USP22 is required for the tumorigenicity of breast cancer stem cells.
USP22 has been suggested as a cancer stem cell gene or death-from cancer signature gene and its high expression often predicts the poor clinical outcomes of cancer patients 7, but its role in maintaining CSC stemness remains to be defined. We sorted CD24−CD44+ breast CSCs from the patient-derived Luc2-eGFP (L2G)-labeled breast triple-negative (TN1) cancer cells 6 as well as in the murine breast cancer 4T1 cells (Figure s1A), and found a significantly higher USP22 expression in breast CSCs compared to that in the CD24+CD44− non-CSCs by western blotting (Fig. 1A & 1B and Figure s1B). To decipher the USP22 functions in generating and/or maintaining breast cancer cell stemness, we generated USP22 targeted deletion in mouse 4T1 and human breast cancer-derived L2G+ TN1 cells by a CRISPR-Cas9 approach. A complete USP22 deletion was validated by immunoblot analysis (Fig. 1C & s1C). Importantly, silencing USP22 dramatically reduced the CD24−CD44+ breast CSCs population in L2G+ TN1 and 4T1 cells (Fig. 1D and s1D & s1E), indicating that USP22 is important for breast CSCs self-renewal. We then utilized a well-established tumor sphere formation assay 21, 22 to evaluate the role of USP22 in breast CSCs self-renewal. Indeed, the tumor sphere formation from both patient derived L2G+ TN1 and mouse 4T1 breast cancer cells was largely impaired by USP22 CRISPR deletion, which was further confirmed by an in vitro extremely limiting dilution assay (Fig. 1E-G, s1F-H). Consistently, USP22 inhibition in 4T1 cells resulted in a substantial reduction in colony formation (Figure s1I & J). Therefore, these results indicate that USP22 is required to maintain an optimal CSC population, possibly by controlling CSC self-renewal in vitro.
CSCs are a critical small population of cancer cells with potent capability for tumor initiation. To test USP22’s function in promoting tumor initiation in vivo. We orthotopically injected 102, 103, and 104 USP22 knockout or control 4T1 breast cancer cells into BALB/c mice. Surprisingly, in contrast to the fact that five out of eight mice implanted with 102 WT 4T1 cells developed cancer three months after implantation, none of the eight mice received with USP22-deficient 4T1 cells developed breast cancer. Even when a higher number of 4T1 cells, 103 and 104, were orthotopically injected, USP22 deletion dramatically inhibited the development of syngeneic tumors (Fig. 1H & 1I), indicating that USP22 is critical for in vivo tumor initiation. Cancer metastases, per prevailing theory, are predominantly initiated by rare cancer cells that bear stem cell properties 23, 24. We then determined whether USP22 exerted a driving role in breast cancer metastasis by intravenously injection of 4T1 USP22-null or its control WT cells into BALB/c mice. As expected, USP22 deletion dramatically inhibited 4T1 cancer colonization to the lung by reducing more than 60% of tumor nodules with further reduced metastatic foci size (Fig. 1J-L). Immunohistochemistry staining confirmed the deletion of USP22 and detected a significant decrease in the levels of stem cell marker CD44 expression in lung metastasis (Fig. 1M & 1N). As a consequence, USP22 ablation significantly improved the overall survival of the mice with 4T1 lung metastasis (Fig. 1O). Collectively, our results revealed that USP22 play an important role in breast CSC maintenance, which is critical for breast cancer initiation and metastasis.
USP22 promotes breast CSC self-renewal through upregulating ITGB1 expression
Integrin family members are known as key regulators in cancer cell stemness, epithelial-mesenchymal transformation and extracellular matrix to initiate the metastatic process for multiple cancer types including breast cancer 19, 25. Importantly, unbiased analysis of the public database TMIER2 26 revealed a strong and statistically significant positive correlation nearly in all types, 37 out of total 40 of human cancers (Fig. 2A and Supplementary table 1). Further analysis of integrin a and b family revealed a positive correlation of several integrin family members in particular integrin a1, 2, 8 & 9 (Figure s2A and Supplementary table 1). These results suggest a possibility that USP22 regulates cancer stem cell self-renewal through regulating the transcription of some of integrin family members. Indeed, comparison analysis by western blotting and flow cytometry of integrin family members between WT and USP22-null breast cancer cells detected a dramatic reduction in integrin b1 expression by USP22 inhibition in both mouse 4T1 and patient-derived L2G+ TN1 cells (Fig. 2B-D). USP22 appears to positively regulate integrin b1 expression at the transcriptional level because its targeted deletion resulted in a more than 70% reduction in ITGB1 mRNA levels (Fig. 2E). In addition to integrin b1, USP22 deletion led to a modest but statistically significant reduction in several additional integrin family members including integrin a1-6 and integrin b2-3, b5-7, but not b4, b8 and a7-8 expression (Figure s2B), In contrast, integrin b6 expression is slightly increased in USP22-null breast cancer cells (Figure s2B). Therefore, USP22 appears to regulate the expression of multiple integrin family members with b1 as the dominant one.
We then focused on studying the functional consequences of USP22-mediated integrin b1 upregulation and assessed whether USP22 maintains breast CSC self-renewal and promotes breast cancer metastasis through integrin b1 upregulation by ectopic reconstitution of ITGB1 in USP22 knockout cells (Figure s2C-E). Indeed, ectopic ITGB1 expression partially rescued tumor sphere formation from in both mouse 4T1 and patient-derived L2G+ TN1 USP22-deficient breast cancer cells (Fig. 2F-H). Consequently, expression of integrin b1 largely, but not totally restored 4T1 breast cancer lung metastasis of USP22-null cells as documented by analyzing both lung tumor nodule numbers and the metastatic foci size (Fig. 2I-K). Collectively, these results demonstrate that USP22 enhances BCSCs tumorigenic potential, in part, through integrin b1 upregulation.
USP22 functions as a de novo FoxM1-specific deubiquitinase in breast cancer cells.
The fact that USP22 deletion reduced ITGB1 mRNA expression suggest that USP22 regulates integrin b1 expression at transcriptional level. Indeed, western blotting analysis revealed a significant reduction in the protein expression of FoxM1, a critical transcription factor for ITGB1 expression 27, in USP22-null breast cancer cells (Fig. 3A). In contrast, USP22 ablation did not alter FoxM1 mRNA levels (Fig. 3B). Together with the fact that USP22 is a deubiquitinase, these results imply that the USP22 exerts its regulatory function on FoxM1 protein expression at the post-transcriptional level. Indeed, treatment of USP22-null cells with the proteasome inhibitor MG132 largely restored FoxM1 expression to a level comparable to that in WT breast cancer cells (Fig. 3A). By contrast, treatment with NH4Cl, an inhibitor of endosome-lysosome degradation pathway, fails to protect FoxM1 from degradation (Figure s3A), suggesting that USP22 promotes FoxM1 level through inhibiting its proteasomal degradation.
As a deubiquitinase, USP22 exerts its biological function largely through protecting its downstream substrates from ubiquitination-mediated degradation 28. Accordingly, we speculated that USP22 could be a deubiquitinase of FoxM1. Indeed, USP22 interaction with FoxM1 was detected in HEK-293T cells transiently transfected Myc-USP22 and Flag-FoxM1, but not in control cells transfected with Flag-FoxM1 or Myc-USP22 alone (Fig. 3C). The endogenous interaction between USP22 and FoxM1 in patient-derived breast cancer L2G+ TN1 cells was further validated (Fig. 3D and s3B). USP22 protein carries an N-terminal zinc finger and C-terminal U19 peptidase catalytic domain (Fig. 3E). Truncated mutation analysis revealed that the zinc finger-containing N-terminus is sufficient for USP22 interaction with FoxM1, while the C-terminus ubiquitin-specific peptidase domain is not involved in mediating its FoxM1 interaction (Fig. 3F). These results indicate that FoxM1 physically interacts with USP22 in breast cancer cells.
A ubiquitin-specific peptidase often inhibits ubiquitination of its interacting proteins 29. Thus, we determined the effect of USP22 on FoxM1 ubiquitination. Higher molecular weight bands were detected in FoxM1 immunoprecipitants, indicating FoxM1 is ubiquitinated possibly by its endogenous E3 ubiquitin ligases such as FBWX7 30. Importantly, transient USP22 expression largely diminished FoxM1 ubiquitination (Fig. 3G). Conversely, loss of USP22 expression resulted in a significant increase in FoxM1 ubiquitination in both mouse 4T1 and patient-derived breast cancer cells (Fig. 3H). Our data indicate physical interaction between USP22 and FoxM1 is required for USP22-mediated suppression of FoxM1 ubiquitination, because mutation of cystines 61 and 63, which disrupts the zinc finger structure and its interaction with FoxM1 (Fig. 3I), totally abolished USP22 activity in suppressing FoxM1 ubiquitination (Fig. 3G). As expected, expression of the catalytically inactive deubiquitinase, through C185A mutation of USP22, failed to inhibit FoxM1 ubiquitination despite not altering its interaction with FoxM1 (Fig. 3F & 3G). These results indicate that USP22 is a bona fide FoxM1-specific deubiquitinase in breast cancer cells. In concordance with this conclusion, USP22 overexpression dramatically prolonged FoxM1 half-life as measured by pulse-chase analysis (Fig. 3J & 3K). Consistent with the ubiquitination data, neither USP22 C185A nor C61/63A mutant sustained FoxM1 stability (Fig. 3J & 3K). In line with this, USP22 ablation dramatically decreased FoxM1 half-life (Fig. 3L & 3M). Consistently, re-expression of WT USP22, but not its mutants restored integrin b1 expression in USP22-null breast cancer cells (Fig. 3N & 3O). These results define USP22 as a de novo FoxM1 deubiquitinase in breast cancer cells to protect FoxM1 from ubiquitination-mediated proteasomal degradation for upregulating integrin b1 expression.
USP22 promotes integrin b1 expression through FoxM1 stabilization.
FoxM1 has been identified as an integrin β1 transcription factor thereby promoting breast cancer progression 27, implying a possibility that USP22 controls breast cancer cell ITGB1 expression through FoxM1 stabilization. Indeed, reconstitution of FoxM1 expression fully restored the endogenous integrin b1 expression in both USP22-null 4T1 and L2G+ TN1 breast cancer cells as determined by western blotting and qRT-PCR (Fig. 4A & 4B), which was further confirmed by flow cytometry (Fig. 4C & 4D). In contrast, we observed that FoxM1 expression fails to rescue integrin b2-7 expression (Figure s4A). These results support our hypothesis that USP22 specifically promote integrin b1 expression through FoxM1 stabilization. Consistent with this, we observed that ectopic expression of FoxM1 largely restored the tumor sphere formation ability of USP22-deficent breast cancer cells (Fig. 4E-G). Likewise, the impaired ability in colony formation of 4T1 breast cancer cells by USP22 depletion was largely rescued by exogenous FoxM1 expression (Figure s4B & s4C).
We also noticed that, while FoxM1 expression fully rescued integrin b1 expression both in USP22-null 4T1 and TN1 breast cancer cells, but their sphere and colony formation were only partially restored by FoxM1 re-expression (Fig. 4E-G). We then utilized the lung metastasis model to further illustrate the role of USP22-FoxM1-integrin b1 pathway in breast cancer tumorigenesis in BALB/c mice. Indeed, in contrast to the fact that USP22 deletion resulted in a more than 50% reduction in lung metastases 4T1 cancer nodules, FoxM1 re-introduction restored USP22-null 4T1 cancer lung metastasis to a level of about 85–90% of the WT (Fig. 4H-J). As a consequence, FoxM1 expression dramatically attenuated but not totally abolished the protection of mice from lung metastasis-induced lethality by USP22 targeted inhibition (Fig. 4K). Collectively, these results indicate that USP22 promotes breast cancer metastasis at least partially, through promoting FoxM1-mediated integrin b1 expression.
Pharmacological inhibition of USP22 abrogates BCSCs tumorigenicity.
Our discovery that genetic USP22 deletion hindered breast cancer stem cell self-renewal and inhibited their lung metastasis provides a rationale for USP22 targeting in anticancer therapy. We first analyzed the effects of pharmacological USP22 inhibition on BCSCs self-renewal using a small molecule inhibitor USP22i-S02 that we recently identified (Fig. 5A) 31. Similar to our observation from USP22 CRISPR KO studies, treatment of breast cancer cells 4T1 and TN1 significantly inhibited both integrin b1 and FoxM1 expression. Consistent with our previous observations, S02 treatment also reduced USP22 expression levels presumably because USP22 is a deubiquitinase of itself (Fig. 5B). Further addition of the proteasomal inhibitor MG132, but not with lysosome inhibitor NH4Cl, largely rescued FoxM1 protein levels from USP22i-S02 treatment (Figure s5A & s5B), confirming our observation that USP22 inhibition facilitates proteasomal FoxM1 protein degradation. In line with this, treatment of 4T1 cells with S02 dramatically shortened FoxM1 protein half-life (Figure s5C & 5D). As expected, S02 treatment suppressed ITGB1 and other stemness related genes expression, including CD44, ALDH, and NANOG (Fig. 5C). In contrast, S02 treatment did not alter FoxM1 mRNA transcription (Figure s5E). These results confirm that USP22 is a positive regulator for FoxM1-mediated ITGB1 expression in breast cancer cells by an orthogonal pharmacological approach.
We next set out to determine the effects of USP22 pharmacological inhibition on breast cancer stem cell self-renewal. As expected, S02 treatment reduced breast CSCs population for more than 80%, to a level that is comparable to USP22 knockout (Fig. 5D and s5F). Importantly, treatment of USP22-null breast cancer cells did not further reduce the frequency of breast cancer stem cells (Fig. 5D and s5F), supporting the high specificity of this USP22-specific small molecule inhibitor. Consequently, treatment with S02 significantly impaired breast cancer cell sphere and colony formation capability (Fig. 5E & F and s5G & H). Further in vitro extremely limiting dilution assay confirmed that S02 inhibited breast CSCs self-renewal (Fig. 5G), implying for its great therapeutic potential in treatment breast cancer. We then used the preclinical 4T1 breast pulmonary metastasis model to illustrate the potential anti-metastatic effect of S02 (Figure s5L). Of note, a six-day treatment with S02 after tail vein injection of 4T1 breast cancer cells resulted in a significant reduction in 4T1 breast cancer lung metastasis and prolonged mice survival (Fig. 5H-K). Further immunohistochemistry analysis of the lung metastatic cancers detected a reduction in both integrin b1 and FoxM1 levels in the S02 treatment groups (Fig. 5L). Similar to our recent study, administration of S02 did not show any detectable toxicity because the mice body weight was unaltered (Figure s5J), and further hematoxylin-eosin (H&E) staining did not detect obvious liver damage in S02 treatment mice (Figure s5K). Therefore, these results indicate that pharmacological USP22 targeting is a safe and effective therapy in treatment of triple negative breast cancers.
We then further evaluated the therapeutic potential of USP22i-S02 in a patient-derive xenograft model by orthotopically implanting TN1 cells to immune compromised RAG1 mutant mice (Fig. 5M). Intriguingly, a 3-day treatment of pre-established PDX tumor significantly hindered patient derived xenograft tumor growth (Fig. 5N & 5O). Further characterization by IHC staining show that the levels of USP22, FoxM1 and integrin b1 protein expression by USP22i-S02 treatment, which consequently inhibited the breast cancer cell growth because the percentage of Ki-67+ proliferative cells was dramatically decreased. Importantly, we detected a significant reduction in CD44+ breast cancer cells in the S02 treat group, implying that USP22 pharmacological inhibition attenuates either the breast cancer stem cell self-renewal or their survival (Fig. 5P). Therefore, pharmacological inhibition of USP22 achieves represents a potentially efficacious treatment for breast cancer and metastasis.
Positive correlation of USP22, FoxM1 and integrin b1 in human breast cancer.
Our data collectively documented that USP22 maintain breast cancer stemness in part through stabilizing ITGB1 transcription factor FoxM1 to promote breast cancer growth and metastasis, which define a previously unknown USP22-FoxM1-ITGB1 pathway in breast cancer pathogenesis. Further analysis of the sorted integrin b1low, integrin b1middle and integrin b1high 4T1 cells revealed a gradual elevation in USP22 and FoxM1 expressions (Fig. 6A & 6B). We then generated a green fluorescent protein (GFP)-USP22 fusion knock-in in 4T1 cells with the endogenous USP22 ablation (Fig. 6C). Consistently, the expression of both FoxM1 and ITGB1 are profoundly increased in USP22high comparing to that in USP22low 4T1 knock-in cells (Fig. 6D & E). Further, a significant increase in integrin b1 and FoxM1 in breast CSCs versus none breast CSCs population was observed (Fig. 6F).
To further determine the critical roles of the USP22-FoxM1-integrin b1 in breast cancer pathogenesis, we utilize the immunohistochemistry staining determined the expression of USP22, FoxM1, and integrin b1 protein in human breast cancer tissue microarray (Supplementary table 2). As expected, the protein levels of USP22, FoxM1, and integrin b1 was markedly higher in the breast tumor tissues than those in begin tumors (Fig. 6G, 6H & s6A-D), and levels were even further elevated in metastatic tissues (Fig. 6G, 6H), further supporting our discovery that upregulated USP22 in breast cancer stem cells though FoxM1-mediated ITGB1 gene transcription for promoting breast cancer lung metastasis. To support this notion, the protein expression levels of USP22, FoxM1 and integrin b1 are strongly correlated in human breast cancers (Fig. 6I & s6E). Collectively, our study identified USP22 as a FoxM1-specific deubiquitinase which promotes FoxM1 transcriptional activation for ITGB1 expression, which consequently promotes breast cancer stem cell self-renewal and drives breast cancer metastasis to distal organs including lung (Fig. 6J).