Scientists at Dana-Farber Cancer Institute and Mass General Cancer Center have created molecular ON-OFF switches to regulate the activity of CAR T cells, a potent form of cell-based immunotherapy that has had dramatic success in treating some advanced cancers, but which pose a significant risk of toxic side effects.
CAR T cells are immune cells genetically modified to recognize and attack tumors cells.
Once given, these “living drugs” proliferate and kill tumor cells over weeks to months, in some cases causing life-threatening inflammatory reactions that are difficult to control.
In this way, CAR T cells are unlike more established forms of cancer therapy – chemotherapy or radiotherapy for instance – whose dose can be precisely tuned up or down over time.
The scientists reported in Science Translational Medicine the development of switchable CAR T cells that can be turned on or off by giving a commonly used cancer drug, lenalidomide.
In the laboratory, the researchers designed OFF-switch CAR T cells that could be quickly, reversibly turned off by administering the drug, after which the CAR T cells recovered their anti-tumor activity. Separately, the researchers also reported ON-switch CAR T cells that only killed tumor cells during lenalidomide treatment.
In the future, switchable cell therapies might allow patients with their physicians to take a pill – or not – to tune the amount of CAR T cell activity from day to day, hopefully reducing toxic side effects.
“From the start, our goal was to build cancer therapies that are less hard on people. Having built these switches using human genetic sequences and an FDA-approved drug, we are excited for the potential to translate this research to clinical use,” said Max Jan, MD, Ph.D., first author of the report. He is affiliated with the laboratories of Benjamin Ebert, MD, Ph.D., and Marcela Maus, MD, Ph.D., the report’s senior authors. Other authors include researchers from the Broad Institute of MIT and Harvard, and Harvard Medical School.
CAR T cells are created by harvesting immune T cells from the patient and reprogramming them in the laboratory to produce a finely-tuned receptor molecule, termed a CAR (for chimeric antigen receptor), that recognizes a distinctive protein on the surface of the patient’s cancer cells.
The CAR T cells, after being engineered in the lab and returned to the patient, circulate through the body and home in on the cancer cells by binding to the distinctive surface protein they have been engineered to recognize. This binding event stimulates an immune attack, destruction of the cancer cells, and proliferation of the CAR T cells.
A drawback, however, is that uncontrolled proliferation of the CAR T cells sometimes triggers cytokine release syndrome (CRS), the release of inflammation-causing signals throughout the body that can cause toxicities ranging from mild fever to life-threatening organ failure.
Current management of these toxic reactions relies on intensive care unit support and drugs including immunosuppressive corticosteroids, while many researchers are trying to develop methods of controlling the activity of CAR T cells in order to prevent these toxic side effects.
“CAR T cells can be fantastically effective therapies, but they can also have serious toxicities and can cause significant morbidity and mortality,” said Ebert who is Chair of Medical Oncology at Dana-Farber. “They are currently difficult to control once administered to the patient.”
CAR T cell therapy has had most success in blood cancers.
Three CAR T agents have been approved: Kymriah for children and young adults with B-cell precursor acute lymphoblastic leukemia (ALL), both Kymriah and Yescarta for treatment of adults with diffuse large B-cell lymphoma and Tecartus for adults with mantle cell lymphoma. Scientists are investigating an array of different approaches with the aim of extending the reach of CAR T therapies to other blood cancers and to solid tumors, if a number of hurdles can be overcome, including the problem of treatment toxicity.
To create the ON and OFF switch systems for CAR T cells, the scientists used a relatively new technique known as targeted protein degradation.
It exploits a mechanism that cells use to dispose of unwanted or abnormal proteins; the proteins are marked for destruction by a structure within cells that acts like a garbage disposal. A small number of drugs, including lenalidomide, act by targeting specific proteins for degradation using this pathway.
The researchers used this technique to engineer small protein tags that are sent to the cellular garbage disposal by lenalidomide. When this degradation tag was affixed to the CAR, it allowed the tagged CAR to be degraded during drug treatment, thereby stopping T cells from recognizing cancer cells.
Because CAR proteins are continually manufactured by these engineered T cells, after drug treatment new CAR proteins accumulate and restore the cell’s anti-tumor function. The researchers propose that the switch system might in the future allow patients to have their CAR T cell treatment temporarily paused to prevent short-term toxicity and still have long-term therapeutic effects against their cancer.
The scientists also built an ON-switch CAR by further engineering the proteins that physically interact with lenalidomide. This system has the potential to be especially safe, because the T cells only recognize and attack tumor cells during drug treatment.
If used to treat cancers such as multiple myeloma that are sensitive to lenalidomide, ON-switch CAR T cells could allow for a coordinated attack by the immune cells and the drug that controls them.
“The long-term goal is to have multiple different drugs that control different on and off switches” so that scientists can develop “ever-more complex cellular therapies,” explained Ebert.
The ability to control protein–protein interactions with small chemical compounds can open up exciting applications across various fields such as cell biology, immunology, and immunotherapy. These switchable systems are commonly known as chemically induced dimerization (CID) systems (1, 2).
In general, in a CID the interaction between two proteins can be triggered by a small molecule, and therefore, CID systems can also be regarded as molecular ON-switches. The only molecular ON-switch that has been used in humans in vivo is based on a mutated version of FK506 binding protein (FKBP) 12, which is homodimerized upon administration of the small molecule AP1903 (3).
However, for many applications it is necessary to regulate the interaction of two different proteins. Indeed, various systems have been introduced that enable such conditional heterodimerization (4⇓⇓⇓–8), including the FRB/FKBP system that is used extensively in vitro. However, their clinical translation is limited due to unfavorable characteristics of the small molecule or the nonhuman origin of protein components (7, 9⇓⇓–12). Thus, an effective molecular ON-switch that can induce heterodimerization in a clinically relevant setting is still lacking.
One important application of ON-switches is the regulation of T cells that are genetically engineered to express chimeric antigen receptors (CARs). CARs consist of an extracellular antigen-binding moiety that is fused via a transmembrane region to an intracellular signaling domain derived from the T cell receptor complex and from costimulatory molecules (13, 14).
Upon recognition of specific antigens on target cells, CARs trigger the release of cytokines and cytotoxic mediators. CAR T cells targeting the B cell marker CD19 have been impressively effective in the treatment of B cell malignancies such as acute lymphoblastic leukemia and lymphomas, recently gaining US Food and Drug Administration (FDA) approval (13).
However, a significant limitation of this therapy is the inability to control CAR T cells after they are administered to the patient. This often leads to severe adverse events, such as neurological toxicities, organ dysfunction, and cytokine release syndrome (13, 15⇓–17). Therefore, molecular tools which enable regulation of CAR T cell activity in vivo are urgently needed.
In this study, we aimed at generating a type of molecular ON-switch that matches two important design criteria: the usage of 1) an orally available small molecule with a favorable safety profile in vivo and 2) a human protein that undergoes a drug-induced conformational switch. We hypothesized that human lipocalins are ideally suited for such a molecular ON-switch.
Lipocalins possess a β-barrel fold with an internal hydrophobic ligand-binding pocket, which can bind a range of different hydrophobic small molecules, depending on the shape and biochemical property of the binding pocket (18). Moreover, some lipocalins undergo conformational change upon binding to a small molecule (19⇓⇓⇓⇓–24).
Thus, we hypothesized that other proteins could be engineered to specifically recognize the small molecule-loaded conformation of a lipocalin, forming the basis of a molecular ON-switch. In this proof-of-concept study, we used two different binder scaffolds: 1) reduced charge Sso7d (rcSso7d), which is a charge-neutralized version of a small (7 kDa), hyperthermostable protein derived from the archaeon Sulfolobus solfataricus (25, 26), and 2) the tenth type III domain of human fibronectin (FN3) with a molecular weight of 10 kDa (27⇓–29).
Here, we demonstrate that lipocalin-based molecular ON-switches can be designed to be specifically regulated with an orally available small compound. We present ON-switches in which the affinity between the human lipocalin retinol binding protein 4 (hRBP4) and its engineered binders is increased up to 550-fold upon addition of the small molecule drug A1120.
The crystal structure of the assembled ON-switch showed that the engineered binder specifically recognizes A1120-induced conformational changes in hRBP4. Finally, we show that this molecular ON-switch can be used to regulate cytotoxic activity and cytokine production of primary human CAR T cells, illustrating a potential future application of lipocalin-based ON-switches.
Discussion
In this study we generated a molecular ON-switch system, in which the interaction between a human lipocalin and an engineered binder scaffold can be controlled with an orally available small molecule. Molecular ON-switches for conditional heterodimerization are currently limited with regard to in vivo applicability, lack of orthogonality, and/or potential toxicities (4, 9⇓–11).
For example, the FRB/FKBP system can be regulated by rapamycin.
However, due to its immunosuppressive activity, rapamycin is considered to be suboptimal. A very recent preclinical CAR study showed promising results with lower concentrations of this drug (39). Nevertheless, if available, a safe compound without any immunocompromising effect—like A1120—would be preferred, especially for immunotherapeutic applications.
An alternative to rapamycin is the usage of its derivatives (so-called rapalogs), such as AP21967, which also have several drawbacks. Apart from residual immunosuppressive activity, the synthesis of rapalogs is difficult and cost intensive, and potential contamination with rapamycin is a risk potentially resulting in enhanced immunosuppression. Therefore, rapalogs are considered to be suboptimal for broad clinical application (9⇓⇓–12, 39).
In contrast, the small molecule A1120 was originally developed for long-term treatment of insulin resistance and was later also tested for treatment of age-related macular degeneration (AMD) (31, 40). hRBP4 is the transport molecule for retinol in human plasma. Due to its relatively small size (21 kDa), hRBP4 would be rapidly cleared from the circulation by the kidneys. This is prevented by complexation with another plasma protein called transthyretin (TTR) (41, 42). A1120 was developed for blocking this interaction of hRBP4 with TTR (31, 32), resulting in increased filtration of hRBP4 through the kidneys.
This is the mechanistic basis for the original applications of A1120 mentioned above, where the overall goal was a reduction of hRBP4 and/or retinoid levels in the plasma. Importantly, three different research groups have shown that oral administration of A1120 to mice does not cause any systemic toxicities, even at high doses up to 30 mg/kg and for up to 5 mo (31, 32, 40, 43).
Although the free plasma concentration of A1120 has not been reported in the literature, the data of Du et al. suggest that low micromolar serum RBP4 levels could be virtually saturated with A1120 (44). There is thus substantial evidence for efficient loading of hRBP4 with this drug in vivo, which is the critical parameter for the function of such an ON-switch. Note that A1120 is approved by the FDA for testing in patients as an investigational drug, and clinical trials for long-term treatment of AMD and inherited Stargardt macular dystrophy are under development (40, 45).
Another recently introduced ON-switch is based on the human protein BCL-xL and the small molecule ABT-737 (4). In that system, antibody fragments (Fabs) were successfully engineered to bind to a newly generated epitope consisting of both BCL-xL and the solvent-exposed portion of ABT-737. These Fabs bind to BCL-xL with high affinity only in the presence of ABT-737.
However, since ABT-737 blocks the antiapoptotic function of BCL-xL and other Bcl-2 family members, administration of ABT-737 is associated with platelet and lymphocyte toxicities (46). Moreover, ABT-737 is not orally available (47), as is often the case with small molecules with a molecular weight > 500 Da. Finally, since a large portion of ABT-737 contributes to the epitope, this small molecule is also recognized by the Fab when bound to the homologous protein BCL-W, albeit with lower affinity. This suggests that using solvent-exposed small molecules, which form part of the recognized epitope, may limit the achievable specificity of the resulting ON-switches.
In contrast, the small molecule A1120 used in our lipocalin-based ON-switch is almost completely hidden in the ligand-binding pocket of hRBP4 (SI Appendix, Fig. S7B). For that reason, it could have been anticipated that this precludes efficient discrimination between the ligand-bound vs. the unbound state of hRBP4.
Remarkably, the binders recognized hRBP4 with up to 550-fold higher affinity in the presence vs. absence of A1120. More broadly, this demonstrates that in molecular ON-switches the small molecule does not need to be solvent exposed for efficient switching behavior, provided that a conformational switch is triggered in the protein which enables allosteric recognition. We confirmed this hypothesis by analyzing the hRBP4-A1120-RS3 complex by X-ray crystallography and showing that RS3 indeed mostly interacts with the two loops of hRBP4 (36), which have been reported to change their conformation upon binding of A1120 (31, 32).
Notably, comparison of those switching loop regions in the retinol-bound vs. either of the two A1120-bound structures revealed RMSD values of around 3.3 to 3.9 Å, despite only minor differences (∼0.4 Å) between the overall structures (Fig. 4C). For comparison, RMSDs between two protein cores with only 20% sequence identity are typically in the range of 2 Å (48), indicating that the A1120-induced structural deviations in those loop regions are substantial, potentially explaining the high efficiency of the ON-switch.
Furthermore, an overlay of the RS3-bound hRBP4/A1120 complex with the retinol-bound hRBP4 structure shows that in the retinol-bound state loop 2 of hRBP4 would sterically clash with residues 23 and 25 of RS3 (SI Appendix, Fig. S4), further supporting the hypothesis that the binder recognizes the A1120-induced conformation of hRBP4. Importantly, this allosteric mechanism automatically avoids direct recognition of the small molecule and therefore off-target interactions with other proteins bound to the same compound, limiting unwanted side effects.
The fact that the engineered binders recognize a specific conformational state of hRBP4 bears the risk that other small molecules binding to hRBP4 may also trigger the ON-switch system. However, we demonstrated that the interaction of RS3 with hRBP4 bound to other known natural and synthetic ligands is almost undetectable. This confirms that lipocalin-based ON-switches can be designed to be specifically activated by a given small molecule but not by other small molecules, even if they bind to the hydrophobic pocket of the lipocalin. Together, these data strongly suggest that our ON-switch is orthogonal, i.e., largely independent of other small molecules.
One important application of molecular ON-switches is the regulation of CAR T cell activity for cancer immunotherapy. ON-switches based on human proteins and an orally available drug would facilitate the regulation of CAR T cell function in vivo. We demonstrated that the hRBP4-A1120-RS3-based ON-switch can turn on primary human CAR T cells using A1120 in vitro. This was achieved by splitting the CAR into two polypeptide chains, which only assemble upon interaction of hRBP4 and RS3. The chosen design could be considered a challenge as both chain I and chain II are expressed by the T cell. This means that chain II (which is a soluble protein) needs to be captured on the surface—or in the endoplasmic reticulum (ER) or Golgi during secretion—in order to promote assembly of the CAR. Alternatively, the soluble protein might accumulate locally and opsonize the target cells before being captured by the CAR T cells. Nevertheless, the activation levels achieved with the ON-switch CAR were comparable to those of an anti-CD19 control CAR.
Our CAR assembly system is different from that recently used by Lim and colleagues, where both chains were expressed in a membrane-anchored version (10) (SI Appendix, Fig. S7A). While the strategy with two membrane-anchored constructs prevents loss of any soluble CAR chain due to diffusion, it harbors the disadvantage that the interaction with the target cells is not regulated. Instead, only the activation of CAR signaling, and not binding to the antigen, can be controlled with the small molecule. This might potentially result in CAR T cells becoming trapped in antigen-positive tissues despite the absence of the small molecule. In our CAR system, however, both CAR signaling and the physical interaction of the CAR T cells with the target cells are prevented in the absence of A1120, which provides an additional layer of safety compared with just turning off CAR signaling. Apart from regulation of CAR T cells based on protein switches, alternative approaches were recently introduced, such as suppression of TCR and CAR signaling by dasatinib (49⇓–51). Despite not being specific for CAR T cells, the application of this kinase inhibitor represents an elegant alternative approach for controlling CAR T cells in vivo.
Summing up, we introduced an ON-switch system based on a human lipocalin, an orally available drug, and two different engineered binder scaffolds. Notably, whereas the engineered binding sites on rcSso7d-based binders are located on a rigid β-sheet (25), the binding surfaces on FN3-based binders are composed of flexible loop regions (27, 28) (Fig. 1C). While ON-switches based on the human scaffold FN3 could be engineered successfully, those based on the nonhuman scaffold rcSso7d showed even higher dependency on A1120. Therefore, in this first proof-of-concept study we focused on the nonhuman and potentially immunogenic scaffold rcSso7d. However, our findings that completely different binding sites can be engineered to specifically recognize small molecule-induced conformational changes in a lipocalin illustrate the flexibility of the system with regard to the choice of the engineered interaction partner. Consequently, we anticipate that other human binder scaffolds such as scFvs or fynomers (52) can also be used to construct hRBP4-based ON-switches. In addition, since small molecule-induced conformational changes have also been described for other lipocalins (19, 20, 22⇓–24, 53), we expect that this type of molecular ON-switch is not limited to hRBP4 either. Finally, it has been observed both in nature (18, 54) and during protein engineering experiments in the laboratory (55, 56) that the lipocalin structure is highly tolerant to mutations, enabling adaptation for binding to different small molecular compounds. Thus, we anticipate that lipocalin-based ON-switches are flexible with regard to the choice of all three components: the lipocalin, the binder scaffold, and the regulating small molecule.
reference link : https://www.pnas.org/content/117/26/14926
More information: M. Jan el al., “Reversible ON- and OFF-switch chimeric antigen receptors controlled by lenalidomide,” Science Translational Medicine (2020). stm.sciencemag.org/lookup/doi/ … scitranslmed.abb6295