Overview of experimental design and methods.

A) Schematic of the experimental design. Replicate groups of juvenile clownfish (Amphiprion percula) housed as size-matched pairs (n = 30 pairs) and size-matched solitary individuals (n = 30 individuals). Standard length (SL) was measured at the beginning and end of the five-week experiment, and individuals were denoted as dominant (P1), subordinate (P2), and solitary (S) at the end of the experiment based on their emergent social positions. B) Feeding behavior scoring method for pairs and solitary fish. During feeding, pairs were separated by a divider to prevent competition, divider was included in solitary tanks as a sham control, and individuals were provided with equal amounts of pre-portioned food daily. Feeding behavior was quantified weekly from video recordings by counting the number of bites taken by each fish over a 5-minute period starting from the first bite. Initially, paired fish were denoted as P-max and P-min based on the total number of bites taken, because individuals could not be reliably identified due to their similar size and lack of distinguishing markings. From week 3 onwards, bites taken were scored separately for dominant (P1) and subordinate (P2) fish, in pairs where fish could be reliably identified based on clear size differences. C) Agonistic behavior scoring method for pairs. Agonistic behaviors (aggression, submission) were quantified during a five-minute observation period from video recordings of paired interactions. Aggression was quantified by the number of chases, headbutts, and bites, while submissive behavior included flees and shakes. Initially, total aggression and submission were measured at the tank level, because fish were similar in size and lacked distinguishing markings. From week 3 onward, agonistic behaviors were scored separately for dominant (P1) and subordinate (P2) fish, in pairs where individuals could be reliably identified due to the emergence of clear size differences. In the accompanying illustrations, the focal fish is shown in orange, with its partner depicted in black and white. D) Gene expression profiling for pairs and solitary fish. At the end of the experiment (week 5), whole-body samples were collected from a subset of individuals [n = 15 per social role: dominant (P1), subordinate (P2), solitary (S)]. To maximize detection of gene expression patterns associated with social roles, replicates with the largest size differences between dominant and subordinate fish were selected for transcriptomic profiling. The resulting RNA-seq data were used to compare molecular signatures across emergent social roles. In all panels, the timeline at the top indicates when (week 1 – week 5) and how (calipers, videos, tissue samples) data was collected.

A) Growth over time. Standard length (mm) of individuals of each social position (pair rank 1 = P1; pair rank 2 = P2; solitary = S) over time with mean and standard error of mean (SEM). A.1) Overall growth (mm) over 5 weeks of each social position are shown with mean (dashed lines) and SEM. B) Feeding behavior over time. Individual food intake, measured in number of bites of size-matched pairs (P-max and P-min) vs solitary (S) individuals over time with mean and SEM. B.1) Food intake for a subset of replicates in which dominant (P1), subordinate (P2), and solitary (S) individuals could be reliably identified due to the emergence of clear size differences (week 3, n = 5; week 4, n = 17; week 5, n = 19), shown over time with mean ± SEM. C) Agonistic behavior of pairs over time. Aggression (the total number of chases, bites, and headbutts) and submission (based on the total number of flees and shakes) within pairs over time with mean and SEM. C.1) Aggression and C.2) Submission for a subset of pairs in which dominant (P1) and subordinate (P2) individuals could be reliably identified based on clear size differences (week 3, n = 5; week 4, n = 17; week 5, n = 19), shown over time with mean and SEM. Data points shown jittered on all figures for visual aid, but collected at the same time.

Overall gene expression profiles showed no significant differences between P2 and S, compared to P1 individuals.

A) Principal Component Analysis (PCA) of all genes across social positions, accounting for clutch ID (genetic background), with PERMANOVA results showing a significant effect of social position. Ellipses show the estimated confidence intervals from a 95% multivariate t-distribution. B) Venn diagram illustrating differentially expressed genes (DEGs) from pairwise comparisons of social positions (FDR-adjusted p-value of 0.05). Orange upward arrows and corresponding numbers represent significantly upregulated DEGs in the first term, relative to the second, while blue downward arrows and numbers indicate downregulated DEGs (Supplementary Table S4).

Gene ontology (GO) enrichment of the brown module associated with growth and ossification.

An exhaustive list of GO enrichment results for genes assigned to the brown WGCNA module (Fig. 5). Bubble size reflects numbers of genes associated with each term, color indicates GO category. Terms are ordered by adjusted p-value significance on the x-axis (–log₁₀ scale), with more significant terms on the right.

P1 individuals showed upregulation of genes associated with growth and ossification.

A) Average eigengene values across social positions for the brown module. B) Heatmap of significant DEGs (FDR p-value ≤ 0.05) of growth and ossification genes in the brown Biological Process (BP) and Molecular Function (MF) GO terms across the three social positions. Samples are grouped a priori by social position (dominant, P1; subordinate, P2; solitary, S) and ordered within each group by column dendrograms reflecting expression similarity within social positions. The row dendrograms show relatedness among genes based on expression profile similarity across all samples. Genes that are significantly differentially expressed between individuals of P1 vs P2 are denoted with an asterisk (*), those between P1 vs S with a double asterisk (**), and three asterisks (***) represent both comparisons. Genes shown in bold and italics indicate significance at an FDR adjusted p-value < 0.1.

P1 individuals showed downregulation of metabolic and appetite suppression genes.

Heatmap of significantly differentially expressed genes (DEGs) (FDR p-value ≤ 0.01) of genes associated with appetite regulation (APT), Krebs cycle (TCA), and glycolysis (GLY) pathways across the three social positions indicating genetic background of individuals (clutch ID). Samples (columns) are grouped a priori by social position (dominant, P1; subordinate, P2; solitary, S) and ordered within each group by the column dendrogram, which reflects expression similarity among samples within each social position. Genes (rows) are grouped a priori by pathway (APT, TCA, or GLY), and within each pathway group, the row dendrogram reflects relatedness among genes based on expression profile similarity across all samples. Genes that are significantly differentially expressed between individuals of P1 vs P2 are denoted with an asterisk (*), those between P1 vs S with a double asterisk (**), and three asterisks (***) represent both comparisons. Genes shown in bold and italics indicate significance at an FDR adjusted p-value < 0.1.

Clutch assignment validation via hierarchical clustering of microsatellite genotypes.

The dendrogram illustrates genetic relatedness among 36 individuals from three clutches, based on microsatellite genotypes (60 loci) processed using HotShot genotyping. A) Initial assignment of individuals to their presumed clutches, with pairs incorrectly assigned highlighted in blue and orange. Branch lengths reflect genetic divergence; tightly clustered branches indicate sibling relationships within clutches. B) Corrected clustering after reassignment of clutches for mismatched pairs.

Clutch assignment validation using TagSeq-derived SNPs and hierarchical clustering.

The dendrogram illustrates genetic relatedness among 45 individuals from three clutches, based on SNPs from TagSeq data. A) Initial assignment of individuals to their presumed clutches, with pairs incorrectly assigned highlighted in colors. Branch lengths reflect genetic divergence; tightly clustered branches indicate sibling relationships within clutches. B) Corrected clustering after reassignment of clutches for mismatched pairs.

Mapped read count data.

A. percula replicate, social position (P1 – pair rank 1; P2 – pair rank 2; S – solitary) and clutch ID. Read counts of raw, filtered and trimmed sequences, and number of uniquely mapped reads against A. percula genome using STAR, as well as the corresponding mapping efficiency.

Complete list of Amphiprion percula genes associated with appetite regulation, glycolysis, and the TCA cycle, retrieved based on orthologs with A. ocellaris genes (Herrera et al., 2025).

Orthologs were identified using a reciprocal BLAST approach, and genes present in the whole-body gene expression dataset were retained. The table includes each gene and its associated p-values from pairwise contrasts between social positions (P1 – pair 1; P2 – pair 2; S – solitary). Genes with p-value ≤ 0.01, which were visualized in the heatmap shown in the main text (Fig. 6), are indicated in bold italics.

Post-hoc pairwise contrasts of total aggression (upper triangle) and total submission (lower triangle), displaying p-values of all comparisons. Significant differences are denoted with *.

List of significantly differentially expressed genes shared between the pairwise comparisons of P1 vs P2 and P1 vs S (main text Fig. 3B).

A total of 25 genes were consistently downregulated in P1 (relative to both P2 and S), and 84 genes were consistently upregulated in P1. Gene names are shown where available; otherwise, Ensembl gene IDs are provided.

Principal component analysis (PCA) of gene expression across gene subsets and PC axis combinations.

Each panel shows a PCA scatter plot of individual samples (n = 45) colored by social position (P1, P2, S), and shaped by genetic background (clutch ID). Rows represent gene subsets used for PCA: all genes (A–C), top 50% (D–F), top 25% (G–I), and top 5% most variable genes (J–L). Columns represent PC axis combinations: PC1 vs PC2 (left), PC2 vs PC3 (center), and PC1 vs PC3 (right). Ellipses show 95% confidence intervals estimated from a multivariate t-distribution for each social position. PERMANOVA p-values indicate the effect of social position on sample distribution, accounting for the genetic background of individuals (clutch ID). Panel B corresponds to Figure 3 in the main text.

Overall gene expression profiles showed no significant differences between P2 and S, compared to P1 individuals.

Principal Component Analysis (PCA) of all genes across social positions, accounting for clutch ID (genetic background), with PERMANOVA results showing a significant effect of social position. Samples are labeled, and replicates sharing the same numeric ID are connected by lines. Ellipses show the estimated confidence intervals from a 95% multivariate t-distribution.

Volcano plot of differential gene expression between pair-wise comparisons of social positions (P1: dominant, P2: subordinate, S: solitary) of A. percula.

Pairwise comparisons of differentially expressed genes (DEGs) between A) P1 and P2, B) P1 and S, and C) P2 and S individuals. Significant DEGs (adjusted p < 0.05) are colored according to the social position in which they are upregulated. The top 10 most significant DEGs in each pairwise comparison are labeled with gene name or Ensembl gene ID.

Modules of genes associated with phenotypes of interest.

A) Heatmap (merging threshold of 0) of Weighted Gene Correlation Network Analysis (WGCNA) to correlate gene expression pattern with social position (P1, P2, S), growth, appetite, orange hue and saturation of individuals. Only significant correlations are shown while non-significant correlations are marked with dash. B) The number of genes associated with each module found in WGCNA.

Gene Ontology (GO) terms identified in WGCNA modules showing correlation to phenotypes of interest.

Trees generated from GO enrichment analyses using Fisher’s Exact Tests (presence/absence in a module) to identify over-represented pathways within each module across GO divisions of Biological Processes (BP), Molecular Functions (MF), and Cellular Components (CC). Key GO terms, marked with stars, were explored in greater detail by extracting their associated genes and visualizing expression patterns of genes (FDR < 0.05) using heatmaps. Terms with red stars indicated GO terms with clear pattern (heatmap shown in main text), while yellow stars indicate terms which did not show clear patterns of DEGs dependent on social position (heatmaps not shown).

P1 individuals showed upregulation of genes associated with growth and ossification.

Heatmap of significant DEGs (FDR p-value ≤ 0.05) of growth and ossification genes in the brown Biological Process (BP) and Molecular Function (MF) GO terms across the three social positions. The bottom panel shows the size ratio (P2 SL/P1 SL) at week 5 for each replicate pair. Black bars indicate pairs with a size ratio ≤ 0.8 (more dissimilar in size), and red bars indicate pairs with a size ratio > 0.8 (more similar in size). The dashed line denotes a size ratio of 0.8. Samples (columns) are grouped a priori by social position (dominant, P1; subordinate, P2; solitary, S) and ordered within each group by the column dendrogram, which reflects expression similarity among samples within each social position. The gene (rows) dendrogram reflects relatedness among genes based on expression profile similarity across all samples. Genes that are significantly differentially expressed between individuals of P1 vs P2 are denoted with an asterisk (*), those between P1 vs S with a double asterisk (**), and three asterisk (***) represents both comparisons. Genes shown in bold and italics indicate significance at an FDR adjusted p-value < 0.1.

P1 individuals showed downregulation of appetite suppression and metabolic genes.

Heatmap of significantly differentially expressed genes (DEGs) (FDR p-value ≤ 0.01) of genes associated with glycolysis (GLY), Krebs cycle (TCA), and appetite regulation (APT), TCA cycle and glycolysis pathways across the three social positions. The bottom panel shows the size ratio at week 5 (P2 SL/P1 SL) for each replicate pair. Black bars indicate pairs with a size ratio ≤ 0.8 (greater size divergence), and red bars indicate pairs with a size ratio > 0.8 (more similar in size). The dashed line denotes a size ratio of 0.8. Samples (columns) are grouped a priori by social position (dominant, P1; subordinate, P2; solitary, S) and ordered within each group by the column dendrogram, which reflects expression similarity among samples within each social position. Genes (rows) are grouped a priori by pathway (GLY, TCA, or APT), and within each pathway group, the row dendrogram reflects relatedness among genes based on expression profile similarity across all samples. Genes that are significantly differentially expressed between individuals of P1 vs P2 are denoted with an asterisk (*), those between P1 vs S with a double asterisk (**), and three asterisks (***) represent both comparisons. Genes shown in bold and italics indicate significance at an FDR adjusted p-value < 0.1.

Principal component analysis (PCA) of gene expression in paired individuals (P1, P2) based on heatmaps of A) growth, B) appetite, and metabolic genes.

PCA was performed on specific gene sets (growth: Fig 5B; appetite and metabolism: Fig. 6), (P1 and P2; n = 30) using row Z-score scaled expression values, consistent with the heatmap scaling. Each point represents one individual. Ellipses represent 95% confidence intervals around each social position group. A PERMANOVA (adonis2; 999 restricted permutations; shown bottom right corners), found that size ratio was not a significant predictor of gene expression variation after accounting for social position and genetic background (A: psize_ratio = 0.858; B: psize_ratio = 0.844).

A) Hue of individuals of three social position (pair rank 1 = P1; pair rank 2 = P2; solitary = S) over time with mean and standard error of mean (SEM). B) saturation of individuals of social position (pair rank 1 = P1; pair rank 2 = P2; solitary = S) over time with mean and standard error of mean (SEM). Data points shown staggered for visual aid, but collected at the same time. time was a significant predictor of saturation (DF=1, F-value = 870.04, p<0.01), all fish showed an increase in saturation over the experiment (Supplementary Fig. S11B).