Both p16^Ink4a and the p19^Arf-p53 pathway constrain progression of pancreatic adenocarcinoma in the mouse

Abstract

Activating KRAS mutations and p16^Ink4a inactivation are near universal events in human pancreatic ductal adenocarcinoma (PDAC). In mouse models, Kras^G12D initiates formation of premalignant pancreatic ductal lesions, and loss of either Ink4a/Arf (p16^Ink4a/p19^Arf) or p53 enables their malignant progression. As recent mouse modeling studies have suggested a less prominent role for p16^Ink4a in constraining malignant progression, we sought to assess the pathological and genomic impact of inactivation of p16^Ink4a, p19^Arf, and/or p53 in the Kras^G12D model. Rapidly progressive PDAC was observed in the setting of homozygous deletion of either p53 or p16^Ink4a, the latter with intact germ-line p53 and p19^Arf sequences. Additionally, Kras^G12D in the context of heterozygosity either for p53 plus p16^Ink4a or for p16^Ink4a/p19^Arf produced PDAC with longer latency and greater propensity for distant metastases relative to mice with homozygous deletion of p53 or p16^Ink4a/p19^Arf. Tumors from the double-heterozygous cohorts showed frequent p16^Ink4a inactivation and loss of either p53 or p19^Arf. Different genotypes were associated with specific histopathologic characteristics, most notably a trend toward less differentiated features in the homozygous p16^Ink4a/p19^Arf mutant model. High-resolution genomic analysis revealed that the tumor suppressor genotype influenced the specific genomic patterns of these tumors and showed overlap in regional chromosomal alterations between murine and human PDAC. Collectively, our results establish that disruptions of p16^Ink4a and the p19^ARF-p53 circuit play critical and cooperative roles in PDAC progression, with specific tumor suppressor genotypes provocatively influencing the tumor biological phenotypes and genomic profiles of the resultant tumors.

Pancreatic ductal adenocarcinoma (PDAC) ranks as the fourth leading cause of cancer mortality in the United States and causes >200,000 deaths worldwide annually (1, 2). Histopathological analyses have identified precursor lesions, pancreatic intraepithelial neoplasias (PanIN), which appear to progress through increasingly severe stages of cellular atypia leading to invasive PDAC (3). These lesions show multistep molecular progression that includes early activating KRAS mutations and telomere attrition, and subsequent inactivation of p16^Ink4a, p14^ARF, p53, and/or SMAD4 tumor suppressors in a high percentage of cases (4–6).

The Ink4a/Arf locus (hereafter denoted p16^Ink4a/p19^Arf) encodes tumor suppressors p16^INK4A and p14^ARF (p19^Arf in the mouse). p16^INK4A is a G₁ cyclin-dependent kinase (CDK) inhibitor that binds to CDK4 and CDK6 and prevents their association with D-type cyclins (7), thereby facilitating CDK4/6-cyclin D-mediated phosphorylation and inactivation of retinoblastoma protein (RB) and S-phase entry. p16^INK4A-mediated tumor suppression may relate to its induction by activated oncogenes and consequent promotion of oncogene-induced senescence (8, 9). p14^ARF inhibits MDM2-mediated degradation of p53 (10, 11); thus, loss of p14^ARF results in reduced p53 protein levels (12). Mounting evidence suggests that p14^ARF also possesses p53-independent functions including the inhibition of ribosomal RNA processing (13, 14).

The central role of p16^INK4A in PDAC is evidenced by its inactivation in 80–95% of sporadic cases (15) and by increased PDAC risk associated with germ-line p16^INK4A mutations (16, 17). Whereas mutations exclusively targeting p16^INK4A and sparing p14^ARF have been identified in human PDAC, p14^ARF-specific mutations have not been reported. However, the pathogenetic relevance of p14^ARF is suggested by the occurrence of homozygous p16^INK4A/p14^ARF deletions in a subset of PanIN lesions, as well as in ≈40% of PDAC (6, 18). An important unresolved issue is the extent to which this correlation reflects the functional benefits of eliminating both p16^INK4A and p14^ARF, or rather a bystander phenomenon whereby p14^ARF loss occurs as a consequence of targeting the overlapping p16^Ink4a coding sequences.

p53 regulates target genes governing diverse tumor suppressor processes (19). p53 is mutated in 50–75% of human PDAC coupled with loss of the remaining WT allele (6). These mutations typically occur in advanced PanIN lesions that have previously incurred KRAS activation and p16^INK4A loss (20, 21). p53 mutations and p14^ARF deletions coexist in ≈38% of human PDAC cases (6, 18, 20, 22). Although such data may imply nonoverlapping tumor suppressor roles for these proteins, the distinct requirements for p53 vs. p14^ARF inactivation in PDAC development have not been explored by genetic means.

The frequency and temporal occurrence of mutations in KRAS, p16^INK4A/p19^ARF and p53 in human PanIN and PDAC support the view that activated KRAS cooperates with defects in the RB and p53 tumor suppressor pathways to drive the initiation and progression of the disease. This hypothesis has received additional support from genetically engineered mouse models of PDAC. Endogenous Kras^G12D expression in the mouse pancreas promotes development of PanINs (23, 24) that can progress to PDAC after a long latency (23). Furthermore, pancreas-specific expression of Kras^G12D in the setting of homozygous deficiency at the p16^Ink4a/p19^Arf locus results in rapid advancement of PanIN to invasive PDAC (24). These observations indicate that activated Kras initiates PanIN and p16^Ink4a/p19^Arf-deficiency promotes PanIN-to-PDAC progression. The relative contributions of p16^Ink4a and p19^Arf to tumorigenesis was not addressed in this model.

A recent model combining Kras^G12D and p53^R172H alleles produced invasive PDAC, demonstrating that p53 normally functions to suppress the emergence of PDAC (25). Notably, a functional p16^Ink4a/p19^Arf locus was retained in the Kras^G12D p53^R172H tumors. Because both p53 and p16^INK4A are characteristically lost in human PDAC, these data suggest that there may be cross-species differences in the role of the p16^INK4A-RB pathway in cellular transformation (26) or, intriguingly, raise the possibility that antecedent loss of p53 function or the gain-of-function properties of the p53^R172H protein may serve to neutralize rate-limiting components of the RB pathway. Taken together, the aforementioned studies establish a critical need to genetically examine the cooperative contribution of p16^INK4A inactivation in PDAC progression.

Results

p53 or p16^Ink4a Cooperate to Constrain PDAC Progression.

To study the genetic requirements for PDAC progression, we crossed mice with Pdx1-Cre (27) and LSL-Kras^G12D alleles (28) and engineered null mutations in p53 and p16^Ink4a or in p16^Ink4a/p19^Arf (24, 29, 30). Each of these tumor suppressor mutations, alone or in combination, cooperated with Kras^G12D activation to promote invasive pancreatic cancers, although with differing tumor latencies and histopathological and genetic properties (Table 1). Hereafter, because all tumors analyzed possess Pdx1-Cre and LSL-Kras^G12D alleles, specimens are referred to by their tumor suppressor genotypes. p53 nullizygosity (p53^lox/lox) caused the most rapid progression, yielding lethal tumors by 8 weeks of age, which is comparable to the latency of homozygous p16^Ink4a/p19^Arf deletion (23). Notably, in this homozygous-null p53 model, progression kinetics were not appreciably altered by p16^Ink4a status (Table 1), suggesting either cross-species differences or that developmental inactivation of p53 may diminish the need for p16^Ink4a loss (see below). Thus, it is important to note that p16^Ink4a−/− animals, with WT p53 and p19^Arf germ-line status, also developed lethal pancreatic tumors (mean 18.3 weeks).

Table 1.

PDAC Incidence, Latency, and Histological Phenotype

Genotype	No. of tumors	Average latency, weeks	Metastasis, %	Histology			aCGH‡
				Adenocarcinoma	Sarcomatoid	Anaplastic	Kras	Myc
p16/p19lox/lox	27	8.5	11	48* (81)†	26 (37)	26 (70)	13/16	1/16
p16/p19lox/+	12	34.2	69	57 (57)	43 (43)	0 (0)	5/8	1/8
p53lox/lox;p16+/+	3	6.2	0	100 (100)	0 (33)	0 (33)	1/3	0/3
p53lox/lox;p16+/−	5	6.5	0	80 (100)	0 (0)	20 (100)	3/4	3/4
p53lox/lox;p16−/−	5	7.2	20	40 (80)	0 (0)	60 (100)	0/3	0/3
p53lox/+;p16+/+	3	21.8	33	100 (100)	0 (33)	0 (33)
p53lox/+;p16+/−	16	14.7	25	81 (88)	19 (19)	0 (44)	1/6	2/6
p53lox/+;p16−/−	4	13.1	25	75 (100)	25 (50)	0 (50)	—
p53+/+;p16−/−	3	18.3	33	0 (33)	100 (100)	0 (0)
KrasG12D	3	57	67	0 (33)	100 (100)	0 (0)

*Percentage of tumors in which the particular histology predominates (see Materials and Methods).

^†The percentage of tumors in which the particular histology is present in any proportion.

^‡Number of tumors with copy number gains of Kras or Myc.

Because dual inactivation of p53 and p16^Ink4a occurs in most human PDACs, we assessed Pdx1-Cre LSL-Kras^G12D mice with heterozygous mutations in either or both of these tumor suppressor genes. Tumor latency in p53^lox/+ p16^Ink4a+/− animals was significantly reduced compared with those with p53^lox/+ alone (see Table 1; 14.7 vs. 21.8 weeks), suggesting cooperative roles for p53 and p16^INK4A in PDAC suppression. Finally, with regard to the relative impact of p53 vs. p19^Arf in constraining PDAC progression, we observed that p16^Ink4a/p19^{Arf lox/+} mice developed PDAC as well, but with longer latency relative to p53^lox/+ p16^Ink4a+/− animals (34.2 vs. 21.8 weeks), suggesting that p53 functions as a more potent barrier to PDAC progression. Overall, the clinical presentation across the various PDAC models was similar and typified by weight loss and jaundice.

Influence of Genotype on Histopathological Presentation.

As in other LSL-Kras^G12D models (31), histological analysis revealed that the tumors in this study were carcinomas, predominantly ductal adenocarcinomas, defined by the presence of neoplastic glandular (ductal) cells in a dense fibrous stroma or were alternatively, sarcomatoid or anaplastic carcinoma variants, characterized respectively by spindle-cell morphology or by detached tumor cells with marked nuclear and cytoplasmic pleomorphism, occurring in conjunction with ductal differentiation. The distribution of histological phenotypes was different in the various genetic backgrounds. For example, the proportion of ductal adenocarcinomas was highest in the context of p53^lox genotypes (Fig. 1A and B). In the p53^lox/lox background, p16^Ink4a deficiency was associated with a higher frequency of tumors with anaplastic features (Table 1 and Fig. 1C). Nevertheless, mice with a p16^Ink4a deficiency but WT p53 germ-line status developed ductal adenocarcinomas with regions of sarcomatoid differentiation (Fig. 1 D–F). In compound heterozygous mice, ductal adenocarcinomas were more common in the p53^lox/+ p16^Ink4a+/− mice (81% vs. 57% in the p16^Ink4a/p19^{Arf lox/+} cohort) (Table 1 and Fig. 1 G and J), whereas sarcomatoid histology was more common in the p16^Ink4a/p19^{Arf lox/+} tumors (43% compared with 19% in the p53^lox/+ p16^Ink4a+/− tumors) (Table 1 and Fig. 1L). Finally, the anaplastic features often observed in tumors from homozygous p16^Ink4a/p19^{Arf lox/lox} mice (24) were absent in the heterozygous p16^Ink4a/p19^{Arf lox/+} cohort but were often present focally within ductal adenocarcinomas arising in the p53^lox/+ p16^Ink4a+/− colony (44%)(Table 1 and Fig. 1I). Immunohistochemical analysis confirmed the ductal phenotypes of the well differentiated tumors, with positive staining for ductal markers (CK-19 and DBA lectin) and negative for markers of acinar cell (chymotrypsin) or islet cell (insulin) lineages (Fig. 1 H and K and data not shown). Each model showed invasion and metastasis, although these features were more pronounced in the heterozygous models (Fig. 1 M–O). Overall, these results demonstrate significant roles for p16^Ink4a, p19^Arf, and p53 in suppressing pancreatic cancer development and suggest specific impact of mutations in these genes in regulating the differentiation state of the ensuing tumors, most prominently indicating that losses of components of the p16^Ink4a/p19^Arf locus facilitate the development of more poorly differentiated tumors.

Fig. 1.

Deficiency in p53 or p16^Ink4a cooperates with oncogenic Kras^G12D to produce PDAC. (A) Hematoxylin/eosin stain of a PDAC (T, tumor) arising in a p53^lox/lox p16^+/+ mouse. Note invasion of duodenum (Du). (Scale bar: A and D, 200 μm; I, N, and O, 100 μm; B, C, E–H, and J–L, 50 μm.) (B) High-magnification view of the tumor in A showing features of ductal adenocarcinoma. (C) Tumor from p53^lox/lox p16^−/− mouse showing poorly differentiated adenocarcinoma (yellow arrows) admixed with anaplastic epithelioid cells characterized by giant tumor nuclei (black arrows) and eosinophilic inclusions (arrowheads). (D) Invasive PDAC arising in Pdx1-Cre LSL-Kras^G12Dp16^Ink4a−/− mouse. (E) High-magnification of D showing ductal adenocarcinoma histology. (F) Another region of the tumor in E showing sarcomatoid differentiation. (G) PDAC from p53^lox/+;p16^Ink4a+/− mouse. (H) Positive staining of the tumor in G for the ductal marker cytokeratin 19. (I) Anaplastic histology in PDAC from the p53^lox/+;p16^+/−model. (J) Well differentiated p16^Ink4a/p19^{Arf lox/+} PDAC. (K) Positive staining of tumor in J for cytokeratin 19. (L) p16^Ink4a/p19^{Arf lox/+} tumor showing sarcomatoid histology. (M) p53^lox/+p16^+/− tumor invading the duodenum. (N and O) p16^Ink4a/p19^{Arf lox/+} tumors with metastases to the lung (N) and liver (LV; O).

Somatic Inactivation of PDAC Tumor Suppressor Genes.

Because our genetic data implicated germ-line lesions in p16^Ink4a, p19^Arf, and p53 in promoting PDAC progression, we determined the presence of somatic alterations in these genes in primary tumors and early passage tumor cell cultures. In the p53^lox/lox colony, elimination of p53 sequences was verified in all samples analyzed, and 10 of 10 tumors showed robust p19^Arf expression and WT p19^Arf coding sequences (Fig. 2 A and B and data not shown). Examination of p16^INK4A status showed that four of four p53^lox/lox p16^Ink4a+/+ tumor cell lines retained p16^INK4A expression, whereas two of four p53^lox/lox p16^Ink4a+/− tumors showed markedly reduced or absent p16^INK4A expression (Fig. 2B and data not shown). These observations, coupled with retention of p53 in all PDAC specimens from p16^Ink4a/p19^{Arf lox/lox} mice (24), suggest that p53 and p19^Arf serve critical yet largely redundant roles in constraining PDAC progression in the mouse and indicate that p16^Ink4a loss is not required for PDAC progression in the context of early p53 inactivation. As noted above, however, a cooperative tumorigenic effect of p53 and p16^Ink4a lesions was observed in the p53^lox/+ p16^Ink4a+/− model. Analysis of derivative tumor cultures from these animals revealed loss of the WT p53 allele in six of six cases (Fig. 2C and data not shown) and low/absent p16^INK4A expression in five of six cases (Fig. 2D, lanes 1–6). In contrast, all tumor cell lines expressed p19^Arf at comparable levels to the p53^lox/lox model (Fig. 2D). Consistent with the reduced/absent p16^INK4A expression, there was deletion of WT p16^Ink4a sequences in one of six tumor cultures (Fig. 2E, lane 5) and hypermethylation of the p16^Ink4a promoter in three of six cases, as determined by methylation-specific PCR (Fig. 2F, lanes 1, 4, and 5). Similar findings were obtained in a larger group of tumors that were not exposed to cell culture (see the supporting information, which is published on the PNAS web site).

Fig. 2.

Molecular analyses of PDAC cell lines and primary tumor specimens from mice with p53 and p16 mutant animals. (A) PCR reactions to detect the p53-WT (+) and p53^lox alleles (Upper) and p53-null (−) allele (Lower) in normal tissue from p53^lox/lox (lane 1) and p53^lox/+ (lane 7) mice and tumor cell lines (lanes 2–6). All tumor cell lines show only the p53-null allele. Germ-line p16^Ink4a status is indicated at the top. (B) Western blot for p16^Ink4a and p19^Arf expression in cell lines derived from p53^lox/lox mice with various p16^Ink4a genotypes. The negative and positive controls are in lanes 9 and 10, respectively. α-Tubulin is shown as a loading control. (C) PCR analysis of the p53⁺ and p53^lox alleles (Upper) and the p53⁻ (/) allele (Lower) demonstrates loss of the p53⁺ allele in all tumors from p53^lox/+ p16^+/− mice (lanes 2–6). Lane 1 shows the WT control specimen. (D) Western blot analysis shows low or absent p16^Ink4a expression in five of six tumor cell lines from p53^lox/+ p16^+/− mice (Right, lanes 1–6), whereas all retain p19^Arf expression. Negative controls are in lanes 7 and 12. p16^Ink4a expression in PDAC cell lines from p53^lox/lox p16^+/+ (lane 8) and p53^lox/lox p16^+/− (lane 10) mice is shown as a reference (Right). (E) PCR analysis of the p16^Ink4a− and p16^Ink4a+ alleles demonstrates LOH in one of six samples (lane 5). Normal tissue specimens are shown as controls for the p16^Ink4a+/+, p16^Ink4a−/−, and p16^Ink4a+/+ alleles (lanes 1–3). (F) Methylation-specific PCR assay to detect methylated CpG islands in the p16^Ink4a promoter region reveals hypermethylation in three tumor lines (lanes 1, 4, and 5). Upper and Lower show the methylated (M) and unmethylated (U) alleles, respectively. Negative and positive controls are in lanes 7 and 8.

Comparable molecular analysis was performed in the p16^Ink4a/p19^{Arf lox/+} model. As expected, low-passage p16^INK4A/p19^{Arf lox/+} tumor cultures uniformly lacked p16^INK4A and p19^Arf protein expression (Fig. 3A; n = 8), and seven of eight samples sustained loss of the WT p16^Ink4a/p19^Arf allele (Fig. 3B). p53 showed low baseline expression, whereas γ-IR led to increased p53 protein levels and Ser-15 phosphorylation consistent with a physically intact p53 allele (Fig. 3C). Additional molecular analysis was performed on two PDACs arising after 1 year in Pdx1-Cre LSL-Kras^G12D animals without predisposing tumor suppressor mutations. Early passage tumor cultures and primary tumor lysates showed lack of both p16^INK4A and p19^Arf expression by Western blot analysis (Fig. 3D and data not shown). Furthermore, extinction of expression was associated with homozygous loss of the p16^Ink4a/p19^Arf locus as detected by PCR analysis (Fig. 4E). Taken together, our results are in keeping with a model of PDAC tumor suppression in which specific barriers to Kras^G12D-directed tumorigenesis are provided by the p19^Arf-p53 and p16^Ink4 pathways.

Fig. 3.

Molecular analyses of PDAC cell lines from p16^Ink4a/p19^{Arf lox/+} and p16^INK4A/p19^Arf+/+ mice. (A) Western blot analysis p16^Ink4a and p19^Arf expression in lysates from tumor cell lines from p16^INK4A/p19^{Arf lox/+} mice. (B) PCR analysis of p16^INK4A/p19^Arf alleles in tumor cell lines from p16^Ink4a/p19^{Arf lox/+} mice shows only the recombined p16^INK4A/p19^Arf allele (Left, lanes 2–8). Normal tissue from p16^Ink4a/p19^{Arf lox/+} mice shows WT (+) and unrecombined (lox) alleles. Tumor no. 738 (Left, lane 6; Right, lane 2) shows a biallelic deletion of the entire p16^Ink4a/p19^Arf locus. (C) Western blot analyses of γ-irradiated tumor cell lines (+) shows induction of total p53 protein levels and phosphorylation of p53 on Ser-15, consistent with present and functional p53 protein. −, untreated. Cdk4 protein levels are shown as a loading control. (D) Western blot analysis shows absence of p16^Ink4a and p19^Arf expression in Pdx1-Cre LSL-Kras^G12D cell lines (lanes 1 and 2). Lanes 3 and 4 show positive and negative controls. (E) PCR analysis for the presence of p16^Ink4a/p19^Arf exon 2, p16^Ink4a exon 1, and cytokeratin 19 sequences in cell lines from Pdx1-Cre LSL-Kras^G12D mice (lanes 2 and 3) demonstrates biallelic deletion of p16^Ink4a/p19^Arf. Lane 1 shows normal control DNA.

Fig. 4.

aCGH analysis of mouse PDAC. (A) Hierarchical clustering of aCGH profiles reveals three distinct clusters. Sample names are given above the profile followed by an underscore and the genotype class (i.e., “106_6” indicates the profile for tumor no. 106, which is in genotype class 6. The genotype classes are: 1, p53^lox/lox;p16^{Ink4a +/+}; 2, p53^lox/lox;p16^{Ink4a +/−}; 3, p53^lox/lox;p16^{Ink4a −/−}; 4, p53^lox/+;p16^{Ink4a +/−}; 5, p16^Ink4a/p19^{Arf lox/+}; 6, p16^INK4A/p19^{Arf lox/lox}. (B and C) Focal amplifications of Kras2 in p16^Ink4a/p19^Arf-deficient (B) and Myc in p53-deficient (C) tumors are shown. Minimal common regions of amplification are shown by blue lines.

Genomic Aberrations in Mouse Models of PDAC.

The availability of classes of tumors that either retained or lost expression of p53, p16^INK4A, and p19^Arf enabled an assessment the impact of these lesions on the level of genomic instability and on the acquisition of specific regional gains or losses of chromosomes. To investigate the possible existence of clonal cooperating genetic lesions, we performed array-comparative genomic hybridization (aCGH) using a high-resolution oligonucleotide microarray on early passage PDAC cells lines from the p16^Ink4a/p19^{Arf lox/lox} (n = 16), p16^Ink4a/p19^{Arf lox/+} (n = 8), p53^lox/+ p16^Ink4a+/− (n = 6), and p53^lox/lox (n = 10) models. Intergenotype comparisons of overall genomic instability revealed subtle differences among the various PDAC models, with p53 mutant tumors trending toward an increased genomic complexity (see the supporting information). It is notable that analysis of eight human PCAC cell lines using the same aCGH platform revealed significantly higher numbers of chromosomal copy number alterations (CNAs) compared with the mouse tumors (P < 0.001; see the supporting information).

The aCGH analyses revealed distinct classes of profiles between the mouse genotypes. Hierarchical clustering of the aCGH profiles by using the Pearson correlation as a distance metric yielded three distinct clusters that largely segregated according to p16^Ink4a/p19^Arf or p53 status (Fig. 4A). p16^Ink4a/p19^Arf-deficient tumors mainly fell into two clusters (6/8 tumors in cluster 1 and 13/14 tumors in cluster 2), whereas p53-deficient specimens predominated in the final cluster (13/16 tumors in cluster 3). A notable feature of cluster 2 was copy number increases in chromosome 6 containing the Kras locus. Although gains of the entire chromosome were noted, in a number of cases, the alterations were highly focal, spanning <1 Mb, indicating that Kras was the target of this event (Fig. 4B). Amplification of chromosome 6 occurred with much greater frequency in p16^Ink4a/p19^Arf mutant tumors (75%) than in p53 mutant tumors (31%) (Table 2). Recurrent and genotype-specific alterations also occurred in the form of focal gains in chromosome 15 spanning the c-Myc locus (Fig. 4 C and Table 2; 31% of p53 mutant tumors vs. 8% of the p16^Ink4a/p19^Arf mutant tumors) and deletion of chromosome 4 in the vicinity of the p16^Ink4a/p19^Arf locus (Fig. 4A; 50% of tumors from p16^Ink4a/p19^{Arf lox/+} mice and one of six tumors from p53^lox/+ p16^Ink4a+/− mice). These genomic distinctions within the data set were validated by using an alternative unsupervised clustering approach (D. Carrasco, unpublished work; see supporting information). These approaches demonstrate that CNA patterns including amplifications of Kras and c-Myc effectively identify classes of tumors that reflect the underlying tumor suppressor genotype. Other CNAs were also documented in these mouse PDAC genomes, including recurrent gains on chromosomes 1q, 4q, 8q, 16q, 17q, and 18q as well as losses on chromosomes 9q, 10q, 14q, and 17q (Fig. 4A and supporting information). Importantly, several focal and recurrent CNAs, such as 16q amplification and 17q deletion, are syntenic to recurrent CNAs in sporadic human PDAC (32, 33). Thus, these CNAs provide both a measure of molecular validation of murine PDAC models and a potential means of facilitating identification of PDAC progression genes.

Table 2.

Gain/amplification of Kras2 and Myc in mouse PDAC

Genotype	Kras2, %	Myc, %
p53lox/lox (n = 10)	40	30
p53lox/+ (n = 6)	17	33
All p53 (n = 16)	31	31
p16Ink4a/p19Arf lox/lox (n = 16)	81	6
p16Ink4a/p19Arf lox/+ (n = 8)	63	13
All p16Ink4a/p19Arf (n = 24)	75	8

Both p16^Ink4a and Arf-p53 Are Barriers to Mouse PDAC Progression.

This study provides several lines of evidence establishing that both components of the p16^Ink4a/p19^Arf locus provide critical barriers to PDAC progression. Specifically, mice with homozygous mutations in either p53 or p16^Ink4a developed tumors with short latency; moreover, mice with combined heterozygous mutations in p53 and p16^Ink4a developed PDAC that predominantly lost expression of products of both genes but showed robust expression of p19^Arf. Tumors arising in p16^Ink4a/p19^{Arf lox/+} animals lost both products of this locus but retained p53. Two aspects of this analysis are notable. First, although sharp distinctions have been drawn between Arf and p53 function in several other models and p53 has Arf-independent activities and vice versa (12), our PDAC models demonstrate that the roles of these tumor suppressors overlap significantly with regard to PDAC progression in the mouse. A second observation relates to the role of p16^INK4A loss in the setting of p53 inactivation. Specifically, whereas heterozygous deletion of p53 and p16^Ink4a during Kras initiation begets subsequent inactivation of the WT alleles of both genes during PDAC progression, loss of p53 function in the context of intact p16^Ink4 is not associated with subsequent p16^INK4A loss (this study and ref. 25). Possible explanations for this discrepancy include: (i) early and/or uniform homozygous loss of p53 coincident with Kras activation enables bypass of p16^INK4A -mediated RB pathway tumor suppressor functions that would otherwise be operative; or (ii) there are species-specific differences in the tumor suppressor functions between human and mouse p16^INK4A and p53, as suggested by other studies (26). Irrespective, the double-heterozygous models reflect well the human situation in regard to the complimentary benefits of their concomitant loss and, thus, will provide a highly appropriate model system for studying aspects of PDAC progression.

Tumor Suppressor Gene Status Influences Tumor Biology.

Different combinations of tumor suppressor gene mutations, in conjunction with Kras^G12D expression, all promoted the progression of PanIN to PDAC but produced tumors with varying spectra of clinical and histological features. Although tumors in all genotypes showed extensive local invasion and micrometastases, gross metastases were only prominent in mice with engineered heterozygous tumor suppressor deletions (p16^Ink4a/p19^{Arf lox/+} or p53^lox/+ p16^{Ink4a lox/+} mice) but not in mice with engineered homozygous deletions. Because the homozygous animals rapidly developed a lethal tumor burden, often with multiple primary tumors, these results raise the possibility that factors relating to tumor latency rather than the specific p16^INK4A-RB or p19^Arf-p53 pathway lesions are primarily responsible for modulating metastatic behavior.

With respect the effect of genotype on tumor histology, the p53-deficient cohorts showed the higher prevalence of well differentiated ductal adenocarcinoma compared with the p16^Ink4a/p19^Arf-deficient animals (Table 1). Conversely, undifferentiated sarcomatoid histology was significantly reduced in p53-deficient models. These results are consistent with observations that Kras^G12D and p53^R172H alleles produce primarily ductal adenocarcinomas (25). Anaplastic carcinoma, although common in the homozygous models, was never the principal histological component in the heterozygous p53^lox/+ p16^Ink4a+/− tumors and was completely absent in the p16^Ink4a/p19^{Arf lox/+} model. In humans, ductal adenocarcinoma histology predominates, and the sarcomatoid and anaplastic subtypes are considered uncommon variants of PDAC with more aggressive clinical behavior, although they all appear to have comparable spectra of genetic lesions (34, 35). Our mouse models collectively recapitulate these different histologic variants, albeit at different frequencies than seen in spontaneous human tumors. Overall, these observations suggest that the set of tumor suppressor lesions strongly influences the cell differentiation phenotypes of the resulting tumors.

Recurrent Genomic Aberrations Arise During PDAC Progression.

Genomic analysis revealed an effect of tumor suppressor genotype on the profile of tumor-associated chromosomal alterations and only modest trends toward increased genomic instability in p53 mutant tumors on the basis of the number of CNAs per tumor. Notably, the degree of genomic instability occurring in each of these PDAC mouse models is significantly less than that observed in human specimens (this study and refs. 32 and 33). This may reflect a bias imposed through the use of genetically engineered mouse models (i.e., p53 loss and KRAS activation may obviate the need for many further cytogenetic aberrations) or cross-species differences in chromosome biology and structure, particularly telomeres (5, 36, 37). The patterns of regional alterations in chromosomal copy number differed between tumors of different genotypes, suggesting that chromosomal alterations may harbor genes in which altered expression may specifically cooperate with distinct tumor suppressor gene mutations. Notable copy number gains at chromosome 6 encompassing the Kras locus were more commonly observed in the p16^Ink4a/p19^Arf mutant tumors. This correlation may reflect the capacity of Kras to activate both the p16^INK4A and p19^Arf promoters leading to cellular senescence in vivo (8, 38). p16^INK4A/p19^Arf-deficient tumors may be especially permissive for Kras^G12D amplification, facilitating the cellular transformation associated with increased Kras signaling. In contrast, chromosome 15 gains overlapping the Myc locus were much more common in the p53 mutant tumors. The selective amplification of Myc in association with p53 mutation has been noted in genomic analysis of melanomas harboring either p53 or p16^Ink4a/p19^Arf mutations (39). In these contexts, Myc amplification may serve to bypass the inhibition of CDK4 activity and consequent cell-cycle arrest conferred by elevated p16^Ink4a expression (39, 40). It is notable that, in the subset of p53^lox/lox and p53^lox/+ tumors retaining robust p16^INK4A expression, other RB pathway components remained intact as evidenced by the absence of p16^Ink4a and Cdk4 mutations in full-length sequence analyses and the lack of perturbations in expression of RB and cyclin D1 (data not shown).

Amplifications of KRAS on chromosome 12q12 and MYC at 8q24 as well as deletions of p16^INK4A on 9p21 are commonly observed in human pancreatic adenocarcinoma (32, 33). In addition to CNAs harboring validated pancreatic cancer genes, we have also identified additional mouse CNAs that are syntenic to human PDAC loci (see supporting information). Recurrence of these loci in multiple different tumor specimens and their evolutionary conservation indicate that they may harbor cancer genes with prime importance in PDAC pathogenesis. Moreover, the targeting of syntenic loci in these mouse PDACs highlights the potential of mouse models of cancer to serve as effective filters in the identification of cancer genes present in complex human copy number data sets.

Materials and Methods

Mouse Strains, Histopathology, and Establishment of Primary PDAC Cell Lines.

The mouse strains in this study included the following alleles: LSL-Kras^G12D (41), Pdx1-Cre (27), conditional p53^lox (30), conditional p16^Ink4a/p19^{Arf lox} (24), and a germ-line p16^Ink4a-specific null allele (that retains WT p19^Arf) (29). All experiments were performed on >87.5% FVB/n background. Some mice with the p16^Ink4a/p19^{Arf lox} allele developed lymphomas (latency 28–60 weeks) that occurred independently of concurrent expression of Pdx1-Cre or LSL-Kras^G12D. The p16^Ink4a/p19^{Arf lox} allele shows intact expression and function of the p16^Ink4a/p19^Arf locus in cell culture-based assays in vitro, and it appears that the lymphomas arising in these mice are attributable to hypomorphic activity of p16^Ink4a/p19^{Arf lox} allele. In addition, a subset of mice from both the p16^Ink4a/p19^{Arf lox/+} and p53^lox/+ p16^Ink4a+/− colonies developed progressive wasting in the absence of evident neoplastic growth, whereas some others developed cutaneous papillomas. These phenotypes are likely due to the extrapancreatic expression of Kras^G12D induced by the Pdx1-Cre transgene because they are not observed in the context of the p48-Cre strain in which expression pattern is more tightly restricted to the pancreas (data not shown). Tissue processing, immunohistochemical staining, and establishment of primary PDAC cell lines were performed as described in ref. 24. The histological classification of each tumor reported in Table 1 was determined by evaluation of cross-sectional slides from two axes. The histological type representing the largest cross-sectional area was considered the primary component.

Molecular Analysis.

RNA and DNA isolation, protein extract preparation from PDAC cell lines, and primary tumors was performed as described in ref. 24. For loss of heterozygosity analysis of p16^Ink4a, p53, and p16^Ink4a/p19^Arf, DNA was extracted from the early passage cell lines and amplified by PCR. Methylation-specific PCR analysis was performed as described in ref. 42.

Genomic Analysis.

aCGH and data analysis was performed as described on Agilent mouse development or human 1A oligonucleotide microarrays (see supporting information) (32).

Abbreviations

PDAC

pancreatic ductal adenocarcinoma

PanIN

pancreatic intraepithelial neoplasias

CDK

cyclin-dependent kinase

RB

retinoblastoma protein

aCGH

array-comparative genomic hybridization

CNAs

copy number alterations.