Fig 2: MLX knockdown disrupts redox balance and promotes ferroptosis in osteosarcoma cells.A Determination of cystine in 30 pairs of osteosarcoma and normal control tissues. Data are represented as mean ± SD. Paired t-test was used. B, C The measurement of Cysteine and GSH in 143B upon MLX knockdown. Data are represented as mean ± SD, n = 3. Unpaired t-test was used. D The measurement of ROS level in shCont- and shMLX-143B. Data are represented as mean ± SD, n = 3. Unpaired t-test was used. E Detection of MLX and 4-HNE protein levels in 143B cells upon MLX knockdown by western blotting (WB). F WB analysis of transferrin, transferrin receptors, ferritin and ferroportin protein levels in 143B cells with shMLX knockdown. G Measurement of intracellular ferrous iron levels in 143B cells upon MLX knockdown. Data are presented as mean ± SD, n = 3. Unpaired t-test was used. H Percentage of cells with increased levels of lipid peroxidation in shCont- and shMLX-143B cells. Data are presented as mean ± SD, n = 3. Unpaired t-test was used. I 24-hour cell viability of osteosarcoma cells treated with MLX knockdown and ferroptosis inhibitors (10 μM Liproxstatin-1, 10 μM Ferrosatin-1, 50 μM Deferasirox, and 100 μM DFO). Data are presented as mean ± SD, n = 3. Unpaired t-test was used. ***: P ≤ 0.001, **: P ≤ 0.01, *: P ≤ 0.05, n.s.: not significant.

Fig 3: MLX is essential for the growth and metastasis of osteosarcoma cells in vitro and in vivo.A Relative growth curves of 143B and SJSA1 cells stably transduced with nontargeting scrambled control shRNA (shCont) or two MLX shRNAs (shMLX#1 and shMLX#2). Data are represented as mean ± SD, n = 3. Unpaired one-way ANOVA test followed by Dunnett’s test was used. B Invasion assay conducted in shMLX versus shCont in 143B (upper panel) and SJSA1 (lower panel). Data are represented as mean ± SD, n = 3. Unpaired t-test was used. C Sphere formation assay in 143B (upper panel) and SJSA1 cells (lower panel). Data are represented as mean ± SD, n = 3. Unpaired t-test was used. D Representative images of tumor growth in 143B-derived orthotopic model (n = 5 per group). Unpaired t-test was used. E The representative images of metastasis in 143B-derived orthotopic model (n = 5 per group). Unpaired t-test was used. F Representative images of H&E, Ki-67 and MLX staining of 143B-shCont and 143B-shMLX tumors and the respective analyses (scale bar = 50 μm). Data are represented as mean ± SD, n = 3. Unpaired t-test was used. ***: P ≤ 0.001, **: P ≤ 0.01, *: P ≤ 0.05, n.s.: not significant.

Fig 4: Analysis of MLX binding and regulation of essential spermatogenesis genes in testes.(A) Venn diagram showing binding overlap of loci occupied by MLX and MAX in WT and MLXKO testes identified by ChIP-Seq. (B) Examples of individual ChIP-Seq tracks for Txnip and male GC-specific genes (Prm and Tnp) showing occupancy by MLX and MAX in WT and MLXKO testes. (C) Volcano plot showing LFC in mRNA expression between WT versus MLXKO testis with genes that are MLX bound (dark blue dots) compared with those genes not bound by MLX (light blue dots) (KS test p = 0.000000e+00). (C’) Violin plots of LFC data from (C). (D) Volcano plot showing LFC in mRNA expression between WT versus MLXKO testis with genes that are MAX(WT)-Bound in WT testes (dark blue dots) or not bound (light blue dots) (KS test p = 3.853399e-02). (D’) Violin plots of LFC data from (D). (E, E’) Venn diagrams showing (E) overlap of MLX bound genes with CREM and DMRT1 bound genes (E’) overlap of MAX with known CREM and DMRT1 target genes in WT testes. (F, F’) Violin plots comparing LFC of WT versus MLXKO testes for: (F) CREM-bound loci and (F’) for DMRT1-bound loci (t test). (G) Venn diagrams showing binding overlap of MAX(MLXKO)-bound genes in MLXKO testes with known CREM and DMRT1 target genes in testes. (H) Volcano plot showing LFC in mRNA expression between WT versus MLXKO testis with genes that are MAX(MLXKO)-bound in MLXKO testis (KS test p = 7.366175e-05). (H’) Violin plots comparing LFC of WT versus MLXKO testes for MAX(WT)-bound versus unbound loci. (I) Diagram summarizing modes of MLX and MAX binding to subsets of target genes. De novo targets refers to MLX-bound targets only occupied by MAX in MLXKO testis. In WT cells, MLX is predicted to either directly or indirectly exclude MAX from these sites. For Venn diagrams, p-values shown are derived from hypergeometric tests. For volcano plots, a KS test was used, and for violin plots, a t test was used. The underlying data for Fig 7A–7H can be found in S1 Data. ChIP-Seq, chromatin immunoprecipitation and sequencing; GC, germ cell; KO, knockout; LFC, log fold change; MLX, MAX-Like protein X; WT, wild-type.

Fig 5: Characterization of hepatic adenomas originating in MlxKO and DKO livers. (A) Number of mice ages 14–16 months of the indicated genotypes with visible liver tumors at the time of death. (B) Gross appearance of tumor-containing livers from panel A. Tumors were generally small but multifocal (arrows). (C) H&E-stained sections showing the typical appearance of MlxKO and DKO livers. The first 2 MlxKO images show well-differentiated and myxoid-type adenomas, respectively, whereas the third panel shows a focus of well-differentiated HCC embedded within an adenoma. The first 2 DKO images show regions of microvesicular steatosis with balloon cells and nuclear enlargement, respectively, whereas the third image shows another adenoma that is quite similar in appearance to those arising in MlxKO livers. (D) H&E and Ki-67 immunostained sections of the Myc-expressing adenoma from panel D showing regions of inflammation (arrowhead), adjacent to those resembling well-differentiated HCC (arrow). Ki-67 staining is more intense within the nodular adenoma (lower right) compared with adjacent normal liver (upper left). Quantification of Ki-67 staining from 4 different adenomas and adjacent normal tissue was performed on 300–2000 cells from each region. The mean Ki-67 index in adenomas was 30.08% ± 6.5% vs 3.2% ± 3.1% in adjacent nonadenomatous tissues (P = 3.4 × 10-5). (E) Immunoblots showing Mlx expression in WT livers and its absence in adenomas arising from MlxKO and DKO livers. Only 1 large adenoma from an MlxKO mouse with hepatomegaly expressed detectable levels of Myc (lane 2). (F) Principal components analysis (PCA) of whole transcriptomes from WT livers, HBs generated by the overexpression of mutant forms of β-catenin+YAPS127A, and the earlier-described adenomas. (G) Whole-transcriptome profiles of the tissues from panel F. Note that control liver samples were derived from 2 different strains: C57B6 and FVB. (H) Heat maps of 15 of the 22 transcripts that are dysregulated in murine HBs and correlate with poor outcomes in human HBs and other human cancers.47 GAPDH, glyceraldehyde-3-phosphate dehydrogenase; PC, Principal Component.