Summary of General Meeting Presentation 7-25-12
2015-01-13azim58 - Summary of General Meeting Presentation 7-25-12
Summary of 7-25-12 General Meeting Powerpoint
I’ll talk about discovering tumor specific antigens which are proteins
unique to cancer that the immune system can recognize.
We know that vaccines are an effective method for inducing the immune
system to attack specific targets. To make a vaccine, you just need to
identify certain proteins unique to a pathogen, and then inject these
into an organism along with any adjuvants which raise the immune
response. We have already eradicated smallpox from the planet, and polio
has nearly been eradicated. There are still many diseases for which we
don’t have effective vaccines such as AIDS and other infectious
diseases such as malaria. Cancer is another disease which we don’t have
effective vaccines for, but there have been promising results. If you
vaccinate a mouse with a self tumor mixutre, whole cell tumor lysate, or
a tumor associated antigen, some level of protection against that same
type of cancer can be observed. Now we need to discover more tumor
specific antigens to make really robust cancer vaccines.
Here is the method I plan to use to discover tumor specific antigens. The
first step is to construct a cDNA library from the tumor of a mouse. RNA
is extracted from the tumor, reverse transcribed to DNA, and cloned into
plasmids. These plasmids can then be transformed into bacteria, and these
bacteria will then produce proteins found in the original tumor. At this
stage, I then want to find out which tumor proteins the immune system of
that mouse can recognize. To do this I can take antibodies found in the
blood of the mouse, and determine whether they bind any tumor proteins in
my library.
There are also some additional steps that could be taken in order to
enhance this discovery process. For example instead of just using
antibodies, I could create a phage library from the RNA of the B cells
which make the antibodies. Phage are viruses that infect bacteria. When I
put the antibody genes into the phage, they can express antibodies or
mini forms of antibodies on their surface. The advantage of doing this is
that I can produce as much antibody as I would like. Another additional
step which may help me discover tumor specific antigens is to purify the
antibody libraries using the random peptide arrays. When I screen my cDNA
library with a smaller set of antibodies, I may obtain fewer false
positives.
Once a tumor specific protein is found, this can be vaccinated into a
mouse and tested for immunogenicity.
Here is a timeline to show some of the work performed at the beginning
stages of my library construction. Mice were injected with mouse breast
tumor cells (7000 4T1 cells), and later on the spleens and tumors were
collected from these mice. RNA was extracted from the tumor to create the
tumor cDNA expression library. The spleen was used to obtain B cells. The
RNA from these B cells was then used to create a phage antibody library.
In addition to creating a tumor phage library, I also immunized mice with
a protein which Shen works with frequently, SMC1fs. I use the samples
collected from these mice for many of my positive controls since I know
which protein exactly that these mice produced an immune response against.
This ELISA graph shows that all of the mice in the group truly did
produce antibodies against the SMC1fs protein, and I use the antibodies
in later experiments in this presentation.
This is an outline of how the cDNA library was constructed from the RNA
of the tumor tissue. This kit has some neat features such as the
switching mechansim at the 5' end of the RNA template, which allows full
length RNA transcripts to be reverse transcribed into DNA instead of just
truncated products. This Clontech kit that I used also makes use of
In-Fusion technology which allows fragments of DNA to be joined together
if they have 15 bp of homology at each end. Therefore, it is not
necessary to perform a traditional ligation which is not as efficient.
Once I had produced my tumor cDNA library, I wanted to make sure that the
library was not too biased. Sometimes when people make cDNA libraries
there can be many repeats of the same gene, or there could be just very
short truncated genes, and therefore there would not be a good
representation of all of the different transcripts that were originally
in the tumor cell. On this gel I looked at the sizes of the genes from 15
different colonies chosen at random. There are many different sizes here
so the library looks like it is not too biased. Ideally it would have
been better to screen more than 15 colonies, and I plan to do that when I
make a human cDNA library in the future. I also sequenced these genes,
and found that there was not too much bias. Due to some features of the
kit I used, I actually only expected 1/6 of all of my cDNA library genes
to be in the correct orientation and in frame. When I examined the 15
sequences I have here I did find that about 1/2 of the genes were unique
full length genes in the correct orientation. About 1/3 of those are
expected to be in the correct frame which leaves me with about 1/6
correct genes. If I had screened more than 15 colonies than I probably
would have seen a number lower than 1/6 (maybe 1/10) since there would be
some duplications. These out of frame and out of orientation problems are
also issues that I will correct when I make the human cDNA library.
Now that I had made my cDNA library, I could have bacteria produce the
proteins, and then screen the library for antibody binding. Normally,
people will screen tumor cDNA library by spreading colonies out onto
these large square plates testing sera to see if it binds any colonies.
This is a very cumbersome process, and if I wanted to screen 1 million
clones with 3X representation I would need 450 of these plates which is
not very practical. Therefore, we wanted to see if we could screen
colonies on small arrays. We ended up printing the lysate onto
nitrocellulose arrays.
Next I had the bacteria produce the tumor proteins, and then lysed the
cells. The cell lysate with the tumor proteins was spotted onto
nitrocellulose slides, which are about the same size as the 10K random
peptide slides we use. I wanted to see how dilute the protein could be
and still be detectable. Here you can see the SMC1fs spots on the array
with their different dilutions. In this graph you can see that I can
detect a 1/1000 dilution of the SMC1fs lysate into regular lysate, and
the signal is still detectable over lysate alone. This is important since
I later planned to print the tumor cDNA library so that there was 1000
colonies in each spot.
I also performed a titration with secondary antibody with sera and
without sera to see at what level I could still detect signal from the
sera, while the secondary signal alone was much lower. With the random
peptide arrays I normally always use 5 nM, but with these arrays, 1 nM
appears to be more optimal. Also, the intensities on these nitrocellulose
slides is always much lower than the intensities I see on the random
peptide arrays, but the standard deviation of signals on the array is
always very small and consistent. These results suggest that the
nitrocellulose slides have a very different behavior.
Since the array surface and conditions seemed to be working reasonably
well, I could try to print a whole library onto an array. I set
everything up so that there would be 1000 colonies per spot with 3000
spots per slide. This would require 32 96 well plates which would later
be transferred to 8 384 well plates. If I identified a spot which had
high binding, I could then spread this spot containing 1000 colonies out
onto a whole new array and then screen that to determine which spot
contained the tumor protein of interest. This protein could be a
potential protein for a tumor vaccine.
In order to accomplish this printing of the library with so many plates,
I used several high throughput methods. For example, I used the HiGro
shaker to shake many plates with bacterial cultures at once.
I used the Biomek robot to automatically transfer and mix liquids
necessary to produce the protein.
Here’s another picture of the Biomek robot.
I also used this program called ImageJ to automatically count bacteria
colonies on plates which I needed to determine to make appropriate
dilutions for my cultures.
Here is what my whole cDNA library looks like from the mice with the
tumor on the slide when screened with different sera. On the right you
can see what the slide actually looks like.
Here is the tumor sera and the SMC1fs sera on the slide. Of course the
SMC1fs sera has much more intensity across the whole array since this was
a hyperimmunized mouse. Also note that I intentionally put a few spots on
the array which contained the SMC1fs protein, and these spots had very
high intensity as expected.
This slide just shows that the sample correlations between the slides
were good. In the table on the right, I show that not every sample
correlates well with any other, just to its duplicate.
This bar graph shows that the SMC1fs spot is detected very well be SMC1fs
sera, but not as well by naïve sera. The intensity is on the y axis. The
PUC19 negative control spot does not show any significant difference at
all between the two types of sera. Therefore, the controls for this
library screening are behaving as expected.
Here are the top 10 tumor lysate spots that exhibited some type of change
from the intensity with the naïve sera to the intensity with the tumor
sera. Notice that most of these top changing spots are all spots that
decreased in intensity from naïve to tumor. So the tumor has lower
intensity.
Here are the top 5 tumor lysate spots that have a higher intensity in the
tumor sample.
I then chose my very best increasing tumor spot, and I expanded this out
onto a whole new array. On this new array, instead of 1000 different
tumor protein clones, there was just about 1 clone per spot. I then
screened the library with the tumor sera, and here are the spots with the
best p values from naïve to tumor. There are 3072 spots so one potential
cutoff p value is 1/3072 which is 0.000326. Unfortunately, I only had 1
spot below this p value. However, I ended up continuing with each of
these 3 spots. These spots each contain one unique colony ideally, but
this was determined by diluting a large culture. Therefore, in reality
each spot just contains approximately one colony. I spread colonies out
from each of these spots (from the glycerol stock, not from the lysate of
course) onto plates, and then sequenced several colonies (about 10). From
these sequencing results, I found that each spot contained about 2-3
colonies. I then produced protein from each of these colonies for a total
of about 8. I now have this protein lysate, and this week I plan to print
this lysate onto nitrocellulose slides, and screen with tumor sera. I
think there is a possibility that I may not discover a tumor protein from
this round of screening since the initial 1000 colony spot that I chose
to expand could have been a false positive. There is also a possibility
that the colony of interest in the 1000 colony spot did not make it into
my new array with one colony per spot just by chance. Screening the
individual colony array may not have gone well also since I have only
done this one time so far. There are many things that could have gone
wrong. Nevertheless, there is also a chance that I will screen the
proteins from these spots, and find some protein that binds the tumor
sera more than the naïve sera. So I’ll see what happens.
Now I would like to talk about producing a phage antibody library. You
can see which part of my original scheme this refers to here. The benefit
of producing a phage library is that you can produce as much phage as you
want, whereas you cannot obtain as much antibody as you want from one
patient. Here is a cartoon of what a phage would look like. They have a
capsid containing DNA like many viruses. Then on one end they would have
3-5 proteins sticking out which contain mini forms of antibody. These
mini antibodies are called scFv which stands for single chain Fragment
variable region. They are called this because only the variable antigen
binding part of the antibody is included into the genes of the phage to
express on the surface. In order to get these genes from the B cells into
the phage you have to do a lot of PCRs to amplify and link genes.
This diagram here is kind of small, but I just wanted to give you a
general impression of what is involved to make these scFv genes. First
you need to amplify the variable domains from the light chain and heavy
chain genes of the antibody. You may recall that an antibody has four
different chains: two identical light chains, and two identical heavy
chains. I then amplified these initial PCR products, and then linked them
together for my final scFv genes. This whole library of different scFv
fragments could then be cloned into a plasmid.
Here is a gel showing that I had correctly produced my scFv genes, which
are expected to be about 750 bp, which these fragments are. I produced
the scFv libraries from SMC1 immunized mouse, naïve mice, 1-78 immunized
mice (1-78 is just another frameshift protein similar to SMC1fs),
and the tumor mice.
Now that I had my scFv, I could clone them into a plasmid. I actually
spent a lot of time trying to clone these genes into a plasmid we
received from a lab, but then I eventually found out that this plasmid
did not have the sequence that I thought it did. Now we have received a
new plasmid from the Scripps Research Institute which seems to be
behaving much better. However, I can’t immediately clone the scFv genes
into this plasmid. This plasmid has some extra restriction sites that we
don’t want, and doesn’t have some restriction sites that we do want.
Therefore, Andrey and I planned how we wanted to modify the plasmid.
I’ve made some of the modifications already, and there’s just one
more thing left for me to verify that I’ve done, and then I’ll be
able to clone my genes into the plasmid. I’m not going to go into any
of the details about these modifications here though. Once the scFv are
cloned into the plasmid (actually phagemid), then I can transform them
into bacteria, and have the bacteria start producing the phage.
Here is my last topic. I will discuss how I might purify antibodies using
the random peptide array before screening the cDNA library. If I use
purified antibodies, I might be able to obtain much stronger signals, and
fewer false positives. The purification is possible because we know that
there are some random peptides which specifically bind to tumor sera
compared to naïve sera. Therefore, these random peptides can be used to
bind and elute just the antibodies of interest.
Here is some data from an experiment in which I applied tumor and naïve
sera to the random peptide array as people have done many times in our
lab. However, no one has done it with my particular mouse sera which came
from the mouse that I made my cDNA library with. Here you can see that
the tumor sera intensity is slightly higher than the naïve sera which is
exactly the opposite result I observed on my cell lysate arrays. The
correlations are also reasonable, but the intensities are a little low.
Here are the peptides which had the best p values when comparing the
intensities with the tumor sera compared with the naïve sera. We chose
to have these 6 peptides synthesized onto tentagel beads so that we could
put the beads in a column, and purify the tumor specific antibodies.
After the antibodies were purified, I put them back onto the array. I
expected that the purified and eluted antibodies would bind to the same
peptides used to purify them. However, my purification did not work.
These graphs show that the flow through which did not bind to the
peptides could still bind to the 6 peptides on the array, but the
purified fraction did not bind to these 6 peptides. Therefore, we had to
rethink how to do the purification. After some discussion we felt, that
the major issue was that the peptides on the array are at a much higher
density than the peptides on the Tentagel mesh. However, the orientation
of the peptides may also be an issue.
In our lab we have 3 different types of random peptide arrays. The v1
slides have the array bound to L amino acid peptides (the naturally
occurring sterioisomers) in a C to N orientation. Therefore the carboxy
terminus is closest to this maleimide group bound to the amine on the
serface, and the free amino group at the N terminus is farthest from the
array surface. However, in version 2, the peptide is still in the L form
but runs N to C. Version 3 also runs N to C, but the amino acids are in
the unnatural D form. Our lab produced an array like this to choose
peptides which the body would have a hard time breaking down when they
were incorporated into synbodies. When I screened my tumor sera, I
screened on the version 2 array. This presents a problem though.
When peptides are synthesized directly onto the Tentagel beads, they can
only be synthesized from C to N like the peptides we have on the v1
arrays. However, I wanted my peptide to be from N to C like on the v2
arrays. There is a possibility that an antibody would still bind to the
peptide regardless of the direction of the peptide since it may just
recognize some specific adjacent functional groups. However, there is
also a possibility that it could matter very much. Therefore, to keep
everything consistent, the next time I do an antibody purification, we
will try to make sure that the antibodies are in the correct orientation.
In order to do this with the Tentagel beads, the peptide first needs to
be synthesized which Donnie just finished a day or two ago. Then it needs
to be HPLC purified which Fatyon did the other day. Now I need to add a
maleimide group to the Tentagel bead mesh, and then conjugate the
correctly oriented peptide to this maleimide group.
In addition to getting the peptide in the correct orientation, we also
plan to increase the density of the peptide 4 fold. There’s a trick
that Zhao and Donnie know about which allows four peptides to be
synthesized from one original amino acid when two lysines are at the
beginning of the sequence. Lysines have to amines which can be used to
bind to two more things. Therefore, in this scheme, one lysine binds to
two others. Then each of these two lysines can bind to the maleimide
group which can be used to conjugate to the N to C oriented peptide. Once
I reach this point, I can try to purify antibodies from the sera once
again, and then see if the different fractions bind to these 6 peptide
sequences on the array.