Temporal and spatial cooperation of Snail1 and Twist1 during epithelial-mesenchymal transition predicts for human breast cancer recurrence

Abstract

Epithelial-mesenchymal transition (EMT) is a normal developmental program that is considered to also play an important role in cancer metastasis. Ultimate inducers of EMT are transcriptional repressors that individually can induce experimental EMT, yet in many cells, particularly cancer cells, multiple inducers are expressed simultaneously. Why, and if, and how they interact to regulate EMT is unanswered. Using RNA interference technology to affect protein knockdown and avoid potential overexpression artifact coupled with transient TGFβ treatment to better mimic in vivo conditions we show, in both nontumorigenic and tumorigenic epithelial cancer cells, that Snail1 is uniquely required for EMT initiation, whereas Twist1 is required to maintain late EMT. Twist1, present in resting epithelial cells, is dispensable for EMT initiation. Mechanistically, in response to transient TGFβ treatment, transient Snail1 expression represses Twist1 transcription directly, which is subsequently upregulated, as Snail1 levels decrease, to sustain E-cadherin downregulation and growth arrest of EMT. Persistent Twist1 expression is associated with a p38 and extracellular signal-regulated kinase signal feedback loop that sustains growth-inhibitory signals characteristic of quiescent micrometastatic tumors. This Snail1-Twist1 temporal and spatial cooperation was also observed in vivo during human breast cancer progression to metastasis. Twist1 level, but not Snail1 level, and Twist1:Snail1 ratio in disseminated micrometastatic bone marrow tumor cells was found to correlate with survival and treatment resistance and is highly predictive of metastatic or recurrent disease.

Figure 1. Transient TGFβ1-induced EMT in MCF10A cells

(A) Phase contrast photographs and immunofluorescent photographs for indicated proteins at indicated times of MCF10A cells treated with transient TGFβ1 for 4h. DAPI (blue) was used to detect nuclei in all panels. Representative results of 3 independent experiments are shown. (B – D) Western blotting for indicated proteins of whole cell lysates (WCL) from MCF10A cells treated with 2 ng/ml TGFβ1 either continuously or transiently for 4h and cells collected at indicated times (B) Graphs of densitometric intensity of Snail1 protein (C) and Twist1 protein (D) from MCF10A cells treated with TGFβ1 as in (B). (E – G) RT-PCR for indicated mRNA of total RNA from MCF10A cells treated with 2ng/ml TGFβ1 either continuously or transiently for 4h and cells collected at indicated times (E). Graphs of densitometric intensity of Snail1 mRNA (F) and Twist1 mRNA (G) from MCF10A cells treated with TGFβ1 as in (E). Error bars represent mean +/− s.e.m. from at least 3 independent experiments. See also related Figures S1 and S2.

Figure 2. Snail1, but not Twist1, is uniquely required for the initiation of TGFβ1-induced EMT

(A and B) Western blotting for indicated proteins of WCL from MCF10A cells transduced with a lentivirus expressing either the control luciferase or Snail1 shRNA (A) or Twist1 shRNA (B), then treated with 2 ng/ml TGFβ1 continuously for indicated times. Representative results of 3 independent experiments are shown. See also related Figures S3 and S4.

Figure 3. Twist1, but not Twist2, is required for EMT maintenance in MCF10A cells

(A and B) Western blotting for indicated proteins of WCL from MCF10A cells transduced with a lentivirus expressing either the control luciferase shRNA or Twist1 shRNA (A) or Twist2 shRNA (B), and treated with 2 ng/ml TGFβ1 either continuously or transiently for 4h, and cells collected at indicated times. Representatives of 3 independent experiments are shown. See also related Figure S4. (C) Western blotting for indicated proteins of WCL from MCF10A cells transduced with a lentivirus expressing either the control luciferase or Twist1 shRNA, treated with 2ng/ml TGFβ1 transiently for 4h, and then either not retreated or retreated on day 3 with 2ng/ml TGFβ1 either transiently for 4h or continuously, and cells collected at indicated times. Representatives of 3 independent experiments are shown. See also related Figures S5–S7.

Figure 4. Snail1 transcriptionally represses Twist1 expression

(A) Western blotting for indicated proteins of WCL from MCF10A cells transduced with a lentivirus expressing either the vector pBABE or Snail1. (B) Schematic of the human Twist1 promoter showing 13 consensus E-boxes grouped into 3 main clusters A to C. Green box: minimal Twist1 promoter; Red box: E-box; ATG: Transcription start site; Numbered blue double arrows: primer sets 1–7. (C) Luciferase reporter assay in HEK293T cells using mutants (P1-P4) of the human Twist1 promoter fused with the luciferase reporter plasmid pGL2 (left), and co-transfected with either a pCMV-3Flag-Snail1 plasmid or the empty vector (right). hEcad: Human E-cadherin promoter used as a positive control for Snail1-dependent repression. Fold changes in luciferase activity over the empty pGL2 vector are shown. Each experiment was performed in triplicate. Error bars represent means +/− s.e.m. Representatives of 3 independent experiments are shown. *, p < 0.05, Student’s T-test. (D) Upper: Chromatin immunoprecipitation using a Snail1 monoclonal antibody in nuclear extracts from MCF10A cells either untreated or treated with 2ng/ml TGFβ1 for 2 days. Immunoprecipitated chromatins were subjected with PCR using primer sets 1 to 7 as shown in (B). Lower: PCR using primer sets 1 to 7 of total chromatin inputs. Representatives of 3 experiments are shown. See also related Figure S8.

Figure 5. The Twist1:Snail1 ratio in bone marrow micrometastases is highly predictive for distant recurrences in human breast cancer

Total RNAs were isolated from paired samples of primary tumors and BM DTCs (micrometastases) and subjected to microarray analyses. Quantitative RT-PCR (qRT-PCR) for Snail1, Twist1, and Snail2 (internal control) mRNA was used to validate DTC data, which was normalized against Gapdh mRNA. See also related Figure S9. (A) Snail1 array signals in primary tumors showed a trend toward correlation with distant recurrences (p= 0.056). (B) Normalized Snail1 mRNA in BM DTCs did not correlate with distant recurrences. (C) Twist1 array signals in primary tumors did not correlate with distant recurrences. (D) Normalized Twist1 in BM DTCs correlates with distant recurrences (p= 0.008). (E) Twist1:Snail1 ratio in primary tumors does not correlate with distant recurrences. (F) Twist1:Snail1 ratio in BM DTCs highly correlates with distant recurrences (p=0.0001).

Figure 6. Twist1 induces growth arrest and is associated with a low ERK:p38 activity ratio

(A) Cell counts in triplicate in a 24-well plate at indicated times from seeding of MCF10A cells transduced with a lentivirus expressing either the control luciferase, Twist1, or Twist2 shRNA, then treated with 2 ng/ml TGFβ1 continuously or transiently for 4h, and cell numbers determined at indicated times. Error bars represent mean +/− s.e.m. Representatives of 3 independent experiments are shown. (B and C) Western blotting for indicated proteins of WCL (B) from MCF10A cells similarly manipulated as in (A). The ratio of p-ERK1/2 normalized to total ERK1/2 over p-p38 normalized to total p38 was determined by densitometry (B) and shown as percent of the ratio in untreated control cells (C). Error bars represent mean +/− s.e.m. from 3 independent experiments. Student’s T-test, * p<0.05 for each comparison between Twist1 shRNA and each other experimental condition. (D) Schematic of a potential Twist1-p38-ERK signal feedback loop. Directions where direct evidence is provided are represented as solid lines. The p38 to Twist1 direction is proposed and shown as broken arrow. The p38 to ERK1/2 inhibitory crosstalk has been reported elsewhere. See also related Figure S10–S12.

Figure 7. Model of the temporal and spatial cooperation of Snail1 and Twist1 during cancer EMT and the metastatic cascade

In response to a transient EMT stimulating signal (e.g. TGFβ1), Snail1 level rises briefly to initiate EMT in cancer cells near or at the invasive, leading edge of a primary tumor. As metastatic cells migrate away from the primary tumor, they lose Snail1 expression. In contrary, Twist1, present throughout the primary tumor, first experiences transient repression by Snail1 at the invasive edge during EMT initiation, then begins to rise again in late stages of EMT to maintain cancer EMT, and possibly dormant micrometastases. One potential scenario for the emergence of macrometastases is the downregulation of Twist1 expression in dormant micrometastases, thus relieving growth arrest associated with Twist1 and EMT maintenance.