The E.coli was cited only in quotations of a publication, of which I
only mentioned to borrow concepts from, because of the ability of
bacteria to produce a polymer, not a conducting polymer, seemed
relevant, but that's not the same as suggesting that I want to use
E.coli to do that. I I can see the confusion. My writings are very
dense and interdisciplinary (Many of my concepts are suggesting to
test and apply meta-ideas to other fields) That paper is here:
"Users of 3D printers such as the RepRap that use polylactic acid
> (PLA) have potentially an inexpensive and robust way of harvesting the
> bioplastic as has been reported: "Biosynthesis of polylactic acid and its
> copolymers using evolved propionate CoA transferase and PHA synthase" (Nov
> 2009) and "Metabolic engineering of *Escherichia coli* for the production
of
> polylactic acid and its copolymers" (Nov 2009)."
http://blog.reprap.org/2009/11/bugs-brew-pla-direct.html
http://www3.interscience.wiley.com/journal/123188458/abstract
http://www3.interscience.wiley.com/journal/123188463/abstract
From the first link: "However, they are presented as granules in the
e.coli, it is not deposited. The researchers lysed the cells and used
a centrifuge to extract them. On top of that the PLA was identified
with mass spec, so I'm not sure what the purity is like."
Thus, what I would want to borrow from that technique towards a
hybridization with the techniques behind this paper:
http://x-journals.com/2010/microbes-reprogrammed-to-ooze-oil-for-renewable-biofuel/
is exploring membrane secretion of polymers, to understand how
conducting monomers of polypyrrole can polymerize, and how to achieve
a high degree of polymerization for it to become insoluble in water as
Dr. Albert Mihranyan has replied. Alternatively, if there is a
mechanism to line up bacteria or yeast in a row so they simltaneously
secrete monomers, which are readily polymerized with enzymes designed
for the lattice of polypyrrole (or the conducting polymer of
interest).
This is a bit like searching for two completely brand new techniques,
and applying discrete, modular parts to a new cell system/factory that
allows two isolated machinery aspects (biosynthesis and membrane
secretion) to be combined to produce a new function- secreting not oil
for renewable biofuel, but a polymer, and thirdly, a conducting one.
"but what I wonder about, is how
> easily a mutation can knock-out the promoter sequence, and so, make
> the RNA readable by natural ribosomes."
That is something I plan to look into too and I discuss that below.
> Ok, so this is a little like the arsenic phosphate switch found in
> some high arsenic environments? The only way to spread the micro-
> organism, is to spread the specific environment it is based on?
Correct, that sounds like an applicable analogy. From that recent
news, phosphate in the DNA backbone apparently is replaced with
arsenic at rates of up to 11%, in addition to lipids and ATA
(adenosine-tri-arsenate?) The analogy works even without using arsenic
(i.e. a silicon based organism instead of carbon, though that hasn't
been observed) If knocking out an amino acid synthase, thus requiring
addition of amino acids such as L-glutamine (or proline/arginine,
etc), then natural uptake of a plasmid or lysed bacteria that contains
Glutamine synthetase on a auto-(self-excising)-transposon in the
genome/plasmid it would be less likely.
"I suggest that the real problem is to limit the number of
> locations to those where prophylactic mechanisms are fairly strong,
> and monitored for failure."
"That's what I was referring to when I wrote:
"I would look to the history of biologically-produced
pharmaceuticals, transgenic organisms/crops, and such compounds
produced by bacteria- antibiotics, insulin, etc, for guidance as to
the likelihood of a microbe escaping a bioreactor factory and
hybridizing with non-GMOs. I think some safeguards could be used such
as using less than minimal media, which could prevent the growth of
the bacteria unless certain minerals were added manually, rather than
automatically."
Using combinatorial/permutational genetic BioBrick parts could
decrease the mutation count.
There are computational mechanisms that I studied, such as the
likelihood of a point-mutation as a function of the number of base
pairs, and the repair accuracy of the DNA polymerase's error checking
subunits. If I recall correctly, they're in the 1x10^-6 range:
"In general, the mutation rate in eukaryotes and bacteria the rate is
roughly 10-8 per base pair per generation[4]. The highest mutation
rates are found in viruses, which can have either RNA or DNA genomes.
DNA viruses have mutation rates between 10-6 to 10-8 mutations per
base per generation, and RNA viruses have mutation rates between 10-3
to 10-5 per base per generation[4]. Human mitochondrial DNA has been
estimated to have mutation rates of ~3×10-6 or ~2.7×10-5 per base per
20 year generation (depending on the method of estimation)[5]; these
rates are considered to be significantly higher than rates of human
genomic mutation at ~2.5×10-8 per base per generation[1]. Using data
available from whole genome sequencing, the human genome mutation rate
is similarly estimated to be ~1.1×10-8 per site per generation [6].
RNA has a drastically higher mutation rate than DNA because of several
DNA repair systems that can correct changes before they become fixed
in the genome as mutations.[7]"
http://en.wikipedia.org/wiki/Mutation_rate
Those mutations would have to be addressed for each susceptible gene
at risk of mutation towards having genetic and organismal
transferrability to other (external) microenvironments.
Combining abiotic conditions with some synthetic parts could decrease
the mutation rate, though I'm not sure if that number would have to be
extreme (e.g 10^-90 or 10^-900+) as a many-fold reduction, but I'm
very curious as to what computational biology can integrate into it
and if such algorithms are used by industries that use bacteria/yeast
to yield synthetic/natural compounds.