Overview
On this page, we will introduce the engineering design cycle of our entire project in chronological order. In our project, we improved and refined our design over three full engineering cycles, finally culminating in multi-targeted CAR-NK92 cell library that were invincible in “Whac-a-Mole” game.
In the initial phase of the project,the first engineering cycle,we attempted to get a single immune cell which can express multiple CAR in response to multiple antigen targets due to tumor heterogeneity. However, with the continuous progress of our project, we finally rejected this plan through modeling and literature study.
In the second engineering cycle,we took the opposite approach and attempted to construct a population of CAR-NK cells expressing different types of CAR.At the same time, we did modeling and cell experiments to verify our ideas and achieve initial success.
In the third engineering cycle, we tried to use antibody library technology to further expand the variety of CAR carried by our CAR-NK cells to address unpredictable tumor targets.However, due to the serious pandemic situation in Shanghai, our wet experiment had to be canceled.But! through rigorous and substantial modeling, we still get complete results to confirm the feasibility of our design !
Next,we'll detailedly explain each step of our engineering cycles as follows.
Research 1: How can we deal with multi-target tumors?
Nowadays,traditional treatment including surgery, chemotherapy, and radiotherapy are still recommended more often than novel treatment.However, these treatments not only bring serious side effects, but often fail to solve the metastasis and recurrence of the tumor.Tumor heterogeneity is a major cause of these problems.How to solve the heterogeneity of tumor without bringing side effects to patients has become our research goal.
So far there are no good effective means to deal with tumor heterogeneity. At this moment, an article regarding multi-target CAR-T therapy caught our attention (Han et al., 2019). In this paper, the authors constructed a CAR T cell that simultaneously carries three CAR and achieved good efficacy in experiments.
Imagine 1: Can we build CAR immune cells which can target more antigens?
Inspired by this paper, we imagine that can we build CAR immune cells which can target more antigens? At this point we thought the most direct way was to express multiple CARs on a single immune cell.In this way, as soon as any CAR binds to its antigen, the corresponding CAR immune cells concentrate near the antigen and kill the tumor cells that express it.
Design 1: Expressing multiple CARs on a single immune cell and making amplification.
So we immediately started to design our circuit.The design of this idea is simple. We only need to transfer the scfv sequences of each antibody designed in advance into the same chassis cell to achieve the purpose of expressing multiple antibodies on one immune cell at the same time.When any CAR on a cell is stimulated by a tumor antigen, the stimulated immune cells proliferate and kill the tumor. In this way, the problem of multiple antigen targets caused by tumor heterogeneity is also solved.
(At this stage, the tentative chassis cells are T cells, as there are a number of clinical trials and applications of CAR T cells engineered from T cells.)
Build 1: Using modeling to test our design
However, we are also concerned about the maximum number of foreign genes (CAR) that can be expressed simultaneously in a single cell.So before we started the cell experiment, we first tested it with modeling.
Test 1: Modeling test results
Modeling: Limitations on the number and types of CARs that can be expressed in a single cell.
Studies have shown that multi-targeted CAR-T shows positive efficacy in the treatment of multiple malignancies and becomes an effective way to prevent antigen evasion, but this type of CAR-T cells can be recognized by multiple target antigens, leading to an excessive increase in the level of NFAT activation and over-activation of immune cells, which can easily over-kill normal cells expressing low target antigens. Through bioinformatics analysis, we derived the relationship between the number of types of CARs expressed by individual immune cells and the rate of antigen evasion and over-misfiring of normal cells. Taking colon cancer cells as an example, the number of CAR types should be controlled within 4-8 types.
Furthermore, research has shown that transfection of multiple CARs on a single cell is challenging and expression of multiple CARs on a single cell is prone to ligand non-dependent tonic signaling.
The result of our model showed that expressing multiple CARS on a single immune cell,especially when the number of CAR types is large, is not feasible.
Learn 1: Expert advice and online literature study
Then we consulted Dr.guo, director of immunology Department of Naval Medical University, about our ideas and modeling results.He shared with us his experience in CAR research : In general, the human immune system works in concert. If you want to express multiple CARs on a single immune cell,you have to consider the stimulation of individual CAR on downstream signaling pathways.In other words, if only one or two CARs on these membranes are activated by tumor antigens, they do not form a "cluster effect" that activates downstream signals less strongly, and the immune cells are less capable of killing tumor cells.And it's not clear how many side effects this design might have, given the combined suppression of the tumor microenvironment and other immune cells.
At the same time, he also put forward an opposite way of our design : Can multiple CARs be expressed on many immune cells? And expressing only one CAR per immune cell?
At the same time, we searched online for literature on the feasibility of this design : Although many research data shows, more targeted CAR - T the curative effect of treatment of various malignant tumors showed positive, become the effective means to prevent antigen to escape, but this type of CAR - T cells can be multiple target antigen recognition and high strength activation of T cells, easy to cause a cytokine storm, and the lower expression of normal cells of the target antigen excessive destruction, Therefore, further optimization in structural design and function is needed to effectively play the role of tumor killing under the guarantee of safety.In addition, in the development process of CAR T cells, the positive rate of CAR transfection is a required test item for the quality control of CAR T cells. However, for multi-target CAR T cells, the expression of 2-3 CARS needs to be detected. Compared with single CAR, the detection of multiple Cars is more difficult. The use of different secondary antibody staining, prone to non-specific background strong, two CAR difficult to distinguish and other problems.The above information further confirms the infeasibility of our first design.
Improve 1: Try the opposite way to achieve our goal.
After listening to Dr.guo's advice, we began to try the opposite way to solve the tumor heterogenity ---expressing multiple CARs on a large number of immune cells but each immune cell expresses only one kind of CAR.
Research 2: How can we implement our new idea?
Expressing different kinds of CAR on a large number of immune cells is not an easy task.Through the meeting and discussion, we identified two important issues to be solved.
First,because the number of engineered immune cells we can transfuse to patients is limited, the number of immune cells expressing each CAR will be small.It is possible that even after the immune cells come into contact with the corresponding antigen, they are unable to proliferate due to insufficient cardinal number.
Second, due to the characteristics of cellular immunotherapy, in order to ensure its safe and controllable in human body, we need to design an apoptotic pathway to control their expansion in vivo.
Imagine 2: The selection of chassis cell and the effect of gene circuit
The selection of chassis cell and the effect of gene circuit are the key to solve the above two problems.
the selection of chassis cells :
In the course of cellular immunotherapy, the chassis cells commonly used are T cells.However, T cells are not the best candidates because their proliferative capacity is not very prominent, and T cells can cause severe GVHD and CRS if they accumulate in large numbers in the body.(Zhao et al., 2021)
So we looked at the T cell's neighbor, the NK cell.NK cells have a number of characteristics that T cells do not have: First, allogeneic NK cells did not induce GVHD; Secondly, NK cells do not secrete inflammatory cytokines; Thirdly, in addition to CAR targeted killing, CAR-NK cells can also nonspecifically recognize tumor cells to improve the effect of immunotherapy. Finally, allogeneic NK cells come from a wide range of sources, including peripheral blood, NK cell lines, cord blood, etc.(Valipour et al., 2019)
These features not only make allogeneic NK cell transplantation safer, but also enable our product to become an off-the-shelf product, bypassing the personalized customization process of traditional CAR T products, thus effectively reducing costs ! (Please see Cost analysis for more details!)
Finally, we selected the tumor-derived NK-92 cell line as our chassis cells because we needed our chassis cells to have strong proliferative capacity. The NK92 cell line, derived from a patient with acute non-Hodgkin's lymphoma, has the advantage of being able to proliferate indefinitely.(Suck et al., 2016)
The effect of gene circuit :
Due to the strong proliferative capacity of the NK-92 cell line, we need to design a suicide circuit to ensure the safety of our product.
So we imagined that the final effect of our gene circuit would be something like this : CAR-NK92 cells that recognize cancer cells survived and proliferated,while CAR-NK92 cells that did not recognize the antigen were induced to apoptosis.
Design 2: The design of gene circuit
So we initially designed three interconnected genetic circuits
CAR structure
Gal4-KRAB---UAS promoter inhibitory protein
iCASP9---Induced suicide switch
※ Engineering optimization
Although the three gene circuits designed by us can theoretically operate independently of each other and do not interfere with each other, the success rate of transfection will be relatively low because the three lines need to be transduced by three independent plasmids.
So we started to think about whether we could combine the latter two gene circuits to achieve engineering optimization.
So we designed the following optimization scheme
we generated the opposite construct, in which the two expression cassettes were cloned in such a manner that both promoters were at opposite ends and at a long distance.The expression of Gal4-Krab does not inhibit P-NFAT because the two promoters are respectively on the positive and negative strands of DNA.The number of plasmids we use is reduced by one and the success rate of transduction is greatly improved, reflecting our engineering optimization.
Build 2: Building a small CAR-NK92 library
Next, in order to verify the feasibility of our designed circuit, we buillt a CAR-NK92 library containing ten kinds of established CARs.
We used this part (BBa_K4421003) to gain ten kinds of CAR.This part is a phage display scfv library which consists of ten kinds of individual colonies:BBa_K4421011, BBa_K4421012, BBa_K4421013, BBa_K4421014, BBa_K4421015, BBa_K4421016, BBa_K4421017， BBa_K4421018, BBa_K4421019 and BBa_K4421020.
Test 2: Cell experiments and mathematical modeling
Cell experiments
1.Verification of specific killing effect of CAR-NK92 cells
The specific killing effect was confirmed using fluorescent target array killing assays with MCF-7 cells and derivative cells.
2.Verification of the ability of AP1903 to cause apoptosis when the engineered NK cells are not stimulated
Our data showed that AP1903 had no effect on the viability of NK-92 cells but caused apoptosis/necrosis of the engineered NK92 cells with no stimulatory cells. However, after 24h of stimulation with specific antigens, the engineered NK92 cells were notably affected by the reduced pro-apoptotic effect of AP1903, and both the EGFR-specific and HER2-specific CAR-NK92cells were resistant to AP1903 after antigen stimulation for 72h.
3.Verification of the growth dynamics of logic-gated NK-92 cells
The gene circuit was effective in the cell-based CAR screening method and adding AP1903 was essential for the enrichment of CAR NK-92 cells.
mathematical modeling
1.CAR-NK92 suicide gene pathway feasibility model
Based on the original McKeithan kinetic proofreading model, our CAR activation model was developed by relaxing various constraints, including the range of dephosphorylation rate and the order of phosphorylation and dephosphorylation, to obtain a more consistent phosphorylation model for CAR. In terms of the results, the biological properties of CAR-NK92 cells in terms of sensitivity, specificity and time efficiency in the recognition of exogenous molecules were reflected more successfully and realistically. It has implications for predicting the activity of CAR-NK92 cell gene circuits in animal and clinical experiments.
2.AP1903 pro-apoptosis model
Optimize the linear model equation and obtain a relatively complete effect change equation to describe the relationship between efficacy, time and concentration of AP1903.
(the linear model equation)
(a relatively complete effect change equation)
When the initial concentration was set at 0.05, 0.1, 0.5 and 1.0mg/kg and AP1903 was injected at an interval of 12h, the relationship of plasma pharmacodynamics with time was observed.
When the initial concentration was set at 0.05, 0.1, 0.5 and 1.0mg/kg and AP1903 was injected at an interval of 24h, the relationship of plasma pharmacodynamics with time was observed.
3.CAR-NK92 proliferation model
Build Lotka-Volterra predator-prey model，importing logistic elements
(Picture indicates how their numbers change over time)
In these experiments, we determined the kinetics of the genetic circuit in which NFAT was coupled with KRAB activation in engineered NK-92 cells after CAR stimulation.The effectiveness of the circuit was verified by both cell experiments and mathematical modeling
Learn 2: Reflection and Summary
Through the above experiments and modeling verification, we preliminarily proved the feasibility of our design.
Our ultimate goal is to solve the problem of multiple antigen targets caused by tumor heterogeneity.To accomplish this, we also need to determine how many CARs are needed in our constructed CAR library to ensure that CAR-NK92 recognizes all tumor antigens (both native and mutated neoantigens).
Through our previous bioinformatic analysis, we were able to determine the number of tumor-surface antigens that were upregulated in tumor cells.(The bioinformatic analysis results are listed below with heat maps, volcano maps, and tables.Please see Model for more details !) The results show that there are more than a thousand antigens that may be highly expressed on the surface of tumor cell membranes (including many different tumor models).
Therefore, only relying on the prespecified ten kinds of CAR in our previous cell experiments is far from enough to cope with the huge antigen repertoire.Moreover, because of the uncertainty of tumor cell mutation, we can't predict what the new antigen will be like.
Improve 2: How do we get a CAR library consisting of more CARs?
We have just analyzed the type and quantity of tumor surface antigens by bioinformatics,but the key question for these surface antigens is how many CARs should be used to cover the antigen repertoire and How do we get a CAR library consisting of more CARs.
※In our project, we envision a giant hammer, our CAR library, that can cover all tumor antigen.No matter what antigen the tumor cells produce, there will always be a CAR that recognizes the tumor antigen.Just like whack-a-mole, no matter where the moles come from, this giant hammer can always hit them.
Research 3:Diversity of antibodies
Then we further studied and learned about the knowledge related to antibodies.We learned that Burnett's clonal selection theory could explain the diversity of antibodies.
This theory holds that there are many clones of immunocompetent cells in animals, and the cells of different clones have different surface receptors and can complement each other with the corresponding antigenic determinants. Once an antigen enters the body and binds to the receptor of the corresponding clone, it selectively activates the clone, causing it to expand and produce a large number of antibodies (i.e., immunoglobulins), molecules of the same specificity as the receptor on the surface of the selected cell (Cohn et al., 2007).All in all, the human body can produce at least billions of antibodies against foreign antigens (including tumor antigens).
Imagine 3: Using a whole library of antibodies,is it possible?
Therefore, following the clue of clonal selection theory, we conducted in-depth research on the concept of "antibody library".
The antibody library technique is actually a very old technology.In a narrow sense, the so-called antibody library technology is a technology that uses gene cloning technology to clone the full set of variable region genes of heavy and light chains of antibodies, and then recombine them into specific prokaryotic expression vectors, and then transform Escherichia coli to express functional antibody molecular fragments, and obtain specific variable region genes of antibodies through affinity screening. The antibody genes screened by antibody library technology will be used to construct and express genetically engineered antibodies.
At present, the main application of antibody library technology is to screen a variety of important antibodies from antibody library, such as membrane protein antigen, self antigen, virus antigen antibodies, for clinical treatment or to modify existing antigens to improve the affinity.
But few people would have thought of using an entire library of antibodies without screening for specific antibodies.Therefore,we imagined that we could use the antibody library technology to combine the sequences of all the light and heavy chains of antibodies to make scfv, and then express them on NK-92 cells to make CAR-NK92, so as to truly realize the CAR library.In this way, both native tumor antigens and newly mutated antigens can be recognized by one of the CARs in the CAR-NK92 library.
Design 3: Antibody library Vs Antigen library
Based on our previous modeling and data from the literature, it is generally believed that millions of kinds of CARs are sufficient to recognize all possible tumor antigens.
In terms of gene circuit design, we continue to use the circuit designed in the second cycle.
Although the antibody library technology has matured, the digital antibody library on the Internet has not yet been realized, so we would use the phage antibody library to make the CAR-NK92 library.We need to build a library with a million kinds of antibodies that we can use before we actually use them in clinical practice.
Build and Test 3: Cellular automata-Visual dynamic simulation of the CAR-NK92 library competing with tumour cells
At this stage, due to the impact of the epidemic in Shanghai and our concern for the safety of animal experiments, we cancelled the original experiments in vivo.
But we can still rely on modeling to prove the effectiveness of our CAR-NK92 library !
Our ultimate goal is to address the many neoantigens that come with tumor heterogeneity.Since the process of tumor mutation and proliferation is very similar to the principle of cellular automata, we decided to use cellular automata for modeling.So our team modeled the killing effect of 10^6 kinds of CAR-NK92 cells against tumors and their mutants.
We also draw a whole flow chart to illustrate this process The results showed that a very small number of tumor cells survived at the end in most simulations !
Learn 3: Is it enough for our design?
Through the above design, experiment and proof, we have tentatively constructed the seemingly invincible CAR-NK92 library.Through meeting after meeting, we found that there was a fatal flaw in our project.Just building a CAR-NK92 library that contained all the antibodies could not be used directly for therapy.The reason is that the antibody library contains almost all kinds of antibodies, so there must also be antibodies that can bind with HLA antigens in normal human tissues.This can cause a severe allergic reaction, which can be very dangerous.
Under the guidance of our mentor, we realized that the solution to this problem was not as difficult as we thought.Using phage display technology can solve this problem.(Phage display technology is a biotechnology in which the DNA sequence of foreign protein or polypeptide is inserted into the appropriate position of the structural gene of phage coat protein, so that the foreign gene can be expressed with the expression of coat protein, and the foreign protein can be displayed on the surface of phage.)
The derived phage display antibody library technology can display several different antibodies simultaneously on the phage surface. As long as we use healthy human blood cells mixed with the phage display library, after several screening, we can get a phage antibody library that will not bind with healthy human HLA antigens.
Such a screening process, consistent with the body's own production of mature T cells, is called "negative selection".
Improve 3: Outlook on the future of this project
Despite the long way to go from experiment to clinic, we designed a preliminary outlook on the future of this project.
1. Draw 200 copies of peripheral blood from healthy volunteers.
2. Extract mRNAs encoding antibodies in B cells and use RT-PCR and SOE-PCR to form an scFv cDNA library.
3. Select leukocytes from volunteers on a large scale and obtain their HLA antigens as material for negative selection.
4. After negative selection, reserve the negative ones to form the final library that can be put into use.
5. Before the patient is treated, confirm that the scFv library has no affinity to the patient's HLA antigen.
6. Construct CAR-NK92 cells carrying the scFv library and infuse them into the patient for treatment.
Engineering Success !
References
1. Han, X., Wang, Y., Wei, J., & Han, W. (2019). Multi-antigen-targeted chimeric antigen receptor T cells for cancer therapy. Journal of Hematology & Oncology, 12, 128. https://doi.org/10.1186/s13045-019-0813-7
2. Zhao, C., Zhang, Y., & Zheng, H. (2021). The Effects of Interferons on Allogeneic T Cell Response in GVHD: The Multifaced Biology and Epigenetic Regulations. Frontiers in Immunology, 12, 717540. https://doi.org/10.3389/fimmu.2021.717540
3. Valipour, B., Velaei, K., Abedelahi, A., Karimipour, M., Darabi, M., & Charoudeh, H. N. (2019). NK cells: An attractive candidate for cancer therapy. Journal of Cellular Physiology, 234(11), 19352–19365. https://doi.org/10.1002/jcp.28657
4. Suck, G., Odendahl, M., Nowakowska, P., Seidl, C., Wels, W. S., Klingemann, H. G., & Tonn, T. (2016). NK-92: An “off-the-shelf therapeutic” for adoptive natural killer cell-based cancer immunotherapy. Cancer Immunology, Immunotherapy: CII, 65(4), 485–492. https://doi.org/10.1007/s00262-015-1761-x
5. Cohn, M., Mitchison, N. A., Paul, W. E., Silverstein, A. M., Talmage, D. W., & Weigert, M. (2007). Reflections on the clonal-selection theory. Nature Reviews. Immunology, 7(10), 823–830. https://doi.org/10.1038/nri2177