Teaching seagrass to withstand warmer seas — using epigenetics to switch on the heat tolerance it already carries in its own genome. No foreign genes, no GMO.
Seagrass is one of nature's most valuable assets — for carbon, for coastal ecosystems, and for the fisheries and coral reefs that depend on it — and it is dying. We are working to save it.
Seagrass meadows bury carbon faster than tropical forests, anchor coastal food webs, and shelter an outsized share of marine life from a fraction of the seafloor. They are also a foundation of the coastal ecosystems that sustain coral reefs. And they are in serious peril: warming seas are driving mass die-offs faster than meadows can recover, and many of the plants that remain lack the heat tolerance to withstand what is coming.
The traits that let seagrass survive heat already exist within the species — the same plant carries both warm-adapted and cold-adapted populations. We read the heat tolerance out of the plants that have it and switch on those same traits in the plants that need it, working entirely within the plant's own genome. It comes down to teaching the non-adapted plants what the adapted plants already know.
Reversing seagrass decline will take a concerted, intentional effort. This project is built to be part of that effort.
Seagrass meadows are among the most productive and ecologically critical ecosystems on Earth — yet they are disappearing faster than tropical rainforests.
Declines of many seagrass species have been attributed to thermal stress associated with rising sea surface temperatures and extreme temperature events. Many seagrass species persist in low energy and shallow water environments such as coastal embayments, inlets and fjords. These environments are particularly susceptible to warming and extreme temperature fluctuations."Seagrass Restoration Is Possible: Insights and Lessons From Australia and New Zealand"
Robust, climate-resilient seagrass connects to several established and emerging markets. These define the applied value of the biological work.
This is a funded federal priority, not a speculative one. A 2024 USDA/NOAA report to Congress places heat-tolerant seagrass inside an active government agenda, and the numbers behind it are concrete:
Federal figures from the 2024 USDA/NOAA report to Congress (Farming Seagrasses and Seaweeds); full technical treatment, including the ReefGen planting robotics, in Findings §9.
No seagrass has ever been genetically engineered. We're pioneering epigenetic triggering — a faster, non-GMO approach that works with the plant's own gene-regulation machinery.
Genome + epigenome from multiple populations — typical habitats vs. heat-stressed environments.
Open-access genomes and epigenomes for public science. Building the best publicly accessible seagrass genomic repository.
Switch on heat-tolerance traits with a targeted RNA signal — RNA-directed DNA methylation (RdDM) tunes the plant's own genes, with no foreign DNA required.
Treated seeds planted at scale by seafloor planting robots (ReefGen) that inject seed directly into sediment in degraded meadows.
CRISPR would rewrite the plant's DNA to force heat tolerance — and for seagrass that path is both slow and unusable in practice. Editing requires tissue culture and whole-plant regeneration, which have never been established for seagrass, so the pipeline stalls before it starts. What comes out the other end is a GMO: a high regulatory bar, a decade-plus timeline, and the public resistance that has kept engineered plants out of conservation. Epigenetic triggering skips all of it — no cut to the genome, delivered as a seed treatment, non-GMO, and deployable in years rather than a decade.
| Gene editing (CRISPR) | Epigenetic triggering (RdDM) | |
|---|---|---|
| Changes the DNA sequence? | Yes | No |
| Tissue culture + regeneration | Required — not established in seagrass | Not needed |
| Regulatory status | GMO (high burden) | Non-GMO (low burden) |
| Timeline | 10+ years | 2–5 years |
The obvious alternative is to move a warm-adapted seagrass from a hotter region into the meadow you want to save. That trades one climate problem for an ecological one: introducing non-native genotypes risks outbreeding, disease, and displacement of the local population — the same harm conservation programs work to prevent, and a hard sell to the communities and regulators who protect these sites. Our approach keeps the local genotype and switches on the resilience that population already carries, so the meadow stays genetically itself.
Heat tolerance is not something we invent; warm-adapted seagrass ecotypes already carry distinct DNA methylation patterns that persist across generations. The task is to identify those marks and switch them on in local plants. Our 2026 analyses confirm the mechanism is available: the complete RdDM pathway is intact in Zostera marina and conserved across all four seagrass families (see Latest Findings).
Results from computational analysis of 20 Zostera marina Nanopore methylomes (Decibel Bio sequencing). Full detail, figures, and downloadable data on the companion Findings page.
All 20 core RNA-directed DNA methylation genes are present in the Zostera marina genome (13 strong, 6 moderate, 1 weak ortholog; none absent), and the full pathway is conserved across six seagrass genomes spanning all four seagrass families (Zostera, Thalassia, Halophila, Cymodocea, Amphibolis, Posidonia). The natural, non-GMO machinery this strategy depends on is an ancestral feature shared across the seagrasses, not something specific to Zostera.
Feature-level analysis shows promoters carry the highest CHH (non-CG) methylation of any gene region (3.3%) — higher than intergenic or repeat DNA. Because CHH methylation is maintained by ongoing RdDM, promoters are exactly where a targeted RNA trigger would act to tune a gene without altering its sequence.
Across the 20 individuals, DNA methylation follows the canonical plant hierarchy (CG 70.5%, CHG 27.0%, CHH 1.44%) and is remarkably uniform. All 20 are genetically distinct (no clones), and methylation similarity tracks genetic similarity (Mantel r = 0.77) — so genotype-linked marks must be separated from facultative ones when choosing targets.
From field collection to ocean deployment — an integrated pipeline connecting sequencing, analysis, and robotic planting.
Additional methylation sequencing across populations → epiGWAS association study → identify novel epigenetic factors → validate → target. More rigorous, builds the field.
Use published candidate genes (HSP70 family, known methylation marks) → go directly to epigenetic targeting on priority species. Faster, leverages existing literature.
Five deployment-target seagrass species spanning freshwater to marine, tropical to temperate — chosen for ecological importance, genomic tractability, and thermal vulnerability — plus one comparative reference species that anchors the heat-tolerant end of the spectrum.
Our species fall into two roles. Deployment targets are the species we aim to make more heat-resilient and eventually replant. Comparative references are species we sequence and analyze to learn from rather than to treat — a naturally heat-adapted seagrass shows us what thermal tolerance looks like at the sequence and methylation level, giving us the "answer key" for the marks we want to switch on elsewhere.
| Species | Common Name | Habitat | Genome | Thermal Optimum | Lethal Temp | Reference Genome | Carbon Seq. |
|---|---|---|---|---|---|---|---|
| Thalassia testudinum | Turtle grass | Marine tropical | 4,866 Mb | 28–30°C | 36–39°C | Chromosome-level | 122 g C/m²/yr |
| Halodule wrightii | Shoal grass | Marine tropical | Unknown | 23–32°C | >42°C | In preparation | ~5–10 g C/m²/yr |
| Zostera marina | Eelgrass | Marine temperate | 260 Mb | 15–23°C | 30–35°C | Chromosome-level | 25 g C/m²/yr |
| Posidonia oceanica | Neptune grass | Marine temperate | 3,192 Mb | 15–25°C | ~35°C | Chromosome-level | 84 g C/m²/yr |
| Vallisneria americana | Wild celery | Freshwater | ~3,600 Mb | 28–36°C | >38°C | JGI draft in progress | Not quantified |
| Species | Common Name | Habitat | Genome | Thermal Optimum | Lethal Temp | Reference Genome | Role |
|---|---|---|---|---|---|---|---|
| Cymodocea nodosa | Little Neptune grass | Marine temperate–subtropical | 375 Mb | ~30°C (up to 34–35°C) | >34°C | Chromosome-level | Heat-tolerant benchmark; 4th RdDM lineage |
From most to least thermally tolerant — every species shows ecotype-dependent variation in thermal tolerance.
Samples from three biogeographic regions enable comparative sequencing between typical and heat-stressed populations.
Primary site — Private island property. Multiple collection trips completed with Pati Mar and Sociedad Ambiente Marino volunteers.
Heat-tolerant populations — Private island property. Shallow mangrove location with naturally heat-stressed seagrass for comparative sequencing.
Warm-edge eelgrass — Southern range limit populations. Successfully collected and sequenced (20 samples analyzed, 2026). Small genome makes it the most cost-effective for initial work.
Chromosome-level reference genomes now exist for three deployment targets plus our comparative-reference species, Cymodocea nodosa. We are building the most comprehensive public seagrass genomic repository — four species held (Zostera, Thalassia, Cymodocea, Posidonia), two targets still awaiting assembly.
| Species | Size | Chromosomes (n) | Assembly | Key Accession / Source | Reference |
|---|---|---|---|---|---|
| Zostera marina (v3.1) | 260 Mb | 6 | Chromosome-level | Phytozome (PRJNA701932) | Ma et al. 2021 |
| Zostera marina (v2.1) | 204 Mb | scaffold | Scaffold-level | GCA_001185155.1 | Olsen et al. 2016 |
| Thalassia testudinum | 4,866 Mb | 18 | Chromosome-level | Marine Angiosperm Genome Initiative | Ma et al. 2024 |
| Posidonia oceanica | 3,192 Mb | ~20 | Scaffold/Chr-level | GCA_037176725.1 (PRJNA1041560) | JGI 2024 |
| Cymodocea nodosa | 375 Mb | 28 | Chromosome-level | GCA_036874045.1 (PRJNA1041560) | JGI 2024 |
| Halodule wrightii | Unknown | 24–39 | SRA data only | SRA runs (prep) | No assembly yet |
| Vallisneria americana | ~3,600 Mb | 20 | JGI draft | PRJNA566556 | JGI (raw reads) |
Genome + epigenome data available for Thalassia testudinum and Zostera marina from our sequencing runs with Decibel Bio. Twenty Z. marina Nanopore methylomes were fully analyzed in 2026 (see Findings page). Data location: Google Drive — Good Ancestor Shared Drive > Genome, with a dedicated Analysis folder for computed results.
All seagrasses share an ancient whole-genome triplication (~87 million years ago) and remarkable convergent adaptations to marine life:
Comprehensive thermal tolerance profiles across the deployment targets plus our heat-tolerant reference (C. nodosa). Heat tolerance is ecotype-dependent in every species studied — meaning the alleles and epigenetic marks already exist within species.
| Parameter | V. americana | T. testudinum | H. wrightii | Z. marina | P. oceanica | C. nodosa (ref) |
|---|---|---|---|---|---|---|
| Thermal optimum | 28–36°C | 28–30°C | 23–32°C | 15–23°C | 15–25°C | ~30°C |
| Photosynthesis inhibition | — | >33°C | Unaffected at 34–35°C | >25°C | >27°C | Tolerates 34–35°C |
| Chronic stress onset | — | 32–33°C | — | 25°C | 27°C | >32°C |
| Upper lethal (sustained) | ≈38°C | 36°C (4 wk) | >42°C (short-term) | 28–30°C | ≈35°C | >34°C |
| Thermal safety margin | Large | <3°C (Gulf) | Moderate | Very small | Small | Large (warming winner) |
| Ecotypic variation | Yes | Limited data | Yes | Yes | Yes | Yes |
| Epigenetic data | None | Moderate (our data) | None | Extensive (our data) | Moderate | RdDM confirmed (our data) |
Two technology partners enable our epigenetics-first pipeline: Decibel Bio reads and writes the epigenome, and Chris Oakes brings marine seed-delivery and pelletization for deployment at sea.
Berkeley-based crop epigenetics platform. Reads, understands, and writes the plant epigenome using small RNA/DNA molecules that direct DNA methylation via the natural RdDM pathway. Our 20 Zostera marina methylomes were sequenced by Decibel (Oxford Nanopore, 5mC).
Chris Oakes (former CEO/founder of ReefGen) brings robotic-planting and seed-delivery expertise to the marine deployment problem, including a pelletization technology for controlled seafloor release of treatments and seed.
Core team, science and industry advisors, and field collaborators.
Marion Prieler brings a coral assisted-evolution advisory network to the project; these advisors flow through her collaboration rather than advising the project directly.
The analysis tools used to produce the 2026 findings, and the autonomous-research tools we are evaluating for the next phase.
Protein-to-genome alignment used to map the 20 Arabidopsis RdDM machinery proteins onto six seagrass genomes, calling ortholog presence and confidence per gene. Basis of the conservation figure.
De novo transposable-element discovery and genome masking. Produced the 68.9% repeat annotation and the LTR/DNA/LINE class fractions for Z. marina v3.1 (the RdDM substrate).
Per-sample VCFs merged and reduced to biallelic SNPs; IBS distances and clustering established that all 20 samples are distinct genotypes (no clones).
pandas / numpy / scipy / matplotlib on the 20 Nanopore methylomes: weighted-methylation landscape, feature-level (promoter/exon/intron/TE) profiling, and the methylation↔genotype Mantel test (r = 0.77).
65,932 georeferenced occurrence records across the six species, joined to sea-surface-temperature envelopes and marine ecoregions to build the collection-site and thermal-niche maps.
Data-driven discovery agent. Accepts research objectives plus complex datasets (up to 5 GB): literature search, high-dimensional analysis, hypothesis generation. ~$200/run; we have credits. Advisor: Sam Rodriques.
Local-first autonomous research studio for long-horizon scientific work: git-native tracking, Bayesian optimization, research-map visualization. Open source.
Where we are as of mid-2026.
Foundational literature and our own 2026 analyses. See the Findings page for the full RdDM literature review.