The Biomarker Paradox at the Heart of Qilu's CLDN6 ADC

When Qilu Pharmaceutical presented Phase I data on its CLDN6-targeted antibody-drug conjugate at AACR 2026 last week, the top-line number looked solid: a 50 percent objective response rate in platinum-resistant ovarian cancer. Drill into the details, and the picture shifts. The confirmed response rate is 28 percent. Just 12 of 28 enrolled patients were CLDN6-positive. And two patients with zero detectable CLDN6 expression had partial responses anyway.

That last detail is the one worth sitting with.

CLDN6 is a tight junction protein normally expressed in testis and fetal tissue, largely absent in adults. The therapeutic hypothesis is straightforward: surface the protein on tumor cells, deliver a cytotoxic payload, watch the cancer die. QLS5132 is designed to do exactly that. It showed activity in patients who did not have the target the drug was designed to hit. Qilu's investigators have offered a plausible explanation — a bystander effect, where the payload diffuses into neighboring cells regardless of whether the original cell expressed CLDN6. That is biologically plausible and consistent with how some other ADCs behave. It is also the kind of explanation that holds until it does not.

The competitive landscape does not make the paradox any easier to resolve. Daiichi Sankyo abandoned its CLDN6 ADC, DS-9606, in January 2026 after a strategic portfolio review. The company did not cite safety concerns. The 15 percent ORR looked uncompetitive in a space where investors expect more. BioNTech went further, terminating its CLDN6 CAR-T program BNT211 and reporting disappointing results for its mRNA-encoded T-cell engager BNT142 at ASCO 2025. The field's two largest players exited the same target within eighteen months of each other.

TORL Biotherapeutics is the remaining CLDN6 bet. Its ADC, ixotatug vedotin (TORL-1-23), posted a 50 percent confirmed ORR in the same indication. The critical difference: those responses came exclusively from CLDN6-positive patients. The biomarker did the work it was supposed to do. That is the standard against which Qilu's results have to be read, and against which they look puzzling.

QLS5132 used a topoisomerase-1 inhibitor payload, a mechanism that has powered several successful ADCs including trastuzumab deruxtecan. TORL-1-23 carries an MMAE warhead. Daiichi's abandoned DS-9606 used a PBD prodrug. Three different payloads, three different outcomes — which raises the possibility that the payload, the linker, or the dosing schedule matters more than the target itself.

Qilu has not yet published translational data explaining why CLDN6-negative patients responded. Until it does, the bystander explanation remains a placeholder. If confirmed, it would mean CLDN6 is a useful address label but not the actual driver of efficacy — a meaningful distinction for any developer building a companion diagnostic or selecting patients by expression level. If the responses in CLDN6-negative patients are later shown to be unconfirmed or non-durable, the paradox dissolves and the story becomes a standard ADC trial with a disappointing confirmation rate.

The 28 percent confirmed ORR, while lower than the headline number, is still clinically meaningful in a population that has received a median of five prior lines of therapy. Platinum-resistant ovarian cancer has no shortage of treatment-refractory patients. The disease control rate across all dose levels was 94.4 percent. Safety looked manageable: 32 percent of patients had grade three or higher treatment-related adverse events, primarily hematologic, with no discontinuations or deaths attributed to the drug.

What Qilu does next will determine whether this is a paradox or a puzzle. A biomarker analysis showing that CLDN6-negative responders share some other molecular feature — Fc receptor activity, tumor microenvironment, a different claudin family member — would reframe the entire program. Silence on that question will leave the field wondering whether QLS5132 is a CLDN6 drug that works for reasons the developers do not yet understand, or a drug that works despite the target. Those are not the same story.

For now, the most honest reading of the data is that Qilu has a drug that appears to work regardless of its ostensible target, in a field where the two largest developers concluded the target was not worth pursuing. Whether that is an opening or a warning sign depends on answers Qilu does not yet have.

QLS5132 is in Phase I. TORL-1-23 is in Phase I. Neither has answered the question that matters most: why does this work at all?

Sources: AACR News Release | ADC Review | ApexOnco | Fierce Biotech | Fierce Pharma | ASCO 2025