1
Reducing Fluorescent Cross-Talk Noise
with an Arduino and AOTF
Bachelor Project: Bonno Meddens Daily supervisor Zhiqing (Andy) Zhang Main Examiner/Supervisor: Iddo Heller Secondary Examiner: Erwin Peterman
2 Abstract: The goal of this project was to develop a low-cost and versatile electronic triggering system for temporal control of excitation laser beams in single-fluorescence microscopy applications. The triggering system was developed with a microcontroller in the Arduino eco-system. This system was used in a proof-of-principle experiment to lower the fluorescent cross-talk between two channels, using alternating excitation of fluorescent molecules. In this proof-of-principle experiment, each excitation laser beam was controlled with an acoustic-optic tunable filter (AOTF). The Arduino switched the AOTF channels on and off and lastly, the Arduino was controlled with the opensource automated microscopy software, micro-manager. Micro-manager already had built-in support for the Arduino Uno, but the Arduino Uno did not have the required features to be used with this project. This problem was solved by writing new Arduino micro-manager firmware that work on newer Arduino-compatible boards that offered more features and a new micro-manager device-adapter was created, called ‘Arduino32BitBoards’. The firmwares are available in the GitHub repositories bonnom/Arduino32BitBoards and bonnom/Micro-manager-Arduino. And lastly, the Arduino implementation was tested with a proof-of-principle experiment by imaging two-types of motor proteins in the cilia of the C. elegans worm. The triggering system successfully reduced fluorescent cross-over. Therefore, this triggering system can be used in future experiments need to solve the cross-over problem.
Thesis Content
1. Introduction ... 3
2. Fluorescence ... 4
3. Hardware Triggering System (Arduino) ... 6
3.1 Introduction ... 6
3.2 Micro-Manager and Arduino ... 8
3.3 Details of Implementation with AOTF ... 9
3.3.1 AOTF setting changes ... 9
3.3.2 DAC vs PWM control of the AOTF ... 9
3.3.3 AOTF – Arduino interface ... 9
3.3.4 Arduino boards ... 9
3.3.5 Addon boards ... 11
4. C. Elegans, Proof-of-Principle Experiment ... 12
4.1 Introduction, Proof-of-Principle Experiment ... 12
4.2 Method Proof-of-Principle Experiment ... 13
4.3 Results ... 15
4.3.1 Determining the improvement of switching illumination ... 15
4.3.2 Images of Alternating Imaging ... 15
4.4 Conclusion of Results ... 20
5. Discussion ... 20
5.1 Limitations: Implementation and Experiment ... 21
5.2 Alternative solutions ... 21
5.3 Possible improvements ... 22
6. Conclusion ... 22
7. References ... 22
8. Appendix ... 24
3
1. Introduction
Fluorescence microscopy is a useful tool for gaining insight in the mechanics of the biomolecular mechanism and therefore plays a significant role in gaining better understandings of diseases (Chalfie, 2009). Sometimes it is important to image multiple fluorescent molecules simultaneously, for example, imaging the interactions between different proteins. When designing a fluorescent imaging experiment that requires multiple different fluorescent molecules (fluorophores) it is important to choose fluorophores that have as little emission cross-over as possible. eGFP and mCherry have only little cross-over in emission spectrum and this makes them an ideal combination for co-localization of proteins in fluorescent microscopy (Doherty, Bailey, Lewis, 2010). However, this little cross-over may not be ignorable when there are many more fluorescent eGFP molecules compared to mCherry molecules. The emission spectra of eGFP are displayed in figure 1. The cross-talk problem becomes even worse in fluorescent imaging of living organisms because living organisms contain many other molecules that are also fluorescent and have a cross-over in their emission spectrum with eGFP or mCherry. A more detailed explanation of fluorescence and fluorescence microscopy will be explained in chapter 2)
Figure 1: Emission Spectra of mCherry and eGFP. The overlap between the spectra is highlighted in purple. This overlap becomes a significant problem when much more eGFP molecules are
4 The goal of this bachelor project was to solve the eGFP-mCherry cross-talk problem, and it was achieved by developing a low-cost and versatile electronic triggering system for temporally alternative control of excitation laser beams of eGFP and mCherry. This thesis describes how alternating imaging can be achieved by switching consecutively between the two excitation laser beams with an acousto-optic tunable filter (AOTF) that is controlled by an Arduino and the micro-manager software. Micro-micro-manager is an open-source software package specifically designed to control automated microscopes and related ancillary equipment.
The entire setup was tested with a proof-of-principle experiment. In the experiment, two motor proteins, i.e. kinesin II and cytoplasmic dynein 2, were imaged alternatively in the primary cilia of C. Elegans. Kinesin II and dynein motor proteins were tagged with mCherry and eGFP, respectively. Several other methods have been considered to control the laser light in a fluorescent setup. Eventually it was chosen to use an Arduino with an AOTF to control the laser light. One advantage of using the Arduino microcontroller is that a microcontroller can work in real-time on a specific task, which is different from a PC that needs to perform a lot of tasks. A PC can’t perform all those tasks simultaneously and therefore it must schedule the tasks that it needs to perform. This leads to variable execution time, what can be problematic when proper timing is needed. In this case proper timing is needed to synchronize the two laser beams with certain camera exposure time.
An AOTF was used to control the amount of light instead of directly modulating the intensity of the lasers with pulse-width-modulation (PWM; PWM is explained in chapter 3). An AOTF is essential for controlling the 561nm laser, because it is a diode-pumped solid-state laser and it can’t be switched on-off very quickly.
This thesis is organized into six sections. The first is the introduction, the second is a general introduction of fluorescence microscopy. The third section contains information about the triggering system written in Arduino and the technical details. The fourth section is about the proof-of-principle experiment within C. elegans and the results of the experiment. The fifth section is the discussion. The last section is the conclusion.
2. Fluorescence
The name fluorescence was first used by the famous mathematician and physicist George Stokes to describe the ability of fluorspar and uranium glass to change ultra-violet invisible light into visible light (Stokes, 1852). Nowadays, fluorescence plays an important role in microscopy as a tool for cellular biology research. It enables researchers to visualize the locations and movements of biomolecules inside the cell, and thereby it helps researchers to determine the functions of biomolecules. One recent important discovery of fluorescent microscopy is the discovery of the green fluorescent protein (GFP) and its DNA sequence. With genetic manipulation, the DNA sequence of GFP can be added as a tag to the end of proteins. This GFP tag makes the protein fluorescent and therefore visible under a fluorescent microscope (Strack & Jaffrey, 2014). This opened many new research possibilities for cellular biological research. Because of the high impact of GFP on today’s cellular biology, Nobel prize was awarded to the discoverers of GFP and the DNA sequence. Many new GFP variants have been made since the discovery of GFP. It leads to the discovery of new fluorophores that have different colours and/or better characteristics than the original GFP (Day & Davidson, 2009). One popular GFP variant is the enhanced GFP (eGFP).
Besides GFP and their derivatives, there are also other ‘families’ of fluorescent proteins discovered and an important one is the mFruit family (Day & Davidson, 2009). An important member of the mFruit family is mCherry. One of the biggest advantages of mCherry is that its emission peak lies far away of the emission peak of eGFP. This makes eGFP and mCherry an ideal combination to be
5 imaged simultaneously (Doherty, Bailey, Lewis, 2010). To better understand fluorescence, it is important to know what fluorescence exactly is.
Figure 2: The Jablonski diagram that illustrates the electron energy levels of a molecule. S denotes a singlet state and T denotes a triplet state. The straight arrows represent the emission or absorption of
a photon. The dot lines represent nonradiative electron transitions. An important nonradiative transition is the intersystem crossing (ICS) since this transition can cause a molecule to be in a long
dark state.
Fluorescence is defined as light emitted by a molecule after absorption of light by the same molecule and involves a spin-allowed, singlet-singlet electronic transition (Wildenberg, Prevo & Peterman 2011). The transitions between the energy levels can be visualized in a Jablonski diagram (figure 2). The diagram shows an example of fluorescence. A photon denoted by a blue arrow raises the energy level of an electron from the S0 state to the S2 state. Then the electron energy decreases from S2 to S1 state due to internal conversion, this energy is converted into heat. And lastly, the electron then returns from S1 to the ground state S0 by emitting a photon. Because part of the energy gets converted into heat has the emitted photon a lower energy compared to the absorbed one. This phenomenon is called the stoke shift (displayed in figure 3).
Also displayed in the Jablonski diagram are also phosphorescence and photochemical reactions. They phenomena also play a large role in fluorescence microscopy because they both cause the fluorophores to lose its fluorescent capability. With a photochemical reaction, the exited molecule undergoes a chemical reaction, which can cause the molecule to permanently lose it fluorescent ability. This process is called bleaching. With phosphorescence, an exited electron converts via intersystem crossing (ICS) to a triplet state, the triplet state can then radiate to the ground state and emit a photon. This transition is ‘forbidden’ under the dipole selection rules and therefore requires a higher multipole transition, what is a much slower compared to the fluorescent transition (Jones, 1998). During this transition, the molecule is not fluorescent and therefore temporarily not visible under a fluorescence microscope (Mamontova et al., 2017). When designing a Singe-molecule fluorescence microscope, it is important to supress the amount of bleaching from photochemical reaction and phosphorescence, this can for example be done by limiting the amount of light that the fluorophores receive. This however increases the need to reduce background-noise and the accurate detection of photons emitted by one or a few fluorescent molecules (Wildenberg, Prevo & Peterman, 2011). This brings us back to the goals of this thesis, to develop a low-cost and versatile electronic triggering system for temporal control of excitation laser beams in single-fluorescence microscopy applications and use this system to solve the cross-over fluorescence problem.
6 Figure 3: Stoke Shift of eGFP and a laser wavelength used for excitation of eGFP at 491nm (green).
The difference in wavelength between excitation (blue) and emission (red) can be clearly seen.
3. Hardware Triggering System (Arduino)
3.1 Introduction
A single-molecule fluorescent microscopy setup contains many components that need to be controlled. Most of the modern laboratory equipment can be controlled with computers. Unfortunately, computers do not work in real-time, which means that tasks need to be scheduled for execution. This leads to a variable execution time and it can be a problem when fast responsiveness and proper timing is needed. This problem can be solved by handing some tasks over to a micro-controller. A microcontroller is a small computer that is specifically designed to control equipment in real-time.
One of the most popular hobbyist microcontrollers ecosystem is Arduino. Arduino is an open-source hardware, software and ecosystem from the company that bears the same name. The company has the goal to make electronics available to a wider audience. The Arduino company provides software and hardware themselves but because of the open ecosystem, there are a lot of third-party companies that also make and provide hardware and software that can be used with the Arduino ecosystem. This gives a wide range of development boards and components that can be used with Arduino and therefore a viable option to use with laboratory equipment and software. In general, the Arduino boards are programmed in the Arduino language, what itself is a subset of the C/C++ language.
7 The most popular and basic Arduino board is the Arduino Uno. The Uno uses an 8-bit 16Mhz microcontroller that has 5V digital out and inputs and up to 5V analog inputs (analog-digital-converter, ADC) but it lacks an analog output (digital-analog-(analog-digital-converter, DAC). An external DAC chip can be added to the setup to solve this problem, and in other situations, it can be solved by using Pulse-Width-Modulation (PWM). With PWM, the digital outputs get switched between on and off at a certain frequency. The ratio between on and off time (duty ratio) determines the average power that is transmitted. The Arduino Uno runs at PWM frequencies at around 500Hz, what is be a bit low for imaging. When choosing an Arduino board, it is important to choose a board with a high PWM frequency.
Figure 4: Graphs of Pulse Width Modulation. The x-axis represents time and y-axis represents voltage. The duty cycle is the ratio between high and low signal (Hirzel, n.d.).
Besides the Arduino Uno, there are also other boards available that are much faster and do have DACs, but they usually run at 3.3V instead of 5V. This can lead to some problems when using these boards with the laboratory equipment since those usually send signals and receive signals at 5V. Most 3.3V Arduino board will get damaged when it receives a 5V signal unless it is specifically made to handle the 5V voltages. The other way around, sending a 3.3V digital signal to a device that is made for 5V, usually works just fine. When using an Arduino that can’t handle 5V it is important to use a voltage level converter that converts a 5V signal to 3.3V. There are multiple ways to convert the 5V signal to 3.3V but one of the simplest ways is to use a bi-directional voltage converter made with two resistors and a BSS-138 MOSFET. The mechanism is explained in the application note AN10441 from NXP. One of the advantages of this bi-directional voltage level converter is that many online electronics shops sell breakout boards that have all components directly soldered.
8
3.2 Micro-Manager and Arduino
Micro-Manager has built-in Arduino support. The support was mostly focused on the Arduino Uno. Unfortunately, the Arduino Uno does not provide some critical features needed for this setup. In this experiment, the intensity of the laser light must be controlled and switched quickly. The analog outputs can be added with an external chip, but this would make the setup more complicated. Therefore, it was chosen to use different Arduino compatible boards that offer more functions than the Arduino Uno.
The micro-manager website does provide Arduino code (firmware) that implements the functions and the communication with micro-manager but unfortunately the firmware used specific commands that only work on the Arduino Uno board and therefore the code would not work on any other Arduino boards. Therefore, the firmware had to be rewritten so that it can be used with the newer Arduino Boards. The rewritten firmware does provide some new and enhanced features compared to the old Arduino Uno firmware.
Features of rewritten firmware:
• Easier to adapt for new Arduino boards
◦ Newer Arduino boards are much faster
◦ Have internal DAC
◦ Faster and better control of PWM
• Ability to use DAC with digital pattern and blanking
• Higher communication speed with micro-manager
While the rewritten firmware did provide new features, but it still has some limitations. For example, the power could only be set for two output channels. This can be a problem when more than two lasers are used. Another problem is that if users want to control different output channels on the AOTF, they would have to physically switch the wires on the Arduino or change the configuration of the AOTF channels in the software and the last problem is that the control slider bar for the DAC in micro-manager can only be set from 0 to 5. These problems were solved by writing a new device adapter and firmware for the Arduino 32-bitboards. The Arduino32bitboards device adapter is now available with the nightly builds of micro-manager and provides extra features on top of the regular Arduino device adapter:
• Eight output channels instead of six
• Added PWM support
• Unified DAC/PWM controls in Micro-manager
• DAC/PWM control slider is from 0 to 100
• Ability to change output power on all eight outputs
◦ Currently, channel 1 and 2 are DACs
◦ Channel 3 to 8 are PWM
Important notes for implantation of Arduino in micro-manager.
There are two Arduino device adapters for micro-manager, one is called ‘Arduino’ the other is called ‘Arduino32bitBoards’. They each have their own Arduino firmware and cannot be used interchangeably. Micro-manager will give an error if the firmware version doesn’t match. Arduon32bitBoards is the latest Arduino device-adapter and comes with the micro-manager versions 1.4.23 and 2.0. The firmware for Arduino32bitBoards can be found in the GitHub repository: ‘bonnom/Arduino32BitBoards’. The newly written firmware’s for the regular Arduino device-adapter can be found in the github repository: ‘bonnom/Micro-manager-Arduino’.
9 There is currently also another problem, the Arduino can’t control the AOTF from Opto-electronics if the AOTF is added to the same micro-manager configuration. This is because the AAAOTF device adapter automatically sets the AOTF channels on ‘internal’. This means that the AOTF can’t be controlled with external signals from the Arduino.
3.3 Details of Implementation with AOTF
3.3.1 AOTF setting changes
The Arduino works at 3.3V but the AOTF inputs scale from 0 – 5V. This means that the Arduino cannot enable the full power of the AOTF. This problem can be partly solved by increasing the dBm on the AOTF. The dBm is an absolute power range where 0 dBm stands for 1.0mW. Since power grows quadratically with voltage this means that the decibel must be increased with 3.6 dBm to get the same value. This was calculated with the formulate Log ((3.3/5)2) where the log is the logarithm of 10.
3.3.2 DAC vs PWM control of the AOTF
The AOTF power can be controlled with the DAC and PWM outputs on the Arduino. There is a big difference in AOTF behaviour between PWM and DAC control. With PWM, the output strength of the lasers behaves entirely linear. For example, when the PWM channel is at 50% power the laser light at the microscope is at 50% of the brightness. The DAC does not behave this way, at 50% power the laser light is not necessarily at 50% of the max brightness. Therefore, it is not recommended to use PWM and DAC together since the brightness of the laser light on the objective behaves differently. Unfortunately, PWM wasn’t tested in any experiments, therefore it is unknown what the effect will be on bleaching.
3.3.3 AOTF – Arduino interface
The Arduino is attached to the MDS/MPDS with a DB-25 connector. The pinout of the DB-25 connector on the AOTF can be found in the appendix. The Arduino must be connected the desired channels to ground. The blanking must be connected to 5V, unless the user wants to control the blanking itself.
The Arduino sends a signal over a specific pin on the DB-25 channel. This signal gets amplified by the dBm set in the MDS for that channel.
3.3.4 Arduino boards
As mentioned in the introduction of this section, because of the openness of the Arduino ecosystem there are many other companies that make boards that can be programmed in Arduino. Some of the boards from third-party manufacturers exhibit extra features that the official Arduino boards don’t have. All the Arduino boards that I wrote firmware for are from third-party manufactures, this was mostly done because of the extra features and ease of use.
There are multiple Arduino boards that can be used with the software, with each having their own pros and cons. In this section multiple boards will be discussed. Firmware has been written for the following boards: Feather M4 (Adafruit), Metro M4 express (Adafruit), ESP32, Teensy 3.2, 3.5, 3.6 and 4.0
The Feather M4 (figure 5) from Adafruit uses the feather system. The great advantage of the feather system is that many add-on boards are available, and it is relatively simple to design your own add-on board.
10 The Metro M4 Express from Adafruit is similar to the Feather M4 Express except that it is in the larger Arduino Uno formfactor (figure 5). This board is especially useful if someone needs to make an add-on board that needs more space.
The ESP32 is an Arduino compatible chip from Espressif. The ESP32 has two DACs and the PWM frequency can be adjusted. There are many different boards ESP32 available in all kind of form-factors. I will discuss two ESP32 boards here, the TinyPico and the Huzzah32. The TinyPico uses its own small form-factor. The Huzzah32 from Adafruit uses the standard Featherwing format and therefore can be used with all the Featherwing add-on boards.
The Teensy boards have a bit more advanced features compared to the boards from Adafruit. The PWM frequency and resolution can be changed for the mentioned boards. The Teensy 3.2 has 5V tolerant inputs but only has only one DAC. The Teensy 3.5 also have 5V tolerant inputs and two DACs. The Teensy 3.6 is similar to the teensy 3.5 except that it is faster and has some additional features that are not relevant to microscopy at the moment, but it does not have 5V tolerant inputs. There is also the new Teensy 4.0, the teensy 4.0 does not have any DACs but it is much faster than any other Arduino compatible board and a good option if only PWM is needed. All the teensy boards have a similar formfactor and there is a feather breakout board available that allows the Teensy boards to be used with the Feather system (figure 6). Just note that channel 1 and the DAC channels are not available when using the feather breakout board with the Teensy 3.5, 3.6 and 4.0.
Figure 5: Arduino Uno (above) from Arduino and Feather M4 board from Adafruit (below). There are many add-on boards available for both formfactors. [Top image from Arduino.cc, bottom from
11
3.3.5 Addon boards
The initial setup used loose wires that connected the DB-25 connector from the AOTF with the Arduino. This makes the entire setup not very robust and prone to failure. Three PCBs were designed for multiple Arduino formfactors to make the entire setup less prone to failure. There is an Addon-board that uses the Featherwing footprint (figure 6), one with the regular Arduino UNO footprint and one for the TinyPico board. Every PCB has a DB-25 connector, a BNC connector, and a voltage level converter soldered. The voltage level converter lowers the trigger signal voltage from the camera to 3.3V. On every PCB the first channels of the AOTF are connected to the DAC outputs, and the rest are connected to the PWM. The addon boards can be directly connected to the M(P)DS (AOTF controller). Please note that the AOTF must first be turned off before connecting the Arduino with add-on board, otherwise there will be some sparks flying around.
The schematics for the PCBs can be found in the supplementary files.
Figure 6: Teensy 4.0 on a Featherwing adapter (above) and an AOTF Featherwing PCB (below). The boards are connected to each other on a ‘doubler’ board (not visible).
12
4. C. Elegans, Proof-of-Principle Experiment
4.1 Introduction, Proof-of-Principle Experiment
As stated in the introduction, simultaneous imaging of eGFP and mCherry causes extra background for the mCherry channel because of auto-fluorescent cross-talk. This problem was solved by exciting eGFP and mCherry in an alternating manner during imaging. The alternating imaging was achieved by switching the two laser beams with an AOTF. The channels and the output power of the AOTF were controlled with an Arduino. The Arduino got a trigger signal from the microscope camera. The camera sends this trigger digital signal to the Arduino when it starts capturing an image. This causes the Arduino to switch the output channels of AOTF and thereby the laser light. Unfortunately, the setup couldn’t be tested with the Arduino32bitBoards device adapter and only with the regular micro-manager device adapter.
The channel trigger pattern and the output power per channel of the Arduino were set within the micro-manager software.
The alternating setup was tested on the primary sensory cilia of the nematode C. elegans. C. elegans are often used as an organic model for biological research, because it is a relatively simple model and low cost to maintain. Short reproduction cycle makes it great to get large quantities in a short amount of time, and their translucent and their small bodies (~1mm) make them ideal for fluorescent microscopy (Kaletta, & Hengartner 2006; Leung et al., 2008).
Proteins and other molecules need to be transported to specific places inside a cell. Prokaryotic organism, such as bacteria mostly use diffusion to distribute the molecules and large complexes within cells. Eukaryote cells are usually much larger and more complex and therefore cannot solely rely on diffusion to distribute their components, which is especially the case for cells that have long protrusions such as neurons. This problem is solved by having motor proteins to move cargos inside cells. The motor proteins walk with the cargo on microtubules or actin filaments, consuming ATP in the process (Howard, 2001). The actin and microtubules have two distinct ends and the motor proteins can only walk unidirectionally towards one of the ends. Kinesin motors walk on the microtubules towards the plus-end and dynein motors walks the other way around.
The cilium is the protrusion from a cell body that contain a microtubule base cytoskeleton. Cilia can further be divided into motile cilia, nodal cilia and primary cilia. The motile cilia are cilia that have controllable movement. In humans they are found for example in the respiratory epithelium where they have the function of sweeping mucus and dirt out of the lungs and the fallopian tubes where they transport the egg cell (Enuka et al., 2012). Disfunction of motile cilia can lead to infertility and complications of pregnancy where the embryo attaches outside of the uterus (Horne & Critchley, 2012).
The nodal cilia are only present in developing embryos and generate a leftward flow. This leftward flow is detected by primary cilia and this activates a signalling cascade that establishes left-right sidedness. Disfunction of nodal cilia can lead to abnormal arrangement of organs and therefore cause serious health problems (Horani & Ferkol, 2018).
The primary cilia are found on nearly every cell and play an important role in cell chemical sensation, signal transduction and control of cell growth. Because of these functions, the primary cilia play a critical role in homeostasis, development and senses such as smell and sight. Disfunction of primary cilia can lead to a variety of severe diseases such as polycystic kidney disease which leads to complete loss of functions of the kidneys. Therefore the C. elegans nematode is used as model organism in research to get a better understanding of the cilia and ciliopathies (Müller et al., 2011).
13 Cilia consist of many different parts that are not yet fully understood. One of them is to understand the behaviours of the motor proteins in the intraflagellar transport (IFT). The IFT plays a critical role in maintenance of cilia and has been implicated to play a role in certain ciliopathies (Ou et al., 2005). To understand the behaviour of the motor proteins it is important to image different proteins simultaneously.
Since this thesis explains a way to reduce crossover noise between different fluorophores, it makes sense to do a test experiment by imaging two motor proteins simultaneously in the cilia of the C. elegans. The application would especially be useful in this instance, since all the proteins and other biomolecules cause even more severe crossover noise.
Three different motor proteins have been imaged in the proof-of-principle experiment, kinesin-II, OSM-3 and dynein. In the experiments, the motor protein dynein was imaged together with either kinesin-II or OSM-3. Dynein was always tagged with eGFP. Kinesin-II and OSM-3 were tagged with mCherry. The mCherry protein was excited with 561nm and the eGFP was excited with 491nm laser light.
4.2 Method Proof-of-Principle Experiment
The goal of this bachelor thesis was to reduce the crosstalk noise when doing simultaneous fluorescent imaging of eGFP and mCherry. The cross-talk was eliminated by enabling alternating imaging of eGFP and mCherry. The alternating imaging was achieved by using an Arduino switching the laser light. The setup was tested on the motor proteins in the sensory cilia of the C. elegans where there were two different type of motor proteins labeled with eGFP or mCherry. The goal was to visualize the single molecule trajectory of single motor protein tagged with mCherry, in the presence of bulk dynein motors tagged with eGFP. The motor proteins have a high expression level, therefore there are many motor proteins on the cilia, and it becomes impossible to identify single molecule trajectories. The problem was solved by bleaching the mCherry tagged motors until only a couple were visible, without bleaching the eGFP tagged motors. To make the single motor proteins visible it was important to increase the strength of the laser light. This in turn also caused the remaining motor proteins to bleach therefore, it was important to find the balance between strong illumination and bleaching.
The general equipment used in this experiment can be found in chapter 8 of the book “Single-Molecule Analysis” written by Jaap van Krugten and Erwin Peterman. Added to the setup is an AAOTF from Optoelectronics and the Arduino-compatible board, ItsyBitsy M4 with the Arduino bootloader. A schematic of the setup is displayed in figure 7. The ItsyBitsy controlled the AOTF with its two analog outputs. PWM wasn’t tested in this proof-of-principle experiment. Also, the setup still used an older micro-manager version and not a newer one. This unfortunately meant that the Arduino32bitBoard device adapter couldn’t be tested in this experiment.
14 Figure 7: Setup schematic of the proof-of-principle experiment, the line arrows represent the light
path. The other lines are data lines. The Arduino is highlighted in orange.
Acoustic-Optical-Tuneable filter is a device that deflects light at certain angles when vibration at certain frequencies, the AOTF is controlled by an Arduino.
C. elegans strains were grown at 20 degrees Celsius on nematode growth medium that was seeded with E. coli as a food source. Two strains of c. elegans were used: kinesin-II::mCherry + Dynein::eGFP, OSM-3::mCherry + Dynein::eGFP. Every four Days, 5 worms at L4 were picked from an old plate to a new one. Four days after that the plates were full of nematodes that were ready to be picked and imaged. When selecting the best worms for imaging it is important to pick adult worms because only, they have fully matured in the cilia. The nematodes can’t be too old because the older adults have more auto fluorescent background noise. Adults with ~10 eggs are best for imaging. When taking the worms from a plate, it is important to avoid taking the growth medium and the E. coli as much as possible. This is because the growth medium and the E. coli cause auto-fluorescent background noise.
The picked worms were laid in a solution of 5mM Levamisole in M9 salt solution, on the centre of a microscope cover slip. Levamisole is agonist of the nicotinic acetylcholine receptor. The activation of the nicotinic acetylcholine receptor causes continuous stimulation of the nematode muscles, leading to paralysis. The worms must stay ~10 minutes in the solution to become fully paralyzed. After ~10 minutes a microscope glass with agarose is put on top of the cover slip with levamisole. The agarose layer is needed to prevent the worms from being crushed between the two glasses. The cover slip was than sealed with a mixture of equal parts of Vaseline, lanolin, and paraffin wax, that was heated to 85 degrees Celsius. This is to prevent the cover slip from moving and stops the moisture from escaping. Each cover slip contained 7 worms for imaging.
15 About 20 mins after the microscope slide was sealed, oil was added to the oil objective and the microscope slide was attached on top of the microscope objective with the cover slip down. The next step was to find the cilia of every worm in regular light and setting their position in micro-manager. This makes it easier to find the cilia of later when switching to fluorescence microscopy. After every position of every worm was marked, the lights were turned off and the sample was illuminated with laser light at a low intensity to localize the cilia. A light filter was then removed, the power was increased to 10% for 491nm and 20% for 561nm, and the alternating imaging was started. The 561nm laser light power was increased to 60% after the fast majority of mCherry molecule were bleached. The imaging was continued until no more mCherry molecules could be seen.
The images were then analysed in Fiji, an ImageJ distribution. This was done by first making a new stack of only the 561nm images. All the 561nm images were averaged and a kymograph was made. A kymograph is a graph where one axis represents time and the other axis represents distance. The resulting images will all have time on the x-axis and on the y-axis the location of the cilia. The kymograph was made within the ImageJ plugin KymographClear 2.0, with a 7-pixel line thickness (Mangeol, Prevo, & Peterman, 2016).
4.3 Results
The goal of this bachelor thesis was to reduce the crosstalk noise when doing simultaneous fluorescent imaging of eGFP and mCherry. The cross-talk was eliminated by making alternating imaging of eGFP and mCherry. The alternating imaging was achieved by using an Arduino to switch the laser beams.
4.3.1 Determining the improvement of switching illumination
The goal of using an Arduino was to lower the cross-talk between the 491nm and 561nm light. There are multiple methods to determine the effectiveness of the cross-talk illumination. One method is to compare the kymographs of measurements with and without Arduino. Unfortunately, due to the large difference in trajectories it is hard to do a direct comparison of the improvement in image quality. The second method would be to calculate to take the cross-talk noise of the 491nm image and compare this to the 561nm image. This can however not be used since mCherry does also get excited by 491nm and so therefore not everything of the 491nm image is noise.
Another way is to add the pixels value of 491nm and 561nm images and comparing the difference with the 561nm image. This does give a direct comparison but there is another problem. Because of the camera had to take two images instead of one, the total capture time had to be halved to get the same interval of 150ms between each image. This increases the shot noise.
The method by adding the pixels value of 491nm and 561nm was chosen to determine the improvement of alternating imaging. The 491nm and 561nm pixels were added together with a python script.
Additional to a visual inspection of the results, there is also a line plot made of kymograph sections. This plot is made by using the ‘straight line’ to select a region of interest (e.g. Cross-section of a motor protein path) and using “Plot Profile” to plot the pixel values across pixels.
4.3.2 Images of Alternating Imaging
The results contain pictures that were illuminated with both 491nm & 561nm light and illuminated with just 561nm.
Comparison 1:
Figure 8 contains images taken in the early stages of imaging. The left one was obtained with both the 491nm & 561nm illumination, while the right one only with 561nm. As indicated, the general noise of
16 only 561nm illumination is much less. For example, the silhouette of the worm’s tail is not visible with in the image that is only illuminated with the 561nm laser. It must be noted that this image is taken in the early stages of imaging where there are still many dynein::mCherry molecules visible. In the single molecule stage, the eGFP cross-talk becomes unignorable.
491nm + 561nm 561nm
Figure 8: images of kinesin-II-mCherry and dynein-eGFP fluorescent proteins. Left the sample is illuminated with 491nm and 561nm laser, right only illuminated with 561nm.
Figure 9 is the average pixel value of all frames obtain from the same cilia. The noise is much less apparent in these images but when looking closely the difference can be seen in the structures surrounding the cilia. For example, the microtubules down of the cilia are much clearer in the image that was only illuminated with 561nm.
17
491nm + 561nm 561nm
Figure 9: Z-projection of all frames in one series. The left sample was illuminated with 491nm and 561nm laser, while the right one was only illuminated with 561nm. Notice that the microtubules
outside of the cilium are much clearer on the right image. Comparison 2:
The next step is to compare a kymograph section. The section of the kymograph is displayed in Figure 10. The difference in noise can be clearly seen. The 491nm + 561nm image has much more noise and the trajectory of a motor protein can be seen only vaguely. The trajectory is much better visible in the image that was illuminated with only 561nm.
Figure 10: Kymograph of kinesin-II. The first image was obtained with the 491nm and 561nm illumination, the second one was obtained with 561nm illumination only. The x-axis represents time,
the y-axis represents position.
A plot of pixel values was made over a line in the images of figure 10 to view the variation in pixel value. The line is shown in figure 11.
18 Figure 11: Line over which a plot was made in figure 12, the same line was taken over the 491nm +
561nm image of figure 10.
Figure 12: The pixel plot of figure 11. The same line was taken from 491nm + 561nm image and plotted as blue. The 561nm plot is shown as red. As can be seen in the figure is that the 561nm plot
has less noise than 491nm + 561nm plot.
In figure 12 it can be clearly seen that the 561nm line plot is much smoother compared to the line plot of the 491nm + 561nm image, and the peaks are more visible. This means that there is much less variable noise on the 561nm image. But the trajectory of the right side of the images in figure 10 can still be seen on both line plots as two peaks.
Comparison 3:
Another kymograph comparison, with the same analysis as in comparison 2, was done on more mCherry trajectories.
19 In figure 13 the image illuminated only with 561nm has noticeable less noise compared to the 491nm + 561nm illumination but the difference is much smaller compared to comparison 2.
Figure 13: Kymograph of kinesin-II. The top image contains the 491nm and 561nm illumination, the bottom one is 561nm only. The 561nm image is a bit clearer than the image that was illuminated with
491nm + 561nm.
Just as in comparison 2, a line and a pixel value plot are shown in figure 14 and 15.
Figure 14: Line over which a plot was made in figure 15, the same line was taken over the 491nm + 561nm image of figure 13.
20 Figure 15: The procedure was done as in figure 14 but now with a different kymograph part. The
561nm illumination has two distinct peaks compared to the 491nm + 561nm illumination. The pixel value plotted over the line in figure 14 is depicted in figure 15. The line plot of 561nm is smoother than the 491nm + 561nm line plot. The 561nm line plot has also two distinct peaks that belong to the trajectory of motor proteins. The same peaks cannot be distinguished from noise in the 491nm + 561nm line plot.
4.4 Conclusion of Results
There is a noticeable difference in noise between the obtained with only 561nm light illumination and the images obtained with both 491nm and 561nm illumination. This means that making alternating images between 491nm and 561nm reduces the amount of noise and background considerably. Therefore, the conclusion of the proof-of-principle experiment is that doing alternating illumination imaging with an Arduino and an AOTF can significantly reduce the amount of cross-over and background noise.
5. Discussion
The goal of this project was to develop a low-cost and versatile electronic triggering system for temporal control of excitation laser beams in fluorescence microscopy applications. This was tested on a proof-of-principle experiment by imaging two different motor proteins in the primary cilia of the nematode C. elegans. As shown in the results of the proof-of-principle experiment, using alternating imaging with an Arduino an AOTF does lower the cross-talk noise of fluorescent molecules.
21 While the overall objective of this bachelor project has been achieved there are still some further points that need be discussed. These will be discussed below.
5.1 Limitations: Implementation and Experiment
The current setup has some limitations, the device has only two DAC output channels and the rest are PWM, and the Arduino outputs only at 3.3V. The used PCBs are not flexible in reconfiguring pins, which means that when using the PCBs, the channels must be switched around on the AOTF to channel 1 and 2 if they want to use a DAC.
Reduced exposure time
Alternating imaging requires twice the number of pictures to be taken. It means that either the exposure time needs to be halved, or that the time between images gets doubled. A decrease in exposure time means that the shot noise will be increased. It is important to note that this was not taken into account in the results of the proof-of-principle experiment.
Different exposure times
In micro-manager with multi-dimensional acquisition it is possible to set different exposure times for each AOTF channel. Currently this does not work for an unknown reason. It would require more experimentation to find out if this problem is caused by either the Arduino or the Andor camera that Because it is much harder to image single mCherry molecules than eGFP, it would be a great addition to have separate exposure time for the different molecules.
PWM vs DAC
While the AOTF can be controlled with PWM, the use of PWM was not tested on actual fluorescent molecules. This means that it is not certain what the effect will be on bleaching.
Arduino32bitBoards
The new Arduino32bitBoards device adapter could not be used in the proof-of-principle experiment, this is because the device-adapter is currently only supported in the latest Micro-managers 1.4.23 and 2.0 and I couldn’t get those working with the microscope camera from Andor. Therefore, it is not yet known if there are small bugs in the device adapter.
5.2 Alternative solutions
While in this thesis presents a method that uses an Arduino to control an AOTF for alternating imaging. Two other options will be discussed below.
Using AAAOTF device adapter
In the introduction it was argued that controlling equipment with a PC can be problematic because computers don’t work real-time. It must be noted that the severity of this problem wasn’t actually tested if this was the case. Technically the AOTF could also do alternate imaging within micro-manager. The communication between the computer and the AOTF run at a baudrate of 57600. This means that the computer can send a maximum of 57600 bits a second to the AOTF. The commands micro-manager sends to are something around 64bits, which has no overhead added. At the maximum rate this would mean that it takes a bit more than 1ms for micro-manager to change one AOTF channel. This could be enough for imaging, but it is unknown how many overhead there is inside the AOTF and micro-manager.
AAAOTF device adapter adaptation
In this setup, the Arduino sends analog signals to an MDS, the MDS amplifies this signal with a preconfigured value in dBm at a preconfigured frequency. An option would be to control the amplification of the MDS in micro-manager directly. The Arduino would not have to control the amplitude of an MDS channel but only turn them on and off.
22 Currently this cannot be done in micro-manager. This is because the current AAAOTF device adapter puts all the channels of the MDS channels on internal. This means that the AOTF can’t be controlled with the external signals from the Arduino. This problem can be solved by modifying the AAAOTF device adapter with an extra option to set the AOTF channels on external/internal. The great advantage of controlling the amplification of the MDS in micro-manager would be that the Arduino would only have to send a high-low signal to the MDS. This would solve the problem of having only two DAC outputs on the Arduino boards.
5.3 Possible improvements
Currently, the Arduino only has outputs at 3.3V but the AOTF can take signals up to 5V. One improvement could be to increase the output voltage from 0 – 3.3V to 0 - 5V with an operational amplifier (op-amp), this is especially useful important if the required channel can’t be raised by 3.6 dBm. This happens for example, when an AOTF channel is configured to be at 20dBm, the dBm must be raised by 3.6dBm to get the same power output when using a 3.3V Arduino. But now the problem is that a AOTF channel can only be set to maximum of 22.5dBm. In this case it means that a 3.3V Arduino can’t reach the full power of the AOTF.
For PWM, a single MOSFET can be used to raise the voltage from 3.3V to 5V but it is important to check the implementation with an oscilloscope because the voltages might rise a bit too slowly. This is however not a big problem since at the highest luminosity the bleaching happens very quickly, and this is usually not desired.
Increasing DAC output channels
The current Arduino boards on the market only have a maximum of two analog outputs, this can be extended to eight in multiple ways. One method would be to use an external DAC IC that does have has eight output channels. An IC that might be useful is the DAC43608. Using an external DAC might introduce some new problems, for example new Arduino library must be written to use the DAC.
Limitation of determining the improvement
There is not a standard way to calculate an improvement in signal-to-noise ratio with imaging. This is especially true for high-noise situations where it is difficult to distinguish the signal at all. This formed a limitation on determining the improvement of this setup.
6. Conclusion
In this bachelor project I developed a triggering system with an Arduino that can be controlled with micro-manager. This was done by rewriting the existing firmware and I introduced a new device adapter that is available in the newer versions micro-manager. The Arduino triggering system was tested in a proof-of principle experiment, where it was successfully used to image single molecule imaging in the presence of bulk eGFP. This triggering system can be used in future experiment that need to image single-fluorescent mCherry molecules in a bulk of eGFP molecules.
7. References
Chalfie, M. (2009). GFP: lighting up life (Nobel Lecture). Angewandte Chemie International Edition, 48(31), 5603-5611.
Day, R. N., & Davidson, M. W. (2009). The fluorescent protein palette: tools for cellular imaging. Chemical Society Reviews, 38(10), 2887-2921.
Doherty, G. P., Bailey, K., & Lewis, P. J. (2010). Stage-specific fluorescence intensity of GFP and mCherry during sporulation in Bacillus subtilis. BMC Research notes, 3(1), 303.
23 Enuka, Y., Hanukoglu, I., Edelheit, O., Vaknine, H., & Hanukoglu, A. (2012). Epithelial sodium channels (ENaC) are uniformly distributed on motile cilia in the oviduct and the respiratory airways. Histochemistry and cell biology, 137(3), 339-353.
Hirzel, T. (n.d.). PWM. Arduino: Learning Foundations. https://www.arduino.cc/en/Tutorial/PWM. Horne, A. W., & Critchley, H. O. (2012). Mechanisms of disease: the endocrinology of ectopic pregnancy. Expert reviews in molecular medicine, 14.
Horani, A., & Ferkol, T. W. (2018). Advances in the genetics of primary ciliary dyskinesia: clinical implications. Chest, 154(3), 645-652.
Howard, J. (2001). Mechanics of motor proteins and the cytoskeleton.
Jones, H. F. (1998). Continuous Groups (SO(N)). In Groups, representations and physics (pp. 116). CRC Press.
Kaletta, T., & Hengartner, M. O. (2006). Finding function in novel targets: C. elegans as a model organism. Nature reviews Drug discovery, 5(5), 387.
van Krugten, J., & Peterman, E. J. (2018). Single-molecule fluorescence microscopy in living Caenorhabditis elegans. In Single Molecule Analysis (pp. 145-154). Humana Press, New York, NY. Leung, M. C., Williams, P. L., Benedetto, A., Au, C., Helmcke, K. J., Aschner, M., & Meyer, J. N. (2008). Caenorhabditis elegans: an emerging model in biomedical and environmental toxicology. Toxicological sciences, 106(1), 5-28.
Leung, M. C., Williams, P. L., Benedetto, A., Au, C., Helmcke, K. J., Aschner, M., & Meyer, J. N. (2008). Caenorhabditis elegans: an emerging model in biomedical and environmental toxicology. Toxicological sciences, 106(1), 5-28.
Mangeol, P., Prevo, B., & Peterman, E. J. (2016). KymographClear and KymographDirect: two tools for the automated quantitative analysis of molecular and cellular dynamics using kymographs. Molecular biology of the cell, 27(12), 1948-1957.
Mamontova, A. V., Grigoryev, A. P., Tsarkova, A. S., Lukyanov, K. A., & Bogdanov, A. M. (2017). Struggle for photostability: bleaching mechanisms of fluorescent proteins. Russian Journal of Bioorganic Chemistry, 43(6), 625-633.
Müller, R. U., Zank, S., Fabretti, F., & Benzing, T. (2011). Caenorhabditis elegans, a model organism for kidney research: from cilia to mechanosensation and longevity. Current opinion in nephrology and hypertension, 20(4), 400-408.
NXP Semiconductors, Level shifting techniques in I2C-bus design, AN10441, Rev. 01, June 2007. Ou, G., Blacque, O. E., Snow, J. J., Leroux, M. R., & Scholey, J. M. (2005). Functional coordination of intraflagellar transport motors. Nature, 436(7050), 583.
Stokes, G. G. (1852). XXX. On the change of refrangibility of light. Philosophical transactions of the Royal Society of London, (142), 463-562
Strack, R. L., & Jaffrey, S. R. (2014). Using RNA Mimics of GFP to Image RNA Dynamics in Mammalian Cells. In Fluorescence Microscopy (pp. 83-91). Academic Press.
van den Wildenberg, S. M., Prevo, B., & Peterman, E. J. (2011). A brief introduction to single-molecule fluorescence methods. In Single Molecule Analysis (pp. 81-99). Humana Press.
24
8. Appendix
DB-25 connector on AOTF Pin, (DB–25 connector) Function
1 TX (RS232) – only on MPDS not MDS
2 RX (RS232) – only on MPDS not MDS
3 RAMP OUT 0-3.3V (Sweep FCT)
4 Enable(1) / Latch(0) Pin Profile FCT
5 Channel 8 6 Channel 7 7 Channel 6 8 Channel 5 9 Channel 4 10 Channel 3 11 Channel 2 12 Channel 1 13 Blanking 14, 15, 16 Ground 17 Reset
18 Bit 0 (LSB) Pin Profile FCT
19 Bit 1 Pin Profile FCT
20 Bit 2 (MSB) Pin Profile FCT
21 OPT FCT1 (NC)
22 OPT FCT2 (NC)
23, 24, 25 24 VDC
25 Figure 16: The DB-25 Connector and pin numbering
The DB-25 Connector pins
http://www.electroniccircuitsdesign.com/pinout/rs232-db25-pinout.html
9. Populairwetenschappelijke samenvatting van
Bachelorthesis
Zichtbaar maken van enkele verschillende eiwitten onder microscoop
Dit bachelorproject ging over het zichtbaar maken van enkele eiwitten in een cel. Dit werd gedaan door de eiwitten apart te belichten. Het licht voor ieder eiwit kwam vanaf een laser dat de eiwit zichtbaar maakt. Het licht van de lasers werd aangestuurd door gebruik te maken van een filter dat elektronisch werd aangestuurd door een Arduino. Een Arduino is een relatief goedkoop ontwikkelingsboordje met daarop een microprocessor (eenvoudige computer dat gebruikt wordt om apparaten aan te sturen) dat makkelijk geprogrammeerd kan worden door studenten en hobbyisten.
Bij het bestuderen van cellen wordt vaak gebruik gemaakt van een microscoop. Nu bestaan er veel verschillende soorten microscopen. Een van de meest gebruikte microscopen is de fluorescentie microscoop. Fluorescentie is een fenomeen waarbij bepaalde stoffen een andere kleur licht uitzenden dan die ze ontvingen. Vaak is deze kleur ‘roder’ dan het licht dat de fluorescente stof ontvangt. Een bekend voorbeeld van fluorescentie is disco-kleding die oplicht als er met black-light-lamp op wordt geschijnd. In dit voorbeeld zendt een ‘black-light-lamp’ UV licht uit, dit UV licht wordt geabsorbeerd door de verf. De verf zet dit UV licht vervolgens om een zichtbaar licht. Het verschilt per stof welke kleuren licht ze kunnen omzetten.
Fluorescentie niet alleen gebruikt bij black-light feestjes maar wordt ook gebruikt in biologisch onderzoek en dan met name in de fluorescentie microscopie. Bij fluorescentie microscopie worden bepaalde eiwitten in een cel gelabeld met een fluorescente stof. Door met een speciaal licht te schijnen beginnen de eiwitten op te lichten in de cel waardoor ze onder een microscoop zichtbaar worden. Hieruit kunnen een stel van eigenschappen van het eiwit achterhaald worden zoals de locatie van het eiwit in de cel.
26 Door twee verschillende soorten eiwitten te labelen met verschillende fluorescente stoffen kan ook zichtbaar gemaakt worden of de eiwitten samenwerken in de cel. Er is echter een probleem, veel fluorescente stoffen hebben een overlap in de kleuren die ze uitzenden! Dit is vooral een groot probleem wanneer eiwitten zichtbaar zijn van een soort ten opzichte van de andere soort. Dit probleem kan opgelost worden door de eiwitten om de beurt te belichten en dan een foto te maken onder de microscoop. Hierdoor is per foto maar een eiwit zichtbaar.
Voor het goed zichtbaar krijgen van de eiwitten op de foto’s is het belangrijk de juiste hoeveelheid licht te krijgen. Bij te veel licht worden de eiwitten te snel gebleekt waardoor ze niet meer zichtbaar zijn onder de microscoop. Bij te weinig licht zijn de eiwitten niet zichtbaar onder de microscoop. Voor het regelen van de hoeveelheid licht is gekozen voor een speciaal elektronisch filter. Dit filter werd aangestuurd met een Arduino.
Dit bachelorproject laat zien dat hoe goedkope Arduinos gebruikt kunnen worden voor het aansturen van microscoop apparatuur.
