Preprint: TROP2 ADC Model Separates Antigen Delivery from Payload Sensitivity

Two routes to the same payload

Sacituzumab govitecan can deliver its SN-38 payload after binding TROP2 on a tumor cell, but SN-38 released into the circulation can also enter a tumor. A preprint from Johns Hopkins University and Merck investigators asks how those routes interact with tumor heterogeneity. Their quantitative systems pharmacology (QSP) model combines drug disposition, cancer-cell killing and tumor–immune interactions to examine the sources of variation in response to this antibody–drug conjugate (ADC).

The investigators added an ADC module to an existing triple-negative breast cancer model. Four compartments represent the circulation, peripheral tissues, tumor and tumor-draining lymph node. Within the tumor, two cell clones have high or low TROP2 expression and different intrinsic sensitivities to SN-38. These linked features are central to interpreting the simulations: a response associated with higher antigen expression can also reflect a greater fraction of payload-sensitive cells.

Calibration and a separate trial comparison

The model represents ADC binding and internalization, intracellular payload release, and movement of unconjugated SN-38 between compartments. Sacituzumab govitecan was simulated at 10 mg/kg on days 1 and 8 of a 21-day cycle. Published experimental and clinical data supplied the parameterization.

Latin hypercube sampling varied 34 parameters to generate virtual patients. The investigators calibrated response-category frequencies against the phase I/II basket trial, excluding non-evaluable clinical participants when calculating the reference proportions. An initial 400 virtual patients were used for calibration after a cohort-size check. The calibrated distributions were then compared with ASCENT phase III outcomes without further calibration; confidence intervals used 244 evaluable virtual patients. A separate stability analysis supported 1,200 virtual patients for subsequent sensitivity analyses, with simulated tumors failing the initial size criterion excluded.

In the ASCENT comparison, the predicted partial-or-complete response proportion was 33.2%, against 34.0% in the clinical data. Predicted stable disease was 38.5% versus 39.3%, and progressive disease was 28.3% versus 26.6%. Agreement across these categories tests a population-level output, while the model's mechanistic interpretation comes from additional perturbations of its equations.

Delivery and sensitivity have different effects

To separate payload sources, the investigators disabled transport of circulating free SN-38 into the tumor. Under that comparison, TROP2-mediated delivery accounted for approximately 2.4% of total intracellular SN-38 exposure in the antigen-high clone and 0.4% in the antigen-low clone, relative to the corresponding simulations with plasma-to-tumor transport enabled. The model therefore attributed most intracellular exposure to systemically released payload under the studied assumptions.

Tumor composition still strongly affected response. Simulations containing only the antigen-high, more payload-sensitive clone showed complete regression, whereas a starting antigen-low fraction of 25% allowed regrowth after approximately 3 months. Across virtual patients, response increased with TROP2 expression, but the authors interpret that pattern through its covariation with SN-38 sensitivity. It does not isolate antigen abundance as the principal cause of the modeled efficacy difference.

Sensitivity analyses identified payload potency, cytotoxicity and tumor-vasculature parameters as influential. Immune parameters had more modest effects. Responding simulations also showed higher T-cell densities, attributed in the model to cancer-cell killing, antigen release and subsequent immune activation. The paper discusses how this could inform ADC–checkpoint-inhibitor research, but it does not present a quantitative validation of combination efficacy.

The assumptions bound the mechanistic result

The model omits payload effects on immune cells, intracellular SN-38 glucuronidation, acquired resistance and new metastatic lesions. Clone growth rates were assumed identical, and differences between lines of therapy were not mechanistically resolved. The authors also note uncertainty about direct growth effects from TROP2 blockade.

The study provides a way to interrogate competing explanations for a trial-level response pattern. Its exposure fractions describe this sacituzumab govitecan model, with its linked antigen and sensitivity assumptions. The preprint's trial comparison supports population-level consistency; further experimental discrimination is needed to establish the relative payload-delivery mechanisms in tumors.