Fifty-Five Ways to Recognise a Leukaemia Cell Mapping the AML Surface for Safer CAR T Therapy

This application claims engineered immune cells that carry a receptor pointed at any one of a list of roughly fifty-five proteins found on the surface of acute myeloid leukaemia cells, and a two receptor version in which the cell only switches on when it meets both targets at once. The applicant is Memorial Sloan-Kettering Cancer Center in New York, and the named inventors are Michel Sadelain and Fabiana Perna. The list is the invention: it is the output of a whole body search for a target that a killer cell can hit without also killing the patient.

The Problem

CAR T cell therapy works. The specification says so plainly, citing complete remissions in patients with CD19 positive malignancies for whom chemotherapy had failed, and noting that cancer immunotherapy was named a scientific breakthrough in 2013. Acute myeloid leukaemia is the obvious next target: it is the most common acute leukaemia in adults, it affects more than a quarter of a million adults worldwide each year, the standard induction chemotherapy regimens have not changed substantially in forty years, and overall survival remains very poor.

The obstacle is not the cell engineering. It is the absence of a suitable target. A CAR does not distinguish a cancer cell from a healthy cell, it only distinguishes cells that display a particular protein from cells that do not, so any healthy tissue carrying the same protein is attacked as well. The specification calls this the on-target, off-tumour effect and lists three cautionary cases from the clinical literature. A CAR against carbonic anhydrase IX in renal cell carcinoma caused cholestasis, because the antigen is also on bile duct epithelium. A CAR against carcinoembryonic antigen in colon cancer caused severe colitis from recognition of normal colonic tissue. And a patient treated with a CAR against HER2 developed rapid respiratory failure and multi-organ dysfunction and died, from reactivity against HER2 expressed in lung tissue.

Four CAR targets for AML had been reported when the priority application was filed, including Lewis-Y and CD123, and the results were partial or transient. None, the specification concludes, meets the criteria of an ideal target. It also criticises how the search has been run: candidates are usually assessed against their closest normal counterpart, and mostly from messenger RNA data, which correlates only loosely with how much protein actually sits on the cell surface.

What This Invention Does

The work behind the claims is a large integrated map of what the specification calls the AML surfaceome. Public transcriptomic datasets were combined with new surface specific proteomics, including cell surface biotinylation across a panel of human myeloid leukaemia lines chosen to span the genetic heterogeneity of the disease, from an MLL-AF9 rearranged line to AML1-ETO, FLT3-ITD and PML-RAR lines, alongside normal granulocytes and normal haematopoietic progenitors.

Each candidate was then annotated against reference resources that map protein expression across more than sixty normal tissues and organs, combining immunohistochemical assays with mass spectrometry data, and against a subcellular localisation database to confirm the protein really sits at the cell membrane. Starting from over five thousand gene identifiers, an algorithm filtered for expression on AML cells, absence from vital tissues and confirmed surface localisation. The reported yield was thirty-two candidates, eleven top candidates, four suitable for a single CAR strategy, and fifty-five pairs of which three were top ranked for combinatorial use. Four of the leading targets turned out to be G protein coupled receptors: ADGRE2, also called EMR2, plus CCR1, LTB4R and P2RY13.

Claim 1 is that filtered list turned into property. It covers an isolated immunoresponsive cell comprising an antigen recognising receptor that binds an antigen selected from a group of about fifty-five named proteins, running from EMR2 and CD33 through PIEZO1, ITGA5, KCNN4, LTB4R, CD70, CLEC12A, TNFRSF1B, CD96 and LILRB2 to WT1, PRAME and SLC19A1. Successive dependent claims narrow that list in stages, from eleven, to four, to a final group of EMR2, CCR1, CD70 and LILRB2, and single out EMR2.

Claim 16 is the more interesting construct. It covers a cell carrying two receptors: an antigen recognising receptor that binds a first antigen and is capable of activating the cell, and a chimeric co-stimulating receptor that binds a second, different antigen and is capable of stimulating it. Neither signal alone is sufficient for a full response, so the engineered cell effectively behaves as a logic gate, killing only where both antigens appear together. That is the direct answer to the on-target, off-tumour problem: no single healthy tissue needs to be antigen free, only the combination needs to be unique.

Key Features

• A target list built from proteins, not just transcripts. Surface specific proteomic measurements were generated and merged with transcriptomic data, on the stated grounds that messenger RNA levels are a poor proxy for surface protein abundance.
• Whole body expression as a filter. Every candidate was checked against protein expression across more than sixty normal tissues and organs, rather than only against the closest normal blood cell counterpart.
• Fifty-five antigens claimed as a group. Claim 1 covers an engineered immune cell targeting any antigen from the filtered list, with narrowing claims down to EMR2 alone.
• A two receptor logic gate. A CAR against a first antigen supplies activation and a chimeric co-stimulating receptor against a second supplies stimulation, so only cells displaying both are fully attacked.
• Multiple cell types, autologous or allogeneic. The claims cover T cells, natural killer cells, cytotoxic T lymphocytes, regulatory T cells, NKT cells and pluripotent stem cells from which lymphoid cells can be differentiated.
• A menu of signalling domains. The intracellular signalling domain may be CD3 zeta, CD28, 4-1BB, OX40, ICOS, CD27, CD2 and others, or combinations of them.

Who Is Behind It

Memorial Sloan-Kettering Cancer Center is one of the oldest and largest cancer centres in the world, and one of the institutions where CAR T cell therapy was invented rather than merely adopted. Michel Sadelain, the founding director of its Center for Cell Engineering, is among the small group of researchers credited with the second generation chimeric antigen receptor design that made tisagenlecleucel and the therapies that followed possible. Co-inventor Fabiana Perna was a member of his laboratory and led the surfaceome analysis this specification reports as Example 1.

The priority chain runs back a decade. The present application is a divisional of Australian application 2024200020, itself a divisional of Australian application 2017307610, related to international application PCT/US2017/045632 and claiming priority from United States provisional application 62/371,199 filed on 4 August 2016.

Why It Matters

AML has been waiting for its CD19 for more than a decade. The lymphoid malignancies got lucky: CD19 sits on the malignant cells and on normal B cells, and a patient can live without normal B cells. Myeloid leukaemia offers no equivalent, because the antigens on an AML blast are largely the antigens on the healthy haematopoietic stem cells the patient needs to survive. That is why the search described here had to be reframed as a systems problem rather than a single antigen problem.

The answer this document reaches, combinatorial targeting, has since become one of the main lines of work in the field, and ADGRE2 in particular has moved into clinical development as an AML target. Reading the list in claim 1 now is a little like reading a field’s to-do list written in advance, with several entries having gone on to become programmes in their own right.

The filing strategy is worth noting separately. Claiming a group of fifty-five antigens is an unusually broad way to protect what is essentially a screening result, and it is the kind of claim that gets tested hard during examination. Keeping the family alive through successive Australian divisionals to a 2026 filing, on a 2016 priority date, gives the applicant fresh opportunities to reshape scope around whichever of those antigens turns out to matter commercially.

Related Concepts

• Chimeric antigen receptor – the engineered receptor the claimed immune cells carry.
• Acute myeloid leukaemia – the disease the target list was assembled for.
• CD19 – the target whose success in lymphoid cancers set the benchmark this work is chasing.
• ADGRE2 – the EMR2 receptor that the narrowest claims single out.
• Proteomics – the measurement approach used to map what actually sits on the cell surface.
• Cancer immunotherapy – the broader treatment class this work sits inside.

---

AU 2026202107 was published in the Australian Official Journal of Patents on 9 April 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.