The authors used spatial profiling platforms—Xenium and RNAscope—to chart ligand-receptor relationships between tumor cells and surrounding cell types. They categorized tumor senders into niche-associated transcriptional states (Hypoxic: S2 or S4; Stem-like: S1 or S5) and treated myeloid, neuronal/glial, and vascular populations as receivers in the interaction analysis. The figure legend states ligand-receptor pairs unique to a specific sender-receiver combination were italicized and asterisked in the visualization. The source did not provide quantitative thresholds or exact detection criteria in the figure legend.
An interaction network illustrates where the ADM:CALCRL pair reached significance across samples and sender-receiver combinations. Arrow color and density scaled with the number of samples in which an ADM:CALCRL interaction was detected as significant for that pairing, according to the figure legend. The network therefore conveys recurrence across the cohort rather than absolute counts per sample. The manuscript text or legend did not enumerate specific sample counts per arrow, nor did it include statistical test details for those detections in the figure description.
The figure contains a heatmap showing expression of ADM and CALCRL by cell type across defined tumor niches. This view aggregates per-cell-type expression patterns to make niche-specific expression visible. Data in the heatmap were presented visually; the legend did not provide numeric expression values, normalization details, or thresholds used to define positive expression within the figure caption.
Panel D reports expression of ADM and CALCRL stratified by tumor-region localization using the IVY-GAP resource. The authors therefore probed whether regional tumor architecture correlates with expression of this ligand-receptor axis. The figure caption indicates IVY-GAP as the data source but does not list which IVY-GAP regions were compared or supply numerical comparisons within the legend.
Bar plots summarize the proportion of ADM+ cells among SOX2+ tumor cells and the proportion of CALCRL+ cells among PECAM1+ vascular cells across all samples. The legend specifies sample size as n=5. The figure shows percentage-based bar plots, but the caption does not report exact percentage values, measures of spread (for example, SD or SEM), or statistical comparisons between groups.
The authors computed median distances, on a per-sample basis, from ADM+ SOX2+ tumor cells to the nearest CALCRL- or CALCRL+ PECAM1+ cells; results were presented as bar plots per sample. These median nearest-neighbor metrics provide a spatial readout of proximity between ligand-expressing tumor cells and receptor-expressing vascular cells. The figure legend did not include absolute distance units in the caption, nor did it disclose the image-processing parameters used to determine cell centroids or nearest-neighbor assignments.
A parallel set of panels evaluated the JAG1:NOTCH3 axis. Bar plots show the percent of JAG1+ among SOX2+ tumor cells and the percent of NOTCH3+ among PDGFRB+ or PECAM1+ stromal/vascular cells across the same sample set. The figure legend describes these percent-positive summaries but does not provide the underlying counts, the staining thresholds, or the statistical methods used to compare groups.
For the NOTCH signaling pair, the authors report median per-sample distances from JAG1+ SOX2+ tumor cells to the closest NOTCH3- or NOTCH3+ PDGFRB+ cells. These distance metrics mirror the approach taken for ADM:CALCRL and give a spatial context for potential ligand-receptor engagement in situ. The legend does not state whether distances were measured edge-to-edge or center-to-center, nor does it provide numerical distance summaries in the caption.
Limitations and open items documented in the figure legend
The figure compiles interaction detection, heatmaps, IVY-GAP localization, percentage-positive summaries, and median nearest-neighbor distances. The legend explicitly reports sample counts for the percent-positive plots (n=5) but omits many procedural and quantitative details: specific numeric percentages, confidence intervals, p-values, the image analysis pipeline used for nearest-neighbor distance calculations, normalization methods for expression heatmaps, and exact IVY-GAP region labels compared. These omissions limit direct interpretation of effect sizes and statistical robustness from the figure alone.
Questions remaining from the presented data
Which IVY-GAP regions showed the largest differential expression of ADM or CALCRL? What were the absolute percentages and variance for ADM+ SOX2+ and CALCRL+ PECAM1+ populations? How consistent were the median distance relationships across all samples beyond the per-sample bars? The figure legend does not provide these numeric answers; readers must consult the main text or supplementary materials for further methodological and statistical detail.