ELOWITZ, Michael, Cell Computation, Caltech, May 22, 2018
<p class="MsoNormal" >ELOWITZ, Michael, Cell Computation, Caltech, May 22, 2018
This is a Frontiers in Biology lecture at 4PM at Stanford in Clark Aud.
He is a superstar at Caltech, and this is an SRO audience! (really a squeeze to get in.)
Overall topic is on computation/ communication in cells.
(I am a few mins late.)
he has begun...
there are a small # of pathways...
eg Hedgehog, Wnt, growth factors;
but how do you send a message to a particular cell type...
he shows a cute letter...
"sorry for the slow reply... but your message reached me via diffusion."
here's an example: <b>BMP = bone morphogenetic pathway</b>...
which can induce ectopic bone formation.
but BMP uses a promiscous receptor-ligand interaction...
there are abt <b>20 BMP ligands.</b>
receptors can send 2nd msgrs that lead to phosphorylation.
yes... abt 20 BMP ligands and each can bind to as many as 7 type 1/II receptors...
so seems this is v promiscous...
so, how do achieve specificity?
BMPs are morphogens and operate in difft zones of the embryonic kidney.
he draws an analogy from BMP pathway to Neural networks...
work of Yaron Antebi and James Linton.
Y Antebi in Cell 2017.
this work is in NMuMG mouse mammil. cell line.
some ligands are additive and some are ratiometric.
example of additive is interax of BMP9 with BMP4
but GDF5 is ratiometric with BMP4....
so difft ligands interacts in difft ways.
he shows a model similar to a neural network.
affinities are the Kijk's
our parameter scan reveals maybe 100,000 difft combos.
and depends on relative ligand concentrations.
and depends on RELATIVE levels... as to
whether Additive; ratiometric; or whether imbalance versus balance;
so, affinity is not activity:
and no downstream circuitry is required...
can see these even at 20 mins;
but... is this sensitive to receptor expression...
but difft cell lines show distinct receptor expression profiles...
he contrasts NMuMG cells with mouse ESCs.
he shows combinatorial specificity;
two ligands an orthogonaly address multiple distinct cell types.
letter to "Frances Fibroblast"
99 Epithelium Avenue
Los Ancellus
from Nellie Neuron
32 Ligand Lane...
*
<b>NOTCH pathway... </b>signal from one cell to another adjacent cell.
cell A.... <b>four ligands; cell B has 4 receptors...</b>
but perhaps they all do the same thing... cuz 2nd messenger might be the same...
Sandy NogodaPal
Delta 1 and Delta1 activate Notch with difft dynamics...
one stays on the other comes on then turns off.
continuous case: steady trickle of receptor into the receiving cell;
pulse might be due to a big burst... he shows animation
(Daniel Fisher asks a Q)
do dynamics matter?
Sandy does expt with DAPT...
yes... dynamics do matter...
depends on duration of stimulation (must be > 4 hours) for Hey1 to turn on.
summary: two ligands produce difft dynamics;
work of Nandagopal in 2018.
*
many difft interpretation schemes.... Notch is multiplex;
*
Computational approach:
synthetic circuits...
oscillators, binary logic, signal processing; classifiers;
Drew Endy comic book; Common Signal Carriers:
Adventures in Syn Bio 2005... v boyish Drew Endy...
gene regulation is naturally composable.
but can we do this with ONLY proteins...
could be nice to do this without entering nucleus...
Tobacco Etch Virsu genome TEV
viral proteases might be ideal protein circuit components...
cleavage is all or none.
CHOMP is this project...
look at Citrine...
Degron clipping...
we wanted to make proteases that turn off other proteases...
but ... a gallery of failures.
can add or cleave off lucine zippers.
and can do cascades with these zippers.
can do binary logic;
Does it chompute?
can do OR, AND, NOR gate...
<b>this is CHOMP...</b>
CHOMP can detect and kill the cell.
Ras Kill circuit can be encoded on a single transcript...
<b>CHOMP is circuits of hacked orthogonal modular proteases...</b>
this is all with proteins!
Chomposable proteineases enable design of programmable synthetic protein circuits.
in Nature by Yaron Antebi...
CHOMP Xiaojing Gao..
</p>