Hacking Your Own Fluorescent Yogurt

There is a common conceit among we DIYbio enthusiasts, namely to suggest that one could opt to create “glow-in-the-dark yoghurt” using DIYbio-oriented techniques as a nigh trivial matter. Indeed, this conceit led to my recently being queried by twitter and email about the possibility; where are the guides and how-tos, if it is so trivial?
While a conceit it may be to suggest that glow-in-the-dark yoghurt would be trivial, that’s not to say it’s at all out of reach to the dedicated biohacker. Here, I will lay out a suggested course of action based on the available literature.

Firstly, let use choose our definitions. What do we mean by “glow in the dark”? There are two commonly pursued strategies to choose from; fluorescence and bioluminescence. The former means that the bacteria will glow some colour when exposed to blue or ultraviolet light, usually green. The latter means that the bacteria will literally glow, emitting their own light from within provided they have enough energy from food.

Though bioluminescence is without doubt a cooler trait, for a variety of practical reasons, fluorescence is a more practical choice. Until, that is, one has more experience in yoghurt hacking and more money to burn on ambitious goals! So, here we will explore the transformation of a yoghurt bacterium with a variant of “Green Fluorescent Protein“, which renders organisms fluorescent under blue or UV light.

So, now that we know what we want, what are we working with? Yoghurt is most often composed of a co-culture of Streptococcus thermophilus and Lactobacillus delbrueckii subsp. bulgaricus, shortened hereafter to L.bulgaricus (its former name). The two species work in concert, providing nutrients and cofactors to one another and producing acids and bacteriocins that prevent invasion of other unwelcome species. Together they digest and remove much of the lactose content of the milk, and produce the polysaccharides and volatile molecules that give yoghurt its distinctive mouthfeel and flavour.

Of these two, either could be engineered to fluoresce, though the construction of the DNA and the methods needed to get it into the cells would differ. For this example, let us say that we have chosen to engineer S.thermophilus (chosen at random; I haven’t even researched L.bulgaricus, and it may indeed be a far easier bet for this project! Expect an update later if so.).

A quick literature hunt on DuckDuckGo and Google Scholar reveals that the most popular method for engineering S.thermophilus is through induction of a state called “Natural Competence“. In this process, cells are encouraged to activate a natural system of DNA uptake by exposing them to starvation or stressful conditions. This system is composed of multiple regulatory systems, and involves proteins that actively bind and absorb DNA, and proteins that splice this DNA into the cells’ chromosomes if it matches them at least partially. Not all strains respond to the usual methods used to induce competence, and some strains are considered nontransformable even though they are known to contain a seemingly intact competence gene array.

This poses a problem to DIYbioers, who may not know whether their chosen strain responds to induction of competence. However, a deeper search through the literature reveals a more universal method. This method uses an artificial pheromone called ComS(17-24) (AKA Shp316 which consists of the amino acids “PYFAGCL”), which mimics the natural induction pheromone required for competence to develop. Interestingly, even in strains where competence cannot be induced normally, addition of ComS induces competence provided the necessary genes are intact. Therefore, this seems a safer bet for DIYbioers seeking hacked yoghurt!

You might be thinking at this point, isn’t that an awkward thing to get? Firstly, the alternative involves starving the cells in an extremely awkward “chemically defined medium”, the ingredients for which are likely to be expensive if desired in moderate purity. Secondly, the pheromone is a short peptide (a very small protein), and there are several companies offering cheap peptides; $2 per amino acid seems to be the usual price. For the ComS peptide (“PYFAGCL”), that would amount to $14, although expect to pay tax and refrigerated shipping sufficient to raise the price to perhaps $40 (or more?). For that, you will probably get quite a few transformations done!

Methods employed for induction of competence (see methods in this paper for an example) through ComS seem to bear a few artefacts of prior methods, including the use of chemically defined medium and prolonged incubation times with DNA prior to the addition of ComS. The use of Chemically Defined Media was originally necessary to induce competence by starvation, but with ComS added artificially this shouldn’t be needed anymore. Although I have *not* tested either the referenced methods or my suggested updates, here is a protocol that ought to work based on my understanding of the process and the function of ComS:

  1. Grow overnight culture in a rich broth at 37C.
  2. Overnight culture washed twice in same volume of broth; (to isolate cells between washes, centrifuge at 5000g for 9m at room temperature).
  3. Re-suspend in one volume of 50% diluted broth.
  4. Dilute 1/30 in 50% diluted broth, aliquot to 300ul Volumes.
  5. Add 1uM ComS(17-24) and 25ng linear DNA with 1kbp+ complementary ends to target site.
  6. Incubate for 5h at 37C.
  7. Plate on selective media and incubate overnight at 37C.
  8. Choose colonies and PCR-verify.

“Rich Broth” in this case is usually an exceptionally rich broth called “M17″, which includes beef extract, yeast extract, pre-digested protein, glucose, and a few salts and phosphatase inhibitors, sometimes even with lactose added for good measure. This is almost without doubt overkill. A broth consisting of skimmed milk powder and a bit of Yeast Extract (without added salt) will certainly suffice. Let’s say 20g Skimmed Milk Powder and 3g Yeast Extract in 1L deionised water.

You’ll note that in the above protocol, the DNA is required to have 1kb+ complementary ends to chromosomal DNA. That’s because the uptake of DNA is only the first part of the process; the DNA has to integrate into the cells’ own chromosomal DNA in order to be maintained for longer than a generation or two. To achieve this, you have to provide DNA that has flanking complementary regions long enough to encourage a process called “homologous recombination“, wherein the cell exchanges part of its normal DNA for the provided DNA. In order to select for the subset of cells that do exchange their DNA thusly, it is common to use antibiotic resistance genes, so that non-transformed cells can be killed.

Once cells have been selected, antibiotics are probably not necessary in order to ensure the DNA is inherited by daughter cells, as it is part of the chromosome. Since antibiotics are not needed once the gene is established, it would be responsible to remove the cassette afterwards, even if it introduces additional complications. This is too complicated for a “beginner’s guide to hacking yoghurt”, but should be strongly considered; I’d suggest an inducible Cre/LoxP system, and would be happy to consult and assist anyone wishing to attempt this.

For our project then, we want DNA that resembles the following:

A diagram showing the format of the desired DNA

Colourised and labelled sections of the DNA, presented with the unnamed section of target DNA. This assumes the simplest scenario where no "cleanup" mechanisms are included to remove the resistance gene.

The promoter (Prom) should be a constantly “on” promoter, termed “constitutive”. The terminator (Term) is a region of DNA that prevents the gene from transcribing beyond its normal context, which could cause unintended interruptions of cellular functions; unhappy bacteria could result, and the gene could end up unstable. The antibiotic should be chosen to avoid medically significant antibiotics such as ampicillin; this is a civic responsibility matter, as otherwise your yoghurt could end up assisting dangerous pathogens in becoming resistant to medicines. Ideally your antibiotics would be self-excising once they become unnecessary, leaving a yoghurt containing only harmless fluorescent proteins and nothing else.

The homologous regions (Flank1 and Flank2) you choose are down to where in the chromosome you wish the DNA to end up. There’s a few million letters of DNA there to play with, and quite a few unnecessary genetic bits you could replace if you chose to. There are probably some retroviral genomes hiding away in the average S.thermophilus genome that you could deliberately replace, for example. However, a design consideration to bear in mind is the presence of native restriction enzymes; enzymes that chop up DNA when they detect certain “words”. In S.thermophilus, there are quite a few such enzymes, and the most common “words” they target are: CCWGG; CCCG; GGCC; CCGG; GATC; GCSGC; GCNGC; CCWGG.

Although native S.thermophilus DNA won’t be cut thusly, due to the protective effects of the same restriction systems, any newly added DNA will be, and the efficiency of transformation will suffer greatly. Therefore, choose target DNA that doesn’t contain these words, or contains as few as possible. Also make certain to screen the newly designed DNA for these words, modifying as necessary to remove them. Most codon adaptation tools will offer to exclude sites, and manual removal from other DNA may be required. If you don’t know what you’re doing manually editing promoters and terminators, opt first to use different promoters/terminators rather than risk failure of the DNA entirely.

Once you’ve identified the target sites, order PCR primers to amplify the two flanking regions using PCR. You’ll need a good enzyme like Pfu or KOD rather than old-school “Taq”, because taq can’t reliably amplify large enough regions. Splash out and treat yourself to a good kit, or homebrew some from a Pfu-producing strain.

Then design and order your DNA: choose a GFP derivative of your liking, there are plenty to choose from. Alter the codons in the protein to minimise those that S.thermophilus doesn’t use, so that it expresses well. Don’t bother with a “best codon” optimisation strategy, as these strategies often fail to achieve the expected results. Just use tools like JCAT (downloaded locally, rather than the online tool) to identify which codons are least desirable.

To do this, find the S.thermophilus genome in NCBI, download it, and feed it into JCAT so that JCAT can build a profile. Then provide the desired GFP derivative to JCAT, and let it calculate which codons are least optimal; replace these so each codon is at least above the 50% margin. JCAT is designed to automagically make all the codons near-perfect, but manual alteration of just the limiting codons is all that’s needed. Sometimes, having all-perfect codons can be counterproductive, for reasons not fully understood yet.

Once you have the protein as optimised DNA, attach a constitutive promoter; you can probably find one by searching the literature, nicking it from a critical metabolic gene, or a virus that predates the species. Also add a shine-dalgarno sequence to your protein coding sequence, just in front of the start codon. If you can’t find an appropriate Shine-Dalgarno, the complementary sequence to the last few nucleotides of the 16S rRNA (changing Us to Ts) in S.thermophilus should be ideal. That is, after all, what Shine-Dalgarno sequences usually seem to entail; target sites for 16S rRNA binding.

Finally, add a terminator. Terminators should be fairly species independent, as their function is down to their RNA structure forming as they are transcribed. Therefore, you can probably use a good, well-tested terminator from the parts registry, even one from E.coli. However, if you can find a nice B.subtilis terminator, stick with that; it’s more closely related to S.thermophilus than E.coli. Best be on the safe side?

The resistance cassette can be built likewise, but it’s easier to just find one of the plasmids commonly used for studying S.thermophilus genetics, of which there are several, and to copy/paste the DNA corresponding to the resistance gene out of one of these. Be careful to ensure that the annotated DNA includes a full gene and not simply the coding sequence for the resistance protein; your resistance gene does need to be transcribed in order for it to work! So, if the “resistance cassette” starts with ATG and ends with TAA, be careful to include a wide margin of DNA around it, perhaps 150bp to either end to be on the safe side. If you actually have a plasmid matching one of these in stock, or can source one, just get it and PCR the DNA out to save money.

Once you have the desired DNA for expression of the GFP protein and the resistance protein, you need to flank it with the intended target sequences. In order to achieve this with the PCR-amplified target sequences, you need to use another round of PCR to assemble the two flanking ends to the custom designed DNA; for this to work, there must be primer-length regions of complementarity between the amplified flanking regions and the designed DNA.

In order words, add some DNA to either end of the insertion cassette that matches at least 20bp of the inner ends of the flanking DNA. Then, order your DNA and wait a month or so. You’ll be paying about 28€c/bp; that’s what I was charged by Epoch, at any rate (whom I can highly recommend).

When the DNA arrives, the procedure goes like this: use PCR to amplify the flanking ends of the target site, and to amplify your intended insertion cassette. Use another round of PCR to fuse the “left” flank, the insertion sequence, and the “right” flank together. Use a third round to amplify from this the finished DNA, by using the upstream (“forward”) primer of the “left” flank and the downstream (“reverse complement”) primer of the “right” flank. This will certainly call for a good enzyme, as you’ll be amplifying in excess of 3kb of DNA.

This PCR product can then be directly added, along with 1uM of ComS (which you should also have received in the post by now), to your washed, diluted overnight cells as described above. The cells should, if all goes well, be induced to active competence, absorb the DNA, and include it in their chromosomes. When you apply antibiotic selection after 5/6 hours of incubation, you’ll kill any cells that *didn’t* undergo this transformation; there may be very few cells indeed that survive due to correct transformation.

With any luck, you’ll see colonies on the agar the next day; try flashing a blue LED at them and see if they fluoresce the desired colour! You can filter out blue using an orange filter, which can be as contrived as a sweet wrapper. If you use UV light, you’ll probably have a clearer view of any fluorescence, because UV doesn’t appear very bright if it’s visible at all.. but don’t leave the light on too much or you’ll kill your cells! Definitely avoid using UV sterilisation lamps; stick with the sort you use to check banknotes, which use lower frequency light which is less harmful.

If you get your desired glowing cells, congratulations! Reunite them with L.bulgaricus, and try brewing up your first batch of fluorescent yoghurt! If not, try again and don’t worry too much. The transformation frequencies with S.thermophilus are really low, so perseverance may be called for.

However, in the end of the day, it mightn’t work at all, and there might be no foreseeable reason why it didn’t. If so, you’ll have joined the ranks of biohackers worldwide who’ve tried only to meet with unexplained failure. I hope the project didn’t cost you too much to attempt, but don’t throw it all out; share that DNA with others, see if they can improve the procedure; perhaps a different insertion site? A slight modification to the integrated DNA? A different transformation procedure? Most importantly of all, share your failure; mankind learns most from failure, and not enough failures are shared in science. Be the difference, speak up and tell your cohort what failed for you!

If you want to try this, get in touch with me first for more in-depth advice on some of the details. Before you commit, bear in mind that this project will cost a lot of money at current costs per nucleotide! Unless you can access the DNA for a green fluorescent protein and a resistance cassette (which would save you a lot of synthesis money), you’ll have to order these from scratch, and that won’t be cheap. At least a few hundred Euro. However, the reward is something you can share with others and boast about for years to come; your own fluorescent GMO yoghurt. Very 2012.

19 thoughts on “Hacking Your Own Fluorescent Yogurt

  1. Pingback: Cavalcade of Mammals » Blog Archive » Links for January 2012

  2. Thanks, Cathal – looks very interesting!

    Have you tried any of these steps yourself yet? I wonder why you chose the focus on natural competence and incorporation into the chromosome, as opposed to a plasmid-based approach. Seems like that would make things more complicated (but then, what do I know…)

    Looking forward to hearing from someone who’s tried this method!

    • Hey Patrik!
      I haven’t: I don’t have time right now to experiment with yet another species, but all the steps were pretty well supported by the literature available to me. I’m not sure if you really need a full kilobase of complementarity on either end, but that was the length used by other work..

      As to “why not a plasmid?” – stability is key, or you’ll need antibiotics in your yoghurt. Also, natural competence generally functions far more efficiently on linear DNA rather than circular, and therefore unless your plasmid is multimeric and cut open you’ll either lose bits of it or it won’t recombine into a circle successfully and will get lost. In short, linear DNA with chromosomal complementarity is the most likely approach to work, whether in a big lab or a biohacker’s kitchen!

      Why natural competence? Because that’s apparently the only reliable way to engineer this species without an electroporator. There don’t seem to be any chemical methods of inducing competence as there are with E.coli. I ran into this problem with B.subtilis too; natural competence is the key. Technically, this is a good thing because you should theoretically need fewer external chemicals to get started provided you have the right strain. Really though, if you don’t have the right strain already it’s a pain. Some yoghurts won’t need the competence peptide to absorb DNA at all, for example; it’d be just a matter of timing to get them to transform. To cover all bases and potential strains though, I’ve elected to suggest the peptide in all cases. You get better results anyway.

      I’d actually like to try this sometime; it’d be a lot of fun to work with a more nichey strain of bacteria and open up the dev-kit somewhat! I hope this acts as a suitable first step for others out there. :)

  3. A quick search gave me this paper that mentions a transformation procedure that’s pretty much exactly what you want. The plasmid (pLEM415) replicates in E coli (saves you tons of money), transforms well and is expressed without you having to deal with insertion. This may be a way to get a quick pass at yogurt that glows.

    The problem is you’d have to maintain the plasmid with antibiotics… scanning through other work by that lab and I think they may have an alternate approach in this paper although I haven’t read it through to be sure it’s what you want.

    • Hey, thanks for the refs!
      The choice for chromosomal integration was deliberate; an integrated piece of DNA that isn’t openly harmful to the cell is probably going to be quite stable, whereas as you say yourself the average biotech plasmid needs antibiotics to stick around.

      That’s not something you couldn’t address, of course; if you wanted to go with plasmids, just find a single-stranded origin* for your plasmid (assuming it’s a rolling-circle-mechanism replicon) to enhance stability, and stick a restriction/modification operon on the plasmid.

      The former merely replaces something that was probably hacked out by accident when the “minimal replicon” was originally developed using antibiotics, which is required for single-stranded intermediate plasmids to regenerate into double stranded DNA. The lack of a single-stranded origin is probably responsible for most of the observed instability in biotech plasmids, forcing one to use antibiotics to overcome innate instability.

      The latter, a restriction/modification operon, forces the cell to keep the plasmid or die. The methylases that protect DNA are less active than the restriction enzymes that destroy DNA, so cells that try to replicate their chromosome without a healthy number of methylases around (say, because it lost your plasmid?) tend to fail and die, whereas those that retain the plasmid survive. Disadvantage; the cells are subsequently harder to hack without screening any additional DNA for the new enzyme’s recognition sequence.

      Plasmids would certainly have an advantage for fluorescence because you could have a high copy number of the GFP gene in the cell, allowing for lots of bright glowey goodness. If you can engineer a stable plasmid with passive and active stability factors, you could even sell it as a biotech tool, recouping the costs of your project! :)
      It is *possible* that the plasmid I’m currently working on will work also in S.thermophilus. I’ll be sure to test it and see just as soon as I’m done with B.subtilis!

      • Oh, I just visited your refs; they cover techniques for hacking L.delbruckeii rather than S.thermophilus. That’s awesome; it offers an alternate route to fluorescent yoghurt that I never bothered researching for reasons of time constraint. Thanks! :D

  4. Pingback: How to make glow-in-the-dark or Flourescent Yoghurt | Citizen Science Quarterly

  5. Hey Cathal !

    Very nice article! This project is clearly not for newbies due to the multiple tricky steps that are required before obtaining the saint graal! ;) I tried myself last spring to get along with this cool project, but I faced pretty fast limitations in the isolation of the desired strains of bacteria from yogurt. At that time I had a naive approach as I was working with LB but it turned out that it was not satisfying for growing both L. Bulgaricus and S. thermophilus. I started to read agronomy books on fermentated foods and there was written protocols for optimized isolation of such bacteria but it requires very custom made media, though most of them use milk as base. This is one of the main starting difficulties for me and whoever would like to work on this project should spend time first on that.
    Another tricky point will now also rise at the end of the project, when you already got your GFP expressed in the desired bacteria, it is the making of the starting culture where both bacteria strains are concentrated and mixed in precise quantity and then used on milk to get yogurt. This is also explained in relevant books.
    We, at La Paillasse, will surely continue to work on this project now that our lab is completely set up :) Let’s make 2012 the year of the fluorescent yogurt!
    Cheers!!

    • I haven’t looked much into growth requirements for S.thermophilus yet, but I’m planning to sometime soon; I want to find a media for thermophilus that contains no lactose, so that I can use a lac-operon replacement and re-introduction strategy. Doing so would allow one to get GFP in with antibiotics while replacing the lac operon, and then get the antibiotic resistance genes out again by replacing them with the original lac operon and selecting on lactose. But for this to work, you need a media to grow them on in between containing something other than lactose. I know that thermophilus will grow on glucose, at least, but beyond that.. Perhaps LB with added glucose at 45C would be somewhat selective?

      Would be awesome to see you guys working on this, I hope I haven’t mis-written it though! :) If you’re lucky, the strain of thermophilus you isolate mightn’t need the pheromone, although efficiency will be higher if you can get it.

      As to getting the mix of bugs right, I’d say discard the first few brews and it should work itself out! ;)

  6. Cathal,

    Thanks for such a nice read. I was wondering what could be the applications or advantages over fluorescent E.coli?

    • Hi Dexter,
      There’s no utilitarian reason why yoghurt would be any better than E.coli, except that you can’t make yoghurt with E.coli.
      Although I wouldn’t suggest eating any lab-altered bacteria without proper safety testing, there isn’t likely to be anything in the glowing yoghurt that would harm someone; it’s a novelty, and one which is very compatible with public engagement. In other words, it’s cool!

      • E. coli would taste like … well, what do you think gives human poop it’s smell?

        You wouldn’t need to synthesise the flanking arms; you could amplify them by PCR, then use overlap PCR to put them together with your targeting cassette (but the overlapping portions between the PCR products should be as long as possible; 50 – 200 bp).

        Why not make electrocompetent bacilli? Those extra logs of efficiency are worth some effort.

        And one last thought; if you want to get something into the chromosome, either a transposon or an integrating phage would save you an awful lot of effort. You would have to get the phage either from a stock centre, or by collecting from a likely source and plating on a lawn (looking for plaques), or you could go for long PCR to find one. That said, you would really be well advised to test your flourescent system on a plasmid before you went to the effort of trying to get it into the chromosome.

        (a transposon on a plasmid would potentially be a good hacking tool kit project; if I recall correctly, the Tn10 from E.coli is non host specific. You could have a gutted transposon cassette with a multi cloning site inside, and on another part of the plasmid, an inducible transposase. Would be easy peasy to put together by PCR (you could probably easily recover the transposon from your intestinal flora by PCR); hard part as always would be all the trouble shooting)

  7. oh, and I also meant to say, I thought the idea of putting a restriction/modification addiction cassette on the plasmid to make it stable, was pretty cool.

    • and one last thing, the transposon Tn5 is non-host specific, and could be made into a useful tool (this sounds like a decent hacking project; it is pretty simple, it amounts to hacking a plasmid, and you can probably hack it out of your own commensal flora):
      http://www.epibio.com/item.asp?id=291

  8. oh, and I also meant to say, I thought the idea of putting a restriction/modification addiction cassette on the plasmid to make it stable, was pretty cool.

    the transposon Tn5 is non-host specific, and could be made into a useful tool (this sounds like a decent hacking project; it is pretty simple, it amounts to hacking a plasmid, and you can probably hack it out of your own commensal flora):
    http://www.epibio.com/item.asp?id=291

  9. Pingback: Tuur Van Balen shows how to hack L.delbrueckii on stage | Indie Biotech

  10. Pingback: Hacking yogurt | Molecularist

  11. Pingback: Using GFP to make food of the future!! | Uncanny Valleys

  12. Pingback: 193: TZ Interview – Cathal Garvey / Biohacking

Comments are closed.