Automated monitoring of behaviour in zebrafish after invasive procedures

Fish are used in a variety of experimental contexts often in high numbers. To maintain their welfare and ensure valid results during invasive procedures it is vital that we can detect subtle changes in behaviour that may allow us to intervene to provide pain-relief. Therefore, an automated method, the Fish Behaviour Index (FBI), was devised and used for testing the impact of laboratory procedures and efficacy of analgesic drugs in the model species, the zebrafish. Cameras with tracking software were used to visually track and quantify female zebrafish behaviour in real time after a number of laboratory procedures including fin clipping, PIT tagging, and nociceptor excitation via injection of acetic acid subcutaneously. The FBI was derived from activity and distance swum measured before and after these procedures compared with control and sham groups. Further, the efficacy of a range of drugs with analgesic properties to identify efficacy of these agents was explored. Lidocaine (5 mg/L), flunixin (8 mg/L) and morphine (48 mg/L) prevented the associated reduction in activity and distance swum after fin clipping. From an ethical perspective, the FBI represents a significant refinement in the use of zebrafish and could be adopted across a wide range of biological disciplines.


Illustration of the Fish Behaviour Index (FBI) system
The FBI takes data from the previous 1 minute interval and refreshes every 1 minute. In our system we currently have analysis time periods of 1, 10 and 30 minutes, which means that on a rolling 1 minute basis the data from the last 1 minute is reported at time period 1 minute, data from the last 10 minutes in the 10 minute period, and data from the previous 30 minutes in the 30 minute period. The data is therefore analysed in the following periods: 1, 2, 3, …, 30 minutes; 1-10, 2-11, 3-12, … 21-30 minutes; 1-30, 2-31, 3-32, …, t to t+29 minutes. Our videos were generally stopped at 25 minutes due to a restriction in our software on the size of video file acquired. In this case, the 30 minute FBI period would use minutes 1-25 only. However, the 30 minute FBI was only used as an indicative overall evaluation (e.g. SI figure S3) because shorter periods of 10 minutes, updated every 1 minute, were adjudged to be representative analysis periods. The system was also tested over periods of 40 minutes at a time. The number, extent and updating frequency of the FBI time periods may be further adapted to suit the particular design of experiments and interventions.
Fig. S1 illustrates both the latest FBI for different time periods (Fig. S1(a)) and the history detail of the 10 min and 1 min FBI ( Fig. S1(b)). Fig. S1(a) shows the higher-level overview of FBI as in Fig. 2 of the main paper. Here, the fish is exhibiting normal healthy behavior on the various timescales. At a deeper historical level, the 1 min indicator shown on Fig. S1(b) provides a picture of behavior on a short timescale and shows considerable fluctuations between Normal and Abnormal FBI. The two Abnormal FBI states in this 40 min period are circled on Fig. S1(b). It was observed that the healthy fish unexpectedly had 2 quiescent periods at minutes 33 and 36 (prolonged periods of resting on the tank bottom) which would be more expected from unhealthy fish. By contrast the 10 min FBI remains constant and at maximal rating of Healthy for this fish after the first 12 minutes.
At an even deeper level, Fig. S2 below illustrates for a Control fish and a PIT tagged fish, displaying the extremes of the scale or index, the 'raw' data for their Activity and Distance parameters. Fig. 2 (a) and (d) (main paper) shows the FBI results derived from such underlying raw data of Fig S2. The FBI analysis module is currently contained in a separate Microsoft Excel file for ease of use and can provide real-time functionality or alternatively be used to analyze subjects' test files (coordinate logs) retrospectively (post processing). It generates the FBI over periods of the latest 1 minute, 10 minutes, 20 minutes and 30 minutes (with provision also for latest 60 minutes). The FBI is qualitative (with categories Healthy, Ok, Unhealthy, Abnormal) and is currently based on a combination of Distance Swum (DS) and Activity (A).    S2 shows raw Activity as zones (1-9) visited over time (with the tank divided into 9 zones covering the area available to the fish) and Distance travelled over time for the Control and for the PIT fish of Fig. 2 over 25 minutes at post-treatment time point (3). It also shows the front (side) and top views of the tank using plots of the locations visited by the fish. These data are used in the derivation of the FBI of Fig. 2 (main). The raw data of Fig. S2(b) shows marked differences between a PIT treated fish and the Control fish. Fig. S2(b) shows typical characteristics of abnormal wellbeing: very low raw Activity, low changes in raw Activity level; low Distance swum with several periods of stationary behavior between 'pulses' of movement; low utilization/exploration of the tank.
2. Evaluation of the system The deeper system components, 10-  The 10-minute FBI gives considerable detail on developments in welfare. In period (ii), C8 declined in its normal activity, going through an 'Ok' phase which declined further to Unhealthy and Abnormal for 3 and 4 minutes respectively before recovering through U and O for 2 and 15 minutes respectively. In period (iii) it was again Healthy overall, as expected for an untreated fish. The effect of the fin clip on FC2 was dramatic as the fish immediately went into Abnormal condition throughout periods (ii) and (iii) without recovering. By contrast, FC6 declined over periods (ii) and (iii) but only reached Abnormal for 3 minutes. Otherwise, the fish recovered to Healthy for 7 minutes in period (ii). Overall, it appeared less affected by the treatment than the severely affected FC2. FBI characterizes more gradual variation in behavior over the medium-short term, whereas the longer term 30 min view was far less sensitive to short term fluctuations.   S4 shows 10min FBI updated minute by minute. It further illustrates the interplay between 30min and 10min FBI for several behaviors. Sham1 ( Fig. S4(b)), has a 'quieter' period of 12 minutes in period (iii) seen in the consecutive 10min FBI 'Ok' categorizers extending from minute 3 to minute 14. Nevertheless, overall (25min) it is adjudged Healthy. AL1 ( Fig. S4(c)) has similar dips in period (i), which include Unhealthy and Ok but is overall Healthy. In period (iii) it is overall Unhealthy (but not worse than that) after dipping to Abnormal for 13 minutes extending from minute 6 to minute 18. The 10min FBI characterizes more gradual variation in behavior over the medium-short term. The longer term 30 minute view is far less sensitive to short term fluctuations. (An extended version of Figure S4

Validation in Two Dimensions
FBI dependence on 3D versus 2D monitoring was investigated since many laboratories may not have the capacity to put video cameras above tanks in a typical zebrafish racking system. This affected distance travelled thus the software was modified to operate with one camera (side view) and FBI was adapted (publically demonstrated at the Blue Planet Aquarium 2015 (www.blueplanetaquarium.com/fish-health-monitor-trial-this-saturday-at-blue-planet).
To compare 2D with 3D FBI, the 3D coordinates for the 5 fish of Fig. 2 (main) were also run through the modified 2D FBI system (i.e. 2D FBI utilizing front camera coordinates only).
Fifteen videos from experiment 1 were chosen at random and re-assessed using only 2D data.
Out of 15 only 2 differed from 3D analysis but were within the Healthy/OK or within Unhealthy/Abnormal categories thus the overall wellbeing status was within healthy and

Equipment Setup and Operation
The system runs tracking software 15  Six are detailed in the following Table 1 and an example screen of the system in use is shown in Fig. S6. Additional videos include the tracking system in operation and an example of a fish improving to Healthy from Unhealthy (SI Videos 1 and 2). Overall for this timescale: seems Normal, intermittently exploring full tank volume.
The following screen example (Fig. S6) shows the fish subject (right, identified with pink circles), its trajectory (center right) and the detailed FBI analyses (left half) and figure S7 demonstrates how the FBI assesses zebrafish from each treatment group 2 h after treatment.   Fig S7 presents, for all the subjects in the 5 groups of female zebrafish, mean FBI for the 30 minute timescale at the 3 initial timepoints (pre, 1h, 2h). For clarity, each group is normalized to its mean Pre value. FBI of the Control and Sham groups remained virtually constant and Healthy throughout. FBI for the Fin Clip group, which is most affected, declined by ~50% at timepoint 2 and another ~25% from timepoint 2 to 3, taking this group into the Abnormal FBI category. The other treated groups responded similarly to Fin Clip but with reduced severity. PIT ended at timepoint 3 with FBI ~50% reduced from pre-treatment whereas the least affected group, Acid Lip, has FBI reduced by about 35%. Acid Lip and PIT treated fish reached the Unhealthy category but avoided the Abnormal category. The results from the FBI in Figure S7 reflect the statistically significant results obtained for the Acid Lip 1%, PIT and Fin Clip treatment groups (Table S2) based on individual parameters from the Principle Components Analysis (Figs. 4 and 5).

Discussion
Behavioural effects of treatment: Female zebrafish were profoundly affected by the invasive procedures employed in the present study with reductions in swimming speed and the amount of the tank explored and an increase in the use of the bottom of the tank. The fin clip and 5/10% acid lip groups had not recovered by the end of the 6 h experiment and their behaviour still differed from pre-treatment behaviour. This prolonged, complicated behavioural change over 6 hours with no recovery indicates that this was not a simple nocifensive reflex 1 but was a substantial modification of the animal's normal behaviour 2 . This response was distinguishable from sham-handled fish whose behaviour did not differ discernibly from controls thus the changes in behaviour are not due to the stress of anaesthesia and handling 1 . These changes were ameliorated by the administration of pain-relieving drugs and thus this information can be used to develop analgesic protocols for fish.

Acetic acid:
The injection of increasing concentrations of acetic acid provoked a change in behaviour, similar to that observed in a previous studies 3 , which was dose dependent. The injection of 1% acetic acid had only a minor impact on the percentage of tank-explored post 2 and 3h before exploration returned to normal. An increase in concentration up 10%, however, yielded a much greater behavioural response resulting in a significant change across all three behavioural traits. Other studies have used direct observation rather than a behavioural analysis tool but similarly have demonstrated zebrafish reduced their activity after a potentially painful event 3,4 . A reduction in activity and exploration has been observed in zebrafish and trout during painful stimulation 4 and mirror the immobile states and reduced exploration that higher vertebrates can experience in response to a noxious stimulus 5 . Reilly et al. 4 hypothesised that these behaviours may serve to protect the animal from predation when injured by allowing the conservation of energy that may be needed in a fight or flight response and making the individual less conspicuous.
PIT tagging. The insertion of PIT tags resulted in a departure from control behaviour 2 hours after implantation and was characterized by a reduction in average speed and tank exploration with a clear preference for the bottom of the tank. The potential for this procedure to cause pain is considerable given the risk of damage to the musculature of the abdominal cavity, yet the addition of extra weight via the PIT tag could add an energetic cost to movement resulting in the observed reduction in activity. Previous studies investigating the impact of tagging on a broad spectrum of species found no effects on critical swimming velocity 6-8 even when the weight of the tags reached 6-12% of the bodyweight of the individual 8 which are well above the relative weight of the PIT tags (2% of zebrafish bodyweight) used in this study.
Fin clipping. The fin clip procedure resulted in reductions in average speed, tank exploration and a clear preference for the bottom of the tank after treatment. The fin clip procedure resulted in a significant departure from normal behaviour earlier at 1 h and across more time points suggesting that the impact of the fin clip was more immediate and potentially of greater severity relative to the PIT tag. These changes in behaviour have been observed in previous studies in zebrafish 9.10 . Currently, fin clip is a procedure which is deemed mild severity under EU legislation 11 and is believed to result in mild or acute pain for a few hours but our results show that the responses to fin clipping persists for several hours and as such should be deemed moderately severe. The use of immersion analgesia to alleviate any associated pain would ensure that any pain would be reduced and the procedure was indeed mild.
As with the PIT tag procedure, it is possible that the reduction in activity was related to the physical impediment of having 40% less tail fin as opposed to being a complex response to pain. The complete absence of a tail fin in the no-tail strain of zebrafish resulted in a 65% reduction in critical swimming performance 12 , an effect that could account for the observed changes in behaviours linked to activity. If the changes in behaviour are related to the ability to swim with a shorter tail then the administration of analgesics would have no affect; however, certain analgesics at a specific dose helped to ameliorate the behavioural impact of the fin clip thereby confirming the potential for this procedure to be painful.

Impact of drug use:
The efficacies of four drugs, from three different classes of analgesics (NSAIDs, opioids and local anaesthetics), differed in their ability to prevent the behavioural change induced by the fin clipping in female zebrafish. Lidocaine (5mg/L) was successful analgesic, since it reduced the effect of the fin clip across all behaviours for the duration of the experiment. Lidocaine treated zebrafish exhibited behaviours that consistently aligned with that observed in the control group even though they had been fin clipped. Studies have found lidocaine to be effective in rainbow trout 13 and zebrafish 10 . Together these findings further validate this drug as a promising analgesic in fish as it has now shown effectiveness in alleviating more than one pain type (chemical and mechanical) in both a cyprinid and a salmonid. The second local analgesic tested was not quite as successful: at 6h fish treated with all doses of bupivacaine displayed behaviour akin to fin clip without analgesia since they were significantly different from controls. In mammalian models the local anaesthetic bupivacaine can have a superior duration of activity 14 ; however, this action was not observed here. Instead the highest dose (1 mg/L) did not alleviate the fin clip procedure to the same degree as the lidocaine and can be deemed less effective. Perhaps a higher dose of bupivacaine is required and this should be explored in future studies.
The opioid morphine and NSAID flunixin both proved effective and had a clear dose dependent effect with the highest dose in each case leading to the greatest reduction in fin clip mediated behavioural change. Morphine is effective at ameliorating the impact of noxious stimuli across a large spectrum of vertebrates [15][16][17][18] including teleost fish 19 . In this current study, 48mg/L of morphine treated fish behaviour showed average speeds that were similar to control fish. The reverse, however, was observed with the lowest dose of morphine. Although previous studies have largely focused on the injection of morphine, only two have looked at administering morphine via the immersion route in goldfish (Carassius auratus) 20,21 . Jansen and Greene 21 demonstrated the rapid uptake of morphine from water in goldfish, however, Newby et al. 20 could not replicate these results and found a much slower rate of uptake. Despite this slow uptake the high dose of morphine (48mg/L) still resulted in a reduction in pain related behaviours; morphine within the water was therefore hypothesised to act centrally 20 . The results from this current study demonstrate the effectiveness of morphine administered via the immersive route, albeit at a high dose, although future work should examine the uptake kinetics and potential side effects of morphine in zebrafish to help better determine an effective dosage.
The highest dose (8mg/L) of the NSAID flunixin meant fish behaviour did not differ over time and was similar to controls except at the 6h time point. The behaviour average speed in the 8mg/L group were similar to that observed in controls and elevated compared to that observed in the fin clip group. The efficacy seen in the highest dose steadily decreased with dose with the behaviour of the 2mg/L flunixin group resembling that of fin clipped fish and average speed was significantly different from controls at 2, 3 and 6h. NSAID's function through the inhibition of the enzymes arachidonate cyclo-oxygenase 1 (COX-1) and 2 (COX-2); functional genes for both of these enzymes are conserved in zebrafish 22 and levels of the NSAID diclofenac as low as 1 μg/L are known to reduce COX expression in fish 23 . This would suggest that flunixin at 8mg/L is not as effective as lidocaine or morphine or that future studies should test higher doses.

Conclusion.
Taken together it would seem from the present study that lidocaine (5mg/L) administered via immersion prior to treatment is the most effective drug to prevent fin clip induced changes. Morphine was also effective but given the cost and regulatory restrictions in its use it may not be widely adopted. We recommend that analgesia is provided for all invasive procedures that cause tissue damage as it is likely they may give rise to the sensation of pain to ensure good welfare. However, where the analgesic drug itself may confound data collection and justifiably cannot be used then experimenters should wait 24 hours after an invasive procedure before beginning behavioural data collection to allow the zebrafish to recover. After this period it has been shown zebrafish behaviour returns to normal after fin clipping 10 .