Evidence for type II cells as cells of origin of K-Ras–induced distal lung adenocarcinoma

Abstract

Identifying the cells of origin of lung cancer may lead to new therapeutic strategies. Previous work has focused upon the putative bronchoalveolar stem cell at the bronchioalveolar duct junction as a cancer cell of origin when a codon 12 K-Ras mutant is induced via adenoviral Cre inhalation. In the present study, we use two “knock-in” Cre-estrogen receptor alleles to inducibly express K-RasG12D in CC10⁺ epithelial cells and Sftpc⁺ type II alveolar cells of the adult mouse lung. Analysis of these mice identifies type II cells, Clara cells in the terminal bronchioles, and putative bronchoalveolar stem cells as cells of origin for K-Ras–induced lung hyperplasia. However, only type II cells appear to progress to adenocarcinoma.

Over the past 10–15 y, it has become clear that there is considerable molecular heterogeneity among adenocarcinomas of the lung. This may relate to the different combinations of genes mutated in these tumors. Currently, 26 different genes have been implicated in lung adenocarcinomas (1), with 30% of tumors carrying activated K-RAS (2). Among tumors with K-RAS activation, another potential source of heterogeneity, with clinical implications for targeted therapy, is the cell type in which the mutation arose. For example, similar K-RAS mutations have been identified in pancreatic cancer, colon cancer, lung cancer, thyroid cancer, and myeloid leukemia (3). However, these tumors have markedly different histologies and behaviors. Such cell-of-origin effects have not been exhaustively examined for lung cancer in relation to the different cell types within the respiratory epithelium, and the cell(s) of origin of human lung adenocarcinoma has not been clearly identified.

One approach to this problem is to exploit some of the inducible mouse models of human lung cancer that have been established (4, 5). Based upon K-RasG12D activation by inhaled adenoviral Cre, it was proposed that the initiating cells for lung adenocarcinoma are putative bronchioalveolar stem cells (BASCs) at the bronchioalveolar duct junction (BADJ) (6). These cells are characterized as coexpressing Clara cell antigen 10 (CC10, Scgb1a1) and surfactant protein C (Sftpc). However, whether they function in vivo as stem cells for both the bronchioles and alveoli is not clear (7).

In the present study, we use two mouse lines with “knock-in” Cre recombinase -estrogen receptor fusion protein (CreER) alleles to activate K-Ras in different epithelial cells in the lung. The first line carries a CC10-CreER knock-in allele expressed in secretory (Clara) cells throughout the airways, in the BADJ, and in some alveolar cells (7). Although no K-Ras–induced tumors form in CC10-expressing cells in the airways, CC10⁺ cells undergo hyperplasia at the BADJ and give rise to tumors in the alveoli. The second line carries an Sftpc-CreER knock-in allele (Fig. S4A) expressed in alveolar type II cells and the majority of putative BASCs at the BADJ. By following the incorporation of BrdU in cells that have undergone recombination, we provide evidence that dual-positive CC10⁺, Sftpc⁺ putative BASCs are not the only cells that proliferate in the respiratory epithelium in response to K-Ras activation. Moreover, Sftpc⁺ cells that undergo recombination form adenocarcinomas in the alveoli but not at the BADJ. CC10-CreER; K-RasG12D tumors and Sftpc-CreER; K-RasG12D tumors have very similar transcriptome patterns. These results taken together demonstrate that multiple cells located near the BADJ and in the alveoli proliferate in response to K-RasG12D induction. Moreover, they raise the questions of why cells in the larger airways and nonterminal bronchioles do not develop neoplasia and why cells in the terminal bronchioles do not progress beyond hyperplasia.

Results

K-RasG12D Activation in CC10-Expressing Cells Leads to Hyperplasia at the BADJ and Tumor Formation in the Alveoli, but Not in the Larger Airways and Bronchioles.

To investigate whether oncogenic K-Ras can transform all or only a subset of CC10-expressing cells, we crossed mice carrying the CC10-CreER allele to Lox-stop-Lox (LSL) K-RasG12D mice (8). These mice were also heterozygous for p53 (Materials and Methods). When mice were 6–8 wk old, four doses of i.p. tamoxifen (tmx) were administered to induce recombination in CC10⁺ cells. As demonstrated previously (7), four consecutive doses of tmx label about 80% of CC10-expressing Clara cells, 90% of putative BASCs, and 10% of alveolar type II cells in CC10-CreER; Rosa26R-fGFP mice (Fig. 1 A–C). Significantly, after four doses of tmx, CC10-CreER; LSL K-RasG12D mice develop adenomas and subsequently adenocarcinomas in the alveoli (Fig. 1D). However, almost all larger bronchi and nonterminal bronchioles appear normal even though recombination has occurred throughout these regions.

Fig. 1.

K-RasG12D activation in CC10⁺ cells leads to hyperplasia at the BADJ and adenocarcinoma in the alveoli. (A–C) Sections of CC10-CreER; Rosa26R-fGFP mouse lung 1 wk after tmx injection. (A and B) Green, anti-GFP, lineage label; red, anti-CC10, Clara cells. More than 80% of airway Clara cells were lineage-labeled. (C) Green, anti-GFP, lineage label; purple, anti-Sftpc, type II cells. About 90% of putative BASCs (arrow) and 10% of alveolar type II cells (arrowhead) are double-labeled, as described (7), with higher magnification (Inset). (D) H&E section of CC10-CreER; LSL-K-RasG12D mouse lung 15 wk after tmx. Tumors arise predominantly near the BADJ and in the alveoli but not in the airways.

We analyzed tumors at different times after tamoxifen injection. As shown in Fig. 2 and previously reported by others (6), after 3 wk of K-Ras activation, an increase in the number of dual-positive cells is observed at the BADJ (Fig. 2A). In addition, small regions of hyperplasia are found in the alveoli (Fig. 2 B and C). These are Sftpc⁺ but do not stain with anti-CC10. At 9 wk after tmx, hyperplastic regions contain dual-positive cells at the BADJ (Fig. 2D). In contrast, the alveoli now contain much larger Sftpc⁺ adenomas (Fig. 2 E and F). The BADJ hyperplasia persists through 15 and 21 wk (Fig. 2 G and J). In the alveoli, Sftpc⁺ papillary adenomas (Fig. 2 H and I) are seen at 15 wk and advanced papillary adenocarcinomas (Fig. 2 K and L) at 21 wk. These results demonstrate that, despite expression of oncogenic K-RasG12D in CC10-expressing cells throughout the airways, CC10⁺ hyperplasia develops only at the BADJ, and Sftpc⁺ tumors form only in the alveoli. Thus, CC10-expressing cells at different locations have varying susceptibilities to transformation.

Fig. 2.

Time course of neoplasia at the BADJ and in alveoli. Sections of CC10-CreER; LSL-K-RasG12D mouse lung at various times after four doses of tmx were stained with anti-CC10 (red) and pro-Sftpc (green) (A, B, D, E, G, H, J, and K) or H&E (C, F, I, and L). (Insets) Cells marked by arrows. (A–C) Three weeks. (A) Rare CC10⁺, Sftpc⁺ cells (arrow) at the BADJ. (B) Alveolar hyperplasia consists of Sftpc⁺ cells (arrow). (C) Hyperplasia in alveoli (arrow). (D–F) Nine weeks. (D) BADJ dual-positive cells (arrow). (E and F) Alveolar Sftpc⁺ adenomas. (G–I) Fifteen weeks. (G) BADJ dual-positive cells (arrow). (H and I) Sftpc⁺ papillary adenoma. (J–L) Twenty-one weeks. (J) Dual-positive hyperplasia at BADJ (arrow). (K and L) Sftpc⁺ papillary adenocarcinoma.

At Early Times After CC10-Driven K-RasG12D Induction, Multiple Cell Types Proliferate at the BADJ and in the Alveoli.

To identify which CC10⁺ epithelial cells proliferate soon after K-RasG12D induction, we generated CC10-CreER; LSL-K-RasG12D; Rosa26R-fGFP compound mice and injected either one or four doses of tmx. Mice were killed at 24 h or 7 d after one dose and at 72 h after four doses. BrdU was injected 6 h before sacrifice to label proliferating cells. For the purposes of this analysis, we define the BADJ as the last 25 airway cells before the alveolar space is evident. Cells that are GFP⁺ are assumed to have expressed CC10 at the time they received tmx. However, it should be noted that in CC10-CreER; Rosa26R-fGFP mice and CC10-CreER; LSL-K-RasG12D; Rosa26R-fGFP mice exposed to tmx, immunoreactive CC10 cannot always be detected in GFP-labeled cells in the alveoli. This presumably reflects the very low level of CC10 expression in this region (7).

At 24 h after one dose of tmx, only 25 ± 3.3% of the BrdU⁺, GFP⁺ cells in the BADJ express Sftpc. Significantly, the majority (75 ± 3.3%) of the BrdU⁺, GFP⁺ cells are negative for Sftpc (Fig. 3 A–C). Most of these cells are surrounded by GFP⁺, BrdU⁻, Sftpc⁺ cells (Fig. 3A). This pattern was also seen at 7 d after one dose of tmx; 83 ± 3.2% of BrdU⁺, GFP⁺ cells are Sftpc⁻ and only 17 ± 3.2% are Sftpc⁺. These results indicate that even with a low dose of tmx, the majority of cells that proliferate in response to K-RasG12D activation are GFP⁺, Sftpc⁻ cells and only a minority are putative BASCs (GFP⁺, Sftpc⁺). Even at 72 h after four doses of tmx, 68 ± 5.5% of GFP⁺, BrdU⁺ cells at the BADJ are Sftpc⁻, and 32 ± 5.5% are Sftpc⁺ (Fig. 3C and Table S1). In the alveoli, very few cells incorporate BrdU at a low dose of tmx. After four doses, almost all BrdU⁺, GFP⁺ alveolar cells are Sftpc⁺ (Fig. 3 D–F). In the larger bronchi, smaller numbers of BrdU⁺ cells are noted at both doses of tmx and very rare hyperplasia is seen (Fig. 3 G and H). Taken together, these data demonstrate that a single dose of tmx labels CC10-expressing cells and a smaller number of putative BASCs at the BADJ. Administration of a higher dose of tmx labels more putative BASCs as well as alveolar cells. All of these cells are capable of K-RasG12D–induced proliferation.

Fig. 3.

BADJ CC10⁺ cells and alveolar CC10⁺, Sftpc⁺ cells are the most abundant proliferating cells after K-RasG12D activation. (A, B, D, E, G, and H) Sections of CC10-CreER; LSL-K-RasG12D; Rosa26R-fGFP mouse lung stained with anti-BrdU (red), GFP (green), and Sftpc (purple). (A and B) Twenty-four hours after one dose of tmx. (A) Representative BADJ with one GFP⁺, Sftpc⁻, BrdU⁺ cell (arrow) surrounded by GFP⁺, Sftpc⁺ cells (arrowheads). (B) Representative GFP⁺, BrdU⁺, Sftpc⁺ cell at the BADJ (arrow). (C) Quantification of BrdU-incorporating BADJ cells at different times after K-Ras activation. (D, E, G, and H) Seventy-two hours after four doses of tmx. (D) Alveolar BrdU⁺ cells are GFP⁺, Sftpc⁺. (E) Higher magnification of boxed area in D showing BrdU⁺, GFP⁺, Sftpc⁺ cells (arrows). (F) Quantification of alveolar BrdU⁺ cells at different times after K-Ras activation. (G and H) BrdU⁺, GFP⁺, Sftpc⁻ cells in larger airway (G) and trachea (H). Error bars represent standard variation of three different sections of a mouse lung.

To ask whether other types of cells contribute to early hyperplasia at the BADJ, we stained sections at 72 h after four doses of tmx for Foxj1. This is a marker of ciliated cells and also preciliated cells presumably derived from CC10⁺ progenitors. As Fig. S1 demonstrates, the hyperplastic regions at the BADJ contain three different subpopulations of cells that have undergone recombination (GFP⁺). These are (i) GFP⁺, Foxj1⁺ cells, (ii) GFP⁺, Sftpc⁺ cells, and (iii) GFP⁺ cells that are Sftpc⁻ and Foxj1⁻. No GFP⁺ cells expressing both Foxj1 and Sftpc were found (Fig. S1 A and B). In the alveoli, no Foxj1⁺ cells are detected (Fig. S1 D and E). Cells in subsequent adenocarcinomas stain positively for Sftpc but not for Foxj1 (Fig. S1 G and H).

To characterize the fates of GFP⁺ cells at the BADJ, we analyzed cells at the BADJ at 2, 7, and 16 wk after four doses of tmx in CC10-CreER; LSL-K-RasG12D; Rosa26R-fGFP mice. All three cell types [(GFP⁺, Foxj1⁺), (GFP⁺, Sftpc⁺), and (GFP⁺, Sftpc⁻, Foxj1⁻)] are consistently found in hyperplastic BADJ regions in similar relative proportions at each time. Taken together, these data suggest that, after K-RasG12D induction in CC10-expressing cells, multiple cell types are capable of proliferating. However, none of these cells progresses beyond hyperplasia. In the alveoli, GFP⁺, Sftpc⁺ cells are likely cancer-initiating cells.

A previous study reported that naphthalene treatment, which kills CC10⁺ Clara cells, increases the number of CC10⁺, Sftpc⁺ (dual-positive) putative BASCs at the BADJ and the number of tumors when K-RasG12D is induced by adenoviral Cre (6). We therefore asked whether naphthalene injury alters the distribution of proliferating cells between CC10⁺, Sftpc⁻ and CC10⁺, Sftpc⁺ subpopulations in our model. Intraperitoneal naphthalene was administered 1 wk after the fourth dose of tmx. BrdU was injected 6 h before sacrifice at either 3 d, 1 wk, or 3 wk after naphthalene injury. As Fig. S2 demonstrates, the majority of cells that proliferate in K-RasG12D–expressing cells at the BADJ after naphthalene injury are GFP⁺, Sftpc⁻ cells. The percentage of GFP⁺, Sftpc⁺ cells gradually rises over time from about 10% at 3 d after naphthalene to about 25% at 3 wk after naphthalene. Thus, in our model, naphthalene injury does not lead to preferential proliferation of dual-positive cells compared with CC10⁺ cells in the BADJ.

K-RasG12D Induction in Sftpc⁺ Cells Leads to Tumor Formation in the Alveoli but Not at the BADJ.

To further investigate whether BASCs at the BADJ are exclusively the tumor-initiating cells and which alveolar cells can initiate K-RasG12D–induced tumors, we generated Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP compound mice. When mice were 6–8 wk old, one or four doses of tmx were administrated to activate K-RasG12D in Sftpc-expressing cells. As shown in Fig. 4 A and B, one dose of tmx administered to an Sftpc-CreER; Rosa26R-fGFP mouse can label the majority of alveolar type II cells and 75% of putative BASCs. By 14 wk after one dose of tmx, Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mice develop widespread adenomas and adenocarcinomas in the alveoli. Significantly, at this dose, almost every BADJ is normal (Fig. 4C and Table S2). Similar findings are noted after four doses of tmx.

Fig. 4.

K-RasG12D activation in Sftpc-expressing cells leads to lung adenocarcinoma in alveoli. Arrows in A and B denote putative BASCs. (A) Section of Sftpc-CreER; Rosa26R-fGFP mouse lung 2 wk after one dose of tmx stained with anti-GFP (green) and CC10 (purple). GFP labels the majority of type II cells and putative BASCs. (B) BADJ area in A. (C) H&E section of Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mouse lung 14 wk after one dose of tmx. The majority of tumors arise in the peripheral lung and near the BADJ but not in the airways.

We then analyzed tumors at different locations and at different times after tmx injection. After 2 wk of K-RasG12D activation, BADJs are normal, but small adenomas are found in the alveoli (Fig. 5 D and E). In contrast to CC10-CreER; LSL K-RasG12D mice, an increase in the number of dual-positive cells at the BADJ did not develop (Fig. 5 B, G, L, and Q). However, similar to CC10-CreER; LSL K-RasG12D mice, small adenomas in the alveoli were Sftpc⁺, CC10⁻ (Fig. 5C). Almost all BADJs are histologically normal (Fig. 5 D, I, N, and S and Table S2). The alveolar tumors progressed to adenocarcinoma (Fig. 5 J, O, and T), and tumor cells remained Sftpc⁺, CC10⁻ (Fig. 5 H, M, and R). These tumors are virtually immunohistochemically indistinguishable from the alveolar tumors induced in the CC10-CreER; LSL-K-RasG12D; Rosa26R-fGFP mice. Thus, activation of oncogenic K-Ras in Sftpc-expressing cells predominantly drives tumor formation in the alveoli but not at the BADJ.

Fig. 5.

Time course of lung adenocarcinoma progression in Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mice after one dose of tmx. Insets are higher magnification images of cells denoted by arrows. (A–E) Two weeks. (A) Section of Sftpc-CreER; Rosa26R-fGFP mouse lung stained with anti-Sftpc (green) and CC10 (red). At the BADJ, putative BASCs stain Sftpc⁺ and CC10⁺. (Inset) Putative BASC at higher magnification. (B–E) Sections of Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mouse lungs. (B) BADJ CC10⁺, Sftpc⁺ cells (arrow) do not expand. (C) Representative small tumor is Sftpc⁺, CC10⁻. (D) Normal BADJ (arrow) and hyperplasia in the peripheral lung. (E) Representative small tumor. (F–J) Four weeks. (F) In Sftpc-CreER; Rosa26R-fGFP mouse lung, alveolar CC10⁺ type II cell. (G–J) Sections of Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mouse lung. (G) Normal BADJ. (H) Sftpc⁺, CC10⁻ alveolar tumor cells. (I) Normal BADJ (arrows) and alveolar tumors. (J) Peripheral lung tumor. (K–O) Eight weeks. (K) In Sftpc-CreER; Rosa26R-fGFP mouse lung, alveolar CC10⁺, Sftpc⁺ cells. (L–O) Sections of Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mouse lungs. (L) Normal BADJ staining. CC10⁺, Sftpc⁺ cells (arrow) do not expand. (M) Tumor with few very low CC10⁺ cells (arrow). (N) Normal BADJ (arrows) and alveolar tumors. (O) Representative tumor. (P–T) Fourteen weeks. (P) In Sftpc-CreER; Rosa26R-fGFP mouse lung, alveolar CC10⁺, Sftpc⁺ cells. (Q–T) Sections of Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mouse lungs. (Q) Normal BADJ staining. (R) Sftpc⁺, CC10⁻ alveolar tumor cells. (S) Normal BADJ (arrows) and widespread alveolar tumors. (T) Large adenocarcinoma in the peripheral lung.

After Sftpc-Driven K-RasG12D Induction, Proliferation Occurs in Almost all Alveolar Sftpc⁺ Cells, but the Putative BASCs only Very Rarely Develop Small Hyperplastic Regions.

Because of the tumor-initiating ability of CC10-expressing type II cells in CC10-CreER; LSL K-RasG12D mice, we questioned whether other type II cells were able to initiate tumors. To identify which Sftpc⁺ epithelial cells are induced to proliferate after K-RasG12D induction, we killed Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mice at different times after either one dose or four doses of tmx. BrdU was injected 6 h before sacrifice to label proliferating cells. After one dose (Fig. S3) or four doses (Fig. S4) of tmx, GFP-labeled, CC10-expressing cells in the alveoli incorporate BrdU. The vast majority of BrdU-incorporating cells are GFP-labeled alveolar type II cells. Due to our inability to detect CC10 protein by immunohistochemistry in type II cells, we are unable to definitively determine whether only CC10-expressing type II cells or all type II cells in the alveoli can initiate tumors. However, compared with CC10-CreER; LSL-K-RasG12D mice, which succumb due to tumor burden at 20–24 wk, Sftpc-CreER; LSL-K-RasG12D mice develop much more widespread neoplastic areas that cover very large alveolar regions (Fig. 4C). The longest that any of these mice has lived is 14 wk after tmx injection. For these reasons, we believe that both CC10⁺, Sftpc⁺ and CC10⁻, Sftpc⁺ type II cells in alveoli can initiate tumors.

As shown in Fig. S3 and Table S2, almost all BADJs contain only one or two GFP⁺ putative BASCs after either one dose or four doses of tmx even at later times. GFP-labeled CC10⁺ cells rarely incorporate BrdU, and only two small hyperplastic regions were noted (at 14 wk after tmx injection) in the 314 BADJs examined (Table S2). Taken together, these results suggest that lung tumors in Sftpc-CreER; LSL-K-RasG12D; Rosa26R-fGFP mice are not initiated from putative BASCs at the BADJ but from alveolar Sftpc-expressing cells.

Differential Expression of Sox2 in Hyperplasia and Tumors from CC10-Driven K-Ras and Sftpc-Driven K-Ras Lungs.

To further investigate the cell of origin in the hyperplasia at the BADJ, we stained control lungs and regions of hyperplasia and tumors of both CC10-CreER; Rosa26R-fGFP and Sftpc-CreER; Rosa26R-fGFP mice, with and without the LSL-K-RasG12D allele, with antibody to Sox2. As Fig. S5 demonstrates, CC10⁺ cells at the BADJ and throughout the airways are strongly positive for Sox2. A small percentage of CC10⁺ cells in the alveoli also stain positively for a low level of Sox2 (Fig. S5A, Inset). In contrast, Sftpc⁺ cells in the alveoli appear negative for Sox2. Dual-positive cells at the BADJ stain positively for Sox2 (Fig. S5E). When K-RasG12D is induced in CC10⁺ cells, the resulting hyperplasia and adenomas stain strongly for Sox2 (Fig. S5 B–D). However, when K-RasG12D is induced in Sftpc⁺ cells, the resulting hyperplasia and adenomas are Sox2⁻ both at the BADJ and in the alveoli (Fig. S5 F–H). Larger adenomas and adenocarcinomas that develop subsequently in the alveoli in both models are Sox2⁻. These results suggest that either cells with strong Sox2 expression are unable to progress to cancer or Sox2 is down-regulated as adenocarcinomas progress.

Comparative Microarray Analysis of CC10-Driven and Sftpc-Driven Alveolar Tumors Demonstrates Similar Transcriptome Profiles.

To investigate whether alveolar tumors resulting from K-RasG12D induction in CC10⁺ cells and Sftpc⁺ cells express similar genes, tumors were microdissected from the alveolar region of three CC10-CreER; LSL-K-RasG12D mice and three Sftpc-CreER; LSL-K-RasG12D mice. RNA was isolated and hybridized to Affymetrix mouse 430.2 microarray chips at the Duke Microarray Core Facility (http://www.microarray.duke.edu). Analysis of the transcriptomes using an adjusted P value threshold of 0.1 revealed only 24 differentially expressed genes. Of these, nine probe sets are sex-related due to sex differences between the mice. The remaining 15 probe sets are listed in Table S3. Interestingly, Sox2 is significantly higher in CC10-CreER; LSL-K-RasG12D tumors than in Sftpc-CreER; LSL-K-RasG12D tumors. This is consistent with the immunohistochemistry in Fig. S5.

We next sought to correlate these differentially expressed genes to human K-RAS mutant adenocarcinomas. Because the only gene in the list in Table S3 that has been implicated in lung cancer is Sox2, we asked whether human K-RAS mutant lung tumors vary in prognosis based upon SOX2 expression. However, existing human adenocarcinoma datasets are not annotated as to K-RAS mutation status. Therefore, using gene expression from 143 lung cancer cell lines annotated with K-RAS status, a predictive model of K-RAS mutation was trained that was then applied to the National Cancer Institute Director's Challenge data (9) to select predicted K-RAS mutant samples. Using the predicted K-RAS mutant samples within this dataset, Sox2 did not significantly correlate with survival (P = 0.15) (Fig. S6).

We also assessed whether mouse CC10-CreER; LSL-K-RasG12D tumors shared similarity with K-RAS mutant human lung cell lines. To measure this, we first identified differentially expressed genes between K-RAS mutant and wild type in human lung cell lines within the Cancer Cell Line Encyclopedia (CCLE; http://www.broadinstitute.org/ccle), yielding 121 genes (false discovery rate < 0.01, Benjamini-Hochberg; ref. 10). Of these 121 genes, 105 genes could be mapped to mouse orthologs. We then generated a ranked list of differentially expressed genes between microdissected CC10-CreER; LSL-K-RasG12D tumors (n = 3) and CC10-CreER; LSL-fGFP sorted cells (n = 3), and evaluated whether these 106 genes were enriched with respect to the tails of the rank distribution. We found that these genes were indeed enriched [P < 1e-11, Kolmogorov-Smirnov (KS) test], suggesting that K-RasG12D–induced mouse tumors and human lung cell lines share transcriptional similarity.

Discussion

In the present study, we used two knock-in CreER driver mouse lines to express oncogenic codon 12 mutant K-Ras in CC10- and Sftpc-expressing cells of the adult lung. BrdU incorporation studies show that soon after K-Ras activation in CC10-CreER; LSL-K-RasG12D mice, proliferation occurs in several different cell types. At the BADJ, the proliferating cells are predominantly Sftpc⁻, with lower proportions of Sftpc⁺ and Foxj1⁺ cells. All three cell types appear to be present in about the same proportion in the hyperplasia that develops in the BADJ region, as shown in the proposed model of cancer-initiating cells in CC10-CreER; LSL-K-RasG12D mice (Fig. 6, Lower Left). Fig. 6 (Lower Right) demonstrates that very rare hyperplasia develops from Sftpc⁺ cells at the BADJ in Sftpc-CreER; LSL-K-RasG12D mice. These results indicate that CC10⁺, Sftpc⁺ putative BASCs are not the only cell type that proliferates in response to K-Ras activation in the BADJ. In the alveoli, all cells that proliferate early are Sftpc⁺. Some can also be CC10⁺, as judged by expression of GFP using the CC10-CreER and Rosa26R-fGFP reporter alleles. The tumors that develop in the alveoli are made up of Sftpc⁺ cells. These results clearly demonstrate that CC10⁺ type II cells can give rise to adenocarcinomas in response to K-Ras activation, although whether these are the only cell type that can do so remains to be tested. Finally, despite recombination in larger airways in CC10-CreER; LSL-K-RasG12D mice after tmx administration, we see no tumor formation or consistent hyperplasia in more proximal airways. The lack of tumor formation in the larger bronchi and bronchioles is surprising, given the widespread recombination in the CC10-CreER mouse line.

Fig. 6.

Model for response of various respiratory epithelial cells to K-RasG12D induction. In the steady-state normal lung (Upper), there are type II cells that express Sftpc (red) and those that are dual-positive for CC10 and Sftpc (purple). When K-RasG12D is expressed in CC10-CreER mice (Left) there is hyperplasia in the BADJ that involves not only putative BASCs but also Clara cells. We propose that adenocarcinomas arise in the alveoli exclusively from dual-positive type II cells but that these become only Sftpc⁺. When K-RasG12D is expressed in Sftpc-CreER mice (Right), the BADJ is normal or contains rare small hyperplastic areas, and tumors arise only in the alveoli. We propose that they can arise from either dual-positive or Sftpc⁺ cells.

Previous studies in other laboratories have used inducible oncogenic stimuli to investigate the origin of lung cancer. The earliest studies used inhaled adenoviral Cre to induce expression of oncogenic K-Ras throughout the respiratory epithelium (6, 8, 11). In these models, expansion of CC10⁺, Sftpc⁺ cells (putative BASCs) at the BADJ was taken as evidence for their being cells of origin of adenocarcinoma. The expansion of these cells following naphthalene-induced injury, combined with the increased number of tumors after naphthalene treatment, further supported this idea. The lack of tumor formation at the BADJ in our Sftpc-CreER; LSL-K-RasG12D mice makes it unlikely that BASCs are the only cells capable of giving rise to lung adenocarcinomas. Moreover, we found that, at the BADJ, CC10⁺, Sftpc⁻ cells proliferated early after K-Ras activation. Finally, we saw no evidence of the hyperplasia in this region progressing to adenocarcinoma.

Other groups have also noted expansion of dual-positive cells in response to induction of oncogenic stimuli. Expression of p38-α (12), PTEN deletion with K-RasG12D expression (13), and Gata6 deletion (14) lead to increases in dual-positive cells, but detailed delineation of expansion of other cells and assessment of alveolar tumor formation are often omitted. To assess definitively whether putative BASCs initiate tumors, it will be necessary to develop an allele that targets dual-positive cells specifically. Alternatively, transplantation of BASCs into immunodeficient mice will need to be performed. This will necessitate agreement on BASC cell surface markers, which has been contentious (4, 15).

One tenet of the cancer stem cell hypothesis is that tumors arise from stem cells or from progenitor or postmitotic cells that attain stem cell-like properties. Our data indicate that, at the BADJ, many CC10⁺ cells, not only those that are Sftpc⁺, are capable of proliferating in response to oncogenic K-Ras. This is consistent with recent lineage-tracing experiments in which several CC10⁺ cells in the bronchioles and at the BADJ were able to contribute to epithelial regeneration after injury (7, 16). Although lineage tracing with the Sftpc-CreER allele shows that type II cells can differentiate into type I cells during steady state and after alveolar injury (17), whether type II cells can self-renew over long periods awaits the results of long-term lineage-tracing studies. Although the alveolar tumors that eventually develop in CC10-expressing and Sftpc-expressing cells are identical by immunohistochemical analysis, microarray analysis demonstrated some significant differences between these tumors (Table S3). One of these is Sox2, which has been shown to be oncogenic for lung cancer.

Other previous evidence also points to type II cells as a cell of origin for K-Ras–induced adenocarcinoma. First, human K-RAS mutant tumors are almost exclusively adenocarcinomas (18, 19), and they tend to occur peripherally whereas small-cell and squamous tumors tend to occur centrally (20). In a chemically induced mouse model of lung adenocarcinoma, codon 12 K-Ras mutants were identified, and electron microscopy revealed type II features in the tumor cells (21). Finally, K-RAS mutations were found exclusively in alveolar tumors but not bronchial tumors in a PCR-based analysis of resected human adenocarcinomas (22). The reasons for the susceptibility of type II cells for mutant K-Ras–induced cancer are presently unclear. In addition, the cell of origin will most likely differ based upon oncogenic stimulus. The CreER gene knock-in mouse lines used in this study will be useful in identifying initiating cells in future models.

Materials and Methods

Mice.

The CC10-CreER and Rosa26R-CAG-farnesylated GFP (Rosa26R-fGFP) mouse lines have been described previously (7). To generate Sftpctm1(cre/ERT)Blh (Sftpc-CreER) mice, the coding sequence and 3′ UTR of Sftpc were retrieved from a BAC by recombineering into a vector upstream of a diphtheria toxin (DT) cassette for negative selection in ES cells. An IRES-CreERT2 cassette and a PGKneo cassette flanked with Flp recognition target (FRT) sites were recombined into the 3′ UTR (Fig. S4). The construct was linearized and electroporated into 129S6/SvEvTac ES cells. Ten correctly targeted clones were identified by Southern blot and PCR, and ES cells from three clones were injected into C57BL/6 blastocysts. Mice heterozygous for Sftpc-CreER from one ES cell clone were bred to ROSA-FLP line 129S4-Gt(ROSA)26Sortm2(FLP*)Sor/J mice to remove the neo cassette. All mice that carry an LSL-K-RasG12D allele in this study also carry a Trp53Flox/+ allele. LSL-K-RasG12D; Trp53Flox/Flox mice were kindly provided by David Kirsch (Duke University, Durham, NC) and crossed to CC10-CreER; Rosa26R-fGFP mice or Sftpc-CreER; Rosa26R-fGFP mice. Being heterozygous for p53 did not alter the number of tumors but did slightly accelerate their initiation and progression. For brevity, Trp53Flox/+ was omitted from all mouse descriptions in this study.

Tamoxifen/Naphthalene Administration.

A 20 mg/mL tamoxifen (Sigma) solution was created by dissolving tamoxifen in Mazola corn oil. Six- to eight-week-old CC10-CreER; LSL-K-Ras mice were injected with 0.25 mg/g body weight tamoxifen either once or every other day for four doses. Naphthalene (Fisher) was dissolved in Mazola corn oil and injected at 275 mg/kg for male and 250 mg/kg for female mice 1 wk after the last dose of tamoxifen injection. All mice injected with naphthalene lost 20% of body weight, indicative of the desired effect.

Histology and Immunohistochemistry.

For cell proliferation studies, mice were injected with 10 μL/g weight BrdU 6 h before sacrifice. Lungs were perfused with PBS and then inflated and fixed with 4% paraformaldehyde overnight. Fixed lungs were then washed three times with PBS and gradually dehydrated in ethanol and embedded in paraffin. Paraffin sections were stained with the following primary antibodies: chicken anti-GFP (1:400; Aves Labs; GFP1020), rabbit anti–pro-Sftpc (1:500; Millipore; AB3786), goat anti-CC10 (1:10,000; kindly provided by Barry Stripp, Duke University, Durham, NC), rabbit anti-CC10 (1:10,000; Barry Stripp), rat anti-BrdU (1:200; Accurate Chemical and Scientific; OBT0030), mouse anti-Foxj1 (1:200; eBioscience; 14-9965), and rabbit anti-Sox2 (1:1,000; Seven Hills Bioreagents; WRAB-Sox2). Alexa Fluor-coupled secondary antibodies (Invitrogen) were used at a 1:400 dilution. All images used for scoring cells were captured on a Leica Sp2 laser scanning confocal microscope either with a Z stack of optical sections or single scanning. Multiple optical sections were scored manually to distinguish cell boundaries. Three different whole-lung longitudinal sections containing the main axial bronchial pathway for each mouse at each time point were scored, including all BADJ areas, alveoli, and larger airway bronchi.

Microarray Analysis.

Tumors were microdissected from the lungs of mice after sacrifice at various times. Because the Sftpc-CreER; LSL-K-RasG12D mice had more aggressive tumors, these mice were killed and tumors were microdissected at 2, 3, and 4 wk after the fourth dose of tamoxifen, whereas the CC10-CreER; LSL-K-RasG12D mice were killed and tumors were microdissected at 11, 12, and 16 wk after the fourth dose of tamoxifen. At these stages, the tumors were all primarily Sftpc⁺, CC10⁻ by immunohistochemistry. Each tumor was homogenized and the RNA was extracted using Qiagen protocols. The RNA was submitted to the Duke Microarray Core Facility, where it was reverse-transcribed, and the cDNA was hybridized to Affymetrix 420 mouse microarray chips. The resulting CIMFast Event Language (CEL) files were robust multichip average (rma)-normalized in Bioconductor in the R environment. Differential expression was carried out using the Limma package with multiple comparisons controlled by the method of Benjamini and Hochberg (10). To infer K-RAS status, we built a predictive model of K-RAS mutation status using a large cell line panel (CCLE), where the K-RAS mutation status is known. We split the CCLE data in half—into a training and a validation dataset—and fit a penalized regression model (http://www.stanford.edu/hastie/papers) on the training data using 10-fold cross-validation to estimate the regularization parameters. We validated this model on the holdout data and demonstrated robust predictive performance with an area under the curve (AUC) of 0.9. We then applied this model to the Shedden et al. dataset (10) (after normalizing the data to have the same mean center and variance as the cell line data) to infer K-RAS mutant status. We know on average that 30% of a lung cancer cohort is likely to have K-RAS–mutated, so we chose the top 30% of the predicted samples as our K-RAS mutant cohort (n = 140). Using this subset of samples, we tested whether SOX2 is correlated with survival using a Cox regression model.