Facet-Controlled Zinc and Copper on Clay: Less Metal per Pathogen Killed in Poultry

Farms that stop feeding antibiotics to animals often use copper and zinc minerals to control gut pathogens instead. These metals work, but most of what is fed passes straight into manure and builds up in soil. This proposal asks whether the shape of the metal crystals can cut that waste. Crystals present different faces, or facets, to their surroundings, and some faces are much more chemically active than others. Chemists already steer which faces a crystal ends up showing by adding molecules that stick to one face and slow its growth, a technique called facet-selective growth inhibition. We propose to grow zinc-oxide and copper-oxide deposits on a clay carrier with feed-safe additives, methionine and urea, that favor the reactive faces. The target is to kill the same amount of Salmonella in chicks with substantially less metal fed and excreted.

The second part of the proposal concerns how the animal trials are run. Chick challenge trials are small and have to be repeated in sequential rounds, and the infection strength, chick batches, and clay batches all drift from one round to the next. Comparing a new formulation against a control group from an earlier round is the same "historical control" error that inflates effects in clinical research. We therefore build a concurrent-control design with a negative-control arm into every round. We also measure directly how wrong a historical-control comparison would have been, which has not been quantified for livestock feed-additive trials.

Core Claim

For a fixed metal loading on palygorskite clay, the fraction of high-reactivity facets on the zinc and copper deposits changes how effectively the additive kills bacteria per gram of metal, both in simulated gut fluid and in Salmonella-challenged chicks. The additives methionine and urea can set that facet fraction. The central scientific question is whether this benefit comes from anything beyond faster release of metal ions. We treat that question as the decisive experiment rather than assuming the answer. If ion release explains everything, the result is still useful: facet control then becomes a way to tune release speed. But the claim that the crystal surface itself matters would be refuted.

Grounding and Relation to Prior Work

The starting point is Copper/Zinc-Modified Palygorskite Protects Against Salmonella Typhimurium Infection and Modulates the Intestinal Microbiota in Chickens. Zhang et al. loaded copper and zinc onto palygorskite clay by swapping them into the clay's structure in water (ion exchange). The best material, Cu/Zn-Pal-1, killed Salmonella in vitro at a minimum bactericidal concentration of 5.0 mg/ml. Fed at 5 g/kg, it cut fecal Salmonella from about 10^5.7 to 10^3.2 CFU/g by day 20 after infection and raised body weight of infected chicks from 233 g to 280 g, against 296 g in healthy birds.

Two findings in that paper motivate this proposal. First, the authors did not control or characterize which crystal faces their metal deposits exposed. Second, a plain mixture of metal salts and clay did not kill bacteria, so how the metal is structured on the clay matters. That paper also used a single dose, a single trial batch, and did not measure metal in manure. We add facet control, a dose-response study, excretion measurements, and a multi-round design that accounts for drift over time.

The method for controlling facets comes from outside animal science. Several papers show that one small molecule, sticking selectively to one crystal plane, reshapes the whole crystal:

We reuse this mechanism but change two things. We choose additives that are already accepted feed ingredients (methionine is a feed amino acid, and urea is used in ruminant feed), and we grow the crystals inside a clay carrier rather than in clean solvents or battery electrolytes.

Two limitations noted in the facet literature shape the design. Additive molecules can come off the crystal surface, and they can become trapped inside the growing crystal. Behavior of engineered facets in complex biological fluids has also not been characterized. We therefore measure facet stability after incubation in simulated digesta, rather than assuming facets survive in the gut. Atomistic Modeling of Crystal Nucleation and Growth supplies the idea of a "shape diagram", a map of which crystal shape results from which growth conditions. We build this map empirically, by varying additive concentration, rather than by simulation.

The trial design borrows from human clinical epidemiology.

  • Why historical controls mislead. Clinical studies show that improving background care, operator learning, and outside shocks all make historical controls misleading (Yuchen Bai et al.; Cong Liu et al.). Their livestock analogues are gradually refined challenge protocols, supplier changes in chick batches, and drift in clay batches.
  • Comparing study designs. Comparative performance of the concurrent comparator design compared concurrent-comparator and historical-comparator designs against each other, and used outcomes that should show no effect to check its estimates.
  • Borrowing past controls safely. Doubly Protected Estimation for Survival Outcomes Utilizing External Controls borrows external controls only when their outcomes match the concurrent controls. It uses a penalized regression that down-weights drifted historical subjects.

We take the concurrent-control and negative-control ideas directly from these papers. We use the selective-borrowing idea only as a secondary, pre-registered analysis, adapted from survival time to a continuous shedding outcome. The research file lists no prior work linking livestock trials to temporal confounding, and the temporal-confounding literature notes that these methods have rarely been applied outside human medicine.

Research Program

Synthesis and facet mapping. The materials work starts from Zhang et al.'s ion-exchange protocol. Once the metals are exchanged into the clay, zinc- and copper-oxide deposits are grown in situ, with methionine or urea added at five concentrations while total copper and zinc loading is held fixed. We measure facet exposure with three methods:

  • X-ray diffraction peak-intensity ratios, for example the (002)/(100) ratio for zinc oxide, where (002) corresponds to the polar faces;
  • high-resolution transmission electron microscopy with facet indexing;
  • nitrogen gas adsorption to measure specific surface area, since reactive faces and small particles are easy to confuse.

Metal loading is measured by inductively coupled plasma analysis, and leftover additive trapped in or on the crystals by thermogravimetric analysis.

Release and potency in gut-like fluid. Each material is incubated in simulated chicken cecal fluid containing bile salts, pancreatin, and mucin. We track dissolved zinc and copper over 24 hours and repeat the facet measurements afterward. We then measure the minimum bactericidal concentration against S. Typhimurium in the same fluid, not only in broth. The key quantity is potency normalized in two ways: per milligram of metal, and per unit of free metal ion released over time.

Animal trials. Chick challenge trials follow Zhang et al.'s S. Typhimurium model. Because of animal-facility capacity, they run in four sequential rounds. Every round randomizes chicks, within that round, to five arms:

  • infected and untreated;
  • infected and fed unmodified clay, the negative control, whose true effect should be near zero every round;
  • infected and fed Cu/Zn-Pal-1 at 5 g/kg as the reference;
  • two infected arms fed facet-engineered formulations, each at a lower metal dose.

The primary outcomes are day-20 fecal Salmonella counts and body weight. We also measure copper and zinc concentrations in the excreta, and compute Salmonella reduction per milligram of metal excreted. As exploratory work, we run quantitative PCR on excreta for metal-resistance genes and co-located antibiotic-resistance genes. This speaks to the resistance problem directly, since a feed additive that selects for resistance genes would undo part of the point of replacing antibiotics.

Baselines

The do-nothing reference is the infected-untreated arm run within the same round. It sets the size of the treatment effect under that round's infection strength, and it is the only valid comparison point for that round.

The strongest existing method is Cu/Zn-Pal-1 at 5 g/kg, which Zhang et al. reported cut fecal Salmonella from about 10^5.7 to 10^3.2 CFU/g at day 20. It is the best reported non-antibiotic mineral formulation in the file, tested on the same pathogen model. We will re-synthesize it with their protocol and run it concurrently in every round, rather than quoting their numbers. The benchmark is our own concurrent measurement; their figure serves only as a check that our infection model reproduces theirs.

A facet-engineered formulation beats this baseline if it meets two conditions. First, it must not be meaningfully worse at reducing day-20 shedding, with a pre-specified margin of 0.5 log10 CFU/g. Second, it must reduce excreted metal by at least 40%.

Three ablations isolate facet control as the key ingredient.

  • Surface-area and loading ablation. An additive-free material is size-matched and surface-area-matched to each facet-engineered one, by adjusting synthesis temperature. It separates facet effects from simple particle-size effects.
  • Ion-release ablation. A zinc/copper sulfate solution is dosed to reproduce the facet-engineered material's measured free-ion profile over time. It is tested in vitro, and in the animal trial as the release-matched arm (see below).
  • Additive-only ablation. Methionine or urea is added to the reference material after synthesis. It rules out a direct antimicrobial or nutritional effect of the additive itself.

Why not just…?

Why not just lower the dose of existing Cu/Zn-Pal? This is the cheapest route to less metal in manure, and Zhang et al. tested only one dose, so the full dose-response curve is unknown. We add a half-dose Cu/Zn-Pal-1 arm to rounds three and four. The facet claim only has value if the facet-engineered material at a given metal dose outperforms Cu/Zn-Pal-1 at that same dose. If the half-dose reference already achieves the target, dose reduction alone solves the manure problem, and the facet contribution would be reported as negligible.

Why not just release zinc and copper ions more slowly or faster, for example with a coated salt? This is the most serious threat. Metal-oxide antibacterial action is often attributed largely to dissolved ions, and nothing in the file shows that surface effects beyond ion release exist in gut fluid. The ion-release ablation tests it directly. The in-vitro test is whether potency per unit of released ion differs between facet-engineered and release-matched salt. The animal test is to promote the release-matched salt to a full arm in round four, run head-to-head. If the two show equal potency, we will report that facets act only through release speed. The contribution then becomes a way to program release speed, with no claim that the crystal surface acts on its own.

Why not just use concurrent randomized controls and ignore temporal-confounding methods? For the primary analysis, we do exactly that. Within-round randomization is the design, and no statistical correction replaces it. The temporal part adds two things that concurrent controls alone do not provide:

  • It quantifies how badly a historical-control comparison would mislead. This matters because, according to the file, historical comparisons remain the norm in agricultural evaluation.
  • It tests whether selective borrowing of past-round controls, following the doubly protected estimator, can safely reduce the number of animals per round.

If borrowing adds bias measurable against the concurrent estimate, we will say so and recommend against it.

Decisive Checks

The first check is materials-level. The additive series must produce a monotonic shift in facet fraction at constant loading and matched surface area, and that shift must survive 24 hours in simulated digesta. If facets reconstruct in gut fluid, the animal-trial arms using them are dropped, because the premise fails at the interface.

The second check is the in-vitro ion-normalized potency test described above, which determines whether the claim concerns crystal surfaces or release speed.

The third check is the animal non-inferiority-plus-excretion test against concurrent Cu/Zn-Pal-1.

The fourth check is temporal. For each round from two to four, we compute the treatment effect we would have estimated using round-one controls. We report the gap from the concurrent estimate as the historical-control bias. The negative-control clay arm serves as a calibration: its apparent effect across rounds shows how large pure drift is. No reported result was found on Lacuna for the size of historical-control bias in livestock challenge trials, so any estimate here would be new.

The selective-borrowing analysis is judged by whether it tracks the concurrent estimate more closely than naive pooling does. Its purpose is to reduce the number of animals needed in future rounds, and it succeeds only if it does so without introducing bias.