Kidney Chip Biomarker Kinetics Link Drug Injury to Conditional Human Simulations
A 2019 study paired human proximal tubule chips with pharmacokinetic and biomarker models. Early KIM-1 increases captured drug injury, while later decreases had different meanings across conditions; modeled immune effects helped extend the short-lived chip signal toward reported clinical profiles.
Following an Injury Marker Over Time
Researchers at MIT and the University of Washington combined a human kidney microphysiological system with computational models to investigate drug-induced proximal tubule injury. This 2019 study used the shed protein kidney injury molecule-1, or KIM-1, as a time-dependent injury signal and asked how in vitro measurements might inform human biomarker profiles.
The chip contained primary human renal proximal tubule epithelial cells from one donor, designated HIM-31, in a perfused three-dimensional tubule. The estimated cell number was 5,000 per chip. A parallel static two-dimensional culture used the same cell source, with an estimated 10,000 cells per well. Comparisons of baseline KIM-1 accounted for sample volume and seeded cell number.
Untreated chips shed approximately tenfold less KIM-1 than the static cultures over the measured intervals. The chip cultures also maintained low shedding and viability after the static cultures had deteriorated. These observations provided the baseline context for comparing responses to cisplatin, rifampicin and gentamicin.
Drug experiments began eight days after seeding. Chip cultures received two-day or ten-day exposure, whereas the static comparison used ten-day exposure. KIM-1 was measured every 48 hours, with additional chip samples at six and 24 hours. Metabolic activity measured with PrestoBlue supplied a separate indicator of the cellular response.
A Falling KIM-1 Signal Did Not Have One Interpretation
After six hours, every tested drug and concentration produced a significant KIM-1 increase. The increase was dose-related for cisplatin but not for rifampicin or gentamicin. By 24 hours, many treatment groups had returned to baseline or below, despite continued exposure.
The accompanying metabolic-activity result helped distinguish these responses. In the medium- and high-gentamicin groups, KIM-1 was low at 48 hours and metabolic activity was substantially reduced at the day 10 assessment after the two-day exposure. These separate measurements are consistent with cell loss but do not establish when it occurred. High-dose cisplatin also reduced metabolic activity.
Other conditions showed low KIM-1 without the same metabolic-activity loss. After a two-day exposure, some rifampicin and low-gentamicin groups later returned toward baseline shedding. The authors discuss repair-related processes as one possible explanation, rather than treating every low biomarker result as either recovery or cell death.
Exposure duration changed the rifampicin comparison. High-dose treatment over ten days reduced both metabolic activity and KIM-1 in the chip, whereas the two-day exposure did not produce those sustained changes. The corresponding static culture did not show the same high-dose metabolic-activity response. Across the three drugs, the chip and static formats therefore differed in baseline stress, response magnitude and temporal behavior.
Human Translation Added Pharmacokinetics and an Assumed Immune Component
The computational workflow first used a whole-body physiologically based pharmacokinetic model to estimate renal drug exposure in 100 virtual patients. It then linked exposure to KIM-1 shedding rates interpolated from the chip measurements and scaled the rates to the modeled number of human nephrons. The virtual patients were simulated parameter sets, separate from the single experimental cell donor.
The initial translation generated an early KIM-1 peak for cisplatin and rifampicin, but not gentamicin. The source notes a large gap between tested and typical clinical gentamicin concentrations as a possible reason for that discrepancy. The in vitro system’s lack of immune cells also limited reproduction of sustained clinical biomarker elevations.
To explore the latter issue, the investigators added a putative neutrophil contribution computationally. Its recruitment dynamics came from a mouse injury model, and its effect on shedding came from a separate cell model; comparable human effects were assumed. This addition produced cisplatin and rifampicin plasma peaks of roughly 1,000 pg/mL at 12–24 hours and urinary elevations of roughly 3,000 pg/mL after 24 hours. These were simulated values compared with published clinical profiles, not newly measured patient outcomes.
The study also used rifampicin exposure–effect simulations to illustrate how the workflow could explore dosing-related trade-offs. That analysis was retrospective and conditional on the chip injury thresholds, pharmacokinetic assumptions and drug-effect model. It did not establish a new clinical dosing recommendation.
The authors identify broader drug panels, donor variation, drug/protein binding losses and more detailed renal transport and immune biology as needed extensions. The study’s contribution is the connection among an early cellular biomarker response, sustained injury measures and a transparent translation hypothesis, with the immune and population components remaining computational.