October 20, 2021

Medical Trend

Medical News and Medical Resources

Challenges: Antigen recognition combination in Glioblastoma T cell therapy

Antigen recognition combination to overcome specificity, heterogeneity and durability challenges in Glioblastoma T cell therapy


Challenges: Antigen recognition combination in Glioblastoma T cell therapy.  Although chimeric antigen receptor (CAR) T cells have shown remarkable results in the treatment of hematological malignancies (1), the development of effective CAR T treatments for solid cancers is still a challenge, to a large extent This is because it is difficult to determine the best target surface antigen.

Very few antigens are true tumor-specific antigens, and cross-reaction with normal tissue targets/extra-tumor can cause fatal toxicity (2-5). In addition, even if highly tumor-specific antigens are identified, these targets are often expressed heterogeneously, and selective CAR targeting can allow antigen-negative tumor cells to escape (6).

Therefore, novel tumor identification strategies are generally needed, which can cope with the dual challenges of specificity and heterogeneity in order to expand the therapeutic window for safely and effectively attacking solid cancers.

A specific example of this dual challenge was found in glioblastoma (GBM) (Figure 1A). Epidermal growth factor (EGFRvIII) is a highly GBM specific neoantigen, which exists in some GBM patients (7-10). However, in previous clinical studies, targeting GBM with EGFRvIII CAR resulted in tumor recurrence.

Although EGFRvIII+ cells were effectively killed, the high heterogeneity of EGFRvIII expression allowed EGFRvIII-tumor cells to escape (6, 11, 12) . In contrast, alternative glioma-related surface antigens, including Aphrin type A receptor 2 (EphA2) and IL13 receptor α2 (IL13Rα2), are expressed in most GBM cells (13-15), but their specificity is not perfect .

Although they are not expressed in normal brain tissues, they are expressed at low levels in some non-tumor and non-brain tissues (13-15). All in all, it is challenging to find a single, specific and homogeneous ideal surface GBM antigen.

T cells that recognize multiple antigen combinations provide possible solutions to the problem of antigen heterogeneity and specificity. We have previously developed a “start and kill” circuit, in which synNotch receptor (an engineered receptor, when it recognizes its homologous antigen, activates transcriptional output (16), and induces the expression of CAR against the killer antigen (17, 18). Here we hypothesize that through careful selection of initiating and killing antigens (synNotch and CAR ligands, respectively), such a loop may lead to hybrid recognition behavior, which may provide a gap between antigen specificity and heterogeneity Ways to make trade-offs.

Challenges: Antigen recognition combination in Glioblastoma T cell therapy Challenges: Antigen recognition combination in Glioblastoma T cell therapy


We use two different strategies-priming with tumor-specific but heterogeneous antigens (such as EGFRvIII) (Figure 1B-E) or brain-specific antigens (such as myelin oligodendrocyte glycoprotein (MOG)) (Figure 1B-E) 1F-H). The promoters triggered by these antigens were then used to locally induce the expression of CARs that recognize the more uniform antigens EphA2 and IL13Rα2. The tumor specificity of EphA2 and IL13Rα2 is imperfect, making them non-ideal targets for conventional single-target CAR T cell therapy. However, if the priming antigens provide higher tumor selectivity, these antigens can be used as effective killing targets.

We hypothesize that T cells engineered with priming and killing circuits can induce local CAR-driven cytotoxicity, confined to the vicinity of the priming cell, thereby avoiding random killing in distant normal tissues expressing killer antigens but lacking priming antigens . This type of circuit spatially integrates the recognition of two imperfect but complementary antigen targets: the priming antigen provides specificity, and the killing antigen ensures the uniformity of the treatment attack. In all these circuits, in order to achieve uniform killing and further reduce the possibility of tumor escape, we used tandem CARs that simultaneously target two killing antigens EphA2 or IL13Rα2 (13, 19). Tandem CAR is used as an OR gate, and its extracellular region contains an α-EphA2 single-chain antibody and an IL13 mutein (a variant of IL13 ligand that has a higher affinity for IL13α2 compared to IL13Rα1) (19, 20).

In order to overcome this heterogeneity, a key question is whether the sensitized and killer T cells triggered by one cell (EGFRvIII+) can kill another adjacent target cell (EGFRvIII-) – we call this the counter-priming process/ Killing (Figure 1C) (cis-priming/killing will describe priming and killing based on antigens present on the same cell). For the first test of transkilling, we used U87 “starter” GBM cells engineered to stably express the starter antigen EGFRvIII and natural U87 “target” cells (endogenously express EphA2 and IL13Rα2 killer antigens, but are negative for EGFRvIII) (11, 12, 21, 22). We mixed U87-EGFRvIII+ and U87-EGFRvIII- cells in different ratios to recapitulate the different levels of heterogeneity observed in GBM patients (10-100% of starter cells), and then tested whether the presence of starter cells The killing of target cells was induced (Figure 1D, E).

We found that CD8+ T cells modified with α-EGFRvIIIsynNotchàαIL13Rα2/EphA2 CAR priming and killing circuit can effectively kill EGFRvIII-target cells in vitro, even as low as 10% EGFRvIII+ priming cells (Figure 1D, E). In contrast, in the absence of priming cells, no killing of EGFRvIII-target cells was observed. In these analyses, we tracked the kinetics of killing two different tumor cell populations (EGFRvIII- and EGFRvIII+) within 72 hours (see video S1 for the passage of killing time). Effective killing was observed with as low as 10% of the starter cells, although it was slightly slower than the killing observed with 50% of the starter cells (p=.0149; t-test). We also verified the effectiveness of trans-killing with model antigens, thus showing the robustness of trans-priming/killing (Figure S2 A-F). All these in vitro killing studies have shown that priming/killing is effectively carried out, and the EGFRvIII-induced circuit therefore represents a promising strategy to prevent tumor escape due to heterogeneity.

We also hypothesized that T cells can also be locally triggered by recognizing tissue-specific antigens (expressed on non-malignant cells) (Figure 1F). For example, in the case of GBM, we might design a T cell circuit triggered by brain-specific antigens, and then trigger local killing by inducing CARs against GBM antigens EphA2 and IL-13Rα2. Therefore, the circuit triggered by brain antigens may provide a solution for the treatment of EGFRvIII-negative GBM tumors. We identified two candidate brain surface proteins through bioinformatics, Cadherin10 (CDH10)-a brain-specific cadherin, and myelin oligodendrocyte glycoprotein (MOG)-a neuronal myelin sheath On the surface protein.

The predicted tissue expression of these antigens is shown in Figure 1G and Figure S7A. In this study, we mainly focused on using MOG as the priming antigen, because it has proved to be more specific to the brain (Figure S7A shows the analysis of CDH10 as the priming antigen). We identified antibodies that bind to MOG and used them to construct homologous synNotch receptors that can be activated by cells expressing MOG mouse isoforms (Figure 1H), so that these receptors can be endogenously small. Rat brain tissue triggered. Then, we used the α-MOGsynNotchàαIL13Rα2/EphA2 CAR circuit to engineer CD8+ T cells, and co-cultured them with GBM target cells (here, GBM6 PDX cell line) in the presence or absence of trigger cells. The L929 cells represent MOG).

We found that these T cells can effectively kill GBM cells, but only in the presence of MOG+ trigger cells (Figure 1I). Importantly, these T cells did not show any killing of the initiating cells. In summary, these in vitro studies indicate that there are multiple strategies designed for priming and killing circuits against GBM, which perform counterpriming/killing, and therefore it is possible to overcome antigen heterogeneity while still maintaining high specificity.

Based on these in vitro data, we next evaluated the anti-tumor activity of these initiating and killing CAR T cells in a GBM xenograft mouse model. First, we want to confirm that T cells triggered by EGFRvIII can also transkill EGFRvIII-GBM cells in vivo, but only in the presence of EGFRvIII+ triggered cells.

As a proof of principle, we implanted double tumors into NCG mice-in the brain, we implanted U87 tumors, the ratio of which is 50% EGFRvIII+ and 50% EGFRvIII-; on the side, we implanted U87 tumors ( EGFRvIII only) (Figure 2A).

Challenges: Antigen recognition combination in Glioblastoma T cell therapy

Here, flank tumors represent potentially cross-reactive tissues that normally express killer antigens but do not express priming antigens; on the contrary, brain tumors have both priming antigens and killing antigens. On the 6th day after tumor inoculation, the mice received intravenous injection. Give primed and killer CAR T cells or control untransduced T cells (n=6/group). All mice treated with control T cells showed tumor growth at both sites and quickly reached the end of euthanasia, with a median survival time of 25.5 days. In contrast, mice treated with primitive killer CAR T cells showed significant inhibition of intracranial tumor growth compared to control mice (p<0.001; t-test).

However, it is important to note that mice treated with sensitized and killer CAR T cells did not have statistically significant suppression of flank tumors compared to the control group (p = 0.4; t-test, Figure 2B). The lack of selective killing in non-priming flanking tumors indicates that the cytotoxic activity of priming and killing CAR T cells is spatially limited to tumors expressing priming and killing antigens.

We also performed a systematic comparison of 0%, 50% and 100% EGFRvIII+ U87 cells killing implanted tumors (Figure S3 A, B). We found that sensitized and killer CAR T cells did not show any clearance rates of 0% EGFRvIII-positive tumors, but showed equivalent clearance rates of 50% and 100% EGFRvIII-positive tumors (the control did not show clearance rates). Therefore, in this case, priming and killing CAR T cells can identify and effectively overcome tumors with heterologous EGFRvIII expression, but can be performed in a specific priming antigen-gated manner.

Challenges: Antigen recognition combination in Glioblastoma T cell therapy

We then attempted to evaluate the efficacy of initiating and killing CAR T cells in a tumor model that exhibits the natural heterogeneity of EGFRvIII expression. We identified xenograft (PDX) tumors derived from GBM6 patients as aggressive GBM models, showing inherent EGFRvIII heterogeneity (23) (Figure 2C). Equally important, GBM6 tumors have shown a highly replicable ability to evade EGFRvIII single antigen CAR treatment in vivo (Figure 2F). When the mice were treated with α-EGFRvIIICAR, the GBM6 tumor shrank sharply, but then relapsed slowly and steadily with high reproducibility (Figure 2F, Figure S3C, as described below).

These recurrent tumors showed the loss of EGFRvIII expression (Figure 2J) (we confirmed the presence of a part of GBM6 cells in vitro, these cells showed undetectable EGFRvIII antigen, and are resistant to conventional EGFRvIII CAR killing-Figure S4C, D). Therefore, GBM6 tumors mimic the heterogeneity-based escapes observed in clinical trials of α-EGFRvIIICAR, and therefore represent an ideal tumor model in which alternative circuits that can overcome these problems can be evaluated.

We tested T cells with EGFRvIII priming circuit, T cells constitutively expressing α-EGFRvIIICAR or α-IL13Rα2/EphA2 tandem CAR, and non-transduced (control) T cells have an effect on NCG mice with GBM6 tumors in the brain. Treatment (Figure 2F, G). On day 43 after tumor inoculation, all mice (n = 5) that received non-transduced control T cells died of tumor progression (Figure 2E, F). Treatment with constitutive α-IL13Rα2/EphA2 tandem CAR is largely ineffective.

Treatment with α-EGFRvIIICAR T cells produced initial tumor shrinkage, but consistently (n=6) resulted in EGFRvIII-negative tumor recurrence in all mice (3 out of 6 mice died of tumor progression on day 125) (Figure) 2 E). In sharp contrast, all mice treated with primary and killer CAR T cells (n = 6) showed long-term complete remission of GBM6 tumors. This longer-lasting and more thorough tumor clearance rate has high reproducibility (Figure S3C), which is also reflected in the significantly improved survival rate of mice treated with the priming and killing circuit (multiple doses) Below (Figure 2G).

Challenges: Antigen recognition combination in Glioblastoma T cell therapy

We performed post hoc immunofluorescence analysis on mice treated with EGFRvIII CAR or the trigger-kill circuit. The brains of mice treated with the perfusion and killing circuit showed the absence of GBM6 tumor cells (consistent with tumor clearance), but the persistence of CAR T cells (pink) in the brain parenchyma and meninges (Figure 2I). In contrast, the brain treated with EGFRvIII CAR showed a large number of GBM6 tumor cells (yellow-consistent with recurrence), but the EGFRvIII antigen was lost (red), and CAR T cells did not survive (Figure 2J). In short, the in vivo tumor killing research supports the concept that the dual antigen circuit can combine the specificity of the triggering antigen with the homogeneity of the killing antigen to achieve higher but more specificity than the CAR for a single antigen can achieve. Complete destruction.

In order to directly observe the initiation of T cells in vivo, we fused the GFP tag to the induced α-IL13Rα2/EphA2 CAR. Six days after T cell infusion, analysis of brain slices from recipient mice showed the presence of GFP+-primed and killer CAR T cells (co-stained with human CD45) in the tumor (Figure 3A). In contrast, no triggered T cells were found outside the tumor (near the brain tissue) or in the spleen (Figure 3A). In addition, we performed in vivo imaging of these sensitized and killed T cells two days after the injection, and the results showed that there were a large number of sensitized (green) T cells in the tumor, and T cells that turned green when approaching the tumor (Figure 3B) , video S2). Importantly, these videoes show that the triggered T cells are stably localized within the tumor (presumably interacting with target cells) and do not enter or exit the tumor quickly. This behavior may help explain the high specificity and local killing effect of these T cells.

One of the most surprising findings in these in vivo studies is that compared with the constitutive αIL13Rα2/EphA2 CAR T cells, the priming and killer T cells have significantly better tumor clearance capabilities because both groups of T cells use the same CAR kills molecules, and both are equally effective in killing tumors in vitro (Figure S4D). These observations indicate that the CAR circuit induced by synNotch has other functions and can significantly improve the anti-tumor activity in the body. The general challenge of using CAR T cells to treat solid cancers is the depletion of T cells, which prevents long-lasting anti-tumor activity.

Recent studies have shown that the tonic signal of constitutively expressed CAR can play an important role in improving its fatigue sensitivity (24, 25). Therefore, we checked the differentiation status of different types of T cells through flow analysis and found that all synNotchàCART cells used in this study showed a higher proportion of cells in the naive state (CD62L+ CD45RA+, naive or dry central memory). Compared with equivalent constitutive CAR T cells (Figure 3C, D). In addition, when we directly studied the persistence of T cells in the body (6 days after injection into mice), we observed a large number of synNotchàCAR circuit T cells.

In contrast, we found that there were no viable constitutive tandem CAR T cells at this time (Figure 3E). Taken together, these findings are consistent with a simple model: the synNotchàCAR circuit prevents the tonic signaling that is usually observed in constitutively expressed CARs, thereby keeping T cells in a more naive state and not easily depleted. Therefore, the local transient priming of CAR expression not only increases the target specificity, but also seems to produce a more effective and durable T cell state.

We also want to test whether T cells triggered by brain-specific antigens are effective in vivo. Therefore, we treated NCG mice implanted with intracranial GBM6 PDX tumors with T cells implanted with α-MOGsynNotchàα-IL13Rα2/EphA2CAR circuit (Figure 4A). GBM6 tumor cells do not express MOG, so in order to trigger T cells, they must be triggered by MOG endogenously expressed in the brain of the host mouse (Figure 1I). We found that MOG-induced T cells are very effective in clearing GBM6 tumors and increasing the survival rate of mice (Figure 4 B, C).

In mice where GBM6 tumors were implanted on the side instead of the brain, MOG-induced T cells were unable to clear the tumor (Figure S7F), which is consistent with the need for local brain signals to allow T cells to kill. Mice treated with α-CDH10synNotchàα-IL13Rα2/EphA2CAR circuit also showed effective killing of GBM6, but in this case, it showed poor discrimination in killing brain tumors and flanking tumors, which indicates that MOG is superior to CDH10, because it strictly restricts brain activity, specifically activates the antigen (Figure S7F).

Post-mortem immunofluorescence analysis of mice treated with MOG-primed T cells showed that there were a large number of T cells in the tumor, many of which were in a primed state (observed by GFP fusion CAR) (Figure 4D). In the brain tissue adjacent to the tumor, we observed fewer T cells, but they were also primed (GFP+) (primed T cells were not observed in the spleen). These results are triggered by T cells in the whole brain, and are presumably consistent with the more obvious spread of T cells in the tumor caused by CAR activation and proliferative cytokine release.

The effectiveness of this brain-specific antigen-initiated circuit represents a new advancement in engineering treatment of cells-this circuit not only represents a possible method for the treatment of EGFRvIII-GBM, but may also provide general strategies for engineering cell therapies targeting the brain to treat a wide range of nerves Systemic diseases, including other brain tumors, neuroinflammation or neurodegenerative diseases.

In summary, these results indicate that there are multiple ways to design synNotchàCAR circuits that can effectively combine and recognize imperfect complementary antigens as a single antigen target. Circuits triggered by highly specific neoantigens (such as EGFRvIII) or tissue-specific antigens (such as MOG) can be established. Crucially, these priming antigens need not all be present on all tumor cells, nor do they all need to be present on any tumor cells (in the case of tissue-specific priming).

Once activated, T cells can be programmed to target the homogenous CAR antigen on the tumor, thereby completing complete tumor killing, even if they themselves have imperfect specificity. By integrating information from multiple antigens and multiple cells, these circuits essentially provide us with an improved ability to refine tumors into complex tissues, thereby opening up how to identify and attack tumors in a safer and more specific way. Many new possibilities. Other related strategies combine CAR and bispecific adaptors to integrate multiple antigen combinations (26).

We show here that synNotchàCART cells have multiple characteristics that distinguish them from conventional CAR T cells, which may prove to have a high advantage in the treatment of solid cancers (such as GBM). First, their multiple antigen recognition enhances the distinction between tumor cells and normal cells. Second, the ability of these circuits to mediate trans-priming/killing allows them to overcome the escape of target antigen heterogeneity.

Finally, the simple act of placing CAR expression under controlled control seems to maintain T cells in a more naive state, which is more durable and less prone to fatigue. Similar improved anti-tumor activity was observed for synNotchàCAR T cells targeting cancers other than GBM, indicating that these circuits may provide a very effective general strategy for the treatment of many solid cancers.



(source:internet, reference only)

Disclaimer of medicaltrend.org