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.