Hi! I am an undergrad looking to automate maturation media changes for floating organoids. I am looking for advice from folks with experience in this or lab automation enthusiasts more generally.
As the organoids float freely and move around the well, every media change risks aspirating the organoid along with the spent media, which I cannot allow.
Has anyone reliably automated this media change? How did you do this? And if you haven’t but have ideas, those are very welcome too. Thank you!
Have you tried gently centrifuging down to a pellet? Then you can aspirate most of the media off and add fresh media.
Presumably you’re automating a protocol that someone has already done manually. What does the manual protocol say?
That depends a little bit on what you have available. There are dedicated modules for tilting your labware, which are perfect for your kind of application. You can tilt the labware and aspirate from a dedicated position that is still save for the organoid. This is the recommended procedure for organoids.
If you don’t have that available, I would recommend an off-center aspiration with a save fixed height. You will have to determine what is save experimentally. Depending on the size of your organoids you can also work with pressure detection, as they may clog up your tip. Sadly, that is more probable with small tip sizes, which take longer to aspirate the supernatant.
This type of media exchanges can be done with liquid handlers; the key is to aspirate at a slow rate. An alternative from Agilent is the MultiFlo AMX Automated Media Exchange module. The main downside of this is that there is no way to clean the probes between wells or plates.
@Stefan, I have considered this as I am centrifuging to remove supernatant earlier in the protocol already. The issue with this is that my organoids are in flat-bottom plates, which means I would need to transfer them into a V-bottom or conical and then centrifuge, which I believe would cause mechanical damage and risks errors in automation over time. I should be able to test this down the line, however. I could just use V-bottom plates for maturation, but this is worse for oxygenation as the organoid would be then sitting below a deep amount of media.
The protocol is a media change by hand, so someone manually takes care to not aspirate the organoid.
@PatrickM, the tilt module sounds quite good and I am reading about it now. This seems preferable for its simplicity. Is there any downside to this option (QInstruments TiltStation specifically)?
@andewjmash, the risk with this is that even with slow aspiration, the liquid handler does not know where the organoid is. Doing a fixed XY/Z can be fine for one well and right on top of the organoid in the next (which, even if not causing death of the organoid, would be a confound I’m sure). Even 1% loss per feed across ~40 feeds I will be doing is only ~67% survival which would make the project too expensive for me. To keep loss from this step under 5%, I need per-feed loss under 0.13%, which seems difficult with just slow aspiration. But this just my thoughts, do you have experience with slow aspiration working?
Not sure if I’m missing something so feel free to correct, and thank you for the great responses.
I think it’s worth explaining exactly how the manual operator does it. How do they “take care” not to aspirate the organoid? If you can logically break down exactly what algorithm the operator is running to not aspirate the organoid, then you are most of the way towards having a robot do the same thing.
That make sense. I realized I can do U-bottom plates.
I do wonder if there is a simpler way to do this when freeing ourselves from the assumptions that come up with human operation. Anyhow, this seems pretty straightforward… the computer vision doesn’t worry me. I am sure I will have to add a QC step for edge cases, but this seems like it might work.
Does this seem like the most reliable way to do this? I am willing to trade reliability for complexity (ex. using computer vision adds complexity, but I am willing to do it for reliability over many organoids). Would you opt for another approach for reliability? Thank a lot.
So these organoids are clearly visible and the operator is supposed avoid them? That would be pretty challenging to automate, but not impossible. You would likely need a very high res camera and very good control over the focal height, shooting from below.
Maybe more trouble than it’s worth. In theory the organoids are denser than the media and should remain settled at the bottom unless disturbed. A very slow aspiration speed might get you there. If your media isn’t too expensive (not a given if your differentiating organoids), you could do a 2x partial media exchange. Two 80% exchanges gets you the equivalent of a 96% exchange. Would that be good enough?
One note with the U bottom wells is that the benefit you will get from them (concentrating them at the bottom) can also hurt you. A bunch of organoids at the bottom of the well can end up fighting for nutrients and impacting growth and phenotype.
The only downside I see is that this is not platform-native. I don’t see you mentioning which automation you use, but I assume any of these will have rheir own version, which will be easier to implement. Though I highly doubt any.company offering a tilt module will leave you hanging in regards how to implement it on a given platform. Stefan’s approach is also fine, mind you, and saves on both deck space as well as tech ology to buy. I would recommend first trying to solve this with simple means and then get more complex until the results are satisfactory.
It seems that the next step is to take these few approaches to the lab and experiment to see what works well. I’ll post an update here then 
Interesting prroblem and here is some maybe overengineered ideas:
[edit - saw this was also your take in your earlier post so sorry for the regurgitation]
We have some Hamilton robots equipped with an easypick camera where we have custom software for image recognition and creation of picking coordinates (for sequences). The easypickII software itself is really not optimised for the things we do (thresholding, image processing and selection criterias/normalization are all quite flawed as soon as you are not picking nice little round colonies). We basically subtract a background image of a “blank” plate from our “sample” image - this allows us to threshold even things like watery sodium alginate pearls. we use it for selecting what we want but you could also in theory have a script thresshold the “background” and select the furthest point from your organoid. with a relative good camera and some quick vibe coding you could make a prof of concept for the thresholding and coordinate calculations. This of course is depending on the organoids staying relatively in place, and good control of your light conditions.
for Hamilton(or equivalent) I would prob combine it with tadm(if that itself is not sufficient) - create curves for pipetting the organoid vs pipetting the media with a liquid class that trigger a difference if possible and then have error handling if pipetting of an organoid is detected?
it also seems like there is a fair bit of transwell labware designed to isolate the cells from pipetting operations having dedicated pr. well pipetting positions but i do not have experience with those. (like millicell-microwell-plates-organoid-culture-tb11972en-mk.pdf)
/3 cents.
I will try the approach of subtracting a blank, this seems like it work very consistently and is a very great idea. Yes, I think the liquid handling is very feasible once I know where to pipette. I am familiar with these hydrogel plates but there are a number of predictive validity tradeoffs in using those. Thank you.