Micro-western Arrays for Identification of Novel Mechanisms Underlying Resistance of Cancer Cells to Chemotherapeutic Agents
Richard B. Jones, Ph.D. Assistant Professor, Ben May Department for Cancer Research, University of Chicago Institute for Genomics and Systems Biology, University of Chicago
Friday afternoon session: Bioinformatics and Systems Approaches to Cancer Treatment and Diagnostics, Beckman Auditorium, 1:30 PM -- 2:15 PM
Abstract: A fundamental goal in cancer treatment is the delivery of therapeutics in a manner that takes into consideration genetic differences between individuals. Genome-wide association analyses have discovered many DNA variants that influence complex phenotypes such as response to anti-cancer chemo-therapeutic agents. Expression quantitative trait loci (eQTL) mapping techniques have subsequently been employed to identify gene candidates whose mRNA expression is affected by genomic variation and that may serve as missing links between genome variation and cell phenotypes. To improve our understanding of the genetic basis of differential response to anti-cancer chemotherapeutic agents, we quantified hundreds of cell signaling proteins and transcription factors in genotyped HapMap LCLs. We then identified many novel genome loci associated with protein expression and identified many novel proteins whose expression was correlated with cellular response to anti-cancer chemotherapeutic agents. We subsequently validated several of these proteins as novel causal agents whose expression influenced cell death in response to anti-cancer agents. In this talk, we will speak about our vision for using global micro-array based protein analysis approaches to further the goal of personalized medicine.
Speaker bio: Dr. Jones is an Assistant Professor in the Ben May Department for Cancer Research at the University of Chicago. He is also a Core Fellow of the Institute for Genomics and Systems Biology and is a founder and director of the University of Chicago Micro-western array Protein Analysis Core Facility. Dr. Jones has a long-standing interest in the development of automated protein analysis technologies for interrogating the roles of protein systems in cancer biology. Dr. Jones received his undergraduate training at Indiana University. He received his Ph.D. at the Center for Cancer Biology and Nutrition through Texas A&M University. He trained at Harvard University as an NIH postdoctoral fellow where he participated in the development of protein microarray approaches for studying mapping human protein interaction networks. His lab has since developed the micro-western array methodology for the network-level interrogation of protein abundances and modification states. His lab has recently applied this methodology for identifying protein networks that underlie resistance to anti-cancer agents.
Other and that gives rise to differences in translation of whole classes of proteins that that have that particular codon in them
And these would be kind of invisible to an expression array study
Because many of these snips would actually not exist in the mRNA that you're for the target
You're studying they exist in something that regulates the translation of all proteins, right?
So just to briefly give kind of a rundown of kind of the methods that are are used to do this
Or have been used to do this in the past and why it's been difficult to to scale these up to whole genome studies
There's the traditional 2d gels and mass spec. This is kind of what I was
I got a little bit of training as a biochemistry graduate student a long time ago, and I thought wow
This is really a tough job to
Cut out lots of bands and put them in the mass spectrometer and try and recast the gel and get the same experiment several times
It's really a tough chore if you were to try and do a protein phenotype on a hundred individuals
And try and understand the genetic determinants behind these spots. It's very difficult to quantify some of these and it's very difficult to
To go across many samples
So there's newer methods that have been developed in the last 10 years one of which is the so-called reverse phase lysate array
that's been pioneered by
Lance Leota
and Emanuel Petricoin and now has also been
Pushed forward by Gordon Mills at MD Anderson Cancer Center
You can take cell lysates from your your cells of interest that have been treated with different conditions and and array those cells on to
So-called reverse phase lysate arrays. These are just nitrocellulose coated
slides that you can array your your lysates on and then you you add antibodies to proteins of interest and the notion is if your
Protein of interest goes up or down or gets modified in a certain way
And if you have an antibody directed at that protein or modification you can quantify it and you can use very small amounts of sample
but the real problem here is that there's a lot of cross reactivity that goes on with antibodies and even though you've purchased an antibody to
Protein one it may recognize protein one really well
But it may also recognize a little bit proteins two through fifty
That would obscure your ability to see changes in protein one because there there's background there
So there are a limited number of antibodies that you can use with the approach that are highly specific highly selective
And yeah, there's a lot of validation that goes into that process of finding the antibodies to begin with
So the kind of tried-and-true method that's been used for a long period of time is the immunoblotting or Western blotting method?
Where you electrophoresis?
samples on a gel to reduce the
Complexity of the mixture and then even if your antibody kind of has some other bands that show up you can kind of in many
Cases discriminate between what's what you know is the band you want to look at and the other bands?
But it's it's a lot of work
To measure each individual protein it takes a lot of reagents
It's expensive each one of these antibodies now is on the order of three to four hundred dollars for a hundred
Microliters, and that's that's enough to do five to ten experiments if you wanted to go across the whole transcription factors across hundreds of individuals
This would be impractical
So as a result most proteomic studies trying to relate genetic differences in
in your predisposition to disease or cancer
Have have either examined small numbers of proteins across. Let's say the NCI 60 panel or
across lung cancer samples, etc
Or they've examined a large number of proteins
using mass spectrometry across one human cell line or across three human cell lines and try to ask are their differential expression and
With these one or three you can never have any power to say this genetic variant is responsible for this difference in
In your phenotype, and if you're going across samples just small numbers of proteins really difficult in fact to
discover new variants associated with
Response let's say to chemotherapeutic agent
So the few studies that have kind of pushed this forward trying to understand relationships of genetic variation
With differentially with with mRNA and with proteins of they've used yeast as a model system
And in those particular cases they found
loci that
correspond independently to proteins protein expression they found
loci that associate with transcripts, and they've also found
loci that affect both but
differentially that in many cases the effect on one is much greater at different than the other and
This information suggested to us that if we were able
To measure large numbers of proteins across large numbers of individuals
We might find unique loci that would give rise to differences in proteins and in fact those loci may also have been
previously proposed to be important for your response to diseases and
Therefore you might infer that those proteins associated with the loci that were responsible with the diseases
Might be drug targets for that disease
They may actually causally affect that disease and if you could find ten of those
proteins associated with with a particular with different loci associated with disease you might be able to start building networks of
influence of proteins related to disease
so we thought so I'm in the
Binmay department and and a close collaborator physically and now in terms of of our work together as Eileen Dolan
He's done a lot of work on relating
genetic
variability with your
Differential ability to respond to common chemotherapeutic agents that are used to treat cancer like cisplatin carboplatin
4-hc etc
And she's done a lot of work to relate genetic variants that associate with that and so we thought we could
Actually start to relate protein levels to genome loci
Some there are many genome loci that have been associated with response to chemotherapeutic agents
But in many cases the false discovery rate associated with those associations is very high
it may be that 90% of the variants you found are false positives and
The reason is because there are so many loci in many cases 10 million snips that you're testing
For association that simply by chance you'll pick up many false positives
But the notion is if you have a series of snips associated with protein expression
And you have a series of proteins associated with
With that we found that whose levels are associated with chemotherapeutic agents
Some of which you knock down and you show that now you've induced changes in sensitivity to agents
you might actually be able to to to narrow down the search of
genome loci associated with with with
Responses and you may be able to find new proteins associated with responses, so here's the cast
Ron Hawes is a crazy graduate student that came to my lab and said look. I'm trained in plant genetics
I'd like to actually start doing genome wide protein studies protein associations
And I have little biochemistry experience, but if you could teach me biochemistry
I have a lot of genetics experience and so we could kind of start bridging the gap
Amy Stark is a graduate student in Eileen Dolan's lab
She actually just got her PhD and she's very interested in the genetic determinants of pharmacologic response
Kevin white I'm not sure if I should call him crazy or not. He's the director of our Institute and he
was responsible in large part for recruiting me and the notion he gave to me was
Rich if you could find some way of measuring all the human transcription factors
We have these RNA sequencing methods and nanostring and other things where we can follow mRNAs
But we don't really know what proteins are responsible for transcribing those mRNAs
We're inferring and we're trying to do computational ways of doing that
But if you could have some method of doing that we could potentially start to really increase our understanding of transcriptional responses
so
With Kevin's help we purchased an initial collection of about 5,000 antibodies directed at all human
in fact multiple antibodies directed at all human transcription factors and
The notion was we would start to use our our methods to screen those antibodies for which were the good ones which were the bad
ones and then we would use that collection to start screening HapMap cell lines for
expression of large numbers of transcription factors and relate them to these
Chemotherapeutic differences and so I'll back up a little bit
For those of you that aren't aware of the HapMap
Project the notion is you?
that over the last 10 years or so
blood samples have been collected from
individual populations across the earth
They're Asians and Africans and Chinese and Japanese and and Seth population in Utah
the notion is that these B cells were taken and then immortalized with epstein-barr virus and
And given out the labs all across the world where they're where their genome was was
Profiled for large numbers of snips and now and I'll talk a bit more about that
And there are advantages and disadvantages to this particular system
And we think many of the advantages outweigh the disadvantages, and I'll talk about many of the advantages
It's very easy to experimentally manipulate them. You can very quickly take these panels of
Individual cells and actually profile their response to different agents
And now they've been genotyped, and they're fairly cheap to grow their unlimited resource because they've been transformed with EBV
And there's lots of lots of genetic information, and I'll talk about more information. That's available for those in just a bit
The on the downside the phenotype may not recapitulate in vivo
This is a in vitro system subject to in vitro confounders
Many of the cells have different numbers of epstein-barr virus that have integrated into them and some of them grow slightly faster than others
And so in many cases we have to perform surrogate variable analysis or principal component analysis to try and regress out a non genetic
confounders
Most cell lines lack p450 system for pharmacokinetic studies, and so that's another limitation
So here are the things that are available with this particular system that make it made it pretty attractive for us
There are four million snips directly genotyped and with new thousand genomes data. You can impute much of the missing data
And which allows you to have about 12 million snips at a minor illegal frequency of 0.0 of greater than 0.05
There are expression arrays happy metrics exon arrays that have profiled the expression of these
individuals and in culture
As RNA sequencing now from Jonathan Pritchard's lab at University of Chicago our collaborator
Eileen Dolan has his profile now the micro RNA in addition to mRNA with a postdoc in her lab Stephanie Wong
She's Eileen has now measured the response of these cells to 18 different drugs
And so we can actually ask what proteins are associated with cisplatin sensitivity carboplatin etc for 18 different
different drug treatments
we have a lot of knowledge about
non genetic confounders from other laboratories that have used these cells and
Now we also have a lot of snips from that
That are associated with many complex diseases
so if we find a snip or a series of snips associated with a series of
Novel transcription factors and those same snips are associated with complex disease traits
We can potentially make the gap or bridge the gap and say maybe these proteins
Associated with those loci are also associated with those disease phenotypes. It's it's
hypothesis generating from that
perspective
So to to stop for a second and ask a few overarching questions
How can we measure thousands of protein levels across human populations or a human population of 68?
lymphoblastoid cell lines from the Yoruban population of Hapnap
Is the genetic variation associated with mRNA expression the same are the are the same snips?
That if you regulate an mRNA abundance in cells and culture
Are they also regulating the proteins or are there different snips for each?
Which proteins are associated with resistance to chemotherapeutic agents? It's really cool. There's a lot about what's known about let's say cisplatin
damaging DNA and how you can
detoxify cisplatin or how you can detoxify
cytarabine through
The cell has mechanisms for for converting these particular agents into other agents that are less toxic
There's a lot regarding
Cell signaling and if it should if a cell should choose to die or if it should choose not to die go into a quiescent
State there are many things that aren't exactly known about how that works
Can we find genetic variants previously associated with complex traits?
That are associated with protein expression. There are lots of these there are lots of snips associated with particular
Phenotypes for which we they're in gene deserts, maybe they're in a non-coding RNA. We don't know
But perhaps we can actually start to relate some of these to protein levels and actually find out what the underlying protein mechanisms are
So the first question, how can we measure thousands of proteins across human populations of 68 LCLs?
So here I'll take a step back from all this talk about genetics, and I'll talk about the basic laboratory problem of how do you measure?
Proteins and and measure it in a quantitative way in a in a way that has a large linear dynamic range and doesn't require
thousands of graduate student and postdoc hours and and all the agony that goes along with with measuring proteins, so
My advisor Gavin Macbeth at Harvard University
As a postdoc when I was leaving he had published a paper suggesting that
If you take cells you grow them in culture you lyse them and you array them onto those slides like I mentioned a few minutes
Ago called the reverse phase lysate array, and it's kind of an unfortunate word. It's reverse phase because originally
people put
Antibodies on slides and pull them down with arrays and now with this new with the new method that came out in the year
2000 they reversed that and they're just putting lysates directly onto pieces of nitrocellulose
And so there's no and so they called that reverse
Notion is now you can add antibody to protein 1 antibody to protein 2
These can be pan antibodies meaning that they don't care about modification state
or you can buy antibodies that really do care about specific modification states, and you can profile those across large numbers of
Slides using a minimal antibody and using minimal cell reagents
And he wanted to use this to profile pathways
But there were there were some big problems along the way if you look at this Western blot to alpha
phospho threonine 308 AKT prior to or after stimulation with EGF
You can see at time point one you see great phosphorylation increase on threonine 308
But you have this background thing that doesn't change
It's just giving you signal regardless of the of the change in the other signal
So if you go down here and look at this graph, this is the signal before stimulation on the array
Meaning that this band here actually gave a lot of signal without regard to phosphorylation state of this protein
Versus the Western blot the Western blot showed a huge change on the blue
this array didn't because of this background problem and
He went through lots and lots and lots of antibodies he published
34 he actually I think originally tried about 340
The ones he published were these for 34 antibodies generate equivalent identical information meaning that antibody was so specific that it was giving you very
Quantitative information on the one thing you were looking for
12 of 34 antibodies generated similar trends
So you kind of could see something changing
But now the dynamic range was so compressed that it was it was much less than before
this is a huge problem, and it raised a lot of alarms and and many of the people in the
Performing the art kind of knew about this you have to screen a thousand antibodies to get your good
20 or 30
And that's just something that they had that they did you just screen a lot of antibodies to find a small number
And I thought this is this is actually a problem
That's basically prohibiting the field
Nobody can move forward if you're only if you're stuck on 20 antibodies some of which are related to that thing you want to study
And most of which are not
So we I had the notion if you could just somehow run a Western blot didn't have to have much resolution
but if you could get a little bit of protein separation, maybe you could get over this problem of
Cross reactivity so our notion was you don't print directly on the to nitrocellulose you print on to acrylamide gels
And now you actually put a buffer strip here, so this is a non-contact micro rare
It's picking up like a micro liter of your cell lysate, and then it's depositing a few hundred picoliters
Across the gel so in this case. We have six samples that have been treated with EGF
We've lysed them
We're printing a protein ladder and those six samples
And we put a buffer strip at the top and the bottom these proteins are just sitting on top of the gel they kind of
Kind of absorb a little bit into the gel and after you put the buffer strip on and apply a voltage
They'll all go and they'll all go in this direction downward
And so you can see here after you've electrophoresis with a semi dry electrophoresis
You can see the protein standard and the bromophenol blue at the bottom
And this is the equivalent of 96 individual little Western blots and so after you've electrophoresis them you transfer them
To piece of nitrocellulose put this under a gasket and scan the membrane well first you need to add an antibody
That has a fluorescent label
Or a secondary that has a fluorescent label, and then this is what that looks like these are a whole bunch of beta actin
This is all six time points for
Replicate one and two and three and four and five the same cell lysate same six
cell lysates from different time courses all blotted with with beta actin to show the
The the same loading and then this is a whole bunch 96 different phospho antibodies that are directing
Detecting different things in this case you see phospho epidermal growth factor receptor tyrosine 1068
That gets phosphorylated immediately within a minute after adding epidermal growth factor
But then you see other proteins that respond with completely different kinetics because they're performing different functions in the cell
For example the s6 ribosomal protein responds very slowly after you've changed the growth media
And you have new amino acids new glucose you'll probably activate mTOR and a KT which starts to ramp up the metabolic
Cycle of the cell and you'll start to see the ribosome get phosphorylated
indicating increases in nutrition
So we did a lot of proof of concept in this first nature methods paper to show look that you get really good
Resolutions similar to a standard new page gel from in vitro gen
You can easily resolve 25 37 50
Depending on how long you run run out these micro westerns
And if you simply load a protein ladder you can and do twofold serial dilution
You can see linear response of many different proteins
You do the same thing with a whole bunch of cell lysates and dilute them twofold and then add up many different antibodies you can
See a linear response of all the antibodies and you can pick out
the the particular proteins of interest you're looking at and
Without regard to the other background protein spots. We did a whole bunch of
Comparison side-by-side of a standard western blot with a micro western blot and compared time courses and in all the cases
We we studied the time courses were very similar across them in fact we picked
two from the Macbeth paper that showed
That the antibodies were good
And they they were actually good in our case too
And we also picked many of the ones that actually did not respond with the reverse phase array
And they they actually work okay with with reverse phase lysate array because in many cases you can see
Background bands, and you simply can ignore them
You don't have to quantify them if you don't want to and this enabled us to generate a whole lot of phospho signaling data
This is to my knowledge the largest number of phospho
Proteins that have ever been profiled or examined with antibody based
methods and so we stimulated cells with different levels of EGF
And then we followed all the phosphorylation patterns in time
And we classified lots of membrane proteins that are other receptor tyrosine kinases that were getting phosphorylated very early
And then we had mid mid stage of map kinases and and ribosome response regulators
And then finally at the very end you see a lot more
ribosome proteins and transcription factors
being phosphorylated
So the notion is we could use this method to profile many proteins
unprecedented number of proteins with with antibodies pulled off the shelf from from vent catalogs from vendors and
You only need to use a few nanoliters for each measurement
So you don't need to worry so much in a discovery oriented screen for asking. What's the target of my small molecule?
I'll just add a small molecule, and I'll say aha these proteins
They all went down or they went up after some time with my molecule versus without my molecule
So we work with Doug Lauffenberger at at MIT to start to connect the dots. What are the proteins that are?
Causally responsible for phosphorylating each other and I don't have time to go into this
But the notion was this was kind of an initial phase of how would you collect a lot of protein data?
If you had a lot of protein data, how could you make sense of it in turn a whole bunch of?
heat map numbers into
This is upstream of that and it's upstream of that
I predict if I added an inhibitor to this then this should go down
But then there's this feedback mechanism that may may change right the notion is we want to start putting these together and making sense of the
data
so about
Towards the end of the development of this method is when Ron Hawes joined the lab and suggested that
And and was talking to others in the human genetics department that suggested. This would be really
Powerful if you could start to profile across the genome large numbers of transcription factors and cell signaling proteins
So we thought we would use this method micro Western arrays to quantify the steady-state abundance of several thousand proteins in the urban population of HapMap
We would initially focus on transcription factors and cell signaling proteins
We would then relate these abundances to genome sequence other central dogma intermediates like mrna
micro RNA since as well as a phenotype the sensitivity of
These cells to killing by chemotherapeutic agents, so we're starting measuring proteins
going to start comparing them to to
Genome variants that are and see if the same ones affect mrna versus protein and ask which of the proteins associated with
Chemotherapeutic sensitivity so we have a pipeline we start out with
Three thousand from each of 68
unrelated Yoruban
Individual or Yoruban LCL so the notion is technicians pull 68 HapMap cell lines
These are from 68 different people pull them out of the freezer and grow up big bottles of them
They then freeze pellet those 68 samples make 30 aliquots of each of them for our study and for each of our experiments
we pull out one of the aliquots form our
proteome wide analysis of transcription factors
in the initial stage we had to
Find out first which of all the antibodies we have actually work
So we pull the sample and we began screening these initial four thousand three hundred and six sixty six antibodies
And we asked did we see a band did we see what looked like a band at the initial?
Starting concentration of this antibody well if it gave something like this we would go back and repeat the experiment
But we would use ten or a hundred fold less of the antibody
And we're simply asking which of these antibodies give any signal if in in pools of ten
Individual cell lines and so then we start checking off the box this antibody seems to work it gave a signal
We don't know if that signal really is recognizing the protein
We think it is but in the end if it's a really interesting correlation that we find about that protein
We'll knock it down and ask did the band go away, and that will tell us this antibody really was recognizing the protein
We think it is
So if the antibody gave a single band we could only see one band
That was very high signal
We would then go back and take the whole 68 individuals
And we would quantify them with this method that I told you that has problems with antibodies that don't generate high
Just generate single bands if there were multiple bands
Then we would take all 68 back through the micro westerner a method and quantify all 68 individuals with this method
And then at the end of the day we ended up with a matrix out of the original four thousand three hundred and sixty-six with
a matrix of four hundred and sixty-five protein expression levels per individual across three
Biological replicates and then this whole micro westerner a experiment was also repeated three times so we have an idea of the technical
variation as well as biological variation and
Then the next stage of this was to further validate the antibodies
We think are the good ones by knocking down a subset of them
And then asking the question which of the bands what percentage of the bands actually went away
So the notion is now at the end of the day we have all of these data sets that have already been generated by others
on HapMap
We have now our our expression of three hundred and sixty-five transcription factors
Based on RNA sequencing there should have been a thousand we think
Expressed in LCLs at if you it's always difficult, but if you pick a threshold of like ten reads per
million
RPK and reads per thousand
in any event
There should have been a thousand mRNAs expressed by RNA sequencing we found three hundred and sixty-five transcription factors
And we found a hundred cell signaling proteins that were expressed and many of these are
The total abundances in many cases, it's actually the modification states of signaling proteins
So I know my time is running out
So I will this is all fairly new and we're just writing it up now for publication, but we identified
579 loci
associated in cysts
With 81 protein levels that means that there was a single nucleotide
Polymorphism that was located in the gene model for 81 of the proteins
That whose expression seemed to be highly associated with those and so here's just a top of the list as far as
Alphabetically there are snips associated with the expression of ribosomal components transcription factors
etc
lots of different
proteins and
We found that up to 50% of all cysts
EQTLs that means someone in the past has found a loci
That was associated with the mRNA expression of a gene and we found that if we found that
50% of all of the the snips that were associated with mRNA expression were also associated with the downstream protein expression
And depending on the p-value you pick
Once you get to the highest p-values of RNA sequencing those are the ones where we start to really pick up which means potentially many
Of these that have previously been found by mRNA expression at lower p-values
Many of them are actually false positives
And we identified over a hundred thousand
So-called trans pq TLS. These are things located throughout the genome that are nowhere close to the protein that they're regulating
But nonetheless
At 4 times 10 to the minus 11th. This is genome-wide significant
Most of these actually at the top are genome-wide significant these to my knowledge are the first ever
Trans loci in the genome that have ever been associated with protein expression
The notion is for example if you have a certain allele at this position of chromosome 16
You'll have greatly different levels of the data one protein and it's on a completely different chromosome
So here's an example of a replicating eq TL pq TL. So with this particular loci and ZNF 266
At the mRNA level by RNA seek an expression array. You see the CC individuals have lower levels of the mRNA and if you
Use Western blots of these micro Western blots you see differences in the protein expression in the same direction
and
So we also identified many novel protein specific
QTLs if you for some technical reason there were very few reads actually picked up at the mRNA level on RNA sequencing and now we think
in fact
Many of these early RNA sequencing studies didn't sequence to enough depth
So in fact there are many low
abundance mRNAs that you actually cannot read by RNA seek that actually give very high signal at the protein level and in some cases the
Snip is actually in some of the the probes on the expression arrays as well that were thrown out because of potential mismatches
so
Here's an example of a trans snip if you go across these are
Recombination blocks in the genome and then we found there are a series of snips in this particular recombination block that are highly associated with diet
oh one
diet oh one is is located on a completely different chromosome and
This snip is located in between two genes one of which is a lysyl tRNA synthetase and one of which is an adenosine
Deaminase both of which are responsible for the charging of lysine specific tRNAs
So the notion is if you regulate the expression of a tRNA
Or the or the expression of or the function of a tRNA synthetase perhaps
You may also regulate the expression of many proteins
Downstream that have that particular lysine codon in them and that would be invisible to you by an mRNA expression array analysis
or expression
Analysis, so we've identified hundreds of pQTLs that have overlapped the overlap with previously described snips associated with the disease traits
Many of these are kind of interesting and and would be expected
There's a snippet that we found associated with gab one levels and that's gab one is the adapter protein for hepatocyte growth factor
This snip has has been shown to have it caused you to have a different
disposition to hepatocellular carcinoma
Hox b7 which is important for vertebrate patterning is also
Associated with a snip that's associate that has been previously associated with hip geometry and so we have a series of these
a
Whole bunches of which for lots of different disease traits that we're now following up on to ask the question is this
protein tub
That is associated with a snip associated with adiposity and this tub protein has been associated if you knock it out in mice
It affects their fat metabolism
So potentially now we've actually identified the underlying variant the underlying reason why potential snips give rise to differences in
phenotype
And then lastly and this is I think really cool
We've identified about 200 proteins associated with sensitivity of cells to 18 chemotherapeutic agents
so depending on so I told you that that our
Collaborator Eileen had profiled the the chemotherapeutic sensitivity of 18 different agents and with each of these agents
We've identified between 18 and 30 novel proteins that are highly correlated with those responses
So here are just a few of them
And the notion is now
If we were going back again and saying if we have a protein associated with a particular response
And that proteins associated with the snip is a snip associated with the response and if we get this full triangular pattern relationship
It gives us high confidence that we've identified a bona fide genetic variant
Associated with your response to a particular therapy and we've identified the protein that is mediating that response
So the particularly you could actually make a drug to modify that response
So here's an example and I'll end very soon after this. So the notion is we found snip associated with aracy sensitivity
at a certain p-value we also found
This snip was associated with
ATF to a DNA damage response protein and we found this DNA damage response protein
Associated with the sensitivity of cells to aracy. So we thought okay. Well if you knock down
ATF to you should make the cells more sensitive to apoptosis
So we first asked can we actually measure ATF to being knocked down by si RNA and we found 25 hours that worked the best
so we picked 25 hours is the time we would test for differential sensitivity to aracy and
and so we have two different si RNAs here the the red is the
The si RNA to ATF to and the black is the scrambled and in both cases you see a subtle effect
Although it's statistically significant that shows that yes indeed ATF to if you knock it down you will get greater
sensitivity to the to the chemotherapeutic agent
similarly with paclitaxel if you
we found that
For data one, it's highly associated with apoptosis if you have higher levels of it
You'll have higher apoptosis and if you knock it down, you'll get lower apoptosis. This was a so-called
Functional validation of the associations that we're finding
So finally, we found also a whole lot of micro RNAs that are associated with protein expression
this HSA mirror 491 3p
Highly associated with the trends with the protein expression levels and has no association with mRNA expression level
So here's ZNF 7 one of the associated proteins has a high association with with this micro RNA expression
It has no association with the mRNA expression suggesting that this is potentially regulating the translation of this particular
mRNA
So I'm not going to go through all the potential future directions
You can imagine if you can study lots of proteins
You can start to relate it to whole lots of different things that I think are pretty new and exciting
And we're we're really excited about this
Gonna chip lots of transcription factors and find where they're binding in the human genome
Relate those to human disease
Tut-tut-tut-tut
There's this is just an example of recent use of the method that came out in cell a few months ago
By a V reggae have been near Hakko and at Harvard Medical School to unders that basically to ask how the signaling
Pathways are affecting the maturation of dendritic cells
and with that I'll end and thank
Mark Chacho and Max Chu the co-inventors of the method and
Amy Stark and Ron Haas the graduate students that were responsible for doing the the pharmacology and protein analysis work
Kevin white who is the visionary responsible for purchasing large numbers of antibodies
With the notion that I would find some method or be able to develop some method with which to use
thousands of antibodies
And Eileen Dolan who's the the pharmacogenome assist on the project and also?
Thank the funding sources as well our collaborators at MIT for doing modeling and all the people in my lab and lastly for anybody that
Would like to do micro Western arrays in their own lab you can go to the if you just type in
Micro-western arrays on Google or go to our core facility website you'll find that for five hundred dollars
You can submit your samples and profile six
Samples against a hundred antibodies and you can find all of the antibodies
On our website or at least several hundred of them that are available, or you can give us your own antibodies
So with that I'll take any questions, and thank you for your time
I
Was just wondering
That's a good question, so this is this is the most recent of the things we're working on so I
They've previously very recently
Published series of micro RNAs associated with drug sensitivity and associated with mRNA expression
And I don't know what the overlap is yet
So I haven't haven't looked yet, so what our data do suggest so far is that
If you compare RNA platforms against themselves if you were to take the mRNA expression data set
measured by
DNA microarray versus RNA sequencing the overlap the correlation is R of like 0.2
Among the experiments and the correlation between RNA seq and protein platform is like 0.4
And so from that perspective we correlate better with RNA seq than RNA seq correlates with expression
But we I still don't know if if the noise in the technical measurements would allow you to even know
What is the correlation between proteins and mRNAs? And what are the correlations between?
micro RNAs
mRNAs and proteins given that the technical variation is still pretty large and trying to measure that
And individuals with high micro RNA levels and their mRNA is not effective
at any statistically significant level
But again, there's still I think the possibility that these are individual artifacts in many cases in a large RNA seq data set expression array
data set you filter filter filter because of a probe mismatch or because you didn't get enough reads or whatever and you simply focus on
What correlated and somebody only did the experiment once?
Biologically and technically and there's probably if you were to go back and repeat it eight times you would find a new things in each
each experiment
Okay, so I don't see any questions right away