Ectopic expression of Zmiz1 induces cutaneous squamous cell malignancies in a mouse model of cancer.

Cutaneous squamous cell carcinoma (SCC) is the second most common form of cancer in the human population, yet the underlying genetic mechanisms contributing to the disease are not well understood. We recently identified Zmiz1 as a candidate oncogene in nonmelanoma skin cancer through a transposon mutagenesis screen. Here we show that transposon-induced mutations in Zmiz1 drive expression of a truncated transcript that is similar to an alternative endogenous ZMIZ1 transcript found to be overexpressed in human SCCs relative to normal skin. We also describe an original mouse model of invasive keratoacanthoma driven by skin-specific expression of the truncated Zmiz1 transcript. Unlike most mouse models, Zmiz1-induced skin tumors develop rapidly and in the absence of promoting agents such as phorbol esters. In addition, we found that the alternative Zmiz1 isoform has greater protein stability than its full-length counterpart. Finally, we provide evidence that ZMIZ1 is overexpressed in a significant percentage of human breast, ovarian, and colon cancers in addition to human SCCs, suggesting that ZMIZ1 may play a broader role in epithelial cancers.

Introduction

Non-melanoma skin cancer is the most common human cancer worldwide with ~1 million new cases reported in the United States annually (Alam and Ratner, 2001). While basal cell carcinoma (BCC) is more common than squamous cell carcinoma (SCC), SCC is a more aggressive disease and metastasizes at low frequency (Alam and Ratner, 2001)(Alam and Ratner, 2001)(Alam and Ratner, 2001)(Alam and Ratner, 2001). Current treatments rely upon surgical excision of the cancerous tissue. However, surgical removal can be disfiguring since tumors develop on sun-exposed skin. The major risk factor for developing non-melanoma skin cancer is UV exposure, and UV-induced TP53 mutations are frequently observed in both BCC and SCC (Benjamin and Ananthaswamy, 2007; Brash ). Additionally, 46% of human SCCs have mutations in HRAS, many of which are also UV-induced (Pierceall ). Despite knowledge of these prevalent mutations, targeted therapies are not in widespread use and greater understanding of the genetics contributing to this disease will aid in their design.

We recently performed a Sleeping Beauty (SB) transposon mutagenesis screen in mice in which a variety of epithelial tumor types were produced, the most common being cutaneous malignancies formerly designated as SCCs (Dupuy ). Nearly 90% of cutaneous tumors harbored transposon insertions within the Zmiz1 locus, likely resulting in over-expression of an N-terminally truncated protein (i.e. Zmiz1Δ1-185). Interestingly, Zmiz1 has previously been implicated in lymphoma through retroviral mutagenesis screens (Sauvageau ; Uren ).

The ZMIZ1 locus encodes a protein with a PAT domain, an MSX-interacting zinc finger (MIZ) domain, and a putative nuclear localization sequence. The molecular function of ZMIZ1 is not well characterized, but the MIZ domain is predicted to function as an E3 SUMO ligase with homology to the PIAS (protein inhibitor of activated Stats) family. Prior work demonstrated that ZMIZ1 increases transcriptional activity of the androgen receptor (AR), p53, and Smad3/4 when overexpressed in cultured cells (Lee ; Li ; Sharma ). The increased AR activity was associated with sumoylation of AR, apparently mediated by ZMIZ1 (Sharma ). A constitutive Zmiz1 knockout mouse allele has been described, but homozygous deletion of Zmiz1 causes early embryonic lethality, thus preventing the study of Zmiz1 function in adult tissues (Beliakoff ). Consequently, current knowledge regarding Zmiz1 function is inadequate to explain how its overexpression might lead to skin cancer.

Here we report the results of experiments performed to validate Zmiz1 as an oncogene and to explore how Zmiz1 truncation by transposon insertion might confer selective advantage to tumor cells. We found that overexpression of Zmiz1Δ1-185 in mouse skin was sufficient to induce tumors, thereby creating a transgenic model of cutaneous malignancy. Interestingly, we discovered that a similar endogenous isoform is expressed in a tumor-specific manner in human skin. We also observed nuclear accumulation of ZMIZ1 in human breast, ovarian, and colon cancers, supporting a widespread relevance of our mouse model to other epithelial cancers, not only in SCC.

Results

Tumor specific expression of an alternative ZMIZ1 transcript

We recently identified Zmiz1 as a cancer gene important for the induction of cutaneous SCC using SB transposon mutagenesis (Dupuy ). Significantly, all insertions were located within intron 8, and oriented in such a way as to promote expression of a truncated transcript including exons 9-24 (Figure 1a). Importantly, the SB-induced transcript strongly resembles an endogenous human ZMIZ1 isoform (uc001kag-2), which includes an alternative promoter and exon that is spliced into exon 7 of the full-length transcript (Figure 1a).

Figure 1

Tumor-specific expression of ZMIZ1

(a) Diagram of the ZMIZ1 locus in mouse and human. The human alternative promoter is located in intron 6 and drives expression of the ZMIZ1 transcript. This transcript is similar to the transcript produced by SB transposon insertion into mouse intron 8. (b) The amino acid sequence of human and mouse short-forms differs by approximately 60 amino acids at the N-termini, but are otherwise nearly identical. Note that both full-length and ZMIZ1short contain all currently annotated functional protein domains. (c) RT-PCR analysis confirmed that ZMIZ1 is expressed in MCF7, but not in the other human cell lines tested. (d) Quantitative RT-PCR performed on human SCCs shows increased tumor-specific expression of ZMIZ1. Dotted lines indicate the average Ct values for normal skin.

Transcriptome and chromatin immunoprecipitation (ChIP) data generated by the ENCODE Project from a variety of cell lines are consistent with the presence of an alternative promoter just upstream of the alternative exon (Birney ) (Supplementary Figure S1). Expression of the short isoform (i.e. ZMIZ1short) is likely regulated by the activity of this alternative promoter. All known and predicted functional domains are preserved in the predicted open reading frame of this short isoform (Figure 1b).

We detected endogenous expression of ZMIZ1 in MCF7 breast cancer cells, however, expression was not detected in the other human cell lines tested, including immortalized normal keratinocyte cell lines HaCaT and N-HSK-1 (Figure 1c). ZMIZ1 transcript was also measured in human SCC and, when possible, matched normal skin samples. Interestingly, we observed tumor-specific ZMIZ1 up-regulation in 4 of 6 samples by qRT-PCR (Figure 1d). The full-length transcript (i.e. ZMIZ1) was up-regulated in only 2 of 6 samples.

Zmiz1Δ1-185 expression in mouse skin induces squamous cell tumors

As previously mentioned, we predicted that transposon-induced expression of the Zmiz1Δ1-185 is oncogenic. To test this hypothesis, we engineered transgenic mice that conditionally express Zmiz1Δ1-185 (Figure 2a). A lox-stop-lox strategy was employed to prevent ubiquitous expression of Zmiz1Δ1-185. Cre-mediated recombination will delete the DsRed reporter embedded in the lox-stop-lox cassette and induce expression of the downstream Zmiz1Δ1-185-IRES-EGFP cassette. Two Zmiz1Δ1-185 founder animals showed detectable levels of the DsRed reporter in the skin and were used to establish transgenic lines 207-Zmiz1Δ1-185 and 163-Zmiz1Δ1-185.

Figure 2

Zmiz1Δ1-185 transgenic mice develop skin tumors

(a) Diagram of the Cre-inducible Zmiz1Δ1-185 transgene. (b) Western blot showing skin from Zmiz1Δ1-185 single transgenic mice has nearly undetectable expression of EGFP, whereas double transgenic mice display detectable EGFP expression. (c) Kaplan-Meier survival curve of mouse line 207 showing tumor-free survival of K14-CreTg/+; Zmiz1Δ1-185 Tg/+ double transgenic mice (circles, n=54), K14-Cre+/+; Zmiz1Δ1-185 Tg/+ single transgenic (squares, n=18), and K14-CreTg/+; Zmiz1Δ1-185 +/+ single transgenic mice (triangles, n=35). Double transgenic mice developed skin tumors with an average latency of 14 weeks after birth, while single transgenic controls did not develop skin tumors (log-rank, P<0.0001). (d) Western blot confirms overexpression of the Zmiz1Δ1-185 transgenic protein in mouse tumors. Endogenous Zmiz1 expression is undetectable in normal skin.

Each line was crossed to K14-Cre transgenic mice to induce expression of Zmiz1Δ1-185 in the proliferative basal layer of the cutaneous epidermis (Dassule ). Transgene expression in normal adult skin was assessed by western blot analysis of EGFP expression. As expected, EGFP expression was readily detectable in double transgenic skin by western blot, but undetectable in skin from single transgenic mice (Figure 2b). Cre-mediated expression of 163-Zmiz1Δ1-185 was less efficient, as evidenced by lower EGFP protein levels in double transgenic skin (Supplementary Figure S2). Given that both Zmiz1Δ1-185 transgenes are likely multi-copy concatemers, the discrepancy in EGFP expression is likely caused by Cre-mediated deletion of the transgene. Similar situations have previously been described (Grippo ). Nevertheless, both Zmiz1Δ1-185 transgenic lines show Cre-inducible expression.

A cohort of Zmiz1Δ1-185;K14-Cre double transgenic mice (n=54) was generated and aged alongside Zmiz1Δ1-185 (n=18) or K14-Cre (n=35) single transgenic littermate controls (Figure 2c). Over 85% of double transgenic mice spontaneously developed an average of 2.7 skin tumors per mouse with a mean latency of 14 weeks (p<0.0001). In all cases, the lesions were variably-sized crateriform masses filled with compact keratin and lined by proliferative stratified squamous epithelium, a morphology consistent with keratoacanthoma (KA). Similarly, a more comprehensive study of the SB-induced tumors from Dupuy showed identical histology, so these will also be referred to as KAs for the remainder of this manuscript. Zmiz1Δ1-185-induced masses compressed subjacent tissues and exhibited cutaneous muscle invasion in 23 of 147 tumors (15.6%). Double transgenic mice from line 163-Zmiz1Δ1-185 also developed KAs, but exhibited lower penetrance with increased latency. This is likely due to the lower transgene expression levels observed (Supplementary Figure S2). These results are consistent with the hypothesis that truncated Zmiz1 is an oncoprotein, and suggest that expression of ZMIZ1 may also directly contribute to human SCC.

Gene expression profiling of Zmiz1Δ1-185 induced skin tumors

The Zmiz1Δ1-185;K14-Cre double transgenic mice represent the first published in vivo validation of Zmiz1 as an oncogene, but the biological functions of Zmiz1 are not well studied. Thus, we sought to further characterize the Zmiz1-induced tumors to determine if any mechanistic insights could be gained. We generated gene expression profiles from 5 independent tumors from 207-Zmiz1Δ1-185;K14-Cre transgenic mice along with 6 normal control skin samples from 207-Zmiz1Δ1-185;K14-Cre, 207-Zmiz1Δ1-185, or K14-Cre transgenic mice.

Normalized gene expression profiles were compared to identify genes that showed ≥2-fold change in expression between tumor and normal skin with a false discovery rate less than 5%. This identified 701 and 603 genes that were up- or down-regulated, respectively (Supplementary Table S1). Differential expression of 6 of these genes was verified using qRT-PCR (Supplementary Figure S3). Analysis of these results using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (Huang da ) or gene set enrichment analysis (GSEA) (Subramanian ) did not reveal any significant conservation among the differentially-regulated genes. Repeating these analyses using only genes that showed a ≥5-fold change in expression also failed to reveal any significant trends.

Zmiz1 accumulates in the nuclei of mouse KAs and human SCCs

Zmiz1 protein expression was evaluated in SB-induced KAs by immunohistochemical (IHC) staining of formalin-fixed paraffin-embedded skin sections. As predicted, tumors with transposon insertions in the Zmiz1 locus displayed high levels Zmiz1 expression (Figure 3a). Conversely, Zmiz1 protein was undetectable in normal mouse skin (Figure 3b), although the endogenous full-length transcript is expressed (Dupuy ). Based on these findings, we conclude that the majority of Zmiz1 protein detected in SB-induced tumors is derived from the truncated Zmiz1 transcript.

Figure 3

Aberrant nuclear accumulation of Zmiz1 in tumor tissue

(a) Immunohistochemistry (IHC) analysis reveals overexpression of Zmiz1 in SB. induced tumors with transposon insertions in intron 8. (b) Zmiz1 expression is undetectable by IHC in normal mouse skin. Scale bars, 50μm. Similarly, nuclear accumulation was observed in K14-CreTg/+; Zmiz1Δ1-185 Tg/+ double transgenic tumors (c), but not normal skin from double transgenic mice (d). Scale bars, 100μm. (e) Strong nuclear ZMIZ1 expression was observed in 14 of 19 human cutaneous SCCs when analyzed by IHC. (f) Example of a human SCC without ZMIZ1 overexpression. Scale bars, 100μm.

Similarly, tumor tissue from Zmiz1Δ1-185;K14-Cre double transgenic mice displayed the same strong Zmiz1 nuclear expression pattern that was observed in spontaneous SB-induced KAs (Figure 3c). Interestingly, normal adult double transgenic skin did not express detectable levels of Zmiz1 protein by IHC or western blot (Figures 2d and 3d). This led us to hypothesize that Cre-mediated recombination results in complex transgene rearrangements with high frequency and results in a lack of transgene expression in most cells. Thus, Zmiz1Δ1-185 overexpression is a rare event that drives hyperplasia, and eventually tumorigenesis. If this is true, expression of the Zmiz1Δ1-185 transgene should be tumor-specific.

DsRed and EGFP RNA expression was measured by qRT-PCR in normal skin from double- and single-transgenic mice, as well as in tumors (Figure S4). We found that some level of DsRed expression was retained even after Cre recombination, suggesting that recombination is inefficient. Furthermore, while transgenic Zmiz1Δ1-185 expression in normal skin is not detectable at the protein level, we are able to detect expression of EGFP at the RNA level. Consistent with our hypothesis, EGFP RNA expression is greatly increased in tumors. Likewise, a comparison of the normalized expression values for all probes corresponding to Zmiz1 exons from our exon array datasets showed tumor-specific overexpression of exons encoded by the Zmiz1Δ1-185 transgene. Expression of these exons in normal skin from Zmiz1Δ1-185;K14-Cre double transgenic mice was similar to that of non-transgenic control skin samples (Supplementary Figure S4).

Our results indicate that Zmiz1 is an oncogene involved in skin cancer in mice (Figure 2c). Additionally, ZMIZ1 is overexpressed in human SCC, but not normal keratinocytes (Figure 1c and d). However, protein expression in human SCC has not been assessed. Therefore, we analyzed a panel of human SCCs by IHC to determine the frequency of ZMIZ1 nuclear accumulation. Consistent with our observations in mouse KAs, 14 of 19 human SCCs showed nuclear accumulation of ZMIZ1 (Figure 3e and f). This data, combined with the fact that overexpression of Zmiz1Δ1-185 in mouse skin is sufficient to produce skin lesions, indicates that ZMIZ1 is involved in the etiology of human SCC.

Finally, we obtained a tumor tissue array from the National Cancer Institute (NCI) to assess ZMIZ1 expression in breast, ovarian, colon and lung tumors. Twenty-eight percent of breast, 15% of ovarian, and 10% of colon cancers displayed strong nuclear ZMIZ1 expression when analyzed by IHC (Table 1 and Supplementary Figure S5). This result suggests a conserved role for ZMIZ1 in cancer development across diverse epithelial tissue types.

Table 1

Mixed tumor array ZMIZ1 expression data

Tumor type	% Nuclear Zmiz1	# of tumorsexamined
Breast	28%	32
Ovarian	15%	40
Colon	10%	28
Lung	0%	41

Biological impact of Zmiz1 protein truncation

As previously mentioned, transposon insertions in our SB screen clustered in intron 8 of Zmiz1, demonstrating a strong preference for a transcript mimicking ZMIZ1 rather than full-length ZMIZ1. In an attempt to explain why the short-form might give tumor cells a selective advantage, we examined steady-state protein expression levels of full-length and truncated Zmiz1 expressed from recombinant retroviruses in multiple cell lines (Figure 4). Higher steady-state expression of Zmiz1Δ1-185 was consistently observed across independent cell lines, despite no significant differences in transcript levels as assessed by qRT-PCR (data not shown). This suggests that Zmiz1 protein stability is regulated. Consistent with this hypothesis, Zmiz1 protein stability was increased upon treatment with the proteasomal inhibitor MG132 (Figure 4). This result indicates that increased protein stability might confer a selective advantage to tumor cells expressing Zmiz1Δ during tumorigenesis.

Figure 4

Zmiz1Δ1-185 displays increased protein stability

Western blots show higher steady-state protein expression in Zmiz1Δ1-185-HA-expressing HeLa, MCF7, and N-HSK-1 cells than those stably expressing full-length Zmiz1-HA, despite lacking significant differential transcript expression as analyzed by qRT-PCR (data not shown). Where indicated, MG132 (20μm) treatment increased protein stability, suggesting Zmiz1 and Zmiz1Δ1-185 are degraded by the proteasome and that stability is regulated at the protein level.

Discussion

We previously identified Zmiz1 as a candidate cancer gene in cutaneous squamous cell malignancies using Seeping Beauty (SB) transposon insertional mutagenesis (Dupuy ). The frequency of ZMIZ1 mutation in human cancer is unknown, and though ZMIZ1 transcripts have been sequenced in a variety of tumor types, very few independent tumors have been tested for any single tumor type (Forbes ). However, an alternative ZMIZ1 promoter has recently been annotated and is located just upstream of an alternative exon that splices into exon 7 of the full-length transcript. This endogenous transcript is predicted to produce an N-terminally truncated protein (ZMIZ1short) that is strikingly similar to the Zmiz1 truncation selected by the SB system.

Endogenous ZMIZ1 is expressed in MCF7 breast cancer cells, but not in two immortalized human keratinocyte cell lines (HaCat, and N-HSK-1). When we assessed ZMIZ1 expression status in a panel of human SCC samples, we observed increased tumor-specific expression when compared to normal skin. Furthermore, expression of the short-form in mouse skin leads to tumor formation with remarkably short latency, suggesting that Zmiz1Δ1-185 provides a robust oncogenic stimulus. Taken together, we conclude that expression of the short-form not only correlates with tumor status in both human SCC and mouse models, but also that ZMIZ1 expression may play a causal role in tumor development.

It should be noted that the histology of skin tumors induced by skin-specific Zmiz1Δ1-185 expression was consistent with the diagnosis of keratoacanthoma (KA) rather than SCC. KAs can be defined by their tendency to spontaneously regress. However, it is debated whether these are early-stage tumors that may progress into invasive SCC, or if they remain a distinct class of squamous hyperplasia (Ko, 2010). Efforts to develop histological methods to distinguish KA from SCC have failed to provide reliable and specific markers (Karaa and Khachemoune, 2007; Mandrell and Santa Cruz, 2009). Regardless, skin-specific Zmiz1Δ1-185 expression is clearly capable of producing invasive squamous cell malignancies that did not spontaneously regress. Whether these would become even more aggressive upon the acquisition of additional mutations is unknown.

Until now, the role of Zmiz1 in cutaneous malignancy has been unappreciated, and our transgenic skin tumor model may lend unique insight into the biology of non-melanoma skin cancer for a variety of reasons. For example, the majority of genetically engineered mouse models of non-melanoma skin cancer require promotion with a phorbol ester such as TPA (12-O-tetradecanoylphorbol-13-acetate) and produce mostly benign papillomas, only some of which progress to SCC (Jansen ; Liu ; Vassar ; Wang ). Most of these models have also been constructed on susceptible inbred strain backgrounds (e.g. FVB/n, BALB/c) (Diez ; Proweller ; van Hogerlinden ). Despite these features, many existing mouse models develop tumors with a long latency. In contrast, Zmiz1Δ1-185; K14-Cre transgenic mice rapidly develop tumors on a resistant strain background without the need for promotion with TPA (Figure 2). Furthermore, we did not detect any spontaneous mutations in either Hras (n=14) or Trp53 (n=13) in Zmiz1Δ1-185-induced KAs (Supplementary Figure S6), demonstrating the strong oncogenic potential of Zmiz1. The status of Zmiz1short expression in mouse models of SCC is currently unknown, and it may be useful to investigate whether Zmiz1 also plays a role in these models since KAs are rarely seen in two-stage chemical carcinogenesis experiments.

In an effort to further characterize our model, gene expression profiling was performed on transgenic tumors. Many of the down-regulated genes in tumors are involved in muscle function (e.g. Myh1, Ttn, Acta1). It should be noted that, unlike normal skin, tumors are largely devoid of hair follicles and their associated arrector pili muscles. As a consequence, the decreased expression of muscle genes observed in tumors could be a result of variation in cellular composition and may not directly contribute to Zmiz1-induced transformation.

Among the genes showing the greatest increase in expression in tumor relative to normal skin samples are members of the small proline-rich protein (Sprr) and serine peptidase inhibitor (Serpin) families (Supplementary Figure S3). Some of these family members have previously been shown to be associated with KA and/or SCC (De Heller-Milev ; Kato and Torigoe, 1977). In particular, overexpression of Serpinb3a has been shown to mediate susceptibility to skin tumors in mice (Gariboldi ). Therefore, these genes represent likely candidates for elucidating a mechanism by which Zmiz1 induces tumor formation.

One of the few proposed biological functions of ZMIZ1 is an E3 SUMO ligase activity (Sharma ). A variety of E3 SUMO ligases have been reported to have roles in promoting genomic stability and DNA repair (Bartek and Hodny, 2010; Potts, 2009). We report that Zmiz1Δ1-185 has oncogenic function in the skin, but whether its SUMO ligase activity is required for oncogenesis is currently unknown. Presumably both full-length Zmiz1 and Zmiz1Δ1-185 could function as E3 SUMO ligases, yet there is clearly selection for the short-form in SB-induced tumors. Our data suggest that expression of Zmiz1Δ1-185 could be due to its increased protein stability. Therefore, expression of Zmiz1Δ1.185 or ZMIZ1short would lead to prolonged sumoylation activity, which may ultimately contribute to the oncogenic potential of Zmiz1.

Our experimental results include data directly supporting Zmiz1 as an oncogene, the development and initial characterization of an original mouse model of skin cancer, and the discovery that an endogenous human alternative ZMIZ1 transcript is expressed in a tumor-specific manner. We also report that frequent aberrant ZMIZ1 expression occurs in human SCC, as well as in a subset of human breast, ovarian, and colon cancer, demonstrating a potential role in the development of a variety of human epithelial cancers. While ZMIZ1short expression status in these cancers is unknown, the results of our experiments indicate that ZMIZ1 is causally involved in tumorigenesis. Our data suggests that further study of the role ZMIZ1 plays in cancer could have widespread impact on our current understanding of these diseases.

Materials & Methods

ZMIZ1 and ZMIZ1short qRT#PCR

Total RNA was extracted from the stably-transduced cell lines or human tissues using the PerfectPure RNA Tissue Kit (5-Prime, Gaithersburg, MD). Total cDNA was generated with the SuperScript® III First-Strand Synthesis System (Invitrogen, Carlsbad, CA) using the oligo dT primer. qPCR was performed using the Platinum® SYBR® Green qPCR SuperMix (Invitrogen, Carlsbad, CA) with the following primers: ZMIZ1altexF (5′-GCGACCGGTGCAACTTCTA-3′), ZMIZ1ex8R (5′-GACAGAGTGGGTTTCATGGAG-3′), ZMIZ1Ex6F (5′-AATCCTGCCAACTTCCACAA-3′), ZMIZ1ex7R (5′-AGCAGTCTGTAGCCGAGGTC-3′), TBPex2F (5′-GCTGAGAAGAGTGTGCTGGA-3′), and TBPex3R (5′-GCCATAAGGCATCATTGGAC-3′). ZMIZ1 or ZMIZ1 Ct values were first normalized to TBP Ct values, and then to N-HSK-1 using the ΔΔCt computational method (Livak and Schmittgen, 2001). This study was conducted in adherence to the Helsinki guidelines. Patients provided written consent for studies using human tissues and all procedures were approved by the Institutional Review Board.

Generation of Zmiz1Δ1-185 transgenic mice

Exons 9-24 of Zmiz1 cDNA were PCR-amplified from the NIH Mammalian Gene Collection IRAV clone ID 6856060 (Invitrogen, Carlsbad, CA) using primers Zmiz1-F (5′-GCTAGCCCTATGGCCAATGCCAACAA-3′) and Zmiz1-R (5′-AGATCTGTTGTTCTCAAAGAGAGACA-3′). This fragment was digested with NheI/BglII and cloned into the NheI/XhoI sites of pTraffic (Basheer et al., in prep) along with a linker to add a stop codon [annealed Linker+ (5′-GATCCTAAAGATCTC-3′) and Linker− (5′-TCGAGAGATCTTTAG-3′) oligos]. This construct was sent to the University of Iowa Transgenic Animal core facility to generate Zmiz1Δ1-185 transgenic mice on a B6SJL F1 background. Animals used for aging were backcrossed to C57BL/6 for 3 generations before breeding to K14-Cre mice (#004782, Jackson Laboratory, Bar Harbor, ME). All aging experiments were performed following guidelines approved by the University of Iowa Institutional Animal Care and Use Committee.

Immunohistochemistry

Formalin-fixed, paraffin-embedded skin tumors were cut into 5μm sections and mounted on glass slides, The ImmPRESS Antibody Peroxidase Kit was used for immunolabeling (Vector Laboratories, Burlingame, CA) using anti-RAI17 (AbGent, San Diego, CA) at a 1:50 dilution overnight at 4°C. Sections were counterstained with Hematoxylin QS (Vector Laboratories, Burlingame, CA) and mounted in Permount (Fisher Scientific, Pittsburgh, PA) for light microscopy. Primary antibody incubation was omitted in negative controls.

Generation of stable cell lines and MG132 treatment

MCF7 and HeLa cells were obtained from ATCC (Manassas, VA). N-HSK-1 keratinocytes were a gift from Aloysius J. Klingelhutz (Gourronc ). Retroviral Zmiz1 constructs were produced by co-transfection of GP2-293 packaging cells with the pVSV-G packaging plasmid and pQCXIN, or pQCXIN containing Zmiz1 or Zmiz1Δ1-185. Retroviral supernatants were collected 24 hours later and applied to proliferating cells. Stably transduced cells were selected 24 hours later with media containing 400 μg/ml (HeLa, MCF7) or 100 Wg/ml (N-HSK-1) G418. Selection was maintained for 12 days, whereupon resistant clones were pooled and expanded in selection media.

Total cDNA was generated as previously described. Ectopic expression of Zmiz1 or Zmiz1 mRNA in stably transduced cell lines was assessed by qPCR with primers against the neomycin resistance sequence (Neo forward: 5′-TGAATGAACTGCAGGACGAG-3′ and Neo reverse: 5′-ATACTTTCTCGGCAGGAGCA-3′). Proteasomal degradation was inhibited by culturing cells in 20 μm MG132 (Sigma, St. Louis, MO) for 24 hours.

Western blotting

Total protein was collected by lysing with RIPA. Samples were boiled in a reducing buffer and SDS-PAGE was performed. Protein was transferred to nitrocellulose membranes for western blotting. Antibodies used were anti-HA (1:5000, Covance, Princeton, NJ), anti-RAI17 (1:100, Abgent, San Diego, CA) and anti-β-Tubulin (1:1000, Sigma-Aldrich, St. Louis, MO).