A drug that protects a mouse heart when given four weeks before a heart attack often fails when given four hours after one. This is the prophylactic-therapeutic gap: almost every cardioprotective, anti-ferroptotic, or anti-inflammatory candidate in the preclinical literature is dosed before or alongside the injury, because that is what is easy to measure, while patients in emergency departments can only ever be treated after the injury has begun. The reason the post-injury regime is not mapped is largely instrumental. Mapping the "point of no return" — the latest time a drug still rescues — requires knowing, continuously, where the tissue sits on the dying cascade, yet the standard tools are terminal: Western blots for GPX4, histology, RNA sequencing at fixed endpoints. Functional readouts are worse than useless early on: in septic mice, ejection fraction still reads "normal" at six hours post-injury while Global Longitudinal Strain is already badly impaired and the chemokine cascade is actively recruiting destructive neutrophils (Ferrostatin-1 improves acute sepsis-induced cardiomyopathy via inhibiting neutrophil infiltration through impaired chemokine axis).
This proposal argues that a strained epitaxial oxide film — a single crystal a few nanometers thick, forced to stretch or squeeze to match its underlying substrate — is a genuinely well-suited transducer for exactly this measurement. Correlated oxides such as $\mathrm{NdNiO}_3$ sit near a metal-insulator transition, a cooperative electronic switch in which resistivity changes by over three orders of magnitude in response to tiny structural or chemical perturbations (Strain-mediated metal-insulator transition in epitaxial ultrathin films of NdNiO3). The authors of that work explicitly note that this sensitivity makes such films promising chemical and environmental sensors, but no one has ever placed one in a biological environment. The core idea of this proposal is a two-part claim: first, that epitaxial strain provides a clean, substrate-defined knob for setting and ratiometrically duplicating the chemiresistive operating point of a nickelate film at physiological temperature in cell-culture medium; and second, that such a film, grown under a layer of cardiac or endothelial cells undergoing hypoxia-reoxygenation or endotoxin injury, can report the onset and progression of the injury cascade continuously, allowing post-insult dosing curves to be mapped densely rather than sampled at a handful of destructive endpoints.
Core Claim
Epitaxial strain engineering supplies the mechanism (a strain-tuned, amplifying electronic phase boundary used as a chemiresistive sensor), and rescue therapy efficacy supplies the application regime (mapping post-injury therapeutic windows that current terminal-assay methodology cannot resolve). The specific claim is that a pair of $\mathrm{NdNiO}_3$ or $\mathrm{LaNiO}_3$ films, one tensile-strained on $\mathrm{SrTiO}_3$ ($+2.6\%$) and one compressive-strained on $\mathrm{LaAlO}_3$ ($-0.3\%$), read out ratiometrically, will detect the redox and inflammatory shifts of an acute injury cascade in vitro hours before any fixed-timepoint terminal assay in current use, and that this continuous signal will let us locate the post-insult dosing window for two well-characterized rescue candidates (Ferrostatin-1 and nuciferine) that are currently only validated as pre-treatments.
The choice of two strain states is not incidental. Asymmetric Orbital-Lattice Interactions in Ultrathin Correlated Oxide Films showed that tensile and compressive strain do not simply shift a nickelate's properties in opposite directions along one axis; tensile strain provokes a symmetry-breaking octahedral "breathing" distortion and a $\sim 0.10$ eV charge-ordering gap that compressive strain does not, while strain tunes ligand-hole density at $100$ meV per percent of lattice misfit. Two differently strained films therefore respond to the same chemical perturbation through different microscopic channels, and their ratio is a built-in differential measurement that suppresses drift, temperature, and nonspecific adsorption — the classical failure modes of bare chemiresistors.
Research Program
The program has three workstreams in sequence, with a hard go/no-go gate between the first and the rest.
The first workstream builds the sensor and establishes that it works in liquid at $37^\circ\text{C}$. Films of $10$–$15$ unit cells ($4$–$6$ nm) of $\mathrm{NdNiO}_3$ will be grown by pulsed laser deposition with in situ reflection high-energy electron diffraction, on both $\mathrm{SrTiO}_3$ (001) and $\mathrm{LaAlO}_3$ (001), following the synthesis and surface-quality standards of Strain-mediated metal-insulator transition in epitaxial ultrathin films of NdNiO3, which reports atomically flat surfaces with roughness under $80$ pm. The honest physical risk, stated up front: the metal-insulator transition in $\mathrm{NdNiO}_3$ is a low-temperature phenomenon, and at $37^\circ\text{C}$ the film is comfortably metallic. The sensor therefore does not rely on switching the full transition but on the steep conductance slope near it. Strain shifts the transition temperature (this is the demonstrated content of the Liu et al. work: $+2.6\%$ tensile preserves the transition, $-0.3\%$ compressive quenches it entirely, with metallicity persisting to $10$ K), and electrolyte gating or mild oxygen-vacancy control shifts it further. The go/no-go criterion is quantitative: a two-terminal film pair in phosphate-buffered medium must show a reversible, repeatable differential conductance response of at least a few percent to micromolar-scale hydrogen peroxide and to recombinant CXCL2 chemokine — the two molecular species that anchor the injury cascade stages we care about — with a signal-to-noise ratio above five over a 48-hour soak. If the strained films cannot hold a stable baseline in culture medium for 48 hours, the program stops and reports the negative result, which would itself constrain the "oxide bioelectronics" idea that reviews like Bioelectronic Medicine: a multidisciplinary roadmap from biophysics to precision therapies gesture toward but never ground in a real material platform.
The second workstream is the calibration experiment, the heart of the proposal. Cardiac-derived HL-1 cardiomyocytes and human cardiac microvascular endothelial cells — the same static-culture systems used in Semaglutide attenuates myocardial ischemia-reperfusion injury by inhibiting ferroptosis of cardiomyocytes via activation of PKC-S100A9 axis and Nuciferine reduces vascular leakage and improves cardiac function in acute myocardial infarction by regulating the PI3K/AKT pathway — will be cultured directly on the film pair, then subjected to hypoxia-reoxygenation or lipopolysaccharide exposure. The sensor pair records conductance continuously while parallel wells are sacrificed at fixed time points for the terminal assays those papers used: GPX4 and MDA levels for ferroptotic progression, VE-cadherin localization for endothelial junction integrity, and qPCR for the $Ccrl2$, $Cxcl2$, $Cxcl3$, $Cxcl5$ chemokine axis that Ferrostatin-1 improves acute sepsis-induced cardiomyopathy via inhibiting neutrophil infiltration through impaired chemokine axis identified as the decisive early cascade (its RNA-seq found 243 genes regulated by both sepsis and Fer-1, dominated by leukocyte chemotaxis). The calibration target is a predictive one: sensor features must identify the time of chemokine-axis surge and the time of GPX4 degradation before the assays run at those times confirm them. The key calibration anchor from the literature is that GPX4 degradation has not yet occurred at 6 hours post-injury — so the sensor must be detecting the earlier chemokine-stage signature, not the ferroptotic end-state, to be useful.
The third workstream uses the calibrated sensor to do what the field has not done: map post-insult dosing curves for two drugs whose rescue efficacy is explicitly unknown. Ferrostatin-1 ($5$ mg/kg in the original study) and nuciferine ($10$ mg/kg) will be added to injured cultures at $+30$ min, $+1$ h, $+2$ h, $+4$ h, $+6$ h, and $+12$ h post-insult, with the sensor continuously reporting whether the cascade signature stalls, slows, or proceeds. The endpoint is whether the continuous trace predicts the 48-hour cell-viability outcome, i.e., whether the sensor lets us say, from a live measurement, whether a given dosing time came before or after the in vitro point of no return. One important interpretive caution comes from within the file itself: Fer-1's pre-treatment benefit in sepsis was mediated by chemokine suppression rather than by canonical ferroptosis inhibition, so the dosing curves for the two drugs should diverge in shape, and the sensor should be able to show that divergence mechanistically — Fer-1's window should close when the chemokine surge is already underway, whereas a genuinely anti-ferroptotic agent should retain efficacy until GPX4 degradation begins. That mechanistic contrast is itself a testable prediction of the platform.
A secondary, deliberately modest characterization thread: chronic exposure of the films to medium will be checked structurally, but we will not lean on deep-learning phase retrieval. MAMBA, the U-Net that reconstructs atomic positions from crystal truncation rod data with $R^2$ values close to 1 on simulated $\mathrm{LaAlO}_3/\mathrm{SrTiO}_3$ interfaces (Epitaxial Thin Film Interface Imaging with Deep Learning), was trained on noise-free synthetic data assuming in-plane symmetry and degrades beyond 16 layers — it is exactly the wrong tool for a disordered liquid-solid interface, and the proposal treats that as a known limitation rather than a workstream.
Baselines
The naive reference point is the standard terminal-assay protocol itself: injured cultures with no drug, sampled at fixed time points for GPX4, MDA, and chemokine qPCR, exactly as run in the semaglutide and nuciferine studies. Its published reference numbers, taken from the file, set the scale of the problem: untreated acute myocardial infarction controls reach an ejection fraction of approximately 28%, and untreated cecal ligation and puncture sepsis reaches approximately 70% mortality within 48 hours (Nuciferine reduces vascular leakage and improves cardiac function in acute myocardial infarction by regulating the PI3K/AKT pathway; Ferrostatin-1 improves acute sepsis-induced cardiomyopathy via inhibiting neutrophil infiltration through impaired chemokine axis). Beating this baseline means the sensor detects injury-stage transitions strictly earlier than the earliest fixed time point at which the terminal assays register them, at the same cost per well.
The strongest existing method is the pre-treatment protocol with the most sensitive functional readout in the literature: Ferrostatin-1 at 5 mg/kg administered before cecal ligation and puncture, monitored by speckle-tracking echocardiography with Global Longitudinal Strain rather than ejection fraction, which reduced the roughly 70% 48-hour mortality and detected cardiac impairment at 6 hours when ejection fraction still read normal (Ferrostatin-1 improves acute sepsis-induced cardiomyopathy via inhibiting neutrophil infiltration through impaired chemokine axis). This is the strongest competitor because GLS is the only cited modality that detects the early injury window at all. The comparison will be run in vitro: GLS-style imaging is inapplicable to a culture dish, so we replicate its temporal resolution by high-frequency sampling of terminal assays (every 30 minutes for the first 6 hours post-insult) in a matched cohort, giving the incumbent methodology its best possible temporal resolution at a cost in animals and wells that the sensor approach does not pay. Beating it means matching that dense sampling's detection latency at a small fraction of its consumable cost, while additionally resolving which cascade stage (chemokine surge versus ferroptotic commitment) is underway.
The key ablation isolates epitaxial strain as the operative ingredient. Identical nickelate films will be grown beyond their critical thickness, so that misfit dislocations relax the strain, on the same run of the same deposition system, and otherwise-identical devices will be made from sputtered polycrystalline nickelate. These relaxed and polycrystalline controls will be run in the same chambers, same cell batches, same time points, with the same readout electronics. If the relaxed films detect the injury cascade as early and as specifically as the strained films, then epitaxial strain — and this entire framing — contributes nothing beyond what any nickelate resistor provides, and the honest conclusion would be that the interesting variable is the material's proximity to a Mott instability, not strain engineering. Conversely, if the tensile-on-STO and compressive-on-LAO pair outperforms both controls and the two strain channels respond to different cascade stages, the ratiometric strain-engineering claim is confirmed.
Why not just…?
Why not just use commercial electrodes or PDMS organ-on-chip platforms? This is the most serious objection, and it deserves a straight answer. Gold electrochemical electrodes and stretchable polymer membranes are mature, cheap, and biocompatible, and if a functionalized gold electrode can detect the chemokine surge at the same moment the nickelate pair does, this proposal's advantage evaporates. The case for the oxide rests on amplification: a material near a cooperative phase boundary converts sub-percent surface perturbations into orders-of-magnitude conductance change, whereas a passive electrode reports analyte adsorption linearly and needs a receptor layer to get specificity at all. The published basis for expecting an amplification advantage is that coherent strain alone moves $\mathrm{NdNiO}_3$ resistivity by over three orders of magnitude and modulates ligand-hole density at 100 meV per percent misfit (Strain-mediated metal-insulator transition in epitaxial ultrathin films of NdNiO3; Asymmetric Orbital-Lattice Interactions in Ultrathin Correlated Oxide Films). But that basis is a solid-state result, not a biological one, so the proposal includes the head-to-head as a first-class experiment rather than an argument: a gold interdigitated electrode array in the same chamber, same medium, same cells, same clock. The comparison is decided by detection latency and false-alarm rate over 48 hours, and we commit in advance to reporting it either way.
Why not just run rigorous post-injury dosing time courses in animals? In the long run, yes — no in vitro sensor platform replaces an in vivo time-to-treatment study, and the ultimate medical question is protocol design, not materials science. But the file documents two concrete reasons the brute-force approach under-delivers on its own. First, the standard functional readout is blind exactly where the information is: ejection fraction reads normal at 6 hours post-sepsis while tissue destruction is underway, so an animal study calibrated on ejection fraction will misclassify the injury stage at exactly the dosing times that matter (Ferrostatin-1 improves acute sepsis-induced cardiomyopathy via inhibiting neutrophil infiltration through impaired chemokine axis). Second, post-injury delivery is physically impeded by microvascular collapse and endothelial junction breakdown, which is why the utmd gene-delivery work exists at all (Cardiac-targeted delivery of nuclear receptor RORα via ultrasound targeted microbubble destruction optimizes the benefits of regular dose of melatonin on sepsis-induced cardiomyopathy). What the sensor platform offers is not a replacement for animal dosing studies but a cheap, continuous pre-mapping layer that tells the pharmacologist which post-insult time points carry information and which are already past the point of no return, so the animal study spends its animals on the informative part of the curve. If the in vitro windows turn out not to predict the in vivo ones, that is a decisive negative result we would report.
Why not just relax the "strained" requirement and use any oxide sensor, or skip the sensor and improve the assays? Skipping the sensor reduces to the first alternative, already addressed. The strain-specific version of the question is answered by the ablation above, but there is a deeper reason strain is load-bearing rather than decorative: the file's own evidence shows strain is not a scalar dial. Tensile and compressive strain trigger categorically different lattice responses — a breathing distortion and charge order under tension, preserved metallicity under compression (Asymmetric Orbital-Lattice Interactions in Ultrathin Correlated Oxide Films; Orbital control in strained ultra-thin LaNiO 3 /LaAlO 3 superlattices). That asymmetry is what makes a two-strain ratiometric pair chemically selective rather than merely sensitive, and it is unavailable to a relaxed or polycrystalline film, which has no substrate-defined operating point to shift. Whether that selectivity actually separates chemokine-stage from ferroptosis-stage signals is precisely what workstream two is designed to show; if it does not, the fallback claim is the weaker but still useful one of an early-warning continuous viability monitor, and we would say so.
Decisive Checks
Three results decide the proposal. First, the go/no-go: strained film pairs must hold a stable differential baseline in culture medium for 48 hours and respond measurably to hydrogen peroxide and CXCL2 at micromolar and nanomolar scale respectively; failure ends the program with a published negative result constraining oxide-bioelectronics proposals. Second, the calibration prediction: in at least three independent injury runs, sensor features must flag the chemokine-surge time point before the qPCR assay run at that time point confirms it, and must flag GPX4 commitment no later than the Western blot time point that first shows degradation (which, per the Ferrostatin-1 study, is later than 6 hours post-insult — a concrete, literature-anchored deadline). Third, the dosing-window divergence: Fer-1's post-insult efficacy window in vitro must close measurably earlier than a directly anti-ferroptotic control's, consistent with its chemokine-mediated rather than GPX4-mediated mechanism, and the continuous sensor trace — not post-hoc viability — must be what predicts each drug's closing time. If all three hold, the deliverable is the first strain-engineered solid-state device doing biological work in the post-injury regime, and the first dense, mechanistically resolved post-insult dosing curves for two drugs whose rescue efficacy is today entirely unknown.
References
- Ferrostatin-1 improves acute sepsis-induced cardiomyopathy via inhibiting neutrophil infiltration through impaired chemokine axis
- Strain-mediated metal-insulator transition in epitaxial ultrathin films of NdNiO3
- Asymmetric Orbital-Lattice Interactions in Ultrathin Correlated Oxide Films
- Bioelectronic Medicine: a multidisciplinary roadmap from biophysics to precision therapies
- Semaglutide attenuates myocardial ischemia-reperfusion injury by inhibiting ferroptosis of cardiomyocytes via activation of PKC-S100A9 axis
- Nuciferine reduces vascular leakage and improves cardiac function in acute myocardial infarction by regulating the PI3K/AKT pathway
- Epitaxial Thin Film Interface Imaging with Deep Learning
- Cardiac-targeted delivery of nuclear receptor RORα via ultrasound targeted microbubble destruction optimizes the benefits of regular dose of melatonin on sepsis-induced cardiomyopathy
- Orbital control in strained ultra-thin LaNiO 3 /LaAlO 3 superlattices