Copyright by Sertan Kutal Gokce 2016
The Dissertation Committee for Sertan Kutal Gokce Certifies that this is the approved version of the following dissertation:
Parallel and Serial Microfluidic Platforms for Femtosecond Laser Axotomy in Caenorhabditis elegans for Nerve Regeneration Studies
Committee:
Adela Ben-Yakar, Supervisor
Mikhail A. Belkin
Andrew Dunn
Mark F. Hamilton
Jon Pierce-Shimomura
Preston S. Wilson Parallel and Serial Microfluidic Platforms for Femtosecond Laser Axotomy in Caenorhabditis elegans for Nerve Regeneration Studies
by
Sertan Kutal Gokce, B.S.; M.S.
Dissertation Presented to the Faculty of the Graduate School of The University of Texas at Austin
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
The University of Texas at Austin May 2016 Dedication
Annem Arife Çay’a, memleketim Adana’ya ve Galatasaray’a.
To my mom Arife Çay, my hometown Adana, and Galatasaray. Acknowledgements
First, I would like to thank my advisor Dr. Adela Ben-Yakar for her guidance, encouragement, and financial support throughout my Ph.D. studies at The University of Texas at Austin. Her enthusiasm and passion for multidisciplinary research and neuroscience have inspired numerous great projects such as the ones described here. I would like to thank my committee members Dr. Mikhail Belkin, Dr. Andrew Dunn, Dr. Mark F. Hamilton, Dr. Jon Pierce-Shimomura, and Dr. Preston S. Wilson. They have been supportive throughout my Ph.D. study. In particular, my most sincere appreciation is extended to Dr. Jon Pierce-Shimomura and Dr. Preston S. Wilson for their help and guidance during my Ph.D. studies and their valuable advice for my future career. I wish to thank all my past and present fellow lab members for making this Ph.D. journey interesting and fun and for providing their support when I struggled. I would like to thank especially Dr. Onur Ferhanoglu, Dr. Sudip Mondal, Dr. N. Ghorashian, Travis Jarrell, Aubri Kottek, Dr. Sudip Mondal, Dr. Ryan Doonan, Dr. Ki Hyun Kim, Dr. Neil Everett, Evan Hegarty, Christopher Michael Martin, Dr. Frederic Bourgeois, Sam Aminfard, Dr. Ilan Gabay, Dr. Dan Eversole, Peisen Zhao, Kaushik Subramanian, and Dr. Fred Li.
Next, I would like to thank all JPS lab members, especially Dr. Luisa Scott, Dr. Andres Vidal-Gadea, Sarah Nordquist, and Claire Celia Beron for always being helpful and listening to my problems regarding research. Next, I would like to thank Dr. Sydney Geissler, Anna Warden, and Roberto Ulises Cofresí for helping me with editing my dissertation and for answering my questions regarding biology.
v My sincere appreciation is extended to my dear friends Dr. Onur Kacar, Hande Gerkus, Dr. Yiorgos Zalachoris, Elif Zeynep Iyriboz, Dr. Can Hankendi, Batur Isdiken, Dr. Sanjiv Shah, Funda Donmez, Onur Domanic, Gokhan Yildiz, and Quack’s Bakery crew. I want to thank them for making my life in Austin joyful and sharing my Ph.D. journey. I want to thank my good friends in Turkey, who shared my Ph.D. journey over a very long distance. I want to thank Denizcan Soner, Umut Toklu, Murat Atak, Sabit Sagir, Emre Aytekin, Baran Evliyaoglu, Can Yagli, Fidel Berber, Onder Ucar, Utku Ozkan, Onur Gokpinar, Ilker Gezer, Mert Pala, Arda Bahadir, and Sencer Ucar. Last, but never the least, I would like to declare my deepest appreciation and greatest gratitude to my father, Salih Gökçe, my wonderful mom, Arife Çay, my grandparents who always believed in me and wanted me to pursue higher education, Ali Şahin Çay and Hatice Çay, my super aunt, Mürüvvet Çay, my aunts, Sacide Çay and Ülviye Gürbüz, my uncle, Yasin Çay. I do not believe I would ever have been able to seek higher education without your endless support and love. I also wish to recognize the many people that have made this dissertation a reality. In addition, I apologize in advance for the many people who I forgot to thank here.
vi Parallel and Serial Microfluidic Platforms for Femtosecond Laser Axotomy in Caenorhabditis elegans for Nerve Regeneration Studies
Sertan Kutal Gokce, Ph.D. The University of Texas at Austin, 2016
Supervisor: Adela Ben-Yakar
Understanding the molecular basis of nerve regeneration can potentiate the development of novel and efficient treatments for neurodegenerative diseases. Severing axons in the small nematode Caenorhabditis elegans (C. elegans) with femtosecond laser surgery and then observing the subsequent axonal regrowth is a promising approach to understand the molecular mechanisms of nerve regeneration in vivo. Effective and reversible immobilization of the nematodes is necessary for both axotomy and follow-up imaging. However, conventional worm handling techniques such as using anesthetics or polystyrene beads are labor-intensive and time-consuming processes, hindering high- throughput. Microfluidic devices enable the manipulation of the nematodes on a single chip with unprecedented throughput and integrity. This dissertation introduces two comprehensive microfluidic systems for femtosecond laser axotomy in C. elegans, offering several advantages over conventional techniques in terms of speed and the ability for automation of the tedious axotomy experiments. The first microfluidic system is an automated serial microfluidic platform for vii performing femtosecond laser axotomy in C. elegans. The microfluidic platform along with a custom-developed automation program isolates a single nematode from a pre-loaded population, immobilizes the nematode, and performs femtosecond laser axotomy. The full automation of the axotomy process is achieved by combining efficient image analysis methodologies with synchronized valve and flow progression in the microfluidic chip to perform multiple surgeries in a serial and automated manner. The serial automated microfluidic platform reduces the time required to perform axotomies within individual worms to ~ 17 s/worm. The second microfluidic system is a parallelized multitrap microfluidic platform, “worm hospital”, that allows on-chip axotomy, post-surgery housing for recovery, and imaging of nerve regeneration on a single chip. The microfluidic platform features 20 trapping channels for laser axotomy and subsequent post-surgery imaging, and a perfusion area to house the worms after laser axotomy. This microfluidic platform is a single-flow
Polydimethylsiloxane (PDMS) layer device and has no active control PDMS layer, which reduces the fabrication and operation complexity of the chip, especially for non-expert users. The roles of neurodevelopmental genes in the Wnt/Frizzled pathway on the nerve regeneration was investigated using the “worm hospital”. In summary, the microfluidic platforms presented in this dissertation enabled performing femtosecond laser axotomy in C. elegans in a fast and repeatable manner with a controllable microenvironment. Both microfluidic platforms offer promising methodologies for prospective large-scale screening of genes involved in nerve regeneration with a high throughput in an automated manner.
viii Table of Contents
List of Tables ...... xii
List of Figures ...... xiii
Chapter 1: Introduction ...... 1 1.1 Dissertation overview and objectives ...... 2
Chapter 2: Background ...... 4 2.1. C. elegans as a model organism:...... 4 2.2 Femtosecond laser ablation ...... 5 2.3 Femtosecond laser axotomy in C. elegans ...... 7 2.4 Microfluidics Overview ...... 8 2.5 Microfluidic devices for C. elegans research ...... 11
Chapter 3: The Fully Automated Serial Femtosecond Laser Axotomy Chip ...... 15 3.1 Introduction ...... 15 3.2 System Overview ...... 17 3.3 Microfluidic Device Design ...... 19 3.4 Progression of valve actuation, flow, and worm manipulation processing25 3.5 Image processing methodology for automated identification of neurons and targeting for laser axotomy ...... 28 3.5.1 Step 1: Identification of the worm location ...... 31 3.5.2 Step 2: Detection of a neuronal cell body in the small FOV ...... 33 3.5.3 Step 3: Verification of neuron of interest ...... 35 3.5.4 Step 4: Laser axotomy ...... 38 3.6 Characterization of the automated platform ...... 39 3.6.1 Effect of chip manipulation on worm survivability ...... 39 3.6.2 Timing and axotomy success rates ...... 40 3.6.3 Automated on-chip surgery axonal reconnection rates ...... 44 3.7 Experimental Procedures ...... 46 3.7.1 Device fabrication methods ...... 46 3.7.2 Opto-mechanical setup...... 48 ix 3.7.3 C. elegans maintenance ...... 50 3.7.4 Laser axotomies and post-surgical imaging on agar pads ...... 51 3.7.5 Microfluidic chip priming and preparation for axotomies ...... 51 3.8 Conclusions ...... 52
Chapter 4: A Microfluidic “Worm Hospital” for Axotomy, Recovery, and Imaging of Caenorhabditis Elegans ...... 54 4.1. Introduction ...... 54 4.2. Device design and operation ...... 57 4.2.1. Microfluidic device design ...... 57 4.2.2. Microfuidic flow process ...... 61 4.3. Results ...... 62 4.3.1. Trapping and orientation characterization ...... 62 4.3.2. Viability test ...... 67 4.3.3. Axonal regeneration results ...... 69 4.3.4. Axonal growth and guidance, and neuronal polarity genes ...... 71 4.4 Experimental Methods ...... 77 4.4.1 Device fabrication ...... 77 4.4.2. Opto-mechanical setup...... 78 4.4.3. System operation and control...... 79 4.4.4 Laser axotomy and post-surgical imaging ...... 80 4.4.5 Nematode maintenance ...... 80 4.4.5. Neuronal RNAi hypersensitive strain for axotomy assays ...... 81 4.4.6. On-chip immobilization with the aid of levamisole ...... 82 4.5 Conclusions ...... 83
Chapter 5: Conclusions and Future Work ...... 85 5.1 Summary of the dissertation ...... 85 5.2 Future Work ...... 86
x Appendix A: Supplementary Software Document for the Serial Automated Microfluidic Femtosecond Laser Axotomy Platform for Nerve Regeneration Studies in C. elegans ...... 88 A.1 Software Instructions and List of Hardware Requirements: ...... 88 A.2 Hardware list ...... 88 A.3 Axotomy.vi ...... 89 A.3.1. START Tab (Figure A.1): ...... 91 A.3.2. CCD Camera Tab (Figure A.2): ...... 91 A.3.3 Motorized Stage Tab (Figure A.3): ...... 93 A.3.4 Shutter Tab (Figure A.4): ...... 95 A.3.5 Piezo Stage Tab (Figure A.5):...... 96 A.3.6 Valves Tab (Figures A.6-A.12): ...... 97 A.3.7 ROI Locations (Figure A.13): ...... 107 A.3.8 Automated Process Tab (Figure A.14): ...... 107
REFERENCES ...... 109
xi List of Tables
Table 4.1: Comparison of on-chip and on-agar axonal reconnection percentages of
ALM neurons in single mutant and RNAi experiments...... 71 Table 4.2: Comparison of on-chip and on-agar axonal reconnection percentages of
PLM neurons in single mutant and RNAi experiments...... 71 Table 4.3: On-chip axonal reconnection percentages of ALM neurons in single
mutant animals...... 73 Table 4.2: On-chip axonal reconnection percentages of ALM neurons with RNAi
gene knockdown...... 74 Table 4.5: On-chip axonal reconnection percentages of PLM neurons in single mutant
animals...... 75 Table 4.6: On-chip axonal reconnection percentages of PLM neurons with RNAi gene
knockdown...... 75
xii List of Figures
Figure 2.1: The schematic view of Larval 1 (L1) stage of C. elegans...... 5 Figure 2.2: Illustration of the series of mechanisms involved in plasma-induced
optical breakdown ...... 6
Figure 2.3: Femtosecond laser axotomy in C. elegans ...... 8
Figure 2.4: Fabrication process flow of multilayer soft lithography ...... 10
Figure 2.5: Two basic configurations for multi-layer microfluidic valves...... 11
Figure 2.6: Recent examples of microfluidic laser surgery platforms...... 14
Figure 3.1: Axotomy chip overview...... 18
Figure 3.2: Worm delivery challenges in the trapping area ...... 20 Figure 3.3: Flow direction in T-shaped staging/trapping area and 3D interconnects.
...... 24
Figure 3.4: Automated progression of valve actuation and flow...... 26
Figure 3.5: The soft touch neurons of C. elegans: ...... 29
Figure 3.6: Automation Flowchart...... 30 Figure 3.7: Image processing methodology for identifying the worm location within
the trapping area (Step 1)...... 32 Figure 3.8: The circular object detection methodology to detect a cell body in the
field of view (Step 2)...... 34 Figure 3.9: Fine focusing methodology to verify the neuron of interest and determine
the orientation of the axon...... 36 Figure 3.10: Image processing steps to verify the neuron of interest (Step 3), locate
the target on the desired axon, and perform laser axotomy (Step 4).37
Figure 3.11: Lifespan analysis...... 40
xiii Figure 3.12: Average process timings per worm for different experiments ...... 41
Figure 3.13: Success rate and timing performance of the automation...... 43
Figure 3.14: Imaging of regrowing axons of interest 24 hours after surgery...... 45
Figure 3.15: Axonal reconnection results...... 46
Figure 3.16: Schematic of the serial automated laser axotomy setup...... 50
60
Figure 4.1: Worm Hospital overview...... 60
Figure 4.2: The flow progress on the serial microfluidic chip ...... 62
Figure 4.3: Trapping and retrapping efficiency characterization ...... 66
67
Figure 4.4: Head-tail orientation characterization...... 67
Figure 4.5: Lifespan analysis...... 68 Figure 4.6: On-chip imaging of axonal regrowth and reconnection 24 h after surgery,
and comparison of on-chip axotomy performance to on-agar axotomy
using anesthetics...... 70 Figure 4.7: On-chip axonal reconnection results for the selected genes in
WNT/Frizzled family...... 76
Figure A.1: The front panel of Axotomy.vi with “START” tab chosen...... 90
Figure A.2: “CCD Camera” Tab...... 92
Figure A.3: “Motorized Stage” Tab...... 94
Figure A.4: “Shutter” Tab...... 96
Figure A.5: “Piezo Stage” Tab...... 97
Figure A.6: “Valves” tab and “Main Valve Control” sub-tab chosen...... 100
Figure A.7: Two elements of Pre-Loading Sequence: ...... 101
Figure A.8: Four elements of the Injection Sequence: ...... 102 xiv Figure A.9: Immobilization Sequence: ...... 103
Figure A.10: Two Elements of Ejection Sequence: ...... 104 b) 105
Figure A.11: Two Elements of Flush Sequence: ...... 105
Figure A.12: Elements of cleaning sequence: ...... 106
Figure A.13: The front panel of Axotomy.vi with “ROI Locations” tab chosen. 107 Figure A. 14: The front panel of Axotomy.vi with “Automated Process” Tab chosen.
...... 108
xv Chapter 1: Introduction
The limited ability of neurons to regenerate, especially in the mature central nervous system (CNS), leads to permanent deficiencies in the nervous system after an injury [1, 2]. The inability of neurons to regrow after an injury is attributed to both intrinsic [3-5] and extrinsic [6-8] mechanisms. Despite many efforts to understand the mystery behind nerve regeneration, the molecular mechanisms of neural regrowth still remains elusive. Understanding these mechanisms could lead to the development of novel and effective therapeutics for brain trauma, stroke, Parkinson’s disease, Alzheimer’s disease, and other neurodegenerative illnesses. These disorders affect countless people globally, leading to an excess economic and social burden worldwide [9, 10]. A promising approach of elucidating the molecular basis of regeneration in vivo is to severe axons in the small model organism Caenorhabditis elegans (C. elegans) and observe the post-surgery regeneration. Specifically, nerve regeneration and degeneration processes can be studied by axotomizing the neurons of living and intact C. elegans with ultrafast laser axotomy [11]. Since the first demonstration of the successful regeneration of neurons after femtosecond laser axotomy [11], this technique has spurred many studies to understand the molecular basis of in vivo nerve regeneration [3, 12-20]. The advantage of femtosecond laser surgery is its ability to create a sub-micron ablation volume with minimal damage to the surrounding tissue, which permits the study of nerve regeneration after the surgery in vivo. However, nerve regeneration studies in living C. elegans have been hampered by the duration and complexity of this method, since laser axotomy requires immobilization of the worm so that the axon of interest can be precisely positioned within the focal volume of the laser beam. In addition, a complete immobilization of the nematodes is required for high-resolution imaging of the
1 regenerating axons post injury. Conventional immobilization methods include anesthetics [3, 18, 21], polystyrene beads [22, 23], and gluing worms [24]. Gluing worms is not a viable option for nerve regeneration studies because the worms cannot be recovered for post-injury follow-ups. Other methods require manually orienting nematodes with a platinum tip to get the neuron of interest near the glass cover slip interface. These tasks must be accomplished without losing or harming the animals during their recovery from anesthetics or polystyrene beads via liquid buffer exchange [11, 22]. These conventional immobilization techniques require picking, sorting, and transferring of individual worms, hindering the throughput of the axotomy experiments. Using the conventional techniques, large-scale assays of nerve regeneration studies such as gene knockout and RNA interference (RNAi) would take years to complete. Advancements in the capability of microfluidic technologies of fluid enabled manipulation of handling small volumes an automated, precise, and consistent fashion. Since the dimensions of the nematode’s body are in the micron to millimeter scale, the worms can be processed using microfluidic platforms that provide a controllable experimental environment and can enhance the automation of worm handling processes. To increase the throughput and improve the repeatability of experiments, the microfluidic technology has been used for a variety of C. elegans studies. Consequently, microfluidic platforms along with the automation hold a great promise of increasing throughput of axotomy experiments in C. elegans.
1.1 DISSERTATION OVERVIEW AND OBJECTIVES
The primary objective of this dissertation is to develop and test integrated, novel microfluidic platforms that allow femtosecond laser axotomy and imaging of post-recovery nerve regeneration in C. elegans in a rapid and controllable manner to understand molecular mechanisms behind nerve regeneration.
2 Chapter 2 first gives a brief review of the use of C. elegans as a model organism and examples of femtosecond laser axotomy in C. elegans. Then, microfluidic platforms for C. elegans assays and on-chip axotomy of C. elegans platforms are detailed, along with a comparison of traditional axotomy and on-chip axotomy techniques. Chapter 3 presents the development and implementation of the fully automated serial microfluidic platform. Particular attention is paid to the unique features of the axotomy platform enabling full automation of the process and the development of a custom-developed automation software to sever axons of the nematodes. Then, the success rates of and average timing for each automation step are presented. Finally, a comparison of axonal reconnection rates of on-chip and on-agar experiments using anesthetics are given for two different ablation energy conditions. Chapter 4 presents the design and characterization of the parallelized multitrap microfluidic platform, the “worm hospital”, that allows on-chip axotomy, post-surgery housing for recovery, and imaging of nerve regeneration on a single chip. First, the fluidic design of the chip is detailed. Particular focus is given on the on-chip trapping and retrapping efficiencies. Finally, the role of neurodevelopmental genes in the Wnt/Frizzled pathway on the regenerative capacity of the anterior lateral microtubule cells (ALM) and the posterior lateral microtubule cells (PLM) neurons is presented by utilizing the “worm hospital”. Chapter 5, the last chapter, summarizes the dissertation work and discusses the future research.
3 Chapter 2: Background
This chapter presents the background information on the scientific methodologies used in this dissertation. This chapter introduces the model organism (C. elegans) and emphasizes its use for nerve regeneration studies. Then, an introduction to microfluidic technology and related valve configurations is given. Lastly, information on microfluidic assays that address different biological problems will be presented, highlighting how the development of laser axotomy chips has advanced nerve regeneration studies in C. elegans.
2.1. C. ELEGANS AS A MODEL ORGANISM:
Since its introduction to the scientific community, the nematode C. elegans (see Figure 2.1) has been widely used as a model organism [25-27]. The small roundworm has been studied across a variety of fields of the biological sciences, such as development [28- 30], behavior [31-41], metabolism [42-46] and neurodegenerative diseases [47-51]. The interest in C. elegans research owes to both the simplicity and complexity of the nematode. First, the roundworm C. elegans can be easily bred in laboratory conditions. The animals can grow on agar plates seeded with Escherichia coli [52] or they can also grow in liquid cultures with bacteria (S-medium), which allows high-throughput assays [53-56]. Moreover, the development of worms from egg to adult stage of (~ 1 mm in length) occurs within only 3 days [57]. Second, due to its small size, short generation time, and mostly self-fertilizing reproduction, assays can be easily carried out on large isogenic populations. Third, the nematode has a transparent body that allows the optical interrogation of individual cells throughout the lifespan of the wormss. Fourth, C. elegans have one of the simplest and the best characterized nervous systems of any model organisms, and the entire neural circuitry is mapped [58], making the study of single neurons possible. Hermaphrodite nematodes have 302 neurons and males have 381 neurons along with male-
4 specific neurons [52]. Despite the small number of neurons, nematodes possess a wide range of complex behavioral modalities such as thermotaxis [59], chemotaxis [60], and odortaxis [36, 61], which all be studied in molecular and cellular detail [62]. Fifth, C. elegans is the first multicellular organism with a fully sequenced genome [63]. Moreover, highly significant genetic homology with humans [64-67], many conserved biological processes [49, 57, 68], and genetic amenability of C. elegans make the nematode an emergent model organism for drug discoveries for human diseases.
Figure 2.1: The schematic view of Larval 1 (L1) stage of C. elegans. The schematic shows the distinct body parts and the principal parts of the nervous system. The image is adapted from [69].
2.2 FEMTOSECOND LASER ABLATION
As the light intensity becomes higher, nonlinear effects become prominent, such as self-phase modulation, self-focusing, multiphoton absorption, and laser-induced optical breakdown [70]. Ultrafast laser technology offers very precise ablation capability with its ability to reach high peak power intensities that leads to lower ablation threshold values
[71, 72]. Precise ablation of a variety of materials has been shown using femtosecond near- infrared laser techniques [73-75].
5 The principle mechanism behind the low-energy and precise femtosecond laser ablations is the laser-induced optical breakdown [70, 71, 73]. Femtosecond laser ablation of transparent materials happens through a series of multi-photon absorption, tunneling, and avalanche ionization [72, 76]. First, the multiphoton absorption by the material is required to ionize an electron (Figure 2.2). Ionized electron becomes a seed for avalanche ionization, it starts absorbing more photons linearly and gaining kinetic energy through the multiple Inverse Bremsstrahlung process [76]. When the energy of the ionized electron reaches between approximately 1.5-2 times of the band gap energy, it can produce another free electron by impact ionization [72, 76]. The free electron density increases exponentially through avalanche ionization, leading to formation of a high density plasma within the focal volume. The exponential growth continues until the electron density reaches a critical limit where the frequencies of the plasma and incident laser beam match. Then the electron plasma passes on high kinetic energy to the ions, which leads to meting and vaporization of the material.
Figure 2.2: Illustration of the series of mechanisms involved in plasma-induced optical breakdown The ionization of valence electrons can occur through multiphoton absorption. A high density of ionized electrons are obtained through multiple Inverse Bremsstrahlung absorption, which is followed by impact ionization. The image is adapted from [76]. 6 2.3 FEMTOSECOND LASER AXOTOMY IN C. ELEGANS
It was only recently discovered that C. elegans neurons can regenerate shortly after the severing of an individual axons with the femtosecond laser ablation [11]. By focusing ultrafast laser pulses, Yanik et al. showed that it is possible to precisely sever a single axon in living and intact C. elegans without damaging the surrounding tissue [11]. Interestingly, the injured neurons could spontaneously regenerate within 24 hours of the laser axotomy, accompanied by functional recovery as in Figure 2.2. This method has been successfully adapted by the C. elegans community since its first demonstration twelve years ago [3, 12- 18, 77]. Axons have the ability to regenerate after being severed and regain their functionality in the nervous system. The regeneration process after an injury leads to changes in gene transcription and local protein synthesis. The process is regulated by different factors such as phylogenetically conserved pathways [78, 79], the neuronal environment [80], and the intrinsic state of the neuron [78, 81]. However, the roles of most of these components remain a matter of debate. Several studies have attempted to clarify the role of these components in nerve regeneration. However, these studies were limited to large animal models [82-87], which ultimately did not elucidate the regulatory mechanisms due to the lack of an adequate nerve injury method available only in these organisms. An organism with a simpler nervous system such as C. elegans¸ provides a feasible path towards studying molecular mechanisms behind nerve regeneration. The importance of using C. elegans as a nerve regeneration model is that several pathways affecting vertebrate axon outgrowth in humans have been found to shows similar effects in C. elegans [3, 30, 78]. There have been only two large-scale genetic screenings to identify pathways associated with axon regeneration in C. elegans [3, 17]. Despite the unprecedented scale of these studies, a large fraction of the worm’s genome still needs to 7 be tested for nerve regeneration pathways. This requirement can only be realistically met using high-throughput manipulation and axotomy platforms.
Figure 2.3: Femtosecond laser axotomy in C. elegans Fluorescence images of axons labelled with GFP before, immediately after, and in the hours following axotomy. The image is adapted from [11].
2.4 MICROFLUIDICS OVERVIEW
Microfluidic devices enable manipulation of fluids at the micron to nanometer scale in a controllable environment. Moreover, they provide a platform to perform precise experiments in life sciences with minimal use of chemical and biological reagents. The first microfluidic chips were made of silicon or glass [88, 89]. Typically, wet etching or deep reactive ion etching (DRIE) tools were employed to create channels of varying geometries in the substrate of choice for the desired application [90]. These etching processes were both time consuming, and labor-intensive. This was due to the need to perform photolithography and etching with advanced dry etching tools, and harsh chemicals for every batch of devices made. Dr. George Whiteside’s group developed a new lithography technique based on replicas of the silicon master molds using polydimethylsiloxane (PDMS) [91]. In soft lithography, typically, a pattern is defined in a photolithographic mask that is used to generate the pattern in a photosensitive material 8 (photoresist) that has been spin-coated onto a silicon wafer surface. The photoresist on the silicon wafer is used as the mold for PDMS or another chosen elastomer [92, 93]. The elastomer is cured on the silicon mold for hardening and then is peeled of the master mold (silicon wafer). The hardened elastomer is then bonded to glass, silica or more PDMS. One apparent advantage of the soft lithography is that it allows fabrication of multiple replicas of the PDMS chip from a single master mold. Recent advances in soft lithography techniques made the microfluidic technology an ubiquitous tool for biological sciences over the last couple decades. One of the main breakthrough in soft lithography came when Dr. Stephen Quake’s group introduced multi-layer microfluidic chips with unprecedented robustness and design flexibility [94, 95]. At least two master molds are needed to fabricate multilayer microfluidic devices. One of the layers serves as a flow layer and the other as the control layer. For the bottom layer, PDMS elastomer is spin-coated across the mold so a 10-30 µm PDMS layer rests above the photoresist features. After the PDMS of the bottom layer is partially hardened, the top layer of the device, which is typically fabricated as its own single layer in the single microfluidic device fabrication, is bonded to the bottom layer, (see Figure 2.4). The membranes are formed where two layers intersect.
9 Figure 2.4: Fabrication process flow of multilayer soft lithography Two master molds are fabricated using photolithography. One master mold is for replicating the bottom channel (left) on which a thin layer of PDMS is coated by spin-coating. The PDMS replica that was created by the other master mold is orthogonally aligned on master mold coated with the thin PDMS layer, which creates a deflectable thin membrane between the two mold replicas of the channels. The image is adapted from [94]. There are two valve configurations for multilayer microfluidic device fabrication ; push-up and push down [96]. Push-up valves control the flow through a channel from below whereas push-down valves control the flow through a channel from above. Both valve structures use thin deflectable membranes between fabrication layers to control the flow channel (Figure 2.4). With the push-up valve configuration, completely sealed
10 channels can be achieved while push-down configuration can offer only partially sealed channels.
Figure 2.5: Two basic configurations for multi-layer microfluidic valves. a) A PDMS push-down valve, where the control layer is on top of the flow channel with a semi-round cross-section. b) A push-up valve, which has the opposite structure as in part (a). Thin deflectable membranes are used in both cases to close the flow channel [96].
2.5 MICROFLUIDIC DEVICES FOR C. ELEGANS RESEARCH
The conventional worm handling techniques at present include the use of petri dishes or standard multiwell plates [13, 97]. For more precise interrogation and manipulation of the worms, researchers immobilize the nematodes using anesthetics, polystyrene beads or gluing worms to the substrates.
One of the initial examples of microfluidic chips for C. elegans applications was a chemotaxis chip [98]. In this particular study, the authors designed a PDMS device and placed it on an agar surface. The microfluidic chip created an oxygen gradient, which allowed studies of hyperopia avoidance in wild type and mutant worms. In the last decade, there have been several examples of microfluidic platforms for applications such as behavioral studies [31, 33, 36, 98-110], developmental studies [111-117], phenotyping [115, 118-126], and laser surgery [12, 16, 127-130]. 11 In regards to the aim of the dissertation, the discussion will focus on immobilization of living nematodes on microfluidic chips from here on. Nerve regeneration studies in living C. elegans are time consuming and labor intensive since laser axotomy requires complete immobilization of the worm so that the axon of interest can be precisely positioned within the focal volume of the laser beam. In addition, high-resolution imaging of the regenerating axons post injury is necessary, which also requires a high degree of immobilization. Microfluidic immobilization techniques for in vivo studies typically involve mechanical trapping assisted by either a tapered channel, pressurized membranes [12, 13, 16, 128], exposure to a cold (4 °C) fluid to induce temporary paralysis [118], di- electrophoresis [131], or surface acoustic wave manipulation [132]. In one of the first examples of microfluidic surgery platforms, Chung et al. developed a microfluidic platform to ablate cell bodies at a rate of 110 worms per hour [129]. This microfluidic platform constituted three separate layers, including a cold-fluid layer, restriction valve layer, and flow layer (Figure 2.6a). The microfluidic chip employed two worm-loading channels that operate in parallel. This design enabled simultaneous worm-loading and worm-collecting on single chip. The immobilization of L1 stage worms was achieved using temperature-controlled layer. With this layer, the worms in the loading area were exposed to a cold fluid (4˚C) to induce temporary paralysis. After the immobilization of worm was achieved, the ablation of cell bodies was performed by a custom-developed automation program [129]. The most common on-chip immobilization technique for the laser axotomy is the pressurized membranes to exert adequate force for complete immobilization [12, 13, 128]. Guo et al. showed one of the first examples of an on-chip femtosecond laser axotomy platform. The microfluidic platform consisted two positioning channels, one immobilization membrane, and recovery chambers (Figure 2.6b). To immobilize the 12 worms, pressure was applied in the second layer, deforming the membrane above the trapping area. This actuation pushed the worm against the cover glass, allowing an optimum optical access to the neuron of interest [69]. Utilizing the double layer microfluidic platform, researchers observed that regeneration in mechanosensory neurons was much faster when worms were processed on chip, as opposed to those mounted on agar pads with anesthetics [12]. In another attempt towards on-chip femtosecond laser axotomy on a microfluidic platform, researchers developed a semi-automated platform to screen chemical compounds for nerve regeneration studies [16]. In this method, the worms were loaded from multiwell plates tilted at an angle to condense the worms into the corner of each well instead of delivering them via syringe (right panel on Figure 2.6c). With this study, researchers revealed specific chemical modulators of neurite regrowth in the PLM neurons. To summarize, microfluidic platforms for robust and automated on-chip laser axotomy of C. elegans hold great promise to enable high-throughput axotomy assays for nerve regeneration studies. The following two chapters present my effort on the development, testing, and utilization of two comprehensive microfluidic platforms that provide technology that will move researchers closer to this goal.
13 Figure 2.6: Recent examples of microfluidic laser surgery platforms. a)The left panel shows the microfluidic laser ablation platform filled with dye [129]. The active components are: loading channel (yellow), control membrane channel (red), and temperature-control channel (blue). Right panel shows a worm loaded in the loading channel. The image is adapted from [129]. b) The laser axotomy chip for imaging, laser nanoaxotomy [12], and housing of C. elegans. Left panel is an overview of the nano-axotomy platform. View of the trapping system: Valves 1–4 (yellow rectangles) respectively control inlet regulation (1), fine positioning of the worm (2 and 3) and gating to the recovery chambers (4).The right panel shows the conceptual three-dimensional section renderings of the trap channel without and with a worm (green) immobilized by a membrane. The image is adapted from [12]. c) A few worms are delivered to the device, where a single worm is trapped and immobilized for axotomy and imaging after cleaning steps. c) Left panel shows the microfluidic device with key components of the immobilization area filled with different dyes. The syringe-free loading of the worms into the devices was depicted on the left panel.
14 Chapter 3: The Fully Automated Serial Femtosecond Laser Axotomy Chip
This chapter discusses the fully automated serial microfluidic chip to perform femtosecond laser axotomy in C. elegans. The automated microfluidic platform is capable of isolating a single worm at a time from a large population, immobilizing the isolated worm, identifying the location of the worm, detecting the neuron of interest and relative location of axon, and axotomizing the axon in an automated manner. The content of this chapter has been adapted from reference [127].1
3.1 INTRODUCTION
In vivo nerve regeneration studies in living C. elegans are time consuming because the laser axotomy requires complete immobilization of the worm so that the axon of interest can be precisely positioned within the focal volume of the laser beam. In addition, high- resolution imaging of the regenerating axons post injury is necessary, which also requires a high degree of immobilization. The traditional methods to immobilize C. elegans for surgery or imaging include the use of anesthetics [11, 78, 79, 133], gluing worms to substrates [24] or using polystyrene beads [22]. Using anesthetics for immobilization might interfere with the recovery of the worms [134] and, thus, the regeneration process of the injured axons. Gluing methods are not viable methods since the worms cannot be recovered after gluing [24]. Using polystyrene microsphere beads require use of buffer solution exchange to recover the worms, limiting the throughput to around 5-10 worms per experimental set [22, 23]. These limitations of the conventional immobilization methods
1 S. K. Gokce, S. X. Guo, N. Ghorashian, W. N. Everett, T. Jarrell, A. Kottek, A. C. Bovik, and A. Ben-Yakar, "A fully automated microfluidic femtosecond laser axotomy platform for nerve regeneration studies in C. elegans," PLoS ONE, vol. 9, p. e113917, 2014. S.K.Gokce, S. X. Guo, N. Ghorashian, W. N. Everett, and A. Ben-Yakar conceived and designed the experiments. S.K.Gokce performed all experiments. S.K.Gokce and A. Ben-Yakar analyzed the data. S. K. Gokce, S. X. Guo, N. Ghorashian, W. N. Everett, T. Jarrell, A. Kottek, A. C. Bovik, and A. Ben-Yakar contributed to reagents and tools. 15 hinder the throughput of the axotomy processes. To overcome the limitations of manual techniques, new immobilization methods using microfluidic devices have been developed for both imaging [102, 105, 118, 131, 132, 135-138] and surgical studies [12, 16, 128, 129]. Microfluidic immobilization techniques for in vivo studies include, but are not limited to, mechanical trapping assisted by either tapered channels [102, 135, 136, 139], pressurized membranes [12, 16], pressurized membranes with the addition of suction [128], CO2- induced paralysis [13, 116], exposure to cold (4 °C) fluid to induce temporary paralysis [118], dielectrophoresis [131], gel based [140, 141], and surface acoustic wave manipulation [132]. Automation of these microfluidic platforms and the related imaging, and surgery processes is necessary to enable investigation of nerve regeneration in C. elegans at high speeds and with a high-throughput. To achieve full automation of the axotomy process, a serial automated microfluidic platform has been developed and tested, which is presented in this chapter. The full automation was achieved by redesigning our group’s previous microfluidic immobilization method [12] to enable repeatable and rapid immobilization of individual worms. Our group previously demonstrated that a microfluidic immobilization technique based on a deflectable membrane could provide sufficient immobilization to perform precise femtosecond laser axotomy in C. elegans [12]. However, this first device required manual interventions to reduce errors in immobilization orientation and the number of worms trapped at a given time that limited automated identification of the worm body and neurons of interest. To achieve repeatable and rapid laser axotomy of nematodes in a fully automated manner, a new microfluidic was designed using the deflecting membrane immobilization technique. The fully automated serial axotomy platform utilizes novel image processing algorithms to enable the automated ablation of 300 nm-wide axons
16 with a sub-micron resolution. The entire process of staging, trapping, and axotomy requires, on average, about 17 s/worm.
3.2 SYSTEM OVERVIEW
The serial microfluidic chip is designed to perform femtosecond laser axotomies in a fast and controllable manner on single worms without affecting the worm viability and without user intervention during operation. To simplify hardware manipulation and workflow in the optical setup, and to enhance precision of the axotomy process, a single trapping location for optical interrogation was implemented. Unlike most other nerve regeneration studies, this system uses a high-NA oil immersion objective and well-tuned laser parameters to increase axotomy precision, and decrease the variation of regeneration and reconnection results of the highly stochastic axotomy process between each experimental set. As previously shown, the regeneration capability of injured axons directly depends on the precision of laser injury and extent of damage [142]. When the damage to the surrounding tissues is minimized and the axon is severed at a pre-synaptic location, the severed axons could not only regrow, but also reconnect to their distal parts [18, 143, 144] thus restoring at least the neural connectome [11]. Further, high precision requires minimizing the aberrations by directly immobilizing the worms on the cover glass without an intervening layer of PMDS, air, or liquid. Other nerve regeneration studies, that have relied on more relaxed focusing parameters using low-NA air objectives as opposed to functional recovery [16, 145]. To enable high-precision laser axotomies, our group has previously utilized a pressurized membrane to immobilize worms directly against the optical interface for surgery [12]. By eliminating an extra layer of PDMS or dead channel volume between
17 worms and the cover glass, the deflectable membrane technology offered ideal optical focusing and accuracy to study the complete regeneration process [12].
Figure 3.1: Axotomy chip overview. A schematic top-view of the microfluidic chip. The loading chamber is used to house the worms throughout the automation process. The staging area isolates a single worm from the whole population kept in the loading chamber and sends it to the trapping area for immobilization and axotomy. The T-shaped trapping area with a sieve structure enables rapid immobilization of single worms. 3D interconnects enable access to the fully sealing valves to enhance the repeatable localization of worms against the sieve structure in the trapping area by completely stopping the flow to the exit outlets. After each axotomy, the flushing inlet guides the flow to take the worm through one of the outlets. In this current serial microfluidic design, the chip was redesigned into a unique configuration to enable its full automation along with adopting the same immobilization methodology as previously utilized [12]. The new serial microfluidic chip contains five distinct compartments (Figure 3.1): (1) a loading chamber to house a population of up to 250 worms, (2) a staging area to isolate a single worm from the population in the loading chamber and delivering it to the trapping area, (3) a trapping area to ensure repeatable and rapid immobilization of single worms near the focused laser beam, (4) 3D 18 interconnects to enable transition to completely sealing valves, and (5) flush exit outlets to transition processed worms to an off-chip location. In the following sections, the details of the serial microfluidic chip and unique characteristics of the microfluidic platform are presented. Then, the automation sequence including both automated actuation of valves and on-chip flow; as well as image processing methodology are discussed.
3.3 MICROFLUIDIC DEVICE DESIGN
A key design consideration when developing the automation program to perform femtosecond laser axotomies in C. elegans is the optimization of single worm delivery from a pre-loaded population into the trapping area while minimizing the degree of spatial variability in the trapped position of the worm. More specifically, the ability of trapping a single worm at the same location is essential to have a high success rate of the automation process. To reduce ambiguity in axonal reconnection data, other design considerations include overcoming delivery challenges such as trapping a worm in a folded configuration (Figure 3.2a), multiple worms in the trapping area (Figure 3.2b), and sending a non- axotomized worm into the pool of axotomized worms. With these challenges in mind, the serial microfluidic chip has three new components: 1) a staging area to isolate individual worms from preloaded population, 2) a T-shaped configured trapping area for fast and repeatable worm trapping, and 3) 3D interconnects to incorporate fully sealing valves.
19 Figure 3.2: Worm delivery challenges in the trapping area a) The undesired orientation of the worm inside the trapping area, which was overcome by actuating the trapping membrane, valve V3, in a cyclical manner during the initial trapping procedure while being pushed by the flow from the staging area. b) The microphotograph shows an example of multiple worms injected to the trapping area. Only one of the worms (left) was immobilized. The other one (right) was partially immobilized. Only tail part of the left worm was underneath the pressurized trapping membrane. Staging area: In the co-linear configuration of our previous design [12] , it was observed that transporting the worms at very high speeds to the trapping area, would occasionally result in multiple worms being trapped or some worms being lost when removing an axotomized worm. To isolate a single worm from a pre-loaded population of worms and avoid trapping of multiple worms, a staging area is placed before the trapping area. The new staging area incorporates two partially sealing gate valves (V1, V2) separated by a narrow channel that allows for the containment of only one Larval 4 (L4)
20 stage worm and prevents other worms in the main chamber from entering (Figure 3.3a). The main staging channel is 30 µm × 35 µm with a length of 600 µm, which is the approximate dimensions of an L4-stage worm [146]. It was previously shown by our group that a deflected PDMS membrane could almost completely seal a 30 µm-deep, 120 µm- wide rectangular channel, leaving small gaps in the bottom corners of the channel [12]. Accordingly, the staging channel beneath the gate valves (V1, V2) was designed to be 120 µm wide to prohibit staged or non-staged worms from passing through the actuated valves. T-shaped configured trapping area: The new T-shaped trapping area design enables the automated delivery and trapping of single worms in the desired location (Figure 3.3a). Automatic identification of the neuron of interest in the trapped worm requires processing of high-resolution images within a small field of view (FOV). It is therefore important to immobilize the worm in a location close to the desired FOV to find the neuron rapidly using a single step image acquisition. In the co-linear trap design, positioning the worm underneath the trapping membrane was extremely difficult, hindering the throughput of the axotomy assays [12]. To overcome this problem, our research group and other research groups incorporated complex manipulation methods for fine-positioning of the worms using, for example, side channels [12, 16]. However, such complex manipulations required manual operation interventions. The proposed T-shaped channel configuration of the trapping area eliminates the need for such complex manipulations by providing two major advantages. First, the structure provides a repeatable immobilization location by injecting the worm close to the middle of the sieve structure. Second, the channel shape allows for perpendicular decoupling of the injection and flushing channels, which permits a reliable and straightforward worm ejection after axotomy. A worm’s entire body is pushed against an array of short and narrow flow outlets that form the microfluidic sieve structure (Figure 3.3b). The pressure drop across these channels immediately straightens 21 the delivered worm into an elongated body position just before actuation of the trapping membrane. 3D interconnects: Microfluidic valves actuated in channels with rectangular cross- sections do not block the channel flow completely, even at very high valve pressures. While the staging area configuration took advantage of the leaky nature of these valves, such structures downstream from the trapping area resulted in inconsistent positioning of worms during immobilization. Worms delivered to the center of the trapping area showed less motility than the ones immobilized around the corners of the trapping area. The bending membrane of the trapping area restricts the worm’s whole-body better at the center of the trapping area. This step is critical for minimizing worm motility due to pharyngeal pumping and led to an improvement in ablation accuracy. This type of positioning error could be overcome with valves that completely sealed the exit flow channels. Valves actuated in channels with hemispherical cross-sections have been known to seal completely. However their fabrication would necessitate a thin layer of PDMS [118, 129] or a control layer between the worm and the cover glass in the trapping area [147]. The presence of this additional PDMS layer in the optical path would introduce an index of refraction mismatch and spherical aberrations at focusing distances beyond 30 – 40 µm, reducing the system’s effective NA and compromising the precision of both imaging and surgery. To avoid the additional PDMS layer, the fluid flow was diverted to another microfluidic flow layer and both flow layers were connected using new 3D interconnects. Interconnects were created on both sides of the trapping area by punching thru-holes at overlapping microfluidic structures within the first and second layers, as show in Figure 3.3c. To prevent media from flowing out of the device from the punched thru-hole, a permanent metallic plug (Figure 3.3c) was inserted into the thru-hole at a depth that it did not interfere with flow within the interconnect. Fabricating the fully sealing valves in a 22 different layer than the axotomy layer and connecting them with a permanent 3D interconnect, eliminated optical distortion and degradation of the image and ablation quality of the system. These three features simplified the image processing-based approach to identify anatomical features of the worm and speed up the process of automated analysis.
23
Figure 3.3: Flow direction in T-shaped staging/trapping area and 3D interconnects. (a) Optical image of a dye-filled microfluidic device with black arrows indicating the direction of fluid flow. Orange dye fills the control layer and the blue dye identifies the flow channels. The loading chamber holds pre-loaded worms before their serial transportation into the staging area. The synchronized actuation of valves V1 and V2 in the staging area allows loading of a single worm against the microfluidic sieve in the immobilization zone. A partially sealing trapping membrane, actuated by valve V3 in the control layer, then physically immobilizes the worm for laser axotomy. (b) Schematic cross-section referring to the sectioning arrows A-A' in (a) that shows the flow direction through the sieves during membrane deflection, location of the worm in the trapping area during delivery and after membrane deflection, and the relative heights of the microfluidic sieve and channel within the immobilization zone. The objective is placed underneath the glass interface. Avoiding any additional PDMS layers in the optical pathway eliminates optical distortion for imaging and ablation. (c) Top view and cross-sectional view (y- y') of a 3D interconnect that enables transitioning the flow from the 1st layer (controlled by partially sealing valves) into the 2nd layer controlled by completely sealing valves. A metallic pin (red) is plugged half-way into the through hole that connects the layers. The cross-sectional view of a partially sealing valve is given along x-x' while a completely sealing valve is shown along z-z'.
24 The final design consideration is related to the capture of un-wanted debris during device operation. Despite efforts to remove particulate matter from the worm suspension in M9 buffer prior to loading worms into the device, there was still an accumulation of microscopic particles within the chip during the automated process. To prevent the sieve structures from clogging in the staging and trapping areas, which could negatively affect chip performance and could eventually lead to total chip failure, an array of staggered filter structures with gaps ranging from 10 µm down to 5 µm at the entrance of each flow channel were built [148]. Additionally, an array of pillars spaced 30 µm apart were fabricated. The array of pillars allowed for the passage of worms into the loading chamber, but blocked a large fraction of debris from entering the trapping area and clogging the sieve structures in the staging and trapping areas.
3.4 PROGRESSION OF VALVE ACTUATION, FLOW, AND WORM MANIPULATION PROCESSING
The automation software, developed on a LabVIEW platform, controls the external solenoid valves maintaining flow into the device and the on-chip valves to synchronize the automation steps. Specifically, it stages worms for serial processing, traps them individually for laser surgery, and controls the remainder of flow throughout the device. While the worm is immobilized in the trapping area, femtosecond laser axotomies are carried out on immobilized worms by the custom-developed automation software (details of the automation program are given in the next section). For the entire duration of
25 automated operation, a constant head pressure of ~15 kPa is used to continually drive flow through the loading chamber and move worms into the staging area.
Figure 3.4: Automated progression of valve actuation and flow. (a) A schematic representation of the chip, showing the valves, channels, and chambers/areas indicated by their reference numbers. Valves V1 and V2 are used to stage and inject individual worms serially into the trapping area (4). Valve V3 serves as the immobilizing membrane. Valves V4 and V5 are used to open and close the ejection and flush channels, respectively. The ejection channel is used to deliver worms that have been successfully axotomized to a recovery area, and the flush channel is used to jettison worms that failed at any step of the automation process. Two-way flow through the side channels (3) in the staging area aid in serial staging and injection of worms into the trapping area. A two-way channel (5) in the trapping area (4) allows the flow through the microfluidic sieve structure to push the worm against the sieve structure during loading and away from the sieve structure during unloading of the worm along with the one-way side channel (6) leading to the trapping area. (b) Filling the loading chamber (2) with a population of worms by opening side channels (1) and closing all other valves and channels. (c) Staging of a single worm by opening valve V1, closing valve V2, and pushing the worm into the staging channel where flow is directed out through the side channels (3). (d) Preventing additional worms from entering the staging area by closing valve V1. (e) Injecting the worm into the trapping area by opening valves V2 and V3 and reversing the flow in the staging side channels (3). Valve V1 is also opened to prevent other worms from entering the staging area prematurely. (f) Immobilizing the worm by closing valve V3 in a cyclical pumping manner and then performing automated axotomy on the trapped worm using image- processing algorithms. (g) Transporting the successfully axotomized worm into the recovery area through the ejection channel (7). (h) Starting the next cycle by loading the next worm in the staging area. The sequence of valve and flow progression at each step during automation is shown in Figure 3.4. The numbering of valves and chambers/channels is provided in Figure 3.4a. First, a population of worms is loaded into the device by blocking all flow channels
26 except the small flow exits (1) in the side of the loading chamber (2) (Figure 3.4b). After the loading chamber has been filled with worms, the gate in the staging area is used to withdraw a single worm and inject it into the trapping area by opening and closing the two valves located at the front (V1) and back (V2) of the staging zone (see Figures. 3.4c –d). An array of small channels (3) on both sides of the staging channel acts as a sieve-like fluidic path to direct the worm between valves V1 and V2 during staging. Once a worm is located and isolated from the rest of the population in the staging area, it is injected into the trapping area (4) by reversing the flow through the staging sieve with a head pressure of ~65 kPa (back into the staging channel and opening valves V1, V2, and V3 for 1000–1200 ms (Figure 3.4e). This valve sequence pushes the worm against the sieve (5) in the trapping area, preparing it for trapping and surgery. Reversing flow in the staging sieve and opening valve V1 during this injection period also helps to prevent worms within the loading chamber from prematurely entering the staging area. During the trapping phase, valve V3 is actuated in a cyclical manner (close, open, and close) to avoid unfavorable worm folding (Figure 3.2) as the membrane begins to immobilize the worm. Immediately after immobilizing the loaded worm and recognizing its body centroid via image processing algorithms, the system moves the translation stage automatically to align the center of the worm body within the center of the FOV. After switching the objective from 5× to 63×, the automation program proceeds with positioning the laser beam on the axon of interest automatically (Figure 3.4f). Details of the image processing algorithm and axon ablation are described in the following sections. After performing the axotomy, the system switches back to the 5× objective (large FOV) and the software simultaneously opens valves V3 and either V4 (flush) or V5 (ejection) (Figure 3.4g). The automation software makes an intelligent decision as to whether or not the outcome of each individual automation step during trapping and 27 axotomy is successful. If the outcome is negative, the corresponding worm is flushed through the waste channel immediately after the unsuccessful step. Successfully axotomized worms are collected through the ejection channel. An external solenoid valve is simultaneously actuated to reverse flow through the sieve structure in the trapping area and to deliver fluid through the one-way-flow flush channel, both with a head pressure of ~135 kPa. After the valve sequence rapidly removes the worm from the trapping area and delivers it through either the flush or the ejection channel, the next cycle starts with staging the another worm from the pre-loaded population in the housing chamber (Figure 3.4h, just as in Figure 3.4c). This automated process is repeated until axotomies are performed on the desired number of worms from the loaded sample population, and any remaining worms are then removed from the chip by opening all on-chip valves and reversing the flow through the sieve structure in the trapping area. The entire process of staging, trapping, and axotomy requires, on average, about 17 s/worm.
3.5 IMAGE PROCESSING METHODOLOGY FOR AUTOMATED IDENTIFICATION OF NEURONS AND TARGETING FOR LASER AXOTOMY
Soft touch is sensed by six touch receptor neurons, which are the anterior lateral microtubule (ALM) pair, anterior ventral microtubule (AVM), the posterior lateral microtubule (PLM) pair, and posterior ventral microtubule (PVM),(see Figure 3.5) [149,
150]. Half of these neurons respond to gentle touch in the anterior part and the other half neurons are responsible for the gentle touch in the posterior half of the body. A pair of these neurons (the ALM pair) runs anteriorly along both sides of the worm with the somas located close to the worm’s centroid. The custom image processing methodology was developed to perform axotomies automatically on GFP-labeled mechanosensory ALM neurons.
28
Figure 3.5: The soft touch neurons of C. elegans: Left lateral view (a) and ventral view (b) of fluorescent images of an adult hermaphrodite C. elegans expressing green fluorescent protein (GFP) in six touch receptor neurons. The image is adapted from [151]. c) Top schematic shows approximate locations of six touch receptor neurons (TRN). The axes legend refers to ventral, posterior, anterior, and dorsal directions. The bottom shows the cross-sectional view of the worm body. The image is adapted from [103]. The automation software uses a four-step image processing procedure to automatically identify the neuron, locate the target on the desired axon, and perform laser axotomy for injury. The software is also able to detect and flush away worms that are not immobilized in a way that allows the image processing algorithms to pursue a successful axotomy. A flow chart describing the whole automation process, including image- processing steps, is given in Figure 3.6. It is important to note that Step 1 in the flow chart involves all the valve actuation steps to trap a single worm, including staging, injection and
29 trapping, in addition to the image processing step that are discussed in the following sections.
Figure 3.6: Automation Flowchart. Flow chart illustrates the steps of the automation process, including the image processing algorithms used to find and ablate the axon of interest.
30 3.5.1 Step 1: Identification of the worm location
Once a worm is immobilized, which can occur anywhere within the trapping area (Figure 3.7a), it is necessary to identify the relative location of the centroid of the trapped worm. The knowledge of the centroid location helps to verify the worm location immobilized under the deflected membrane/in the trapping area, and provides valuable information on the relative location of the neuron of interest (ALM for this work). The stereotyped neuroanatomy of C. elegans enables accurately positioning of the high- magnification, small FOV near the desired neuron for the subsequent automation steps, where fine focusing of the target axon and ablation occurs. By processing the low- magnification (5×), bright-field images of the trapping area, the software finds the relative location of the worm within the trapping area. The image processing involves background subtraction and thresholding. Before trapping a worm, the software saves a baseline image of the similar FOV without a worm but with the pressurized trapping membrane (Figure 3.7b). By subtracting the baseline image from the snapshot of the trapped worm, the background is removed and only the worm remains in the resultant image (Figure 3.7c). The program next extracts a pre-defined region of interest (ROI) from the resultant image, based on the perimeter of the trapping area. An image-thresholding filter is applied to enhance contrast (Figure 3.7d). To eliminate noise in the processed image, a particle filter removes all features spanning areas smaller than 300 pixels. If the detected center of the worm is found near the edge of the ROI, the trapped worm is flushed (rejected) and the
31 software proceeds with staging the next worm. This simple subtraction algorithm eliminates the need for complex and time-consuming pattern recognition algorithms.
Figure 3.7: Image processing methodology for identifying the worm location within the trapping area (Step 1). (a) Bright-field image of an immobilized worm. (b) Background image of the trapping area with the same FOV as shown in (a) with the trapping membrane fully actuated. The image in (b) is then subtracted from the image in (a), resulting in the image given in (c). (d) Image thresholding and noise filtering is then used to generate a high degree of contrast so that the centroid of the worm can be accurately located.
32 3.5.2 Step 2: Detection of a neuronal cell body in the small FOV
After the centroid of the trapped worm is defined and translated into the center of the FOV, the 63× objective lens moves into place and the illumination is switched to fluorescence mode. Here, finding the centroid of the worm in Step 1 ensures that there are only ALM and possibly AVM neurons present in the small FOV. Other anatomical features need to be integrated to perform axotomies on other neurons such as the PLM. To find a cell body in the small FOV, the coarse z-stage then begins to advance the focal plane in steps of 2.5 µm into the body of the worm (Figure 3.8a), collecting fluorescent images at each increment and looking for circular features. Each step moves the focal plane towards either the left or right ALM neuron, reaching to a neuron positioned closest to the cover glass-worm interface, and until a circular shape corresponding to the soma is detected. It is important to note that when perform surgeries are carried out on ALMs, no distinction is made between the left and right ALM (Figure 3.9a). To find a cell body, images collected at each z-step are thresholded to a pre-determined intensity cutoff (determined empirically to be 8× the mean intensity) until a circular feature drops to a radius between 2 and 6 µm (Figure 3.8a). If the program cannot detect a circular feature in more than 10 iterations, it flushes the trapped worm and proceeds to stage the next worm. This cell body-locating process provides a fast method to find the neuron soma and adequately brings it closer to the focal plane, avoiding the need for slower fine-focusing approaches with higher resolution capabilities that are unnecessary at this stage of the automation program.
33 Figure 3.8: The circular object detection methodology to detect a cell body in the field of view (Step 2). (a) Fluorescence image of the detected cell body in the 63× FOV. (b) Thresholded image of the fluorescent cell body in the 63× FOV. In the thresholded fluorescence images, the automation program looks for objects with circular shapes having diameters between 2 µm (red circle) to 6 µm (dashed circle). The white object shows the detected cell body that is a little larger than the lower limit of desired size of the cell body. (c) The fluorescence image of the cell body is relocated to the location of the laser spot, which is close to the center of the FOV to aid in further image processing.
34 3.5.3 Step 3: Verification of neuron of interest
The next step in the automation is the verification of the neuron of interest and the axon’s orientation relative to the anterior/posterior body axis of the worm. By fine-focusing on the cell body and looking for axonal processes, the program determines if the detected cell body is the neuron of interest and finds its orientation. For fine-focusing, a z-stack of images (0.7 µm steps) is collected via piezoelectric actuator translation (Figure 3.9). The image with the highest variance of pixel intensity correlates to the most in-focus z-position