Raman and Fluorescence Imaging of Core Amyloid Beta Plaques in Human Alzheimer Diseased Brain Tissue

(1)

Raman and Fluorescence Imaging of Core Amyloid Beta

Plaques in Human Alzheimer Diseased Brain Tissue

Sander Verheul

December 9, 2020

STUDENT NUMBER 11689730

DAILY SUPERVISOR dr. Benjamin Lochocki SUPERVISOR dr. Freek Ariese 2ND_R

EVIEWER prof. dr. Marloes Groot STUDY Physics and Astronomy COURSE Bachelor Project; 15EC

(2)

Scientific Summary

Alzheimer disease is a cognitive condition where amyloid beta molecules accumulate in the brain tissue and form plaque deposits. The goal of this project is to identify these plaque locations using Raman spectroscopy. We found out that specific vibrational modes experience a Raman resonance enhancement (RRS). This is only visible in the plaque locations and not in background or lipofuscin tissue. The Raman enhancement of these location specific peaks corresponds to the Raman signal of carotenoids. In this project we examined the RRS signal of Lutein and Beta-Carotene with different excitation sources; 785 nm, 532 nm and 413 nm lasers. These carotenoids have a similar RRS enhancement at specific wavenumbers and therefore are a new location indicator of plaque formation in AD brain tissue.

Secondly, by using Stimulated Raman Scattering (SRS) we are able to determine and locate DNA, protein and lipid molecules in the brain tissue. This gives insight into the kind of molecules in and around a core amyloid beta plaque. We observed that plaque and lipofuscin locations predominantly consist of DNA and proteins, whereas surrounding tissue consist mainly of lipids and proteins. These results are measured using an in house built SRS setup with a pump laser of 1064 nm and an adjustable Stokes laser.

Another goal of this project is to determine the fluorescence spectrum of a core amyloid beta plaque compared to surrounding lipofuscin areas. We found out that plaques have a distinctive fluorescence spectrum which peaks around 540 nm when excited with a 488 nm source. The lipofuscin fluorescence signal peaks around 564 nm. In addition we determined that if the same location is excited with 405 nm the fluorescence signal of the plaque peaks around 447 nm and lipofuscin around 474 nm. However, the spectral shape of the fluorescence signal of the plaque looks significant different. Therefore we assume there are multiple fluorescent molecules present in the plaque location with different absorption and emission spectra.

Keywords: Alzheimer Disease, Amyloid Beta Plaque, Resonance Raman Spectroscopy (RRS), Stimulated Raman Scattering (SRS), Fluorescence

(3)

1 Introduction

Alzheimer’s disease (AD) is a chronic neurodegenerative disorder where over time brain functions are lost. Mainly elderly people suffer from it. It is common to experience cognitive function decline including memory, judgement, orientation and reasoning. In 2015 more than 46 million people world wide were living with Alzheimer’s disease. It is estimated that in 2050 this number will increase to 131 million (14). Currently there is no cure for AD.

Brain tissue consist of many different cells and molecules. Neurons, dendrites, stem cells, blood vessels and various glial cells. These glial cells contain lipofuscin deposits which are present in healthy and diseased brain tissue. When someone is suffering from Alzheimer’s disease amyloid beta molecules are being misfolded into anti-parallel beta sheets. These are insoluble and therefore hard to be cleared by the brain. These folded amyloid beta molecules might accumulate and form cored plaque locations. These plaques are currently the primary indication of AD, post mortem. Nowadays there is no in vivo detection method of plaque locations in AD brain tissue.

In this project, I investigated whether Raman spectroscopy can be used to detect plaque locations in ex vivo human brain tissue. Different excitation source wavelengths are used at 785 nm, 532 nm and 413 nm. In this regard I identified the specific Raman resonance enhanced vibrational modes. To get more information about the surrounding tissue and molecules inside the plaque, I measured stimulated Raman spectra to examine if DNA, lipid and protein molecules are present. Beside this I investigated the fluorescence spectra of the plaque locations with 405 nm and 488 nm excitation wavelengths.

(5)

2 Theory and Methods

2.1 Tissue preparation

Post mortem brain tissue was provided by the pathology department of the Amsterdam UMC hospital. Brain tissue was snap frozen and cut at 20 micron thickness and placed on superfrost microscope slides. The investigated area was the gray matter from the frontal lobe. The donor was a female of 90 years and had cored amyloid beta deposits.

2.2 Raman Spectroscopy

In spectroscopy the interaction between light and matter is studied. This interaction depends on the energy of the incident photons of the light. The energy of a photon is depending on the wavelength and is given by the following equation (1).

E =hc

λ (1)

In this Equation h is the Planck constant with a value of 6,62607015 * 10-34_{J*s. c is the speed of}

light in a vacuum. This is constant at 299792458 in m/s. Finally, the energy is inversely proportional to the wavelength of the light (λ). The wavelength is the spatial period of a periodic wave. λ is given in meters.

In spectroscopy photons interact with all types of atoms, molecules and crystals. The type of interaction can be measured and give us information about the matter. One type of interaction is absorption. This is a phenomena where the matter takes up the energy of the photon. It transforms the electromagnetic energy to internal energy for instance thermal energy. The amount of energy that is absorbed is molecule specific.

Another type of interaction is scattering. When there is no energy exchange between the matter and the photon this is called elastic scattering, otherwise it is called inelastic scattering. An example of elastic scattering is Rayleigh scattering. Examples of inelastic scattering are Stokes Raman and Anti-Stokes Raman scattering. Both can be best explained with a Jablonski diagram in Figure 1.

In Figure 1 we see the ground state S0 and the first excited singlet state S1. There is an energy difference between them. Scattering photons can excite a molecule into a so called ’virtual excited state’. This is a virtual state between the S0 and S1 states. When the molecule falls back to the absolute ground state S0 there is no energy exchange and the photons will scatter elastically. This is called Rayleigh scattering, see center transition in Figure 1. When the molecule falls back to a vibrational state, above the ground state, there is an energy exchange and the photons will scatter inelastically. This is called Raman scattering. There are two types of Raman scattering, Stokes and anti-Stokes Raman. When the photon has less energy than before the scattering at the virtual state it is called Stokes Raman scattering. Conversely it is called Anti-Stokes Raman scattering. Anti-Stokes Raman happens when the molecule is already in a vibrational mode. Because of Boltzmann statistics we know that molecules are less likely to be in this vibrational ground state

(6)

Figure 1: Jablonski diagram of S0 and S1 electronic state with vibrational levels. Photons can excite a molecule into a virtual excited state. Then the molecule falls back to its ground state (S0). If there is no energy exchange the photon is Rayleigh scattered, see center transition. If the molecule absorbs part of the energy of the photon, it is Stokes Raman scattered, see right transition. If the molecule provides energy the photon is anti-Stokes Raman scattered, see left transition. (Courtesy Maaike Straathof, accessed 10 November 2020) (3).

and consequently it is less likely to see Anti-Stokes Raman scattering than regular Stokes Raman scattering.

To measure the Raman signal we look at the energy difference between the induced laser light and the measured scattering of the photons. In order to achieve this we need a filter to block the laser light. This is commonly achieved with a sharp long-pass filter. Consequently we also block the Anti-Stokes Raman signal. We choose to see the Stokes Raman signal because it is more likely to occur and therefore stronger.

The difference in energy and hence the difference in wavelength is called the Raman shift. This shift is normally displayed in wavenumbers with a unit of cm-1. The Raman shift∆ν is given by Equation 2. ∆ν = (_λ1 0− 1 λ1) ∗ 10 7 ₍₂₎

λ0 is the wavelength of the laser light in nm,λ1 is the wavelength of the inelastically scattered

photons in nm.

In this research project I used the inViaTM confocal Raman microscope from Renishaw. This setup has two lasers, at 532 nm and 785 nm. The laser light is focused onto the sample with an objective. The Renishaw has multiple objectives ranging from 5x, 20x, 40x and 63x. The scattered light is collected with the same objective. Then it is passed through a filter and optical grating. This is specific for each laser wavelength. After the grating the light is captured by a CCD camera and displayed on the PC as a Raman spectrum.

(7)

2.2.1 Resonance Raman Spectroscopy (RRS)

In this research project I make use of a special enhancement effect of specific Raman signals. When the virtual exited state is getting close to the actual electronic transition state specific Raman signals are strongly enhanced. This is called the Resonance Raman Spectroscopy effect (RRS). The electronic transition of a molecule changes the structure of the molecule and hence changes the bond length or bond forces of a molecule. Virtual states which get really close to these structural changes can enhance specific vibrational modes. It is shown that this effect can increase the intensity of specific Raman peaks from ten to thousands of times (5).

RRS is in this project achieved by changing the excitation source wavelength. The Renishaw has two lasers at 532 nm and 785 nm. Both excitation RRS measurements are recorded making 10 accumulations with a measurement time of 20 seconds. The laser power is in both cases 100%, 532 nm at 60 mW and 785 nm at 82 mW. Another additional in house built Raman setup has a laser wavelength of 413.1 nm. This setup is in addition to the Renishaw also used measuring the RRS signal of Lutein and Beta Carotene dissolved in hexane, depicted in Section 3.2.1. Measurements are performed using 20 accumulations with a measurement time of 20 seconds. The laser power was 0.6 mW at the sample plane.

2.2.2 Stimulated Raman Scattering (SRS)

Another type of Raman scattering used in this project is Stimulated Raman Scattering (SRS). This special scattering process makes use of two different picosecond lasers. Firstly the sample gets illuminated by a ’pump’ photon. This has the same effect as spontaneous Raman scattering and roughly one in every 1 million photons will be scattered inelasticlly. However, a second ’Stokes’ photon is also focused on the sample. When the energy difference between the pump and Stokes photon corresponds to a specific vibrational mode the Raman signal of this mode can be highly enhanced. For this to happen both laser beams needs to be focussed at the same time and place on the sample.

The SRS setup in this project uses a Stokes laser of 1064 nm and an adjustable pump laser. The Stokes laser has a repetition rate of 80 MHz and around 15 mW power at the sample plane. The pump laser has an adjustable wavelength due to the ’Optical Parameric Oscillator’ (OPO). Inside the OPO, a piëzo element and temperature control let us choose the wavelength of the pump photon. In this project the pump laser has a wavelength of 808.9, 811.6, and 816.6 nm or 2967, 2926, 2850 cm-1. This covers the C-H stretch vibrations of DNA, Protein and Lipid vibrational modes in human brain tissue according to Fa-Ke Lu (2). The laser power at the sample plane is kept constant at 50 mW. Both beams are focussed with a Zeiss, C-Achroplan W, 32x, 0.85 NA, water immersion objective onto the sample. The 1064 nm laser is blocked with a dichroic mirror and a shortpass filter. The SRS signal is collected with an ADET36A photodetector. Then the remaining signal is demodulated using an HF2LI lock-in amplifier (Zurich instruments).

(8)

2.3 Fluorescence

Fluorescence is another spectroscopic entity. When a photon with enough energy is absorbed by a molecule it could excite this molecule to a higher excited state (S1), see Figure 2. After some time (nanoseconds) the molecule falls back to its ground state (S0) through the emission of photons, this is called fluorescence. The molecule can relax into different vibrational modes of the ground state. Due to matrix effects the emitted wavelength of the photons are broadened. The difference in energy of the fluorescence compared to the absorption of the molecule is called the stokes shift. The fluorescence, emission, curve has lower energy, higher wavelength, than the absorption curve of a molecule, see Figure 3. Because of this shift we are able to see the fluorescence signal of molecules with the use of a simple filter.

Figure 2: Photons can excite a molecule to an excited state S1, this is called absorption (blue arrows). After nanoseconds (10-7-10-10 sec.) the molecule can relax into different vibration modes of the ground state S0 through the emission of photons, this is called fluorescence (green arrows).

Figure 3: Absorption and emission (fluorescence) spectra of a molecule. Photons at different wavelengths are absorbed and excite the molecule in an excited state. After nanoseconds the molecule falls back into different vibrational modes of the ground state with the release of photons. They form the broad emission, fluorescence curve. The difference in energy between the absorption and emission curve is called the stokes shift. (Courtesy Wikipedia accessed 10 November 2020)

(9)

2.3.1 Plaque location

To obtain the location of a plaque in brain tissue we acquired fluorescence images with a Leica DM2000 fluorescence microscope. The sample is uniformly illuminated with a blue LED of 470 nm with the use of different objectives of 5x, 10x, 20x and 40x. The reflective light is collected and passes a dichroic longpass filter at 500 nm. Everything below 500 nm is removed from the signal including the LED. The remaining signal is detected by a Leica DFC450 C camera and displayed on a PC. The plaque area is fluorescent in the green whereas lipofuscin appears in the orange. This instrument can not provide more specific spatial information. The background tissue is dark and basically not fluorescent. This makes the plaque deposit clear to distinguish and to make a scratch in nearby tissue that will help us find the same spot in different measurement set-ups again.

2.3.2 Fluorescence spectrum

To measure the distinguishable fluorescence curve of the plaque location a Nikon A1 spectral scanning microscope was used. Here, the tissue was raster scanned with a 405 nm and 488 nm excitation source. The 488 nm wavelength was chosen because it is close to the illumination wavelength of the LED in the Leica fluorescence microscope, which we used to find plaque locations. With these settings we expected to see similar green fluorescence in plaques as we saw before. The 405 nm source was chosen to examine if the plaque might be fluorescent below 500 nm. This is where the obligated filter is located in the Leica microscope.

In the Nikon setup the laser light is focused with a 20x, 0.8NA objective onto the sample. The fluorescence light is collected after a filter where the laser light is removed. The light enters a spectrometer and is detected by 32 different emission channels all 6 nm apart. From the intensity in the different channels a fluorescence spectrum can be obtained.

(10)

2.4 Thioflavin-S staining

Staining is a technique used to bring contrast in samples. Thioflavin-S is a staining agent which is fluorescent and binds to amyloid beta molecules. Therefore it can be used to confirm plaque locations in AD brain tissue. The staining agent makes the sample unusable for further Raman measurements, therefore this is the final step performed on the samples in this project.

In this research the brain tissue is fixed by immersing the sample in a 4% formalin solution for 10 minutes. This is performed to secure the tissue on the microscope slide. The excess solution is washed off. This is followed by an incubation into a 1% Thioflavin-S solution for 10 minutes. Now the fluorescent agent has been bound to the amyloid beta plaque locations. Afterwards, the remaining Thioflavin-S is rinsed off using 70% alcohol. Then the tissue sections are prepared into a TBS/glycerol medium and covered with a coverslip. The fluorescent Thioflavin-S can be seen using the 470 nm LED microscope. It appears bound to the amyloid beta plaques as bright yellow fluorescence. The lipofuscin fluorescence signal is almost invisible because of the more intense Thioflavin-S fluorescence.

2.5 Data processing

When spectral data is received from one of the Raman instruments the following data processing steps were performed. First, cosmic rays are removed. They were excluded using the matlab function ’filloutliners’. Cosmic rays are easy to detect because they differ significantly from the Raman signal. Secondly, a baseline is removed using the arPLS function (3). This removes most of the fluorescence signal beneath the Raman spectrum. When the 413 nm laser for Raman measurements was used, the signal is analytical smoothed using a weighted average with 7 data points. All data processing steps are performed using the matlab graphical user interface designed by Maaike Straathof (3).

The fluorescence data from the Nikon A1 spectral scanning microscope is analysed using a 3rd or 5thorder polynomial fit. To quantify this fit a R2test is performed. The closer the R2value to 1 the better the fit. The polynomial fit and R2test are calculated using OriginPro 2019.

(11)

3 Results

The results are presented in the following order: First I confirm that we actually look at plaque locations in the brain. This is done by comparing the fluorescence images of plaque locations prior any staining with the thioflavin-S stained tissue at the same location with the same microscope and settings. This can be seen in section 3.1.

After this, I present Raman spectra of plaque locations. In section 3.2 the resonance effect of specific vibrational modes is shown. In section 3.2.1, I identify these specific modes as carotenoids. Here I will also take a closer look at their particular absorption and Raman spectra.

In section 3.3, SRS images are shown to examine if DNA, protein and lipids are present in the surrounding tissue and/or plaque locations. I present stacked z-scans of the individual structures and combinations of the three to see where which structure is present.

Finally, in section 3.4 I present the fluorescence spectra of different plaques recorded with the Nikon A1 spectral scanning microscope. I make a distinction between the 488 nm excitation and the 405 nm excitation source. Different fluorescence spectra are observed under the different excitation sources.

3.1 Confirmation of plaque locations

In Figure 4 I compare the fluorescence images of unstained brain tissue with the images of Thioflavin-S stained tissue of the same location. Thioflavin-Thioflavin-S is a staining agent which binds to amyloid beta molecules. In these Figures one can clearly see that the stained yellow bright spot in Figure 4b, 4d and 4f corresponds to the green fluorescence locations in Figures 4a, 4c and 4e. Therefore, we confirm that the green fluorescence is coming from the same area as the amyloid beta core plaques. Both images are made using the 470 nm LED leica microscope and 20x objective. The scale bars are 100µm in all images. In section 3.2, 3.3 and 3.4 the fluorescence images were acquired with the 40x objective.

(12)

(a) Fluorescence image #1 (b) Thioflavin-S stained image #1

(c) Fluorescence image #2 (d) Thioflavin-S stained image #2

(e) Fluorescence image #3 (f) Thioflavin-S stained image #3

Figure 4: Comparison between unstained and stained brain tissue. In (a,c,e) we see the unstained fluorescence image of AD brain tissue, location #1, #2 and #3. In the center of the images a green fluorescent deposit is visible. We assume these are plaque locations. Surrounding orange fluorescent locations are also visible. They are coming from lipofuscin locations. In (b,d,f) we see the Thioflavin-S stained tissue of the same location. In bright yellow the amyloid beta plaque locations are fluorescent because of the staining, confirming that the green fluorescent locations in (a,c,e) are indeed plaques. All images are made with the 470 nm LED leica microscope with an 20x objective. Scale bar is 100µm.

(13)

3.2 Resonance Raman Spectroscopy (RRS)

To examine the Raman signals of plaque locations we used the Renishaw Raman microscope. We illuminated the brain sample containing core amyloid beta plaques with 532 nm and 785 nm. When using the 785 nm laser there is no clear Raman signal visible, as shown in (10). However when illuminating the same spot with a 532 nm laser it is possible to distinguish the Raman signal of the plaque from background or lipofuscin. This is because strong resonance enhanced Raman signals are occurring. We observe 3 enhanced peaks in the plaque locations which are not visible in the background or lipofuscin. Those peaks occur at wavenumbers 1512, 1149 and 1002 cm-1. See Figures 5b, 5d and 5f. In Figures 5a, 5c and 5e the selected areas of the plaque and background or lipofuscin locations are presented.

In all three cases we see a strong RRS signal coming from the plaque location. The RRS is visible at three different wavenumbers. The background or lipofuscin signal is not the same in the 3 different cases. This is because this tissue consists of different molecules and cells which were accumulated and give different Raman signals. However, in all background signals there is a strong Raman protein peak around 1660 cm-1. This peak is also visible in the plaque regions. Hence the different spectra are normalized on this peak to be able to compare them.

(14)

(a) Fluorescence image #1, excitation: 470 nm LED

(b) Raman signals #1

(c) Fluorescence image #2, excitation: 470 nm LED

(d) Raman signals #2

(e) Fluorescence image #3, excitation: 470 nm LED

(f) Raman signals #3

Figure 5: Core amyloid beta plaque in AD brain tissue. Plaque location #1, #2 and #3. Raman signals of plaque and lipofuscin in (b,d,f) are normalized at the 1660 cm-1peak present in all signals. Data processing performed as explained in Section 2.5. A strong RRS effect in the Raman spectrum of the plaque regions at 1001-2, 1149 and 1511-13 cm-1 can be observed. In (a,c) the region of interests of the plaque and background locations are presented in the blue circles. In (e) a lipofuscin area is selected instead of background. 14

(15)

3.2.1 Carotenoids

Carotenoid compounds could be the reason for the RRS peaks at 1002, 1149 and 1513 cm-1 found in the Raman spectrum recorded at the plaque locations. Previous studies show that carotenoids like Lutein and Zeaxanthin are present in the human brain (11). Whereas Beta-Carotene is used as a reference spectrum for carotenoids. In this project I compared the RRS spectrum of Lutein and Beta-Carotene with the obtained spectra from the plaque. In Figure 6 the molecular structure of Lutein and Beta-Carotene is presented.

Figure 6: Molecular structure of Lutein and Beta-Carotene

To examine Lutein and Beta-Carotene they are dissolved in hexane. Lutein’s concentration is 1.63*10-5 _{M, Beta-Carotene’s concentration is 1.76*10}-5 _{M. The available amounts were too small}

for weighting, therefore the concentration is calculated following the Lambert-Beer law using the absorbance values of the solutions. The absorptivity (²) of Lutein is 1.47*105 (L*M-1*cm-1), Beta-Carotene’s² is 1.39*105(L*M-1*cm-1), when dissolved in hexane (13). In Figure 7 the absorbance spectrum of Lutein and Beta-Carotene are presented.

Figure 7: Absorbance spectrum of Lutein and Beta-Carotene. Both spectra have similar absorbance spectra. Arrows correspond to the laser wavelengths which were used to examine the RRS Raman signals. 1 cm quartz cuvette, reference spectrum of hexane.

(16)

In Figure 7 we see three different arrows which correspond to the laser wavelengths used to measure the Raman and possible RRS effects of Lutein and Beta-Carotene. As explained in Section 2.2.1, the Raman intensity of specific vibration modes are enhanced if the virtual state gets closer to the actual first excited state. The absorbance spectrum gives information about where this state is. 413 nm laser light is the closest wavelength to the peak of the absorbance spectrum and therefore the first excited state, followed by 532 nm. Both wavelength are expected to induce a RRS enhancement on the carotenoids. 785 nm is located far away from this excited state peak and therefore will not enhance the vibration modes of the carotenoids (10).

In Figures 8a and 8b we see the Raman spectra of Lutein and Beta-Carotene dissolved in hexane excited with three different wavelengths. All spectra are normalized on the big 1450 cm-1 Raman peak of hexane. This is the most noticeable ’neutral’ peak and therefore used for normalization. The signal is processed as explained in Section 2.5. The concentration of Lutein is 1.63*10-5M, the concentration of Beta-Carotene is 1.76*10-5M.

In Figure 8a we see a clear RRS enhancement effect of three peaks at 1007, 1159 and 1528 cm-1 in the Lutein solution. The shorter the excitation wavelength the greater the enhancement of these peaks. The peak at 1528 cm-1has the greatest enhancement. Between the use of 532 nm and 785 nm this peak is enhanced 30 times. Between the use of 413 nm and 532 nm 8.5 times and between 413 nm and 785 nm we see an enhancement of 225 times.

In figure 8b we see the Raman signal of Beta-Carotene dissolved in hexane, excited with three different wavelengths. The results are similar to Figure 8a. In Figure 8b a clear RRS enhancement effect of three peaks at 1008, 1160 and 1527 cm-1 is evident. Again, the shorter the excitation wavelength the greater the enhancement of these peaks. The peak at 1527 cm-1 has the greatest enhancement. Between the use of 532 nm and 785 nm this peak is 38 times enhanced. Between the use of 413 nm and 532 nm 5.5 times and between 413 nm and 785 nm we see an enhancement of 209 times.

In the Supplementary Information I present the fluorescence signal of Lutein dissolved in hexane illuminated with a 450 nm excitation source. Measurements are performed using a Cary Eclipse Spectrofluorimeter, see Figure 16. The red line shows the fluorescence signal of the Lutein solution dissolved in hexane with a concentration of 6.871*10-6M. The black line is the emission spectrum of pure hexane. Due to Raman background the blue line shows the calculated spectrum of pure Lutein. This is calculated after scaled subtraction of the hexane background (black) from the Lutein solution dissolved in hexane spectrum (red).

(17)

(a) Raman signal of Lutein dissolved in hexane, C = 1.63*10-5M.

(b) Raman signal of Beta-Carotene dissolved in hexane, C = 1.76*10-5M.

Figure 8: Raman signals of Lutein and Beta-Carotene excited with 785 nm, 532 nm and 413 nm. Signals are normalized on the 1450 cm-1peak of hexane. A clear RRS enhancement of 3 peaks at 1007-8, 1159-60 and 1528-27 cm-1can be observed.

(18)

For general comparison, the enhanced Raman vibrational modes in the plaque spectrum presented in Figures 5b, 5d and 5f are at 1002, 1149 and 1512 cm-1. Whereas the enhanced vibrational modes of Lutein are at 1007, 1159 and 1528 cm-1and Beta-Carotene at 1008, 1160 and 1527 cm-1. From this we can conclude that the enhanced RRS peaks in plaque locations are due to carotenoids. Which carotenoid is responsible for this is hard to say, since both, Lutein and Beta Carotene have a very similar RRS Raman signal. From previous studies it is known that Lutein is present in the human brain whereas Beta-Carotene is not (11). Thus it is more likely we see Lutein’s RRS signal in Figure 5b, 5d and 5f in the plaque locations. The slight changes in wavenumber is probably due to the surrounding liquid environment. The Raman signal is known to slightly change depending on its surrounding.

Due to photo bleaching I was not able to measure Raman signals and possible RRS signals of the AD brain tissue using the in house built 413.1 nm excitation microscope. Due to this higher energy per photon the risk of destroying the brain sample is high. This is not a problem when determining Raman spectra of carotenoids in hexane solutions. In solution the molecules are freely moving around and are not fixed on a microscope slide as in the AD brain samples.

(19)

3.3 Stimulated Raman Scattering (SRS)

To obtain more information about what types of molecules are present in and around the plaque locations, stimulated Raman scattering (SRS) is performed. The Stokes laser has a wavelength of 1064 nm. The adjustable pump laser had in order to enhance DNA vibrational modes a wavelength of 808.9 nm, for protein 811.6 nm and for lipids 816.6 nm. Or respectively in wavenumbers: 2967, 2926 and 2850 cm-1.

The SRS system performs measurements in different z planes, all 1µm apart. The laser intensities were kept constant, 1064 nm at 15 mW and the pump laser at 50 mW, at the sample plane. The dwell time was 177.32µs per pixel with two averages of the image. The z planes were maximum intensity stacked together and shown in Figure 9. In this Figure we see images of the three stokes wavelengths specific molecule locations in and around the plaque. The images have a size of 250 x 250µm.

(a) DNA - 2967 cm-1 (b) Protein - 2926 cm-1 (c) Lipid - 2850 cm-1

Figure 9: Maximum intensity stacked image of 49 z-planes, 1µm apart. Three different Stokes wavelengths are induced to enhance different molecule specific vibration modes. (a) DNA locations, (b) protein locations and (c) lipids locations with respectively 2967, 2926 and 2850 cm-1.

In Figure 10 the fluorescence image made with the 470 nm LED and a combination of the three different SRS signals of Figure 9 is shown. Based on our experience, in Figure 10a we see in green fluorescence, at the center of the image the plaque location. The surrounding orange fluorescence spots are lipofuscin locations. The background is not fluorescent and therefore black. In Figure 10b the three images of Figure 9 are combined into a RGB image. Protein is Red, DNA is Green and lipids are Blue. We see that the lipofuscin and plaque are predominantly yellow, which means that it mostly consist of Proteins and DNA. The background tissue is purple, which means that it consist of Lipids and Proteins.

(20)

(a) Fluorescence image (b) RGB image of protein (R), DNA (G) and lipids (B)

Figure 10: Comparison between, (a) fluorescence image of plaque area with an excitation of 470 nm LED and (b) RGB image of protein (Red), DNA (Green) and lipids (Blue). In (a) we see the plaque is fluorescent in green and lipofuscin in orange. In (b) we see the plaque and lipofuscin areas consist predominantly of protein and DNA. The background consist of lipids and protein.

In Figure 11 we perform the same SRS experiment, as shown earlier, on a different plaque. Instead of 49 z-planes this plaque was measured with 39 z-planes at 1µm apart. The dwell time, laser wavelengths and intensity, averaging and type of plaque are the same. Again, in Figure 12 the fluorescence image and stacked RGB image of the three different compounds, now in the different plaque location is shown. Morelike in Figure 12, we see that the plaque and lipofuscin areas are yellow which indicates they consist of proteins and DNA. The background tissue is purple which tells us that it consist of proteins and lipids. These are similar results compared to the previous plaque in Figure 10.

(21)

(a) DNA - 2967 cm-1 (b) Protein - 2926 cm-1 (c) Lipid - 2850 cm-1

Figure 11: Maximum intensity stacked image of 39 z-planes, 1µm apart. Three different stokes wavelengths are induced to enhance different specific molecule vibration modes. (a) DNA locations, (b) protein locations and (c) lipids locations with respectively 2967, 2926 and 2850 cm-1.

(a) Fluorescence image (b) RGB image of protein (R), DNA (G) and lipids (B)

Figure 12: Comparison between, (a) fluorescence image of plaque area with a 470 nm LED and (b) RGB image of protein (Red), DNA (Green) and lipids (Blue). In (a) the plaque is fluorescent in green and lipofuscin spots are orange. In (b) we see the plaque and lipofuscin areas consist predominantly of proteins and DNA. The background consist of lipids and proteins.

(22)

3.4 Fluorescence spectrum

To evaluate the fluorescence spectrum of amyloid beta core plaques, three brain samples were prepared and measured with the Nikon A1 spectral scanning microscope. The samples were excited with 488 nm and 405 nm. The 488 nm has a similar excitation wavelength as the LED 470 nm used to locate plaques for the Raman measurements. The 405 nm is used to get a complete overview of the spectrum, also below 500 nm which is not visible in the LED 470 nm microscope because it is below the excitation wavelength.

In the spectral detector 32 channels measure the fluorescence spectrum of the brain tissue at 6 nm apart. With the use of the 488 nm excitation we end up with a spectral range between 507 nm and 699 nm. The 405 nm excitation will give a spectral range between 415 nm and 607 nm. Both excitation measurements use a dichroic mirror to remove the laser wavelength from the signal. The difference in excitation wavelength and first channel of the spectral detector is another precaution step to eliminate the laser light even better. Both excitation measurements are made using a 20x, 0.8 NA objective, laser power at 1 mW, a pinhole of 0.9µm, 8 averages and a pixel dwell time of 2.4 µs.

In Figure 13 the core plaque locations #1, #2 and #3 are shown. In (a,c,e) we see the fluorescence images made with the 470 nm LED. In green we see the fluorescence of the plaque and in orange bright spots the surrounding lipofuscin. In (b,d,f) we see the grey-scaled, stacked images of the 32 channels from the spectral detector. In red circles the region of interest are selected for the upcoming results in Section 3.4.1 and 3.4.2. 1 is the selected area of the plaque, 2 is the lipofuscin and 3 is the background signal.

When we illuminate amyloid beta core plaques with 488nm and 405nm we see different fluorescence spectra for each excitation wavelength. This indicates that there are different fluorescent molecules present in the plaque which have different absorption and emission spectra.

(23)

(a) Fluorescence image #1 (b) Grey-scaled, stacked image of the 32 channels #1

(c) Fluorescence image #2 (d) Grey-scaled, stacked image of the 32 channels #2

(e) Fluorescence image #3 (f) Grey-scaled, stacked image of the 32 channels #3

Figure 13: Amyloid beta core plaque location #1, #2 and #3. In (a,c,e) the fluorescence images made with the 470 nm LED are presented. In green the plaque and in orange the lipofuscin locations are visible. In (b,d,f) the grey-scaled, stacked image of the 32 channels measured 6 nm apart are presented. Three regions of interest are selected, 1 the plaque, 2 the lipofuscin and 3 the background location.

(24)

3.4.1 488 nm excitation

In Figures 14a, 14b and 14c I present the fluorescence spectrum of plaque #1, #2 and #3 and lipofuscin locations with an excitation of 488 nm. The intensity at the different channels, wavelengths, are presented of the selected regions of interests seen in Figure 13 (b,d,f). In Figures 14a, 14b and 14c we see in green the fluorescence spectrum of the plaque and in orange the fluorescence spectrum of lipofuscin. The raw data points are fitted with a 3rdorder polynomial. A R2test is performed to evaluate the quality of this fit using OriginPro software. In black we see the raw data points of the background.

In Figure 14a the peak intensity of the plaque is at 541 nm, the R2value of the fit is 0.980. The peak intensity of the lipofuscin is at 567 nm, the R2 value of the fit is 0.947. In Figure 14b the peak intensity of the plaque is at 541 nm, the R2value of the fit is 0.977. The peak intensity of the lipofuscin is at 564 nm, the R2value of the fit is 0.963. Finally, in Figure 14c the peak intensity of the plaque is at 539 nm, the R2value of the fit is 0.979. The peak intensity of the lipofuscin is at 560 nm, the R2value of the fit is 0.962.

The averaged results from Figures 14a, 14b and 14c give the fluorescence peak of a plaque location, with an excitation of 488 nm, at 540 nm. The fluorescence spectrum of lipofuscin peaks at 564 nm.

(a) Fluorescence spectrum plaque location #1

(25)

(b) Fluorescence spectrum plaque location #2

(c) Fluorescence spectrum plaque location #3

Figure 14: Intensity at different wavelengths recorded 6 nm apart. Data and regions of interests selected from plaque location #1, #2 and #3 as presented in Figure 13. In green the fluorescence spectra of the selected plaque areas are presented. In orange the fluorescence spectra of lipofuscin. Raw data points are fitted with a 3rdorder polynomial. In black the raw data of the background are presented.

(26)

3.4.2 405 nm excitation

In Figures 15a, 15b and 15c I present the fluorescence graphs of plaque locations #1, #2 and #3, excited with 405 nm. The intensity at the different channels, wavelengths, are presented of the selected regions of interests seen in Figure 13. In blue we see the fluorescence spectrum of the plaque and in green the fluorescence spectrum of lipofuscin. Raw data points are fitted with a 5th order polynomial. A R2test is performed to evaluate the quality of this fit using OriginPro software. In black we see the raw data points of the background.

When using the 405 nm excitation laser a filter is used to remove the laser light. Unfortunately the 405 nm filter was always combined with a 488 nm filter. This makes specific wavelength channels unusable. To correct for this, 4 data points around the 488 nm wavelength channel are masked and not included in the fit. These points are marked as red in the graphs.

In Figure 15a the peak intensity of the plaque is at 454 nm, the R2 value of the fit is 0.984. The peak intensity of the lipofuscin is at 482 nm, the R2value of the fit is 0.983. In figure 15b the peak intensity of the plaque is at 442 nm, the R2value of the fit is 0.978. The peak intensity of the lipofuscin is at 463 nm, the R2value of the fit is 0.990. Finally, in Figure 15c the peak intensity of the plaque is at 445 nm, the R2value of the fit is 0.981. The peak intensity of the lipofuscin is at 476 nm, the R2value of the fit is 0.985.

The averaged results from Figures 15a, 15b and 15c give the fluorescence peak of a plaque location, with an excitation of 405 nm, at 447 nm. The fluorescence spectrum of lipofuscin peaks at 474 nm.

(a) Fluorescence spectrum plaque location #1

(27)

(b) Fluorescence spectrum plaque location #2

(c) Fluorescence spectrum plaque location #3

Figure 15: Intensity at different wavelengths recorded 6 nm apart. Data and regions of interests selected from plaque location #1, #2 and #3 as presented in Figure 13. In blue the fluorescence spectra of the selected plaque area is presented. In green the fluorescence spectra of lipofuscin. Raw data points are fitted with a 5thorder polynomial. In black the raw data of the background is presented.

(28)

From the different excitation wavelength results presented in section 3.4.1 and 3.4.2 I expect that multiple fluorescent molecules are present in the plaque location. The plaque consist of an accumulation of different molecules, which are fluorescent under different excitation wavelengths. This can be seen in the different emission spectral shape of the raw data points in graphs 14 compared to graphs 15. This spectral shape is substantially different especially in the beginning of the fluorescence spectrum for the different excitation wavelengths. This difference might be due to the different absorption spectrum of the various molecules present in the plaque location. Beside this, the relative difference in intensity between the plaque and lipofuscin locations at the various excitation wavelengths are also an indication of the presence of different molecules. Also, if we would see only one fluorescent molecule in the plaque location the fluorescence spectrum will then be at the same wavelength, independent of the excitation wavelength. Due to internal relaxations above the first excited state. This is not the case. However, more thorough measurements need to be done in the future to verify this.

(29)

4 Discussion

In this project we were able to distinguish plaque locations using a 532 nm excitation laser in the Renishaw Raman microscope. However I was not able to perform RRS measurements on brain tissue with the in house build 413 nm laser. This is partly due to some limitations in this set-up. The highest magnification is a 40x objective in the 413 nm laser, whereas the Renishaw has a 63x objective. Next to this, the 413 nm microscope is less sophisticated in collecting all the light. In this setup a glass dichroic mirror, which depends on the angle of the light’s wavefront, is responsible for removing the laser wavelengths. The Renishaw uses a specific wavelength dichroic filter, which fits the laser wavelengths perfectly. This will deteriorate the SNR of the spectrum. This is visible in Figures 8a and 8b. The SNR is worse for the 413 nm laser compared to the 532 nm and 785 nm laser.

Beside this, the brain tissue was fixed on superfrost microscope slides. These slides have an unwanted strong fluorescence signal. Therefore, when performing RRS measurements, a longer measurement time with more averages is needed to collecte a proper Raman signal. When using the 413 nm laser, this longer measurements time or more averaging can destroy the tissue, due to photo bleaching. This was not necessarily a problem when using the 532 nm or 785 nm lasers in the Renishaw. Brain tissue is a delicate sample and risking to destroy tissue while other spectroscopy measurements still need to be performed was not desired.

Receiving new brain tissue on different microscope slides was pretty hard. Because of Covid-19 rules less people are allowed to be present during working days in the pathology lab of the VU. Therefore they consequentially have less time to cut and fixate tissue, beside their own research.

Next to this, in Figures 14a, 14b and 14c the fluorescence signals are recorded with 32 channels which are 6 nm apart. This will result in a spectral range from 507 nm until 699 nm. However, in these results we see a steep drop off in intensity around the 621 nm channel and no clear signal any further. This is because the detection resolution gets significantly worse the longer the wavelength. In this wavelength range, no clear difference between the plaque, lipofuscin and background intensity can be detected. Therefore, I omitted this part of the fluorescence spectrum and it is not shown in Figures 14a, 14b and 14c.

In Figures 9, 10, 11 and 12 the chosen wavenumbers to enhance the vibrational modes of DNA, protein and lipid in tissue via SRS, are based on F.K. Lu’s paper (2). In Figure 1 of this paper the C-H stretch region of these different molecules are shown. Ideally, we would want these modes to be as distinctive as possible. However, the peaks are up to a certain point correlated with each other. Therefore, the SRS images are not a pure enhancement of one mode but a combination of multiple. However, as F.K. Lu shows, it is still possible to enhance the different modes with the given wavenumbers and distinguish the different molecules.

(30)

The fluorescence of Lutein shown in Supplementary Information Figure 16 could be one of compounds responsible for the green fluorescence at the plaque location. The emission spectra of Lutein fits the spectral data of the 488 nm excitation source. However, Lutein dissolved in hexane is poorly fluorescent and the environment, which influences the emission spectra, is completely different. More research needs to be done to confirm which molecules are responsible for the fluorescent signal of the plaque.

Furthermore, I hypothesize from the SRS results in Figure 10 that the core amyloid beta plaque could be a hollow core. The stacked images of the AD brain tissue indicates that the plaques consist of a DNA and protein core. However when I look at a single 1µm slice I could propose that it might be hollow inside. To advocate with more certainty, thicker slices of brain tissue with a smaller step size resolution in the z direction is needed. See the Supplementary Information in Figure 17 and 18 for the 1µm single z-scan of the imaged core plaque. In these galleries, the protein and DNA seem to form a ring like structure which is not visible a fewµm above and below this center slice (Figure 17(c, d) and 18(c)). This all indicates the plaque might be indeed hollow. Additional DNA, protein and lipid staining could also be performed to determine where these molecules are present and if indeed core amyloid beta plaques are hollow.

(31)

5 Conclusion

To conclude, with Raman scattering we were able to identify specific plaque locations in fixed AD brain tissue. When examining the Raman spectrum of plaque locations with the 532 nm laser, molecule specific RRS peaks are visible which are not visible with the 785 nm laser. These RRS peaks correspond to the Raman spectra of carotenoid molecules. We investigated the RRS spectra of both Lutein and Beta-Carotene with 785 nm, 532 nm and 413 nm lasers. Both molecules see a significant RRS effect at specific wavenumbers analogous to the RRS peaks of plaque locations in AD human brain. Therefore the RRS spectra of an accumulation of carotenoids could be used to determine the plaque location in AD human brain tissue.

Furthermore, when inspecting the SRS spectra of AD brain tissue at 2967, 2926 and 2850 cm-1for individual molecules. The SRS spectra shows that plaque and lipofuscin locations consist predominately of proteins and DNA molecules. Whereas surrounding tissue consist mostly of lipids and proteins.

Finally, plaque locations have a clear fluorescence signal which is different from the surrounding brain tissue and lipofuscin locations. When exciting the tissue with 488 nm the averaged fluorescence peak of a plaque location is at 540 nm. The averaged fluorescence spectrum of lipofuscin peaks at 564 nm. When exciting the AD brain tissue with 405 nm the averaged fluorescence peak of a plaque location is at 447 nm. The averaged fluorescence spectrum of lipofuscin peaks at 474 nm. The use of different excitation wavelengths on amyloid beta core plaques, result in different spectral shape of the fluorescence spectra. This indicates that there are various fluorescent molecules present in the plaque location which have there own absorption and emission spectra. More research needs to be performed to identify the fluorescent molecules we observe.

(32)

6 Supplementary Information

Figure 16: Fluorescence emission spectrum of Lutein dissolved in hexane illuminated with a 450 nm excitation source. Measurements are performed using the Cary Eclipse Spectrofluorimeter (10 mm quartz cuvette, exc/em slit widths of 10/10 nm). The red line shows the fluorescence signal of the Lutein solution dissolved in hexane with a concentration of 6.871*10-6M. The black line is the emission spectrum of pure hexane. Due to Raman background the blue line shows the calculated spectrum of pure Lutein which peaks at 519 nm. This is calculated after scaled subtraction of the hexane background (black) from the Lutein solution dissolved in hexane spectrum (red).

(33)

(a) (b) (c)

(d) (e) (f)

Figure 17: SRS signal of the DNA vibrational mode at 2967 cm-1. Gallary of images at the same plaque location. Starting from (a) scanning until (f) measured at 1µm apart in the z direction. We see that the core plaque in the center of the images might be a hollow shell. Image (c) and (d) suggest the plaque to be hollow, the shell closes at (a) and (f).

(34)

(a) (b) (c)

(d) (e) (f)

Figure 18: SRS signal of the protein vibrational mode at 2926 cm-1. Gallary of images at the same plaque location. Starting from (a) scanning until (f) measured at 1µm apart in the z direction. We see that the core plaque in the center of the images might be a hollow shell. Image (c) suggest the plaque to be hollow, the shell closes at (a) and (f).

(35)

Acknowledgements

I would like to thank Benjamin Lochocki and Freek Ariese for their guidance during this Bachelor project. Both Ben and Freek explained a lot to me and helped me with all my questions. I did this project at an unusual time in the semester, however I felt that this was not a problem at all and it really gave the opportunity to learn and explore even more spectral imaging devices. Beside the fact that we experienced the second wave of the covid-19 pandemic I was still able to do a lot of lab work and was able to proceed with my project.

I would also like to thank Liron Zada for explaining the complete calibration method of the SRS setup. This helped me understand the setup with a more hands on feeling.

Likewise I want to thank the pathology department of the Amsterdam UMC. Especially, Jeroen Hoozemans for the Thioflavin-S staining and Baayla Boon for helping with the fluorescence measurements at the Nikon A1 microscope and with interpreting of the SRS results that show hollow core plaques.

Finally, I want to thank Johannes de Boer and everyone else who joined the weekly zoom section meetings. This was a great way to present and discuss my newly found results. Also it was very interesting to listen and learn from the experimental updates that everyone was presenting.

(36)

References

[1] B. Lochocki, T.H.J. Morrema, F. Ariese, J.J.M. Hoozemans, J.F. de Boer The search for a unique Raman signature of amyloid-beta plaques in human brain tissue from Alzheimer’s disease patients, The Analyst, (2020).

[2] F.K. Lu, S. Basu, M.P. Hoang, et. al. Label-free DNA imaging in vivo with stimulated Raman scattering microscopy, Proceedings of National Academy of Sciences USA, 112(37): 11624-11629,(2015).

[3] M. Straathof, Creating a Graphical User Interface in Matlab for processing of Raman mapping data., Bsc thesis Physics, Vrije Universiteit Amsterdam, LaserLab, (2020).

[4] D.W. Shipp, F. Sinjab, I. Notingher Raman spectroscopy, techniques and applications in the life sciences, Advances in Optics and Photonics, (2017).

[5] B.E.A. Saleh, M.C. Teich. Fundamentals of Photonics, 2nd Edition, Wiley, 2007, ISBN: 978-0-47135832-9

[6] J. den Haan, T.H.J. Morrema, A.J. Rozemuller, F.H. Bouwman,J.J.M. Hoozemans. Different curcumin forms selectively bind fibrillar amyloid beta in post mortem Alzheimer’s disease brains: Implications for in-vivo diagnostics. Acta Neuropathologica Communications, 6, 75-87, (2018). [7] J.A. Hardy, G.A. Higgens. Alzheimer’s Disease: the amyloid cascade hypothesis, American

Association for the Advancement of Science, 256, (1992).

[8] J.W. Erdman, J.W. Smith, M.J. Kuchan, E.S. Mohn, E.J. Johnson, S.S. Rubakhin, L. Wang, J.V. Sweedler, M. Neuringer. Lutein and Brain Function. Foods, Multidisciplinary Digital Publishing Institute, 547-564, (2015).

[9] M. Ji, et al, Label-free imaging of amyloid plaques in Alzheimer’s disease with stimulated Raman scattering microscopy. Science Advances, 4, 11,(2018).

[10] B. Lochocki, B.D.C. Boon, J.J.M. Hoozemans, F. Ariese, J.F. de Boer. Multimodal, label-free fluorescence and Raman imaging of amyloid deposits in snap-frozen Alzheimer’s Disease human brain tissue, ’in review’ (2020).

[11] E.J. Johnson, A possible role for lutein and zeaxanthin in cognitive function in the elderly. The American Journal of Clinical Nutrition 96, 1161S-1165S (2012).

[12] L. van Haasterecht, L. Zada, et al., Label-free stimulated Raman scattering imaging reveals silicone breast implant material in tissue, Journal of Biophotonics, 13, 5, (2020).

[13] N. Craft, J. Soares, Relative Solubility, Stability, and Absorptivity of Lutein and Beta-Carotene in Organic Solvents. Journal of Agricultural and Food Chemistry, 40, 10, 10-21, (1992).

(37)

[14] A. Wimo, G. Ali, M. Guerchet, M. Prince, M. Prina, Y. Wu, World Alzheimer Report 2015. Alzheimer’s Disease International, (2015).

Raman and Fluorescence Imaging of Core Amyloid Beta Plaques in Human Alzheimer Diseased Brain Tissue