NEW YORK – Identifying which patients will benefit from CAR T-cell therapy has historically been challenging, but a group of researchers from MD Anderson Cancer Center have conducted research that suggests that the unique molecular and cellular features of individual CAR T-cell infusion products may play a role in driving variable responses in patients.
The CAR T-cell therapy axicabtagene ciloleucel (Gilead Sciences' Yescarta) was approved by the US Food and Drug Administration in 2017 as a treatment for patients with large B cell lymphoma (LBCL). The CAR T-cell therapy is considered a highly personalized treatment since the product is manufactured using each patient's immune cells.
Still, this type of autologous CAR T-cell therapy doesn't benefit everyone who receives it. Some patients may experience durable responses, while others may not respond or have to stop taking their treatments due to severe toxicities. In fact, Gilead's drug was approved on the basis of clinical trial results in which roughly half of patients experienced a complete response to treatment, and over half of patients experienced serious adverse reactions following infusion.
MD Anderson researchers, including Sattva Neelapu, Linghua Wang, and Michael Green, among other collaborators, published a paper recently in Nature Medicine identifying three findings that may shed light on this variability. First, they found that transcriptomic features of the CAR T-cell product correlated with whether patients achieve a complete response following therapy. Second, they reported that patients' cell-free DNA (cfDNA) responses one week following axicabtagene ciloleucel infusion may predict their subsequent longer-term clinical response. Third, the researchers noted that the presence of a unique population of cells with monocyte-like features in patients' CAR T infusion products may be associated with their chances of experiencing treatment-related neurotoxicity.
Homing in on T-cell state
To identify the features that may be predictive of patients' responses and toxicities, the researchers performed single-cell sequencing on residual cells that remained in the infusion bags that held the anti-CD19 CAR T-cell treatments given to 24 LBCL patients. The study participants included 16 with diffuse LBCL, six with transformed follicular lymphoma, and two with primary mediastinal B cell lymphoma.
Each patient received the standard-of-care dose of axicabtagene ciloleucel. They first had their immune cells harvested. The cells were then altered in the lab to bind to CD19-expressing cancer cells and infused back into the patient as a one-time treatment.
To home in on the characteristics of these T-cell infusion products that correlate with patients' treatment responses, the MD Anderson researchers performed whole-transcriptome single-cell RNA sequencing on 137,326 cells left in the infusion bags following axicabtagene ciloleucel treatment. They used a clustering method and marker genes to group cells by their functional state and linked those data to T-cell receptor sequencing data. In the process, they employed a new method they developed called CapID+ to boost the sequencing saturation of informative genes, allowing for a more accurate assignment of CAR status and cellular gene expression.
Ultimately, the researchers found, patients who experienced a complete response three months after their infusion were generally those who had received an infusion product that was enriched for CD8 memory T cell phenotypes. Patients who experienced disease progression or partial response at three months post-infusion, on the other hand, had mostly received infusion products that had a significant enrichment of cells with a CD8 T cell exhaustion signature. Relative to memory CD8 T cells, exhausted CD8 T cells tend to have reduced T-cell function, including less robust and durable cytotoxic activity against cancer cells.
Furthermore, between the CD8 T cells of products given to patients who experienced complete response versus partial response or disease progression, the researchers found a series of differentially expressed genes. In the exhausted CD8 T cells associated with disease progression or partial response, these included LAG3, BATF, ID2, IFNG, GZMA, GZMB, GNLY, and MHCII.
"The features associated with activation and exhaustion in cells predominantly originating from patients with [partial response or disease progression] are, therefore, linked with a profile of CD8 T-cell dysfunction," concluded the authors. Among patients who experienced a complete response, on the other hand, the infusion product tended to have a high expression of genes associated with the central memory phenotype, including CCR7, CD27, and SELL.
According to Green, one of the Nature Medicine paper's lead authors, the jury is still out as to what exactly accounts for the difference in these T-cell exhaustion signatures or memory signatures in the infusion product. Are these differences an intrinsic part of the cells harvested from the patients or are they the result of the manufacturing process? This will be an important question to answer before researchers can take the next step and determine how to mitigate these differences to improve patients' responses.
"Our hypothesis is that [the difference] is likely pre-existing in the apheresis of the product taken from the patient and then is perpetuated during manufacturing," Green said. If indeed the signatures reflect inherent differences in the cells harvested from patients, he added that it may be beneficial to test patients' apheresis products for these signatures before manufacturing. If patients' cells show features associated with poorer outcomes, doctors can pursue other options, such as investigational allogeneic, or off-the-shelf CAR T-cell products, which are based on T cells from healthy donors and are expected to have fewer toxicities than autologous treatments.
Green and collaborators will conduct research to validate their hypothesis with the support of a recent grant from the Cancer Prevention and Research Institute of Texas (CPRIT).
Potential utility of cfDNA
In addition to scrutinizing the cellular and transcriptomic features of the infusion products, the researchers sought to evaluate whether changes in cfDNA in patients' plasma one week after infusion could predict their subsequent clinical responses at the three-month mark.
Normally, Green explained, patients undergo the first assessment by PET/CT imaging one month after receiving their CAR T-cell infusions, at which point many patients show a partial response. Then, unpredictably, patients' responses start to diverge at around three months, with some having a complete response and others progressing. As such, the one-month, post-treatment PET/CT scan is not the strongest predictor of patients' longer-term responses. By interrogating early molecular responses in patients' plasma-derived cfDNA, the researchers found, it may be possible to predict patients' clinical responses earlier and more accurately.
To determine this, Green and colleagues sequenced 22 evaluable patients' cfDNA with a hybrid capture panel and compared changes in variant allele frequency of somatic mutations present in samples at the point of infusion and one week post-infusion. Nearly all — 16 out of 17 — patients who had three or more somatic mutations at the onset had a decline in variant allele frequency at the one-week mark, but the magnitude of this decline differed.
Specifically, researchers found that the majority of patients with more than fivefold reductions in variant allele frequency ended up achieving a complete response at the three-month mark. None of the eight patients whose variant allele frequency did not reduce by at least five-fold after treatment experienced a complete response. Though the predictive capability of cfDNA responses at one week would need to be validated in a larger cohort of patients, the researchers suggested that it could potentially provide a useful method for evaluating patients' clinical outcomes early on in the wake of CAR T-cell infusion.
Moreover, Green and colleagues happened upon a striking overlap between these two poor response features they identified in their research: Patients who did not to achieve the five-fold reduction in variant allele frequency at one week also appeared to have received infusion products expressing the CD8 T-cell exhaustion signature.
"These two things go hand-in-hand," Green said. "Patients with poor early molecular response tend to have more T-cell exhaustion."
When researchers compared patients who did not achieve a fivefold variant allele frequency reduction in cfDNA against those who did, they noted that the former group received treatments where a higher proportion of CD8 T cells expressed the genes associated with T-cell exhaustion. Of these, they noted, the most commonly expressed receptor genes were LAG3 and TIM3. Cells that co-expressed LAG3 and TIM3 were especially associated with T-cell exhaustion. "This co-expression is important, because those are co-inhibitory molecules that dampen the activity of T cells," Green explained.
Although further research is needed to confirm this finding as well, Green proposed the possibility that patients may be treated with LAG3- and TIM3-blocking antibodies should their one-week cfDNA analysis predict poor outcomes.
"Right now, as part of the study that is funded through CPRIT, we are exploring these differences," Green said. "We want to look at the origin of T-cell exhaustion, validate the ability of cell-free DNA to predict early progression, and evaluate the activity of LAG3- and TIM3-blocking antibodies to try and rescue CAR T-cell exhaustion."
Predicting, mitigating neurotoxicity
As a final element of their research, Green and colleagues sought to assess whether certain features of patients' CAR T-cell infusion products might play a role in the degree to which patients experience treatment-related toxicities. They chose to look into immune effector cell-associated neurotoxicity syndrome (ICANS), which impairs patients' cognitive capabilities, and along with cytokine release syndrome, is a common and potentially life-threatening toxicity associated with axicabtagene ciloleucel.
Researchers stratified patients by the severity of their ICANS following treatment. A dozen patients experienced ICANS grade 0-2, and the same number of patients experienced more severe grade 3-4 ICANS.
While the gene expression of the CD4 and CD8 T cells in the products that were found to be associated with responses were not all too different between the patients with lower- and higher-grade ICANS, the researchers did find a small group of cells that were over-represented in the products of patients who ended up experiencing higher-grade ICANS. These cells shared distinct transcriptomic features similar to those of monocytes (a type of large white blood cell), but interestingly did not express genes encoding canonical monocyte markers. Given this uncertain cell lineage, the researchers referred to the cells as "monocyte-like," but stopped short of definitively calling them monocytes.
"We're still not entirely clear on what the origin of these cells are," Green said. "It was a very surprising result to see those cells there in the data, and we worked a lot to get as much clarity as we could regarding their origin, but still couldn't really nail it down to a specific subtype."
Further research will be needed to drill down on the exact origin of these cells and whether they actually play a causative role in the more severe ICANS that patients experienced. Down the line, however, if researchers do define the lineage of these cells and determine a causal relationship between the presence of these cells in the infusion product and patients' severity of ICANS, Green said it may be possible to mitigate this phenomenon.
One way to accomplish this might be for Green and researchers to potentially work with Kite (the Gilead subsidiary that makes axicabtagene ciloleucel) to tweak the protocols of their manufacturing "to allow them to remove these cells from the infusion products." After all, Green said, cell types other than T cells are not, theoretically, supposed to exist in the CAR T infusion product in the first place.
Beyond tweaking the manufacturing process, another potential mitigation technique may be to give patients the immunosuppressive IL-1 receptor agonist, anakinra (Sobi's Kineret) alongside axicabtagene ciloleucel to reduce these toxicities. "Monocyte-derived IL-1 is a central driver of neurotoxicity," Green explained, "and IL-1 was one of the most highly expressed genes in the monocyte-like cells that we observed in our products as well."
The addition of anakinra to CAR T-cell infusions as a prophylactic method of preventing toxicities is currently being evaluated in a Phase I/II trial led by one of Green's MD Anderson colleagues, Paolo Strati. Importantly, Green pointed out that this early-phase trial is currently taking an all-comer approach to enrollment rather than stratifying patients by any sort of features related to their infusion products, focusing chiefly on evaluating anakinra's safety and ensuring that it does not interfere with the efficacy of the CAR T-cell infusion product.
But even so, a closer analysis of patients' potential decline in ICANS with the addition of anakinra may end up shedding further light on the role that these curious monocyte-like cells play in CAR T-cell-related neurotoxicities.