ADAMTS-5 still looks central in osteoarthritis, but the clinic has not rewarded the biology
Osteoarthritis remains one of rheumatology’s clearest examples of a disease in which a persuasive molecular target has not yet produced a dependable therapy. ADAMTS-5 sits near the center of that mismatch. In experimental osteoarthritis, the enzyme has strong genetic and pharmacological support as a major aggrecanase, and that has long made it an attractive place to intervene when cartilage breakdown is still underway. In human disease, though, the enzyme’s contribution is more conditional than absolute, shaped by tissue context, disease stage, and the broader network of inflammatory and mechanical stress that drives joint failure.
The review places ADAMTS-5 in that tension between mechanistic certainty and therapeutic uncertainty. Its role in aggrecan degradation gives it real biological authority, because aggrecan loss strips cartilage of its ability to resist compression and preserve joint function. Yet a target can be central in a pathway without being sufficient to change the course of the disease once structural damage, compensatory proteases, and altered joint mechanics are established. Osteoarthritis is not a single-enzyme disorder; it is a whole-joint process in which cartilage, synovium, subchondral bone, and mechanical loading all push one another forward.
One sentence captures the central problem: enzyme inhibition has not reliably become symptom relief.
Why a strong aggrecanase target can still fail in human osteoarthritis
The review frames ADAMTS-5 inside a wider network that includes inflammatory signaling, mechanotransduction, extracellular-matrix remodeling, and tissue adaptation. That framing matters because aggrecan degradation does not occur in isolation. In an osteoarthritic joint, inflammatory mediators can amplify catabolic signaling, abnormal loading can reshape matrix turnover, and downstream tissue adaptation can blunt the effect of blocking a single protease. Even when ADAMTS-5 is active, other degradative pathways may compensate, especially if the joint has already moved beyond the stage where one dominant effector drives most of the damage.
This is where translational optimism often weakens. A target that looks dominant in preclinical models may only account for part of the pathology in people, particularly when patient populations are heterogeneous. Some patients may present with inflammation-heavy disease, others with mechanically driven degeneration, and still others with a mixture of both. If ADAMTS-5 activity is greatest only in a subset of these states, then broad systemic inhibition will dilute any signal of benefit across a trial population that is too mixed to respond uniformly.
The review emphasizes disease stage as a major reason for this disconnect. Early cartilage injury may be more vulnerable to protease blockade than advanced disease, where structural attrition and maladaptive remodeling have already become self-sustaining. Once the joint architecture has changed, stopping one enzymatic contributor may not reverse the biologic momentum already in place. In that setting, the treatment may be mechanistically sound and still clinically underwhelming.
ROCCELLA exposed the gap between structural rationale and measurable benefit
The clinical record described in the review is direct. The Phase II ROCCELLA trial of GLPG1972/S201086 showed no significant structural or symptomatic benefit. That result is not a minor footnote. It marks the boundary where a compelling target meets the demands of patient-centered outcomes and fails to deliver a detectable advantage.
For osteoarthritis therapeutics, this is especially consequential because structural biomarkers and symptomatic outcomes do not always travel together. A drug can plausibly influence matrix turnover without producing meaningful pain reduction, and it can alter a biochemical marker without changing the mechanical environment that patients actually experience as stiffness, swelling, and disability. Biomarker-outcome dissociation is one of the most important reasons cartilage-targeting therapies have struggled. If the treatment touches the pathway but not the outcome that matters to patients, the clinical value remains limited even when the mechanism is well established.
A different development path has produced a different kind of evidence for the anti-ADAMTS-5 nanobody M6495. So far, it has been evaluated mainly for Phase I safety, pharmacokinetics, and pharmacodynamics rather than efficacy. That distinction is important because early-phase studies can confirm exposure and target engagement without proving that the intervention changes joint outcomes. In osteoarthritis, pharmacodynamic activity is only the opening move; the harder question is whether that activity survives contact with the complexity of a living joint over time.
Redundancy, delivery, and patient selection limit translation
Protease redundancy sits near the center of the translational problem. Even if ADAMTS-5 is a major aggrecanase, it is not the only enzyme involved in matrix breakdown. The joint has multiple escape routes, and when one pathway is blocked, others may preserve the degradative state. Local tissue adaptation can also reduce the apparent impact of inhibition, especially if the extracellular matrix has already reorganized around chronic injury.
Delivery limits deepen the challenge. Uniform systemic inhibition may not generate the tissue-level concentration needed where cartilage damage is most active, while also raising the possibility of exposure in tissues where the enzyme may have other roles. A locally controlled intervention could, in theory, fit the biology more closely than a broad systemic approach. The review points toward that possibility without claiming it has already been solved.
Precision also depends on identifying the right patients. Molecular endotyping, single-cell profiling, spatial profiling, and other data-driven stratification tools could separate ADAMTS-5-driven disease from forms of osteoarthritis in which other pathways dominate. If an enzyme is truly relevant only in a biologically defined subset, then the trial design has to reflect that reality. Broad enrollment may be efficient for recruitment, but it can be inefficient for signal detection.
Local control and combination therapy may fit the biology better than blanket inhibition
The review argues for stage-specific and locally controlled intervention rather than uniform systemic inhibition. That approach follows the logic of the disease itself. A cartilage protease is most likely to matter when cartilage is still biologically salvageable, when matrix turnover is active, and when the broader inflammatory and mechanical environment has not yet overwhelmed the tissue response. Treating late disease as if it were early disease has repeatedly produced disappointment in osteoarthritis drug development.
Combination therapy may also be more plausible than single-agent blockade. A protease inhibitor paired with a therapy that addresses inflammation, loading, or downstream tissue remodeling could, in principle, cover more of the disease network than any one intervention alone. The review does not overstate this as a proven path; it positions combination therapy as a rational strategy derived from the biology rather than from a completed clinical proof.
ADAMTS-5 therefore serves as a model for osteoarthritis translational failure and not merely as one more disappointing target. Its story explains how a biologically compelling mechanism can lose clinical force when timing, tissue context, redundancy, delivery, and patient selection are not aligned. The open question now is not whether the enzyme matters at all, but in which patients, at which stage, and with what kind of local or combination strategy it might still be worth pursuing.