Facility microbiota can shape zebrafish results

Two zebrafish from the same line, held at the same temperature and fed under comparable protocols, do not necessarily inhabit the same biological environment. Water, biofilters, feed, animals and local husbandry practices combine into a facility-specific microbial landscape. A multicentre study published in 2026 found that this signature was detectable in both tank water and fish intestines. The practical lesson is not to pursue an impossible state of sterility. It is to recognise a variable that often remains invisible, report it more consistently and account for it when results need to travel between laboratories.

Five facilities, 50 water samples and 134 fish

Kayla Evens, Ingrid Bakke and Brendan Bohannan compared five research aquaculture facilities: two in Eugene, Oregon, and three in Trondheim, Norway. Four relied mainly on recirculating aquaculture systems, while one used flow-through water. The team paired tank-water sampling with intestinal sampling and characterised bacterial communities by 16S ribosomal RNA gene sequencing.

The final analysis included 50 water samples and 134 fish. Microbial community composition differed significantly by geographic region and facility. In water, facility identity explained substantial additional variation after location was considered. Gut communities also differed among facilities, including shifts in the relative abundance of Cetobacterium, Vibrio and Aeromonas.

Those names are not a diagnostic list. A bacterial genus detected by sequencing can contain species and strains with very different properties, and relative abundance alone demonstrates neither infection nor virulence. The study describes communities and their differences. It does not establish a universal “healthy” microbiome for a zebrafish facility.

Tank water contributes, but does not explain everything

Microbial source tracking estimated how much of the gut community could be associated with the sampled water. A contribution was measurable in several facilities, but its strength varied markedly. Sources not represented by the water samples dominated the estimates, accounting for 71–99% across analyses. These unknown sources may include diet, host physiology, line history, system surfaces and other compartments that were not sampled.

It would therefore be inaccurate to claim that tank water alone builds the intestinal microbiome. This was an observational study, and fish genotypes were not replicated evenly across every facility. Genotype and facility effects remained partly confounded. One facility was also represented by only five fish from a single tank, limiting statistical power for comparisons involving that site.

The well-supported conclusion is narrower: water and fish from different facilities are not microbiologically interchangeable. That variation may matter whenever an endpoint depends on immunity, metabolism, intestinal barrier function, infection, antimicrobial exposure or host–microbe interactions.

A stable system is not a static system

A 2021 longitudinal study provides useful temporal context. Aaron Ericsson and colleagues followed a newly installed recirculating system for 18 weeks, sampling six tanks, the biofilter, several filtration points and post-UV water. Richness and broad microbiome characteristics became relatively stable around three to four weeks after fish were introduced. Community composition nevertheless continued to change, while the fluidised-bed biofilter and post-UV water retained microbial profiles distinct from the other compartments.

This matters after commissioning, major maintenance, a source-water change or transfer of a colony. Acceptable physicochemical readings do not prove that the microbial ecosystem has returned to its previous state. Conversely, a community shift does not by itself show that a system is malfunctioning. Biofilms and biofilters perform essential work; the operational objective is to track changes that affect interpretation, not to remove microbial life.

Water chemistry adds another layer of variability. A survey of 40 facilities in 12 countries documented different recipes and operating ranges for temperature, pH and conductivity. Experimental work in the same paper showed that very low ionic conditions could alter growth and developmental outcomes in embryos and larvae. Microbiology and physicochemistry therefore need to be recorded together. Neither can stand in for the other.

What should facilities record without turning into sequencing centres?

Routine studies do not all require systematic microbiome sequencing. The first step is to capture variables that can explain a break in comparability: source-water treatment, system architecture, volume and turnover, biofiltration, UV, temperature, pH, conductivity, stocking density, feed, line, age, health status, medication, disinfection and technical events. Feed batch and change dates may be as informative as the average value displayed by a sensor.

For microbiome-centred work or microbiome-sensitive phenotypes, the sampling plan should be fixed before the experiment starts. It needs to define the compartments — water, gut, feed, biofilter or surfaces —, negative controls, extraction method, 16S region or metagenomic approach, sequencing depth and bioinformatic rules. Comparing relative percentages without controlling low-biomass contamination or analytical batch effects can manufacture apparent differences.

Common-garden acclimation after a transfer may reduce some contrasts, but it cannot be assumed to erase them. Depending on the question, a stronger design may block or randomise animals by origin, replicate lines across tanks, include facility as a factor in the statistical model and archive reference samples. Multicentre studies should build this heterogeneity into the protocol instead of discovering it after outcomes diverge.

Standardise reporting before trying to standardise microbes

The 2026 study does not justify a target diversity value, prophylactic treatment or more aggressive disinfection. Abruptly changing a community may disrupt biofiltration or select different organisms without improving reproducibility. Any intervention should distinguish an ecological signal from a technical drift or a demonstrated health problem.

The immediately useful intervention is transparency: describe the facility and husbandry context well enough for another team to understand what was actually compared. Reproducibility does not require every facility to become microbiologically identical. It requires important differences to be visible, incorporated into experimental design and discussed as plausible limitations.

Vetofish can support research facilities with system mapping, proportionate monitoring plans, selection of sampling points and joint interpretation of health, physicochemical and microbiological data. The aim is to turn a hidden variable into usable evidence without confusing a microbial profile with infection risk or proof of causality.