Fig 2: Lowering ATF3 improves the effect of 5-FU on CRC MCTS.A Plot for protein interactions among genes ATF3, ATF4, DDIT4, TRIB3 and FGD6 by the STRING interaction network. B Expression levels of ATF3, ATF4, DDIT4 and TRIB3 in HCT116, DLD-1 and HT-29 cells treated with DMSO or 100 nM sangivamycin for 24 h under monolayer conditions. Actin is the loading control. C Expression levels of ATF3, DDIT4 and TRIB3 were assessed in COAD tissues and normal adjacent tissues using RNA-Seq data from TCGA (* p < 0.05). Cohort size: (num)normal adjacent tissues = 349; (num) COAD tissues = 275. D Western blot analysis of the co-IP complex from cells treated with DMSO or 100 nM sangivamycin for 24 h with ATF3 antibody immobilized on Dynabeads®. Tubulin is the loading control for the input conditions. E Expression levels of ATF3, ATF4, DDIT4 and TRIB3 in HCT116 ATF3parental and ATF3KO cells treated with DMSO or 100 nM sangivamycin for 24 h under monolayer conditions. Actin is the loading control. F Illustration of the interaction between ATF3, ATF4, DDIT4 and TRIB3 based on our results and findings from other groups [79–81]. G ATF3 expression levels in HCT116 and HT-29 cells treated with the indicated concentrations of rapamycin for 72 h under monolayer conditions. Actin is the loading control. H ATF3 expression levels in HCT116 MCTS cells treated with rapamycin (10 µM) or Torin-1 (0.5 µM) for 72 h. Actin is the loading control. I Dose response of HCT116 and HT29 MCTS to 5-FU. The data is expressed as the mean ± SD (n = 3). The half-maximal inhibitory concentration (IC50) was calculated using nonlinear regression, [Inhibitor] vs. normalized response analysis. J Combination of rapamycin (10 µM) or Troin1(0.5 µM) with 5-FU under HCT116 MCTS conditions for 72 h. The data is expressed as the mean ± SD (n = 3, unpaired t-test, two-stage step-up (Benjamini, Krieger, and Yekutieli); * p < 0.05, *** p < 0.001). K Combination of rapamycin (10 µM) or Troin1(0.5 µM) with 5-FU under HT29 MCTS conditions for 72 h. CDI = AB/(A × B) where: AB = cell viability value for the combination of rapamycin/torin-1 and 5-FU. A and B = cell viability value for the single treatment rapamycin/torin-1 or 5-FU. A CDI value <1, =1 or >1 indicates that the drugs are synergistic, additive, or antagonistic, respectively. A CDI value less than 0.7 indicates a significant synergetic effect [31].

Fig 3: Cells lacking ATF3 show upregulated mitochondrial function and are more sensitive to anticancer agents.A Expression levels of ATF3 in ATF3parental and ATF3KO cells under monolayer and MCTS conditions. Actin is the loading control. Three independent experiments were performed. B Images of ATF3parental and ATF3KO MCTS taken by Incucyte, Scale bar: 300 µm. C GSEA enrichment profile of the top-altered hallmark pathways in HCT116 MCTS lacking ATF3 compared with HCT116 ATF3parental MCTS. D Reactome analysis showing upregulation of the respiratory electron transport pathway and mitochondrial translation elongation in ATF3-deficient HCT116 MCTS. Table lists genes involved in complex I biogenesis and mitochondrial translation elongation. E Flow chart of hit compound identification using mechanistic screening set (Library description: Mechanistic Set VI, https://dtp.cancer.gov). F NSC numbers and chemical structures of the hit compounds. G Dose response to hit compounds in HCT116 ATF3parental and ATF3KO cells under monolayer conditions. The half-maximal inhibitory concentration (IC50) was calculated using nonlinear regression, [Inhibitor] vs. normalized response analysis. H Dose response to hit compounds in HT-29 ATF3scramble and ATF3KD cells under monolayer conditions. The half-maximal inhibitory concentration (IC50) was calculated using nonlinear regression, [Inhibitor] vs. normalized response analysis. I Expression levels of ATF3 in HCT116, DLD-1 and HT-29 cells treated with DMSO or 100 nM sangivamycin for 24 h under monolayer conditions. Actin is the loading control. J Expression levels of ATF3 in ATF3parental and ATF3KO cells treated with DMSO or 100 nM sangivamycin for 24 h under MCTS conditions. Actin is the loading control. K Dose response to sangivamycin. HCT116 ATF3parental and ATF3KO cells were treated with sangivamycin in monolayer condition for 72 h. The data is expressed as the mean ± SD (n = 3, unpaired t-test, two-stage step-up (Benjamini, Krieger, and Yekutieli); * p < 0.05, ** p < 0.01, *** p < 0.001). The half-maximal inhibitory concentration (IC50) was calculated using nonlinear regression, [Inhibitor] vs. normalized response analysis. L MCTS treated with indicated concentrations of sangivamycin was monitored in Incucyte for 3 days. Images for each spheroid were taken every day. The volume of spheroids was calculated by equation Volume = (4/3)*3.14*R3 (R = radius). The data is expressed as the mean ± SD (n = 3, using multiple-unpaired t-test, * p < 0.05).

Fig 4: TMEM106C regulates apoptosis pathway and LINC00238 is transcriptionally regulated by ATF3 in HCC. (A) The Enriched KEGG pathway analysis showed that TMEM106C suppressed the apoptosis signaling pathway. (B) The results of western blotting showed that protein expression of ASP7, TIMP2, PDCD4 and ATF3 in HCCLM3 and MHCC-97h cells was increased by LINC00238 overexpression while decreased by TMEM1106C overexpression. *P<0.05 vs. NC group; #P<0.05 vs. pc-LINC00238 group. (C) Recognition motif of ATF3 from the JASPAR database. (D) Venn diagram showing ATF3 is upstream transcription factor of LINC00238 of LINC00958 predicted by PROMO and KEGG. (E) The results of western blotting showed that ATF3 protein expression in HCCLM3 and MHCC-97h cells was decreased by ATF3 siRNAs. (F) The results of RT-qPCR showed that LINC00238 expression in HCCLM3 and MHCC-97h cells was decreased by ATF3 siRNAs. (G) The results of RT-qPCR showed that ATF3 expression was significantly lower in HCC tissues than that in corresponding normal tissues. (H) Scatter plots comparing ATF3 expression in HCC samples (n=396) and normal tissue samples (n=160) showed that ATF3 was downregulated in HCC tumor samples compared with normal samples. (I) Kaplan-Meier analysis indicated an association between high TMEM106 and poor overall survival in patients with HCC. (J) The correlation analysis among ATF3, LINC00238 and TMEM106C showed that the ATF3 expression was positively correlated with LINC00238 and negatively correlated with TMEM106C expression. *P<0.05. TMEM106C, transmembrane protein 106C; LINC, long intergenic non-protein coding RNA; HCC, hepatocellular carcinoma; KEGG, Kyoto Encyclopedia of Genes and Genomes; ATF3, activating transcription factor 3; CASP7, caspase-7; TIMP2, tissue inhibitor of metalloproteinase 2; PDCD4, programmed cell death 4; pc-, plasmid; si, short interfering; RT-qPCR, reverse transcription-quantitative PCR.

Fig 5: KLF10 knockdown promotes activation of LX-2 human HSCs via ATF3. LX-2 cells were transfected with siKLF10 for 48 h and further treated with 10 ng/mL TGF-β1 for another 24 h. The chemical chaperone 4-PBA was applied 2 h before TGF-β1 treatment. (A) Cell viability was determined using MTT assays. Data are expressed as means ± SEM. (B) Levels of α-SMA, fibronectin, and Col1α1 were determined by Western blotting. (C) Collagen accumulation was evaluated by Picro-Sirius red staining. Collagen, red. Scale bar = 100 μm. (D) LX-2 cells were transfected with siKLF10 for 48 h and further treated with 10 ng/mL TGF-β1 for the indicated times. The chemical chaperone 4-PBA was applied 2 h before TGF-β1 treatment. The levels of ATF3, p-SMAD2, p-SMAD3, and SMAD2/3 were determined by Western blotting. GAPDH was used as an internal control.