I sincerely appreciate the openness. Providing the methods, data, interpretations, and next steps is a great way to lean into scientific communication! Thanks for the post
This is a very nice write-up! Thanks for sharing your work. I have a few questions:
1) When cloning the DNA+barcode into the E.Coli strains, is it always placed in the same location?
2) You show good correlation between replicates across your time course, I was wondering if any genes showed positive correlation across time points. Particularly your high fitness effect barcodes at day 5.
3) You averaged your fitness per position, but I wonder if you analyzed data after collating positions/inserts by operons you would have more consistent results. Particularly when using two very different microbes where you may miss interesting things if the full operon isn't included in the DNA insert (thus making any one part non-functional)
1) in this dataset the cloning was actually done into plasmids, so technically yes, but it's not in the genome. Going forwards we're working with random genome integrations.
2) Definitely, there's correlation across time, but asymmetrically. Things that win at day 5 are high fitness at all timepoints -- but things that are high fitness at day 1 aren't necessarily still high fitness at day 5. This is an advantage of sequencing, it reveals high fitness species early on that may not be discovered experimentally via fixation!
3) This is something we're looking into -- examining the data from a more agnostic viewpoint and letting the data speak for itself to e.g. discover putative operons.
Very cool work - thanks for the accessible explanations!
Do the donor extremophile fragments always integrate in the same spot, or potentially anywhere in the host? If the latter, I wonder if there are significant second order effects of not only which fragments provide a fitness benefit, but if they provide *more* benefit in certain genomic locations! Maybe once helpful fragments are found, there could be a follow up experiment evaluating the best integration location for the best fragment?
For this post, we were actually using fragments integrated into plasmids rather than the genome! Although we are now working on getting the genomic integrated version of that up and working. Technically they are randomly integrated so yes -- location of integration could potentially be an important thing to keep track of.
Ah, right, that makes sense! I'll check more into the BOBA-seq details. And that's cool to hear the genomic integrated version is next. I know engineered landing pads are a common strategy for prepping a host to receive genetic inserts, but perhaps leaning into the selective pressure mechanism of the screen will be a nice way for the organism to tell you where it actually prefers a helpful fragment to integrate :)
It's great to see lots of data starting to come through!
Just a thought about chassis choice - given the multitude of choices, maybe taking the experiment to a lower level and doing some sort of emulsion-based cell free assays could work? Take a cell-free mix and add plasmids expressing GFP with extremophile gDNA added and generate picolitre droplets (fake cells) which are exposed to a high-salt condition - after an incubation time sort on GFP levels (proxy for 'still alive'). Obviously this sort of assay has limitations like probably not finding salt pumps etc. but it might be able to find genes that directly help an organism keep transcribing and translating at least!
It would be great if cell-free assays were a simplification compared to assays in cells, but currently cells are substantially easier for these tests!
Microfluidics assays have challenges, such as the variation in droplet size/composition sometimes varies more than the effect we’re trying to measure.
Cells also give us a convenient mechanism of doing pooled screens. Following transformation and outgrowth the majority of cells contain one DNA-barcoded element, whereas the loading of one DNA per droplet across millions of DNA is a major challenge in microfluidics.
People are doing good work on these in vitro challenges (RevivBio comes to mind) but this is not our focus now, especially as we hope to screen for traits in the actual organisms we hope to use, in the most relevant conditions we can create- which would not be possible with droplet-based screening.
In vitro methods can probably help us solve some problems faster, but we haven’t found a good use for them here.
I sincerely appreciate the openness. Providing the methods, data, interpretations, and next steps is a great way to lean into scientific communication! Thanks for the post
This is a very nice write-up! Thanks for sharing your work. I have a few questions:
1) When cloning the DNA+barcode into the E.Coli strains, is it always placed in the same location?
2) You show good correlation between replicates across your time course, I was wondering if any genes showed positive correlation across time points. Particularly your high fitness effect barcodes at day 5.
3) You averaged your fitness per position, but I wonder if you analyzed data after collating positions/inserts by operons you would have more consistent results. Particularly when using two very different microbes where you may miss interesting things if the full operon isn't included in the DNA insert (thus making any one part non-functional)
Hi!
1) in this dataset the cloning was actually done into plasmids, so technically yes, but it's not in the genome. Going forwards we're working with random genome integrations.
2) Definitely, there's correlation across time, but asymmetrically. Things that win at day 5 are high fitness at all timepoints -- but things that are high fitness at day 1 aren't necessarily still high fitness at day 5. This is an advantage of sequencing, it reveals high fitness species early on that may not be discovered experimentally via fixation!
3) This is something we're looking into -- examining the data from a more agnostic viewpoint and letting the data speak for itself to e.g. discover putative operons.
Thanks for the reply! This is really cool work - good luck on your next steps
Very cool work - thanks for the accessible explanations!
Do the donor extremophile fragments always integrate in the same spot, or potentially anywhere in the host? If the latter, I wonder if there are significant second order effects of not only which fragments provide a fitness benefit, but if they provide *more* benefit in certain genomic locations! Maybe once helpful fragments are found, there could be a follow up experiment evaluating the best integration location for the best fragment?
For this post, we were actually using fragments integrated into plasmids rather than the genome! Although we are now working on getting the genomic integrated version of that up and working. Technically they are randomly integrated so yes -- location of integration could potentially be an important thing to keep track of.
Ah, right, that makes sense! I'll check more into the BOBA-seq details. And that's cool to hear the genomic integrated version is next. I know engineered landing pads are a common strategy for prepping a host to receive genetic inserts, but perhaps leaning into the selective pressure mechanism of the screen will be a nice way for the organism to tell you where it actually prefers a helpful fragment to integrate :)
It's great to see lots of data starting to come through!
Just a thought about chassis choice - given the multitude of choices, maybe taking the experiment to a lower level and doing some sort of emulsion-based cell free assays could work? Take a cell-free mix and add plasmids expressing GFP with extremophile gDNA added and generate picolitre droplets (fake cells) which are exposed to a high-salt condition - after an incubation time sort on GFP levels (proxy for 'still alive'). Obviously this sort of assay has limitations like probably not finding salt pumps etc. but it might be able to find genes that directly help an organism keep transcribing and translating at least!
It would be great if cell-free assays were a simplification compared to assays in cells, but currently cells are substantially easier for these tests!
Microfluidics assays have challenges, such as the variation in droplet size/composition sometimes varies more than the effect we’re trying to measure.
Cells also give us a convenient mechanism of doing pooled screens. Following transformation and outgrowth the majority of cells contain one DNA-barcoded element, whereas the loading of one DNA per droplet across millions of DNA is a major challenge in microfluidics.
People are doing good work on these in vitro challenges (RevivBio comes to mind) but this is not our focus now, especially as we hope to screen for traits in the actual organisms we hope to use, in the most relevant conditions we can create- which would not be possible with droplet-based screening.
In vitro methods can probably help us solve some problems faster, but we haven’t found a good use for them here.