Polymorphic genetic control of tumor invasion in a mouse model of pancreatic neuroendocrine carcinogenesis

Matthew G. H. Chuna,b,c,d, Jian-Hua Maoc,1, Christopher W. Chiub,c, Allan Balmainc, and Douglas Hanahana,b,c,2,3

aDepartment of Biochemistry and Biophysics, bDiabetes Center, and cHelen Diller Family Comprehensive Cancer Center, University of California, San Francisco, CA 94143; and dProgram in Biological Sciences, University of California, San Francisco, CA 94158

Contributed by Douglas Hanahan, August 31, 2010 (sent for review August 2, 2010) Cancer is a disease subject to both genetic and environmental hallmark capability for invasive growth in the RT2 mouse model influences. In this study, we used the RIP1-Tag2 (RT2) mouse of cancer. model of islet cell carcinogenesis to identify a genetic locus that influences tumor progression to an invasive growth state. RT2 Results mice inbred into the C57BL/6 (B6) background develop both non- PNET Progression to Invasive Carcinoma Is Modulated by Genetic invasive pancreatic neuroendocrine tumors (PNET) and invasive Background. Following anecdotal observations that PNETs de- carcinomas with varying degrees of aggressiveness. In contrast, veloping in RT2 mice inbred into the C3H background were RT2 mice inbred into the C3HeB/Fe (C3H) background are compar- predominantly noninvasive, we carefully examined the distribu- atively resistant to the development of invasive tumors, as are RT2 tion of the distinctive invasive phenotypes in de novo PNETs C3HB6(F1) hybrid mice. Using linkage analysis, we identified a 13- arising in RT2 mice inbred into either the B6 or C3H genetic Mb locus on mouse 17 with significant linkage to the backgrounds, as well as in C3HB6(F1) hybrids (F1), to determine development of highly invasive PNETs. A residing in this whether the parameter of tumor invasiveness was indeed af- locus, the anaplastic lymphoma kinase (Alk), was expressed at fected by genetic background (Fig. 1 A–C). significantly lower levels in PNETs from invasion-resistant C3H The development of invasive carcinoma lesions (IC) was mice compared with invasion-susceptible B6 mice, and pharmaco- strongly suppressed in RT2 C3H mice. Whereas IC lesions con- logical inhibition of Alk led to reduced tumor invasiveness in RT2 stitute more than half of all tumors in RT2 B6 animals at 14 wk, B6 mice. Collectively, our results demonstrate that tumor invasion less than 15% of all tumors could be classified as invasive in RT2 is subject to polymorphic genetic control and identify Alk as a ge- C3H mice (Fig. 1D). This reduction occurred in both the focally fi netic modi er of invasive tumor growth. invasive IC1 and the widely invasive IC2 subclasses of invasive RT2 tumor lesions (Fig. 1E). The development of IC lesions was also fi anaplastic lymphoma kinase | cancer modi er | malignant suppressed in RT2 F1 mice, and the overall distribution of invasive progression | pancreas cancer | transgenic mouse lesions in RT2 F1 mice was similar to that in RT2 C3H mice (Fig. 1 D and E). These data indicate that the C3H genetic background ancer is a complex disease governed by environmental and is resistant to the development of invasive RT2 PNETs, whereas Cgenetic factors, including genetic mutations and polymor- the F1 phenotype demonstrates that the resistant C3H back- phisms that modulate cancer susceptibility (1). Although many ground is dominant over the susceptible B6 background. investigations have focused on identifying factors that affect in- We also examined other parameters of PNET tumorigenesis itial tumor development (2, 3), data from both human and in the B6 and C3H backgrounds to determine whether additional mouse studies have demonstrated that genetic polymorphisms phenotypes were similarly affected by genetic background. The can modulate multiple aspects of tumorigenesis, such as tumor average tumor burden per animal was significantly higher in both – progression (4 6) and response to therapy (7). RT2 C3H and RT2 F1 mice as compared with RT2 B6 mice, In this study, we investigated the effects of genetic background whereas the average number of macroscopic tumors per animal on tumor progression to an invasive growth state, motivated by was higher in RT2 C3H mice as compared with RT2 B6 and RT2 a provocative observation that mice carrying the same oncogenic F1 mice (Fig. S1). However, there were no significant differences transgene but differing in genetic background developed tumors with regard to either the rate of tumor proliferation or tumor that were markedly distinctive in their invasiveness. This model, apoptosis (Fig. S1). There was no indication that the driving on- the RIP1-Tag2 (RT2) mouse model of islet cell carcinogenesis, cogene was responsible for these phenotypic differences because develops multiple pancreatic neuroendocrine tumors (PNET) in the levels of the Tag oncoprotein were similar in tumors isolated a relatively synchronous and predictable multistage progression – from RT2 mice in the different genetic backgrounds (Fig. S2), pattern by 12 14 wk of age owing to the expression of the SV40 consistent with a previous assessment (10). Additionally, the ex- T antigen oncoprotein (Tag) in the pancreatic β cells (8). The tumorigenesis pathway has predominantly been studied in RT2 mice inbred into the C57BL/6 (B6) background, and the PNETs Author contributions: M.G.H.C. and D.H. designed research; M.G.H.C. and C.W.C. per- that arise in this genetic context display a spectrum of invasive formed research; M.G.H.C., J.-H.M., C.W.C., A.B., and D.H. analyzed data; and M.G.H.C. phenotypes and can be classified as noninvasive islet tumors (IT), and D.H. wrote the paper. focally invasive type-1 carcinomas (IC1), and broadly invasive The authors declare no conflict of interest. type-2 carcinomas (IC2) (9). Surprisingly, we observed that when Freely available online through the PNAS open access option. RT2 mice were inbred into a second strain, C3HeB/Fe (C3H), the 1Present address: Life Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, tumors that arose were predominantly noninvasive, despite being CA 94720. otherwise similar in their tumorigenesis phenotype. The impli- 2Present address: Swiss Institute for Experimental Cancer Research, Swiss Federal Institute cation that the invasive phenotype was influenced by genetic of Technology Lausanne, Lausanne CH-1015, Switzerland. background prompted our investigation, which was aimed at 3To whom correspondence should be addressed. E-mail: dh@epfl.ch. fi assessing the hypothesis that a polymorphic modi er locus (or loci) This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. mediated the susceptibility or resistance to the acquisition of the 1073/pnas.1012705107/-/DCSupplemental.

17268–17273 | PNAS | October 5, 2010 | vol. 107 | no. 40 www.pnas.org/cgi/doi/10.1073/pnas.1012705107 Downloaded by guest on September 25, 2021 SNPs that cover the mouse genome and discriminate between the B6 and C3H backgrounds (Dataset S1). Statistical analysis was subsequently performed using R/qtl to determine whether there was evidence of linkage to the development of invasive lesions or to any of the other RT2 tumor phenotypes. Log of odds (LOD) scores of ≥1.9 and ≥3.0 were considered suggestive and significant linkage, respectively (15). Using the development of IT, IC1, or IC2 PNETs as quanti- tative traits, we observed significant linkage to four SNPs on chromosome 17 for the development of IC2 lesions, with a peak LOD score of 3.52 (Fig. 2A and Dataset S2). The 95% confi- dence interval was located from 63.7 to 76.4 Mb, a 13-Mb region that contains more than 50 annotated genes and one miRNA, mir-1195 (Fig. 2B). Interestingly, we did not identify any locus that was linked to the IC1 phenotype, despite the different fre- quencies in the development of this class of tumors in RT2 B6 and RT2 C3H mice (Fig. S5 and Dataset S2). – Fig. 1. PNET invasion is dependent on genetic background. (A C)H&E Additionally, we observed significant linkage to the X chro- staining of a noninvasive IT PNET, a focally invasive IC1 PNET, and a broadly invasive IC2 PNET from an RT2 B6 mouse. T indicates tumor region, mosome to the development of IT lesions and to the metric of and Ex indicates exocrine pancreas. Dashed lines demarcate tumor mar- tumor number (Fig. S5 and Dataset S2). In both situations, the gins. (Scale bars, 200 μm.) (D) Quantification of tumor invasiveness rep- linked region essentially spanned the entire chromosome, which resented as the percentage of IT tumors or total IC tumors (IC1 + IC2) in complicated our efforts to analyze this region in further detail. RT2 mice on the B6, C3H,andF1 genetic backgrounds at 14 wk of age. A We therefore proceeded to investigate the genes in the minimal minimum of 117 tumors per group was graded. *P < 0.001 by Fisher’sexact region of chromosome 17 that showed significant linkage to the test. (E)SameasD except IC lesions are separated into the IC1 and IC2 development of IC2 tumors. subclasses. *P < 0.001 by the χ2 test. GENETICS Anaplastic Lymphoma Kinase Resides in the Chromosome 17 Minimal B6 C3H pression of cadherin 1 (Cdh1, also known as E-cadherin), a known Region and Is Differentially Expressed in the and Genetic regulator of invasion in the RT2 model as well as other cancers Backgrounds. It has previously been suggested that genetic poly- fl (11), was not obviously different (Fig. S2). morphisms can in uence the levels of in the context of phenotypic modifiers of complex traits (16, 17). We Invasive Modifier Does Not Act in the Bone Marrow–Derived Tissue therefore asked whether any of the genes located within the Compartment. Because bone marrow–derived (BMD) inflam- minimal chromosome 17 region might be differentially expressed matory cells that supply matrix-degrading enzymes such as ca- thepsin proteases and heparanase are functionally implicated in the invasive phenotype in this model (12–14), we examined the possibility that the reduced invasiveness in RT2 C3H and RT2 F1 mice was due to deficiencies in the invasion-promoting func- tionality of BMD cells. We transferred bone marrow from B6 or F1 donor mice into RT2 F1 animals with the rationale that B6 but not F1 bone marrow would “rescue” the invasive phenotype in recipient RT2 F1 mice if the invasive modifier operated in this tissue compartment. RT2 F1 mice were chosen as recipients because they develop invasive PNETs at a reduced frequency (Fig. 1 D and E) and should also be capable of receiving bone marrow from either B6 or F1 donors without host/donor in- compatibility complications. In brief, we did not observe any differences in the invasive phenotype or in any other parameter of RT2 tumorigenesis in RT2 F1 mice whose immune systems had been rendered B6 (Fig. S3). These results suggest that the poly- morphic difference is operative in the cancer cells themselves or possibly in other cellular compartments of the stroma. In light of the evident genetic differences in the frequency of developing invasive carcinomas in RT2 mice, we next sought to map the putative polymorphic locus/loci associated with sus- ceptibility vs. resistance to the invasive phenotype using standard genetic linkage analysis. Fig. 2. The IC2 tumor phenotype is linked to a region on chromosome 17. Linkage Analysis Identifies a Region on Chromosome 17 That Is (A) LOD scores for the IC2 phenotype across the mouse genome. The IC2 Associated with the Development of Invasive Carcinomas in RT2 Mice. phenotype shows significant linkage (LOD ≥3.0) to a region on chromosome 17. Dashed line demarcates LOD-3.0 significance cutoff. (B) Physical map of To identify the genetic locus/loci that modify the invasive phe- fi notype in RT2 mice, we performed a genome-wide linkage study. the 95% con dence interval on chromosome 17. Map was constructed using data from the University of California, Santa Cruz Genome Browser (http:// One hundred forty-three RT2 N2 backcrossed mice, resulting genome.ucsc.edu/) and the National Center for Biotechnology Information from crossing RT2 F1 male mice with B6 female mice (Fig. S4), MapViewer (http://www.ncbi.nlm.nih.gov/projects/mapview/) for the mouse were scored for the incidence of IT, IC1, and IC2 tumor le- genome. Red and green arrows indicate genes that are expressed at sig- sions in addition to the other parameters of RT2 tumorigenesis nificantly higher and lower levels, respectively, in tumors isolated from RT2 (Dataset S1). Constitutional tail DNA was genotyped across 561 C3H mice as compared with RT2 B6 mice.

Chun et al. PNAS | October 5, 2010 | vol. 107 | no. 40 | 17269 Downloaded by guest on September 25, 2021 between the parental strains and therefore contribute to the explore the potential role that Alk might play in the development observed differences in the invasion phenotypes. of invasive RT2 tumors. RNA from RT2 B6 and RT2 C3H tumors were profiled by quantitative PCR for the genes located within the minimal re- Pharmacological Inhibitor of Alk Inhibits Invasion and Other Param- gion on chromosome 17. This analysis revealed that a small eters of PNET Tumorigenesis. We used a small molecule inhibitor subset of the resident genes—Alk, Dlgap1, Emilin2, Lbh, Ltbp1, of Alk kinase activity, NVP-TAE684 (TAE684) (20), in an ex- Rab31, and Spdya—showed significant differential expression perimental therapeutic trial in RT2 mice, aiming to assess the between the B6 and C3H genetic backgrounds at the mRNA effects of reduced Alk activity on RT2 tumorigenesis, particularly level (Figs. 2B and 3A and Dataset S3). with regard to the parameter of tumor invasion. We were particularly intrigued by the Alk gene, which encodes RT2 B6 mice were treated for 4 wk with TAE684 or vehicle the anaplastic lymphoma kinase. Alk mRNA levels were ∼60% using a previously defined dose regimen (20) beginning at 10 wk lower in RT2 C3H tumors vs. RT2 B6 tumors and ∼40% lower in of age when incipient tumors are first observed in RT2 mice (21). RT2 F1 tumors vs. RT2 B6 tumors, which was also reflected at RT2 B6 mice were used because they develop IC lesions at sig- the level (Fig. 3 A–C). Alk expression was also reduced in nificantly higher levels than RT2 C3H mice, and they also express WT islets from C3H mice as compared with B6 mice, consis- Alk in the pancreatic islets and PNETs at significantly higher tent with Alk being expressed at higher levels in the B6 back- levels than RT2 C3H mice (Figs. 1 D and E and 3 A–C). This is ground vs. the C3H background regardless of the neoplastic state also the stage of RT2 tumorigenesis when there is an appreciable of this tissue (Fig. 3B). Alk levels were higher in tumors com- increase in Alk expression levels (Fig. 3D). TAE684 was well pared with WT islets in both genetic backgrounds, and Alk ex- tolerated, and we did not observe any fluctuations in body mass pression showed a progressive increase during the course of RT2 in either TAE684- or vehicle-treated mice during the course of tumorigenesis (Fig. 3 B and D). Notably, there are no poly- the trial (Fig. S6). morphisms in the exonic regions of the Alk gene that differen- At the defined endpoint of the trial, TAE684-treated mice tiate the B6 allele from the C3H allele (http://www.informatics. proved to have developed ∼25% fewer macroscopic tumors than jax.org/strains_SNPs.shtml), and therefore the Alk protein is not control mice (Fig. 4B); there was a concomitant trend toward intrinsically different in structure or function in these different reduced tumor burden in TAE684-treated mice, which, however, genetic backgrounds. Interestingly, Alk belongs to the insulin- was not statistically significant (Fig. 4A). receptor superfamily of receptor tyrosine kinases (18), members Notably, TAE684-treated mice developed significantly fewer of which are known to influence PNET tumorigenesis in RT2 invasive lesions than control mice. There was a clear reduction mice, including tumor invasion (9, 19). Given this association in the frequency of total IC tumors (49.7% vs. 33.3% of total and our observation that Alk expression levels were significantly tumors in control vs. treated mice), which was accompanied by different between the B6 and C3H backgrounds, we sought to a concomitant increase in the frequency of IT tumors (50.3% vs.

Fig. 3. Anaplastic lymphoma kinase (Alk) is differentially expressed in the B6 and C3H genetic backgrounds. (A) Real-time quantitative PCR values for Alk, Dlgap1, Emilin2, Lbh, Ltbp1, Rab31, and Spdya obtained using a TaqMan array to profile tumors isolated from RT2 mice on the B6 and C3H backgrounds. Seven independent tumors per genotype were analyzed in triplicate. Data shown are mean plus SE. *P < 0.05 and **P < 0.01 by the Mann–Whitney test. (B) Real-time quantitative PCR values for Alk in a pool of islets isolated from normal WT B6 or C3H mice or in tumors (T) isolated from RT2 mice on the B6, C3H,or F1 backgrounds. *P < 0.05 by the Mann–Whitney test. (C) Western analysis on tumor pool lysates (T) from RT2 B6 and RT2 C3H mice for Alk and β-actin.(D) Real-time quantitative PCR values for Alk during the stages of RT2 tumorigenesis [normal WT (N), hyperplastic (H), angiogenic (A), tumor (T)] on the B6 genetic background. All real-time quantitative PCR values are shown as the percentage expression of the ribosomal protein L19 (L19).

17270 | www.pnas.org/cgi/doi/10.1073/pnas.1012705107 Chun et al. Downloaded by guest on September 25, 2021 Fig. 4. Alk inhibitor NVP-TAE684 (TAE684) reduces tumor invasiveness in RT2 mice in an experimental therapeutic trial. (A) Tumor burden for RT2 B6 mice treated with TAE684 or vehicle from 10 to 14 wk of age. Data shown are mean plus SE. Data are not statistically different. (B) Tumor number for RT2 B6 mice treated with TAE684 or vehicle from 10 to 14 wk of age. Data shown are mean plus SE. *P < 0.05 by the Mann–Whitney test. (C) Quantification of tumor invasiveness represented as the percentage of IT tumors or total IC tumors (IC1 + IC2) in RT2 B6 mice treated with TAE684 or vehicle from 10 to 14 wk of age. A minimum of 83 tumors per group was graded. *P < 0.05 by Fisher’s exact test. (D) Same as C except IC tumors are separated into the IC1 and IC2 subclasses. *P < 0.01 by the χ2 test. (E) Western analysis on tumor pool lysates from TAE684- or vehicle-treated RT2 B6 mice for phospho-Alk, total Alk, phospho-Akt, total Akt, and β-actin.

66.7% of total tumors in control vs. treated mice), in TAE684- backgrounds, suggesting that this locus is of a class of genetic GENETICS treated mice (Fig. 4C). This shift was due to a reduction in the modifiers that is not altered during tumorigenesis. frequencies of both the IC1 and IC2 subclasses of invasive RT2 The invasion modifier locus on chromosome 17 (63.7–76.4 PNETs (Fig. 4D). Mb) contains more than 50 annotated genes. Additionally, one TAE684 functions by interfering with Alk kinase activity (20), miRNA, mir-1195, resides in this locus, although there is no and tumors from treated RT2 mice showed reduced levels of coding change between the B6 and C3H sequences for this phosphorylated Alk (Fig. 4E). We also observed a modest but miRNA (http://www.informatics.jax.org/strains_SNPs.shtml). Of appreciable reduction in the levels of phosphorylated Akt, one the 50 genes in the modifier locus, 7 were found to be differen- downstream Alk target, compared with controls (Fig. 4E), con- tially expressed in the PNETs isolated from RT2 mice inbred into fi firming that TAE684 inhibited Alk activity in the tumors of the B6 and C3H backgrounds. As a rst step toward auditing fi RT2 mice. candidate invasion modi er genes in this locus, we focused on the Alk receptor tyrosine kinase, motivated in part by a series of Discussion studies demonstrating that Alk is activated by mutation or chro- A considerable body of research has identified polymorphic mosomal translocation in human hematopoietic and solid can- – modifier loci scattered across the mouse genome that affect mul- cers, evidently converting it into an initiating oncogene (18, 25 tiple aspects of cancer susceptibility and development (1, 22, 23). 27). On the basis of these and previous studies implicating Alk as fi Our data demonstrate that tumor progression, specifically to an an oncogene, several small-molecule inhibitors speci ctoAlk invasive growth state, is also subject to polymorphic genetic con- have been developed as potential therapeutics for these diseases trol. We identify a polymorphic locus on mouse chromosome 17 (20, 28). Our use of one such kinase inhibitor to probe the pos- [syntenic to human 2 (107.5–110 Mb), 5 (2.5–10 sible roles of Alk in PNET tumorigenesis demonstrated that Alk promoted both tumor growth and progression; most notably, Mb), and 18 (29–34 Mb)], which influences the susceptibility of pharmacological inhibition of Alk activity reduced tumor in- PNETs to progress from solid adenomatous tumors to invasive vasiveness in RT2 B6 mice. These results are consistent with our carcinomas. observation that Alk is expressed at lower levels in the tumors of Using a prototypical mouse model of multistage tumorigenesis, RT2 C3H mice, which are rarely invasive, as compared with the we observed that the propensity to develop an invasive phenotype tumors of RT2 B6 mice, which consistently develop invasive RT2 B6 is affected by genetic background. mice inbred into the PNETs. In comparing the B6 and C3H sequences, we did not background develop PNETs of varying degrees of invasiveness, identify any polymorphism in either the protein-coding or un- whereas RT2 mice inbred into the C3H background are largely translated portions of the Alk mRNA that might suggest a basis resistant to the development of invasive tumors. Furthermore, for Alk’s invasion modifier effects and/or differential expression. RT2 F1 hybrid mice are also resistant, indicating that the C3H However, there are four polymorphisms located within 10 kb of genetic background is dominant-suppressive over the invasion- the 5′-flanking region and two within 10 kb of the 3′-flanking prone B6 background. Linkage analysis of RT2 N2 backcross region, in addition to ∼300 polymorphisms residing in the large mice, produced from backcrossing RT2 F1 mice once to the sus- intron 2 of the Alk gene, that distinguish the B6 and C3H alleles ceptible B6 background, identified a locus on chromosome 17 that (http://www.informatics.jax.org/strains_SNPs.shtml), and one or correlated with susceptibility (when the locus is homozygous B6) more of these polymorphisms may account for the observed dif- vs. resistance (when a C3H allele is present). Previous studies have ferences in allelic expression. Our results associating Alk with documented that tumors isolated from RT2 mice undergo chro- invasion are also congruent with a previous study demonstrating mosomal gains and losses at different frequencies dependent on that single-chain variable fragment antibodies targeting Alk can genetic background (10, 24). Notably, chromosome 17 is not af- reduce tumor cell invasion in an in vitro setting (29). Additionally, fected by copy number abnormalities in either the B6 or C3H pharmacological inhibition of Alk hindered tumor formation in

Chun et al. PNAS | October 5, 2010 | vol. 107 | no. 40 | 17271 Downloaded by guest on September 25, 2021 RT2 mice, in accordance with earlier studies examining the on- creas but resistant for Hras-induced cancers in the skin. A major cogenic properties of Alk (20, 25, 26, 30). Importantly and in determinant of skin tumor resistance is a polymorphism in the contrast to the aforementioned studies in which Alk was the Patched gene (Ptch1), located on mouse chromosome 13, that driving oncogene, our results demonstrate that Alk can also act as introduces a nonconservative coding sequence change at the C a tumor progression factor, being up-regulated during multistep terminus of the protein (37). This polymorphism was not de- tumorigenesis to collaborate with an initiating oncogene (in tected in the present linkage analysis of invasive pancreatic tu- this case the SV40 Tag oncogene that abrogates the tumor- mors. Therefore, both tumor types are governed by polymorphic suppressing activities of pRb and p53). Thus, Alk inhibition modifiers of invasive cancer, albeit distinctive ones. Additionally, may prove to be a useful therapy even in situations in which Alk is yet other phenotypic modifiers of metastasis are implicated in not the initiating oncogene, either as a result of mutation or mouse models of breast cancer (40) and in human breast cancer other means. (41). Given the neuroendocrine nature of the tumor type subject Although our data implicate Alk levels as a determinant of to the invasion modifier reported herein, we wonder whether RT2 tumor invasion, we envision that other polymorphic invasion similar tumor types such as small-cell lung cancer or brain cancers modifier genes may reside in the chromosome 17 locus. The Alk might also be affected by this genetic modifier. Interestingly, Alk inhibitor reduced tumor invasiveness, but not to the degree seen has been implicated in glioblastoma (29), and as such, this tumor in the C3H background, which could reflect incomplete Alk in- type could be subject to this polymorphic modifier. hibition or additional genetic components to the modifier effect. Assessing the existence of polymorphic invasion modifiers Indeed, several other genes residing in this locus also showed in human cancers will be challenging. The availability of in- significant differential expression in RT2 tumors from the B6 and creasingly cost-effective DNA sequencing of individual genomes C3H genetic backgrounds (Fig. 3), and one of these genes, Ltbp1, (both normal and cancerous) may afford inroads to identifying contains a nonsynonymous coding change between the B6 and polymorphisms correlating with progression to invasive carcino- C3H backgrounds (http://www.informatics.jax.org/strains_SNPs. mas, particularly in organs in which both noninvasive adenomas shtml). Ltbp1 encodes the latent TGF-β binding protein 1, a com- and invasive carcinomas are prevalent, such as the colon. Elu- ponent of the TGF-β pathway (31), which is known to influence cidation of such polymorphic modifiers could well contribute to many aspects of cancer progression, including tumor invasion the future of personalized medicine, whereby susceptibility vs. and metastasis (32). Additionally, it has recently been suggested resistance alleles of invasion modifiers might be factored into the that Emilin2, which encodes the elastin microfibril interfacer 2, is treatment for patients diagnosed with early-stage cancers. subject to DNA methylation leading to reduced gene expression in human breast cancers, and Emilin2 hypermethylation is associated Materials and Methods with poorer clinical outcome, in particular relapse and poor sur- Genetically Engineered Mice. The generation and characterization of the RT2 vival (33). Last, elevated expression of Spdya (also known as Spy1), mouse line has been previously described (8). The RT2 line has been back- which encodes the speedy homolog A, accelerates tumorigenesis crossed into the B6 (Charles River Laboratories) and the C3H (Jackson Lab- in a mouse model of breast cancer (34) and has also been associ- oratory) genetic backgrounds more than 20 times and is effectively inbred ated with more aggressive human breast cancers (35). As such, into these backgrounds. F1 hybrid mice and RT2 N2 mice were generated as described (Fig. S4). Beginning at 10 wk of age, all RT2 mice received 50% other genes in this locus merit future investigation. – fl sugar food (Harlan Teklad) to relieve the effects of hypoglycemia caused by Although bone marrow derived in ammatory cells have been the insulin-secreting tumors. All mice used in this study were housed and shown to contribute to the invasiveness of RT2 PNETs (12, 13), it maintained in accordance with the University of California, San Francisco does not seem that their activity is modulated by the invasion institutional guidelines governing the care of laboratory mice. modifier gene(s). Thus, invasive PNETs were still rare in RT2 F1 mice that received bone marrow from an invasion-permissive B6 Genomic DNA Preparation, SNP Genotyping, and Linkage Analysis. Genomic donor (Fig. S3). Although we cannot rule out the possibility that DNA was isolated from mouse tails by proteinase K (Qiagen) digestion fol- this modifier locus operates in other stromal cell types or in lowed by phenol-chloroform extraction using standard methods (42). Four another tissue compartment, it seems most likely that the in- micrograms of genomic DNA per animal was SNP genotyped using the ’ fi vasive modifier acts in the cancer cells. Illumina platform according to the manufacturer s speci cations, and 143 RT2 N2 mice for which tumor phenotype data were available were geno- In addition to proinvasive inflammatory cells, other factors are fl typed by this method. A panel of primers that discriminate between the B6 known to in uence progression to an invasive growth state in this and C3H genetic backgrounds across all 19 somatic chromosomes and the X prototypical model of multistage tumorigenesis. Loss of cell–cell chromosome was used (see Dataset S1 for a complete list of SNPs used in this adhesion complexes, including the adherens junctions mediated study). Statistical analysis was performed using R/qtl (http://www.rqtl.org/) by Cdh1 (11) and desmosomes (36), are associated with the de- for the IT, IC1, IC2, tumor number, and tumor burden metrics. Because these velopment of more-invasive tumors. Signaling through the type-1 metrics were not normally distributed, nonparametric tests were chosen in insulin-like growth factor receptor (Igf1r) can also drive pro- R/qtl. A LOD score of ≥3.0 was considered significant (15). gression to an invasive state (9). The present study now establishes a unique dimension to this multifactorial invasive growth pheno- TAE684 Inhibitor Trial. The characterization of the Alk inhibitor NVP-TAE684 type, involving a polymorphic genetic modifier that can alterna- (TAE684) has been previously described (20). TAE684 was resuspended in 10% 1-methyl-2-pyrrolidinone/90% PEG 300 (vol/vol) (Sigma), and male RT2 tively override or allow these other functional effectors of invasive B6 mice were administered a 10-mg/kg dose once daily or vehicle solution growth. It remains to be determined whether the chromosome 17 fi fi alone by oral gavage from 10 to 14 wk of age. Body mass was monitored invasion modi er locus identi ed in this study modulates any of twice weekly to adjust dosing levels and to assess any toxicity caused by these functionalities or acts in a completely independent fashion. the treatment. Finally, it is pertinent to consider the translational implica- tions of this newly identified invasion modifier. First, we suspect ACKNOWLEDGMENTS. We thank the Mutation Mapping and Developmen- that this polymorphic modifier will prove operative in other tal Analysis Project (Harvard Medical School, Boston, MA; National Institute of Child Health and Human Development Grant U01-HD43430), the Partners cancer types but most likely not in all. Notably, the develop- Healthcare Center for Genetics and Genomics (Harvard Medical School, ment of squamous carcinoma is under distinctive polymorphic Boston, MA), and the University of California, San Francisco (UCSF) Helen control in mice. In this case, the B6 background is largely resis- Diller Family Comprehensive Cancer Center Genomics Core (San Francisco, tant to the development of invasive squamous carcinomas in three CA) for genotyping and genomics services; the UCSF Diabetes Center — Microscopy and Islet Isolation Cores (San Francisco, CA); Nathanael Gray different oncogenic contexts an activated Hras oncogene (37), (Dana-Farber Cancer Institute, Boston, MA) for the TAE684; Stephan Morris the HPV16 oncogenes (38), and chemical carcinogens (39). Thus, (St. Jude Children’s Research Hospital, Memphis, TN) for advice on Alk anti- the B6 background is permissive for invasive cancers in the pan- bodies; Anguraj Sadanandam (Swiss Federal Institute of Technology Lau-

17272 | www.pnas.org/cgi/doi/10.1073/pnas.1012705107 Chun et al. Downloaded by guest on September 25, 2021 sanne, Lausanne, Switzerland) for bioinformatic analysis of the modifier search Professor. This work was supported by National Cancer Institute Grant locus; Susan Cacacho, Ehud Drori, I. Celeste Rivera, Marina Vayner, and 5R01CA45234-24 (to D.H.) and by a graduate research fellowship from the Annie Wang for superior technical support; and Matthias Hebrok, Martin National Science Foundation (to M.G.H.C). A.B. and J.-H.M. were supported McMahon, and members of the D.H. laboratory for advice and encourage- by National Cancer Institute Grant U01-CA84244 and Department of Energy ment at all stages of this project. D.H. is an American Cancer Society Re- Grant DE-FG02-03ER63630.

1. Balmain A, Gray J, Ponder B (2003) The genetics and genomics of cancer. Nat Genet 21. Bergers G, Javaherian K, Lo KM, Folkman J, Hanahan D (1999) Effects of angiogenesis 33(Suppl):238–244. inhibitors on multistage carcinogenesis in mice. Science 284:808–812. 2. Easton DF, et al.; SEARCH collaborators; kConFab; AOCS Management Group (2007) 22. Demant P (2003) Cancer susceptibility in the mouse: Genetics, biology and Genome-wide association study identifies novel breast cancer susceptibility loci. implications for human cancer. Nat Rev Genet 4:721–734. Nature 447:1087–1093. 23. Quan L, Hutson A, Demant P (2010) A locus on chromosome 8 controlling tumor 3. Tomlinson IP, et al.; CORGI Consortium; EPICOLON Consortium (2008) A genome-wide regionality—a new type of tumor diversity in the mouse lung. Int J Cancer 126: association study identifies colorectal cancer susceptibility loci on chromosomes 10p14 2603–2613. and 8q23.3. Nat Genet 40:623–630. 24. Hodgson G, et al. (2001) Genome scanning with array CGH delineates regional 4. Park YG, et al. (2005) Sipa1 is a candidate for underlying the metastasis efficiency alterations in mouse islet carcinomas. Nat Genet 29:459–464. modifier locus Mtes1. Nat Genet 37:1055–1062. 25. George RE, et al. (2008) Activating mutations in ALK provide a therapeutic target in 5. Lifsted T, et al. (1998) Identification of inbred mouse strains harboring genetic neuroblastoma. Nature 455:975–978. modifiers of mammary tumor age of onset and metastatic progression. Int J Cancer 26. Mossé YP, et al. (2008) Identification of ALK as a major familial neuroblastoma 77:640–644. predisposition gene. Nature 455:930–935. 6. Nagase H, et al. (1995) Distinct genetic loci control development of benign and 27. Morris SW, et al. (1994) Fusion of a kinase gene, ALK, to a nucleolar protein gene, malignant skin tumours in mice. Nat Genet 10:424–429. NPM, in non-Hodgkin’s lymphoma. Science 263:1281–1284. 7. Sequist LV, Bell DW, Lynch TJ, Haber DA (2007) Molecular predictors of response to 28. Li R, Morris SW (2008) Development of anaplastic lymphoma kinase (ALK) small- epidermal growth factor receptor antagonists in non-small-cell lung cancer. J Clin molecule inhibitors for cancer therapy. Med Res Rev 28:372–412. Oncol 25:587–595. 29. Stylianou DC, et al. (2009) Effect of single-chain antibody targeting of the ligand- 8. Hanahan D (1985) Heritable formation of pancreatic beta-cell tumours in transgenic binding domain in the anaplastic lymphoma kinase receptor. Oncogene 28:3296– mice expressing recombinant insulin/simian virus 40 oncogenes. Nature 315:115–122. 3306. 9. Lopez T, Hanahan D (2002) Elevated levels of IGF-1 receptor convey invasive and 30. Piva R, et al. (2006) Ablation of oncogenic ALK is a viable therapeutic approach for metastatic capability in a mouse model of pancreatic islet tumorigenesis. Cancer Cell anaplastic large-cell lymphomas. Blood 107:689–697. 1:339–353. 31. Oklü R, Hesketh R (2000) The latent transforming growth factor beta binding protein 10. Hager JH, et al. (2004) Oncogene expression and genetic background influence the (LTBP) family. Biochem J 352:601–610. frequency of DNA copy number abnormalities in mouse pancreatic islet cell 32. Massagué J (2008) TGFbeta in Cancer. Cell 134:215–230. carcinomas. Cancer Res 64:2406–2410. 33. Hill VK, et al. (2010) Identification of 5 novel genes methylated in breast and other

11. Perl AK, Wilgenbus P, Dahl U, Semb H, Christofori G (1998) A causal role for E- epithelial cancers. Mol Cancer 9:51. GENETICS cadherin in the transition from adenoma to carcinoma. Nature 392:190–193. 34. Golipour A, et al. (2008) The Spy1/RINGO family represents a novel mechanism 12. Gocheva V, et al. (2010) IL-4 induces cathepsin protease activity in tumor-associated regulating mammary growth and tumorigenesis. Cancer Res 68:3591–3600. macrophages to promote cancer growth and invasion. Genes Dev 24:241–255. 35. Zucchi I, et al. (2004) Gene expression profiles of epithelial cells microscopically 13. Gocheva V, et al. (2006) Distinct roles for cysteine cathepsin genes in multistage isolated from a breast-invasive ductal carcinoma and a nodal metastasis. Proc Natl tumorigenesis. Genes Dev 20:543–556. Acad Sci USA 101:18147–18152. 14. Joyce JA, Freeman C, Meyer-Morse N, Parish CR, Hanahan D (2005) A functional 36. Chun MG, Hanahan D (2010) Genetic deletion of the desmosomal component heparan sulfate mimetic implicates both heparanase and heparan sulfate in tumor desmoplakin promotes tumor microinvasion in a mouse model of pancreatic angiogenesis and invasion in a mouse model of multistage cancer. Oncogene 24: neuroendocrine carcinogenesis. PLoS Genetics, in press. 4037–4051. 37. Wakabayashi Y, Mao JH, Brown K, Girardi M, Balmain A (2007) Promotion of Hras- 15. Lander E, Kruglyak L (1995) Genetic dissection of complex traits: Guidelines for induced squamous carcinomas by a polymorphic variant of the Patched gene in FVB interpreting and reporting linkage results. Nat Genet 11:241–247. mice. Nature 445:761–765. 16. Quigley DA, et al. (2009) Genetic architecture of mouse skin inflammation and 38. Coussens LM, Hanahan D, Arbeit JM (1996) Genetic predisposition and parameters of tumour susceptibility. Nature 458:505–508. malignant progression in K14-HPV16 transgenic mice. Am J Pathol 149:1899–1917. 17. Hubner N, et al. (2005) Integrated transcriptional profiling and linkage analysis for 39. Hennings H, et al. (1993) FVB/N mice: An inbred strain sensitive to the chemical identification of genes underlying disease. Nat Genet 37:243–253. induction of squamous cell carcinomas in the skin. Carcinogenesis 14:2353–2358. 18. Chiarle R, Voena C, Ambrogio C, Piva R, Inghirami G (2008) The anaplastic lymphoma 40. Winter SF, Hunter KW (2008) Mouse modifier genes in mammary tumorigenesis and kinase in the pathogenesis of cancer. Nat Rev Cancer 8:11–23. metastasis. J Mammary Gland Biol Neoplasia 13:337–342. 19. Ulanet DB, Ludwig DL, Kahn CR, Hanahan D (2010) Insulin receptor functionally 41. Hsieh SM, Look MP, Sieuwerts AM, Foekens JA, Hunter KW (2009) Distinct inherited enhances multistage tumor progression and conveys intrinsic resistance to IGF-1R metastasis susceptibility exists for different breast cancer subtypes: a prognosis study. targeted therapy. Proc Natl Acad Sci USA 107:10791–10798. Breast Cancer Res 11:R75. 20. Galkin AV, et al. (2007) Identification of NVP-TAE684, a potent, selective, and 42. Sambrook J, Russell DW (2001) Molecular Cloning: A Laboratory Manual (Cold Spring efficacious inhibitor of NPM-ALK. Proc Natl Acad Sci USA 104:270–275. Harbor Lab Press, Cold Spring Harbor, NY).

Chun et al. PNAS | October 5, 2010 | vol. 107 | no. 40 | 17273 Downloaded by guest on September 25, 2021