Validation of the Hanabi-PIII Robotic Metaphase Cell Harvester

Validation of the Hanabi-PIII Robotic Metaphase Cell Harvester Juliana Groisman, Richard Hall ACC Spring Conference Liverpool, March 2008

Overview § Introduction and context § Hanabi-PIII Robotic Harvester § Validation, data analysis and preparations § Considerations and limitations § Oncology trial § Conclusions and further work

Introduction § Manual harvesting is physically demanding, time consuming and prone to operator variability. § Automation of this process should increase efficiency and reduce manual routine work, leaving experienced staff free to concentrate on other important tasks. § Robotic harvesters have been around for a number of years; however, there has been relatively little use of these machines in cytogenetics labs in the UK. - There is little evidence available to prospective laboratories regarding the effectiveness of these machines; - Machines have not been fully automated and require operator interaction.

§ Hanabi harvesters were the first harvesters to offer complete automation of the process. The Hanabi PII harvesters, which have a capacity of 24 samples, have been used for some years in Europe, USA and Japan. § We had the opportunity to validate the first Hanabi PIII robotic harvester, which has a sample capacity of 64 tubes. § For a period of two weeks the Hanabi-PIII robotic harvester was trialled in our laboratory. § Our main aims were to evaluate and compare the preparations obtained with the automated Hanabi-PIII robotic harvester against our manual harvest of blood lymphocytes.

Criteria assessed The robotic harvester should: Produce preparations suitable for diagnostic use; Increase sample throughput; Be efficient and require no intervention; Show high reproducibility - removing operator variability; § Reduce manual routine work § § § Be robust and easy to use

Hanabi-PIII Robotic Harvester § The Hanabi-PIII automated suspension harvester (ADStec – ADScience Technologies, Japan) is unique in its mode of action and sample throughput capability: – It includes a centrifuge and a vortex for agitation – It is capable of harvesting 64 specimens simultaneously – Once the specimens have been loaded no operator interaction is required until the harvest is complete

HEPA filter Touch screen menu

Validation Technical aspects Setting the programme using the touch-screen menu was easy; to start with, we changed the settings to suit our current protocol. Good results obtained with the first run Slide making

Validation Technical aspects We carried out experimental runs to optimise the settings • Pre-fixation step (G step) • Second hypotonic step • Different lengths of exposure to hypotonic • • • solution Different vortex speeds Changed the centrifuge times Changed aspiration levels (which determine how much supernatant should aspirated) After assessing how much supernatant Changing the aspiration level to 12 was left in our manual harvest, we had a When we tried aspiration level 10, we positive effect, preparations but started at aspiration level 14 were clean found that preparations were slightly and mitotic our yields were affected found that index was good cytoplasmic

Validation Technical aspects Our settings for the validation: A. Centrifuge for 5 mins (1, 000 rpm); B. Aspirate to level 12 C. Start the vortex; add 5. 5 ml KCl (hypotonic solution); D. Incubate for 6 mins; H. Centrifuge for 5 mins (1, 000 rpm); I. Aspirate the supernatant to level 12; J. Start the vortex; add 5 ml fixative K. slidemaking 3 X Centrifuge for 5 mins (1, 000 rpm); L. Harvest time was comparable to our manual harvest

1. 2. 3. 4. 5. Preparation takes approx 5 mins to prepare and 2 -10 mins to load depending on sample size 6. 7. Switch on the harvester Prepare fix and hypotonic solutions, top-up distilled water Carry out reagent exchange – flushes the system replacing water with reagents (fixative and hypotonic) Select saved protocol (at this point parameters can be changed if required) Set sample number (the number of tubes to be loaded). Dummy tubes containing water must be used to balance the centrifuge. Take samples to Class II hood and add colcemid; start the timer. Load the samples. The air temperature inside the robotic harvester can be set at 37°C – samples can be incubated in colcemid whilst inside the harvester. The lid is removed (and discarded) and each sample is placed in its allocated slot – sensors detect that loading is carried out correctly.

8. Harvest takes 50 mins -1 h 58 depending on sample size 9. 10. 11. Washing takes approx 10 mins 12. 13. 14. When the time in colcemid is up (15 minutes at Guy’s), press main start button and walk away. When the harvest is finished, remove tubes and use fresh lids. Samples are ready for slidemaking. Carry out reagent exchange – flushes the system with water Carry out tube washing (soaking of aspiration probes for 3 mins), remove tubes and allow pump to dry (~5 mins) Swab surfaces with virkon Empty waste containers Switch off the harvester

Quality Assurance § There is no sample transfer; samples are harvested in the same tubes used for culturing § For consistency, hypotonic solution is warmed up to 37˚C during injection § The program ensures that each step is carried out correctly before proceeding to the following step - Reagent exchange must be carried out before loading - After adding sample size, loading positions are indicated. Sensors detect that samples have been loaded in their correct positions to ensure that the centrifuge is balanced - Only then it is possible to start the harvest

Validation 1. 53 patients were processed for the validation. 2. For each patient, test and control cultures were set up and processed exactly the same apart from the harvest; 3. Blind study – All slides were coded prior to data collection; 4. Slides were scanned on Metasystems and the 30 best metaphases were selected; 5. The following parameters were assessed: Quantitative Yield Total number of cells obtained Mitotic index Calculated by Metasystems for the whole slide Qualitative Chromosome length Spreading A total of 3, 000 metaphases were assessed

Results

Validation results Data analysis Comparable Yield Found to be slightly higher in the robotic harvester; Comparable Mitotic Index Found to be lower with the robotic harvester but more consistent; Quality Possibly more consistent with the robotic harvester Comparable Spreading Slightly lower with the robotic harvester but possibly less variable Comparable

Preparations

Specimen 07/12308 Robotic harvester Manual harvest

Specimen 07/12304 Robotic harvester Manual harvest

Operator variability § In our validation we found that some manual harvests were better than others; § Harvest 4 (n=13, operator 2) in the validation was a poor manual harvest, with some broken and cytoplasmic preparations; cultures harvested with the robotic harvester were considerably better. § This is likely to be due to operator error, which is one of the factors we aim to remove using the robotic harvester.

Operator variability Poor Manual harvest (harvest 4) Manual harvest Robotic harvester Specimen 07/12516

Considerations/Limitations § Culture tubes Conical tubes made of polystyrene are not suitable for the robotic harvester. Polypropylene tubes (15 ml) are needed. § We tested polypropylene BD Falcon tubes for toxicity and found that they had no adverse effect on our cultures and Falcon tubes turned out to be cheaper than our current culture tubes; § The lids are slightly bigger than our current lids, although they still fit into our racks and centrifuge buckets; they are not as transparent our current polystyrene tubes. § Lids must be removed and stored during the harvest; an additional mechanism needs to be in place to return lids to the correct tubes; furthermore any material still on the lid would not be fixed and may affect the quality of preparations. We decided to discard the lids and use fresh lids after the harvest. Using a fresh lid currently means using two tubes per culture - we have been in contact with the manufacturer to purchase packs of lids.

Considerations/Limitations § Pre-fixation § The harvester can incorporate a G step, or pre-fixation step, for the addition of small volumes of fixative whilst the sample is in hypotonic. § The G step must be applied to the whole run; it cannot be limited to certain sample types (e. g. newborn babies). § Centrifuge r. p. m. is not readily adjustable § It can be changed by one of the engineers. We did not need to change the setting; we found that 1, 000 r. p. m. was a suitable setting. § Cannot use a third reagent § We currently use 5% acetic acid for one of our two cultures; it is not possible to use a third reagent with the harvester.

Considerations/Limitations § Underspreading and “clumping” with current settings § Some cultures on the robotic harvester had a tendency to be underspread and require more effort during slide preparation; further optimisation of the protocol might improve spreading § Vortex settings §The speed of vortex is fully adjustable; § However it can only be set at two different speeds per harvest. There are two vortex speeds

Oncology trial § The oncology team have not yet carried out a validation trial; they only processed a few samples. § Comments from our oncology team: – Initial results indicate that Hanabi preparations for standard oncology bone marrow samples are equivalent in quality to cultures harvested manually; – Aspiration level cannot be changed throughout the harvest. Tests so far suggest that it may not be possible to include bone marrows with low counts in a routine Hanabi harvest as these generally have large pellets following centrifugation, which would require individual adjustment of aspiration volumes; – There has been no evidence that culturing bone marrows in the Falcon tubes is detrimental.

Fulfilling the criteria The robotic harvester should: - Produce preparations suitable for diagnostic use; – Increase sample throughput; – Be efficient and require no intervention during harvest; – Show high reproducibility - no operator variability; • 24 x 2 operators 64 cultures • Takes ~1 h 58 m to process 64 cultures; • Easy to load, approx 15 -20 mins hands-on time for preparation, tube loading and cleaning of the robot • Validation results confirm that preparations obtained with the harvester are more consistent, and for MI this difference is statistically significant (p =0. 0004) - Be robust and easy to use • There are protocols available for recovering in case of malfunction

Conclusions from the Validation § Our trial only lasted two weeks in October 2007 - there is still scope to optimise the robotic harvester settings and further improve the quality of preparations. § The preparations obtained, even prior to optimizing the protocols, are comparable to our routine preparations and perfectly suitable for diagnostic use. § The robotic harvester is user-friendly, simple to programme, straightforward to load and requires no supervision or intervention.

Since then… § We decided to purchase the harvester and successfully applied for a grant from the Evelina Appeal Charity. § We have been using the robot routinely for one of our two cultures since January; to date, we have carried out over 50 harvests using the Hanabi PIII and harvested over 600 samples. § Chromosome preparations are consistently good quality and often significantly better than our second cultures harvested manually. § We aim to fully automate the harvest of blood lymphocytes in the next few months. § Settings have not been changed since the validation; we will be experimenting with the settings to improve preparations further.

Acknowledgements Guy’s Anne Bergbaum Ian Kesterton Sally Walsh Paul Stevens Shiply Begum Sara Kadir Kamal Uddin Liz Allan Helen Geoghegan Paul Scriven Shehla Mohammed Transgenomic Ben Nouri Paul Hornsby