Monitoring the movement of large populations of worms over time

Multiple technologies have been designed to monitor worm movement. They mostly involve samples of worms (Petri dishes, multi-well plates or microfluidic devices) that need to be moved under a single fixed high-power camera in sequence to record images or videos, or an array of stationary worm samples where a camera moves between them (see ‘Discussion’ and Table 1). These technologies present the challenges of precisely and consistently aligning moving parts. To monitor large numbers of worms simultaneously, we use an array of low-grade cameras each connected to its own single-board computer. This distributed computing approach with no moving parts is designed to be robust and scalable. For example, hundreds of single-board computers can be connected to a single gateway computer using Ethernet, whereas connecting lots of cameras to PCs using USB has lower limits of scaling and creates more issues with space, cabling, cooling and data handling.

The WormGazer™ is designed to use the same agar Petri dishes as used in standard C. elegans laboratory culture. Each 6-cm Petri dish is illuminated from above by permanently-on white LED panels and imaged by a camera below it (Fig. 1A). This camera is controlled by its connected single-board computer. As temperature is an important variable for ageing in C. elegans and since changes in temperature can affect the moisture content of the Petri dishes, the chamber containing the dishes is carefully temperature controlled to within ± 0.1°C using a feedback system of cooling by blowing externally chilled air (from an air-conditioned room for example) to counteract the heat emitted by the lights and electronics. Extra heating is provided by small resistance heaters when necessary.

Fig. 1

Schematic of the WormGazer™. A Petri dishes are placed in a temperature-controlled box with one imaging station per Petri dish. This diagram represents how the camera is positioned to take images and is connected to a single-board computer. The camera takes an image every 0.8 s for 160 s and repeats this process every 5 min. The first level of image processing occurs on the single-board computer. B A transmission image from the camera of a plate with 30 L4 worms (left). The images’ pixel values are averaged across 200 images in each imaging window (middle left), producing a high-contrast image where movement can be detected (middle right). The fraction of animals moving, and speed which the moving objects move are two main outputs (right)

Every 5 min, each camera records 200 images in a 160 s time window (an image every 0.8 s). The contrast between the worm and its translucent medium allows for the measurement of changes in pixel values (Fig. 1B). When the images from an imaging window are averaged and subtracted from each frame, and then overlaid, a high-contrast image is created, representing the movement in the imaging window, where pixels which changed in value during the imaging window due to movement are denoted as white and the unmoving background denoted in black (Fig. 1B). Where and how far a worm moves during the imaging window is therefore determined by following the path of these bright pixels at their centre of mass [24]. A threshold speed for detection is set to 10 µm s−1 to limit noise from other changes to the plate which the camera can pick up, such as worm trails in the bacterial lawn. There are some false positives occasionally resulting in apparently more than 100% worms moving at a single time point, but these anomalies are not at a sufficient level to mask movement at a population level.

All objects moving above the threshold are used to measure the fraction of worms moving (number of objects) and the mean speed of moving worms (length of objects) (Fig. 1B). The position of the worm on the plate (coordinates of object) is also recorded. There is an area limit for worm detection, denoted by the red ring (Fig. 1B). Using a series of sequential images to detect movement, this system can identify worms that move during the imaging window. It is not necessary to track each individual worm because measuring changes in behaviour at a population level is sufficient to measure ageing. Data analysis occurs on locally connected servers, and Python scripts are used to create initial data outputs.

Once an experimental run is started, the boxes are left undisturbed for the duration of the run, which is between 7 and 14 days. The worms are placed on the dishes at the L4 stage, and the box normally runs at 24.0°C. Under these conditions, wild-type worms become adults by 24 h, show a peak in movement around 48 h and then begin their ageing-associated decline in movement and speed. This system allows for the non-invasive monitoring of worms in their normal laboratory setting.

Characterisation of the movement of the long-lived age-1(hx546) mutant

To test whether the system could detect changes due to ageing in C. elegans mutants established to show effects on lifespan, we compared the movement of the long-lived age-1(hx546) and the short-lived daf-16(mu86) mutants with a wild-type (WT) control. A minimum of 298 worms across ten Petri dishes per condition were compared over a 12-day period. Two micromolars of floxuridine (FUdR) was added to the media to prevent eggs from hatching.

The fraction of worms moving at any time is defined as the proportion of the population that moves during the 160 s imaging window. The lines of the graph are smoothed, with shading of the standard error of the mean (SEM) (Fig. 2A). Movement of the worms reaches a peak around day 2 so this time point is considered to be the start of age-related decline and is used as the starting point for the area under the curve (AUC) calculation, which produces the average time the worms spent moving in that time period.

Fig. 2

Movement analysis of age-1, daf-16 and WT worms over 12 days. A The fraction moving graph shows the proportion of worms moving during the imaging window with SEM shading. n ≥ 298 worms, 10–12 Petri dishes per condition. B The area under the curve integration for days 2–7. C The area under the curve integration for days 2–12. D Fraction moving of an independent repeat, with SEM shading. n ≥ 420 worms, 14 Petri dishes per condition. E The mean speed of moving worms for A. F The mean speed of all worms, which is a function of A and E. G The area under the curve integration for days 2-12 for the mean speed of all worms (F). Petri dishes with 2 µM FUdR on DM. ** = p < 0.01, *** = p < 0.002, one-tailed test

The age-1 mutant worms spend 93.5% more time moving between days 2 and 7 compared to the WT (Fig. 2B), and the worms continue to show an increased fraction moving until the end of the experiment (Fig. 2A). A video showing the movement in two representative Petri dishes illustrates this result and how the technology works (Supplementary Video S1). Interestingly at an early stage of adulthood (days 0.5 to 2), when ageing is likely to be insignificant, age-1 mutants show a 12.1% lower fraction moving compared to WT (p < 0.05, Supplementary Table S1). This unexpected result suggests that the age-1 mutant has fewer moving worms in early adulthood compared to the WT.

The daf-16 mutant showed no significant difference in average time spent moving in the first 7 days when compared to the WT, but when the AUC is calculated for days 2 to 12, it spent 23.2% less time spent moving compared to WT (Fig. 2C). Overall, these results are consistent with slower ageing in the age-1 mutant because it stays active for longer than the WT, while daf-16 experiences a faster decline than the WT. The data show excellent reproducibility (Fig. 2D).

The mean speed of moving worms is a function of how far each detected worm moves during the imaging window (distance/time = speed, Fig. 2E) and does not consider those worms which do not move during the window. It is independent of the fraction of worms moving. In this analysis, the age-1 mutant had a 27% lower speed than WT at day 2 (p < 0.002, Supplementary Table S2), as well as reaching its maximum mean speed at day 1.53 compared to day 1.94 for the WT (Fig. 2E). Meanwhile, there was no significant difference in speed between WT and daf-16. After day 10, no more speed data are available for daf-16 as at least one entire dish of worms had zero worms that moved above the detection threshold of 10 µm s−1.

When the “fraction moving” and “mean speed of moving worms” parameters are multiplied, it produces the “mean speed of all worms” (Fig. 2F). This graph produces a representation of how quickly all animals moved in the experiment. In this case, it shows that age-1 mutants have a lower peak in speed but continue moving at higher and faster levels for longer, and shows that daf-16 mutants have an earlier and steeper decline in speed than the WT. The area under these curves represents the average distance the worms move and shows that age-1 mutants moved a significantly greater distance than the WT but daf-16 mutants were not significantly different from the WT (Fig. 2G). A full comparison of the two repeats can be found in Supplementary Fig. S1.

Automated movement analysis vs manual lifespan

To compare movement analysis between the WormGazer™ and a manual lifespan assay, both approaches were performed in parallel. Twenty-six dishes of 30 temperature-sensitive sterile glp-4(bn2) worms per condition were prepared with 50% analysed on the WormGazer™ and 50% by manual lifespan methods (n = 390 worms set up per condition, per method). Worms were transferred to fresh dishes on days 7 and 14 with both methods. SMX was used as a positive control. SMX extends C. elegans lifespan in a dose-dependent manner up to at least 256 µg/mL by inhibiting folate synthesis in OP50 Escherichia coli. Inhibiting bacterial folate synthesis inhibits an E. coli activity that accelerates ageing [25]. In this experiment, we used low concentrations of SMX that, in a previous study, increased lifespan with an effect that was significant (4 µg/mL and 8 µg/mL) or not significant (1 µg/mL) [26].

The manual lifespan experiment (Fig. 3A) showed a statistically significant increase in survival with increasing SMX concentration (p < 0.0001 for all conditions, Wilcoxon test). Mean survival for 1 µg/mL SMX was 20 days, while that for 4 µg/mL and 8 µg/mL was 21 days compared to control which was 18 days (Supplementary Table S1). This experiment took 40 days to complete.

Fig. 3

Manual lifespan versus WormGazer™ movement analysis. glp-4 sterile worms were placed on as L4s and were transferred on days 7 and 14, being placed either into the WormGazer™ or into the 24°C incubator. A Survival curve of the manual lifespan method. B Fraction moving of the worms on the WormGazer™. An interrupted axis has been added to the fraction moving plot to indicate when the worms were transferred in the first two instances, while the last gap indicates when the machine was restarted. All these instances created bumps in movement which were smoothened by omitting 46 imaging periods. C Integration of the area under the curve is shown for days 2–22 and D days 2–7. E The mean speed of moving worms and F the mean speed of all worms which multiply B and E. Manual scoring occurred every other day on weekdays from day 7 onwards. ** = p < 0.01, *** = p < 0.002, one-tailed test. Compound added to DM agar. n ≥ 260 worms, 12 Petri dishes per condition, per technique

In the WormGazer™ assay, worms showed an improvement in movement levels on all SMX concentrations (Fig. 3B) that was statistically significant from day 2 onwards (p < 0.002) (Fig. 3C). Notably, by day 7, the difference was already clear. The AUC between days 2 and 7 showed that all SMX concentrations led to significant improvements in healthspan (p < 0.01 for 1 µg/mL SMX, p < 0.002 for 4 µg/mL and 8 µg/mL SMX) (Fig. 3D). The automated monitoring also collected speed data (Fig. 3E, F), showing that the mean speed of moving worms was significantly slower for the SMX conditions at day 2.5 (p < 0.002, Supplementary Table S3) but speed across the whole experiment was significantly increased with SMX (Fig. 3F, inset).

The WormGazer™ experiment required disturbing the worms only three times over the time period while the manual lifespan required doing so every other day from day 7 onwards. No movement above the automated movement threshold was detected after day 22, whereas the manual method could detect smaller movements and response to prodding, and so lasted until day 40.

Overall, the two methods showed comparable results but with large differences in user effort and the amount of data captured. An AUC from days 2 to 7 showed the same significant result as a 40-day manual lifespan, which indicates that a 7-day healthspan can be used to detect lifespan extensions without requiring the work and time of a lifespan assay.

Testing alpha-ketoglutarate

Next, we tested alpha-ketoglutarate (αKG), a metabolic intermediate in the Krebs cycle shown to extend lifespan in C. elegans [27]. To understand the effect of αKG on ageing, we conducted an experiment to look for a concentration-dependent effect on healthspan from 0.5 to 8 mM, with 8 mM being the reported effective concentration. We found that 8 mM, in fact, significantly reduced worm movement while 0.5 mM and 2 mM had no effect (Fig. 4A).

Fig. 4

Alpha-ketoglutarate improves worm health. A Fraction moving of worms on alpha-ketoglutarate and its integration of the area under the curve (inset). n ≥ 180 worms, 6–8 Petri dishes per condition. B Fraction moving of a smaller concentration range, identifying two positive concentrations with integration of the AUC (inset). n ≥ 330 worms, 11–12 Petri dishes per condition. C Mean speed of all worms for the B experiment and integration of AUC (inset). D Fraction moving of a repeat for B, showing the same results with integration of AUC (inset). n ≥ 300 worms, 10–11 Petri dishes per condition. E Fraction moving of a smaller concentration range, identifying three positively effecting concentrations with integration of AUC (inset). n ≥ 150 worms, 5–9 Petri dishes per condition. * = p < 0.05, ** = p < 0.01, *** = p < 0.002, one-tailed test. Compound added to DM agar. Temperature-sensitive sterile glp-4 worms were used

We tightened the concentration range to 0.5 to 4 mM and used higher animal numbers, to find that 2 mM and 4 mM αKG significantly improved movement by 12.7% and 10.7%, respectively, in the first 7 days compared to control (p < 0.002, Fig. 4B). Furthermore, 4 mM improved the mean speed of all worms (Fig. 4C). The mean speed of moving worms for all experiments can be found in Supplementary Fig. S2. This finding was reproducible (Fig. 4D). An even smaller concentration range was then assayed to find that 3 mM and 5 mM αKG were also effective in improving movement (Fig. 4E). Overall, by using a 7-day automated monitoring experiment, we found αKG is effective in maintaining health across a small range of concentrations.