24x7 Automated Behavior Tracking For Rodent Safety Pharmacology & Phenotyping

1. 24/7 Automated Behavior Tracking for Rodent Safety Pharmacology and Phenotyping Sponsored by:

2. InsideScientific is an online educational environment designed for life science researchers. Our goal is to aid in the sharing and distribution of scientific information regarding innovative technologies, protocols, research tools and laboratory services.

3. Thank you to our event sponsor

4. Presenters: David Craig Actual Analytics dcraig@actualanalytics.com Sara Wells, PhD MRC Harwell s.wells@har.mrc.ac.uk Will Redfern, PhD AstraZeneca will.redfern@astrazeneca.com

5. 24/7 Automated Behavior Tracking for Rodent Safety Pharmacology and Phenotyping David Craig CEO Actual Analytics Copyright 2015 - David Craig, Actual Analytics, and InsideScientific. All rights reserved

6. Actual Analytics • Founded in 2010 • Staffed by Ph.D’s in computer vision, machine learning, neuroscience & biology • Founded to transform the industry • Committed to excellent Innovation contact: David Craig: dcraig@actualanalytics.com +44 (0) 7803 162005

7. The Journey We have worked with Astra Zeneca and the MRC under two “Crack-It” projects with the assistance of the NC3R’s we express our profound gratitude to all three for their significant support, engagement and vision. contact: David Craig: dcraig@actualanalytics.com +44 (0) 7803 162005

8. The Product ActualHCA (Home Cage Analyser)is the only product that permits scientists to gather 24x7 data and automated analysis of the complex behaviours of group housed rodents, identity retained, in their real home cage, with major 3Rs benefits contact: David Craig: dcraig@actualanalytics.com +44 (0) 7803 162005

9. Preliminary Validation of a Home Cage Automated Behavioural Monitoring System in Rats Will Redfern PhD, FSB, FBPhS, DSP Translational Safety Drug Safety & Metabolism AstraZeneca Copyright 2015 – Will Redfern and InsideScientific. All rights reserved

10. Disclaimer The views expressed are those of the presenter and do not necessarily reflect the views of AstraZeneca plc. This presentation should not be interpreted as an endorsement of any technology product mentioned, neither AstraZeneca plc nor the presenter have any financial or commercial interest with the manufacturers of any such product, nor is there any other conflict of interest.

11. Attrition of candidate drugs during development Cook et al. (2014) analysed the causes of failure of candidate drugs in AstraZeneca during 2005-10 (safety; efficacy; strategic portfolio decisions). They found that ~60% of drug project closures were safety- related. David Cook et. al., Lessons learned from the fate of AstraZeneca’s drug pipeline: a five-dimensional framework. Nature Reviews Drug Discovery (2014) 13: 419-431. Safety-related attrition

12. Attrition due to inadequate safety – Why? Shortcoming Impact Solution? 1. Lack of early detection of safety signals ‘Doomed’ compounds enter in vivo toxicology phase Improve frontloaded screening: in silico and in vitro 2. Lack of detection of safety hazards preclinically ‘Doomed’ compounds enter clinical development Improve quality and increase information content of safety pharmacology and toxicology studies 3. Lack of confidence, knowledge, or precision in preclinical- clinical translation Defective risk assessment: ‘Doomed’ compounds may be let through, anticipating a large safety margin; ‘safe’ compounds may be stopped, anticipating an inadequate safety margin. Improve risk assessment and decision- making by better understanding of the translation of the preclinical signals to humans.

13. Attrition due to inadequate safety – Why? Shortcoming Impact Solution? 1. Lack of early detection of safety signals ‘Doomed’ compounds enter in vivo toxicology phase Improve frontloaded screening: in silico and in vitro 2. Lack of detection of safety hazards preclinically ‘Doomed’ compounds enter clinical development Improve quality and increase information content of safety pharmacology and toxicology studies 3. Lack of confidence, knowledge, or precision in preclinical- clinical translation Defective risk assessment: ‘Doomed’ compounds may be let through, anticipating a large safety margin; ‘safe’ compounds may be stopped, anticipating an inadequate safety margin. Improve risk assessment and decision- making by better understanding of the translation of the preclinical signals to humans. *Redfern WS et al. (2013) Functional Assessments in Repeat-dose Toxicity Studies: The Art of the Possible. Toxicology Research 2: 209-234.

14. Impact of adverse effects of drugs by organ function throughout the pharmaceutical life cycle…

15. The various toxicity domains have been ranked first by contribution to products withdrawn from sale, then by attrition during clinical development. Phase ‘Nonclinical’ Phase I Phase I-III Phase III / MKTG Post-Marketing Post-Marketing Information: Causes of attrition Serious ADRs Causes of attrition ADRs on label Serious ADRs Withdrawal from sale Source: Car (2006) Sibille et al. (1998) Olson et al. (2000) BioPrint® (2006) Budnitz et al. (2006) Stevens & Baker (2008) Sample size: 88 CDs stopped 1,015 subjects 82 CDs stopped 1,138 drugs 21,298 patients 47 drugs Cardiovascular: 27% 9% 21% 36% 15% 45% Hepatotoxicity: 8% 7% 21% 13% 0% 32% Haematology/BM: 7% 2% 4% 16% 10% 9% NERVOUS SYSTEM 14% 28% 21% 67% 39% 2% Immunotox; photosensitivity: 7% 16% 11% 25% 34% 2% Gastrointestinal: 3% 23% 5% 67% 14% 2% Reprotox: 13% 0% 1% 10% 0% 2% Musculoskeletal: 4% 0% 1% 28% 3% 2% Respiratory: 2% 0% 0% 32% 8% 2% Renal: 2% 0% 9% 19% 2% 0% Genetic tox: 5% 0% 0% 0% 0% 0% Carcinogenicity: 3% 0% 0% 1% 0% 0% Other: 0% 0% 4% 16% 2% 2% Adapted from Redfern WS et al. SOT 2011 1-9% 10-19% >20%0%   

16. Studying adverse effects of drugs on the nervous system • Neuronal-astrocyte co-cultures • In vitro electrophysiology (ion channels; neurones; slices) IN VITRO IN VIVO POST MORTEM • Behavioural/neurological • Neurophysiological recordings (EEG; ERG; EMG; BAER; nerve conduction velocity) • Neurochemical (in vivo microdialysis; biomarkers) • Neuroimaging (MRI; MRS; PET; SPECT) • Neurohistopathology/ immunohistochemistry Redfern WS & Wakefield ID (2006) Safety Pharmacology. In Toxicological Testing Handbook: Principles, Applications and Data Interpretation, 2nd Edn., pp. 33-78, D Jacobson-Kram & K Keller (eds.). New York: Informa Healthcare.

17. Studying adverse effects of drugs on the nervous system • Neuronal-astrocyte co-cultures • In vitro electrophysiology (ion channels; neurones; slices) IN VITRO IN VIVO POST MORTEM • Behavioural/neurological • Neurophysiological recordings (EEG; ERG; EMG; BAER; nerve conduction velocity) • Neurochemical (in vivo microdialysis; biomarkers) • Neuroimaging (MRI; MRS; PET; SPECT) • Neurohistopathology/ immunohistochemistry Redfern WS & Wakefield ID (2006) Safety Pharmacology. In Toxicological Testing Handbook: Principles, Applications and Data Interpretation, 2nd Edn., pp. 33-78, D Jacobson-Kram & K Keller (eds.). New York: Informa Healthcare.

18. Global nervous system safety assessment: The Irwin test / Functional Observatory Battery • A manual, multi-parameter assessment of nervous system function in rodents • Involves observations in the home cage and in an arena, as well as manual interaction • Limited to ‘snapshot’ observations at specified time points pre- and post-dose

19. Plus: any miscellaneous observations; body weight gain overnight post-dose Redfern WS et al. (2005) J Pharmacol Toxicol Methods 52: 77-82 Ewart L et al. (2013) J Pharmacol. Toxicol. Methods 68: 30-43. • posture • gait • straub tail • body tone • ptosis • exophthalmos • grip strength • traction response • tremor • twitches • convulsions Neurological Autonomic • salivation • acrimation • piloerection • abnormal urination • abnormal defaecation • abnormal respiration • pupil size • rectal temp • touch response • palpebral reflex • startle reflex • pinna reflex • righting reflex Sensorimotor Behavioural • arousal • activity • vocalisation • aggressiveness • Sniffing • Stereotypy • rearing • bizarre behaviour

20. Locomotor activity Automated measurement of ambulatory activity and rearing in a novel arena, usually over 30-60 min. Habituation to the novelty of the arena occurs during the monitoring period, and upon re-testing.

21. • Methodology - 3 dimensional matrix of infrared beams OR videotracking - Beam breaks OR videotracking used to quantify movement - Rats have to be single-housed during the measurement period • Activity in a novel arena - Novel environment to measure spontaneous locomotor activity - Two phases of locomotion: initial exploratory phase (duration dependent on rat strain/age/conditions) followed by habituation phase • Home cage activity - Home cage: avoids novel environment – no anxiety component - 24 hour continuous monitoring/circadian rhythms - Higher activity during dark phase/darkness Redfern WS & Wakefield ID (2006) Safety Pharmacology. In Toxicological Testing Handbook: Principles, Applications and Data Interpretation, 2nd Edn., pp. 33-78, D Jacobson- Kram & K Keller (eds.). New York: Informa Healthcare. Locomotion Activity

22. Spontaneous Locomotor Activity Spontaneous locomotor activity in a novel arena provides 2 levels of baseline activity: initial high exploratory-related locomotor activity followed by low baseline activity due to habituation. Novelty Phase - 0 to 10 min of assessment Habituation Phase -10 to 30 min of assessment 0 5 10 0 (0-5) (5-10) (10-15) (15-20) (20-25) (25-30) TotalDistanceMoved(m) TestTrials(time from startof assessmentinmin) Spontaneous Locomotor Activity Profile Control Habituation Phase Novelty Phase Asymptotic Level Chu A et al. (1999) Proc US Gen Safety Pharmacol Soc

23. Spontaneous Locomotor Activity Profiles 0 5 10 0 (0-5) (5-10) (10-15) (15-20) (20-25) (25-30) Test Trials (time from start of assessment in min) TotalDistanceMoved(m) Control Sedative Habituation Phase Novelty Phase Asymptotic Level Spontaneous Locomotor Activity Chu A et al. (1999) Proc US Gen Safety Pharmacol Soc Novelty Phase high baseline activity in this phase is suitable for the detection of sedative effects

24. Spontaneous Locomotor Activity Profiles 0 5 10 0 (0-5) (5-10) (10-15) (15-20) (20-25) (25-30) Test Trials (time from start of assessment in min) TotalDistanceMoved(m) Control Stimulant Sedative Habituation Phase Novelty Phase Asymptotic Level Spontaneous Locomotor Activity Novelty Phase high baseline activity in this phase is suitable for the detection of sedative effects Habituation Phase drug-induced activity stimulation becomes readily detectable when baseline activity is low Chu A et al. (1999) Proc US Gen Safety Pharmacol Soc

25. So, is it time for new technology?

26. What if...? • You could monitor the ambulatory activity of each individual rat within a group in a standard home cage over 24 hours, or longer...? • Also monitor their temperature...? • Achieve this without surgery...? • Detect convulsions and other ‘abnormal’ behaviours...? • Achieve all this without having to modify the home cage...? • Do this in a standard IVC cage rack, which you could wheel anywhere...? But how...?

27. SPONSOR submits a technological challenge NC3Rs invites innovators to enter the competition INNOVATORS submit proposed solutions EXPERT PANEL selects winning solutions(s) NC3Rs fund development project INNOVATOR, SPONSOR AND NC3Rs deliver the technological solution https://www.crackit.org.uk/ CRACK IT is a funding competition (Challenges) and technology partnering hub (Solutions) designed to accelerate the development, application and commercialisation of technologies with 3Rs potential as they emerge from the research base. CRACK IT has been developed to facilitate active collaboration between the pharmaceutical, chemical and consumer products industries, contract research organisations, small and medium enterprises (SMEs) and the academic sector.

28. • Challenge set by AZ (Will Redfern) in 2011 • Remit was to monitor activity, behaviour and temperature of individual rats when group- housed in standard home cages, 24/7, for up to a month • Winning solution awarded to Actual Analytics (Edinburgh) • Project got underway in 2012 • AZ’s in-kind contribution has been intellectual input and annotation of video to train the behavioural recognition software • Aims also to include detection of convulsions Rodent Big Brother Project Automated home cage behavioural monitoring system funded by NC3Rs CRACK IT scheme

29. Rodent Big Brother Project Automated home cage behavioural monitoring system funded by NC3Rs CRACK IT scheme • Positional information and temperature via a subcutaneous RFID microchip, detected by a baseplate reader under the cage • Behaviour captured via a high-res camera using IR lighting, analysed automatically by behavioural recognition software • A prototype system was installed at AstraZeneca Alderley Park (UK) in 2014 • We are undertaking a full road-testing, evaluation and pharmacological validation • Intention is to incorporate into early investigative toxicology studies in rats

30. Features of the home cage 24 h monitoring system RFID chip on fingertip... ...injected subcutaneously

31. Infrared lighting panel IVC home cage Baseplate RFID chip reader (under home cage) Side-view video camera Mini-computer and power supply (Part of) cage rack Features of the home cage 24 h monitoring system VIDEO CAPTURE • High quality video capture for manual analysis by expert • Automated analysis of common behaviours by behavioural recognition software • Additional behaviours can be annotated for training the software

32. Infrared lighting panel IVC home cage Baseplate RFID chip reader (under home cage) Side-view video camera Mini-computer and power supply (Part of) cage rack Features of the home cage 24 h monitoring system BASEPLATE READER • Automated acquisition of ambulatory activity • Automated acquisition of subcutaneous temperature • RFID data used to ‘ID tag’ each animal in the video

33. Ongoing video annotation and validation work Manual Annotation of Video • Capture of light-dark phase video from different individual animals • Careful manual annotation of individual behaviours to train behavioural recognition software Mechanical Update – Camera Stand-off

34. • Do all modules capture continuous 24/7 baseplate and video data without drop- out/crashes? • Are 24 h activity and temperature data equivalent between modules? • Do X-Y data from baseplate tally with manual X-Y data using ‘bird’s eye view’ camera? • Do activity and subcutaneous temperature show a 24 h periodicity? • Do peaks in baseplate-derived activity data tally with peaks in automated motion detection? • Is temperature data accurate? • Can the system detect changes in temperature induced by non-pharmacological means (eg, single-housing rats)? • Can the system detect changes in activity induced by non-pharmacological means (eg, cage change)? • Can the system detect pharmacologically- induced changes in activity and temperature? • Do software-derived behavioural data tally with manual video analysis? User Acceptance Testing & Validation

35. Manual analysis of behaviours in a single rat over 22 h Performed as part of behavioural annotation of video Top 10 Most Common Behaviours Measured Behaviour Number of events Scratching 593 Rearing 579 Walking 475 Chewing hind paw 334 Licking/chewing coat 298 Immobile 266 Face washing 205 Eating from forepaws 203 Feeding from hopper 190 Drinking 177 • These annotated episodes are being used to train the behavioural recognition software. Others will be added when sufficient episodes have been captured on video and annotated. • Episodes of convulsions will be captured from an ongoing epileptic rat model (ie, no additional animals used).

36. Manual analysis of behaviours in a single rat over 22 h Performed as part of behavioural annotation of video

37. Actigram plot of a single animal over 2 consecutive days A circadian cycle is evident; work ongoing to verify accuracy of above data manually.

38. Group mean temperature and activity data over 3 days • Data derived via the baseplate RFID chip readers over 3 consecutive days • Six Rats housed in two cages (3 rats per cage)

39. Periodogram showing peak in frequency distribution of activity at a 24 h frequency Differences between dark phase and light phase group mean activity for 3 baseplates (n = 5 rats). Baseplates 24 h Periodicity of activity data

40. 24 h Periodicity of activity data Periodogram showing peak in frequency distribution of activity at a 24 h frequency Differences between dark phase and light phase group mean activity for 3 baseplates (n = 5 rats). Baseplates Filter algorithms currently being refined to minimise registering of micromovements between adjacent antennae in baseplate during periods of low ambulatory activity.

41. Effects of single-housing on subcutaneous temperature • Decrease in subcutaneous temperature immediately upon housing singly after being housed in groups of 3 • Possible causes: group housing or rats enables intermittent ‘huddling’ with two cage mates, and may achieve a higher ambient temperature (with two additional rats generating heat) • Illustrates just one of several physiological stressors associated with single housing (not to mention the psychological stressors). Single vs. Group Housed Rats

42. Advantages This is a technological breakthrough. This data collection wasn’t possible before now...  Enables continuous 24 h monitoring over days and weeks  Increases the information content of existing study types  Animals are housed in social groups  Non-invasive (other than a subcutaneously injected RFID microchip)  No modifications to standard housing cages required  Modules fit inside standard IVC cage rack, so do not require a dedicated room  Can be incorporated into existing study types (without the use of additional animals)  Future developments include automated detection of convulsive behaviours

43. Potential applications Preclinical Safety Assessment • The original intent for this technology. • Can slot into existing study types: safety pharmacology studies and repeat-dose toxicology studies • Will detect the previously undetectable: 24 h activity, 24 h temperature, 24 h behavioural analysis, and the occurrence of convulsions outside normal manual monitoring hours • Could identify ‘at risk’ individuals to back-up/pre- warn welfare decision-making Drug withdrawal phenomena • Different classes of drugs with dependence potential cause different patterns of effects on cessation of treatment. However, these usually include changes in activity, behaviour, temperature and food intake. CNS Drug Discovery • Can slot into existing disease models • Will detect 24 h activity, 24 h temperature, 24 h behavioural analysis, and the occurrence of convulsions outside normal manual monitoring hours Academic Research: Disease Models • Various applications -- could identify ‘at risk’ individuals to back-up welfare decision-making Academic Research: Circadian Rhythms, Ageing • Various applications Academic Research: Phenotyping • New CRISPR technology opens up the availability of transgenic rat strains

44. Acknowledgements AstraZeneca Alderley Park, Cheshire, UK Karen Tse Claire Grant Dave Simpson Liz Beard Karen Sefton Lauren Leslie (1-year placement student, University of Glasgow) Victoria Rimmer (1-year placement student, University of Manchester) NC3Rs London, UK Kathryn Chapman Cathy Vickers Vicky Robinson Publications to date: Redfern WS, Armstrong JD, Heward J, Allison B, Lukins T, Grant C, Leslie L, Craig DJ, Vickers C, Chapman K. (2014) Rodent Big Brother: Development and validation of a home cage automated behavioural monitoring system for use in repeat-dose toxicity studies in rats. Eurotox 50th Annual Congress, Edinburgh, UK. Leslie L, Armstrong JD, Heward J, Allison B, Lukins T, Sillitto, R, Grant C, Craig DJ, Vickers C, Chapman K, Redfern WS. (2014) Rodent Big Brother: Development and validation of a home cage automated behavioural monitoring system for use in safety pharmacology studies in rats. Safety Pharmacology Society 14th Annual Meeting, Washington, DC, USA. Actual Analytics Edinburgh, UK Douglas Armstrong David Craig Tim Lukins James Heward Rowland Sillitto Agis Chartsias Emma O’Callaghan University of Strathclyde Glasgow, UK Judith Pratt

46. MRC Harwell Genetics of Disease Sponsors Seeking refinement in the characterisations of mouse models NC3Rs New technologies to support refinement Project Funders Actual Analytics Technology Specialists Developers Using new technologies to solve difficult problems

47.  Dynamic and progressive research facility  Extensive resources for mouse research  Provide a national and international lead in laboratory animal science MRC Harwell : An International Centre for Mouse Genetics Commitment to Reduction, Refinement and Reproducibility

48.  Open in 2004 -- Capacity for 55k mice at any one time  Significant molecular biology and pathology capabilities  Currently running efficiently at full capacity MRC Harwell : Facts and Figures

49. In 2014 alone:  230k regulated procedures  125 transgenic models  321 lines exported MRC Harwell : Facts and Figures Science-led Service Delivery

50. Building a comprehensive functional catalogue of a mammalian genome The International Mouse Phenotyping Consortium (IMPC) www.mousephenotype.org

51. IMPC – The Context • The function of the majority of genes in the mouse (and human) genomes is unknown • KOs have been generated and analysed in about 35% of mouse genes • Data for these genes is patchy – dependent on the interests and experience of the investigator. There is an increased attention on reproducibility and reliability. • Develop approaches for broad based phenotyping, to provide a comprehensive picture of disease states and to integrate with human and clinical genetics. • IRDiRC (rare disease); Biobank;100,000 genomes; MRC Mouse Network

52. Neuroscience and Medical Research 8,000,000+ people living in the UK with a neurobehavioural condition 1,000,000+ people who are disabled because of their condition 350,000+ people require help for most of their daily activities (Neurological Alliance)

53. Using Mice to Research Genes Mice are widely used in genetic research > 95% of mouse genes are similar to humans Mouse physiology and anatomy is similar to humans

54. How Does One Assess Well-Being?

55.  Digging  Climbing  Nesting Normal behaviours... How Does One Assess Well-Being?

56. How can you measure signs of neurological disease?

57. How can you measure signs of neurological disease?  Hyperactivity  Social Isolation Abnormal behaviours...  Disrupted Sleep

58. Measuring Mouse Behaviour The challenge! It changes during the day! They are active during the night! Day Night Day Night

59. Mice like other mice and not new places! Activity can be measured by wheel-running or video tracking AnxietyHoused Alone The environment of the test can be stressful

60. The Challenge!!! 1. Monitor mouse behaviour 24 hours a day 2. In their home-cages 3. With their cage-mates The Goal: 1. Refine tests by reducing stress factors 2. Gain vital scientific information 3. Record more data from less mice CRACKIT Program...

61. Rodent Little Brother – monitoring groups of mice without them knowing • High resolution cameras and advanced computer processing

62. Rodent Little Brother – monitoring groups of mice without them knowing • High resolution cameras and advanced computer processing • RFID chips monitored by base plates for days and weeks

63. Measuring Signs of Neurological Disease • Social isolation • Hyperactivity/ hypoactivity • Sleep disturbances

64. C3H H males 4 week old (n=9; 3 X 3 boxes) 3 day data.AverageDistance(mm)

65. Monitoring Complex Behaviours The Challenge! • Need to automate the system • Requires computer recognition of specific behaviours • We need to train the software first Video Capture Annotation Machine Learning

66. Future Aims and Targets Analyse complex behaviours • Grooming • Play • Mating • Social interactions Gather much data from individual mice • Novel behaviours • No environmental factors • Unknown to the mouse

67. Thank You! For additional information on the ActualHCA 24x7 automated home cage monitoring system, and other video tracking solutions for behavioural research please visit: http://www.actualanalytics.com/

68. InsideScientific is an online educational environment designed for life science researchers. Our goal is to aid in the sharing and distribution of scientific information regarding innovative technologies, protocols, research tools and laboratory services.