How to write a figure legend

Figure 2. Loss of ATG7 blocks autophagic flux in HeLa cells. (A) Immunoblot of LC3-I, LC3-II and p62 in control and ATG7-knockout (KO) cells treated with vehicle or 100 nM bafilomycin A1 (Baf A1) for 4 h; GAPDH is the loading control. (B) LC3-II relative to GAPDH, quantified from blots as in (A). (C) Representative confocal images of GFP-LC3 puncta (green); nuclei are stained with DAPI (blue). (D) GFP-LC3 puncta per cell, with at least 50 cells counted per condition. Data in (B) and (D) are mean ± SD of n = 3 independent experiments; two-way ANOVA with Tukey’s post hoc test; **P < 0.01, ***P < 0.001; ns, not significant. Scale bar, 10 µm.

• The title gives the result, not the method.
• Panels are described in order, and (B) points back to (A) instead of repeating it.
• Colours, abbreviations and the loading control are defined inside the panel descriptions.
• Replicates, the test and the thresholds behind the asterisks come in one sentence at the end, with the scale bar last.

Figure 4. Losartan reduces cortical fibrosis 28 days after unilateral ureteral obstruction (UUO). (A–C) Masson’s trichrome staining of kidney cortex from sham (A), UUO (B) and UUO plus losartan (C) mice; collagen appears blue. Arrowheads indicate dilated tubules. (D) Fibrotic area as a percentage of cortex, measured in five random fields per animal. n = 6 mice per group; mean ± SEM; one-way ANOVA with Dunnett’s test versus UUO; *P < 0.05. Scale bars, 100 µm.

• Grouped panels (A–C) share one description, with the condition named for each letter.
• The one symbol in the images, the arrowhead, is defined.
• The reader learns how the number in (D) was obtained: five fields per animal, six animals per group.

Figure 1. Proposed mechanism of cisplatin resistance through ERCC1-dependent nucleotide excision repair. Cisplatin (left) forms intrastrand DNA cross-links that stall replication. In sensitive cells (top) the lesion persists and triggers apoptosis; in resistant cells (bottom) the ERCC1–XPF nuclease excises the damaged strand and repair synthesis restores the duplex. Solid arrows indicate established steps; the dashed arrow marks the hypothesized induction of ERCC1 by prior cisplatin exposure, tested in Figure 5. Schematic drafted with Scientific Figure Generator from an author-written description; the authors reviewed every label and relationship.

• “Proposed” in the title keeps the figure on the hypothesis side of the evidence line.
• The reading order (left, top, bottom) is written out so the legend works without the picture.
• Solid and dashed arrows are separated into established and hypothesized, and the hypothesis points to the figure that tests it.
• The AI disclosure is the last sentence, the placement Elsevier asks for; Nature Portfolio generally does not accept such images, and Science journals need the editors’ explicit permission first.

Figure 3. Single-cell RNA-sequencing workflow. Tumour biopsies were dissociated into single cells, sorted for viable CD45-negative cells and loaded on the 10x Chromium platform (steps 1–3); libraries were sequenced and processed with Cell Ranger, then clustered and annotated (steps 4–6). Numbers in circles refer to the Methods subsections. Grey boxes mark quality-control checkpoints and list the thresholds applied.

• A workflow legend explains the sequence and the conventions (numbers, grey boxes), not the parameters of each step.
• Pointing the numbers at the Methods subsections keeps the legend short and auditable.

Figure 5. Temperature dependence of the ionic conductivity of the LLZO electrolyte. Arrhenius plot of conductivity measured by impedance spectroscopy between 25 and 150 °C for pellets sintered at 1,100 °C (circles) and 1,200 °C (squares). Lines are linear fits; activation energies are given in the key. Error bars show the standard deviation of three pellets and are smaller than the symbols where not visible.

• Symbols map to conditions in the sentence that names them.
• What the lines are, and what the error bars are, is stated even when they are hard to see.
• Units and ranges match the axes; nothing in the legend contradicts the plot.