue.

Minor adjustments between images taken of the same objects, with similar microscopy settings, are essential to get a good output. As illustrated in Figure 21 the effect of the filter is influenced by blurriness from fluorescent compounds that shape output from the filter. Comparing picture A and B to C, where the dark parts in picture C are blurred with yellow light only to a small extent, this become evident. In image A and B, the first setting has modest reduction on dark background noise, whereas in image C the lower cut-off limit of -1σ has almost no impact on reduction of dark background. This is because the yellow fluorescence that is evenly spread out on the whole picture except from the lower left corner.

A method for unbiased analysis of fluorescence microscope images of Alzheimer’s disease related amyloids