(A and B) The left heatmap shows data quality assessment by hierarchical sample-to-sample distance clustering. Variance-stabilizing transformation of the RNA-Seq read count for all the samples was used to calculate the sample-to-sample Euclidean distance (color scale) for hierarchical clustering. Controls (C1–C3), radiated (IR1–IR2), or chemotherapy-treated (CT1–CT3) whole tibial bones from young mice. GSEA of the radiated bones in comparison to bones from control mice. The most significant biological processes were assessed by GAGE, with q < 0.000001. The most differentially downregulated biological processes are shown in the bar graphs. The y axis shows the GO terms, whereas the x axis represents the enrichment scores of these terms. The right heatmap in A shows 4136 differentially expressed genes between the radiated bones in comparison to bones from the control mice (FDR-adjusted P value cutoff <0.01, log₂ fold change ±1). The right heatmap in B show 3119 differentially expressed genes between carboplatin-treated and control bones (FDR-adjusted P value cutoff <0.01, log₂ fold change ±1). Color intensity represents row-scaled-normalized log₂(cpm) expression, whereas the columns represent the replicate of each samples. (C) qPCR analysis of Bmp4, Bmp6, Bmp7, Kitl, Tgfb2, and Thbs2 expression (normalized to Actb) by radiated or chemotherapy-treated tibiae relative to controls from young mice. Data represent mean ± SD (n = 8 replicates), and 2-tailed unpaired t tests. (D) Bubble graph shows the qPCR analysis of Bmp7, Cxcl12, Kitl, Tgfb2, Thbs2, Bmp4, and Bmp6 expression (normalized to Actb) by different cell types (as indicated in the figure) in BM from young mice in comparison to levels in osteoclasts. Data represent mean (n = 5 replicates). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.