The trouble with a good cancer target is that it may have an entirely respectable day job. CAIX, a protein found on many clear cell kidney cancers, also appears in healthy bile-duct tissue. An engineered immune cell sent to attack it can therefore arrive at the correct molecular address and do the wrong thing. The identification was accurate. The permission to kill was too generous.
- The problem: cancer markers can also occur on healthy cells.
- The rule: Link’s lead therapy requires CAIX and ENPP3 on the same cell.
- The milestone: Phase 1 dosing has begun at MD Anderson.
- The stakes: a useful recognition circuit still needs evidence in patients.
Link Cell Therapies is built around that distinction. Its proposed solution sounds almost bureaucratic: ask for a second form of identification. In the company’s lead candidate, LNK001, the second marker is ENPP3. The intended killing response requires both markers together. A cell carrying only one should fail the check.
That little word, “and,” does considerable work. It changes the search from finding a protein exclusive to cancer to finding a combination that separates malignant cells from normal ones. Each member of the pair may be imperfect. Their overlap can be much more useful. This is the company’s central wager, and kidney cancer is now its first clinical examination.
A good target with bad company
CAR-T treatment turns T cells into engineered recognizers. A chimeric antigen receptor gives the cell a way to detect a chosen target and respond. The approach has changed treatment for certain blood cancers. Solid tumors present an awkward recognition problem: many attractive antigens are shared with tissues a patient needs to keep.
The phrase researchers use is “on-target, off-tumor toxicity.” It describes an attack on the intended protein in an unintended place. Improving the accuracy of protein recognition alone cannot settle it. A perfectly obedient cell can still be dangerous if its instructions include healthy tissue.
Link’s 2024 kidney-cancer poster made the problem concrete. Conventional CA9-directed CARs, using another name for the CAIX target, caused liver injury in an experimental model. LINK cells targeting the pair did not show that elevated liver-injury signal in the model. The distinction helped justify further development. It was a preclinical comparison, with the limits that implies.
Two markers are the instruction.
The first gate leaked
The foundational research did not begin with a perfectly behaved switch. Its first LINK designs could respond to cells carrying only one antigen. For a system meant to insist on two, this was a rather inconvenient habit.
The researchers worked with LAT and SLP-76, proteins involved in propagating signals inside T cells. They incorporated them into separate engineered receptors. The cooperation of those signaling components offered a way to make activation depend on two external inputs.
But cooperation inside a cell is complicated. The team modified cysteine residues in one receptor component and removed GADS-binding regions to tighten the gate. Those changes reduced unwanted responses to single-marker cells while preserving activity against double-positive targets in the experimental systems.
The lesson sits in the revisions. Writing AND on a diagram does not make a biological circuit obey it. The researchers had to test the near misses, inspect the leakage, and change the machinery. Their 2023 Nature paper described rapid, reversible logic-gated activity, supported by laboratory and animal experiments. Human treatment was still a separate assignment.
“Drawing insights from native T cell signaling networks”Robbie Majzner, co-founder, March 2023
The practical idea is available to anyone designing a selective system: study the almost-right inputs. A demonstration against the ideal target tells you what a circuit can do. The single-marker controls tell you whether it understands the rule.
The pair had to earn its place
A circuit that recognizes two markers is useful only if the right pair exists. Link’s target-discovery work therefore asks two questions together: where do the markers coincide in cancer, and where might they coincide in normal cells?
A September 2026 research abstract described a curated dataset containing more than 29 million normal human cells and 390,000 malignant clear cell renal cancer cells. The team scored over 28 million possible surface-protein pairs. CAIX and ENPP3 emerged as the selected combination.
Target-discovery research, reported September 2026
Computational selection was followed by tissue assays. Immunohistochemistry and flow cytometry examined expression in malignant and normal samples. In the abstract’s analysis of 419 tumor cores, the group with high CAIX expression was almost universally high for ENPP3. That observation supports the pairing. It does not mean the same proportion of patients will respond to treatment.
The distinction between nearby proteins and proteins on the same cell matters. A healthy tissue can contain both markers in different cells. Link’s recognition rule is intended to depend on their simultaneous presence on one target. A neighborhood is not an address.
From a paper to a patient’s cells
Link was formed in 2022 around Stanford research by Robbie Majzner and Crystal Mackall, with biotech entrepreneur Brian Slingerland and Samsara BioCapital. In a public post about the company’s launch, Majzner credited Aidan Tousley with the prototype idea. The scientific origin belongs to a research team as well as the names on the company’s door.
The operational team adds different kinds of experience. Chief Scientific Officer Mark Wallet has worked on engineered cell therapies at Century and Fate. Chief Operating Officer Jerry Cacia brings manufacturing and technical-development experience from Roche and Genentech. The move from a publishable experiment to a repeatable medicine demands both kinds of expertise.
LNK001 is an autologous therapy, engineered outside the body using the patient’s own cells. CTMC manufactures it. That relationship puts the experimental treatment within a development and manufacturing operation connected to MD Anderson, where clinical dosing began. The cell’s instructions matter; so does the ability to make the cells for a patient.
On September 28, 2026, Link announced the start of dosing, a new chief executive, Jie D’Elia, and a larger Series A. LNK001 is being evaluated in Phase 1 in advanced or metastatic clear cell renal cell carcinoma. The movement into patients changes the nature of the evidence the company must produce.
Money buys the next question
The first disclosed seed financing was $12 million in May 2022. By the December 2025 stealth-launch announcement, Link reported $92 million raised across seed and Series A, including a $60 million Series A led by Johnson & Johnson’s venture organization. The September 2026 expansion added $30 million to that round, taking the Series A to $90 million.
$90 million total. Capital raised is not a measure of development spending or a treatment price.
Link’s business is therapeutic development: advance its own candidates and work with strategic partners. Its Johnson & Johnson collaboration extends that approach to additional LINK programs. It is not a service patients can order from a website. For a potential participant, the useful next step is the study record and a discussion of eligibility with the clinical team.
A colorectal-cancer program is planned to reach a first-in-human trial in 2027. The broader opportunity is to find other pairs that turn unsuitable single targets into plausible therapeutic combinations. The constraint is built into the idea: if cancer cells do not display both markers, the AND rule can exclude them. If healthy cells do display both, the intended separation may disappear.
Link has made a precise bet rather than answered every problem in solid-tumor therapy. Its distinctive contribution is a recognition rule, a target-search method, and an experimental product now being tested in people. The elegant conjunction has left the diagram. It must now make itself useful.