Adaptor protein p62 promotes skin tumor growth and metastasis and is induced by UVA radiation

Skin cancer is the most common cancer, and exposure to ultraviolet (UV) radiation, namely UVA and UVB, is the major risk factor for skin cancer development. UVA is significantly less effective in causing direct DNA damage than UVB, but UVA has been shown to increase skin cancer risk. The mechanism by which UVA contributes to skin cancer remains unclear. Here, using RNA-Seq, we show that UVA induces autophagy and lysosomal gene expression, including the autophagy receptor and substrate p62. We found that UVA activates transcription factor EB (TFEB), a known regulator of autophagy and lysosomal gene expression, which, in turn, induces p62 transcription. Next, we identified a novel relationship between p62 and cyclooxygenase-2 (COX-2), a prostaglandin synthase critical for skin cancer development. COX-2 expression was up-regulated by UVA-induced p62, suggesting that p62 plays a role in UVA-induced skin cancer. Moreover, we found that p62 stabilizes COX-2 protein through the p62 ubiquitin-associated domain and that p62 regulates prostaglandin E2 production in vitro. In a syngeneic squamous cell carcinoma mouse model, p62 knockdown inhibited tumor growth and metastasis. Furthermore, p62-deficient tumors exhibited reduced immune cell infiltration and increased cell differentiation. Because prostaglandin E2 is known to promote pro-tumorigenic immune cell infiltration, increase proliferation, and inhibit keratinocyte differentiation in vivo, this work suggests that UVA-induced p62 acts through COX-2 to promote skin tumor growth and progression. These findings expand our understanding of UVA-induced skin tumorigenesis and tumor progression and suggest that targeting p62 can help prevent or treat UVA-associated skin cancer.

(UV) radiation, namely UVA (315-400 nm) and UVB (285-315 nm), is the major risk factor for the development of skin cancer. Of these, UVA accounts for ϳ95% of UV in sunlight, and tanning beds emit UVA in doses 12-fold higher than the sun (1). However, UVA is significantly less effective in causing direct DNA damage than UVB, which led many to believe that it was non-tumorigenic. It has since been shown that UVA can induce skin carcinogenesis in vivo (2)(3)(4), and indoor tanning, even intermittently, significantly increases skin cancer risk (5,6). However, the mechanism of UVA's contribution to skin cancer remains unclear.
Macroautophagy (hereafter referred to as autophagy) is a cellular self-eating process that targets unwanted or damaged organelles and proteins to lysosomes for degradation through autophagosomes (7,8). The protein p62, a multidomain protein also known as sequestosome 1 (SQSTM1), acts as an autophagy adaptor and substrate for the selective inclusion of cargo. During autophagy, p62 binds to LC3 in the autophagosomal membrane through the LC3-interacting region, as well as to polyubiquitinated proteins and protein aggregates bound for degradation through the ubiquitin-associated (UBA) 2 domain (9,10). In addition to its role as an autophagy adaptor, p62 forms interactions with a number of proteins to activate protumorigenic signaling pathways. p62 is found to be up-regulated in several human cancers, including lung cancer, breast cancer, melanoma, and skin squamous cell carcinoma (11)(12)(13)(14)(15). Recent studies have demonstrated that p62 promotes tumor formation and progression (16,17) through regulating NF-B (18,19) and NRF2 (20 -22). Furthermore, p62 expression was induced by Ras activation during tumorigenesis (19). Recently, we have found that p62 binds to the oncogenic transcription factor Twist1 and promotes Twist1 stabilization (15). This interaction promotes the epithelial-mesenchymal transition and, hence, skin tumor growth and metastasis (15). Identifying the underlying molecular and cellular mechanisms of p62 regulation and function may elucidate mechanisms key to skin tumorigenesis and tumor progression.

cro ARTICLE
One p62 regulator is a member of the microphthalmia-associated transcription factor family, transcription factor EB (TFEB), a master regulator of autophagy and lysosomal gene expression (23)(24)(25). TFEB binds to the coordinated lysosomal expression and regulation (CLEAR) binding site found in the promoter of many autophagy and lysosomal genes to activate gene transcription and ultimately the degradation of autophagy substrates (23,25,26). TFEB activation is regulated primarily through phosphorylation. Under nutrient-rich conditions, TFEB is primarily cytosolic, phosphorylated, and inactive (27). Upon nutrient deprivation, TFEB rapidly translocates to the nucleus to induce transcription of autophagy and lysosomal genes (27). However, the role of TFEB in UVA response is unknown.
Here, we show that TFEB is a UVA-responsive factor responsible for the activation of autophagy-lysosomal genes, including p62, in keratinocytes following UVA exposure. p62 bound to COX-2 and stabilized COX-2 through the UBA domain of p62. p62-mediated COX-2 stabilization promotes increased PGE 2 production and may be responsible for increased tumor growth and metastasis in vivo. Elucidating this link between p62 and COX-2 has therefore uncovered a novel oncogenic pathway that may be a key to the development and metastasis of skin cancer.

RNA-Seq analysis of UVA radiation response identifies autophagy-lysosome pathway
To identify genes regulated by UVA radiation, RNA-Seq was performed on sham-or UVA-irradiated normal human epidermal keratinocytes (NHEKs). A comparison of sham-with UVAirradiated NHEKs identified Ͼ4,000 differentially expressed genes ( Fig. 1, A and B). Pathway analysis shows that one of the up-regulated pathways following UVA exposure is the lysosome/autophagy pathway (Fig. 1C). UVA was found to induce the expression of a number of autophagy-related genes, including TFEB, p62, LC3, ULK1, ULK2, LAMP1, and LAMP2 (Fig.  1D). These data indicate that UVA activates transcription of the autophagy-lysosomal pathway, including the expression of p62.

p62 regulation is independent of autophagy in UVA response
Because p62 is up-regulated in skin cancer and regulates skin tumor growth and metastasis (15), we elected to focus on p62 induction by UVA. UVA irradiation increased the protein levels of p62 in a time-dependent ( Fig. 2A) and dose-dependent ( Fig.  2B) manner. In addition, UVA also increased p62 protein levels in mouse epidermis in vivo (Fig. 2C) and skin cancer cells (Fig.  2D). To determine the mechanism by which UVA regulates p62 expression, we assessed the role of autophagy, because p62 is a selective autophagy substrate, and thus inhibition of autophagy leads to an increase in p62 protein levels (9, 10). We next These data indicate that suppression of autophagy did not impair UVA-induced p62 up-regulation.

TFEB is a UVA-responsive transcription factor regulating p62 expression
To determine the mechanism by which UVA regulates p62 expression, we assessed the role of the transcription factor TFEB (Fig. 1C), because TFEB expression was induced by UVA in our RNA-Seq analysis (Fig. 1C). TFEB controls expression of a number of autophagy-lysosomal genes, including p62, through the CLEAR element within the promoter (25). TFEB activation is negatively regulated by phosphorylation, with dephosphorylation triggering the nuclear translocation of TFEB and activation of CLEAR network/TFEB target genes. Following UVA exposure, TFEB protein exhibited a shift in molecular weight suggestive of dephosphorylation (Fig. 3A). Furthermore, there was also an increase in TFEB nuclear localization after UVA irradiation (Fig. 3B), supporting an increase in UVA-induced TFEB activation. TFEB knockdown prevented UVA-induced p62 up-regulation at the protein level (Fig. 4A). Knockdown of TFEB also prevented UVA-induced expression of p62 as well as LC3 (Fig. 4, B-D). Inhibition of RNA synthesis by actinomycin D (Fig. 4E) blocked UVA-induced up-regulation of p62 mRNA, indicating that p62 is transcriptionally upregulated by UVA. Furthermore, UVA induced p62 up-regulation in PAM212 skin cancer cells (Fig. 4F). ChIP analysis of the p62 promoter indicated that UVA increased the binding of TFEB to the p62 promoter (Fig. 4G). These findings demonstrate that TFEB activation is required for UVA-induced p62 expression.

p62 regulates COX-2 expression
Next, we determined the function of p62 induction in UVA response. Using a candidate gene approach, we found that COX-2 was up-regulated in response to UVA, in parallel with p62 induction in NHEK cells (Fig. 5A) and PAM212 skin cancer cells (Fig. 5B). This was accompanied by a moderate increase in COX-2 transcription following UVA (Fig. 5, C and D). Because previous reports have shown that COX-2 induction is required for the development of skin cancer (36), we next asked whether p62 regulates COX-2 in skin cancer cells. Knockdown of p62 led to a decrease in COX-2 protein levels in PAM212 cells (Fig. 5E). Basal COX-2 transcription was also reduced in p62-deficient cells (Fig. 5F). These data indicate that p62 regulates COX-2 protein and transcription.

p62 in tumor growth, metastasis, and UVA response p62 binds to COX-2 and regulates COX-2 stability
Because increased COX-2 protein levels and activity are required for skin cancer development (36 -38), we further examined the regulation of COX-2 protein levels by p62. We have shown that p62 stabilizes Twist1 via the direct interaction of the UBA domain of p62 with polyubiquitinated Twist1 (15). Therefore, we asked whether p62 similarly regulated COX-2 protein stability through the p62 UBA domain. Expression of p62 in HeLa cells, which lack endogenous p62, COX-2, and Twist1 expression, was sufficient to increase the protein stabil-  p62 in tumor growth, metastasis, and UVA response ity of exogenously expressed COX-2 (Fig. 6, A and B). This increase in COX-2 protein stability was lost in cells transfected with a mutant p62 construct that lacks the UBA domain (Fig. 6, C and D), suggesting that the regulation of COX-2 by p62 is dependent on the p62 UBA domain. It appears that the UBA domain loss also decreased basal COX-2 protein level (Fig. 6D). Co-immunoprecipitation (co-IP) of endogenous p62 and COX-2 in skin cancer cells shows that p62 interacts with COX-2 (Fig. 6E). Western blotting of p62 co-IP results showed the binding of p62 with COX-2 (Fig. 6E). We were not able to detect the binding of p62-dUBA with COX-2 (data not shown). This could be due to the loss of binding due to UBA deletion or due to the low basal COX-2 protein levels (Fig. 6D). In addition, p62 also regulates COX-2 transcription through the UBA domain (Fig. 6F). These results suggest that p62 regulates COX-2 transcription and protein stability.

p62 regulates PGE 2 production
COX-2 activity leads to production of PGE 2 , which promotes tumorigenesis through autocrine and paracrine signaling. To determine whether p62 regulates PGE 2 production as a result of COX-2 regulation, we examined PGE 2 levels in skin cancer cells following p62 knockdown. We found that p62 knockdown decreased PGE 2 production (Fig. 7A), and this deficit was reversed by re-expression of COX-2 (Fig. 7A). Similarly, expression of both HA-p62 and COX-2 in HeLa cells increased PGE 2 levels beyond the PGE 2 levels in cells expressing COX-2 alone (Fig. 7B). Co-expression of COX-2 with mutant p62 lacking the UBA domain reduced PGE 2 levels to that of cells lacking COX-2 (Fig. 7B). Therefore, p62 regulates PGE 2 production by stabilizing COX-2 through the UBA domain of p62.

p62 does not regulate proliferation or migration in vitro
Considering the known role of PGE 2 production in proliferation and migration, we assessed whether p62 knockdown impacted these functions in skin cancer cells. In both the Cell Counting Kit-8 assay (Fig. 7C) and the MTS assay (data not shown), p62 knockdown in skin cancer cells had no effect on proliferation. Similarly, p62 knockdown had no effect on migration in Transwell migration assays (Fig. 7D), even when PGE 2 was used as a chemoattractant (data not shown). Whereas this data suggests that p62-mediated effects on PGE 2 production do not affect the cancer cell proliferation and migration, PGE 2 is a known paracrine signaling mediator and may affect neighboring cells in the tumor microenvironment.

p62 is required for tumor growth and metastasis in vivo
To test the requirement for p62 in skin tumor growth and progression, we utilized a syngeneic mouse model of skin cancer. In this model, we injected control and p62-knockdown PAM212 skin cancer cells into BALB/c mice. Measurement of tumor growth over 10 weeks showed that p62 knockdown inhibited tumor growth (Fig. 8A) and metastasis to the lung (Fig. 8B). These data indicate that p62 is required for tumor growth and metastasis. Histological analysis of tumor samples suggested that p62-knockdown tumors exhibit decreased immune cell infiltration and increased cell differentiation (Fig.  8C, arrows). Therefore, we next performed immunohistochem-

p62 in tumor growth, metastasis, and UVA response
ical (IHC) analysis using immune and differentiation markers. Staining for myeloperoxidase (MPO), a marker of myeloid cells, showed that p62-knockdown tumors exhibited lower levels of MPO-positive cells (Fig. 8D). Staining for differentiation marker keratin 10 (K10) also showed that p62-knockdown tumors had an increase in K10-positive differentiated cells as compared with controls (Fig. 8E). These findings support an oncogenic role of p62 in tumor growth and progression.

Figure 8. p62 regulates tumor growth, metastasis, and immune infiltration in vivo.
A, shCon and shp62 PAM212 cells were injected into syngeneic BALB/c mice, and tumor growth was measured over 10 weeks. n ϭ 6 mice/group. *, p Ͻ 0.05, Student's t test, significant difference as compared with the shCon group. B, metastasis to the lungs of mice injected with shCon and shp62 PAM212 cells was measured using H&E staining. n ϭ 6 mice/group. *, p Ͻ 0.05, Student's t test, significant difference. C, H&E staining of tumors from mice injected with shCon and shp62 PAM212 cells.

p62 in tumor growth, metastasis, and UVA response
Discussion UVA radiation is known to cause skin cancer. However, the signaling pathways underlying UVA response are unknown. Here we show that UVA radiation activates the transcription of an autophagy and lysosomal gene program, including p62, through the transcription factor TFEB. p62 regulates COX-2 by two mechanisms: 1) promoting COX-2 expression and 2) stabilizing COX-2 protein by binding to COX-2. Knockdown of p62 inhibits skin tumor growth and metastasis, in association with a decrease in immune cell infiltration in the tumors. Our findings suggest that targeting p62 may be an effective method to prevent tumor growth and metastasis after UVA.
We found that UVA irradiation up-regulates p62 at the transcription levels, independent of autophagy, because 1) UVA increased the level of p62 mRNA, 2) inhibiting RNA synthesis abolished the UVA-induced p62 mRNA increase, and 3) UVA induced the up-regulation of p62 in lysosome-inhibited cells and cells with a genetic autophagy deficiency. Previous studies have shown that in keratinocytes, UVA induces autophagy (40), and this work suggests that p62 functions independently of autophagy in UVA response.
In examining the potential functions of p62 in UVA response, we have identified a novel relationship between p62 and COX-2. As we previously reported with Twist1 (15), p62 binds and stabilizes polyubiquitinated COX-2 through the p62 UBA domain. COX-2 is primarily degraded at the proteasome via ER-associated degradation (29,41). The interaction between COX-2 and the UBA domain of p62 may prevent this degradation of COX-2 at the proteasome, because we have demonstrated that p62 can prevent Twist1 degradation at the proteasome (15). This emerging pattern of positive regulation of protein stability by p62 suggests a mechanism by which p62 differentiates between polyubiquitinated proteins bound for degradation in autophagy and those to be stabilized. Further work will determine how p62 differentiates between these two sets of proteins and whether other oncogenic proteins are similarly stabilized by p62. COX-2 stabilization by p62 also increases the production of PGE 2 . PGE 2 is the key effector of COX-2 activity and acts to promote proliferation (30), invasion (31), and pro-tumorigenic immune infiltration (32,33) while inhibiting cancer cell differentiation. Similar to COX-2 knockdown (36), our studies show that knockdown of p62 led to decreased tumor growth, metastasis, and immune infiltration as well as increased cancer cell differentiation. It is possible that p62 is acting through COX-2mediated PGE 2 production to promote skin cancer growth and metastasis. Further study of the p62-COX-2 signaling axis could determine whether this pathway is critical for skin tumor progression and could provide a novel target for the prevention and treatment of skin cancers.
In addition to p62, our RNA-Seq analysis in primary human keratinocytes identified multiple genes in the autophagylysosome pathways. Whereas we have shown that p62 is required for the transcription and protein stability of COX-2 and for skin tumor growth and metastasis in vivo, other UVAregulated genes may also have important roles in UVA-associated skin cancers. In particular, the function of the autophagy-lysosomal genes is unknown and needs to be investigated in vitro and in vivo. In addition to p62, UVA-induced TFEB nuclear translocation also mediates LC3 up-regulation at the mRNA levels. The roles of TFEB and LC3 in skin cancer are unknown and deserve to be investigated in the future. Recent studies have shown that TFEB is up-regulated in pancreatic cancers and is required for pancreatic cancer growth. The MiT/ TFE family of transcription factors, including TFEB, mediates cancer cell metabolic reprograming to maintain amino acid pools (42,43). LC3 has been shown to be up-regulated in several cancers, including esophageal, gastric, and colorectal cancers (44). However, future studies are required to investigate the mechanism and function of LC3 up-regulation in these cancers and skin cancers.

Cell lines and drug treatments
NHEKs were grown in KGM-Gold BulletKit medium (Lonza) according to the manufacturer's protocol. NHEKs were cultured for less than four passages. PAM212 (squamous cell carcinoma), HeLa, and iBMK (immortalized mouse baby kidney epithelium, kindly provided by Dr. Eileen White, Cancer Institute of New Jersey, New Brunswick, NJ) cells were maintained in DMEM supplemented with 10% fetal bovine serum (Hyclone), 1% non-essential amino acids (Invitrogen), 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen). PAM212 and HeLa cells were maintained in monolayer culture at 37°C and in 95% air, 5% CO 2 (v/v). iBMK cells were maintained at 38.5°C in 8.5% CO 2 .
To inhibit transcription, NHEKs were treated with 1 g/ml actinomycin D (Fisher) for 1 h prior to sham or UVA radiation. To inhibit autophagic flux, cells were treated with 25 nM bafilomycin A1 (Sigma) for 1 h prior to UVA radiation. Protein stability was assessed by treating cells with 100 ng/l cyclohexamide (CHX) for the indicated times. Cells were treated with 10 M MG132 (Sigma) for 6 h before lysing cells for co-IP.

siRNA and plasmid transfection
NHEKs were transiently transfected with siRNA targeting TFEB (Santa Cruz Biotechnology) or p62 (Dharmacon) using an Amaxa Nucleofector kit according to the manufacturer's protocol. Mouse shCon, shp62, shAtg5, and shAtg7 constructs in pLKO.1 vector were purchased from Sigma. Lentivirus was produced by co-transfecting HEK-293T cells with the lentiviral construct, pCMV8.2 packaging plasmid, and pVSV-G envelope plasmid. Supernatant was collected 24 -48 h after transfection and used to infect cells in the presence of 8 g/ml Polybrene (Sigma). Stable clones were selected using 2 g/ml puromycin (Santa Cruz Biotechnology) treatment for 2 weeks as described previously (15).

UVA treatments
For UVA irradiation, four parallel PUVA lamps were used, and doses were measured using a Goldilux UV meter equipped with UVA and UVB detector (Oriel). Contamination from UVB irradiation was eliminated using a 0.13-mm Mylar filter material from Cope Plastics. This filter limits UVB exposure to 0.003% of the total emitted UV radiation (47,48). We have found that these lamps emit no UVC radiation (100 -280 nm). The UVA dosages used here are relevant to human exposure. The dose of UVA that will cause erythema is 30 J/cm 2 for human skin (1), and the UVA dose (20 J/cm 2 ) used in the in vitro studies equates to about 1 h in the midday sun during the summer at latitude 48°north in Paris, France (49). In our laboratory, obtaining 20 J/cm 2 of UVA irradiation requires ϳ1 h. Therefore, the UVA dose used in our application is relevant to human exposure.

RNA-Seq analysis
NHEKs were exposed to sham or UVA irradiation (20 J/cm 2 ). 6 h after irradiation, RNA was prepared from these cells using an RNeasy Plus kit (Qiagen), according to the manufacturer's protocol. Two biological repeats were included for each treatment group. RNA quality assessment, library preparation, and sequencing were performed by the University of Chicago Functional Genomics Facility. RNA quality assessment was performed using an Agilent Bioanalyzer. An oligo(dT)-selected library was prepared, and sequencing was performed on an Illumina HiSeq2000 platform with 50-bp single-end reads.
The established Tuxedo protocol was used to analyze the RNA-Seq data (50). In this pipeline, quality control analysis of raw RNA-Seq data were performed in FastQ Groomer. Reads were then aligned to the human reference genome (hg38) using Tophat2. Transcript assembly was performed using Cufflinks. Cufflinks assemblies were combined using CuffMerge for all treatments. CuffDiff was used to calculate the difference in gene expression between sham-and UVA-irradiated samples, with a false discovery rate of 10% and p Ͻ 0.05 as a standard of significance. CummeRbund was used to perform analyses of differentially expressed genes. Gene ontology and KEGG pathway analysis were performed using DAVID.

Western blotting
Prior to Western blot analysis, cells were treated as indicated and then lysed using radioimmune precipitation lysis buffer containing protease and phosphatase inhibitors (Thermo). Supernatant was removed for analysis or frozen for later use. Normalization of total protein levels was performed using the BCA assay. Lysate was separated on 4 -12% gradient SDS-PAGE gels (Novex) and blotted onto nitrocellulose membranes (Novex). For expression analysis, the following antibodies were used: p62 (GP62-C, Progen), GAPDH (FL-335, Santa Cruz Biotechnology), COX-2 (ab15191, Abcam), LC3A/B (4108, Cell Signaling), and TFEB (A303-673A-M). Anti-HRP secondary antibodies (Cell Signaling) were used for visualization of proteins. Film was used for visualization.

ChIP
IP was performed on 5 ϫ 10 6 cells following treatment with sham or UVA (20 J/cm 2 ) irradiation. ChIP was performed using an EZ-Magna ChIP A Kit (Millipore), according to the manufacturer's protocol. The lysate was sonicated 12 times for 10 s each, with a 30-s rest between each sonication. IP was performed with TFEB (ab2636, Abcam) antibody as well as with positive and negative control antibodies included in the EZ-Magna ChIP kit. qPCR analysis of IP samples was performed using the following primers: 5Ј-CAC AGG CCT TCC TTG TGT C-3Ј (forward) and 5Ј-GCA GAG GCT GTG GCC TA-3Ј (reverse).

Immunofluorescence microscopy
For immunofluorescence analysis, cells were seeded onto glass coverslips overnight prior to treatment. Following treatment, cells were fixed using 10% neutral buffered formalin, permeabilized using Triton X-100, and incubated with primary antibody overnight. Alexa Fluor fluorescence-conjugated secondary antibodies were added for 2 h before imaging samples. Samples were examined using an inverted microscope with fluorescence function and dedicated analysis software (Olympus, model IX71).

PGE 2 production assay
PGE 2 production was measured using a PGE 2 parameter assay kit (KGE004B, R&D Systems). HeLa cells were transiently transfected with empty vector (EV), EV and HA-tagged p62, EV and COX-2, HA-p62 and COX-2, or HA-p62-dUBA and p62 in tumor growth, metastasis, and UVA response COX-2 using X-tremeGENE 9 transfection reagent. 48 h after transfection, cells were serum-starved for 24 h before taking samples for this assay. PAM212 cells with stable knockdown of p62 (shp62) were transiently transfected with mouse COX-2 (shp62-COX-2). 48 h after transfection, cells were serumstarved for 24 h prior to taking medium for this assay.

Migration and proliferation assays
Cell proliferation was measured using the CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (MTS) (Promega) according to the manufacturer's protocol as described previously (15). Cell viability was measured using the Cell Counting Kit-8 assay (Sigma) according to the manufacturer's protocol. Migration was measured using Transwell inserts (Corning) with serum-free medium above and below the insert, with medium supplemented with 10% FBS below the insert, and with serum-free medium containing PGE 2 below the insert. Migrated cells were measured 16 h after seeding.

Mouse studies
All animal procedures were approved by the University of Chicago institutional animal care and use committee. BALB/c mice were purchased from Charles River Laboratories. 5 million PAM212 cells, which are syngeneic with BALB/c mice, were injected subcutaneously into the flanks of BALB/c mice. Tumor growth was measured over 10 -12 weeks using a caliper, and at 1 cm 3 volume, tumors were harvested along with the lungs for histological analysis. Tumors and lungs were fixed in 10% formalin for IHC analysis. For UVA treatment studies in mice, SKH1 hairless mice received sham or 15-J/cm 2 UVA irradiation every other day for a total of three treatments, and then skin samples were harvested for immunohistochemical analysis 72 h after the final treatment.

IHC analysis
For H&E, MPO, and K10 staining, tumors and lungs (or skin in UVA experiments) were harvested and fixed in 10% formalin. Paraffin-embedding, sectioning, and H&E staining were performed by the Human Tissue Resource Center at the University of Chicago. MPO (Abcam ab45977, 1:500) and K10 (Covance MMS-159S, 1:1000) staining was performed as described previously (51). Stained samples were scanned by the Integrated Light Microscopy Core Facility at the University of Chicago.

Statistical analysis
Statistical analyses for qPCR and migration, proliferation, viability, and PGE 2 production assays were performed using GraphPad Prism version 5. Data were expressed as mean of at least three independent experiments and analyzed by Student's t test. Error bars indicate S.D. p Ͻ 0.05 was considered statistically significant.
Author contributions-A. S., B. Z., and Y. Y. H. conceived and coordinated the study, analyzed the data, and wrote the paper. A. S. and B. Z. designed, performed, and analyzed the experiments shown in all figures. L. Q. provided technical assistance and critical reagents. All authors reviewed the results and approved the final version of the manuscript.