Simultaneous calcium imaging of multiple neurons in Caenorhabditis elegans

(1)

Simultaneous calcium imaging of multiple

neurons in

Caenorhabditis elegans

R. Doornekamp Universiteit van Amsterdam

(2)

A system was developed for imaging of behavior and neural activity of multiple neurons simultaneously in freely moving Caenorhabditis elegans. Due to the low magnification, high numerical aperture objective (10X/0.45NA), combined with a high resolution sC-MOS camera we are able to get the entire worm in the field of view as well as resolve individual neurons spaced 2-3 microns apart. Neural activity is recorded optically using a ratiometric measurement between a calcium sensor, GCaMP3, and fluorescent protein DsRed/mCherry, expressed in several key command interneurons involved in locomotion. Custom-written image analysis software detects both the worm outline and its backbone from the fluorescence images and uses these to rectify the worm and subsequently identify and track individual neurons. The algorithm correctly identifies neurons with an error rate of 3% and is robust against temporal disappearance of parts of the worm. The in-strument and software were used to record activity in command interneuron AVA during forward/reverse transitions, characterise the neuron’s response to transitions and determine limitations of this imaging approach.

(4)

Chapter 1

Introduction

1.1 From neurons to behavior

A fundamental question in systems biology is how behavior arises from underlying neural physiology and neuronal dynamics. Efforts to map complex mammalian brains with over 1010 cells are being undertaken, yet there is still much unknown about how much simpler brains manage to produce complex behavioral repertoires. Caenorhabditis elegans, an approximately 1 mm small nematode worm, is one such example: despite having one of the smallest brains in the animal kingdom it shows a wide variety of behavior, including locomotion, feeding, avoidance, sleeping, mating and even learning. How exactly this behavior is generated and modulated in response to the environment by only a few hundred neurons remains largely a mystery, in spite of the fact that the connectome, the map of all neural connections in the brain, has been known for nearly three decades. Understanding how the nervous system computes thus not only requires information on the physical connections, but also functional information and activity patterns of neurons.

To acquire functional information, early research used selective ablation of neurons by laser microsurgery, gene manipulation and electrophysiology. Though these methods were useful to deduce basic characteristics of, and functional subunits in the brain circuitry, they are mostly restricted to characterising a single or a few neurons at once, and all of them severely perturbed the animal, making them fundamentally unfit for behavioral studies[14, 9, 7]. This has changed with the recent advent of genetically encoded voltage and calcium indicators (GEVI’s, GECI’s): fluorescent proteins that allow for in-vivo, spe-cific and chronic recording of membrane potential and calcium concentration - the latter being a proxy for membrane potential. Early experiments with GECI’s in C. elegans fo-cused mainly on establishing the correlation between activity in single sensory neurons and stimuli and head neuron activity and locomotion [43, 2, 13, 30, 28], providing a framework

(5)

for further study but also showing that understanding the link between activity and behav-ior would require more than single neuron recordings. With voltage and calcium sensors growing increasingly bright and their kinetics improving, focus is shifting towards a more integrated, systems approach in which large amounts of neurons are imaged simultaneously in tandem with behavior. This poses new problems for automatic identification, annotation and tracking of many neurons simultaneously in moving worms, as well as for managing, modelling and interpreting the large amounts of data that are potentially generated by this approach. This study develops a method to simultaneously image multiple neurons along the entire body in freely moving C. elegans and to assess feasibility and current limitations in this approach.

1.2 C. elegans as a model system

Caenorhabditis elegans provides a unique opportunity for studying the relation between neural activity and behavior quantitatively for a number of reasons. It has one of the smallest nervous systems in the animal kingdom, with only 302 neurons on 1000 body cells[40]. This is over an order 10 less than the medicinal leech, and many orders less than other model organisms (see figure 1.1). Besides having a small brain, it also has a small and optically transparent body, facilitating non-invasive imaging of neurons[29], and a relatively short life cycle of 2-3 days, making it ideal for propagation in the lab[46]. It nonetheless shows a wide array of behaviors, including feeding, mating, sleeping and learning [9], mostly expressed by different motility patterns, and these behaviors change on second to minute timescales[25], allowing the use of video microscopy. Furthermore, as it has been a model animal in biology for many years, the community resources available are extensive: its entire genome is sequenced, a wide array of off-the-shelf genetic tools is available and the physical connectivity in the form of chemical synapses and gap junctions has been almost fully mapped.[40, 19]

On agar substrate and in fluids C. elegans moves around by producing dorsoventral bending waves. Both bending frequency and body wavelength are modulated in response to the viscosity of the medium, the two extreme modes being referred to as crawling, in high viscosity media, and swimming, in low viscosity media. Its motility is the result of the propagation of sinusoidal bending waves along the body, either from head to tail when moving forward or from tail to head when moving backward and it can bias this move-ment to make turns [44]. Transitions between forward movemove-ment and backward movemove-ment, referred to as reversals, are either spontaneous or induced by (mechano-)sensory stimuli.

(6)

Figure 1.1: Scatter plot of number of neurons versus size in millimeter for a range of commonly studied animals in biology.

C. elegans will respond with a reversal to a surprisingly wide array of mechanical, chem-ical and other environmental stimuli[5] but there are also spontaneous reversals that are most likely the result of the integration of the internal state of the worm, environmental factors, and noise; they are thus interesting to study as they provide both an insight into how noise can randomize/affect behavior and a window into the internal state of the worm.

Nervous system anatomy

The C. elegans hermaphrodite has 302 neurons that, contrary to neurons in many other organisms, are non-myelinated and in general do not conduct action potentials. Instead they show a variety of different responses, among which graded and plateau potentials are the most prevalent[22, 21, 18] Locations of cells are stereotypical from animal to animal and synaptic locations are 75% reproducable from animal to animal[40, 19, 6] Neurons are divided into four functional classes based on their morphology, anatomical location and connectivity. These classes are sensory-, motor-, inter- and polymodal neurons. Whereas the sensory and motor neurons are obvious specialisations, interneurons, comprising the largest group, form a more generic class receiving and integrating input from and relaying output to other neurons. Most of the neurons are positioned between the body wall muscles and hypodermis and concentrated in head/tail ganglia (referred to as the nerve ring), with

(7)

Figure 1.2: TEM cross-section of C. elegans L4 WT (JSH_101143). Blue area is the region where the nerve ring is situated. Neuron somata are visible in this area and are identified as AVA,AVE,AVD,AVB (and others). This region is approximately 10-15 µm in diameter based on measurements from confocal z-stacks in EG4813 L4 larvae - young adults.

the exception of motor neurons, that generally lie along the dorsal/ventral midline (forming the dorsal/ventral nerve cord) where they synapse directly onto muscle cells.

Neurons are connected by a combination of chemical synapses and gap junctions. The wiring diagram of the C. elegans nervous system, commonly referred to as the connec-tome, consists of over 7000 connections and has been extensively mapped using electron microscopy. It is the first nearly complete map of a connectome and suffers only minor gaps[48]. Within this highly interconnected neural network, laser-ablation and optoge-netics studies have revealed neural subcircuits for specific functional modalities such as locomotion, egg-laying and defecation.[1, 42, 14]

Although it remains unresolved how the sinusoidal bending waves are generated, C. elegans’ motility is the result of a particularly simple neuromuscular system that produces waves of dorsal and ventral muscle contractions that are out of phase with eachother.[47] The innervation of muscles is directly controlled by three classes of motorneurons that can be divided into an excitatory (A- and B-type) and inhibitory (D-type) type. Early laser ablation studies[14] have established that the B-type motorneurons are important for for-ward and the A-type motorneurons for backfor-ward locomotion; although ablation of a single class never managed to completely abolish movement, suggesting at least functional redun-dancy or parallel pathways. More recent research suggests that directionality of movement is determined by an imbalanced output between these two classes of motorneurons and that switching between imbalanced states (A > B or B > A) corresponds to a change in this directionality [27, 50]. Several key components in the locomotion circuit that reflect

(8)

Figure 1.3: Tentative schematic of the locomotion subcircuit showing how command in-terneurons have a central task integrating input from sensory neurons and relaying this to motorneurons downstream. (Taken from: [31, 38]).

this change have been identified [38, 50] to be the command interneurons AVA, AVD, AVE, PVC and RIM. Electrophysiological recordings in AVA show evoked activity in response to nose touch , animals lacking AVA were unable to generate long reversals but maintained the ability to do omega turns and short reversals[42], optogenetic stimulation of either AVA or RIM induced backward locomotion[24, 39, 32] and RIM and AVA activity are correlated through gap junctions[27].

1.3 Techniques for neural imaging in C. elegans

There are many potential signals to study in neural physiology: membrane potentials, ions, neurotransmitters, neuromodulators and second messengers. Membrane potential is generally considered the most fundamental signal. However, as direct measurement of membrane potential is often difficult in-vivo, proxies such as [Ca2+] concentration, are employed to measure it. Calcium is an important signaling ion in C. elegans with several functions: it is involved in synaptic vesicle release, modulation of ion channel activity to generate current across membranes. More precisely, graded potentials across the membrane

(9)

lead to calcium transients in the cell through voltage-gated calcium channels; as the calcium is eventually buffered, extruded and pumped back, this rise is reversed on a timescale of order 100 ms[12]. Calcium is particulary suitable as a proxy for neural activity in C. elegans as the animal lacks the sodium-based currents prevalent in other organisms, and because the magnitude of activity-induced changes is much larger than with other ions[34]. In addition, calcium transients last significantly longer than the changes in membrane potential that trigger them, increasing the signal to noise ratio at the cost of temporal resolution[34]. While a comprehensive review of measurement techniques is beyond the scope of this thesis and has been published[36], it is important to briefly highlight each techniques’ strengths and weaknesses.

1. Patch-clamp electrophysiology (PCE): The most direct way of measuring changes in membrane potential. Adapted from existing patch-clamp techniques for the small ≈2 um neuron cell bodies of C. elegans that are contained within it’s pressurised cuticle. In PCE, the worm is immobilized by glueing it to a surface after which a small incision in the cuticle is made to reduce the hydrostatic pressure and suck a recording pipette onto the neuron membrane. [23]

• Requires immobilization and dissection, eliminating the possibility of studying natural behavior.

• Restricted to one or two neurons simultaneously.

• Can register currents in single-channels, muscles and synapses • High temporal resolution (< 1 ms)

2. Synthetic dyes (small-molecule Ca2+ dyes such as Oregon Green BAPTA-1).

• Requires immobilization and dissection • Photostable (>30 mins)

• Large dynamic range (Ca2+-dependent fluorescence change).

• Labelling is non-specific and without direct loading of cells the high back-ground/neuropil contamination restricts imaging

(10)

• Clearance and accumulation in organelles is problematic

3. Direct voltage imaging (Voltage-sensitive dyes and genetically-encoded voltage indicators (GEVI’s)).

• GEVI’s are specific and have fast kinetics (sub-millisecond), but dim.

• VSD’s are hard to load (patch pipette required) and toxic. Restricted to one or several neurons

• No current strains available with GEVIs[41, 34]

4. Genetically-encoded calcium indicators (GECI’s): fluorescent molecules that respond to the binding of calcium ions by changing their fluorescence.

• Non-invasive, non-toxic chronic imaging (stable expression for months)

• Can be integrated into the chromosome and targeted to specific cells or popu-lations of cells

• Don’t perturb the cell (e.g. buffering of Ca2+₎

• Relatively slow kinetics (GCaMP3: rise/decay time of 95 ± 27 ms/650 ± 230 ms).

• Exhibit linear relationship between changes in |Ca2+_{| and fluorescence change}

in the range of interest

Previous calcium imaging approaches, summarized in table 1.1, have relied on either tracking of single or at most a few neurons at low magnification (considered everything up to and including 20X) or whole sets of neurons at high magnification (>20X) in a setup where the worm was immobilized or restrained. As the field of view is small with high magnifications, only part of the worm is continuously visible, requiring the use of an extra camera to detect worm postures and limiting the number of neurons that can be simulta-neously recorded. Furthermore, the low depth of field makes these setups very sensitive to head swings in the direction of the optical axis, often requiring some form restraining of the worm. Low magnification avoids these issues, as the worm can be completely in the

(11)

pap er indicator restrain ed/free neurons magnification tec hnique Guo (2009) GCaMP immobilized ASH,RIM,A V A,A VD 63X Microscop e Ben Arous (2010) D3cp v cameleon free A V A,A VE,PLM 20X Mi c rosco p e+trac king stage Piggott(2011) GCaMP3/DsRed free A V A,A VE,A VD,RIM 20X+1.6X Microscop e+trac king stage Zheng (2012) GCaMP3/DsRed free A V A/RIM 20X+1.6X Microscop e+trac king stage Sc hrö del (2013) NLS-GCaMP5L restrained 70% of head neurons 40X WF-T eF O W en (2012) GCaMP3/RFP restrained DB6/VB9 20X Microscop e Ka w ano (2011) YFP/CFP free A V A/A VE/A VB/ A VD/VB8-9/V A6-8 16X/63X Microscop e+trac king Chronis (2007) GCaMP3 restrained A V A/A VE/RIM/ASH 32X Microscop e Prev edel (2014) NLS-GCaMP5K restrained up to 74 head neurons 40X Ligh t-field Microscop e Shipley (2014) GCaMP3/mCherry free A V A 10X Microscop e+trac king stage Larsc h (2013) GCaMP2.2/3/5 w orm arena A W A,A W C,ASH,A V A,AIA,AIY 2.5X/5X Microscop e + arena F aumon t (2011) Cameleon/TN-XL free A V A,A VB,AIA,D A4,V C3,VB6 63X Microscop e+trac king stage T able 1. 1: Ov erview of p ublications and calcium image setups in C. ele gans

(12)

field of view and the depth of field is much larger, but signal levels can be low and resolving neurons that are very close together can be problematic; especially in the nerve ring. With the recently improved brightness of GECI’s, such as GCaMP3-6[15] and improved sCMOS camera’s with very high resolution and framerates however, imaging at low magnifications has become feasible.

(13)

Chapter 2

Experimental setup

2.1 Worm chip

Several methods have been developed to restrain the worms for microscopy, among which microfluidics are an emerging technique that provide both spatial confinement and precise control over the environment[45]. For C. elegans, devices have commonly been constructed in agar, PDMS or more recently, polyacrylamide [51]. As restrainment can perturb natural behavior, a set of of different devices was fabricated using different materials, in which worms could move both restrained and unrestrained.

Microfabrication of PDMS/PA chip

Fabrication of master mold

The chip, consisting of a series of differently sized chambers, was designed in illustrator and a high resolution (25400 dpi) photomask was printed by SELBA S.A. Master molds were produced on a silicon wafer using a spin-coated layer of SU-8, a high contrast, epoxy-based resist (Micro Chem). Different viscosities (2050, 2100, 3025) and spin conditions were used to achieve the desired film thickness, ranging from 30-150 um. Prior to spincoating, the silicon substrate was cleaned with a piranha wet etch (H2SO4 & H2O2) followed by a demi-water rinse. The master mold was then made according to standard protocol for UV-lithography (see process flow for all steps and parameters involved).

(14)

Molding of the PDMS and acrylamide

The PDMS (Sylgard 184, Dow Corning) mold was produced according to the manufactur-ers’ protocol by mixing the resin and curing agent in a 10:1 ratio and pouring it onto the SU-8 Master. After pouring, the whole chip was put in a vacuum chamber (KNF Neuberger laboport, VWR international) for an hour, until no air bubbles remained in the solution. The PDMS was then placed in the oven at 65◦C for an hour to cure, after which the PDMS replica of the master was peeled off. Access to the chamber is provided by tubing that was inserted into holes, punched into the PDMS (Harris, Unicore 0.5 - 0.75mm). After punching the holes and cleaning of the replica, it was exposed to O₂ plasma and bonded to a glass surface.

For the acrylamide gel chambers, a cover-glass with a glass-machined cavity was glued to the silicon wafer using silica grease, and a 29:1 ratio of acrylamide/bis-acrylamide (Bio-Rad) with a concentration of 40% was poured into the cavity together with 0.1% APS to initiate polymerization. A silanized glass cover slip was put on top and clamped to the wafer to provide an airtight seal. The mixture was then left to polymerise for 2 hours. After polymerization, the polyacrylamide membrane was cut out and washed with dH2O to remove any excess unpolymerized acryl.

(15)

Figure 2.1: Schematic overview of PDMS and acrylamide chip production. An SU-8 layer is spincoated onto a silicon wafer (a) after which it is illuminated by UV light through a patterned photomask (b). The unpolymerized SU-8 is then washed off. Either PDMS (d1) or acrylamide (d2) is then poured onto the chip ensuring in the latter case that the acrylamide doesn’t contact oxygen during polymerization. After polymerization is complete the PDMS/acrylamide is removed from the chip, rinsed thoroughly in demiwater and attached to a cover slide (e1/2).

Fabrication of the agar chip

Thin 2-4% agarose layers are the laboratory condition under which most assays are con-ducted. For our purpose, 2% agarose (BactoAgar) in M9 buffer [46] was poured into a machined glass sample holder containing three copper rings (22 mm diameter) . The sides

(16)

are formed by 0.17mm (No. 1) thick 24x66 mm cover slides with a glass-machined cavity and the number was varied to create agar layers 0.5 - 1 mm thick. After pouring the liquid agar in the cavity, a glass slide was pressed on top slowly until a tight seal was obtained and all excess agar was pushed out. Slow application is crucial in this step to avoid air bubbles being caught in the agar. After solidification of the agar, the cover slide was carefully removed by pushing it sideways to produce a clean and very flat surface.

Figure 2.2: Side and top view of the agar chip (75 x 26 mm), consisting of a thin layer of 2% agarose in M9 buffer sandwiched between two cover slides. Thickness of the agar layer is determined by the number of 0.17mm cover slide spacers.

Results

Both microchambers, that restricted movement of the worm in the x-y plane, and flat slabs of PDMS, acrylamide and agar were produced and tested. The hydrophobicity of PDMS rendered it impossible to fill the chambers and pick worms into them by hand. It generally required bonding to glass to make a watertight seal and tubing to access the channel[45] and this in turn required the worms to be flushed in through tubes with a syringe. Unless the chamber height restricted the worm in the z-direction, it would swim and move its head around freely in all directions, including the direction of the optical axis. This was undesirable as any movement along this axis moves the neurons out of focus. As the velocity of the worms in liquids is also generally much larger, exposure times < 5 ms were required to prevent motion blurring. This resulted in GCaMP3 light levels being too low to detect.

Polyacrylamide gels were succesfully used to make chambers[51] and flat slabs; they had several advantages over PDMS: they were hydrophilic, thus eliminating the need for

(17)

tubing and could be used similarly to agar, while being more resilient to damage such as small dents and scratches that are often produced when depositing the worms. In addition worms crawl on them, lowering their velocity compared to PDMS and enabling longer exposure times. The elasticity of the material could be tuned over a reasonable range. Worms did not burrow into the acrylamide, however also didn’t show normal movement in the chambers as they were impaired by the thin water film that formed on the surface of the acrylamide; this water film would keep the worms attached to the chamber’ walls upon them touching it. In addition, without an airtight cover to prevent evaporation, the thin acrylamide slabs showed rapid drying out, causing them to curl up within 5-10 minutes.

Agarose pads, being the standard choice for behavioral studies, posed several advan-tages over the other methods: by controlling the percentage of agarose, the velocity of the worms could be tuned at the cost of background light (as scattering increased with agarose concentration). Furthermore, it was easily produced and didn’t require special handling. It also dried out over time, but not as fast as polyacrylamide. The only downside of agarose was its relative vulnerability compared to acrylamide (which can be re-used indefinately): air bubbles easily form during production and the substrate is prone to scratching; occas-sionally worms also tried to burrow into it causing their head to move out of focus. The flatness of and lack of defects in the agarose was critical for keeping the neurons in focus with high NA objectives. The copper rings allowed multiple worms on the same slide and the copper also forms a repellent gradient near the edge, required to keep C. elegans from moving off the surface of the slide.

2.2 Optics

The worm chip was placed under a Nikon TE2000 inverted microscope and illuminated with an X-Cite 120 fluorescent lamp through a set of dual-band excitation filters. The emission from GCaMP/DsRed was then filtered out using a dual-band dichroic, followed by a stage where the red and green components were split using a high-pass dichroic beamsplitter and green and red emission filters. A schematic of the setup is shown in figure 2.3.

(18)

Figure 2.3: Schematic diagram of the imaging system.

All samples were imaged with a Nikon 10X, numerical aperture (NA) 0.45, Plan Apo objective. This choice was motivated by the requirements that (1) the field of view should permit imaging the entire worm (0.6 - 1 mm in length) at least in one direction, (2) the depth of field should be large enough so that neurons on opposing ends of the nerve ring are still simultaneously in focus (≈10 um), (3) closely spaced neurons (1-2 um) should still be resolvable and (4) signal levels should be much higher than background. The depth of field (DOF) and resolution (r) in microscopy can be modelled, according to [26], as

r = 0.61λ NA DOF = λn NA2 + n M · NAe

where n is the index of refraction (n = 1.0 for air), λ the wavelength of interest (514 nm, the GCaMP emission peak), M the magnification and e the sCMOS resolution (6.5µm). As figure 2.5 shows, the Nikon 10X/0.45NA, giving a field of view of 1.4 × 0.7 mm, just meets all requirements; although a lower NA objective would be prefered with regard to the depth of field. However, since low NA also negatively impacts the signal to noise ratio, it was a trade-off between high SNR and being able to have all neurons of interest in the

(19)

focus. The high NA does potentially make the setup very sensitive to movement of the worm along the optical axis.

As GCaMP3 signal levels are low, the filters were specifically chosen to maximize the amount of GCaMP3 emission light getting to the sCMOS camera by doing an extensive search among all Semrock and Chroma filters and finding the combination performing best on these criteria:

1. it maximized the quantity ˆ

DBD(λ)(1 − DBS(λ))EM G(λ)GCAM P (λ)dλ

where DBD,DBS,EMG stand for the transmission spectrum of respectively the dual-band dichroic mirror, the single dual-band dichroic and the green emission filter and GCAM P (λ) is the GCaMP emission spectrum.

2. it minimized cross-talk between GCaMP/DsRed (defined as the amount of DsRed emission light arriving in the green channel)

3. prevent blue-light exposure of the worm as much as possible, as this perturbs non-lite-1 mutants [49]

Figure 2.4: Depth of field (DOF), Field of view and lateral resolution as a function of objective magnification and NA. Shaded areas are the minimum requirements for resolving neurons, having the full worm in the field of view, and having a sufficient depth of field for the entire nerve ring.

The selected set (dual-band excitation/dichroic: Semrock FF01-482/563-25,Di01-R488/561-25x36) was mounted in the microscope. Filter spectra and GCaMP3/DsRed

(20)

emission/ex-citation spectra are shown in figure 2.5. The fluorescence image was split by a beamsplitter system (Cairns Research Optosplit LS) that was fitted with the dichroic (Semrock FF560-FDi01-25x36) and red (Semrock FF01-617/73-25) and green (Semrock FF01-525/45-25) emission filters. Neutral density filters were placed in the light path, after the red emission filter, to prevent saturation of the red channel, as DsRed and mCherry were commonly much brighter than GCaMP3. The two channels were then projected onto a PCO Edge 5.5 sCMOS camera (2160x2180 pixels, 6.5µm pixel size) and all images were captured using a PCO plugin in µManager and custom-written C routines to deal with the high data output rates at this resolution (≈150 MB/s at 10 fps). Occasionally manual refocussing due to unevenness of the agar slab was required (on average 2-3 times per recording).

Figure 2.5: Spectra of filters and emission/excitation spectra of both GCaMP3 and DsRed. Filters were selected to minimize crosstalk between red and green and maximize the amount of GCaMP3 emission light getting to the sCMOS chip .

(21)

Chapter 3

Calcium imaging of neurons

3.1 Strains and nematode handling

Several strains from the literature[32, 38] (see table 2) were selected and obtained from the CGC for imaging, all of them expressing either a single or several neurons with key roles in the locomotory subcircuit. Worms were cultivated at 20◦C on NGM-SR plates (3g NaCl, 24 g agar, 2.5g peptone, 1 ml 5 mg mL cholesterol in EtOH in 975 ml water, with 1 ml 1M CaCl2, 1 ml 1M MgSO4, 25 ml K2PO4, pH6)[25]. Strains with extrachromosomal arrays were screened every two days and individuals with high expression levels were transferred to a new plate each time. Before the experiment, worms were cultivated on NGM-SR plates for two days and both L4 larvae and young adults were individually transferred by worm pick to a drop of M9 buffer; purging the worms’ gut to reduce autofluorescence from gut bacteria, as well as removing any bacteria sticking to its cuticle. Selection of worms was performed based on expression level of GCaMP3 and the worms were transferred to an agar slab minutes prior to any recording. The wormchip was then covered with a raised cover glass to prevent dehydration.

Strain name Strain details Expression

TQ3032 lite-1(xu7) X; xuEx1040 [nmr-1p::GCaMP3.0::DsRed] AV(A/D/E/G), PVC, RIM,DVA TQ2183 lite-1(xu7) X; xuEx705. [npr-9p::GCaMP3.0::DsRed2B] AIB

QW625 zfis42 [Prig-3::GCaMP3::SL2::mCherry] AVA, I1, I4, M4, NSM QW1075 zfEx416 [Prig-3::GFP::SL2::mCherry] AVA, I1, I4, M4, NSM

Table 3.1: Overview of all strains (TQ3032/TQ2183 provided by the CGC. QW625/1075 provided by Frederick B. Shipley).

(22)

Figure 3.1: A) Brightfield and fluorescent image of TQ3032 head with neurons identified B,C) Schematic diagram of the neurons around the pharyngeal bulbs (taken from: Brockie et al. 2001b, [10]) and tail (taken from: Brockie et al., 2001 [11]). D. Dorsal view of the tail of TQ3032 with PVCR/PVCL and DVA neurons visible.

3.2 Image processing

To record neural timeseries in recordings of ≈20 minutes, at 10 fps (12000 frames per recording) a robust system to automatically track and identify neurons from frame to frame is required. However, automated tracking and identification of multiple neurons simultaneously in C. elegans is made difficult by several facts:

1. neural somata are hard to distinguish because they have the same size and shape 2. neurons often overlap partially or are located closely together in the nerve ring and 3. cells occasionally move out of the focal plane because the worm lifts or rotates.

(23)

To resolve these issues a method for identification and tracking in worm coordinates as opposed to camera coordinates was chosen: this is a natural choice as neurons are stereo-typical from animal to animal with high reproducibility and the worm shape can easily be extracted from image data or a separate camera. First, captured fluorescence images were filtered to remove speckle using a median filter with a 3x3 window and then thresholded manually. The resulting image was cleaned by applying a sequence of morphological op-erations and hole-filling to get a smooth worm shape. The worm was then skeletonized using a length-preserving customized skeletonization algorithm (appendix A) and a spline was fit to this skeleton. After parametrisation, the worm body was rectified [37, 35] by applying a series of rigid rotations along the backbone (see figure 3.2).

Figure 3.2: Image processing steps. The raw image is filtered and thresholded (A) and a backbone is fit to the worm shape (B). A series rotations of planes perpendicular to the backbone (C) is then used to rectify the worm (D). The neurons are identified in the rectified worm and mapped to coordinates in the original image, from which the neural time series is extracted (E).

(24)

The rectified worm was then aligned by cross-correlation with an annotated reference worm, manually produced from the first image in the stack. This step is crucial as both at the head and tail, the thresholding and subsequent skeletonization steps often produce errors due to the fact that these parts of the worm body are relatively transparent and hard to distinguish from the background. In the rectified, aligned worm, neurons were identified and mapped to coordinates in the original image. This was performed for each neuron/re-gion of interest and a 100x100 pixel window centered on each neuron was extracted from the raw video data.

Extraction of neural time series

Raw fluorescence intensity was subsequently recorded as the weighted sum of a circular region of interest (ROI) centered on the maximum intensity value in the neuron [31]. The weighting was chosen to minimize the noise in the red channel, caused by rig-3::dsred ex-pressing cells near the target neuron entering the ROI. As local changes in illumination and light scattering from the agar affected the average value as well, background correction was performed by subtracting a median-filtered image where the window size was larger than the worm diameter. To correct for artifactual changes in the signal resulting from motion blur and movement in and out of the focal plane, the ratio (R) between the background corrected green and red channel was used as a neural signal.

R = P

x,yWG(x, y)(IG(x, y) − Ibackground,G(x, y))

P

x,yWR(x, y)(IG(x, y) − Ibackground,G(x, y))

P x,yWR(x, y) P x,yWG(x, y) where WG(x, y) = ( 1 x2+ y2≤ r2 ROI 0 otherwise WR(x, y) =    q ₃ πr2 ROI exp 9(x2_+y2₎ r2 ROI x2+ y2≤ r2 ROI 0 otherwise

The radius of the ROI was between 20-30 pixels and chosen to minimize nearby cells that expressed rig-3 entering the ROI. All images were exposed for 50 ms to avoid motion blur, after which a 50 ms camera delay was employed to fix the frame rate at 10 fps. The duration of the recordings was at most 20 minutes; at this point both photo-bleaching of GCaMP3 and accumulation of phototoxic damage, visible as increased gut auto-fluorescence[17],

(25)

would make extracting a signal impossible, as both neuron tracking became difficult and signal to noise ratio dropped significantly.

To account for photobleaching, all raw green and red data were corrected by detrending the signals, fitting an exponential decay function to the filtered data and then dividing the original signal by this function. Reversals were registered by eye, based on the crawling direction of the worm with an uncertainty of 2 frames (~100 ms), and each frame was annotated as being either forward, reverse or invalid. Omega turns and pirouettes were not registered.

Figure 3.3: A) Time series of AVA neural ratio (∆R/R) in the QW1075 (control) and QW625 strain. The background and trace colors indicate the direction of movement (re-verse: red, forward: blue). B) C. elegans moving in a polyacrylamide chamber C) The x-y trajectory in the lab frame of the AVA neuron in QW625 for 10 minutes. The colors of the trajectory indicate forward (blue) or reverse (red) motion. Arrows indicate start- (S) and endpoint. Discontinuities in the trajectory occur where AVA moves out of view due to tracking a tracking error, these frames are removed from the neural trace.

(26)

Results

30 worms (10 TQ3032, 15 QW625, 5 QW1075) were imaged for a total of 10 hours (3.4 TB of raw data). Strains expressing calcium sensor GCaMP3 and fluorescent protein mCherry or DsRed in command interneuron AVA were used, as well as a control strain that ex-pressed GFP instead of GCaMP3. For each worm, the automated tracking algorithm for the neurons was verified by hand and a behavioral time series was added. The algorithm incorrectly identified neurons in less than 3% of all frames; although performance is gener-ally better because this number includes frames in which the neuron was out of focus or not in view at all. In QW625 and QW1075, both non-lite-1 mutants so sensitive to blue light, behavior was visibly affected by the fluorescent light source, causing a flight response.

(27)

Figure 3.4: Neural time series of two QW625 worms, lasting 10 minutes (at 10 fps). Shaded areas are when the worm is moving in reverse. Large calcium transients are observed during the reversal. Arrows indicate large calcium transients that did not coincide with a reversal. Forward and reverse run durations (see figure 3.5) were compared with earlier reports by producing the cumulative distribution of forward and reverse run durations and fitting an exponential decay function. Decay constants were similar to earlier reports for most worms [25], although a small group showed much slower dynamics, owing to the fact they were at rest (not moving). Contrary to previous experiments, no calcium transients were observed in any neurons in the TQ3032 strain, however they were observed in the AVA neuron of QW625 during the onset of reversals in all worms, in agreement with previous

(28)

reports [8, 32, 31, 38]. In In addition, a few large transients per recording that did not coincide with reversals also occured. The AVA time series in QW625 shows well-separated signal intensities for the two distinct behavioral states (figure 3.6). The control strain QW1075 showed no clear calcium transients, however, GFP in bright rig-3 expressing cells surrounding and occassionally overlapping the target neuron did enter the ROI regularly, affecting the time series.

Figure 3.5: (Left) Distribution of fit parameters for reverse (red) and forward (blue) run durations. (Right) The cumulative distributions of reverse and forward run durations for three QW625 individuals.

To characterise the AVA activity in response to change in forward/reverse state, a per-worm average reversal response was produced by aligning all reversals and averaging them. More specifically, denoting the neural signal in reversal i by s_i(t), the behavioral signal bi(t) (1 when reversing, 0 when moving forward), the average ¯s(j) was computed as

¯ s(j) = P isi(j) P ibi(j)

The average response at a reversal for both the forward-reverse as well as reverse-forward transition for all worms is shown in figure 3.6 and shows a nearly linear increase of fluores-cence during the reversal as well as a nearly linear and much slower decrease in fluoresfluores-cence when the forward direction is restored.

(29)

Figure 3.6: (Top) AVA neural activity state distribution for 4 QW625 worms shows a separation between signal intensity in the forward (green) and reverse (red) state. (Bottom) Average response after a transition (left: reverse to forward, right: forward to reverse) of QW625 shows a nearly linear decrease/increase in GCaMP3 fluorescence following the reversal. Black line is the population mean, gray lines are the average response of individual worms. Due to the low number of measurements at longer timescales, sampling noise increases with time. Restoration of AVA activity to baseline, after a reverse to forward transition, is on a much larger timescale.

(30)

Discussion

A system was developed for imaging of behavior and neuronal activity of multiple neurons simultaneously in freely moving C. elegans. The setup combines a low magnification, high numerical aperture objective with a high resolution camera to get both a large depth of field, yet be able to get neurons in the entire body in view. Spatial resolution was sufficient to distinguish neurons spaced ≈ 1 − 2 µm apart, but the large extent of the somata in the nerve ring already posed a problem in separating cells (see figure 3.7); a problem that has been solved in other studies by using nuclear localization (NLS) tags, limiting the expression of GCaMP to cell nuclei. This is particularly crucial for large-scale imaging of neurons in the head, as the neurons are generally closely packed and have overlapping somata. However, this does further decrease light levels, demanding use of calcium indicators with a even larger brightness and dynamic range than GCaMP3.

Figure 3.7: Maximum z-projection of confocal stack of TQ3032 fluorescence images. Even though cell centers of AVA and AVE are not in the same focal plane and their cell cen-ters are separated by 5µm , their somata are largely overlapping making separation and identification difficult.

(31)

In order to obtain a sufficiently high signal from the neurons, both high light intensi-ties and a relatively low framerate were required. These high light levels, as a by-product, induced phototoxic damage and photo-bleaching in the worms. This limited recording duration to the amount of damage the worms could sustain: after 15 minutes the accu-mulated phototoxic damage resulted in increased background fluorescence that prevented the system from distinguishing neurons, rendering further tracking and recording of neu-rons difficult. Previous research [30] has suggested that this can be improved using pulsed excitation light instead of a continuous illumination.

Ratiometric measurement, originally introduced in neural imaging to compensate for artifactual changes caused by e.g. movement or changes in background illumination[29], posed several problems in the current setup for two reasons: first, there was an unidentified source of chromatic aberration in the optosplit LS (see appendix B) causing the red and green focal points to be separated by (≈ 13 um in the object plane) at the sCMOS chip. Even though this shift was reduced with a corrector lens (f=4000mm) in front of the red emission filter, it was not completely resolved. Second, the rig-3 promotor used caused mCherry to be expressed in a significant number of cells very close to the target neuron (see figure 3.8). These cells would enter and leave the ROI during body waves, adding significant noise to the red channel. A similar problem also occurred in the QW1075 strain with GFP. Ideally, promotors should be used that are specific to neurons only to avoid this problem.

(32)

Figure 3.8: Expression in cells with the rig-3 promotor. mCherry is seen expressed a large cluster of cells in the nerve ring, although most of these cells are not actually neurons. GCaMP3 is only visible in the AVA(R/L) neuron and some gut cells as these are the only cells with a significant [Ca2+].

The automated tracking and identification of neurons was able to succesfully identify the correct neuron in most frames with an error rate of 3% (3 bad frames in 100) using the straightening method; unlike in other particle tracking methods the misidentification or temporal disappearance of the neuron posed no problem for identification in subsequent frames: as long as the worm outline was detected properly. As the worm was slightly larger than one axis of the field of view, occassionally part of the worm body would be out of view, causing problems with the tracking. As these frames were easy to detect because of the sudden drop in worm surface area, they were marked and corrected for manually. Similar problems would occur when light scattering made the background indistinguishable from the worm body and when the worm coiled up (its head touching its body), resulting in misidentification of the proper worm shape. This can be easily avoided by relying on a second, low-resolution camera with IR filter to record the worm shape. Even though parallelization of the worm rectification algorithm (see appendix A) is trivial, the method is computationally very demanding due to the high resolution of the images (2180x2160

(33)

pixels per frame) and took ≈ 0.4 s per frame, resulting in a processing time of about 7 hours per 15 minute of video on a single computer.

Although we observed a correlation between the behavioral state (forward/reverse) and the AVA activity in QW625, there was no observable activity in TQ3032, although this is likely attributed to contamination of the original strain. There were also significant calcium transients in QW625 that were unaccounted for by changes in directionality. Due to the temporal resolution and uncertainty on the precise state around an event, it was not possible to detect whether a change in the neural signal preceded the reversal or not. This was aggravated by the fact that the manual tracking of the worm made it both impossible to extract the velocity of the worm and keep the worm in the center of view, as it was often moving too unpredictable to track for long periods by hand.

Transition amplitude was negatively correlated (Pearson’s r = −0.35) and reversal du-ration positively correlated (r = 0.4) to its time of occurrence in the video, suggesting reversal duration is increasing throughout the recording and transition amplitude decreas-ing. It is visibly apparent from the video, although not possible to quantify due to the lack of automatic tracking, that worm movement is significantly slowing down over time. This may be caused by the agar drying out and becoming more viscous in the course of the recording, or by the gradual habituation of the worm to the excitation light (it should be noted that non-lite-1 mutants will show a negative phototactic response to blue light and will attempt to move out of the excitation bundle [33]). The transition amplitude is possibly decreasing due to the gradual photobleaching of GCaMP3 or the decrease is also related to the lower velocity, as AVA amplitude has been shown to be correlated to worm velocity [32].

(34)

Appendix A

The straightening algorithm consists of 3 steps: filtering the image to produce a clean worm shape, length-preserving skeletonization and finally rectification.

Filtering

Filtering requires some manual input, most notably the threshold and a radius for the structuring element used to morphologically open and close the image. These need to be tuned per recording to give the best result.

% A p p l y m e d i a n f i l t e r

i m a g e = m e d f i l t 2 ( image , [5 5 ] ) ;

% T h r e s h o l d i m a g e and pad it for f u t u r e r o t a t i o n s w o r m . c o l o r _ w o r m = p a d a r r a y ( image , [30 3 0 ] ) ; w o r m . b w _ w o r m = i m a g e > t h r e s h o l d ; % M o r p h o l o g i c a l l y o p e n and c l o s e the i m a g e w i t h a d i s k % t h a t is c l o s e to the r a d i u s of the w o r m w o r m . b w _ w o r m = i m c l o s e ( w o r m . bw_worm , s t r e l ( ’ d i s k ’ , w o r m _ r a d i u s )); w o r m . b w _ w o r m = i m o p e n ( w o r m . bw_worm , s t r e l ( ’ d i s k ’ , w o r m _ r a d i u s )); Length-preserving skeletonization

This is an add-on to MATLAB’s proprietary code for skeletonization. The original code has methods for despurring a skeleton, but without preserving the full length. This code will despur the skeleton and produce a single non-interrupted line between those two endpoints of the skeleton that produce the longest backbone. It works by detecting and storing all endpoints in the skeleton, then applying repeated despurring operations. When the number of endpoints has reduced to two it will remove all previously recorded spurs and restore the path that connects the last remaining two endpoints to their origin. The result is a good approximation to the original backbone.

f u n c t i o n o u t p u t = n e w _ s k e l e t o n i z e 2 ( o u t p u t ) t r a c k b a c k = {}; % Use built - in s k e l e t o n i z a t i o n w o r m _ s k e l = b w m o r p h ( o u t p u t . bw_worm , ’ s k e l ’ , Inf ); % F i x e s s i m p l e s p u r i s s u e s in w h i c h an end is b r a n c h e d w o r m _ s k e l = w o r m _ s k e l | b w m o r p h ( w o r m _ s k e l , ’ e n d p o i n t s ’ ); % F i n d all c o o r d i n a t e s of e n d p o i n t s e n d p t s = f i n d ( b w m o r p h ( w o r m _ s k e l , ’ e n d p o i n t s ’ ));

(35)

% S t o r e e n d s and r e m o v e t h e m f r o m w o r m for i =1: l e n g t h ( e n d p t s ) t r a c k b a c k { i } = [ e n d p t s ( i )]; w o r m _ s k e l ( e n d p t s ( i )) = 0; end % r e p e a t t h i s u n t i l 2 e n d p o i n t s l e f t n r _ e n d p t s = l e n g t h ( e n d p t s ); s p u r l i s t = [ ] ; w h i l e ( n r _ e n d p t s > 2) % f i n d all e n d p o i n t s e n d p t s = f i n d ( b w m o r p h ( w o r m _ s k e l , ’ e n d p o i n t s ’ )); n r _ e n d p t s = l e n g t h ( e n d p t s ); for j =1: l e n g t h ( t r a c k b a c k ) h a s _ a d j a c e n t =0; for i =1: l e n g t h ( e n d p t s ) if ( e n d p t s ( i ) ~ = 0 & l e n g t h ( f i n d ( s p u r l i s t == j ) ) ~ = 1 ) if ( i s _ _ f i r s t _ a d j a c e n t _ t o _ s e c o n d ( s i z e ( w o r m _ s k e l ) ,... e n d p t s ( i ) , t r a c k b a c k { j }( end ))) t r a c k b a c k { j }( end +1) = e n d p t s ( i ); w o r m _ s k e l ( e n d p t s ( i )) = 0; e n d p t s ( i ) = 0; h a s _ a d j a c e n t =1; end end end if ( h a s _ a d j a c e n t ==0 & l e n g t h ( f i n d ( s p u r l i s t == j ) ) = = 0 ) s p u r l i s t ( end + 1 ) = j ; end end end g o o d l i s t = s e t d i f f ( [ 1 : l e n g t h ( t r a c k b a c k )] , s p u r l i s t ); w o r m _ s k e l ( t r a c k b a c k { g o o d l i s t ( 1 ) } ) = 1 ; w o r m _ s k e l ( t r a c k b a c k { g o o d l i s t ( 2 ) } ) = 1 ; o u t p u t . b w _ s k e l e t o n = b w m o r p h ( w o r m _ s k e l , ’ t h i n ’ ); o u t p u t . s k e l e t o n = o r d e r _ s k e l e t o n ( o u t p u t . b w _ s k e l e t o n ); end

The auxiliary function is_first_adjacent_to_second detects whether two points in the skeleton of the worm are adjacent.

f u n c t i o n o u t p u t = i s _ _ f i r s t _ a d j a c e n t _ t o _ s e c o n d ( A , lindex1 , l i n d e x 2 ) [ r1 c1 ] = i n d 2 s u b ( A , l i n d e x 1 );

[ r2 c2 ] = i n d 2 s u b ( A , l i n d e x 2 );

o u t p u t = ( r1 >= r2 - 1) & ( r1 <= r2 +1) & ( c1 >= c2 - 1) & ( c1 <= c2 + 1); end

Rectification

Rectification consists of two steps: fitting a smooth spline to the backbone generated by the skeletonization algorithm and then applying a sequence of rigid rotations along this backbone.

b a c k b o n e = w o r m . s k e l e t o n ; n u m _ b o d y p o i n t s = 50;

x = b a c k b o n e (: ,1); x = r e s h a p e ( x ,1 , l e n g t h ( x )); y = b a c k b o n e (: ,2); y = r e s h a p e ( y ,1 , l e n g t h ( y ));

(36)

p o i n t s = l e n g t h ( b a c k b o n e ); s = z e r o s (1 , p o i n t s ); for i =2: p o i n t s s ( i )= s ( i - 1 ) + s q r t (( x ( i ) - x ( i - 1 ) ) ^ 2 + ( y ( i ) - y ( i - 1 ) ) ^ 2 ) ; end xy = [ x ; y ]; w o r m L e n g t h = s ( end ); c u r v e _ s p l i n e = s p l i n e ( s , xy ); % now we h a v e a s p l i n e fit dS = max ( s )/( n u m _ b o d y p o i n t s + 1 ) ; % s t e p s i z e is f i x e d ... b e c a u s e t h i s is p i x e l b a s e d .. s t e p s i z e = 2 0 ; % WAS 20 e n d _ l = f l o o r ( max ( s )/ s t e p s i z e )* s t e p s i z e ; s c a l e =0: s t e p s i z e : e n d _ l ; s c a l e ( end + 1 ) = max ( s ); s p l i n e _ p o s i t i o n = p p v a l ( c u r v e _ s p l i n e , s c a l e ); y _ s p l i n e = s p l i n e _ p o s i t i o n (1 ,:); x _ s p l i n e = s p l i n e _ p o s i t i o n (2 ,:); % f i n d a n g l e of e a c h s e g m e n t ( 1 0 0 p o i n t s is 99 s e g m e n t s ) a n g l e = []; for seg = 1 : ( l e n g t h ( s p l i n e _ p o s i t i o n ) -1) dy = y _ s p l i n e ( seg +1) - y _ s p l i n e ( seg ); dx = x _ s p l i n e ( seg +1) - x _ s p l i n e ( seg ); a n g l e ( seg ) = a t a n ( dy / dx ); if ( seg >1) da = ( a n g l e ( seg ) - a n g l e ( seg - 1 ) ) * ( 1 8 0 / pi ) if ( da > 40) a n g l e ( seg ) = a n g l e ( seg ) - pi ; e l s e i f ( da < -40) a n g l e ( seg ) = a n g l e ( seg ) + pi ; end end end v e r t i c a l s = []; n r _ s e g m e n t s = l e n g t h ( y _ s p l i n e ) -1; for j =1: n r _ s e g m e n t s % n r _ s e g m e n t s Y = y _ s p l i n e ( j +1) - y _ s p l i n e ( j ); X = x _ s p l i n e ( j +1) - x _ s p l i n e ( j ); s t e p s _ p e r _ s e g m e n t = 20; dx = X / s t e p s _ p e r _ s e g m e n t ; dy = Y / s t e p s _ p e r _ s e g m e n t ; for k =1: s t e p s _ p e r _ s e g m e n t if k <= s t e p s _ p e r _ s e g m e n t /2 t h e t a =(( a n g l e ( j ) - a n g l e ( max ( j - 1 , 1 ) ) ) / s t e p s _ p e r _ s e g m e n t ) * . . . ( k + s t e p s _ p e r _ s e g m e n t /2) + a n g l e ( max ( j -1 ,1)); e l s e t h e t a = (( a n g l e ( min ( j +1 , n r _ s e g m e n t s )) - a n g l e ( j ))/ s t e p s _ p e r _ s e g m e n t ) * . . . ( k - s t e p s _ p e r _ s e g m e n t /2) + a n g l e ( j ); end x = x _ s p l i n e ( j )+ dx * k ; y = y _ s p l i n e ( j )+ dy * k ; c o r r e c t e d _ i m g = i m r o t a t e (( w o r m . f u l l _ w o r m ( y - 3 0 : y +30 , x - 3 0 : x +30)) , t h e t a * 1 8 0 / pi , ’ c r o p ’ );

(37)

v e r t i c a l s (1:61 ,( j - 1 ) * s t e p s _ p e r _ s e g m e n t + k ) = c o r r e c t e d _ i m g (: ,31) ’; end

(38)

Appendix B

There was significant chromatic aberration in the setup causing neurons at different focal planes in the red and green channel to be imaged at the time (see figure 3.9A). As all optical elements in the setup should be corrected for chromatic aberrations, to determine the element introducing the aberration, 3D point spread functions (see figure 3.9B) of 1 µm fluorescent beads were recorded by doing a z-scan over a large range for different objectives (4X, 10X Plan apo and 20X), both with and without the OptoSplit LS attached. This clearly showed the aberration occuring somewhere within the OptoSplit LS. The distance between the peaks in the PSF gives an indication of the distance along the optical axis between the green and red image planes at the sCMOS.

(39)

Figure 3.9: A) Chromatic aberration causes different object planes to be in focus simulta-neously at the sCMOS chip, this is most visible at the head neurons in TQ3032 in image. B) 3D point spread functions for fluorescent beads in the green and red channel, the peak intensities are visibly shifted with respect to eachother.

(40)

Figure 3.10: Intensity plot of the 3D PSF of a fluorescent bead in both red and green image paths along the center line (indicated in figure 3.9B with white arrows). Left image shows the intensity plot for the beads with the OptoSplit LS removed from the setup, right image with the OptoSplit LS in place: there is a clear shift (≈ 13 µm in the object plane) between the peak intensities.

To correct for the aberration a corrector lens can be placed in the OptoSplit after the emission filter. As the standard set of corrector lenses (f=1010mm, f=1346mm, f=2019mm) only made the shift opposite but worse, a custom (f=4000mm) lens was ordered. This reduced the shift to ≈ 3 µm in the object plane: roughly of the order of neural somata. Though this provided a workable solution for imaging neural somata, further improvement should be possible and is required when imaging elements that are much smaller than 3 µm (e.g. cell nuclei).

(41)

Bibliography

[1] Avery, L., and Thomas, J.H. (1997). Feeding and Defecation, In C. elegans II, D.L. Riddle, T. Blumenthal, B.J. Meyer, and J.R. Priess, eds. (Cold Spring Harbor Labo-ratory Press), pp. 679–716.

[2] Chronis, N., Zimmer, M., and Bargmann,

[3] C.I. (2007). Microfluidics for in vivo imaging of neuronal and behavioral activity in Caenorhabditis elegans. Nat. Methods 4, 727-731.

[4] Albrecht, D.R., and Bargmann, C.I. (2011). High-content behavioral analysis of Caenorhabditis elegans in precise spatiotemporal chemical environments. Nat. Meth-ods 8, 599-605.

[5] Altun, Z.F. and Hall, D.H. 2011. Nervous system, general description. In WormAtlas. doi:10.3908/wormatlas.1.18

[6] C.I. Bargmann Genetic and cellular analysis of behavior in C. elegans Annu. Rev. Neurosci., 16 (1993), pp. 47–71

[7] Colbert, H.A., and Bargmann, C.I. (1995). Odorant-specific adaptation pathways gen-erate olfactory plasticity in C. elegans. Neuron 14, 803–812.

[8] Ben Arous, J., Tanizawa, Y., Rabinowitch, I., Chatenay, D., and Schafer, W.R. (2010). Automated imaging of neuronal activity in freely behaving Caenorhabditis elegans. J. Neurosci. Methods 187, 229-34

[9] De Bono, M. and Maricq, A.V. 2005. Neuronal substrates of complex behaviors in C. elegans. Ann. Rev. Neurosci. 28: 451-501.

[10] Brockie, P.J., Mellem, J.E., Hills, T., Madsen, D.M., and Maricq, A.V. (2001b). The C. elegans glutamate receptor subunit NMR-1 is required for slow NMDA-activated currents that regulate reversal frequency during locomotion. Neuron 31, 617–630. [11] Brockie et al. 2001 P.J. Brockie, D.M. Madsen, Y. Zheng, J. Mellem, A.V.

Mar-icq Differential expression of glutamate receptor subunits in the nervous system of Caenorhabditis elegans and their regulation by the homeodomain protein UNC-42 J. Neurosci., 21 (2001), pp. 1510–1522

(42)

[12] Broussard G, Liang R and Tian L, "Monitoring activity in neural circuits with genet-ically encoded indicators", Frontier, Mol. Neuroscience. Dec 5, 2014

[13] Chalasani, S.H., Chronis,N., Tsunozaki, M., Gray, J.M., Ramot, D., Goodman, M.B., and Bargmann, C.I. (2007). Dissecting a circuit for olfactory behaviour in Caenorhab-ditis elegans. Nature 450, 63-70.

[14] Chalfie, M. et al. The neural circuit for touch sensitivity in Caenorhabditis elegans . J. Neurosci. 5, 956–964 (1985).

[15] T.W. Chen, T.J. Wardill, Y. Sun, S.R. Pulver, S.L. Renninger, A. Baohan, E.R. Schreiter, R.A. Kerr, M.B. Orger, V. Jayaraman, et al. Ultrasensitive fluorescent proteins for imaging neuronal activity Nature, 499 (2013), pp. 295–300

[16] Chronis, N. (2010). Worm chips : Microtools for C . elegans biology. Lab Chip 10, 432-437.

[17] G.V. Clokey, L.A. Jacobson The autofluorescent “lipofuscin granules” in the intestinal cells of Caenorhabditis elegans are secondary lysosomes Mech. Ageing Dev., 35 (1986), pp. 79–94

[18] Davis, R.E., and Stretton, A.O. (1996). The motornervous system of Ascaris: eletro-physiology and anatomy of the neurons and their control by neuromodulators. Para-sitology 113, S97–S117.

[19] Durbin, R.M. 1987. “Studies on the development and organisation of the nervous system of C. elegans.” Ph.D. thesis. University of Cambridge, United Kingdom. [20] Faumont, S., Rondeau, G., Thiele, T.R., Lawton, K.J., McCormick, K.E., Sottile, M.,

Griesbeck, O., Heckscher, E.S., Roberts, W.M., Doe, C.Q., and Lockery, S.R. (2011). An image-free opto-mechanical system for creating virtual environments and imaging neuronal activity in freely moving Caenorhabditis elegans. PLoS ONE 6, e24666. [21] Goodman, M.B., Hall, D.H., Avery, L. and Lockery, S.R. 1998. Active currents regulate

sensitivity and dynamic range in C. elegans neurons. Neuron 20: 763-772.

[22] Lockery, S.R. & Goodman, M.B. The quest for action potentials in C. elegans neurons hits a plateau. Nat. Neurosci. 12, 377–378 (2009).

[23] Goodman, M. B., Lindsay, T. H., Lockery, S. R. & Richmond, J. E. Electrophysi-ological methods for Caenorhabditis elegans neurobiology. Methods Cell. Biol. 107, 409–436 (2012).

[24] Guo, Z.V., Hart, A.C. & Ramanathan, S. Optical interrogation of neural circuits in Caenorhabditis elegans . Nat. Methods 6, 891–896 (2009).

[25] Helms S.J., Avery L., Stephens G.J. , Shimizu T.S. (2015) Modeling the ballistic-to-diffusive transition in nematode motility reveals low-dimensional behavioral variation across species. eprint arXiv:1501.00481v1

(43)

[26] Inoue S, Spring K (1997). Video Microscopy: the Fundamentals. Plenum Press, New York.

[27] Kawano, T., Po, M.D., Gao, S., Leung, G., Ryu, W.S, and Zhen, M. (2011). An imbalancing act: gap junctions reduce the backward motor circuit activity to bias C. elegans for forward locomotion. Neuron 72, 572-86

[28] Kimura, K.D., Miyawaki, A., Matsumoto, K., Mori, I. The C. elegans thermosensory neuron AFD responds to warming. Curr. Biol. 2004;14:1291–1295

[29] Kerr, R., Lev-Ram, V., Baird, G., Vincent, P., Tsien, R.Y., and Schafer, W.R. (2000). Optical imaging of calcium transients in neurons and pharyngeal muscle of C. elegans. Neuron 26, 583–594.

[30] Larsch J., Ventimiglia D., Bargmann C. I., Albrecht D. R. (2013). High-throughput imaging of neuronal activity in Caenorhabditis elegans. Proc. Natl. Acad. Sci. U.S.A. 110, E4266–E4273

[31] Leifer, A.M., Fang-Yen, C., Gershow, M., Alkema, M.J., and Samuel, A.D.T. (2011). Optogenetic manipulation of neuroal activity in freely moving Caenorhabditis elegans. Nat. Methods 8, 147-152.

[32] Steven J. Husson, Alexander Gottschalk, Andrew M. Leifer. Optogenetic manipulation of neural activity in C. elegans: from synapse to circuits and behavior. Biology of the Cell (2013).

[33] J. Liu, A. Ward, J. Gao, Y. Dong, N. Nishio, H. Inada, L. Kang, Y. Yu, D. Ma, T. Xu, et al. C. elegans phototransduction requires a G protein-dependent cGMP pathway and a taste receptor homolog Nat. Neurosci., 13 (2010), pp. 715–722

[34] L.L. Looger, O. Griesbeck Genetically encoded neural activity indicators Curr. Opin. Neurobiol., 22 (2012), pp. 18–23

[35] Long, F., Peng, H., Liu, X., Kim, S.K. & Myers, E. A 3D digital atlas of C. elegans and its application to single-cell analyses. Nat. Methods 6, 667–672 (2009).

[36] McCombs J, Palmer A. Measuring calcium dynamics in living cells with genetically encodable calcium indicators. Methods. 2008;46:152-9

[37] Peng H, et al. Straightening Caenorhabditis elegans images. Bioinformatics 2008a;24:234-242.

[38] Piggott, B.J., Liu, J., Feng, Z., Wescott, S.A., and Xu, X.Z. (2011). The neural circuits and synaptic mechanisms underlying motor initiation in C. elegans. Cell 147, 922-933. [39] Schmitt C, Schultheis C, Husson SJ, Liewald JF, Gottschalk A (2012) Specific expres-sion of Channelrhodpsin-2 in individual or single neurons of Caenorhabditis elegans. PLoS ONE 7, e43164

(44)

[40] White, J.G., Southgate, E., Thomson, J.N., and Brenner, S. (1986). The structure of the nervous system of Caenorhabditis elegans. Philos. Trans. R. Soc. Lond., B, Biol. Sci. 314, 1–340.

[41] Grienberger C, Konnerth A. (2012) Imaging calcium in neurons Neuron, 73 (2012), pp. 862–885.

[42] Gray, J.M., Hill, J.J., and Bargmann, C.I. (2005). A circuit for navigation in Caenorhabditis elegans. Proc. Natl. Acad. Sci. U.S.A. 102, 3184–3191.

[43] Faumont S, Lockery SR. (2006). The awake behaving worm: simultaneous imaging of neuronal activity and behavior in intact animals at millimeter scale. Journal of Neurophysiology 95:1976-19.

[44] Riddle, D.L.; Blumenthal, T.; Meyer, B.J.; Avery, L.; Thomas, J.H. Introduction: The neural circuit for locomotion. In C. elegans II, 2nd ed.; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, USA, 1997.

[45] San-Miguel A. and Lu H. Microfluidics as a tool for C. elegans research. (Septem-ber 24, 2013), WormBook, ed. The C. elegans Research Community, WormBook, doi/10.1895/wormbook.1.162.1

[46] Stiernagle, T. Maintenance of C. elegans (February 11, 2006), WormBook, ed. The C. elegans Research Community, WormBook, doi/10.1895/wormbook.1.101.1

[47] Wen, Q., Po, M.D., Hulme, E., Chen, S., Liu, X., Kwok, Sen W., Gershow, M., Leifer, Andrew M., Butler, V., Fang-Yen, C., et al. (2012). Proprioceptive coupling within motor neurons drives C. elegans forward locomotion. Neuron 76, 750-761.

[48] Varshney, Chen, Paniaqua, Hall and Chklovskii. Structural properties of the C. elegans neuronal network. PLoS Comput. Biol. Feb 3, 2011

[49] Ward, A., Liu, J., Feng, Z. & Xu, X. Z. Light-sensitive neurons and channels mediate phototaxis in C. elegans . Nature Neurosci. 11, 916–922 (2008)

[50] Zheng M, Cao P, Yang J, Xu XZ, Feng Z (2012) Calcium imaging of multiple neurons in freely behaving C. elegans. J Neurosci Methods 206: 78–82.

[51] P. Nghe, S. Boulineau, S. Gude, P. Recouvreux, J.S. van Zon and S.J. Tans, Mi-crofabricated polyacrylamide devices for the controlled culture of growing cells and developing organisms, PLoS One 8, e75537 1-11 (2013).

Simultaneous calcium imaging of multiple neurons in Caenorhabditis elegans