Paracrine Interaction of Cancer Stem Cell Populations Is Regulated by the Senescence-Associated Secretory Phenotype (SASP)

Dyskeratosis congenita is a telomere DNA damage syndrome characterized by defective telomere maintenance, bone marrow failure, and increased head and neck cancer risk. The Pot1b−/−;Terc+/− mouse exhibits some features of dyskeratosis congenita, but head and neck cancer was not reported in this model. To model the head and neck cancer phenotype, we created unique Pot1b- and p53-null–mutant models which allow genetic lineage tracing of two distinct stem cell populations. Loss of Pot1b expression depleted stem cells via ATR/Chk1/p53 signaling. Tumorigenesis was inhibited in Pot1b−/−;p53+/+ mice due to cellular senescence. Pot1b−/−;p53−/− tumors also exhibited senescence, but proliferated and metastasized with expansion of Lgr6+ stem cells indicative of senescence-associated secretory phenotype. Selective depletion of the small K15+ stem cell fraction resulted in reduction of Lgr6+ cells and inhibition of tumorigenesis via senescence. Gene expression studies revealed that K15+ cancer stem cells regulate Lgr6+ cancer stem cell expansion via chemokine signaling. Genetic ablation of the chemokine receptor Cxcr2 inhibited cancer stem cell expansion and tumorigenesis via senescence. The effects of chemokines were primarily mediated by PI3K signaling, which is a therapeutic target in head and neck cancer. Implications: Paracrine interactions of cancer stem cell populations impact therapeutic options and patient outcomes.


Introduction
Chromosome ends are protected by telomeres that prevent DNA damage response (DDR) and degradation (1). Telomeres form a large duplex loop mediated by single-strand invasion of a G-rich overhang (2). When telomeres become critically short, the DDR is activated at chromosome ends, which induces cellular senescence or apoptosis (3). Cells can stabilize their telomeres and continue proliferation by upregulation of telomerase (4)(5)(6). Given the positive effects of telomerase on telomere length and cellular proliferation, telomerase activity is commonly upregulated in cancer cell lines and primary tumors (7,8).
Cellular senescence was defined as a state of durable cell-cycle arrest (15). However senescent cells are metabolically active and can have both tumor suppressing and tumor-promoting effects on the tissue microenvironment by expression of growth factors and cytokines, which regulate proliferation of neighboring cells. This property of senescent cells is referred to as the senescence-associated secretory phenotype (SASP).
Dyskeratosis congenita is a human telomere DNA damage syndrome characterized by telomere shortening, early bone marrow failure, skin pigmentation, nail dystrophy, and increased risk of head and neck cancer (16,17). It is believed that defective telomere maintenance is the cause of dyskeratosis congenita, although the reasons for specific tissues being more affected are less clear. Stem cell exhaustion due to telomere DDR may result not only in functional tissue failure but also pathologic repair processes such as lung and liver fibrosis observed in some patients with dyskeratosis congenita.
However, increased cancer risk in patients with dyskeratosis congenita is more difficult to explain by the stem cell depletion model. Approximately 80% of patients with dyskeratosis congenita exhibit oral intraepithelial neoplasia, which is one of the diagnostic features of the disease (16,17). Hypothetically mucosal stem cell depletion should decrease cancer risk. Conversely a mucosal stem cell fraction may survive telomere DDR with resultant genomic instability. The Pot1b À/À ;Terc þ/À mouse exhibits some features of dyskeratosis congenita (18), but head and neck cancer was not reported in this model. The shortened lifespan of this mouse due to Terc heterozygosity may explain the absence of the cancer phenotype. In addition, most head and neck cancers exhibit loss of p53 tumor suppressor function (19), which inhibits DDR, cell-cycle arrest, and apoptosis. To address these issues, we created unique Pot1b-and p53-null-mutant models which allow genetic lineage tracing of two distinct stem cell populations in mucosal epithelium. Our results demonstrate that the telomere DDR regulates senescence-associated paracrine interactions between cancer stem cell populations, dramatically affecting tumor progression and metastasis.

FISH, immunofluorescence, and IHC
Fixed mouse skin, mucosa, and tumor tissue was dehydrated in ethanol, cleared in xylene, and embedded in paraffin. Sections were deparaffinized and stained with hematoxylin and eosin. For telomeric FISH, deparaffinized tissue sections or sorted cells were denatured with Cy3-labeled telomeric peptide nucleic acid probe (TTAGGG) 3 in 70% formamide at 80 C for 10 minutes, followed by overnight incubation at room temperature. After washing, sections and cells were blocked with 10% normal serum, and incubated with anti-53BP1 antibody. After washing, sections were incubated with anti-IgG secondary antibody conjugated to AlexaFluor 488. After washing, colocalized DNA damage and telomere signals were visualized by fluorescence microscopy (Zeiss LSM 710 META). In separate experiments, replication protein A (RPA) was localized at telomeres using the same immunofluorescence/FISH protocol. Phospho-ATR, phospho-Chk1, p53, K15, Lgr6, and histone H3K9me3 proteins were localized in mouse mucosa and tumor tissue using the same immunofluorescence protocol. IHC analysis of proliferating cell nuclear antigen (PCNA) and p16 INK4A expression was performed as described previously (20).

ssDNA analysis
Genomic DNA was extracted from SCC and analyzed for singlestrand telomere overhangs as described previously (20).

Telomere length analysis of stem cells
We used a qPCR method to measure average telomere length ratios of sorted GFP þ and GFP À stem cells from mouse mucosa and tumors as described previously (20).

Cell death analysis
Mucosa and tumor tissue sections were incubated with terminal deoxynucleotidyl transferase and dUTP-fluorescein for 1 hour at 37 C according to the manufacturer's recommendations (Roche Applied Sciences). After washing, apoptotic cells were visualized by fluorescence microscopy. The percent fluorescent cells was determined using quantitative image analysis software. Data were analyzed by ANOVA.

FACS
GFP þ and GFP À mucosal epithelial cells and SCCs were dissociated by trypsinization, washed in PBS, and sorted by flow cytometry (Beckman Coulter MoFlo). Data were analyzed by ANOVA.

Western blot analysis
Protein was extracted from mucosa and tumors in 1Â Laemmli buffer. Western blotting using antibodies was performed as described previously (20).

Senescence-associated b-galactosidase activity
Frozen tumor sections were fixed for 5 minutes in phosphate buffered 2% formaldehyde and 0.2% glutaraldehyde at room temperature. Sections were incubated in staining solution containing 1 mg/mL X-gal. Sections were counterstained with nuclear fast red solution.

Antibodies
Novus
We examined signaling pathways downstream of Cxcr2 in control and Cxcr2-null SCC by Western blot analysis. Cxcr2 protein expression was not detected in Cxcr2 null SCC, nor was p53 detected in p53-null samples ( Supplementary Fig. S4). pAkt1 and total Akt1 protein expression was reduced by 2-fold in Cxcr2-null SCC. pERK1 expression was reduced by 2-3-fold, and total ERK1 was reduced by 2-8-fold in Cxcr2 null SCC. pPI3K expression was reduced 2-fold, and total PI3K levels were reduced by 2-4 in Cxcr2-null SCC. pPKC expression was reduced to undetectable levels in Cxcr2-null SCC, as was total PKC levels in some Cxcr2-null samples. pPLCg expression was undetectable in most samples, and total PLCg expression was reduced by up to 32-fold in Cxcr2-null SCC. These results indicate that loss of Cxcr2 expression inhibits multiple downstream signaling pathways in SCC.
To determine which of multiple signaling pathways was most important in regulating chemokine signaling in SCC, we treated Pot1b À/À ;p53 À/À SCC transplanted to NU/J mice separately with three small-molecule inhibitors currently in cancer clinical trials. After 4-weeks of treatment, the PI3K inhibitor buparlisib produced a 64% reduction in tumor volume (P < 0.007; Supplementary Fig. S5A) compared with vehicle-treated control tumors. The MEK inhibitor trametinib resulted in a 39% reduction in tumor volume, and the Akt inhibitor MK2206 created a 23% reduction. Histopathologic sections of SCC treated with buparlisib or vehicle are shown by hematoxylin and eosin staining in Supplementary Fig. S5B and S5C. The proliferating cell fraction was reduced in buparlisib-treated tumors (23% vs. 37%; P < 0.04; Supplementary Fig. S5D and S5E) as determined by PCNA IHC. Buparlisib-treated SCC exhibited increased apoptotic cells (43% vs. 0.4%; P < 0.00002; Supplementary Fig. S5F and S5G) as determined by TUNEL analysis. Buparlisib-treated SCC exhibited increased senescent cells (69% vs. 1.1%; P < 0.004; Supplementary Fig. S5H and S5I) as determined by senescence-associated b-galactosidase activity. The K15 þ cancer stem cell fraction was reduced in buparlisib-treated SCC (0.3% vs. 2%; P < 0.03; Supplementary  Fig. S5J and S5K). The Lgr6 þ cancer stem cell fraction also was reduced in buparlisib-treated SCC (0.2% vs. 17.1%; P < 0.0005; Supplementary Fig. S5L and S5M). These results indicate that PI3K is an important mediator of cancer stem cell chemokine signaling in SCC.

Discussion
An important finding of our study is that telomere DDR can regulate SASP. Pot1b deficiency induced senescence in both p53 þ/þ and p53 À/À backgrounds; however cancer outcomes were markedly different in these contexts. In the p53-deficient background, K15 þ stem cells secreted Cxcl chemokines, which were primarily responsible for Lgr6 þ cell proliferation and tumor progression. K15 þ cells in the p53 þ/þ background failed to secrete chemokines, which resulted in senescence (stable cell-cycle arrest) and markedly inhibited tumor progression. These results suggest that defective telomere DDRs may lead to cancer stem cell expansion and tumor progression. Nonstem glioma cells with telomere DNA damage undergo senescence and secrete proangiogenic factors which increase tumor initiation of glioma stem cells (23). Although senescence can lead to a tumor promoting inflammatory microenvironment, inflammation may be difficult to sustain when bone marrow failure is one of the pathognomonic features of dyskeratosis congenita. However, the SASP secretome is complex and may trigger tumor cell proliferation in the absence of inflammatory signals.
Chemokines mediate numerous functions in tumorigenesis, including proliferation, survival, angiogenesis, epithelial-mesenchymal transition, and metastasis (24). Cxcl1 expression was induced in stratified epithelia treated with phorbol ester (25), and Cxcl1 itself induced epithelial proliferation (26). Cxcl1 and Cxcl3 were expressed in human head and neck squamous cell carcinoma lines and patient tissue (27)(28)(29). Cxcr2 regulated head and neck cancer cell proliferation and migration in cell lines (30). Cxcr2 expression was associated with cervical lymph node metastasis in human head and neck cancer (31). Ras transformation of mouse stratified epithelial cells induced Cxcr2 ligands, and these cells failed to form tumors when Cxcr2 expression was inhibited (32). While paracrine signaling from stromal cells to epithelial tumors has been extensively studied (33), our paper provides a unique example of high level Cxcl secretion by one stem cell population (K15 þ ) which regulates expansion of a second stem cell fraction (Lgr6 þ ) expressing high levels of the receptor for these ligands (Cxcr2). Depletion of the K15 þ stem cell population or deletion of Cxcr2 receptor expression in Lgr6 þ cells dramatically inhibits expansion of the latter population, and blocks tumor growth and metastasis. These studies indicate an important role for Cxcl chemokines and Cxcr2 in human head and neck cancer.
A striking result of our study was marked inhibition of metastasis in Pot1b À/À ;p53 þ/þ SCC. Given low levels of apoptosis in these tumors, the metastatic phenotype is difficult to attribute to increased programmed cell death, and our previous studies indicate that apoptosis in primary SCC does not correlate with metastatic phenotype (20). In contrast, increased senescence and terminal differentiation of SCC provides a likely mechanism for inhibiting metastasis. Reduction in the proliferating cell fraction likely inhibits genomic instability and therefore creation of potentially metastatic clones. This mechanism has therapeutic implications particularly for late evolving metastatic clones.
Another key finding of our study is that Pot1b deficiency does not induce head and neck squamous cell carcinoma, even in the p53 À/À tumor prone background. Our previous studies on TRF2-and Terc-mediated telomere DDR in stratified epithelium did not reveal spontaneous carcinoma formation (20,34). In fact reduced TRF2 expression and to lesser extent Terc deficiency induce a powerful telomere DDR which results in high apoptotic cell fraction, although neither of these mutants exhibit dyskeratosis congenita features. Telomere DNA damage signaling has been associated with chemokine and cytokine expression in other experimental models. Reduced TRF1 expression in endothelial cells correlated with increased expression of CCL2 and GM-CSF (35). IL8 induced telomerase activity and prevented senescence of human endothelial cells (36). Stromal cell-derived factor 1a inhibited endothelial cell senescence via telomerase activation (37). Telomerase was required for TNFa-mediated target gene expression in airway smooth muscle cells (38). Telomerase deficiency was associated with inflammatory chemokine expression and alveolar stem cell senescence in the mouse lung (39). Cxcr2 inhibition was associated with reduced TERT expression and differentiation of human pluripotent stem cells (40). A more limited telomere DDR may be required for dyskeratosis congenita. Alternatively immunosuppression (a feature of dyskeratosis congenita) is a known risk factor for head and neck cancer (41). The increased risk of head and neck cancer in dyskeratosis congenita may therefore be due to progressive bone marrow suppression (42). However, genetically unstable epithelial cells due to dyskeratosis congenita-associated telomere DDR may escape immune surveillance in the context of dyskeratosis congenita-induced bone marrow suppression. A previous study reported that p53 deletion inhibited bone marrow suppression in Pot1b À/À ; Terc þ/À mice (43). Interestingly Pot1b;p16 INK4A deletion accelerated bone marrow failure with increased ATR activation, telomere shortening, chromosomal fusions, telomere replication defects, and p53-dependent apoptosis (44).
The mouse genome has a Pot1a gene whose conditional deletion also results in telomere DNA damage and p53-dependent cellular senescence (45,46). Pot1a depletion at telomeres induces an ATR-dependent DDR similar to that observed in Pot1b null cells (47). Pot1a and Pot1b did not differ in their ability to repress telomere recombination (48). Reduced