iiinova — where your digital self keeps evolving

Gemma Martínez Redondo

@gemmaeling.bsky.social

219 followers 267 following 21 posts

Postdoc at Trinity College Dublin on whole-genome duplications 🧬🧬 Previously: PhD on the genomic🧬 basis of animal terrestrialization 🦀🐙🐟🌊⏭🕷🐌🦇 at IBE, CSIC-UPF She/her

Posts Replies Media

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

Many questions about animal terrestrialization remain unanswered: how many times did it happen? How did each lineage reshape its molecular networks to survive on land? Our results provide a first glimpse into many understudied animals and, hopefully, a starting point for future research. 13/14

comic from @beetlemoses.bsky.social showing tetrapod leaving water to avoid paying a debt

110

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

So, did animals use the same genomic solution when moving to land? Not really. Instead, each terrestrial lineage independently recruited different sets of genes that converged functionally! 11/14

Illustration of the 15 terrestrialization events in the study. Phylogenetic tree showing their relationships and background as blue (sea/freshwater) and green (land)

100

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

Interesting examples include CYP450s, linked to plant terrestrialization, and FMOs, involved in fish odour! Or Selenoprotein L, lost in all terrestrial species. 10/14

d, Functions from the top ten OGs as ranked by the machine learning model. OG name was obtained with BLASTp. GO terms on the right were predicted using FANTASIA.

100

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

We identified 130 OGs that could distinguish aquatic from terrestrial species, of which 81 were also identified in our stress experiments. Most originated before terrestrialization, but their copy-number patterns were mainly lineage-specific. 9/14

a, Simplified workflow of machine learning analysis (full workflow in Supplementary Fig. S272d). The pie chart shows the number of OGs from the top 130 that contain hubs from stress-related modules in GCN. b, Balanced accuracy for each model trained across all iterations of the dRFE workflow for training and testing datasets of all folds. The dashed dark green line shows the iteration where we select the top OGs. c, Logloss values for each model trained across all iterations of the dRFE workflow for training and testing datasets of all folds. The dashed dark green line shows the iteration where we select the top OGs.

100

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

But stress genes are only part of the picture. What about genes involved in development, anatomy or biochemistry? To look into this, we trained a machine learning model to distinguish between aquatic and terrestrial species based on the number of genes in each OG. 8/14

Simplified workflow of machine learning analysis. Input OGs are fed into XGBoost classifier with Kfold stratified cross validation and dRFE. Output is separation between aquatic and terrestrial, and top OGs

100

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

However, even if not shared, genes had similar functions and were associated with the same biological pathways, suggesting convergence at the functional level rather than at the gene level. 7/14

a, Graphical illustration of constellation plots of gene functions across species. Each dot represents a species and each polygon a distinct functional cluster. On the left side, each species belongs to a different cluster, showing the lack of functional convergence, and on the right side, all species cluster together, representing functional convergence. b, Constellation plots of functions of hub genes across 17 aquatic and terrestrial species for each experimental condition. Each polygon represents a distinct functional cluster formed by species with similar GO annotations, despite no orthogroup-level overlap across species. Cluster exemplars indicate the species that act as centroid in each cluster. The presence of hub genes from multiple species within the same functional cluster demonstrates convergence in stress-response roles across independent lineages. Only six experiments are shown for demonstration purposes, given that all experiments showed the same pattern (see also Supplementary Fig. S141). Functional convergence measured as GO-annotation similarity was inferred by calculating Wang semantic similarity between species’ GO term sets using the Best-Match Average method, followed by clustering with Affinity Propagation.

100

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

Surprisingly, only 42 orthogroups (OGs) out of the 50k were shared across all the species in our stress experiments. These OGs originated before terrestrialization, pointing to independent co-option of pre-existent genes instead of a “universal terrestrialization toolkit”. 6/14

d UpSet plot showing overlap of module-associated orthogroups resulted from species-specific co-expression networks. Dots connect species that have a number of module-associated orthogroups in common, the number indicated in each bar from the barplot on top.

110

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

But genomes are not everything. Led by the brilliant @klarael.bsky.social (co-first author), we exposed aquatic and terrestrial animals from 7 phyla to environmental stressors associated with land (O2, desiccation…), and used gene co-expression networks to identify shared stress responses. 5/14

a Overview of abiotic stress experiments, gene co-expression network construction and proteomic detection of stress-related genes across 17 species from seven animal phyla, exposed to standardised stressors mimicking terrestrial abiotic factors.

120

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

To answer that question, we first surveyed the genomes of almost 1000 animals, with 15 terrestrialization events represented, and found that they were characterised by higher gene loss than gain. 4/14

a Mirror phylogeny of 960 animal species plus outgroups. Circles at each internal branch represent the square-root-scaled number of orthogroups gained (left, blue) and lost (right, brown). Terrestrial lineages are highlighted in yellow; the right phylogeny labels individual terrestrial lineages, while the left phylogeny indicates the phyla containing those lineages. Exact numbers of OGs gained and lost can be found in Supplementary Fig. S1. b Phylostratigraphic origin of all genes in all nodes in a representative species (the earthworm Norana najaformis, Annelida), highlighting the gene gain burst in Protostomia (in yellow). c Comparison of gene gain and loss across the 15 terrestrial clades with independent terrestrialisation events.

100

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

Many naturalists and scientists have studied this transition from morphological and physiological perspectives, but its molecular basis has been much less explored. Did different animals use the same molecular solutions to adapt to life on land, or did each lineage find its own way? 3/14

First tetrapod leaving water saying "C'mon out, folks! It looks like we're the first ones here." On the right, two myriapods say "First ones here? Please! We've been here for 60 million years!" and "I...I...I wonder what they eat?"

110

Gemma Martínez Redondo @gemmaeling.bsky.social · 03/09/2026

Animals transitioned from water to land in at least 9 animal phyla. Tetrapods and arthropods are the best-known examples, but other, much less studied animals, like flatworms or nemerteans changed their habitat too! 2/14

a Animal tree of life highlighting the phyla where there has been at least one terrestrialisation event. b Main animal terrestrialisation events within animal phyla discussed in this work (in bold), with the species used in the stress experiments below. c Simplified view of paths towards terrestrial life in animals, and the differences in some abiotic factors between habitats.

100