Doxycycline-induced exogenous Bmi-1 expression enhances tumor formation in a murine model of oral squamous cell carcinoma

ABSTRACT

B Cell-Specific Moloney Murine Leukemia Virus Integration Site 1 (Bmi-1, Bmi1), an epigenetic protein, is necessary for normal stem cell self-renewal in adult animals and for cancer stem cell (CSC) functions in adult animals. To elucidate the functions of Bmi-1 in the oral cavity we created a transgenic mouse line (KrTBmi-1) that expresses ectopic, Flag-tagged Bmi-1 in tongue basal epithelial stem cells only upon doxycycline (DOX) treatment. Genome wide transcriptomics and Ingenuity Pathway Analysis identified several pathways altered by exogenous Bmi-1 expression in the normal tongue epithelium, including EIF2 signaling (P value = 1.58 x 10⁻⁴⁹), mTOR signaling (P value = 2.45 x 10⁻¹²), oxidative phosphorylation (P = 6.61 x 10⁻³) and glutathione redox reactions I (P = 1.74 x 10⁻²). Overall, our data indicate that ectopic Bmi-1 expression has an impact on normal tongue epithelial homeostasis. We then assessed the KrTBmi-1 mice in the 4-nitroquinoline 1-oxide (4-NQO) model of oral carcinogenesis. We found that 80% of mice expressing exogenous Bmi-1 (+DOX, +4-NQO KrTBmi-1; N = 10) developed tumors classified as grade 3 or higher, compared to 60% and 40% of mice expressing just endogenous Bmi-1 (+DOX, +4-NQO Kr and -DOX, +4-NQO KrTBmi-1 groups, respectively; N = 10/group; P value = <0.0001); and 30% of mice expressing ectopic Bmi-1 mice developed 20 or more lesions compared to 10% of mice expressing only endogenous Bmi-1 (P = .009). This demonstrates that exogenous Bmi-1 expression increases the susceptibility of mice to 4-NQO-induced oral carcinogenesis, strengthening the evidence for Bmi-1 as a therapeutic target in human oral squamous cell carcinoma.

1. Introduction

Throughout 2018, the American Cancer Society estimates 33, 950 new oral cancer diagnoses, comprising ~2% of the total new cancer cases within the United States.¹ Unfortunately, the majority of these diagnoses will be advanced stage oral squamous cell carcinoma (OSCC).² Advanced OSCC requires a rigorous combination therapy involving surgery, chemotherapy and radiation.² Despite this, the 5-year survival rates for advanced OSCC remain poor.³ Consequently, we aimed to identify and characterize a novel target for OSCC therapy.

Bmi-1 is a member of the polycomb group (PcG) family of proteins, a structurally diverse set of proteins which assembles into nuclear complexes.^4,5 PcG proteins regulate a number of biological processes, including stem cell maintenance and cancer development.^4,5 Bmi-1 has been shown to have multiple functions. Bmi-1 is a member of polycomb repressive complex 1, which monoubiquitinates lysine 119 on histone H2A, a mark commonly associated with transcriptional repression.^4,5 Other studies have demonstrated a polycomb-independent function for Bmi-1. In prostate cancer, Bmi-1 was found to competitively bind the androgen receptor, blocking its subsequent ubiquitination and degradation.⁶ In endometrial cells, Bmi-1 binds with the progesterone receptor (PR) and the E3 Ligase E6AP to induce PR ubiquitination and activation.⁷ Bmi-1 also regulates mitochondrial function in cells by associating with polynucleotide phosphorylase and inhibiting mitochondrial RNA degradation.⁸

During development, Bmi-1 deficiency in mice, although not lethal to the embryo, prompts neurological and hematological abnormalities (ie. sporadic seizures and reduced hematopoietic cell counts).⁹ These abnormalities result partially from defects in tissue-specific stem cell self-renewal in Bmi-1 -/- mice, causing a reduction in these adult stem cells (ASCs) postnatally.¹⁰ Such results highlight Bmi-1 as vital for maintenance of adult stem cells, the population of cells that is important for tissue homeostasis and regeneration.¹¹ In the tongue epithelium, lineage tracing of a subset of Bmi-1-positive cells revealed that the Bmi-1 promoter is active in the long-term stem cells.^12,13 Furthermore, depletion of Bmi-1-positive cells compromises repair of the tongue epithelium after radiation damage, demonstrating the importance of Bmi-1-positive cells for tissue regeneration, a feature of adult stem cells.¹² These results suggest that Bmi-1 is expressed in stem cells of the tongue epithelium and that Bmi-1 is needed for stem cell maintenance.

Bmi-1 is often overexpressed in a variety of human cancers, including breast, head and neck, and pancreatic cancer.¹⁴ Because of this, there have been more studies of Bmi-1 than of other Polycomb 1 proteins in the context of cancer. Cell heterogeneity within a variety of tumors has led to the formulation of the cancer stem cell (CSC) model. CSCs are a small, rare subpopulation of tumorigenic, self-renewing cancer cells that are capable of forming a tumor and that often drive the carcinogenesis process. CSCs also exhibit chemo-resistance because many CSCs cycle slowly.¹⁵ Bmi-1 plays a key role in maintaining the cancer stem cell (CSC) phenotype in different cancers, including head and neck squamous cell carcinoma (HNSCC).¹⁶ In HNSCC, Bmi-1 is elevated within CSC-enriched populations, and knockdown of Bmi-1 increases the CSC-enriched cell population’s sensitivity to chemotherapeutics and radiation while diminishing its invasiveness and tumor development in immune-compromised mice.^16-18 In a commonly used murine model of oral carcinogenesis,¹⁹ lineage tracing studies show that Bmi-1-positive cells constitute a small subpopulation of tongue tumor cells which divide to populate a large portion of the tumor area over time.²⁰ These Bmi-1-positive tumor cells are also highly tumorigenic and metastatic, while Bmi-1-negative cells are not, suggesting that endogenous Bmi-1 is highly expressed in the CSCs of these tumors.²⁰ Exactly how high Bmi1 expression drives tumorigenesis and the importance of ubiquitination in these actions of Bmi1 are unclear.

Here, we determined the effects of ectopic, doxycycline-inducible Bmi-1 overexpression in the normal tongue epithelium. Second, we assessed how Bmi-1 overexpression affected the carcinogenesis process in our 4-nitroquinoline 1-oxide (4-NQO)-induced mouse model of oral carcinogenesis.¹⁹ To accomplish these aims, we generated transgenic mice that, when treated with doxycycline, overexpress ectopic Bmi-1 protein within the stem cells of the tongue epithelium. 4-NQO is an ideal carcinogen because its mechanisms of action are comparable to carcinogens from cigarette smoke, a risk factor of human oral cancer. Importantly, the lesions induced in this 4-NQO model are histologically similar and also acquire similar molecular changes to those observed in human OSCC.^19,21 Given the histological and molecular similarities between tumors induced within this 4-NQO model and human OSCC, we used this 4-NQO model to delineate the role of Bmi-1 in the development of human OSCC.

2. Methods

2.1. K14-rtTA;TRE-FLBmi-1 mice

To generate K14-rtTA;TREFLBmi-1 mice, we first cloned pSPORT1 murine Bmi-1 cDNA (Accession No.: BC053708) plasmid from Invitrogen into the pTRE2-hygromycin plasmid (Clontech; #631014) using the BamHI and SalI cloning sites (Supplementary Methods). The XhoI/SapI digestion fragment containing TRE-FLAGBmi-1 (2733 bps) was purified and injected pronuclearly into fertilized murine C57/Bl6 eggs by the NYU Rodent Genetic Engineering Laboratory (Supplementary Figure 1). Pronuclear injection yielded one C57/Bl6 TRE-FLBmi-1- positive male mouse. To generate the K14-rtTA; TREFLBmi-1 mouse line, we mated homozygous K14-rtTA mice (The Jackson Laboratory; #008099) with heterozygous C57/Bl6 TRE-FLBmi-1 mice. Such breeding can yield pups with K14-rtTA (Kr) or K14-rtTA; TREFLBmi-1 (KrTBmi-1) genotypes. Animal care and use in this research was approved by the Institutional Animal Care and Use Committee (IACUC) at Weill Cornell Medical College, IACUC protocol # 0710-677A, and conformed to the US the National Institutes of Health guidelines for the humane care and use of Laboratory animals.

2.2. Drug administration

To induce Bmi-1 overexpression we treated Kr and KrTBmi-1 mice with 2 mg/mL doxycycline hyclate (Sigma Aldrich; D9891) added to the drinking water and replenished weekly. For tumor development studies, we treated mice ± 2 mg/mL doxycycline and ±100 μg/mL 4-NQO¹⁹ supplied in the drinking water (N = 5-10/condition) and treatments were replenished weekly. Doxycycline treatment was continued until the time of sacrifice.

2.3. Histological analysis of tongue lesions

For the tumor development studies we harvested tongue tissue immediately following cervical dislocation. We photographed whole tongues (10X magnification) to document lesions and to count and grade the lesions along the tongue surface. Lesion grades ranged from 0 (no lesions) to 5 (most severe).²² Tongue lesions were graded and counted in a blinded manner. For pathological classification, a portion of tongue tissue was fixed in 4% paraformaldehyde (PFA; Sigma; #P6148) overnight. Tissues were processed and sectioned at 5–7 μm in the Weill Cornell Electron Microscopy and Histology Core. Hematoxylin and eosin (H&E) staining was performed for histological analysis of tongue tumors. Dr. Theresa Scognamiglio, a board-certified head and neck cancer pathologist, analyzed the H&E stained tongue sections in a blinded manner and categorized tongue lesions as: hyperplasia, mild dysplasia, moderate dysplasia, severe dysplasia, squamous cell carcinoma (SCC) or invasive SCC.^19,23

2.4. Immunohistochemistry (IHC) and immunofluorescence (IF)

We harvested tissues for IHC and IF after cervical dislocation and tissues were fixed in 4% PFA overnight and subsequently processed and sectioned as for the tumor development studies. We performed immunohistochemical staining as described previously.^22,24 For immunofluorescence staining, we performed deparaffinization, rehydration and antigen retrieval on paraffin-embedded sections as outlined in the immunohistochemistry methods.^22,24 We blocked sections with 2% Bovine Serum Albumin (BSA; Sigma; A7030), 0.3 M Glycine (Sigma; G7126) and 0.1% triton-x100 (Sigma; T9284). Tissue sections were incubated with primary antibodies overnight in 2% BSA with 0.05% triton-x100 at 4°C, followed by incubating with a secondary fluorescent antibody (Table 1). We stained nuclei with Hoechst and mounted slides with 70% glycerol. Table 1 includes a list of antibodies and conditions used in IF and IHC. Immunohistochemical staining was quantified with the ImageJ software (National Institutes of Health) and further details are given in the supplemental methods.

Table 2.

List of primers.

Primer	Purpose	Sequence	Size (basepairs)
MluI_BamHI_Bmi 1FLAGF	Cloning	ACGCGGGATCCGCCACCATGGACT ACAAAGACGATGACGATAAAATGC ATCGAACAACCAGAATCAAGA	N/A
ClaI_Sall_Bmi1R	Cloning	ACGCGTGTCGACTAACCAGATGAA GTTGCTGATG	N/A
Bmi1_1F (Bmi-1 TG)	RT-PCR	GTCAGCTGATGCTGCCAATG	884
bglobinPA_771R (Bmi-1 TG)	RT-PCR	TGCTCAAGGGGCTTCATGATGTCC	884
m36B4(+)A (F)	RT-PCR	AGAACAACCCAGCTCTGGAGAAA	448
m36B4(-)B (R)	RT-PCR	ACACCCTCCAGAAAGCGAGAGT	448

Table 1.

Antibody List for IF and IHC.

Target	Company	Catalog #	Lot #	Dilution	Technique
Bmi-1	Santa Cruz	sc-390443	I0116	1:50D	IHC
Bmi-1	Cell Signaling	6964S	3	1:50D	IHC/IF
FLAG Tag	Cell Signaling	14793S	4	1:50D	IHC
anti-rabbit IgG, AlexaFluor 488	Molecular probes	A11070	99D1-1	1:500D	IF
anti-rabbit IgG, AlexaFluor 488	Invitrogen	A21206	1874771	1:500D	IF

2.5. Reverse transcription-PCR (RT-PCR)

We homogenized tongue tissue in TriReagent (Sigma; #T9424) to isolate the RNA and utilized 1μg for reverse transcription-PCR as described previously.²⁵ We ran PCR products on agarose gels and quantified the products using the Gene Tools (Syngene) software (Table 2).

2.6. Genome-wide RNA sequencing analysis (RNA-seq)

Six-week old Kr and KrTBmi-1 (N = 10) mice were treated with 2 mg/mL doxycycline for 1 week. Note that the Kr mice, even when treated with doxycycline, express only endogenous Bmi-1, while the KrTBmi-1 mice express both endogenous and exogenous Bmi-1. Tongue tissue was harvested and stored in RNA Later at −70°C. We extracted RNA with the RNAeasy plus mini kit (Qiagen; #74134). RNA Quality Control (QC) and RNA seq analysis and statistics were performed by the Weill Cornell Genomics and Epigenomics Core (Supplementary Methods). One KrTBmi-1 sample was excluded because of low ectopic Bmi-1 overexpression. We performed RNA-seq analysis on tongue tissues and used a criteria of absolute fold-change of >1.1 for altered mRNAs with statistical significance to reduce the impact from the background expression of genes from other cell types in the tongue (ie. muscle). The ratio between epithelial cells and other cells were similar among the samples. Differentially expressed genes were categorized into signaling pathways using the Ingenuity Pathway Analysis (IPA) interface and a heatmap of selected genes was generated with the NetworkAnalyst online platform.²⁶ IPA predicts directional changes in pathways based on z-score, a statistical calculation of the match between observed transcript changes and the expected transcript changes in the activated state. A z-score <0 predicts inhibition and a z-score >0 predicts activation. Z-scores <-2 and >2 are deemed significant.

2.7. Statistics

Within the tumor development experiments, we analyzed statistical significance of overall tumor grade and lesion number in various groups by using one-way ANOVA analysis with the Bonferroni correction method. Statistical significance of the distribution of the tumor grades, lesion numbers, and lesion pathology among the groups was analyzed using a chi-square test. We assessed the statistical significance for endogenous Bmi-1 protein expression using a one-way ANOVA analysis with the Bonferroni correction method. We set p-values for all statistical significance below 0.05.

3. Results

3.1. Ectopic Bmi-1 expression is doxycycline-inducible in the tongue epithelia of Transgenic (TG) K14rtTA;TREFLBmi-1 mice

We generated a transgenic K14-rtTA;TREFLBmi-1 mouse line, labeled KrTBmi-1, in which the reverse tetracycline-controlled transactivator (rtTA) expression is controlled by a truncated form of the human keratin 14 (K14) promoter which is specifically active in the basal layer cells of the stratified squamous epithelia lining the tongue, skin and esophagus.²⁷ In the presence of doxycycline, rtTA binds to the tetracycline response element (TRE) and drives the expression of the downstream Flag-Bmi-1 gene (Figure 1(a)). We incorporated an amino-terminal flag tag into this construct to distinguish ectopic from endogenous Bmi-1. Flag-tagged Bmi-1 is ectopically expressed in cells in vitro²⁸ and in vivo²⁹ and binds with other polycomb repressive complex 1 members, such as Ring1A/B.²⁸ Mice positive only for the K14-rtTA transgene, labeled Kr, served as a negative control.

Figure 1.

Doxycycline-Inducible Bmi-1 Transgene Expression in Epithelial Tissues of K14- rtTA;TREFLBmi-1 (KrTBmi-1) Mice.

(a) Diagram Depicting the TET-ON Doxycycline-Inducible System in KrTBmi-1 Transgenic Mice. (b) Bmi-1 Transgene (Bmi-1 TG) mRNA expression using reverse transcriptase PCR (RT-PCR) in the tongue tissue of K14-rtTA only (Kr) and KrTBmi-1 mice treated with or without a 1 week doxycycline treatment (-/+DOX; N = 4 mice/group). mRNA levels were detected using reverse transcriptase PCR (RT-PCR). (c) Bmi-1 protein expression in the dorsal tongue epithelium of -/+DOX Kr mice (a,b) and -/+DOX KrTBmi-1 mice (c, d) (300X; 33µm scale bar). (d) Representative staining for FLAG in the tongue epithelium from +DOX Kr mice (panels a, b) and +DOX KrTBmi-1 mice (panels c, d) (200X; 50µm scale bar; N = 3mice/group; 5 fields/mouse). Cells expressing ectopic Flagged-Bmi-1 are indicated by arrows. (e) Bmi-1 protein expression in the tongue (300X; 33µm scale bar; a-d), skin (600X; 20µm scale bar; e-h) and liver (200X; 50µm scale bar; i-l) of +DOX Kr and +DOX KrTBmi-1 mice. Bmi-1 protein expression was analyzed from N = 3+/group across 3+ fields per mouse (panels C, D). Dashed lines and arrows point out representative regions and cells with ectopic Bmi-1 overexpression, respectively.

After 1 week of doxycycline treatment (+DOX), we observed increased Bmi-1 expression within the tongue epithelium of only the +DOX KrTBmi-1 mice at both the mRNA (Figure 1(b)) and protein (Figure 1(c)) level, confirming the inducibility of the ectopic FLAGBmi-1 construct. Using reverse transcriptase PCR, the ectopic Bmi-1 transgene transcript is only detectable in +DOX KrTBmi-1 mice, demonstrating that this TET-ON system is not leaky in the absence of dox (-DOX) (Figure 1(b)). Moreover, tongue epithelial cells in the +DOX KrTBmi-1 mice had increased Bmi-1 protein expression (dashed outlines), suggesting that the ectopic Bmi-1 transgene transcript is translated (Figure 1(c)). Furthermore, increased Bmi-1 protein was absent in +DOX Kr mice, demonstrating that the Bmi-1 induction in +DOX KrTBmi-1 mice does not result from the induction of endogenous Bmi-1 by doxycycline (Figure 1(c)). To ensure that increased Bmi-1 expression results from the induction of exogenous, Flagged Bmi-1, we examined the expression of the exogenous Bmi-1 protein in the tongue epithelium of +DOX Kr and +DOX KrTBmi-1 mice by staining with a FLAG antibody. We observed staining of a subset of tongue epithelial cells only in +DOX KrTBmi-1 mice vs. +DOX Kr mice, suggesting that Bmi-1 induction results from exogenous Flagged-Bmi-1 protein (Figure 1(d); panels c, d vs. a, b).

We next tested the tissue specificity of this TET-ON system. +DOX KrTBmi-1 mice showed a subset of cells in the tongue epithelium (Figure 1(e); dashed outline; panels a-d) and skin epithelium (Figure 1(e); black arrows; panels e-h) with increased Bmi-1 protein. This staining was absent in +DOX Kr mice. However, Bmi-1 expression in the liver was similar in +DOX Kr and +DOX KrTBmi-1 mice (Figure 1(e); panels i-l), suggesting that the TET-ON system is oral and skin epithelial tissue specific.

3.2. Ectopically expressed Bmi-1 protein functions in normal tongue epithelium as assessed by RNA-seq analysis

To test the functionality of the ectopically expressed Bmi-1 protein in the normal tongue epithelium, we treated Kr (N = 5) and KrTBmi-1 (N = 5) mice (6-weeks-old) with DOX for 1 week and then performed genome-wide RNA-sequencing analyses on the tongue tissue from these mice. We performed Ingenuity Pathway Analysis (IPA) on all altered mRNAs with statistical significance (p-value < 0.05) and had an absolute fold change of >1.1 to include the majority of differentially expressed mRNAs in the tongue epithelium between these two groups. Bmi-1 transcript levels in +DOX KrTBmi-1 mice were increased compared to +DOX Kr mice, demonstrating the induction of ectopic Bmi-1 expression within these mice (Figure 2(b); Bmi-1). We confirmed the increased Bmi-1 protein expression in the tongue epithelia of +DOX KrTBmi-1 mice included within the RNA-seq analysis by immunofluorescence (Figure 2(c)).

Figure 2.

Functional Analysis of Ectopic Bmi-1 Overexpression in the Normal Tongue Epithelium using RNA-seq Analysis.

(a) Pathway Analysis of genes differentially expressed in tongue tissue expressing ectopic Bmi- 1 (N = 4) versus control tongue tissue (N = 5) using Ingenuity Pathway Analysis (IPA). Color- coding indicates IPA’s predicted outcome on pathway activity with inhibition represented by blue (negative z-score) and activation represented as orange (positive z-score). Ratio denotes the number of genes altered out of total genes in the pathway. (b) Heatmap of selected genes differentially expressed in tongue tissue of +DOX KrTBmi-1 (N = 4) and = DOX Kr (N = 5) mice. Red denotes genes with higher relative expression and blue denotes genes with lower relative expression. Heatmap was created with the NetworkAnalyst interface. (c) Representative immunofluorescence of Bmi-1 staining in the tongue epithelia of Dox-treated Kr and Dox-treated KrTBmi-1 mice.

3.2.1. Ingenuity pathway analysis predicts inhibition of eukaryotic initiation factor 2 (EIF2) signaling

EIF2 signaling was the most statistically significant changed pathway in tongues with ectopic Bmi-1 overexpression (Figure 2(a); p-value = 1.58 x 10⁻⁴⁹). Ingenuity pathway analysis (IPA) predicted the inhibition of EIF2 signaling with a z-score of −6.1 (see Methods). EIF2 is a key, positive regulator of global protein synthesis.³⁰ Under normal conditions, EIF2 forms an initiation complex with EIF2B to promote protein synthesis.^30,31 In response to cellular stressors, the EIF2α subunit is phosphorylated and becomes a competitive inhibitor of EIF2B.^30,31 EIF2α phosphorylation enhances the translation of activating transcription factor 4 (ATF4), which induces the expression of NADPH oxidase 4 (NOX4).³¹ NOX4 inhibits threonine protein phosphatase 1 (PP1) activity, blocking de-phosporylation of EIF2α and further amplifying ATF4 expression.³¹

EIF2 signaling inhibition in tongues expressing high ectopic Bmi-1 is highlighted by differential mRNA levels of some of the key regulators of this pathway. For example, we identified a reduction in transcripts of a subunit of EIF2’s binding partner (EIF2b5), an increase in NOX4 transcripts, and a reduction in mRNAs encoding multiple ribosomal proteins (Rpl3, Rplp0 and Rps2) (Figure 2(b); EIF2 signaling). Transcript levels of the transcription factor ATF4 was not significantly altered; differences may be masked by high ATF4 mRNA expression in other cell types within the tongue tissue (ie. skeletal muscle).

3.2.2. Alteration of mTOR signaling and regulation of eIF4 and p70s6K are shown by ingenuity pathway analysis

Two other statistically significant altered pathways upon ectopic Bmi-1 overexpression were mTOR signaling (p-value = 2.45 x 10⁻¹²; z-score = −1) and the regulation of eIF4 and p70S6K (p-value = 5.97 x 10⁻¹⁹; z-score = −1.34) (Figure 2(a)). Alterations in both pathways were expected, as the mTORC1 complex has been shown to positively regulate the eukaryotic initiation factor, eIF4E, which is involved in protein synthesis progression.³² Alteration of mTOR signaling and the downstream regulation of eIF4 and p70S6K are indicated by the reductions in mRNAs of a number of eIF4E targets, including Myc,³³ Hras1³⁴ and CDK4³⁴ (Figure 2(b); eIF4E targets). General inhibition of protein synthesis, possibly through the suppression of both of EIF2 and mTOR signaling, is further supported by a reduction in a number of transcripts coding for proteins of the 40S ribosomal subunit, such as Rps3, Rps6, Rps28, and Rps4x (Figure 2(b)).³⁵

3.2.3. Ingenuity pathway analysis predicts inhibition of oxidative phosphorylation and redox signaling pathways in tongues with ectopic Bmi-1 expression

Mitochondrial oxidative phosphorylation is the main source of cellular energy, adenosine triphosphate (ATP).³⁶ Oxidative phosphorylation requires the function of the electron transport chain which consists of five protein complexes.³⁶ In tongue tissue with ectopic Bmi-1 expression, ingenuity pathway analysis (IPA) predicted an inhibition of oxidative phosphorylation (p-value = 6.61 x 10⁻³; z-score = −3.32), based upon reduced transcript levels of various components of mitochondrial complex I (NDUFA13, NDUFA6, and NDUFA7), mitochondrial complex III (UQCR11 and UQCRQ), mitochondrial complex IV (COX5A, COX6A1 and COX8A) and mitochondrial complex V (ATP5G2) (Figure 2(a, b).

Components of the electron transport chain can contribute to reactive oxygen species (ROS) production, generating superoxide anions that are quickly converted to hydrogen peroxide (H₂O₂).³⁶ Therefore, alterations in oxidative phosphorylation may influence cellular ROS levels, leading to changes in pathways for conserving proper redox levels. In mitochondria, levels of the reactive oxygen species H₂O₂ are lowered by peroxiredoxins 3 and 5 (Prx3 and Prx5), thioredoxin-2 (Trx2) and glutathione peroxidase-1 (Gpx1).³⁶ IPA predicted inhibition of the following redox pathways, the glutathione redox reactions I (p-value = 1.74 x 10⁻²; z-score = −2), NRF2-mediated oxidative stress response (p-value = 2.29 x 10⁻²; z-score = −3) and glutathione-mediated detoxification (p-value = 4.08 x 10⁻²; z-score = −2) (Figure 2(a)). Altered gene expression among these different pathways included reductions in transcripts for the same enzyme families previously mentioned to lower H₂O₂ within the mitochondria, namely Gpx2, Prdx1, and Prdx6 (Figure 2(b); Redox Signaling).

3.2.4. Tongues with ectopic Bmi-1 expression show increased expression of ABC transporter transcripts

Tongue tissue with ectopic Bmi-1 expression showed elevated transcripts multiple ABC transporters, including Abca1, Abca2, Abca3 and Abcc10 (Figure 2(b); ABC Transporters). The ATP-binding cassette (ABC) transporters are a family of proteins that harness ATP hydrolysis as a means to transport a substrate against its concentration gradient.³⁷ Mammalian ABC transporters have the ability to export a wide variety of compounds, including amino acids, lipids, and metals.³⁷ For example, in keratinocytes, cholesterol homeostasis is a tightly regulated process.³⁸ One ABC transporter increased upon ectopic Bmi-1 expression, ABCA1, is the main mediator of cellular export of cholesterol from keratinocytes.³⁸

3.3. Tumor development in the 4-NQO mouse model of oral carcinogenesis in Kr and KrTBmi-1 transgenic mice

To assess if ectopic Bmi-1 overexpression influences oral cancer development, we treated Kr and KrTBmi-1 mice with: water (−4-NQO, -DOX); 2 mg/ml doxycycline (−4-NQO, +DOX); 100 ug/mL 4NQO (+4-NQO, -DOX); or the combination of 100 ug/mL 4NQO with 2 mg/mL doxycycline (+4-NQO, +DOX) for a period of 10 weeks. Following the termination of 4-NQO treatment, we continued doxycycline treatment. We sacrificed mice 15 weeks post-4-NQO treatment to allow for tongue tumor development (Figure 3(a)).

Figure 3.

Dox-treated K14-rtTA;TREFLBmi-1 Mice Show More Severe Tumor Grades and Highest Lesion Numbers in the 4-NQO Carcinogenesis Model.

(a) Timeline outlining the course of different treatment groups for this experiment. (b) Representative whole tongue images depicting a normal epithelial surface from an untreated mouse and epithelial lesions graded from least (1) to most (5) severe from 4-NQO treated mice (10X magnification; 2 mm scale bar). The arrows highlight tongue lesions and how they progress with increasing grade in size and number. (c, e) Graphs of Tumor grade (c) and lesion number (e) from +4-NQO, +DOX Kr mice (N = 10; green triangles), +4-NQO, -DOX KrTBmi-1 mice (N = 10; blue circles) and +4-NQO, +DOX KrTBmi-1 mice (N = 10; blue triangles). These data are graphed denoting the mean with standard error of the mean (SEM). Statistical significance was determined using one-way ANOVA analysis. (D, F) Graphs of the distribution (%) of tongue lesions by tumor grade (d) and lesion number (f) from +4-NQO, +DOX Kr mice (N = 10), +4- NQO, -DOX KrTBmi-1 mice (N = 10) and +4-NQO, +DOX KrTBmi-1 mice (N = 10). Statistical significance was determined with the chi-square test. * = p-value < 0.05; ** = p-value < 0.01;*** = p-value < 0.001; **** = p-value < 0.0001.

We graded and counted tongue lesions for each mouse in a blinded manner. Representative images of graded tongues are shown (Figure 3(b)). Grade 1 tongues display a few early stage tongue lesions. Grade 2 tongues contain early exophytic papillomas. Tongues above grade 2 are assigned their grade based upon the sizes and numbers of their tongue lesions (arrows; Figure 3(b)). With respect to the overall average tumor grade and lesion number resulting from 4-NQO treatment, +4-NQO, +DOX KrTBmi-1 mice did not differ from either control group, +4-NQO, +DOX Kr mice or +4-NQO, -DOX KrTBmi-1 mice (Figure 3(c,e)). However, the distribution of the tumor grade and lesion number indicated that the +4-NQO, +DOX KrTBmi-1 mice exhibited a higher percentage of mice (80%) graded at level 3 or above versus +4-NQO, +DOX Kr mice (60%) and +4-NQO, -DOX KrTBmi-1 mice (40%) (Figure 3(d); p = <0.0001 in both comparisons). Additionally, +4-NQO, +DOX KrTBmi-1 mice showed the highest percentage of mice (30%) with 20 or more lesions compared to +4-NQO, +DOX Kr mice (10%) and +4-NQO, -DOX KrTBmi-1 mice (10%) (Figure 3(f); p = .0009 and <0.0001, respectively). All mice without 4-NQO treatment (−4-NQO, -DOX KrTBmi-1 mice, −4-NQO, +DOX KrTBmi-1 mice and −4-NQO, +DOX Kr mice) lacked tumor formation, demonstrating that ectopic Bmi-1 overexpression alone is not sufficient to drive oral carcinogenesis (Supplementary Figure 2).

Human OSCC follows an ordered series of histological changes. The tongue epithelium, stratified squamous epithelium, consists of structured layers. During oral carcinogenesis, basal epithelial cells divide more frequently, causing a thickening of the epithelial layer (hyperplasia). Eventually, the epithelium looses its stratified structure, becoming dysplastic and progressing to carcinoma in situ (SCC).^19,39 Invasive SCC occurs when abnormal epithelial cells penetrate the basement membrane into the underlying tissues.^19,39 Tongue lesions in the 4-NQO-induced murine model of oral carcinogenesis share these same pathological stages observed in human OSCC (Figure 4(a)).¹⁹ Since the +4-NQO, +DOX KrTBmi-1 group showed a higher percentage of mice with tumors of grade 3 or higher (Figure 3(d)) and with 20 or more lesions (figure 3(f)), we ascertained if the +4-NQO, +DOX KrTBmi-1 group also developed more progressive tongue lesions, such as SCC and invasive SCC. A certified pathologist, Dr. Theresa Scognamiglio, classified the most severe lesion observed in each mouse (Figure 4(b)). The percentages of +4-NQO, +DOX KrTBmi-1 mice which developed dysplasia (40%), SCC (40%), and invasive SCC (20%) were not different from the +4-NQO, +DOX Kr mice (40% dysplasia, 50% SCC and 10% invasive SCC; p-value = 0.1084) or +4-NQO, -DOX KrTBmi-1 mice (50% dysplasia, 30% SCC and 20% invasive SCC; p-value = 0.2809). However, these data are limited because only two slides of H&E-stained tongue sections were utilized for pathological classification of 4-NQO-treated mice. Sectioning of the entire tongue tissue for pathological classification was not done because we reserved some tissue portions for other assays.

Figure 4.

Pathological Classification and Exogenous Bmi-1 Expression in Tongue Tumors Developed in the 4-NQO-Induced Murine Model of Oral Carcinogenesis.

(a) Representative images (100X magnification; 100µm scale bar) of H&E stained sections for: normal tongue epithelium, dysplasia, squamous cell carcinoma (SCC) and invasive SCC. (b) Distribution of the most severe 4-NQO-induced lesion observed in each mouse by pathological classification. Cohorts include the groups not expressing ectopic Bmi-1 (+4-NQO, +DOX Kr and +4-NQO, -DOX KrTBmi-1) and the group that expresses ectopic Bmi-1 (+4-NQO, +DOX KrTBmi-1 mice) (N = 10 mice/group). (c) Exogenous Bmi-1 expression in the normal tongue epithelium of −4-NQO, +DOX KrTBmi-1 mice (N = 3 mice; 5 views/mouse) compared to exogenous Bmi-1 expression in dysplasia (N = 33), SCC (N = 2) and invasive SCC (N = 3) developed in +4-NQO, +DOX KrTBmi-1 mice (N = 10 mice; 2+ views/mouse). Black arrows indicate Bmi-1 overexpressing cells. Images were taken at 200X with 50µm scale bar.

3.4. Ectopic and endogenous Bmi-1 expression in tongue lesions of Kr and KrTBmi-1 mice

We confirmed the expression of exogenous Bmi-1 protein within -/+4-NQO, +DOX KrTBmi-1 mice (Figure 4(c)). Within the 4-NQO-induced lesions (dysplasias, SCCs, and invasive SCCs), we observed variable Bmi-1 expression in the epithelial cells expressing ectopic Bmi-1 (black arrows; Figures 3 and 4(d)). Furthermore, we noted a striking increase in the number of epithelial cells expressing exogenous Bmi-1 throughout the tongue epithelia of +4-NQO, +DOX KrTBmi-1 mice (N = 10; 2+ fields/mouse) compared to the −4-NQO, +DOX KrTBmi-1 mice (N = 3; 5+ fields/mouse) (Figure 4(c)).

We also examined endogenous Bmi-1 protein levels in the various 4-NQO-induced tongue lesions in +4-NQO, +DOX Kr mice and +4-NQO, -DOX KrTBmi-1 mice which should only express endogenous Bmi-1 (Figure 5(a)). Endogenous Bmi-1 expression was concentrated in the basal layer of the normal tongue epithelium of −4-NQO, +DOX Kr mice and −4-NQO, -DOX KrTBmi-1 mice (Figure 5(a); panels a, e). However, we determined that endogenous Bmi-1 expression was extended into the tumor regions in a subset of dysplasias, SCCs, and invasive SCCs from +4-NQO, -DOX KrTBmi-1 and +4-NQO, +DOX Kr mice (Figure 5(a)). Quantification of endogenous Bmi-1 staining showed no change in the mean Bmi-1 staining intensity (Figure 5(b)), but the percentages of the epithelial areas that stained positive for Bmi-1 in dysplasias and SCCs of +4-NQO, -DOX KrTBmi-1 and +4-NQO, +DOX Kr mice were higher than those in the normal tongue epithelia of −4-NQO, +DOX Kr mice and −4-NQO, -DOX KrTBmi-1 mice (Figure 5(c)). Thus, the expression of both endogenous and exogenous Bmi-1 expanded after 4-NQO treatment.

Figure 5.

Endogenous Bmi-1 Expression in Tongue Tumors Developed in the 4-NQO-Induced Murine Model of Oral Carcinogenesis.

(a) Endogenous Bmi-1 expression in the normal tongue epithelium of −4-NQO, +DOX Kr mice (+DOX Kr) and −4-NQO, -DOX KrTBmi-1 (-DOX KrTBmi-1) mice (N = 6 total; 5 views/mouse; panels a, e) compared to Bmi-1 expression in dysplasia (N = 19), SCC (N = 8) and invasive SCC (N = 2) developed in +4-NQO, +DOX Kr mice and +4-NQO, -DOX KrTBmi-1 mice (N = 13 Total; 2+ views/mouse). Quantification of the mean staining intensity (b) and percentage of the epithelial area positive (c) for endogenous Bmi-1 in the tongue epithelia of −4- NQO, +DOX Kr mice (+DOX Kr) and −4-NQO, -DOX KrTBmi-1 (-DOX KrTBmi-1) mice (N = 6 total; 5 views/mouse; panels a, e) compared to dysplasias (N = 19) and SCCs (N = 8) of +4- NQO, +DOX Kr mice and +4-NQO, -DOX KrTBmi-1 mice (N = 13 Total; 2+ views/mouse). These data are graphed denoting the mean with standard deviation (SD). Statistical significance was determined by one-way ANOVA analysis

4. Discussion

4.1. Exogenous Bmi-1 expression alters pathways involved in protein translation and redox homeostasis in the normal tongue epithelium

We first performed RNA-seq analysis on whole tongue tissue of +DOX Kr and +DOX KrTBmi-1 mice to examine the function of ectopic Bmi-1in the normal tongue epithelium. We discovered that ectopic Bmi-1 expression in the murine tongue alters pathways that are involved in protein synthesis regulation. These pathways include eukaryotic initiation factor 2 (eIF2) signaling, mTOR signaling, and the regulation of eukaryotic initiation factor 4 (eIF4) and p70S6K signaling (Figure 2(a,b)). EIF2, in complex with its binding partner EIF2B, induces mRNA translation.^30,31 Additionally, activation of the mTOR complex, mTORC1, leads to the activation of EIF4 signaling and subsequent progression of protein synthesis.^32,40 The high negative z-score for EIF2 signaling (z-score = −6.1) calculated by ingenuity pathway analysis and reductions in multiple transcripts for ribosomal proteins (Figure 2(b)) suggest that protein synthesis in tongue epithelial cells may be lower upon ectopic Bmi-1 overexpression.

There are few studies of the regulation of protein synthesis by Bmi-1. In proerythroblasts, knockdown of endogenous Bmi-1 lowered multiple ribosomal protein transcript levels, suggesting that Bmi-1 may positively regulate mRNA translation in these cells.⁴¹ These data are not consistent with our RNA-seq data reported here. Potential regulation of EIF2 signaling by Bmi-1 was highlighted by alterations in multiple genes involved in EIF2 signaling regulation with Bmi-1 overexpression in normal human cultured urothelial cells, but whether these genes were increased or decreased was not specified.⁴² In the normal tongue epithelium, Bmi-1 marks long-lived stem cells.^12,13 Neural stem cells, hematopoietic stem cells and skin stem cells exhibit low global protein synthesis.⁴³ Therefore, ectopic Bmi-1 expression may be inducing a stem cell phenotype associated with lowered protein translation.⁴⁴

Ectopic Bmi-1 expression in the normal tongue epithelium also led to reductions in several pathways involved in redox homeostasis, including oxidative phosphorylation and Nrf2-mediated oxidative stress (Figure 2(a)). Bmi-1-deficient thymocytes showed reduced mitochondrial oxygen consumption and elevated ROS levels that may have resulted from disruption in electron transport during the electron transport chain.⁴⁵ Normal human keratinocytes overexpressing Bmi-1 show a reduced induction of ROS upon irradiation.⁴⁶ In Bmi-1-/- thymocytes and normal human keratinocytes, Bmi-1 can regulate ROS through the repression of polycomb target genes implicated in ROS formation, such as lactoperoxidase (Lpo) and arachidonate lipoxygenases (Alox15).^45,46 These reports indicate the ability of Bmi-1 to enhance oxidative phosphorylation and suppress cellular ROS. The reduction of multiple transcripts in antioxidant pathways, such as Nrf2-mediated oxidative stress, suggests ectopic Bmi-1 expression may lower cellular ROS levels within the normal tongue epithelium (Figure 2(a,b)). Interestingly, Bmi-1 overexpression in normal hematopoietic cells did not affect ROS levels,²⁹ suggesting that Bmi-1 regulation of ROS levels could also be cell-type specific and/or context-dependent.

4.2. Mice expressing exogenous Bmi-1 have increased susceptibility to 4-NQO-induced oral carcinogenesis

We examined the effects of exogenous Bmi-1 overexpression on tumor development in the 4-NQO murine model of oral carcinogenesis. We found that the +4-NQO, +DOX KrTBmi-1 mice exhibited the highest percentage of tumors of grade 3 or above and lesions of 20 or more (Figure 3(d,f)). These results demonstrate that the expression of exogenous Bmi-1 makes mice more susceptible to the formation of tongue lesions in response to treatment with the carcinogen, 4-NQO. This conclusion aligns well with previously published literature in which the ablation of Bmi-1-positive cells or the reduction of Bmi-1-positive cells with the Bmi-1 inhibitor, PTC-209, stunted tumor development and metastasis in the same 4-NQO murine model of oral carcinogenesis.²⁰ Additionally, lineage tracing studies demonstrated that cells positive for endogenous Bmi-1 exhibit characteristics of cancer stem cells in tumors developed in the 4-NQO-induced murine model of oral carcinogenesis.²⁰

Of note, +4-NQO, +DOX KrTBmi-1 mice have more epithelial cells which express exogenous Bmi-1 compared to −4-NQO, +DOX KrTBmi-1 mice (Figure 4(c); Dysplasia, SCC, invasive SCC vs. normal). Epithelial cells expressing exogenous Bmi-1 are identified by their intense staining for Bmi-1 versus cells expressing endogenous Bmi-1 (Figure 4(c); arrows). 4-NQO treatment induces apoptosis in keratinocytes in vivo.²⁷ In cultured normal keratinocytes, ectopic Bmi-1 overexpression inhibits apoptosis in response to multiple environmental stressors, thus promoting cell survival.⁴⁷ Therefore, epithelial cells that express exogenous Bmi-1 may be more resistant to 4-NQO-induced apoptosis. Epithelial cells expressing exogenous Bmi-1 could then expand in the tongue epithelium. Furthermore, ectopic Bmi-1 expression in head and neck squamous cell carcinoma cells increases the portion of cells in the S phase of the cell cycle.⁴⁸ Therefore, ectopic Bmi-1 expression may also increase the proliferation of epithelial cells in 4-NQO-treated mice. Either possibility could explain the increased numbers of epithelial cells expressing ectopic Bmi-1 upon 4-NQO treatment that we observed (Figure 4(c)).

Additionally, endogenous Bmi-1 expression expanded in various lesions of +4-NQO, -DOX KrTBmi-1 and +4-NQO, +DOX Kr mice compared to −4-NQO, -DOX KrTBmi-1 and −4-NQO, +DOX Kr mice (Figure 5(a,c)). However, the expansion of Bmi-1 expression was not observed in every lesion (Figure 5(a,c)). These results are in agreement with Bmi-1 expression in human OSCC. Elevation of endogenous Bmi-1 levels in various stages of human OSCC development has been described.⁴⁹ In other studies, differences in endogenous Bmi-1 expression were not seen in human OSCC because of the large variation in Bmi-1 expression in various oral cancer lesions.⁵⁰

Our current RNA-seq analysis highlights some preliminary possibilities as to how ectopic Bmi-1 overexpression may contribute to chemo-resistance. The expression of a number of ABC transporters was increased upon ectopic Bmi-1 overexpression in the normal tongue epithelium (Figure 2(b)). Given the wide variety of compounds, including chemotherapeutic drugs, that ABC transporters are capable of exporting, cancers will often increase expression of such genes as a means for acquiring multi-drug resistance (MDR).⁵¹ For example, Bmi-1-positive cancer stem cells in the 4-NQO-induced murine model have elevated levels of the ABC transporter, Abcg4.²⁰ In normal tongue epithelia that express exogenous Bmi-1, the ABC transporter, ABCA3, transcript is increased (Figure 2(b)). ABCA3 effluxes at least 11 different chemotherapeutic drugs, including cisplatin and paclitaxel.⁵¹ Bmi-1 overexpression also increases the expression of the ABC transporters ABCC1 and ABCG2 in mammary epithelial cells in vitro.⁵² Overall, our findings lend further support for the epigenetic protein, Bmi-1, as a potential therapeutic target in human OSCC and reveal preliminary insights into Bmi-1 function in the normal tongue epithelium.

Funding Statement

This research was financially supported by [R01 DE024394] (LJG), [T32 GM073546] (LJG) and [F31 DE024925] (JMK) and Weill Cornell Funds.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Author contributions

JMK: Conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing.

XT: Collection and/or assembly of data, data analysis and interpretation, final approval of manuscript.

TS: Collection and/or assembly of data

TZ: Collection and/or assembly of data

LJG: Conception and design, data analysis and interpretation, financial support, final approval of manuscript

Data accessibility

RNA-seq data has been uploaded to NCBI (GSE121881) and is embargoed until publication.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.