Juno: we could have saved JCAR015
It has been a year since Juno’s Rocket trial of JCAR015 was halted, and only now are details emerging about the root cause of the deaths due to cerebral oedema that had led to the asset being scrapped.
The message from the company is clear: had it known then what it knows now it might not have discontinued JCAR015, its chief medical officer, Mark Gilbert, revealed to EP Vantage. With the world’s first two CAR-T therapies, Novartis’s Kymriah and Gilead’s Yescarta, now approved the findings will be welcomed by a field in which product safety is still largely hit and miss.
“There would have been a way to actually take JCAR015 forward if we had wanted to,” Mr Gilbert said as detailed findings of Juno’s Rocket investigation were presented at the SITC meeting today. But he stressed that there was no need to, because waiting in the wings the company had JCAR017, which is now its lead CAR-T project.
The basis of the group’s optimism lies in ways it reckons it can now identify patients at especially high risk of developing serious neurotoxicities. The suggestion is that eliminating such high-risk individuals from trials, or modifying their treatment, could perhaps mitigate the occurrence of cerebral oedema.
Juno says high-risk subjects might be identified by one or more biomarkers, including: high levels of the inflammatory cytokine IL-15; low levels of blood platelets, which could otherwise protect blood vessels from endothelial damage; and the presence of tumours not carrying the Philadelphia gene signature.
The IL-15 hypothesis is based on the finding that severe neurotoxicity is associated with rapid T-cell expansion in patients with high baseline blood levels of this cytokine – something also related to the intensity of the lymphodepleting and bridging chemotherapy given.
The body’s reaction to chemo is to increase IL-15 levels, said Mr Gilbert. “Then when we add the CAR-T cells you’re giving them a direct growth factor for expansion.”
No brain entry?
The protective effect of blood platelets on the endothelium, meanwhile, goes to the root cause of cerebral oedema. Until now this toxicity was thought to result from CAR-T cells crossing the blood-brain barrier into the CNS and becoming activated there.
But, remarkably, Mr Gilbert said autopsies on two cerebral oedema subjects in Rocket revealed “no evidence of activated T cells entering into the CNS.”
What was seen was a complete breakdown of the blood-brain barrier, something he now thinks was caused by an inflammatory coagulation involving endothelial damage by angiopoietins and von Willebrand multimer.
But any CNS infiltration is a “much later event”, said Mr Gilbert, adding: “We think damage is occurring at the vascular barrier level. The T cells are probably not inside the CNS, they’re in the vascular space.”
Children's Hospital of Philadelphia’s Dr Stephan Grupp, who has been involved in the development of Novartis’s Kymriah, agreed that T cells were unlikely to get into the CNS in sufficiently high numbers. The Juno analysis “suggests that this is not a clear, local inflammatory effect where T cells are doing this directly”, he told EP Vantage today. “It still points the way back onto cytokines.”
Another of Juno’s key findings was the association of toxicity with younger patients who had had fewer therapies, resulting in healthier T cells and too high a proportion of CD8+ (cytotoxic) T cells in the final product. The strongest association with risk, said Mr Gilbert, was high CD8+ T-cell levels.
“Early expansion of these cells and their activity certainly was associated with fatal cerebral oedema, and may have actually impacted other patients [who subsequently recovered].” Of course, this strongly backs Juno’s subsequent focus on JCAR017 – a product comprising a more exact and controllable CD4+/CD8+ T-cell ratio.
Dr Grupp said this, as well as the involvement of IL-15, was not something he had come across, adding: “Thank god [Juno] is doing this analysis – this is the first report at a scientific venue about what happened,” in Rocket.
|Cerebral oedema deaths*
|Flu/cy or bendamustine
|None in ~300 subjects treated
|1 in ~300 subjects treated
|1:1 ratio of CD4/CD8
|Flu or cy or flu/cy or cy/etop
|1:1 ratio of CD4/CD8
|2 in <200 subjects treated
|CD4 + CD8 co-culture
|Flu/cy, then cy only
|5 in <100 subjects treated
|*As reported to the US FDA, and at scientific meetings, relating to estimated numbers of clinical trial subjects.
Less clear-cut are differences between CAR-T constructs; “The CD28 co-stimulatory domain, relative to 4-1BB, probably does change the kinetics of expansion of the cells in vivo,” Mr Gilbert said, but added that there was no common factor.
Dr Grupp continues to argue that evidence points to neurotoxicity risk being higher in CD28-co-stimulated CARs, which tend to “burn brighter”. There is also a potential role for manufacturing differences among constructs that are otherwise very similar, he said.
JCAR014, a Juno project that uses an identical construct to JCAR017, was associated with two cerebral oedema deaths, but Mr Gilbert put these down to too high a dose: “If you give a high enough dose, I think you’re going to see cerebral oedema with every CAR construct that exists.”
Conversely, only one case of cerebral oedema was seen in trials of Yescarta, a very similar construct to JCAR015, but Mr Gilbert cautioned that “we don’t know all of the cases of cerebral oedema or fatal neurotoxicity that have actually occurred in this space”.
If there was underreporting it would have happened despite thorough safety reviews that competitors carried out after JCAR015’s problems (Spotlight – Putting a number on CAR-T deaths, June 26, 2017).
Still, anyone seeking clear-cut answers will be disappointed. Mr Gilbert said cerebral oedema seemed to result from a combination of product and patient characteristics: “You have to have multiple risk factors, a composite.”