Fluorescence Measurements Over Time

[Nuha]

I’m measuring GFP and mScarlet fluorescence over time but notice strange graphical behavior: When O.D and GFP/mScar increase, the real-time fluorescence graph on the GUI has a decreasing slope, instead of an increasing one. I figured this may be due to the baseband normalisation, so I went on the csv file and multiplied the baseband values to the normalised emission values. Only then the graphs started to make sense (As O.D/GFP/mSCAR increase, the slope increases over time).
I wonder if this is a present issue? And if what I’m doing is correct, as in are these real fluorescence values I’m seeing when I do the multiplication?

My understanding is that baseband normalisation is normalising by noise, and I’m guessing when taken overtime the noise increases negatively affecting the real fluorescence measurements. It would be nice to have an option where I can get rid of the real-time normalisation if what I’m doing is correct. Appreciate any feedback/ suggestion! I’m also using LB as my media for the time being.

[harrison]

Hello,
Baseband normalisation is essentially calibrating for the excitation intensity as well as OD. Therefore, the baseband-normalised data should stay approximately constant if one was to hold fluorescence fixed and change OD.

If you are allowing BOTH OD and GFP/mScar to change at the same time, then you MIGHT see a reduction in fluroescence essentially because the fluorescence/cell is decreasing. This often happens if you have a system in which it has grown overnight (i.e. to high fluoprescence stationary state) then you set it going back to exponential phase in the morning (i.e. diluted into fresh media), after which the cells which previously had time to achieve massively high fluorescence/cell are now having this diluted out.

I would say in general by far the best results come from keeping OD constant (i.e. in turbidostat mode) then looking at changes in fluorescence.

You could adjust the software to show you non-normalised data if you like. Have you taken a look at the section that discusses normalisation in the Supplementary of the paper?

One other thing – if you use LB you will get generally a TERRIBLE signal (i.e. in terms of signal-noise ratio) due to the fact that it is very auto-fluorescent and generally also has large variability batch-to-batch and within a batch over time. Potentially unless you have VERY fluorescent cells this kind of chaos will completely obscure the actual fluorescence changes youw ant to measure. I would strongly recommend using M9 media at least…

[Nuha]

I just replied but don’t seem to see it. I hope you received it?

[Nuha]

Apologies for getting back late to you, there was a delay in running the experiments and we recently received the latest chibios. I have switched to M9 media, thanks for the feedback.I’m also running the experiment starting at very low OD, and have the OD regulation set at a target OD 0.5; so I’m not diluting. Here is the interesting observation:

I noticed that for low GFP/RFP induction, the normalized fluorescence over time starts at a higher GFP value (OD 0.05). When the cells approach OD 0.4, GFP fluorescence starts dropping until it reaches the steady state as the OD is set at 0.5.

This behavior is not seen for medium and high GFP induction. Both medium and high normalized fluorescence values show the expected increase in slope from OD 0.05 to OD 0.5, and reaches steady state when the target OD is set at 0.5.

When I remove the normalization, low GFP induction is corrected and shows the expected graph like the medium and high GFP inductions. I have attached the graphs for your reference:

https://drive.google.com/file/d/1VpzAD–2-R3vQupKj3H7jrQUFTmgvTE3/view?usp=sharing

https://drive.google.com/file/d/1yTL252cIFNBfMEs7_12JxM5Qp2kXAhoG/view?usp=sharing

[Nuha]

The link doesn’t seem to be working for low induction. Here it is again:
https://drive.google.com/file/d/1VpzAD–2-R3vQupKj3H7jrQUFTmgvTE3/view?usp=sharing

[harrison]

One thing I often observe is a fluorescence DECREASE at the start of experiments if inoculating with cells that have come for stationary phase, or even from some other condition where fluorescence levels might be high. Even for promoters that are theoretically uninduced, often at stationary phase there is a lot of leakage, and if they haven’t been diluted this gives lots of time for them to accumulate fluorescence. Then, when you put them into a reactor (or plate reader) to grow you see fluorescence decrease initially as this previously built up level is diluted. This is less obvious if you are doing high levels of induction since the (desired) production is greater compared to the leak production.
DO you think this might explain what you are observing? Potentially I misinterpreted your description…

[Nuha]

The decrease in fluorescence might make sense if we are inoculating cells at relatively high OD – closer to the target OD. But in my case I even started experiments directly from a colony on a plate and set the chibio to regulate it at OD 0.5, once it reaches that point. So there is no chance for cells to be at stationary – the cells are always in exponential phase.

I think it’s best if you view the 2 images I shared. I can see the ‘low induction’ link has been changed by the forum to 1 dash instead of 2 dashes between AD and 2, that is why it is not opening. Here it is in link format (I hope that works, else please just type it):

If you compare medium GFP induction with low GFP induction images I hope you’ll understand what I mean. I think the normalization is not sensitive enough at low fluorescence inductions and may potentially bias fluorescence measurements.

[harrison]

Hello,

Yes, by adding that dash I can now see it.
I think, contrasting your previous post, to me it looks like the EXACT same thing is happening in both the low and medium inductions for early times (i.e. until you start regulating OD). They behave basically the same (i.e. initial decrease), but the medium one is at least induced enough so that you can see it rise (whereas the low one is not enough for this) once they reach fixed OD.

Another possible explanation could be other non-linearities in the normalisation. For example that light scattering properties at the two wavelengths change as the OD changes, hence it appears to go down (slightly) over time. In general, the system won’t be great if measuring either very low fluorescence levels per cells, or measuring fluorescence at very low cell densities, since in both cases the bleed-through of the excitation to emission filters is drowning out the signal.

Generally our team does not find this a problem since usually what we do is grow cells to a fixed OD and hold them there for at least 5-10 generations (so that any dependence on previous culturing or state is mitigated), and THEN add inducer chemicals (or light, whatever). Is there a reason why you are not doing it this way? Trying to separate out the behaviour of genetic circuits from the impact of changing cell density is always challenging – hence why we make turbidostats in the first place.

[Author]

Posts