NemaLife chip: a micropillar-based microfluidic culture device optimized for aging studies in crawling C. elegans

In this study, we report a microfluidic device for the whole-life culture of the nematode Caenorhabditis elegans that allows the scoring of animal survival and health measures. This device referred to as the NemaLife chip features: (1) an optimized micropillar arena in which animals can crawl, (2) sieve channels that separate progeny and prevent the loss of adults from the arena during culture maintenance, and (3) ports that allow rapid accessibility for feeding the adult-only population and introducing reagents as needed. The pillar arena geometry was optimized to accommodate the growing body size during culture and emulate the body gait and locomotion of animals reared on agar. Likewise, feeding protocols were optimized to recapitulate longevity outcomes typical of standard plate growth. Key benefits of the NemaLife Chip include eliminating the need to perform repeated manual transfers of adults during survival assays, negating the need for progeny-blocking chemical interventions, and avoiding the swim-induced stress across lifespan in animals reared in liquid. We also show that the culture of animals in pillar-less microfluidic chambers reduces lifespan and introduces physiological stress by increasing the occurrence of age-related vulval integrity disorder. We validated our pillar-based device with longevity analyses of classical aging mutants (daf-2, age-1, eat-2, and daf-16) and animals subjected to RNAi knockdown of age-related genes (age-1 and daf-16). We also showed that healthspan measures such as pharyngeal pumping and tap-induced stimulated reversals can be scored across the lifespan in the NemaLife chip. Overall, the capacity to generate reliable lifespan and physiological data underscores the potential of the NemaLife chip to accelerate healthspan and lifespan investigations in C. elegans.


Results and discussion
Optimization of the NemaLife chip design. C. elegans lifespan measurement in a microfluidic environment requires mimicking plate-like behaviors with the capacity to remove progeny. Prior micropillar devices [48][49][50] provide some guidance on construction of environments that can recapitulate plate-like crawling behavior, while lifespan devices with swim chambers 35,36 offer some insight into the design of sieve channels. At the same time, configuring micropillar devices for life-long culture with progeny removal and survival analysis requires new design considerations and optimal culture conditions. In this section, we discuss the device design considerations and the optimal micropillar geometry for scoring lifespan and healthspan measures. We then present optimization of cultured conditions in the next section "Optimization of worm culture conditions".
Basic device design. We designed the NemaLife culture device (Fig. 1a-d) with design objectives that feature: (1) a micropillar arena that can accommodate the growing body size as young adults are cultured and can enable www.nature.com/scientificreports/ pockets that can develop at the inlet of the device during overnight storage in the incubator. During washing/ feeding the fluid first passes through the vent port, forcing the air out before the introduced fluid enters the pillar arena. Fluid manipulation is performed manually using hand-held syringes, an aspect of design that we felt would make the device more accessible to use in a wide range of laboratories than would a complex pumping set-up.
Micropillar arena design. Worm habitat chambers are designed to have an approximate footprint of ≈ 60 mm 2 for a population size of 10-15 animals (4-6 mm 2 per animal), compared to an average footprint of 2828 mm 2 in standard studies on agar for 30-50 animals (60-90 mm 2 per animal). We were able to accommodate nine of these chambers on a 50 × 75 mm 2 glass slide (Fig. 1a) that could be used to generate survival data on ≈ 100 animals. The geometry of the micropillar lattice in the NemaLife chamber is crucial for successful measurement of lifespan of C. elegans. The lattice structure must accommodate changes in body size during reproduction and aging, maintain the natural crawling gait of C. elegans, and allow non-invasive removal of progeny while retaining the sample population. To identify the optimal micropillar lattice that simultaneously meets these requirements, we fabricated devices with square arrangement of pillars with different pillar diameter (a) and edge-edge gap (s). The nominal dimensions we tested are: Device I, a = 40 μm, s = 60 μm; Device II, a = 50 μm, s = 80 μm and Device III, a = 60 μm, s = 100 μm. The measured dimensions of the three pillar devices are reported in Table S1. During the greatest period of growth, C. elegans body diameter varies from ~ 50 to 100 µm and length varies from ~ 900 to 1500 µm (Fig. S1). Thus, Device I provided the tightest, and Device III the leanest, confinement for the animals during the lifespan measurement.
In all the devices, the pillars had a uniform height of ≈ 75 μm and a clearance from the floor of the habitat chamber of approximately ≈ 25 μm, allowing the pillars to be moved aside by the animal to adjust gait and accommodate changes in body size. The gap between the pillars and the clearance between the pillar tip and the floor also helps in size-based separation of the adults from the larvae and eggs. An adult's body diameter is much larger than the clearance and the animal interacts with approximately 10-16 pillars along its body, which prevents the animal from being washed away during fluid flow. Specifically, animals maintain their posture and crawl normally using the pillar support while the progeny, eggs and other smaller particles are removed via fluid flow.
Sieve channel design. Our optimized sieve channels have a width of 25-30 μm, length of 750 μm, and are separated from each other by 75 μm × 750 μm rectangular blocks. We designed the length of the rectangular blocks long enough so that the animal cannot generate natural waves and propagate through the channels. In addition, we found that the placement of the sieve channel relative to the pillars is important as animals tend to use the pillars to generate thrust to force themselves through the channel. Placing the sieve channels at a distance of 1000 μm away from the nearest pillar mitigated this behavior and allowed the animal to quickly reverse back to the arena when it approached the sieve channel area.
Selection of the optimal micropillar geometry based on locomotion and lifespan outcomes. We initially measured the crawling gait in terms of wavelength, amplitude and crawling speed in the three devices for day 4 animals (from hatching). The amplitude and wavelength corresponding to the worm undulatory motion in the three devices were similar (Fig. 2a). However, the crawling speed was diminished in Devices I and II compared to Device III due to their tighter confinement.
Also of interest is the comparison of data in the test devices to that of animals crawling on agar 55 . We found that the amplitudes are similar between the two 52 . However, the wavelength was about 30% higher on agar. Animals do crawl in Device III with a similar speed to animals crawling on agar surfaces 55 . Thus, among the three devices tested, animals housed in Device III have a locomotory gait that is most similar to that observed on agar plates.
We conducted survival experiments in the test devices at the optimal feeding protocol of 100 mg/mL of E. coli OP 50 with animals being fed once a day (see "Optimization of worm culture conditions" section for the feeding optimization study). Figure 2b compares the survival data for the animals reared in the three devices. Devices with tighter pillar spacing resulted in a reduction of worm lifespan, possibly due to restraints in natural locomotion and stress conferred through body confinement. We found that median and maximum lifespan in Device III were most consistent with studies on agar (see "Validation of the NemaLife chip" section) We conducted all subsequent aging investigations using Device III with s = 100 µm.
Effect of population density on animal lifespan in the micropillar chamber. Population density in each micropillar chamber may limit food availability as well as accumulation of small molecules such as pheromones that help animals sense overcrowding 56 and makes them enter into a non-aging dauer-stage 57 or extends their lifespan by reducing the insulin signaling 58 .
To investigate the effect of population density, we conducted lifespan experiments in which we varied the number of animals per chamber and manipulated the micropillar chamber size to vary the population density (see Table 1). We maintained the culture in the chamber by daily washing and feeding with 100 mg/mL of E. coli OP50. In Table 1 we compare the median lifespan of animals experiencing population densities ranging from 4.0 to 12 mm 2 /animal. We note that population density is typically defined as number of animals per unit area, but here we chose the inverse unit to highlight the pillar area needed for chamber design. We find that the median lifespan corresponding to different population densities is not significantly different compared to our standard NemaLife chip condition of 4-6 mm 2 /animal. This data suggests that in the explored range of animal density, the NemaLife chip does not introduce deleterious effect on animal lifespan.
Additionally, Table 1 shows the lifespan data when an individual animal is housed in the chamber which prevents transmission of the chemical/hormonal signals due to physical isolation. Still, we find that lifespan www.nature.com/scientificreports/ data from the single-animal chamber is comparable to that of our standard NemaLife chip suggesting that in the explored range of population densities, animal crowding does not have a major influence on lifespan. Recognizing that the population density in the chip is much higher than on plates, we believe that daily washing and supply of fresh food may be the reason that animal density may not be as influential in the chip as it is on plates.
Although the different population densities in the chips show comparable lifespan, the live/dead scoring time increases significantly as the population and chamber size increases. Alternatively, having too few animals per chamber will require more chambers to reach an adequate population size for a particular study. We chose to use 10-15 animals per chamber as an optimal trade-off between ease of scoring and achieving a sample size of ≈ 100 animals per assay condition.
Optimization of worm culture conditions. We established culture maintenance protocols to achieve robust and reproducible lifespan data while increasing the overall efficiency of conducting survival analyses. Initially, we focused on identifying the optimal washing conditions necessary to remove all progeny. To do this,  www.nature.com/scientificreports/ we cultured reproductive adults within the NemaLife device for 24 h to allow progeny production and growth. We then tested progeny removal by delivering 200 µL aliquots of S-complete media via the loading port of the device. Figure 3a, b shows the habitat chamber with adults and their progeny (see SI movie 1). After washing, the retained adults are shown in Fig. 3c. In the event that the animals are already near the exit, sieve channels retain them inside the chamber (Fig. 3d, SI movie 2). We found that all adults were retained, and all progeny were effectively removed using a total wash volume of 1 mL for three different loading conditions ( Fig. 3e; SI movie 2). The washing operation took approximately 90 s.
All trials were successful in removing progeny. We washed each chamber and used a stereomicroscope to determine for efficacy of progeny removal. During washing, the progeny that are washed out are eggs and L1 larvae. In case any progeny are left out accidently, these progeny can be L2 (diameter is ~ 25-35 μm) or L3 (diameter is ~ 40 μm) and can still be washed out the next day through the sieve channel of width 25-30 μm. Occasionally when bagging (internal hatching of progeny which can kill the mother) occurred, more repeated washes (2-4 mL) were necessary to remove the progeny of bagged mother or the resulting advanced larval stage progeny. We note that during the washing process, animals in the chamber respond by exhibiting faster crawling momentarily, probably due to the stimulation by fluid forces (see SI movie 2). www.nature.com/scientificreports/ Next, we sought to establish a robust feeding protocol that provides the worms with an adequate supply of food. E. coli OP50 is widely used as the standard diet for C. elegans cultures. Previous lifespan studies in multi-well plates 16,59 and pillar-less microfluidic devices 34,36 used a bacterial concentration of 10 9 -10 10 bacterial cells per mL of S-medium and added the bacterial suspension to the microfluidic devices either continuously 36 or once a day 34 .
In our study, we used the bacterial concentration from previous studies as a starting point to optimize the feeding protocol and evaluate the lifespan. The feeding conditions we tested include 100 mg/mL E. coli OP50 once a day, twice a day, or every other day, where 100 mg/mL E. coli OP50 was found to be equivalent to ≈ 2.4 × 10 11 colony forming units (cfu)/mL. In addition, we also increased food by testing 200 mg/mL E. coli OP50 once a day. We note that the microfluidic chamber can hold ≈ 6.75 μL and ≈ 250 μL of food volume was injected into the device, indicating there was no significant food dilution occurring in the chamber.
We find that worms fed every other day at 100 mg mL −1 had an overall extension in maximum lifespan but a significant decrease in median lifespan as well as a high death rate during reproduction (Fig. 4a). Alternatively, we noted that feeding twice per day did not significantly alter lifespan (Fig. 4a) or age-dependent changes in body size (see SI Fig. S1). Figure 4b shows that animals fed once a day at 100 mg/mL or 200 mg/mL had similar survival curves suggesting that feeding 100 mg/mL daily is sufficient for measuring survival curves.
Next, we assessed whether the animals might become dietary restricted when supplied with a daily dose of 100 mg/mL bacterial solution. We measured the density of the bacterial solution at the end of a feeding cycle by collecting the washed fluid from each NemaLife chip. Each chip had 9-14 animals and 8 chips were used as replicates. Figure 5 shows the residual bacterial density measured daily during the reproductive period, starting from day 3 until day 7, to account for the consumption of bacteria from hatched progeny. We found that the bacterial density dropped significantly after day 3 and remained approximately the same (≈ 1.5 × 10 11 cfu/mL) during the rest of reproductive period.  www.nature.com/scientificreports/ Dietary restriction has been shown to occur on plates at bacterial densities of 1 × 10 8 cfu/mL 31 . The food levels remaining after daily feeding are substantially above the dietary restriction levels even during the peak reproductive period when adults are consuming significantly more and hatched progeny also contribute to the depletion of food sources 60 . Thus, we conclude that a diet of 100 mg/mL E. coli OP50, added once daily, is sufficient for whole-life culture of C. elegans in the NemaLife chip.

Whole-life C. elegans studies in pillar-less microfluidic chambers. Previous microfluidic works
focusing on lifespan measurement in C. elegans used chambers without pillars [34][35][36][37][38] . In these pillar-less microfluidic devices, animals experience swim-induced physiological stress, potentially affecting lifespan and healthspan outcomes. To clarify the potential impact of life-long swimming, we used the same overall design of NemaLife chip but without the pillars in the arena. We used the optimized washing and feeding protocol and conducted in parallel whole-life studies in the pillar-less and pillar-equipped chips. Specifically, for animals cultured in both environments, we measured survival curves and scored for the age-related vulval integrity disorder (avid) phenotype as a measure of physiological health.
Comparing the progeny-washing process between devices with and without pillars, we found that in the pillar-less chip, progeny removal is challenging and requires additional wash steps since the animals accumulate at the sieve channels blocking the flow of fluid (see SI movie 3). Thus, additional washes were needed to ensure removal of progeny in pillar-less chips. Another important distinction is that although animals in the pillarequipped NemaLife chip crawled naturally (Fig. 6a, i), many animals in the pillar-less chip were found to assume a stiff (black arrow in Fig. 6a, ii) and stationary posture as well as show occasional coiling behavior (red arrow in Fig. 6a, ii). These unnatural body postures in pillar-less microfluidics chambers suggest that animals suffer from swim-induced fatigue 40,42 .
In Fig. 6b, we show the survival curves for animals in both the pillar-free and pillar equipped environments. The median lifespan of animals cultured in the pillar-less environment was 11 days compared to 14 days in the pillar environment. However, maximum lifespan was similar for the two populations. Thus, animals experiencing swim-induced physiological stress have shorter mean lifespan.
The larger death rate during early reproductive period in the pillar-less chip is due to animals showing agerelated vulval integrity disorder (avid). Avid is characterized by vulval protrusion and expulsion of intestinal fluid or tissues (see Fig. 6c) which has been shown to increase with hypoxic stress, decreasing temperature, and poor reproductive health 61 . In the pillar-laden chips, we observed the first instances of avid after the reproductive period, which is consistent with studies on plates 61 . However, we find that worms cultured in the pillar-less chambers experience avid as early as day 4 and avid occurs more frequently compared to animals in the pillarenvironment (Fig. 6d). In sum, the reduced lifespan and the frequent occurrence of the avid phenotype in the swim-chambers observed in the absence of pillars, suggest that the pillar environment of the NemaLife chip is crucial for mitigating stress during life-long liquid culture of C. elegans.
Validation of the NemaLife chip. Next, we discuss studies that were conducted in the NemaLife chip to validate our microfluidic approach for lifespan measurement in C. elegans. Specifically, (1) we compared the lifespan data for animals cultured in our microfluidic devices versus those maintained on agar plates, (2) we www.nature.com/scientificreports/ compared the stress induced in the NemaLife chip to that on agar plates, (3) we measured lifespan of mutants with known aging pathways, and (4) we tested efficacy of RNAi interventions in the device. The lifespan data corresponding to all these conditions is included in Table S2 in the Supplementary Information.

Comparison of C. elegans lifespan in device and on agar.
To evaluate if our optimized NemaLife device generates C. elegans lifespan data consistent with that of standard agar plate assays we conducted parallel lifespan analysis of young adults using established protocols. Figure 7a shows that housing worms in the NemaLife chip does not significantly alter lifespan (p > 0.11). The median and maximum lifespan from three replicates on agar were 15.67 ± 0.58 and 24.33 ± 0.58 days respectively. Likewise, the median and maximum lifespan in the microfluidic device were 16.0 ± 0.0 and 24.67 ± 2.08 days, respectively. Here, the maximum lifespan is taken as the day before the last animal of a population is dead. Although we find that the lifespan curves are in good agreement between NemaLife chip and agar assays, we find that the NemaLife chip offers considerably less animal loss during a survival assay. On agar, we found 5-30% animal loss over the course of the assay. It is common to lose animals on agar plates due to (1) crawling www.nature.com/scientificreports/ up along the side wall and death from desiccation, (2) burrowing into the soft agar which precludes scoring, and (3) bagging. Animal loss due to desiccation and burrowing depends on the type of strain, sex, mutation, and intervention used, and sometimes may account for as much as 50% of the total animal population 62 . As death events due to desiccation and burrowing cannot be scored, whether agar-based lifespan is a measure of a selected subset of a population becomes a question. Moreover, loss of animals increases the number of worms needed to initiate each experiment as loss must be anticipated. We found that the NemaLife chip eliminates the incidences of animal loss from desiccation and burrowing due to culture in an enclosed microfluidic chamber. We observed a 0-6% animal loss, which could be attributed to washing mistakes (human error of not plugging the loading port, which lacks a sieve channel, or of excess pressure application for fluid flow such that animals close to the sieve channel squeeze out). Animal loss/censoring from bagging was 1-6% in the NemaLife chip compared to about 2-10% on agar plates in this study.
We also evaluated whether the NemaLife chip induces environmental stresses such as starvation in the animals. We chose a strain that expresses the stress reporter DAF-16::GFP 63 , which exhibits DAF-16 nuclear localization under caloric restriction, heat, and oxidative stresses [64][65][66] . We first determined that the strain harboring this reporter (TJ356) exhibited similar lifespan on both agar plates and in the microfluidic device, indicating that culture in the liquid environment of the microfluidic pillar device does not induce deleterious effects on TJ356 survival (Fig. 7b, see additional trial data in SI Fig. S2).
We then performed fluorescence imaging to assess DAF-16 localization. In the device at 20 °C, day 3 and day 8 animals, or on agar at day 8, animals did not exhibit such DAF-16::GFP localization (images in Fig. 7c-i, ii, iii). We quantified the visible puncta in day 8 animals cultured on agar plates and in microfluidic chambers and found no significant difference (Fig. 7c-iv). Thus, although animals cultured in our microfluidic device can www.nature.com/scientificreports/ induce stress responses similar to those cultured on agar, the standard growth conditions we use do not elicit daf-16-dependent translocation to the nucleus. Over the course of the study, we conducted 20 separate lifespan assays of WT worms in the NemaLife chip (see SI Fig. S3), allowing us to account for seasonal changes in laboratory environments. We found that variation in lifespan is limited to 13%, a level of replicate variation comparable to that found in lifespan assays conducted in LSM technolology 62 .
Consistent with previous reports, daf-2, eat-2 and age-1 mutants exhibited robust extension of lifespan (Fig. 8a, b). For this data set, daf-2, age-1, eat-2, and wild-type shows a 100%, 64.2%, and 15.4% increase in median lifespan in the microfluidic pillar environment, respectively (Fig. 8a, b). For daf-16 mutants, in Fig. 8a we did not observe a statistically significant difference, but we observed a significant decline in the trial shown in Fig. 8b and an additional trial shown in Table S2.
We note that the maximum lifespan of daf-2 is 44 days, which demonstrates that the microfluidic culture environment can adequately support long duration longevity studies in C. elegans. We have also successfully measured lifespan of mutants with locomotory defects in the NemaLife chip indicating that fluid stress due to daily washing/feeding and our death scoring protocol does not elicit undesirable outcomes (Fig. S4). We conclude that NemaLife chip can both support long term culture of C. elegans and yield longevity outcomes that parallel those reported on agar plates.
We also tested whether genetic screens using RNAi can be pursued in the NemaLife device. In C. elegans targeted post-transcriptional RNA knockdown can be achieved by feeding animals bacteria that harbor clones expressing specific double stranded RNAs 72,73 . We found a 15.4% extension and a 15.4% reduction in median lifespan of worms fed RNAi targeting age-1 and daf-16, respectively, when compared to worms fed an empty vector (PL4440) as a control (Fig. 8c). Likewise in terms of maximum lifespan, we found a 25% extension and a 8% reduction in age-1 and daf-16 respectively. RNAi efficacy in the NemaLife chambers establishes that food adequate to elicit the RNAi-mediated silencing of the targeted gene is ingested by the animal, addressing a potential concern on the efficacy of RNAi intervention in the microfluidic environment.

Scoring healthspan measures in C. elegans.
In addition to lifespan measurement, the transparency and shallow depth of the PDMS worm-habitat chamber offer the opportunity to evaluate physiological characteristics and fluorescent biomarkers of healthspan in C. elegans. In this study, we focused on manual scoring of pharyngeal pumping and stimulus-induced forward and reversal speed. The ability to score such phenotypes across lifespan demonstrates the capacity of NemaLife as a device for healthy aging investigations in C. elegans.
Pharyngeal pumping. The pharynx of C. elegans is a heart-like organ that uses rhythmic contraction and relaxation to facilitate bacterial uptake 74 . Pharyngeal pumping rates depend on several factors, such as food availability and environmental quality, and significantly decline with age, making pumping evaluation an attractive physiological marker for evaluating C. elegans health status 45,47 .
Lockery et al. 46 developed a microfluidic device to measure pumping rates by recording the electrical activity of the pharynx while the animal is immobilized. Scholz et al. reported an image-based pharyngeal pumping measurement technique in which the movement of the grinder of an immobilized worm in a microfluidic device (WormSpa 75 ) is tracked 76 . However, in both cases, the microfluidic devices are specifically designed to immobilize animals of a given body size, and as a result these approaches are not conducive for whole-life culture and recording pharyngeal pumping measurements across the lifespan of the animal. Likewise, in prior pillar-less microfluidic devices in which animals swim continuously, it is difficult to record pharyngeal pumping due to 3D motion.
Using our NemaLife platform, we confirmed previous reports of age-dependent reduction of pharyngeal pumping in wild-type C. elegans (Fig. 9, see Movies S4, S5). Additionally, we observe temporal changes in pharyngeal pumping rates throughout various life stages. In animals maintained within NemaLife, we report that pharyngeal pumping rates increase up to the end of the reproductive period, reaching a maximum of ≈ 283 cycles/min on day 8. Pumping rates decrease gradually starting from day 10, reaching ≈ 117 cycles/min on day 20, at which point, degeneration of the pharynx makes it difficult to score pumping frequency.
Maximum pharyngeal pumping rates we observed are similar to that reported on agar 45 , however, the rate of age-dependent decline in the pharyngeal pumping was relatively slow in the microfluidic device compared to agar. Old animals (day 15) grown on agar exhibited a significant reduction in pharyngeal pumping rate (20-30 cycles/min 45,47 ) while animals housed in the microfluidic device maintain a relatively high pharyngeal pumping rate (150-160 cycles/min) in late life. Studies show that C. elegans increases rate of pharyngeal pumping with food concentrations in liquid environments [76][77][78] . It is possible that the abundance of high-quality food even after reproduction makes the decline less pronounced. It is also possible that less bacterial biofilm is formed in the microfluidic environment and therefore animals are less prone to bacterial colonization in the pharynx. Nevertheless, it is clear that the age-associated pumping rate decline can be monitored using the NemaLife chip. The high pharyngeal pumping rate in the NemaLife chip may be advantageous for enhanced uptake of compounds in pharmacological assays. www.nature.com/scientificreports/ Stimulated locomotion. Locomotory vigor is commonly used as a healthspan measure in C. elegans 45,47,79 . Forward crawling of C. elegans is accompanied by pauses and reversals at a speed and frequency that are dependent upon food availability and the crawling environment. As a result, temporal fluctuations in locomotion make it difficult to evaluate true crawling speed. Extended tracking and long-term analysis of time-lapse images are required to properly assess forward locomotion dynamics. Reversals are observed during natural locomotion and are important for escape in response to gentle touch, a behavior in which the worm quickly reverses and suppresses head movement [80][81][82][83]  www.nature.com/scientificreports/ crawling bouts. Reversal behavior is thus an indicator of neuro-muscular function 82 that can be scored reliably in a short observation time.
On agar plates, reversals are induced by applying gentle touch to the animal with an eyelash or by prodding the worm with a platinum pick or by simply tapping the plate 80,84 . Here, we replicate the gentle touch stimulus in the microfluidic device with a hex key to induce stimulated reversals. We induced reversals mechanically by gentle tapping on the top surface of the device 3 times at a location slightly away from the worm pharynx. Stimulus is transferred to the worm as mechanical vibration through the pillar and the fluid. This stimulus-induced locomotory response cannot be evaluated in prior pillar-less microfluidic devices because it is difficult for the animals to exhibit reversals in a fully liquid environment. Thus, the tap-induced reversal is unique to the NemaLife chip.
Stimulated responses generally involve an initial reversal (first reversal), a change in direction, followed by a final forward movement (Fig. 10a). In nearly all cases, we observed that animals immediately respond to a tap by exhibiting a first reversal followed by a turn. Occasionally, we observed brief interruptions in forward crawling motion that appeared to be independent of the stimulus. Inset of Fig. 10a shows the average speed calculated from the different modes of crawling. As expected, the first reversal showed the highest speed.
Using the NemaLife chip, we show that both reversal and forward speed vary significantly with age and we find that reversal speed is always greater than forward speed in worms of all ages (Fig. 10b). Most importantly, the rate of decline in reversal speed is accelerated at the end of reproduction. More specifically, between day 8 and day 12, the reversal speed measure is 50% reduced, whereas stimulated forward speed is only 17% reduced. Interestingly, the decline in the stimulated reversal speed is correlated with an equally rapid decline in survival rates following the reproductive period, an observation that underscores the importance of expanding lifespan assays to include the evaluation of additional health span metrics. Overall, the NemaLife chip enables measurement of stimulated reversal speed as a novel biomarker for aging and healthspan, which couples with other health measures to establish a powerful platform for analysis of C. elegans healthspan and lifespan.

Assessment of throughput of NemaLife chip.
In this study, we developed NemaLife chip for life-long culture of C. elegans focusing on lifespan assays. Standard lifespan assays on NGM plates involve manual steps of animal transfers and scoring animal deaths. The operational protocol for NemaLife chips also involves manual steps of washing, feeding and scoring for live/dead animals. Therefore, it is useful to evaluate the throughput of conducting lifespan assays with the NemaLife chip.
To estimate the labor time taken for conducting a lifespan assay with NemaLife chip and plates, we considered a sample size of 100 animals and the maximum lifespan as 4 weeks. This sample size approximately corresponds to one NemaLife chip (9 chambers × 10-15 animals per chamber) or 4 NGM plates (4 plates × 30 animals per plate). Table 2 shows the estimated labor time when the assay is conducted on NGM plates or NemaLife chip.
Overall, we find that a lifespan assay takes about 2X-4X less labor time with NemaLife chip than on plates. The significant savings in time come from eliminating manual picking of animals as well as from scoring. In the NemaLife chip, progeny washing is done by a hand-syringe (connected with a tubing and a pin) allowing the user to wash one chamber and move to the next. The scoring in NemaLife chip also takes less time since the population of 100 animals is compartmentalized into 9 smaller chambers with fewer animals compared to plates. On NGM plates, animals explore a larger footprint and with a limited field of view which requires searching. In www.nature.com/scientificreports/ addition, death events can be problematic to score in translucent agar, whereas on the chip the transparency of PDMS, narrow chamber height and cleaner background reduces the time required to score. In Table 2, we have only listed the labor time for conducting the assay, but we did not consider the preparation time which includes making plates or chips, obtaining synchronized animals, growing bacteria and storing plates or chips in incubators. The preparation time for animal and bacteria culture are similar for the two approaches, however, the preparation time for NGM plates may be longer than the NemaLife chips since a single Tap stimulus was applied (frame # 10, red arrow) to induce reversal. Animal executed reversal with a very high speed followed by second reversal. Animal turns (blue arrow), crawls in forward direction (black brackets) for relatively long time with pause (amber bracket). Naturally occurring reversal period is shown by the pink bracket. Average speed of the full episode, only forward (including pauses), only reversal and stimulated reversal are shown in the inset. Error bars are standard deviations. n = 5. (b) Decline in forward and first reversal speed as wild type C. elegans ages in the micropillar arena. Error bar is the standard deviation. n > 10, food: 100 mg E. coli OP50/mL S complete, feeding frequency: once/day. www.nature.com/scientificreports/ chip is required for a full lifespan assay while 15-20 NGM plates are required depending on the frequency of animal transfers. In general, other considerations such as device failure, contamination, and clogging can affect the throughput of NemaLife chip. During the course of this study, we used 33 chips for lifespan assays and only 11 out of 247 chambers (4.45% chambers) failed during the experiment. The failure is due to human errors such as forgetting to block the worm loading port with a solid pin which causes animals to escape during washing/feeding, or adding ethanol accidentally in one chamber. Despite these inadvertent errors, we were still able to obtain experimental data since the chip has 9 chambers and censoring of one or two chambers provides meaningful data although at a reduced sample size. We did not observe any major failure of NemaLife chambers due to clogging. The NemaLife chambers are designed symmetrically with respect to inlet and outlet. Our protocol involved using each port alternatively and thus keeping the sieve channels free of clogging. In addition, contamination of chips was not observed due to pre-sterilization of chambers prior to experimentation and also because the enclosed chamber minimizes exposure of the small amount of culture fluid (≈ 6.75 μL) to ambient environment. Finally, we note that the NemaLife chips cannot be reused for lifespan as the chambers accumulate chemicals and bacterial residue over the lifespan assay potentially influencing aging biology and lifespan of C. elegans if chips are re-used.

Conclusions
We demonstrated that C. elegans can be effectively maintained in our NemaLife microfluidic device across its lifespan without using chemicals (progeny-blocking drugs, antibacterial agents, antifungal compounds, etc.) in an environment that recapitulates longevity on agar plates. The closed microfluidic environment keeps animals inside the arena boundary and eliminates the events of death by desiccation. Being fully occupied with fluid, the culture environment maintains uniform humidity and minimizes temperature fluctuations. Micropillars in the microfluidic device enable the animals to maintain natural crawling gaits and eliminate swim-induced stress. The well characterized survival outcomes of "cornerstone" longevity mutants with altered insulin-like signaling or dietary restriction pathways grown on agar are reproduced in the device; RNAi can also be executed.
The capacity for manual injections of fluids and food enables investigator control over feeding conditions or introduced drugs. For compound testing, the liquid environment in the NemaLife chip ensures the uniform availability of the compound which is difficult to ensure in agar-filled plates. In the NemaLife chip, we replace the old environment (larva, eggs, 24 h old bacteria and buffer, pheromones, excrete etc.) daily with fresh food. As a result, animals can be exposed to a fresh dose of compounds daily, similar to dosing schedules with mammalian models. An added benefit of the device is that progeny and effluent (containing any chemicals such as, pheromones etc.) might be collected for downstream analysis.
The NemaLife chip was also designed to improve the scoring of live/dead animals. The microfluidic chip consists of 9 smaller chambers on a 2" × 3" glass slide. Small footprint of an individual chamber is entirely visible at 10 × or 20 × magnification on a stereomicroscope. Each chamber was optimized to house 10-15 animals so that one can score animal surival easily. Thus, we achieve a large enough sample size for drawing meaningful statistical inferences without employing a computer vision system and complicated imaging system otherwise required for a larger chamber. These considerations make the NemaLife chip easily implementable in research labs conducting lifespan assays and investigator scoring options provide quality control on data.
In sum, the NemaLife chip is a simple and low-cost means to obtain reliable lifespan and physiological data on aging animals. The NemaLife chip can be developed into a fully automated platform, that can offer progeny removal, feeding, drug delivery, and phenotypic scoring using automated pump systems and an appropriate software. Thus, the NemaLife chip has a significant potential both as a low-cost "manual" device that can be used in many labs and as a core component of a fully automated platform.

Experimental procedures
Worm culture. All animals were cultured on 60 mm petri dishes containing nematode growth medium (NGM) at 20 °C before loading into the microfluidic chamber. The NGM filled petri dishes were seeded with 300-400 μL of bacteria Escherichia coli OP50 and incubated for 48 h at 20 °C. For age synchronization, 20-25 gravid adults were placed on seeded plates to lay eggs for 2-4 h. After eggs were laid, the animals were removed from the plates, eggs were incubated for 60-72 h. The day the eggs were laid was scored as day 0. Device fabrication and preparation. All microfluidic devices were fabricated in poly(dimethyl)siloxane (PDMS) using soft lithography 85 . A mold was fabricated using two-step SU-8 photolithography such that the chamber height is ≈ 100 μm and the micropillar height is ≈ 75 μm, as described previously 52 . A 4-6 mm thick PDMS (Sylgard 184 A and B, 1:10 by weight, Dow Corning) layer was casted on to the mold and the inlet/outlet holes were punched with a 1 mm hole puncher. The PDMS device was then bonded on a glass surface irreversibly and rendered hydrophilic by plasma treatment (Harrick Plasma Inc.). Before using the device for lifespan experiments, the device interiors were filled with 70% ethanol for 5 min to sterilize them. Subsequently, the device was rinsed 4-5 times with S-complete solution. Devices were then treated with 5 wt% Pluronic F127 (Sigma-Aldrich) for 30 min to prevent protein and bacterial build-up 36 . In addition, Pluronic treatment also assists with removal of air bubbles if any are trapped. After incubation, excess Pluronic was removed by washing with S-complete. The Pluronic-treated devices were stored in moist petri dishes at 20 °C for immediate use and at 4 °C for future use. Bacterial density measurement and food consumption. To assess whether the diet of the animals becomes restricted during overnight incubation, we determined the amount of food consumed by the animals during reproduction, a period of peak food consumption. We first obtained a calibration curve relating bacterial density in cfu/mL to optical density readouts. Using this calibration, we determined the daily consumption of food by the animals. A Synergy HT plate reader was calibrated for E. coli OP50 culture using standard microbiological techniques. Briefly, following the growth of overnight culture at 37 °C, 100 mg/mL E. coli OP50 in S-medium was serially diluted to 2 11 . Density of diluted suspensions was determined by measuring the absorbance of 150 µL aliquots in a flat-bottom 96 well plate. Each diluted suspension was subsequently serially diluted to 10 -7 and 0.1 mL of bacteria was plated onto standard LB plates. Plates were incubated at 37 °C, and individual colonies were counted after 24 h. A plot of cfu/mL against OD600 readings of the original sample was plotted to obtain a linear regression. 9-14 Young adult (65 h) worms were loaded into NemaLife housing chambers by using 1 mL syringe. 100 µL of E. coli OP50 suspension (100 mg/mL) was pushed through the chamber to eliminate potential dilution by the liquid buffer (~ 6.75 µL) that remains in the chamber. To determine the concentration of bacterial cells upon feeding each chamber was washed with buffer and the solution collected immediately after introducing bacteria to the chamber. Microfluidic chips were stored at 20 °C for 24 h, then chambers were washed with buffer and fluid is collected for each. Fresh E. coli OP50 is then replaced as described above. Each sample from before and after incubation is centrifuged at 3500 RPM for 10 min and resuspended in 1 mL of fresh buffer. Bacterial density (cfu/mL) was determined by measuring the optical density (wavelength 600 nm) of 150 µL of the fluid collected from the chamber and resuspended in buffer. OD600 measurements are applied to our calibration curves (see below). We determined an OD600 of ~ 1.0 correlates to 4.1 × 10 11 cells/mL. Fluorescence imaging. We imaged the C. elegans strain containing stress reporter gene zls356IV inside the device without immobilizing them using a Nikon Ti microscope at 10X magnification. Movies were captured using fast time lapse imaging with a camera (Zyla 5.5 sCMOS from Andor Inc.). For imaging worms cultured on agar plates, worms were loaded into a fresh chamber and imaged immediately. Movies were analyzed using ImageJ (NIH) software.
Scoring animal death. Worms were counted manually, and lifespan was scored daily. An animal was scored as dead if it failed to respond to (1) gentle flow of fluid throughout the chamber or (2) gentle tapping of the device by a 3/8″ Allen key. If there is no movement in the pharynx or in the tail 1 min after the stimulus has been applied, we scored the animal as dead. Each death event was scored as 1 and unaccounted deaths (missing, washing error, matricides) were scored as 0. A lifespan curve (Kaplan-Meier) was then generated. Kaplan-Meier curves and Log-Rank statistics were generated using the Statistics Toolbox in MATLAB.

Locomotory measures.
Reversal and forward speed were scored after a gentle tap on the PDMS device with a 3/8″ Allen key, at a location close to the tip of the pharynx. Only the initial, instantaneous reversal following stimulation were scored. Continuous frames of a spontaneous start of reversal and end of reversal were taken as a complete reversal episode. Reversal episodes with a pause, stop, or intermittent reversal were not included. Continuous forward locomotion until a spontaneous reversal was scored to determine forward speed. Images were captured using SVSi streamview camera at 10 frames/s. Locomotion speed was calculated by tracking the displacement of the pharynx in time. The maximum speed in an episode (forward/reversal) of a worm was recorded as the maximum reversal/forward speed.
Pharyngeal pumping. Movies of 10 individual worms were captured at a rate of 20 frames/s with SVSi streamview camera on a Zeiss stereo microscope at 5 × magnification, 15 min after the addition of food to the device. Complete cycle time of contraction/retraction of the isthmus/terminal bulb of pharynx was manually www.nature.com/scientificreports/ computed using ImageJ (NIH). The number of pharyngeal pumping cycles were counted over a 10 s period for each single animal and then reported as cycles per minute.
Data analysis. All the survival analyses were conducted in MATLAB (Mathworks, R2014b). Log-rank (Mantel-Cox) test was used to compare survival between treatment groups. Two-sample t test was used to compare growth of the worm and crawling kinematics in the optimization study.