Research Signal – Source date 2021-10-05
A striking response can make engineered immune cells look close to a breakthrough. The review published on October 5, 2021 describes one patient whose glioblastoma lesions shrank by 77–100% after treatment with cells carrying a receptor aimed at interleukin-13 receptor alpha 2. The cells were given into the brain after tumour resection, followed by further infusions into the brain’s fluid spaces. The response lasted 7.5 months before the disease recurred.
That experience shows both the reach and the difficulty of this field. A treatment can produce dramatic shrinkage in an individual patient without yet offering dependable, lasting control for people with glioblastoma. The review brings together several approaches that might close that gap, including cell design, tumour-target selection, gene editing, delivery routes and treatment timing.
A striking response with a limited lease on control
Glioblastoma is difficult to treat because it is not one uniform mass. Different cells within the same tumour can carry different molecular changes, display different surface features and respond differently to immune pressure. The surrounding tissue also tends to suppress immune activity. Acidity, altered metabolism, suppressive macrophages and microglia, regulatory T cells, and signals such as transforming growth factor beta can all make the tumour environment hostile to attacking lymphocytes.
The single-patient CAR T result therefore matters as a demonstration of what targeted cells can sometimes do in the right setting. It does not show how frequently a similar response would occur, how long control would last across a larger group, or whether the same route and target would work for other patients. The review uses the case to frame a broader question: how can an immune cell keep finding and attacking a tumour that changes as treatment proceeds?
Why a powerful immune attack can lose its target
A chimeric antigen receptor, or CAR, gives an immune cell an added recognition system. Its outer portion is derived from an antibody and binds a selected molecule on the tumour surface. Its inner portion contains signalling elements that can activate the cell after recognition, without relying on the usual major histocompatibility complex pathway used by conventional T-cell receptors.
The design is only useful if the target remains present on enough tumour cells and is sufficiently restricted from healthy tissue. Glioblastoma creates problems on both sides. Its molecular diversity means that a target may be abundant in one region and absent in another. Some targets are also found outside the tumour, raising the risk that an attack aimed at cancer cells could damage normal tissue.
The review discusses EGFRvIII as a clear example of the first problem. In a phase I study of EGFRvIII-directed CAR T cells, a single infusion was followed by loss of the target antigen and signs of adaptive immune resistance, including increased PD-L1, IDO1 and IL-10 alongside FoxP3-positive regulatory T cells. The engineered cells had a defined target, but the tumour environment changed in ways that could blunt the response.
One answer is to recognize more than one tumour feature. The review discusses combining IL13Rα2 and EGFRvIII as a possible way to reduce the chance that cells escaping one receptor would evade the entire treatment. IL13Rα2 is attractive because it is overexpressed in glioblastoma and has little detectable expression in healthy brain tissue in the studies discussed. HER2 presents the opposite difficulty: it can be found in glioblastoma, but its expression in normal brain cells complicates the search for a tolerable treatment.
T cells and natural killer cells start from different strengths
CAR T cells are made from the patient’s own T cells in the usual autologous approach. Blood is collected through apheresis, the cells are given the genetic instructions for the chosen CAR, and the modified cells are selected, expanded and frozen before reinfusion. Viral delivery or electroporation can introduce the construct, but the process takes time and requires enough usable T cells to begin with.
That requirement matters in glioblastoma, where previous treatment and the disease itself can be associated with lymphopenia. A patient with too few T cells, or with rapidly progressing disease, may not have enough cells to manufacture a therapeutic dose in time. The manufacturing interval is therefore part of the treatment problem, not a detail that comes after the biology.
Natural killer, or NK, cells begin from a different recognition architecture. They are innate lymphocytes that use germline-encoded activating and inhibitory receptors, including natural cytotoxicity receptors and NKG2D, to detect stressed, transformed or infected cells. Adding a CAR gives them an extra tumour-directed receptor while leaving those native recognition pathways available.
That combination could matter if a tumour loses the molecule recognized by the CAR. A CAR T cell may lose its main route to the target, whereas a CAR NK cell may still respond through its intrinsic receptors. NK cells also do not induce graft-versus-host disease in the way allogeneic T cells can, and the review describes them as having a potential safety advantage because they are not known to induce the same cytokine-release and neurological syndromes associated with CAR T therapy.
The trade-offs are practical. NK cells are only about 5–15% of circulating lymphocytes in healthy adults, require extensive expansion and generally have a shorter lifespan than T cells. They can also be harder to transfect. The review describes progress with electroporating CRISPR-associated Cas9 protein and guide RNA complexes into primary NK cells, but these engineering and persistence challenges mean that the evidence does not identify a universally superior cell platform for glioblastoma.
Gene editing changes more than tumour recognition
A CAR adds a recognition-and-activation system. Gene editing can alter the cell’s wider behaviour, including how strongly it responds to inhibitory signals and how long it remains functional. CRISPR/Cas9 can be directed to a selected DNA sequence, where it creates a break that the cell repairs. The common repair pathway can introduce small insertions or deletions that disable a gene, while more precise repair is harder to control and can produce unwanted edits.
The review describes a multiplex experiment in glioblastoma-bearing mice that combined an EGFRvIII-directed T-cell design with disruption of the PD-1 gene, the endogenous T-cell receptor alpha chain and beta-2 microglobulin. The edited cells improved survival in those mice. This is important evidence that several barriers can be addressed in one cell design, but it remains an animal result rather than evidence that the same edits improve outcomes for people with glioblastoma.
The review also separates that experiment from an early human editing study in other cancers. Three patients with refractory cancer received autologous T cells edited to remove endogenous T-cell receptor components and PD-1, with a new receptor aimed at the cancer-testis antigen NY-ESO-1. The engineered cells persisted for up to nine months. Those findings show that edited cells can be produced and detected in people, but the patients and target were not a glioblastoma population.
Prime editing is presented as a possible next-generation tool, not as a validated glioblastoma treatment. By combining a modified Cas9 enzyme with reverse transcriptase and an extended guide RNA, it is designed to make more specific nucleotide changes than standard cutting-and-repair approaches. The review proposes that it might eventually be used to adjust exhaustion-related genes or inflammatory signalling while reducing unwanted insertions and deletions. That proposal still requires testing in the setting where tumour heterogeneity, brain exposure and neurological safety all matter at once.
The route can decide where cells act
The blood–brain barrier complicates delivery because it restricts the movement of cells and large molecules from the circulation into brain tissue. Glioblastoma can disrupt that barrier unevenly, which creates an opening for treatment in some regions while leaving infiltrating tumour cells harder to reach elsewhere. The location of the tumour, the integrity of nearby vessels and the distribution of disease all affect how an administered cell population might travel.
Systemic infusion is therefore only one possibility. The review discusses intracranial administration after resection and delivery through cerebrospinal-fluid spaces, including intraventricular and intrathecal routes. Cerebrospinal fluid may help distribute cells toward multifocal lesions that are difficult to reach from a single cavity. The reported IL13Rα2 case, in which later infusions were intraventricular and intrathecal, provides an example of that approach in one patient rather than a general delivery rule.
Local delivery also changes the safety question. A concentrated dose near the tumour might limit some systemic exposure, but cells still need to encounter dispersed disease without injuring healthy neural tissue. The review describes meningeal inflammation with some intrathecal drugs and poor brain-tissue distribution with intraventricular chemotherapy; those observations do not establish the safety or distribution of engineered cells. Targets such as HER2, which are not confined to tumour cells, make the balance between access and off-tumour injury especially important.
Treatment timing is part of the design
Cell therapy does not operate in isolation from standard treatment. Chemotherapy can cause immunogenic death of tumour cells, releasing material that may help the immune system recognize the cancer. Radiation can also alter the tumour environment. At the same time, temozolomide and radiation are lymphotoxic and can deepen lymphopenia, potentially damaging the very cells that clinicians hope to transfer or activate.
The review therefore treats timing as a design variable. One proposed approach is to give engineered T or NK cells after standard treatment, when newly released tumour material might support immune recognition while the cells are not being exposed at the same moment to the full lymphotoxic effect of chemotherapy or radiation. Another proposal places immune-modulating treatment before surgery and cell therapy, using anti-PD-1 with or without 4-1BB agonism before a later CAR-cell infusion.
These schedules are hypotheses about how separate treatments might cooperate, not established glioblastoma regimens. The sequence could affect cell survival, tumour visibility, lymphocyte availability and the risk of inflammation. A useful combination will have to coordinate those effects without turning improved immune activity into intolerable neurological toxicity.
What engineered immune cells could change for recurrent glioblastoma
For people facing recurrent glioblastoma, engineered immune cells could eventually provide an additional way to control disease for longer after standard options have become limited. The practical promise is not a single ideal receptor or a guaranteed replacement for existing treatment. It is the possibility of tailoring the cell’s recognition system, its resistance to suppressive signals and its route of delivery to the characteristics and distribution of an individual tumour.
That benefit remains conditional. It depends on showing sustained activity across tumour sites while keeping neurological and other treatment effects tolerable in patients. If cell design and treatment timing can be coordinated well enough to maintain an attack after the tumour changes its targets, the field could move from isolated episodes of shrinkage toward longer control of recurrent disease.
Key Sources
- Primary source: Challenges and Prospects for Designer T and NK Cells in Glioblastoma Immunotherapy.
DOI or Journal Page · Review or Meta-Analysis · 2021-10-05
Keep exploring