Robert's Stochastic thoughts: 2025

Wednesday, February 12, 2025

CAR VII

Here I return to the possibility of devloiping off the shelf killer T-cells with chimeric antigen receptors (CARs). Off the shelf means that the CAR cell does not have to be custom made for each patient using the patient's killer T-cells. I presenting my (nonexpert) thoughts on how to make them here. Now I note an article which explains how off the shelf CAR T-cells were made Matsuzaki et al (2024) Blood volume 144 supplement 1 page 4811.

Recall the problem is that the immune system (mainly the host' killerT-cells) reacts violently to foreing HLA A HLA B and HLA C and while Natural Killer (NK) cells kill cells which don't display HLA proteins. My thought was to delete the genes for HLA A,B and C and also insert an HLA E gene with a strong promoter. HLA E is a don't kil me signal which prevents NK cell killing by bindiong to the receptors NKG2A and NKG2B.

From Matsuzaki et al, I infer that this doesn't work, and that HLA E is not enough to stop NK cells. Matsuzaki et al take another approach, they delete the bene for Beta 2 microglobulin preventing the expression of HLA A, B, C and E (this is actually an idea I heard from my late father who was a very emminent immunologist and pioneer of the immunotherapy of cancer). The HLA proteins have 3 subunits which form a shape like a capital gamma (or a gallows). Beta 2 microglobulin is a separate protein about the size of one of the subunits which nests in they place of a man hanged by the gallows. It is required for expression of the HLA on the cell surface. But it is not key to the functioning of HLA A B and C. Instead beta 2 microglobulin size pieces of all the proteins in the cell replace the beta 2 microglogulin and are displayed where Killer T-cell receptors can bind to them and recognize the HLA antigen bit of another protein complex. If presented by a properuy activated antigen presenting cell (usually a dendritic cell which has spikes maximizing surface area and looks like a sea urchin) the Killer cells goes off and kills all cells which display the HLA bit of another protein antigen. This is very useful if the other protein is a part of a virus and also if the other protein is strrange because it is mutated because the cell making it is cancerous (cancerous cells usually make such "neoantigens" especially if they have a defect in a system which detects missmatched DNA strands are replaces the abnormal DNA with normal DNA from the homologous chromasome gotten from the other parent (this is not just a digression is is very important below).

Matsuzaki et al them make the cells display HLA E and HLA A by making chimeric proteins with normal HLA E and then the protein sequenc continuing with beta 2 microglobulin all attached as one protein and by making chimeric proteins with normal HLA A then the continuing with beta2 microgolbulin. I understand that the HLA A is needed in addition to the HLA E to convince the NK cells not refrain from killing.

This seemed very odd to me. In the end the cells display HLA A and E but not HLA B and C. It seemed much simpler to delete HLA B and C rather than deleting Beta 2 microglobulin (so the cells don't display A,B,C or E) then engineer a modified HLA A and HLA E which don't need beta 2 microglobulin as they have their own attached.

Then I understood. the beta 2 microglobulin covalently attached to the HLA A will not be replaced by the bits of other proteins. Normal HLA A would desplay all of the donor's proteins and the recipients immune system would react (not as ferouciously as to foreign HLA but still react). This is why the immune system must be suppressed with transplants even from HLA matched donors.

The modified HLA A which just won't let go of the beta 2 microglobulin does not display bits of other proteins. Ah yes. brilliant.

There still are, in principle 2 problems. HLA A from different people is different and HLA E from different people is different. In fact there are about 7000 alleles of the HLA E gene (that is 700 kinds of HLA E in different people). This is not a terrible problme because Killer T-cells receptors do not deal with HLA E. It remains true that the donor and recipient must have the same HLA A. A problem but not a huge one. The problem of HLA matching is that (the number of HLA A alleles)(the number of HLA B alleles)(the number of HLA C alleles) is very large. Just the number of HLA A alleles is not so huge.

Consider the analogy implied by "off the shelf" not all off the shelf pants are identical - they come in different sizes, Whjat matters is that the number of sizes (or HLA A alleles) is not too huge.

Matsuzaki et al discuss their off the shelf CAR T-cells in a very specific context (which I don't really understand) reactivating exhausted killer T-cells. THat is very exciting too, but the off the shelf technology seems to me to have very general applications.

Sunday, February 09, 2025

CAR VI I am again tying to figure out how to make CAR T-cells which can overcome reactive oxygen species (ROS) and immune checkpoints, and how to make them cheaply enough that they can be custom made for each patient (as they are). The earlier posts in this series (including one entirely pulled back from a comment by someone who knows a lot more about CAR T-cells than I do) are here , here , here , and here .

Always remember that I don’t know much about CART-cells so don’t take anything I write too seriously. The problem is that killer T-cells with Chimeric Antigen Receptors (CAR T-cells) are an effective treatment of leukemia and lymphoma but do not work well against solid tumors. The problems include the high levels of reactive Oxygen species (ROS think hydrogen peroxide) in solid tumors and immune checkpoints or “don’t kill me” signals which are very useful, but also abused and displayed by cancer cells.

Actual scientists might be able to make CAR T-cells which overcome these problems in two ways: adding genes or deleting genes. The first is used to make a CAR T-cell. A human designed T-cell receptor which is partly made of a part of a monoclonal antibody which sticks to a tumor specific antigen is added to the T-cell. It then activates the T-cell when the tumor antigen binds to the part of the monoclonal antibody.

On the other hand, I (belatedly) noted that at least one gene should really be deleted. CAR T-cells still have their original receptor. Furthermore CAR T-cells are made starting with a large number of unmodified Killer T-cells which have different receptors. This isn’t a problem with current CAR T-cells but would be a problem with CAR T-cells modified to ignore checkpoints, which evolved to prevent our Killer T-cells from attacking us (one of many aspects of the immune system meant to prevent such autoimmunity all of whlch together don’t always work). This means at least one deletion is needed. That would be of the alpha chain of the T-cell receptor (the alpha and beta chains are the variable parts which stick to the antigen and the beta chain is the one modified in the CAR so it is hard to delete the unmodified beta chain without deleting the chimeric antigen receptor. This was done (for another reason) in research described in the paper cited here. The other goals of defending against ROS and inhibiting checkpoints can be managed by adding genes, but the potentially costly need for two different cell modification techniques remains.

That was a long introduction. For those who are still reading, I now think that the ge ne destruction CRISPR-CAS technology can be used to add the chimeric receptor (this has been done) and to delete undesired genes. The amazing technology makes it possible to cut DNA at a very specific site. The cutting anzyme (CAS) is derected to the site by a guide RNA which binds to the DNA sequence where the cut is desired. This can be used to knock out genes. It can also be used to insert a bit of DNA. Basically the DNA to be inserted includes some sequence from before the cut, then the desired insert, then some from after the cut. The normal DNA repair process then pastes in the DNA to be inserted. In this article the use of CRISPR/CAS technology to insert a Chineric Antigen Receptor is described. Importantly the same procedure eliminates the natural unmodified T cell receptor, because the cut and insertion are made there. The problem is partly solved. In fact in the case described it is entirely solved as the CAR T-cell penetrates a solid tumor. This does not happen with all solid tumors (that is natural non CAR T=cells which react with tumor cells do not always penetrate the tumor.

So I think there is still a case for the further use of CRISPE/CAS described in CAR III and in CAR V

The idea presented in CAR III is to delete or inactivate the genes for KEAP1, PDL1, CTLA-4, TIGIT, and the TGF Beta receptor, KEAP1 suppresses the cells natural antioxidant system. PDL1, CTLA-4, TIGIT, and the TGF Beta receptor, are receptors for don’t kill me signals (that is checkpoints). It is necessary to add the Chimeric receptor, and seems necessary to me to add the gene for Herpes TK. Cells with no checkpoints are dangerous. Cells producing Herpes TK can be kill with Ganciclovir. SO two additions and 5 deletions. A lot of work with two systems

The idea in CAR V is to do it all with additions. In particular inhibiting Checkpoins by having the CAR produce the sticky part of antibodies (Fab) which then stick the checkpoint receptor on the CAR T-cells and nearby T-cells (but which have a short half life in the blood so they don’t build up to high and potentially dangerous levels systemically.

The combined inactivate a gene and insert a new gene technology (really pretty classic CRISPR/CAS) makes it possible to do this in 4 steps.

1. Insert the CAR receptoir and inactivate the normal T Cell receptor (this has been done in the cited article)

2. Inactivate the TGF beta receptor and insert Herpes TK

3. Inactivate PD1 and insert the anti TIGIT Fab

4. Inactivate CTLA-4 and inserte Herpes TK to be sure.

I think only 3 steps are really needed. The FDA has agreed that it is ok to infuse two checkpoint inhibiting antibodies so the local not systemic inhibition is needed only for 2 more.

One (of many) things which I don’t know is whether the 4 or 3 modifications can be made at the same time. It seems to me that it should be possible to get 4 different guide RNAs and 4 different pieces of DNA with inserts and the sequences before and after the site of cutting. It seems to me this should make all 4 (or 3) modifications at once. If so, the production of the Super CAR would be almost as cheap as production of an ordinary CAR.

Even if the process has to be repeated 3 or 4 times, it still seems to me much cheaper than having the equipment (and both CRISPR/CAS and modified retrovirus) to do two different procedures.

Again I don’t know much. If you have read so far, I thank you and hope you enjoyed the read

Wednesday, February 05, 2025

CAR V

This post criticizes and presents and alternative to the approach presented in CAR III. update: I am no longer convinced that the approach described here is an improvement, or even as good as the approach presented in CAR III end update In Car III I proposed further genetic manipulation of a CAR T-cell, that is a killer T-cell to which a chimeric antigen receptor which targets cancer cells but not normal cells has been added. The proposed modifications consisted of deleting5 genes. One logistic problem is that the procedure for deleting a gene is different from the procedure for adding a gene. Since the CAR T-cell or further modified CAR T-cell is custom made for each patient, this could be a serious problem.

In any case, this post discussing modifying a CAR T-cell to make it able to function in a tumor micro-environment by adding various genes, one for the CAR receptor, one to make the CAR prepared for reactive oxygen species (ROS think hydrogen peroxide) and 4 to make the cell ignore checkpoints – signals to not kill which have a normal function in healthy people but which can be used by cancer cells to avoid the reaction of the immune system to neo-antigens (new strange proteins basically) which they display.

Our cells handle ROS with various proteins the production of which is stimulated by the transcription facor Nrf2. Nrf2 is bound and inactivated by KEAP1. I proposed deleting the KKEAP1 gene. Amther approach is to add a gene (with a potent promoter and enhancer) which makes lots of Nrf2 so that the normal KEAP1 can’t bind all of it. One issue is that one of the genes stimulated by Nrf2 is the KEAP1 gene (a bit of homeostasis) so the added gene would have to be transcribed more than the normal Nrf2 gene is.

I am not sure that this is easy to manage or strictly necessary. Stimulation of Nrf2 activity in vitro might be enough to give the CAR T-cell a head start. Once the cell is exposed to ROS the natural mechanism (hydrogen peroxide changes the shape of the KEAP-1 protein so it no longer binds Nrf2) might be enough if the CAR T-cell is not quicky killed or inactivated.

Now consider checkpoint inhibitors. Many have been developed and they have been used with dramatic but not wholly satisfying results. They cure a minority of patients (a fact recognized with a Nobel prize) but other patients are not cured. The checkpoints involve receptors on T-cells which tell the cell not to kill (except in one case to kill itself). The receptors are called PD1 CTLA-4 TIGIT and the TGF-beta receptor. The checkpoint inhibitors are monoclonal antibodies which bind to and block the receptor or which bind to and block the molecule which binds to and activates a receptor. As far as I know from my efforts to survey the literature, in clinical trials no more than 2 ckeckpoints are inhibited in any single patient. I assume this is because doctors know that the checkpoints evolved for a reason and fear that blocking too many would lead to autoimmunity (the immune system attacking the normal cells of the patient). I have a crazy idea for inhibition of the checkpoint near the CAR T-cell but not systemically (everywhere in the patient). The idea is for the CAR T-cell to produce blocking antibodies with a short half life, because they are fairly qyuckly removed from the blood and digested by the liver. Most proteins have a short half life for this reason. A key exception is IgG antibodies which have a very long half life because part of the constant region of theantibody signals to the liver to recycle them. The constant region is a tail of the antibody which does not bind the target and which is not highly variable (to make things complicated there are 4 different constant regions of IgG antibodies but nothing like the immense number of variable regions which stick to specific antigens (targets)).

A reduced size molecule called an Fab can bind to and block a receptor even without the constant region. It consists of two proteins stuck together (in genetic engineering they are often one protein which a highly flexible link made of the tiny amino acid glycine which does not get in the way of the two non-glycine parts of the protein aligning and sticking together). These Fab molecules could block receptors of the CAR T-cell which made them and of nearby T-cells without building up to high levels in the blood steam (and lymph) of the patient and making trouble.

An advantage is that nearby T-cells include the patient’s killer T-cells which infiltrate the tumor. This often happens. Also these cells often do not kill the cancer cells because of checkpoints and/or reactive oxygen species. update: This is an advantage of adding genes for anti checkpint Fabs even if some genes, such as part of the normal T-cell receptor have to be deleted. end update. This could be an important added benefit as the tumor would be targeted by different T-cell receptors reacting to different neo antigens, so the cancer cells can’t evade immune surveillance by mutating to lose a specific antigen.

Also the Fab stuck to the receptor on the CAR T-cell or other nearby T-cells will not send the signals that full antibodies send to other immune cells which include “kill this cell” sent to natural killer cells, swallow and kill this cell sent to macrophages and will not fix complement which kills cells. I think this should be an approach which works as well as the checkpoint inhibitors which are currently being used.

update: It is possible that the CAR itself will attack normal cells – CAR T-cells have the added chimeric antigen receptor but they also have the original T-cell receptors. Self reacting killer T-cells should be eliminated in the fetal thymus, but Killer T-cells do sometimes cause autoimmune problems (hence the use of short half live Fab’s) update: this makes it necessary to delete a gene for part of the normal T-cell receptor to prevent attack on patient. That elimiates the original logic of this approach vs that presented in CAR III and explains the update that I am no longer convinced that the approach preseented here is a better appraoch end update. Also one of the cells or all of the cells might become a leukemia (this can and has happened with addition of genes with a retrovirus). For both reasons it would be wise to add the gene for herpes thymidine kinase which makes cells vulnerable to Ganciclovir. end update.

In any case it is an idea which I find interesting.