A multi-trap microfluidic chip enabling longitudinal studies of nerve regeneration in Caenorhabditis elegans

Several sophisticated microfluidic devices have recently been proposed for femtosecond laser axotomy in the nematode C. elegans for immobilization of the animals for surgery to overcome time-consuming and labor-intensive manual processes. However, nerve regeneration studies require long-term recovery of the animals and multiple imaging sessions to observe the regeneration capabilities of their axons post-injury. Here we present a simple, multi-trap device, consisting of a single PDMS (polydimethylsiloxane) layer, which can immobilize up to 20 animals at the favorable orientation for optical access needed for precise laser surgery and high-resolution imaging. The new device, named “worm hospital” allows us to perform the entire nerve regeneration studies, including on-chip axotomy, post-surgery housing for recovery, and post-recovery imaging all on one microfluidic chip. Utilizing the worm hospital and analysis of mutants, we observed that most but not all neurodevelopmental genes in the Wnt/Frizzled pathway are important for regeneration of the two touch receptor neurons ALM and PLM. Using our new chip, we observed that the cwn-2 and cfz-2 mutations significantly reduced the reconnection possibilities of both neurons without any significant reduction in the regrowth lengths of the severed axons. We observed a similar regeneration phenotype with cwn-1 mutation in ALM neurons only.

capacity 24 h later 9,14 . These steps are very time-consuming and hinder the throughput and consistency of laser axotomy experiments in C. elegans.
In the last decade, microfluidic platforms have improved the throughput and the repeatability of the experiments in various C. elegans studies [18][19][20] . The assays those microfluidic platforms address include, but are not limited to, lifespan analysis 21,22 , neuro-developmental studies 23,24 , phenotyping 25,26 , behavioral studies [27][28][29] , electro-physiology 30 , and laser surgery 4,13,26,31 . A few different microfluidic platforms have been developed by our group and others to immobilize the nematode specifically for on-chip laser surgery and interrogation at the cellular and subcellular resolutions 4,8,13,[31][32][33] . For on-chip cellular surgery, Chung et al. developed a three-layer microfluidic chip to ablate cell bodies at a rate of 110 worms per hour 31 . To achieve a higher degree of immobilization for subcellular resolution surgery, such as laser axotomy, the most common methodology has been pressurized membranes to exert adequate force for complete immobilization 4,8,13,33 , as fabricated using double-layer soft lithography methods 34,35 . Membrane based immobilization methods enabled rapid axotomies in a larger number of animals. In a recent study, our group showed that membrane based immobilization microfluidic chips integrated with automation techniques could perform rapid and serial axotomies at a rate of 17 s/worm 13 . Despite the improved immobilization potential and processing rate, multi-layer surgery platforms are difficult to fabricate and operate. Moreover, they require transportation of processed animals out of the chip after the surgery for recovery.
In another attempt of on-chip laser surgery, a single-layer multi-trap microfluidic device has been developed and used to ablate single synapses 32 . This simple chip used a tapered channel geometry to immobilize the animals, however, post-surgery housing and imaging were limited to a couple of hours due to the lack of a separate housing and feeding areas. More importantly, the tapered geometry led the animals to rotate on the anterior-posterior axis plane of their bodies away from the glass interface, as the channel width becomes smaller than its height 36 . This body orientation requires the laser beam to go through the cover glass and a significant portion of the animal body, leading to unsuccessful axotomies and low-contrast imaging of the lateral neurons.
To overcome these limitations, we developed a new integrated microfluidic device called "worm hospital chip" that allows on-chip axotomy, post-surgery housing, and imaging all on a single device. The chip features 20 immobilization channels, based on a modified tapering channel geometry that we recently proposed and developed 37 , which enables for laser axotomy and subsequent post-surgery imaging, and a perfusion area for housing the worms after laser axotomy. The height of the immobilization channels is reduced in three steps in addition to the continuously tapering channel widths to maintain the aspect ratio around one. This geometry helps the animals to stay in their lateral orientation as they push inside the channels during immobilization. The channels' dimensions are specifically designed to accommodate both larval stage 4 (L4) animals for on-chip laser axotomy, as well as young adult (YA) stage animals for optical interrogation of neuronal regeneration after 24 h following on-chip housing (Fig. 1a,b). The optimized channel length and reduced flow rates once an animal is trapped ensured immobilization of one single animal inside the immobilization channels. The worm hospital chip is a single-flow layer device and has no active control layers. Its simple design reduces the fabrication and operation complexity of the chip making it more user-friendly than previous multi-flow layer chips, especially for non-expert users. We utilized this chip to test the effect on axonal regeneration of mutations in neurodevelopmental genes of the Wnt/Frizzled family of ligands and receptors. While these molecules are known to be involved in axonal guidance and axonal outgrowth during development [38][39][40][41][42][43][44][45][46] , only a limited number of studies have been performed to understand their effect after regeneration 9 . Here, we studied in detail how these candidate genes affect axonal repair following a severing injury. The cwn-2 and cfz-2 mutations significantly reduced the reconnection rates without any significant effect on the regrowth lengths of the severed axons for both ALM and PLM neurons, whereas the cwn-1 mutation had a similar effect selectively on ALM neurons.

Experimental Methods
Device fabrication. Standard soft-lithography techniques were used to fabricate the single PDMS layer axotomy chip 47 . The resultant photoresist pattern, produced using Mylar masks (FineLine Imaging), was used to create the master mold for all polydimethylsiloxane (PDMS; Sylgard 184, Dow Corning Corp) structures. There were four photolithography steps to create the master mold. To begin, a 4′′ wafer was spin coated with SU-8 2007 photoresist (Microchem Corp.) at 1600 rpm for 33 s to obtain a uniform height of ~10 µm. This layer was exposed to UV light through the first patterned photo-mask to achieve Zone 1 features (Z1, yellow, Fig. 1a) using a mask aligner (MA6/BA6, Suss MicroTec). This layer was then hard baked and developed to remove the unexposed SU8 photoresist. A second layer spun at 4100 rpm atop the Zone 4 feature using SU8-2025 photoresist to obtain a height of ~20 µm. The second layer was exposed to UV light through a photo-mask with the pattern for Zone 2 (Z2, red, Fig. 1a). After the hard bake and development of the second layer, a third layer was spin coated on top of the first two layers using SU8-2025 at 2850 rpm to obtain a uniform height of ~35 µm. This layer was exposed through a unique photo-mask, hard baked, and developed to achieve Zone 3 (Z3, green, Fig. 1a). The fourth layer was spin coated using SU8-2025 photoresist at 1700 rpm to obtain a uniform height of ~60 µm. Similarly, it was exposed through a unique photo-mask, hard baked, and developed to achieve Zone 4 (Z4, blue, Fig. 1a). The resultant four-layer SU8 master mold was treated with tridecafluoro-1,1,2,2-tetrahydrooctyl-1-trichlorosilane vapor (United Chemical Technologies) in a vacuum chamber at −67 kPa and 40 °C to create a hydrophobic layer on the wafer to reduce surface adhesion during the soft-lithography process.
The hydrophobic SU8 master mold was then used to fabricate the PDMS microfluidic chips with soft-lithography techniques. The PDMS (Dow Corning) was mixed at a ratio of 10:1 (base to curing agent), degassed in a vacuum chamber for 20 min to remove bubbles, and slowly poured onto the master to a height of 5 mm. The PDMS layer was cured at 70 °C for 2 h and peeled off from the SU8 mold after cooling. Individual chips were cut out using a razor and fluid interconnect holes were obtained with a 22 gauge punch using a manual punching machine (Accu-Punch MP10-UNV, Syneo). The PDMS chip was then cleaned with tape and irreversibly bonded to a 25 × 50 mm #1.5 cover glass (#2450-1½, Brain Research Laboratories) via oxygen plasma treatment at 100 W for 30 s (PE-50, Plasma Etch Inc.). The bonded device was then cured at 70 °C for 6 h to enhance the glass-PDMS bonding.
Opto-mechanical setup. The laser axotomy setup incorporates custom optics to deliver femtosecond laser pulses for surgery into a home-built epi-fluorescence microscope as described in our previous works 4, 13 . Briefly, we performed the axotomies and post-surgery imaging using an oil-immersion objective (Plan-Apochromat, 63×, NA = 1.4, Zeiss). For high-resolution imaging of green fluorescent protein (GFP) labeled axons, a mercury arc lamp (XCite 120, EXFO) passing through a FITC filter set (Semrock) provided the excitation light. A three-axis translation stage with three individual actuators (LTA-HS, Newport) operated by a single controller (ESP301-3, Newport), positioned the samples. We used a three-axis piezoelectric actuator (MAX 302, Thorlabs) with a minimal theoretical step size of 25 nm to perform the fine translation. A CCD camera (1392 × 1040 pixels with 6.45 µm pixel size, CoolSnap ES, Photometrics) captured all the images. We carried out all axotomies using a train of 300 pulses with a pulse width of 250 fs and a pulse energy of 7.5 nJ at a center wavelength of 805 nm generated at a repetition rate of 1 kHz (Spitfire, Spectra Physics). The beam energy was measured with an energy meter (PJ10, Ophir) prior to performing all axotomies and adjusted with two pairs of half wave plates and polarizing cube beam-splitters. The high numerical aperture objective tightly focused the laser beam to a spot size of approximately 620 nm based on the theoretical 1/e 2 diameter of focused Gaussian beams 48 . System operation and control. We used a modified version of our previously demonstrated automation software to control the hardware 13 . To control fluid flows in the worm inlet, perfusion inlet, and outlet fluidic channels, we used pressurized external fluid chambers controlled by three-way solenoid valves (The Lee Co., LHDA0521111H) through a manifold (The Lee Co., LFMX0510418). These chambers contained M9 buffer solution (22 mM KH 2 PO 4 , 22 mM Na 2 HPO 4 , 85 mM NaCl, 1 mM MgSO 4 , in dH 2 O). To minimize debris and avoid clogging of the device, all M9 buffer solution was passed through 1.2 µm in-line filters (Acrodisc, Pall Corp.) prior to entering the microfluidic chip. The solenoid valves in fluidic connections were independently actuated with a multichannel amplifier (Automate Scientific). The digital control of the amplifier was carried out by DAQ card (USB6501, National Instruments).
Before each experiment, we primed the device by first introducing M9 buffer solution into the worm inlet and perfusion inlet and outlet. We let M9 buffer solution flow through the exit outlet for ~2 min, which was followed by plugging the exit outlet for ~10 min. This priming procedure helped to eliminate any residual air bubbles within the tubing and the microfluidic chip.
Laser axotomy and post-surgery imaging. We studied four elements of the axonal regeneration properties of the anterior lateral microtubule cells (ALM) and the posterior lateral microtubule cells (PLM): axonal regrowth (defined by observation of a growth cone stemming from the proximal axon), axonal reconnection (defined as visual contact between proximal and distal axonal fragments), axonal fusion (defined as preservation of the distal fragment after reconnection), and regrowth lengths (measured from the site of surgery to the end of proximal fragment). Depending on the orientation of the immobilized animal, the axotomies were performed on ALM and PLM neurons located on either the left (ALML and PLML) or right (ALMR and PLMR) side of the animal's body. Axons were injured approximately 60 µm away from the cell body for animals immobilized both on-chip and on-agar using the same opto-mechanical setup. Agar pads were prepared by sandwiching 0.25 mL of melted 4% agar between two microscope slides that were then pulled apart upon cooling to create a flat and uniform surface. We used 10 µL of 10 mM levamisole solution to immobilize ~15 animals for each set of experiments. A coverslip (18 × 18 mm 2 , no. 1.5) was placed on top of the animal just prior to laser axotomy.
For post-surgery nerve regeneration scoring, a z-stack of 14 images (0.8 µm step size) around the severed axon was acquired and saved automatically for each worm. Analysis of the stack of images enabled us to assess the nerve regeneration parameters of the severed axons independent of whether they were located closer to the glass interface or not during the imaging session.
Nematode maintenance. For mutant studies, C. elegans were maintained at 16.5 °C on NGM agar plates seeded with HB101 E. coli bacterial culture using standard procedures 49  . These strains express GFP in the six mechanosensory neurons. Populations of age-synchronized animals were prepared by collecting and isolating embryos following hypochlorite treatment. Gravid adults were lysed with a small volume of a 2:1 mixture of sodium hypochlorite and 4 M sodium hydroxide, and the collected eggs were suspended in M9 buffer overnight on a rotator to aerate. The embryos were hatched overnight and were arrested in the larval stage 1 (L1) until the food was reintroduced. The L1 stage worms were then placed on HB101-seeded agar plates and allowed to grow for 48-50 hours until the worms reached the early to the middle L4 stage.
On-chip immobilization. Laser axotomy and post-surgery imaging of C. elegans need a high-degree immobilization. The new single layer axotomy chip provides immobilization of animals by geometrically restricting them inside tapering micro-channels. Individual animals are pushed inside these channels until the walls of the shortest channel section (Z2 in red, Fig. 1a,b) prevent any sinusoidal body movement such as crawling or swimming 22 . Preventing the sinusoidal body movement provides enough immobilization for axotomy on ALM neurons, located near the middle part of the worm body. However, PLM neurons are located near the tail, having space to move (Supplementary Movie 1). To eliminate the motility of the tail of trapped animals, we used a low concentration of levamisole (0.625 mM levamisole) to perform axotomy on the PLM neurons (Supplementary Movie 2). We prepared the animal suspension in M9 buffer solution with 0.625 mM levamisole as follows. First, 1 ml of M9 solution was pipetted onto animals growing on an agar pad. Second, 0.5 ml of this M9 buffer solution suspended with worms was drawn into a syringe tube containing 0.5 ml of 1.25 mM levamisole. As a result, we obtained ~50-75 animals suspended in 1 ml of 0.625 mM levamisole solution. In addition, we used 0.625 mM levamisole in an external fluid reservoir to provide buffer flow through the worm inlet during on-chip PLM axotomy and imaging.
Data availability. The data that support the findings of this study are available from the corresponding author on request.

Results and Discussion
Worm hospital microfluidic chip design. The chip constitutes two main compartments, enabling complete nerve regeneration studies inside a single device without the need for transferring them between the axotomy, recovery, and follow-up imaging procedures. The first compartment includes a set of 20 parallel immobilization channels for axotomy and imaging of the nematodes, and the second compartment includes a perfusion area for their housing and feeding after the axotomy and during post-surgery recovery (Fig. 1a).
The immobilization channels were carefully designed to accommodate three experimental requirements. (1) Immobilize the animals with a desired lateral orientation to ensure their axons are in close proximity of the glass interface for optimal optical access, (2) achieve a high degree of immobilization of L4 as well as YA stage animals, which have substantially different dimensions, and (3)  To fulfill these conditions, we adapted a modified tapered geometry design that we had recently developed, where we could trap and immobilize thousands of animals in a large scale microfluidic chip with a pressure-driven flow 37 . While the early design of tapered channel geometry ensured the high-degree immobilization 22 , we found that the arbitrary orientation of the worms was unfavorable for axotomy and post-surgical imaging 36 . The monotonically decreasing aspect ratio of the channels 21,22,32 caused the animals to rotate their lateral axis plane away from the cover glass interface 36 . To get the best optical access to the lateral neurons (ALM or PLM) and avoid rotation of the animals during immobilization, we reduced the height of the channels in addition to their width, keeping the aspect ratio around one 37 (Fig. 1b).
The perfusion area incorporates the fourth layer in the perfusion inlet and outlet (shown in yellow, Fig. 1a) to successfully house the animals over a 24 h period for follow-up imaging of the regenerating axons. This layer consists of 20 sieve channels on the sides of the perfusion area through which it connects the perfusion inlet to the outlet. The sieve structures on each arm have an opening of 8 μm × 10 μm (width × height), which are small enough to block animals from passing through them but allow the liquid culture medium to perfuse food through the feeding chamber.
We designed the perfusion arms asymmetrically to obtain different hydraulic resistance values. The low resistance arm (perfusion outlet) was essential to remove animals carefully from the immobilization channels after surgery without using high pressure. The high resistance arm (perfusion inlet) was necessary to reduce the food perfusion flow rate to approximately 1 nl/s, which was determined empirically. The food perfuses through a positive pressure gradient between the inlet and outlet perfusion arms. A 5-mL fluid reservoir with a bacterial food suspension provides the pressure gradient to the perfusion inlet. The reservoir fluid line is 4 cm above the opening of the tubing that is connected to the perfusion outlet. The slow perfusion rate allows the animals to swim freely and feed without excessive stress or having their bodies pushed up against the sieve structures during the post-surgery incubation period. The circular pillars in the perfusion area and side perfusion arms provide structural stability to avoid the collapse of the large PDMS chambers during fabrication or operation of the device.
We fabricated a single SU-8 master mold incorporating features at four different heights. This single layer PDMS microfluidic chip offers simple fabrication and lower device-to-device variation than multilayer PDMS microfluidic chips (Fig. 1c,d). Worm inlet and exit channels were fabricated to a thickness of 60 μm (shown in blue; Fig. 1a) to reduce physical stress on the animals inside the chip during the loading process and post-surgery incubation, and to avoid any undesired clogging in the exit area. The narrowest channel section (Z2 in red, Fig. 1a) was designed with the longest channel length of 1.8 mm (Fig. 1b). The relatively long channel ensured the trapping and immobilization of animals at different developmental stages, ranging from young L4 (for initial trapping and axotomy) to YA (for re-trapping and imaging subsequent axon regeneration).
Microfluidic flow sequence. The sequence of valve actuations and flow progressions for each step are shown in Fig. 2. A population of ~50-70 animals suspended in M9 buffer solution is loaded through the main inlet into the individual immobilization channels by blocking the flow at the perfusion connections (Fig. 2a). This step takes approximately 3 min. After filling most channels with a single animal, extra ones in the perfusion area are pushed out of the chip by opening the worm inlet to atmosphere and pressurizing the perfusion inlet and outlet with a pressure of ~100 kPa (Fig. 2b). Before axotomies, the trapped animals are pushed further inside the channels and immobilized by pressurizing the main inlet at ~55 kPa and closing the perfusion connections (Fig. 2c). The pressure gradient between the worm inlet and waste outlet provides the required level of immobilization to perform precise laser axotomy on each immobilized animal (Fig. 2c). Upon completion of axotomies, the animals are returned into the perfusion area for post-axotomy recovery by pressurizing the waste outlet and letting the flow go through the sieve structures. The axotomized animals are housed and fed on-chip (Fig. 2d) via a pressure gradient between perfusion inlet and outlet while the main inlet and the waste outlet connections are blocked for 24 h (Fig. 2e). To provide a constant supply of food (bacteria) in S medium, the perfusion inlet is connected to a 5 ml syringe, filled with bacteria in S medium, and the perfusion outlet is connected to atmospheric pressure 21 . After 24 h of on-chip housing and feeding, the animals are pushed once more into the immobilization channels by pressurizing the main inlet for high-resolution post-surgery recovery imaging of axon regeneration (Fig. 2f).

Trapping efficiency and orientation characterization. The overall success of this on-chip process is
defined by the efficiency of initial trapping of animals prior to axotomy and the efficiency of re-trapping housed worms for imaging 24 h later. For achieving high efficiencies for both trapping and re-trapping, we iterated the chip design several times. For example, we optimized the length of the immobilization channels to overcome the problem of multiple animals filling the same channel, and also incorporated a fourth PDMS layer to shorten the height of the sieve structures to keep the axotomized worms inside the perfusion area while providing an optimum flow rate for feeding (Fig. S1). These design iterations led us to a single layer microfluidic device (Fig. 1) achieving the desired performance for trapping both small L4 (30 μm in diameter, 600 μm in length) and larger YA stage animals (40 μm in diameter, 800 μm in length), while allowing one single animal in each immobilization channel.
In general, we observed a very high initial-trapping efficiency of the L4 stage animals and achieved desired lateral orientation during initial immobilization, which was necessary to perform the highest precision axotomy (Fig. 3a). The trapping efficiency slightly reduced for YA animals after 24 h recovery as their volume nearly doubled due to 25% increase in both their diameter and length (Fig. 3b). Figure 3c shows the quantitative results of the initial trapping and re-trapping efficiencies. Since we used a small amount of levamisole for axotomies of PLM neurons, we characterized the trapping efficiencies for both immobilization conditions, without levamisole and with 0.625 mM levamisole over 158 different sets of experiments. On average, each chip was used for 4 experiments until the cleaning of small sieve structures was insufficient to obtain a fully functioning chip.
The initial trapping efficiency was 94.6 ± 1.1% (N = 67 experiments) for the no-levamisole condition, and 91.9 ± 1.0% (N = 91 experiments) for the 0.625 mM levamisole condition (Fig. 3c). The main reasons for unsuccessful trapping were (i) clogging of the immobilization channels during animal loading, and (ii) trapping of more than one animal in a single channel. Clogging was mainly caused by the introduction of agar particulates, whereas multi-animal trapping in a single channel was due to varying size of animals within the population.
The re-trapping efficiencies were slightly lower than the initial trapping ones, ranging from 83.3 ± 2.0% (N = 67 experiments) to 85.3 ± 1.5% (N = 91 experiments) for the no-levamisole and 0.625 mM levamisole conditions, respectively. The main reasons for unsuccessful re-trapping were (i) the difficulty of re-trapping YA animals that were substantially larger than the L4 stage animals, and (ii) the loss of animals during the 24 h recovery time. Relatively low perfusion rates were needed to keep the animals healthy in the housing chamber but it resulted in loss of animals as they could freely move up the tubing connected to the main inlet.
There was no statistically significant difference between two trapping efficiencies of both groups. The overall efficiencies of the entire on-chip processes were 78.8% and 78.4% for no levamisole and with 0.625 mM levamisole, respectively. In other words, on average we could study nerve regeneration in 16 animals on chip with 20 immobilization channels. All trapped animals were successfully oriented on their lateral side as desired, and with an equal probability for their head/tail to be towards the exit outlet (Fig. S2).
Effect of on-chip immobilization and housing on worm viability. We carried out a viability test to assess the effect of on-chip immobilization and 24 h on-chip housing on worm survival, with and without the use of levamisole, and compared them to a control group of zdIs5 (mec-4::GFP, expressing GFP in six mechanosensory neurons) animals that were not subjected to any on-chip processing. We subjected both groups to an initial immobilization for 20 min with a pressure of 55 kPa, on-chip housing for 24 h, and re-immobilization for 20 min again with a pressure of 55 kPa. After re-immobilization, we flushed both trial groups out of the chip and collected them on NGM plates. We checked the viability of the worms in each population every 24 h and transferred them to new NGM plates. The average lifespans of the immobilized groups were 19.8 ± 3.7 and 19.1 ± 5.0 days for on-chip immobilization with no levamisole and 0.625 mM levamisole, respectively, whereas the average lifespan of the control group was 19.2 ± 3.8 days. We used the log-rank test to determine the difference between the viabilities of each group. We found no statistically significant differences (p > 0.90) in the outcome for both groups (Fig. S3).
Effect of on-chip immobilization and housing on axonal regeneration rates. Before starting the axonal regeneration studies on the new chip, we analyzed whether on-chip axotomy by itself affected this process by comparing the regeneration parameters to those axotomies performed on-agar as a control (Fig. 4). We determined four regeneration parameters using the following criteria: (1) axonal regrowth, as the appearance of a growth cone (Fig. 4a,b), (2) axonal reconnection, as proximity of re-growing proximal axon with the distal We used a two-tailed t-test for statistics (ns -not significant). axon (Fig. 4b,c), (3) axonal fusion, as the lack of beading or fragmentation in the distal axon (degeneration) 10,14 (Fig. 4d), and (4) axonal regrowth lengths, as the length of those regrowing axons that did not fuse. The regrowth length measurement included the regrowth length from the injury site as well as the lengths of filopodia.
To determine the effect of the worm hospital on axonal regeneration, we specifically used the transgenic strain zdIs5, expressing GFP in six touch receptor neurons, as our control. The results presented in Fig. 4e-g show no statistically significant differences for any of these parameters between axotomies performed on-agar using anesthetics and those carried out on-chip for both ALM and PLM neurons.
Regeneration properties of axonal growth and guidance, and neuronal polarity genes. The connectivity and formation of the nervous system are regulated by phylogenetically conserved molecules 38,39,50 . The Wnt ligands and their canonical Frizzled receptors have been shown to regulate in different species several developmental processes, including cell migration, neuronal polarity, axonal guidance, and axonal branching 38-43, 50-53 .  53 . We hypothesized that some of the molecules of Wnt/Frizzled family might also be involved in the axonal regeneration process. Only a few of these developmental genes were screened for their effect on the neuronal regrowth capability of PLM neurons 9 , and we decided to determine their role in axonal regeneration, reconnection and fusion in two different mechanosensory neurons, ALM and PLM. We studied single mutant strains of five genes in the zdIs5 background: cwn-1, cwn-2, egl-20, mig-1, cfz-2 using the worm hospital chip following the protocol described above (Fig. 5). First, we analyzed whether these genes affected the regrowth rate defined by the formation of a growth cone after severing the axons. For every mutant tested, we observed no significant effect in regrowth percentages for both ALM and PLM 24 h after the laser axotomy (Fig. 5a,d, Supplementary Tables 3 and 4). Next, we analyzed whether the neurons that displayed a successful regrowth also underwent a visible reconnection with their separated distal fragment 24 h after the laser axotomy. The reconnection could only be visualized within the available optical resolutions of 350-400 nm, where the proximal and distal ends looked visually in close proximity to each other. We found that in both ALM and PLM neurons the percentage of reconnection was significantly decreased in every mutant tested except egl-20 for the ALM neurons (Fig. 5b,e, Supplementary Tables 5 and 6). We then assessed whether the neurons that displayed a successful reconnection to their distal part also showed a successful axonal fusion. Most reconnected axons in both ALM and PLM neurons showed healthy distal ends, indicating that none of the five genes significantly reduced the axonal fusion rates (Fig. S5, Supplementary Tables 7 and 8). The mutations significantly reduced the reconnection percentage in most genes, led to a low number of animals to be studied for the axonal fusion phenotype. Thus, a larger scale axotomy assay may be pursued to understand the effects of the studied genes on the axonal fusion phenotype.
In summary, our results suggest that the capacity of sprouting a growth cone, as well as fusing with their own separated distal process, are not altered by mutations in any of these individual genes. Second, similarly to what happens during development, these molecules might be involved in regulating the correct guidance of the regenerating axons, except for EGL-20 in the ALM neurons.
Lastly, we measured the regrowth lengths, focusing on those axons that regrew but did not undergo a successful fusion. In both ALM and PLM neurons, mig-1 mutation significantly reduced the regrowth lengths of the severed axons (Fig. 5c,f, Supplementary Tables 9 and 10). In addition, mutations in cwn-1 and egl-20 significantly reduced the regrowth lengths in PLM neurons. Previous studies also showed a significant reduction in regrowth length of severed axons in egl-20 9 , but not in cwn-1. Interestingly, cwn-1 mutant animals did not reduce the regrowth length in ALM neurons despite the significant reduction in the reconnection rates. A similar phenotype was observed with the cwn-2 and cfz-2 mutant animals, where a significant reduction in the reconnection rates did not reduce significantly the regrowth length of the severed axons. These results suggest that the cwn-2 and cfz-2 mutations in both neurons, along with the cwn-1 mutation in ALM neurons reduce the reconnection rate by possibly affecting the guidance of the severed axons without inhibiting the capabilities of regrowth length (Supplementary Tables 11 and 12).

Conclusions
We designed, implemented, and characterized a worm hospital chip with multiple arrays for longitudinal studies of nerve regeneration in C. elegans. The microfluidic platform allows on-chip axotomy, post-axotomy housing, and post-surgical recovery imaging of nerve regeneration after 24 h, all inside the same chip. The trapping efficiency for axotomy is at least 91% with and without using levamisole as anesthetics, while orienting all of them in their lateral side for best imaging and axotomy conditions. After 24 h of on-chip housing, the re-trapping efficiency of axotomized animals exceeds 83%.
Using the worm hospital chip, we studied the effect of several developmental genes of the Wnt/Frizzled family for their role on the axonal regeneration capability of the touch receptor neurons (ALM and PLM). We observed that most, but not all, genes in Wnt/Frizzled family involved in developmental process are also essential for the axonal regeneration and successful axonal reconnection after an injury for both ALM and PLM neurons. Specifically, the mutations cwn-1, cwn-2, cfz-2, and mig-1 significantly decreased the reconnection rate for both ALM and PLM neurons, whereas egl-20 mutation significantly reduced reconnection rate only in PLM neurons. Additionally, we observed that the cwn-2 and cfz-2 mutant animals showed a significant reduction in the reconnection rates without any significant reduction in the regrowth lengths of the severed axons for both neurons. A similar effect was observed for cwn-1 mutant animals in ALM neurons only.
The new worm hospital chip offers an easy-to-use microfluidic platform especially for non-expert users. Its versatility enables nerve regeneration studies of any other neurons such as AVM or PVM neurons, and the neuronal processes located along the ventral cord. The unique capability of integrating the axotomy, housing, and post-recovery surgery imaging in a single chip is promising for future high-throughput studies.
Statistical analysis. Error of proportions was used to assess variation across a single population and two-way comparison was performed using the t-test. For more than two groups, one-way ANOVA with Dunnett's post-hoc test for multiple comparisons was used. For those data sets that did not follow a normal distribution, we used the Mann-Whitney test (Figs. 4g and 5c,f).