Microbial Ecology and Resistome of the Comau Fjord, Patagonia (42ºS)

ecastron@utalca.cl (Eduardo Castro) — Thu, 27 Apr 2023 00:00:00 +0000

The Comau Fjord is an estuarine system in Northern Patagonia (42ºS), Chile: roughly 45 km long and nearly 490 m deep, with steep temperate-rainforest walls dropping straight into the water. High rainfall and snowmelt create a shallow brackish surface layer over a cold, saline body of modified Subantarctic water, producing steep vertical gradients in temperature, salinity, nutrients, and pH. The fjord region is difficult to reach and has almost no laboratory infrastructure — the Huinay Scientific Research Station is the exception that makes this work possible.

We use Comau as a model of anthropogenic impact. The land surrounding the fjord is protected (national parks and private conservation) and there are no major human settlements, but the fjord itself hosts open-cage salmon and trout aquaculture. Feed is medicated with florfenicol and oxytetracycline — the two antibiotics that define the Chilean industry — and the excess reaches fjord waters. The system therefore offers a gradient from relatively undisturbed to heavily human-impacted within a single, well-characterized water body.

Our work here asks three questions: what structures microbial communities in the fjord, what antibiotics do to them, and whether a wild sentinel species registers the same signal.

1. What structures the fjord’s microbial communities?

Season sets the community, not geography — and the fifteen metres between our two sampling depths separate two opposite rules of assembly.

Between 2016 and 2019 we sampled 12 sites along a 35-km transect at 5 m (surface, above the pycnocline) and 20 m (below it), through the calendar year. From that effort we have released:

93 metagenomes and 677 MAGs of medium to high quality, spatially and temporally explicit (Castro-Nallar et al. 2023, Microbiology Resource Announcements). Dereplication yields the 513 representative genomes used in the analyses below.
48 viral metagenomes and 5,656 viral OTUs across the same transect, depths, and seasons (Castro-Nallar et al. 2023, Microbiology Resource Announcements).

MAGs in the Comau Fjord. Most MAGs belonged to Proteobacteria, Cyanobacteria, Actinobacteriota, and Bacteroidota. Archaea were represented by Thermoplasmatota with members belonging to the Class Poseidonia A and B

The fjord has its own microbial signature. Community composition is distinct from global ocean surveys such as Tara Oceans, yet converges on the same cosmopolitan marine lineages — Proteobacteria, Bacteroidota, and Actinobacteriota. Viral communities are even more distinctive, with up to 50% of their protein content unique to this fjord (Guajardo-Leiva et al. 2023, Microorganisms).

The fjord runs on a seasonal clock. Communities split sharply by season (PERMANOVA R² = 0.27), and every season differs from every other. Proteobacteria and Bacteroidota persist year-round, but Cyanobacteria bloom from April to August at 5 m while Actinobacteriota hold the 20 m layer. The environment itself cycles: temperature and pH fall together at the autumn–winter transition and then recover, nutrients accumulate through autumn and winter and are drawn down by summer, and the 20 m layer stays consistently colder, saltier, and more nutrient-rich than the surface.

Community structure varies by season and depth across the fjord. (A) Twelve sampling sites along the fjord, with the Huinay Scientific Research Station marked. (B) Communities split by season in ordination space (PERMANOVA R² = 0.273, P = 0.001); Summer vs Fall is the sharpest contrast. (C) Phylum composition through the calendar year at 5 and 20 m — note the Cyanobacteria pulse at the surface from April to August. (D) Abundance-weighted mean pairwise distance shifts by season and depth.

Chemistry, not distance, tunes the community. Phylogenetic diversity rises with nitrate, nitrite, phosphate, and silicate, and falls with pH, temperature, and chlorophyll a. With depth and season statistically conditioned out, environmental gradients still order the communities, and pH is the strongest single association — six MAGs track it with clean dose–response curves. Distance–decay analysis is unambiguous on the geography question: only seasonal time distance drives community turnover, while geographic and environmental distance are not significant.

Metabolism cycles like a calendar. Whole pathways rise and fall together by season and depth — methane, carbon, sulfur, vitamin B₁₂, nitrogen, and phototrophy — with sentinel genes tracking their pathways (Mcr, DsrA/Sqr, BluB/CobU, PsbA). Autumn favors sulfur cycling and B₁₂ biosynthesis, summer the methane cycle, winter nitrogen metabolism and photosystems. This capability is inherited rather than random: collapsing the MAG tree to families shows functional potential lighting up by lineage, with Rhodobacteraceae, Thioglobaceae, and Flavobacteriaceae acting as multifunctional hubs while Planctomycetaceae and Pirellulaceae sit consistently below average.

Two assembly rules in one fjord. At 5 m, communities are phylogenetically and functionally over-dispersed — closely related taxa avoid co-occurring, the signature of competition and niche partitioning. At 20 m the pattern reverses to convergence, indicating habitat filtering for organisms adapted to colder, saltier, darker water. Traits agree: phototrophy scatters at the surface and clumps at depth. Competition sorts the lit surface; filtering sorts the dark deep.

2. What do aquaculture antibiotics do to these communities?

A single dose at the concentrations the industry uses cut diversity within two days — and eleven days later the community had not come back.

The design. To isolate the antibiotic signal from everything else that varies in the fjord, we ran a mesocosm experiment at Huinay — closed ~3,000 L seawater bags moored in the fjord itself, which keeps the biological realism of in situ water while controlling the dose. Four bags received the four treatments: fish feed alone (C1), water alone (C2), florfenicol plus feed (C3), and oxytetracycline plus feed (C4), at the two drugs and quantities used by the local industry. The disturbance is a single pulse at day 0, followed for 11 days. Each time point was profiled with three marker loci (16S, 18S, ITS) across two size fractions (3 µm and 0.22 µm), plus shotgun DNA and RNA. One bag per treatment, so the experiment is descriptive rather than replicated — a constraint we state plainly in every result below.

Mesocosm design. Four closed seawater bags receive fish feed, water only, florfenicol plus feed, or oxytetracycline plus feed. A single pulse disturbance at day 0 is followed over 11 days, with three marker loci across two size fractions at each time point.

Exposure was real and persistent. LC-MS/MS confirmed both drugs in the dosed tanks and never in the controls. Florfenicol showed no measurable decay over 10 days; oxytetracycline peaked and then declined.

Diversity collapsed and stayed down. Shannon diversity dropped in every bag by day 2 — a pulse effect common to all treatments — but the two antibiotic arms fell furthest and stayed suppressed while the water control recovered. Oxytetracycline was lowest and still declining at day 11. Community composition moved away from the day-0 baseline fastest under antibiotics, with both arms saturating near-completely and tracking each other.

Shannon diversity through the mesocosm experiment. All four bags drop by day 2, but the controls rebound while the antibiotic treatments stay suppressed — oxytetracycline lowest and still falling at day 11.

Turnover was lineage-specific, not a uniform decline. Florfenicol produced a Moraxellaceae takeover; oxytetracycline a late Enterobacteriaceae and Shewanellaceae bloom. Rhodobacteraceae and Sphingomonadaceae were the biggest losers.

Selection acted within days. Resistance gene abundance surged by day 2 — tetracycline efflux under oxytetracycline, floR plus tetracycline efflux and tet(X) under florfenicol. Notably, the resistome then declined by day 10 even as drug continued to accrue, so gene abundance is decoupled from cumulative exposure: selection saturates and the community restructures around it. We also recovered a plasmid co-locating floR, estT, and tet(X) with mobility signatures (IS91, integrase, VirD2) — phenicol and tetracycline resistance travelling on a single element, a route to co-selection of multidrug resistance.

Resistance gene abundance (copies per genome, normalized by SingleM genome equivalents) across the four mesocosm treatments. Controls bloom transiently and crash; the antibiotic arms sustain elevated resistomes, with the highest levels under oxytetracycline.

The genes are transcribed, not latent. Metatranscriptomes show floR strongly induced in the florfenicol treatment by day 2, roughly three orders of magnitude over controls, with the plasmid firing as a unit. Oxytetracycline expression awaits deeper resequencing.

3. Can a wild species act as a sentinel?

A top predator that eats no salmon at all still carries the salmon industry’s two resistance classes in its gut, year-round.

The tanks show what antibiotics do under controlled dosing. A top predator ranging the whole fjord integrates the same signal in nature — a One Health sentinel. We profiled the gut resistome of the South American sea lion (Otaria flavescens) from 30 temporally explicit scats collected at the head of the fjord over a full year, June 2017 to June 2018 (Guajardo-Leiva et al. 2023, Microbiology Resource Announcements).

The wild gut resistome mirrors the aquaculture pharmacopoeia. Tetracycline and phenicol resistance dominate — the two classes that define Chilean salmon farming — while every other class is minor. Tetracycline determinants are near-universal across scats; phenicol genes are prevalent but patchier, and their abundance peaks at specific months, consistent with pulsed external input rather than a stable resident pool.

Gut resistome of the South American sea lion by month, resolved by antibiotic class and resistance mechanism. Tetracycline and phenicol resistance — the two classes used in Chilean salmon farming — dominate and are present year-round.

The source is not human sewage, and not the diet. Clinical and wastewater-associated classes are absent. And two independent methods recovered zero salmonid reads across all 30 scats: the diet is a coherent Patagonian assemblage of hake, hoki, anchovy, jack mackerel and relatives. The sea lions are not acquiring these genes by eating farmed salmon.

Dietary composition recovered from the same scat metagenomes. The prey assemblage is a coherent southeastern Pacific / sub-Antarctic one — and contains no Salmonidae in any of the 30 scats.

Reading the host from its own scat. The same shotgun reads that yield the resistome also carry the animal. Using a target-restricted assembly strategy, we recovered complete mitochondrial genomes of the sea lion from all 30 eDNA scat samples at greater than 40× coverage — 13 protein-coding genes, 22 tRNAs, two rRNAs and the control region — and showed that they support genuine population genetics: AMOVA and pairwise Φ_ST recover statistically significant differentiation between Pacific and Atlantic populations (Baeza et al. 2026, Environmental DNA). Non-invasive scat sampling can therefore do population genomics on a declining marine mammal without ever handling one.

The genes are mobile, and not inherited. Aquaculture-associated resistance genes are strongly enriched on plasmids relative to other resistance genes. Those same mitogenomes let us reconstruct the host maternal lineages behind the 30 scats — 16 distinct genotypes — and place resistance carriage on that tree. It is scattered: high- and low-carriage animals sit side by side within a single maternal lineage, so the resistome is acquired, not vertically inherited.

Host maternal-lineage tree reconstructed de novo from mitogenomes assembled out of the scat metagenomes, with per-cell abundance of the aquaculture resistance genes alongside. Carriage does not follow host lineage.

What it does follow is the gut microbiota. Resistome dissimilarity is congruent with microbiota composition, and null against both diet and host genetics — which bacteria are present sets the resistome, and that community turns with the seasons.

Why it matters

Patagonian fjords cover roughly 240,000 km² and are among the least-studied aquatic systems on Earth, yet they sit downstream of expanding aquaculture, increasing maritime traffic, microplastic pollution, and climate-driven changes in rainfall and glacial melt. This work provides the first multi-year, spatially explicit metagenomic baseline for how microbial communities are structured in the system, experimental evidence of how that structure responds to antibiotic pressure, and a wild sentinel that registers the same chemical signature at the top of the food web.

Dig deeper

For a general audience (video and illustrated poster, in Spanish): Adventures in the Comau Fjord: Exploring the Invisible World Beneath the Surface
For analysts: Seasonal community dynamics in Comau Fjord — analysis walkthrough, with the reasoning and complete R code behind every figure
Data and scripts: github.com/ecastron/comau-reproducibility

MAGs on Castro Lab