LPs have higher levels of replication stress compared with other mammary cell populations.

(A) Replication stress pathway scores of different human normal mammary cell populations. The gene expression data were from Lim et al. (3).

(B) T-distributed statistical neighbor embedding (t-SNE) plots visualizing the replication stress scores and ELF3 expression levels in breast cancer and normal mammary tissue from a BRCA1-mutated triple-negative breast cancer patient. The cells are colored by cell type (top) and replication stress score (bottom).

(C) Monocle-generated cell trajectories visualizing the change in replication stress scores during tumorigenesis. The tumor cells and compared normal luminal cells were obtained from a triple-negative breast cancer patient (case #3) carrying a BRCA1 germline mutation. LP, luminal progenitor cells; ML, mature luminal cells.

(D) Volcano plots showing DEGs in MCF10A-shBRCA1-Tet-on cell RNA-seq results after 1 μg/mL DOX treatment for 2 (left), 5 (middle) and 10 (right) days versus the Ctrl group. DEGs are genes with |log2-fold change| > 2 and FDR < 0.001.

(E) PCA plot of MCF10A-shBRCA1-Tet-on cell RNA-seq results.

(F) Mfuzz clustering analysis of Day 10 versus Ctrl DEGs. The heatmap (left) shows the differential expression mode of all genes in each cluster. The line chart (right) shows the global changing mode of each cluster.

(G) Overlap of HU treated MCF10A RNA-seq results and MCF10A-shBRCA1-Tet-on cell RNA-seq Cluster1 genes.

ELF3 is upregulated in BRCA1-associated breast cancers and is related to a worse prognosis.

(A) ELF3 expression levels in BRCA1-associated breast cancers and BRCA1-nonassociated breast cancers in the TCGA and METABRIC databases. BRCA1-associated breast cancers comprise samples with BRCA1 mutations, heterozygous loss and homozygous deletion (Mann–Whitney U test).

(B) Correlation of ELF3 and BRCA1 expression levels in TCGA breast cancer datasets.

(C) ELF3 expression levels of different molecular subtypes of breast cancers in the TCGA and METABRIC databases (one-way ANOVA).

(D) ELF3 expression levels of different IntClusts of breast cancers in the METABRIC database (one-way ANOVA; for IntClust 5 and 10, Mann–Whitney U test, ns, no significance).

(E) Immunohistochemistry images of human triple-negative, Her2+ and ER+ PR+ Her2+ breast cancer samples stained with the ELF3 antibody.

(F) ELF3 expression levels of the indicated types of human breast cancers (Mann– Whitney U test, *, p < 0.05).

(G) Survival analysis of the indicated subtypes of human breast cancer in the KMplot database. For the TNBC group, samples with negative expression of all hormone receptors (ER, PR and Her2) are included. In the TNBC group, high expression (n=54), low expression (n=62); in the basal group, high expression (n=85), low expression (n=211); in the Lum A group, high expression (n=121), low expression (n=101); in the Lum B group, high expression (n=133), low expression (n=67); and in the Her2 group, high expression (n=72), low expression (n=126).

ELF3 upregulation is induced by replication stress via the ATR-Chk1-E2F axis and by BRCA1 deficiency via GATA3 transcription.

(A) Spearman correlation analysis of ELF3 expression levels and replication stress scores in normal mammary tissue from noncarriers (left) and BRCA1 mutation carriers (right).

(B) Western blot and RT–qPCR results of MCF10A cells treated with the indicated drugs for 12 h. Drug concentration: HU 1 mM, ATRi (VE821) 10 μM, ATMi (KU5593) 10 μM.

(C) Western blot and RT–qPCR results of MCF10A cells treated with the indicated drugs for 12 h. Drug concentrations: HU 1 mM, ATRi (VE821) 10 μM, Chk1i (GDC-0575) 50 μM, Chk2i (BML-277) 10 μM.

(D) Western blot results of MCF10A cells transfected with the empty vector or Myc-E2F6 and treated with DMSO or 1 mM HU for 12 h.

(E) ChIP-Seq enrichment signal (green plot, ENCODE ENCFF858GLM; data are representative of two independent experiments) and the location of detected peaks (black line, ENCODE ENCFF692OYJ; irreproducible discovery rate (IDR) thresholded peaks) of E2F1 in MCF7 cells from the ENCODE database.

(F) ELF3 expression changes as BRCA1 deficiency duration increases. ELF3 expression levels are indicated by the CPM of the RNA-Seq results (means ± SD, n=3).

(G) Western blot and RT–qPCR results of MCF10A cells transfected with siNC or siBRCA1 for 48 h.

(H) RT–qPCR results of MCF10A cells transfected with siNC or siGATA3 for 48 h.

(I) ChIP-Seq enrichment signal (yellow plot up, ENCODE ENCFF384CPN; yellow plot down, ENCODE ENCFF342GNN. Data are representative of two independent experiments) and the location of detected peaks (black line up, ENCODE ENCFF352QVM; black line down, ENCODE ENCFF437NQS; both are irreproducible discovery rate (IDR) thresholded peaks) of GATA3 in MCF7 cells from two independent studies (ENCSR423RTK and ENCSR000BST) from the ENCODE database. The motif sequences were obtained from the JASPAR database.

(J) RT–qPCR results of MCF10A cells transfected with the indicated siRNA for 48 h.

(K) Western blot and RT–qPCR results of MCF10A cells transfected with siNC or siBRCA1 for 24 h followed by treatment with DMSO or 1 mM HU for 12 h.

(L) RT–qPCR results of MCF10A cells transfected with siNC or siGATA3 for 24 h followed by treatment with 1 mM HU for 12 h. In all panels of RT–qPCR results, data are presented as the mean ± SD, n=3, *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns, no significance, by paired two-tailed Student’s t test.

ELF3 helps suppress excessive genomic instability.

(A and B) Cell viability curve of MCF10A cells transfected with siNC or siELF3 for 48 h, treated with the indicated concentration of drugs (HU: μM; cisplatin: nM) and measured by the CCK-8 assay (mean ± SD, n=3, two-way ANOVA, **, p < 0.01; ***, p < 0.001; ****, p < 0.0001).

(C and D) Cell viability of HCC1937 (C) and SUM149PT (D) cells transfected with siNC or siELF3 for 5 days detected by the CCK-8 assay (mean ± SD, n=4, paired two-tailed Student’s t test, **, p < 0.01; ***, p < 0.001; ****, p < 0.0001).

(E and F) Cell viability curve of HCC1937 cells transfected with siNC or siELF3 for 48 h, treated with the indicated concentration of drugs (HU: μM; cisplatin: nM) and measured by the CCK-8 assay (mean ± SD, n=3, two-way ANOVA, **, p < 0.01; ***, p < 0.001; ****, p < 0.0001).

(G and H) Correlation of ELF3 expression levels and cell sensitivity of the indicated drugs in the CellMiner (G) and GDSC (H) databases.

(I) Nude mouse tumors of SUM149PT cells infected with shCtrl of shELF3 lentivirus (mean ± SD, unpaired two-tailed Student’s t test, *, p < 0.05).

ELF3 helps maintain the stability of DNA replication.

(A) Spontaneous γH2AX foci of HCC1937 cells transfected with siNC or siELF3 for 72 h (mean ± SEM, n > 100, unpaired two-tailed Student’s t test, **, p < 0.01; ****, p < 0.0001).

(B and C) γH2AX and 53BP1 foci of MCF10A cells transfected with siNC or siELF3 for 48 h and treated with 2 mM HU for 4 h followed by release for 20 h (mean ± SEM, n > 100, unpaired two-tailed Student’s t test, *, p < 0.05; **, p < 0.01; ***, p < 0.001).

(D) Proportions of cells with γH2AX (⩾ 3) and 53BP1 foci from (B and C) (mean ± SD, unpaired two-tailed Student’s t test, *, p < 0.05; **, p < 0.01; ***, p < 0.001).

(E) Enrichment plot of DNA replication pathways in MCF10A ELF3 knockdown RNA-seq.

(F) RT–qPCR results of MCF10A cells transfected with siNC or siELF3 for 72 h (mean ± SD, n=3, paired two-tailed Student’s t test, *, p < 0.05; **, p < 0.01; ns, no significance).

(G) Replication fork velocity indicated by DNA fiber assay of MCF10A cells transfected with the indicated siRNA for 72 h (mean ± SD, n > 80, unpaired two-tailed Student’s t test, ****, p < 0.0001).

(H) Replication fork symmetry indicated by the DNA fiber assay of MCF10A cells transfected with the indicated siRNA for 72 h (mean ± SD, Mann–Whitney test, ***, p < 0.001; ****, p < 0.0001).

(I) Replication fork stability of MCF10A cells transfected with the indicated siRNA for 72 h (mean ± SD, n > 100, unpaired two-tailed Student’s t test, **, p < 0.01; ****, p < 0.0001).

BRCA1 deficiency disturbs the differentiation of LPs.

(A) LP signature score of samples of MCF10A-shBRCA1-Tet-on RNA-seq. The signature score was generated using data from Lim et al. (3).

(B) Barcode plot of LP gene set enrichment in Day 10 vs. Ctrl. LP gene sets were obtained from data from Pellacani et al. (33).

(C) GSEA plots of the transcriptional profile of MCF10A cells with BRCA1 deficiency, MCF10A cells with ELF3 overexpression and normal human LPs. The LP gene set was obtained from the data of Pellacani et al. (33).

(D) Violin plots displaying the distribution of ELF3 expression levels in distinct cell types in normal mammary tissue from noncarriers (left) and BRCA1 mutation carrier (right).

(E) ETS subfamily transcription factor motif enrichment results of LP gene promoters.

(F) Expression levels of ETS subfamily transcription factors in LPs from BRCA1 mutation carrier and noncarrier of the single-cell sequencing data.

(G) ETS subfamily transcription factor enrichment of ATAC-Seq of LPs of BRCA1-deficient mice and WT mice from Bach et al. (37).

Replication stress inducing ELF3 upregulation promotes BRCA1-deficient breast tumorigenesis.

LPs have higher replication stress levels under normal conditions (left) and thus are more likely to exceed the tumorigenesis threshold under BRCA1 deficiency (right). Replication stress and BRCA1 deficiency results in further ELF3 upregulation (top right). ELF3 can suppress excessive replication stress and help LPs maintain moderate levels of genomic instability, which is tolerable for cells to proliferate and enough to fuel cancer evolution (a). In addition, ELF3 upregulation can boost LP gene transcription, leading to LP dedifferentiation and expansion (b). Finally, as an EMT promoter in the mammary tissue, ELF3 can accelerate the transformation process from LPs to malignant cells (c).

LPs have higher levels of replication stress compared with other mammary cell populations.

(A) Heatmap showing the expression pattern of replication stress-related genes in BRCA1-mutated triple-negative breast cancer cells. The cancer cells were ordered based on Monocle-generated cell trajectories of tumorigenesis (Figure 1C).

(B) RT–qPCR (mean ± SD, n=3, paired two-tailed Student’s t test, ****, p < 0.0001) and Western blot assay results of MCF10A-shBRCA1-Tet-on cells with or without 1 μg/mL DOX treatment for 48 h.

(C) Colony formation assay results of MCF10A-shBRCA1-Tet-on cells treated with or without 1 μg/mL DOX for 14 days.

(D) Neutral comet assay results of MCF10A-shBRCA1-Tet-on cells with or without 1 μg/mL DOX treatment for 48 h (mean ± SD, unpaired two-tailed Student’s t test, ****, p < 0.0001).

(E) Schematic of the experimental procedure of MCF10A-shBRCA1-Tet-on RNA-Seq.

(F) Scatter plot of the PCA results. The horizontal axis represents different principal components.

ELF3 expression upregulation in basal-like breast cancers is not due to amplification.

Proportions of ELF3 amplification in different molecular subtypes of TCGA breast cancers. The numbers represent the percentage of ELF3 amplification in each subtype.

ELF3 upregulation is induced by replication stress via the ATR-Chk1-E2F axis and by BRCA1 deficiency via GATA3 transcription.

(A) Scores of E2F1, GATA3 and MYC motif enrichment in the ELF3 promoter sequence.

(B) RT–qPCR knockdown results related to Figure 3J.

(C) RT–qPCR knockdown results related to Figure 3K.

(D) RT–qPCR knockdown results related to Figure 3L. In all panels of RT–qPCR results, data are presented as the mean ± SD, n=3, *, p < 0.05; **, p < 0.01; ***, p < 0.001, by paired two-tailed Student’s t test.

ELF3 helps suppress excessive genomic instability.

(A) Western blot knockdown results related to Figure 4A and B.

(B and C) Western blot knockdown results related to Figure 4C and D.

(D and E) Cell viability curve of SUM149PT cells infected with shCtrl or shELF3 lentivirus and treated with the indicated concentration of drugs (HU: μM; cisplatin: nM) and detected by the CCK-8 assay (mean ± SD, n=3, paired two-tailed Student’s t test, *, p < 0.05).

(F) Cell viability curve of HCC1937 cells transfected with siNC or siELF3, treated with indicated concentration of olaparib (nM) and measured by the CCK-8 assay (mean ± SD, n=3, two-way ANOVA, **, p < 0.01; ***, p < 0.001).

(G) Correlation of ELF3 expression levels and cell sensitivity to PARPi in the GDSC database.

ELF3 helps maintain the stability of DNA replication.

Neutral comet assay results of MCF10A cells transfected with siNC or siELF3 for 72 h (mean ± SEM, unpaired two-tailed Student’s t test, **, p < 0.01; ***, p < 0.001).

BRCA1 deficiency disturbs the differentiation of LPs.

(A) Violin plots displaying the distribution of ELF3 expression levels in distinct cell populations in normal mammary tissue from two noncarriers.

(B) ELF3 expression levels (RPKM) in different cell populations from human normal mammary tissue in data of Pellacani et al. (33).

(C) Scatter plot of all transcription factors identified by motif enrichment of the ATAC-seq data of Bach et al. (37).