Rebuilding a Melanoma Antibody Into a CAR T Cell A Claim Written Entirely in Amino Acids

This application claims a chimeric antigen receptor, the engineered protein that turns an ordinary T cell into a guided weapon, defined not by what it does but by two specific stretches of amino acids that make up its antibody-derived nose. The target is CSPG4, a proteoglycan found on melanoma, glioblastoma and several other solid tumours. It was filed jointly by the University of North Carolina at Chapel Hill and The General Hospital Corporation, which is the corporate name of Massachusetts General Hospital, and is a divisional of Australian application 2019287676.

The Problem

CAR T cell therapy works spectacularly well against a small number of blood cancers and has been frustratingly difficult to extend to solid tumours. The reason is mostly about the target.

The blood cancer CARs that reached the clinic aim at CD19, a marker sitting on the surface of B cells. CD19 is close to an ideal target for three reasons. It appears on essentially every malignant B cell, it appears on almost nothing else in the body, and the one normal tissue it does appear on, healthy B cells, can be wiped out and the patient can live without them with immunoglobulin replacement. Very few solid tumour antigens tick any of those boxes, let alone all three. Most are also present at some level on normal tissue, which turns a potent therapy into a toxicity problem, and most are present unevenly across a tumour, so the cells that survive treatment are the ones that were not expressing the target in the first place.

CSPG4 is the specification’s candidate for a solid tumour antigen that comes closer. The background section sets out the case. Analysis of about 2000 surgically removed melanoma tumours showed CSPG4 expression in more than 70 per cent of melanoma lesions. Later flow cytometry and immunohistochemistry work found it on glioblastoma, squamous cell carcinoma of the head and neck, triple negative breast cancer, mesothelioma, renal cell carcinoma, chondrosarcoma, osteosarcoma, soft tissue sarcomas and subsets of acute leukaemia. In most of those, the specification says, expression is high on the malignant cells with limited variation within and between lesions. The specification also notes it is expressed on cancer initiating cells in head and neck cancer, triple negative breast cancer and melanoma, though it is careful to add that this is not universal, since CSPG4 was not detected on the CD133-defined cancer initiating cells of glioma tumours.

There is a second problem underneath that one, and it is the reason this particular application exists. Having a good target antigen is not the same as having a good binder for it. The antibody fragment used as the CAR’s recognition element has to fold correctly when fused to a T cell signalling tail, and small changes to it change how well the resulting cells kill.

What This Invention Does

Claim 1 is unusual in shape. It reads: a chimeric antigen receptor comprising the amino acid sequences QIQLVQSGPELKKPGETVKISCKASG and so on, identified as UNC 763.74 Vh1 and given as SEQ ID NO:1, and a second sequence beginning DILLTQSPAILSVTPGETVSLSCRASQ, identified as UNC 763.74 Vk2 and given as SEQ ID NO:2, linked together in any orientation. Claim 2 is the same form with a different pair, 225.28 Vh1 and 225.28 Vk, SEQ ID NOs 3 and 4.

Those two sequences are the heavy chain and light chain variable regions of an antibody. Joined by a flexible linker they form a single-chain variable fragment, or scFv, which is the part of a CAR that recognises the antigen. Two things about claim 1 are worth spelling out. First, it does not recite CSPG4 anywhere. The target is identified only through the description and through the later method claims, such as claim 21, which is a method of treating a subject having a disease associated with elevated expression of CSPG4. Second, claim 1 does not recite a hinge, a transmembrane domain, a costimulatory domain or a CD3 zeta signalling tail, even though the description says a CAR joins the antibody moiety to intracellular signalling domains from the CD3 chain of the T cell receptor in tandem with costimulatory endodomains such as CD28, 4-1BB or OX40. Claim 1 claims the recognition end of the molecule and leaves the rest open.

The phrase “linked together in any orientation” is doing deliberate work, and the dependent claims show why. Claims 3 and 4 give the two full scFv sequences built from the claim 1 pair, one in the Vk2 then Vh1 order and one in the Vh1 then Vk2 order, each with a leader peptide at the front and a glycine-serine linker in the middle. Claims 5 and 6 do the same for the claim 2 pair. Claim 7 adds a detectable moiety and claim 8 adds an effector molecule chosen from a drug, a toxin, a small molecule, an antibody, a cytokine, an oncolytic virus, an enzyme, a nanoparticle, a biomaterial or a scaffold.

From claim 9 onward the claim set walks out through every layer of the technology. Claims 9 to 13 cover nucleic acid molecules encoding the receptor, with the specific nucleotide sequences recited. Claim 14 covers a vector carrying them. Claims 15 to 17 cover cells containing the receptor, the nucleic acid or the vector, with claim 17 listing alpha-beta T cells, natural killer cells, cytotoxic T lymphocytes, regulatory T cells, natural killer T cells and Th17 cells. Claim 18 covers a composition. Claims 19 to 25 are methods of stimulating a T cell response to CSPG4-expressing cells, providing anti-tumour immunity, treating a CSPG4-associated disease, generating and expanding engineered cell populations in a subject, and treating cancer. Claims 26 to 28 cover targeting a cancer cell or cancer initiating cell carrying the CSPG4 antigen, in vitro or in vivo. Claims 29 and 30 are diagnostic: contacting a cell sample with the receptor and detecting whether a binding complex forms, which is offered as an indication of cancer cells in the sample.

The experimental support consists entirely of cell culture. In one set, human melanoma cell lines WM115 and Sk Mel 2 were plated at 250,000 cells per well and T cells added at an effector to target ratio of 1 to 1, with counts taken by flow cytometry on day 5 using counting beads. The specification states plainly that T cells expressing the CAR built with the original 763.74 scFv do not completely eliminate the target tumour cells. In the parallel set, run at 1 to 1 and 1 to 5, T cells expressing the CAR built with the UNC 763 scFv, in either the Vh1-Vk2 or the Vk2-Vh1 orientation, completely eliminated the target cells. That contrast between the old binder and the reworked one, across three donors, is the heart of the disclosure and explains why the claim is written around specific sequences rather than around the antigen.

A second set measured cytokines. After 24 hours of coculture the UNC 763 CAR T cells produced interferon gamma and interleukin-2 in response to the CSPG4-positive lines and not in response to the CSPG4-negative line M14, which is the specificity control. A further five experiments covered the 225.28 constructs: transduction efficiency measured across five T cell donors, then four and five day cocultures against the glioblastoma lines U138MG and U87 and the melanoma lines SKmel2 and WM115, in which the older and newer versions of the 469 and 470 constructs were found equally effective at eliminating the target cells.

Key Features

• A claim defined by sequence, not by function. Claim 1 recites two specific variable region amino acid sequences and nothing else. It names no antigen, no signalling domain and no cell type, which makes it narrow in one sense and unusually broad in another.
• Either orientation is covered. The heavy and light chain sequences may be linked in any order, and claims 3 to 6 recite both orientations of both scFvs in full, with leader peptides and glycine-serine linkers.
• A reworked binder that outperforms the original. The culture data distinguish the original 763.74 scFv, which the specification says does not completely eliminate target cells, from the UNC 763 version, which does, in both orientations and across three T cell donors.
• A second, independent scFv. Claim 2 covers the 225.28 pair, derived from a separate anti-CSPG4 antibody, giving the applicants two distinct binders within the same filing.
• Coverage from gene to diagnostic. The claim set runs from the receptor through nucleic acids, vectors, transduced cells, compositions and treatment methods to laboratory detection of cancer cells using the receptor as a binding reagent.
• Cells other than T cells. Claim 17 puts natural killer cells, natural killer T cells, regulatory T cells and Th17 cells alongside conventional T cells as vehicles for the receptor, and claim 32 specifies an autologous cell.

Who Is Behind It

There are two applicants. UNC Lineberger Comprehensive Cancer Center at the University of North Carolina at Chapel Hill is one of the larger academic cell therapy centres in the United States. The General Hospital Corporation is the legal entity behind Massachusetts General Hospital in Boston, and appears under that name on a great many patent filings.

The lead inventor, Gianpietro Dotti, is a professor at UNC and directs the Lineberger immunotherapy program. He trained in haematology in Italy and spent more than a decade at the Center for Cell and Gene Therapy at Baylor College of Medicine before moving to North Carolina in 2015 to help establish its manufacturing facility for cell therapies. His stated research interest is developing CAR T cells for solid tumours, which is exactly what this application is.

The second named inventor, Soldano Ferrone, was a professor in the Department of Surgery at Massachusetts General Hospital and spent roughly six decades working on tumour antigens and the antibodies that recognise them. The 225.28 antibody in claim 2 is one of the reagents his laboratory is known for. He died on 10 January 2023, some years after this family was filed, and is remembered in an in memoriam notice published in Pigment Cell and Melanoma Research. Cristina Ferrone, also named as an inventor, was a surgical oncologist at Massachusetts General Hospital at the time of filing and is now chair of surgery at Cedars-Sinai. Xinhui Wang was a colleague in the same Boston group. Hannah Reid Hudson and Elena Dukhovlinova were researchers on the North Carolina side.

Unlike many specifications in this position, this one states its own priority position clearly. The text records that the application is a divisional of Australian application 2019287676 filed 24 May 2019, is related to PCT/US2019/034027, and claims priority from United States provisional application 62/684,513 filed 13 June 2018. That makes the priority country the United States.

Why It Matters

The interesting thing about this application is how narrow it is prepared to be. A filing that claimed “a CAR that binds CSPG4” would be far broader and almost certainly unallowable, because anti-CSPG4 antibodies have existed for decades and the background section itself catalogues previous attempts to use them, including short hairpin RNA delivery, an scFv fused to TRAIL to deliver a pro-apoptotic signal, and a bi-specific T cell engager binding CSPG4 and CD3. Instead the claim is pinned to particular sequences, which is a bet that the sequences themselves are the contribution.

The culture data support that reading. Two CARs aimed at the same antigen, built from versions of the same parent antibody, behaved differently: one left surviving tumour cells at a one to one ratio, the other cleared the plate at one to five. Small changes in the binding domain are known to change a CAR’s behaviour through mechanisms that have nothing to do with affinity, including how readily the scFv clumps together on the cell surface and causes the T cell to signal in the absence of any antigen, which exhausts it. The specification does not diagnose why the reworked binder is better. It simply shows that it is.

Readers should note what is absent. Every result in this document is a cell culture coculture. There are no animal experiments, no tumour models in mice, no persistence or trafficking data and no clinical results of any kind. That gap matters more for a solid tumour CAR than for a blood cancer CAR, because the hard parts of solid tumour CAR therapy, reaching the tumour through disorganised vasculature, surviving a hostile microenvironment and persisting long enough to matter, are exactly the parts a culture dish cannot test. There is also a loose end in the drafting: the definitions section refers the reader to an Examples section for a detailed description of the production and characterisation of the antibodies, and no such section exists in this specification. All the experimental detail sits in the brief description of the drawings.

None of that makes the filing unreasonable. Claim 1 is a composition of matter defined by sequence, and the standard for supporting that kind of claim is different from the standard for supporting a therapeutic promise. But a reader who wants to know whether these cells work in a person will not find the answer here.

Related Concepts

• Chimeric antigen receptor – the engineered receptor family this application claims a member of, joining an antibody fragment to T cell signalling machinery.
• CSPG4 – the chondroitin sulfate proteoglycan the claimed receptor binds, long known as a melanoma-associated antigen.
• Proteoglycan – the class of sugar-decorated surface molecules CSPG4 belongs to, which is part of why it is awkward to target.
• Tisagenlecleucel – the first approved CAR T therapy, aimed at CD19, and the benchmark solid tumour CARs are measured against.
• Melanoma – the cancer in which CSPG4 expression was first characterised and the source of the cell lines used in the culture experiments.

---

AU 2026202199 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.