Functional characterization of pulmonaryneuroendocrine cells in lung development,injury, and tumorigenesisHai Song, Erica Yao, Chuwen Lin, Rhodora Gacayan, Miao-Hsueh Chen, and Pao-Tien Chuang1

Cardiovascular Research Institute, University of California, San Francisco, CA 94158

Edited* by Matthew P. Scott, Howard Hughes Medical Institute, Stanford University, Stanford, CA, and approved September 14, 2012 (received for reviewApril 30, 2012)

Pulmonary neuroendocrine cells (PNECs) are proposed to be thefirst specialized cell type to appear in the lung, but their ontogenyremains obscure. Although studies of PNECs have suggested theirinvolvement in a number of lung functions, neither their in vivosignificance nor the molecular mechanisms underlying them havebeen elucidated. Importantly, PNECs have long been speculated toconstitute the cells of origin of human small-cell lung cancer (SCLC)and recent mouse models support this hypothesis. However, a ge-netic system that permits tracing the early events of PNEC trans-formation has not been available. To address these key issues, wedeveloped a genetic tool in mice by introducing a fusion protein ofCre recombinase and estrogen receptor (CreER) into the calcitoningene-related peptide (CGRP) locus that encodes a major peptidein PNECs. The CGRPCreER mouse line has enabled us to manipulategene activity in PNECs. Lineage tracing using this tool revealed theplasticity of PNECs. PNECs can be colabeled with alveolar cellsduring lung development, and following lung injury, PNECs cancontribute to Clara cells and ciliated cells. Contrary to the currentmodel, we observed that elimination of PNECs has no apparentconsequence on Clara cell recovery. We also created mouse modelsof SCLC in which CGRPCreER was used to ablate multiple tumorsuppressors in PNECs that were simultaneously labeled for follow-ing their fate. Our findings suggest that SCLC can originate fromdifferentiated PNECs. Together, these studies provide unique in-sight into PNEC lineage and function and establish the foundationof investigating how PNECs contribute to lung homeostasis, in-jury/repair, and tumorigenesis.

progenitor | naphthalene | cell of origin | tumor suppressor gene

One of the major unresolved issues in the lung field is howundifferentiated epithelial cells generate specialized cell types

in the postnatal lung (1–3). Another pressing challenge is to un-cover the cell types and molecular mechanisms that underlie theenormous regenerative capacity of the lung during injury (4). Thelung produces more than 40 cell types to fulfill the importantroles of mucociliary clearance, gas exchange, and metabolic andendocrine function. The major cell types in lung epithelium in-clude basal cells, ciliated cells, Clara cells, neuroendocrine cells,and type I and type II pneumocytes (Fig. S1 A–F). Pulmonaryneuroendocrine cells (PNECs) are proposed to be the first spe-cialized cell type to appear within the epithelium, implying thatprogressive cell-type specification in the lung starts with pro-genitors of neuroendocrine nature (5). However, the ontogeny ofPNECs and their relationships to other lung cell types duringnormal homeostasis and lung injury remain unclear. Unlikesome other cell types in the mammalian lung, PNECs are foundin most species examined, including amphibians, reptiles, birds,mammals, and even in gill filaments of fish, suggesting an evo-lutionarily conserved function of PNECs in lung physiologyacross phyla.Single PNECs are scattered in the respiratory epithelium,

whereas clustered PNECs [called neuroepithelial bodies (NEBs)]are usually found in the intrapulmonary airways, often at air-way bifurcations or bronchioalveolar duct junctions. PNECs are

innervated at the basal part and have microvilli that protrude intothe airway lumen at their apical surface. NEBs contain secretorygranules and dense-core vesicles, which store various amines andpeptides, including gastrin-releasing peptide (GRP), serotonin(5-HT), calcitonin gene-related peptide (CGRP), calcitonin,substance P, somatostatin, chromogranin A, and synaptophysin(SYP). These contents are released by physiological stimuli, suchas hypoxia, and many of them have potent physiological effects.The diverse and often opposite effects of secreted amines andpeptides from NEBs reveal the complex nature of PNEC func-tion that remains to be elucidated. Decades of studies on PNECsculminate in the current view that PNECs function in airwayoxygen sensing, regulate pulmonary blood flow, control bronchialtonus, modulate immune responses, and maintain a stem cellniche (5, 6). No other cell type in the lung exhibits such a diversearray of activities. However, many of these conclusions are de-rived from experiments using ex vivo tissue/organ culture or cell-based assays. Rigorous tests in a genetic model organism (e.g.,mouse) to validate and pinpoint the major physiological func-tions of PNECs and uncover the molecular mechanisms havenot been performed. The evolutionary conservation of PNECsimplies that conclusions reached in mice can be extrapolated toother mammals, including humans.NEBs are found in close association with secretory nonciliated

Clara cells. Following naphthalene-induced lung injury (7, 8), whichkills most Clara cells, the only surviving Clara cells (termed variantClara cells in some literature) are located near NEBs and at thebronchioalveolar duct junction. These survivors are capable ofrestoring the damaged lung epithelium. These observations led tothe hypothesis that PNECs maintain a stem cell niche essentialfor Clara cell regeneration during lung injury (9, 10). Whethersuch a niche actually exists or has functional importance has neverbeen critically evaluated.PNECs have also been implicated in a number of lung dis-

eases. Perhaps the most important clinical connection comesfrom the observation that characteristic markers of PNECs arefrequently expressed in human small-cell lung cancer (SCLC), ahighly aggressive form of lung cancer related to cigarette smoking(11). This finding led to the speculation that PNECs could be thecells of origin for SCLC, an idea that has been supported bymouse studies (12–15). Despite this insight, the ability to reliablytrace the early events that underlie changes in PNEC behavior andtheir progression toward tumor development in a defined geneticsetting has not been achieved. As a result, a definitive answer towhether SCLC is derived from progenitors or differentiated cellsis not available. Filling in the major gaps in our molecular un-derstanding of PNEC transformation requires the development

Author contributions: H.S., E.Y., and P.-T.C. designed research; H.S., E.Y., C.L., R.G., M.-H.C.,and P.-T.C. performed research; H.S., E.Y., and P.-T.C. analyzed data; and H.S., E.Y., andP.-T.C. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1To whom correspondence should be addressed. E-mail: pao-tien.chuang@ucsf.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1207238109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1207238109 PNAS | October 23, 2012 | vol. 109 | no. 43 | 17531–17536

GEN

ETICS

of additional mouse models of SCLC that recapitulate criticalaspects of human tumors.To regulate gene activity specifically in PNECs, we introduced

CreER into the endogenous mouse CGRP locus by gene targeting(Fig. S2A) because CGRP is the major peptide produced inPNECs. The resulting mouse line is designated CGRPCreER.CreER encodes a fusion protein of Cre recombinase and es-trogen receptor (ER), and CreER is active only when tamoxifen(TM) is present. Thus, TM administration to CGRPCreER miceprovides an efficient way to control Cre activity in a temporallyand spatially specific manner.

ResultsInducible Expression of CreER from the Mouse CGRP Locus ConfersSpatial and Temporal Control of Gene Expression in PulmonaryNeuroendocrine Cells. To test the efficiency and specificity ofCreER expression in CGRPCreER mice, we bred CGRPCreER micewith ROSA26mTmG reporter mice to generate mice carryingthe genotype of CGRPCreER/+;ROSA26mTmG/+. Cre activationupon TM administration resulted in eGFP expression from theROSA26mTmG allele. PNECs were identified using multiplemarkers, including anti-Ascl1 (Mash1), anti-CGRP, anti-SYP,anti-chromogranin A, and anti-PGP9.5. We detected GFPexpression in 50–90% of CGRP+ cells, and in some NEBs theefficiency reached almost 100% (Fig. S2 B–G). The specificityof CreER expression in PNECs was demonstrated by the ob-servation that no eGFP expression was detected without TM(Fig. 1F and Fig. S2 H–J) or in CGRP– cells in CGRPCreER/+;ROSA26mTmG/+ lungs (Fig. S2 B–G). R26R reporter mice yieldedsimilar results. These findings indicate that CreER expressionfrom CGRPCreER mice is specific and sufficient to permit ma-nipulating gene activity in PNECs.

Pulmonary Neuroendocrine Cells Are Colabeled with Alveolar CellsDuring Lung Development. Pulmonary neuroendocrine cells havebeen proposed to originate from the lung epithelium (instead ofthe neural crest). To test this, we took advantage of Shh-Cre,Nkx2.1-Cre, and Sox9-Cre mouse lines in which Cre is expressedbroadly in the lung epithelium but not in the neural crest. Wegenerated Shh-Cre/+;R26R/+ mice in which constitutive epi-thelial Cre expression from the endogenous Shh locus wouldresult in LacZ expression from the R26R reporter. Shh-Cre–mediated recombination can be detected in the lung epithelium asearly as 9.5 days postcoitum (dpc). We found that PNECs to-gether with other epithelial cell types yielded positive stainingfor β-gal (Fig. 1A), suggesting that epithelial cells give riseto PNECs. Similar results were obtained using Nkx2.1-Cre/+;ROSA26mTmG/+ (Fig. 1B) or Sox9-Cre/+;ROSA26mTmG/+ (Fig.1C) mice. Our findings are consistent with the previous reportthat Id2+ tip epithelial cells labeled at the pseudoglandular stageof mouse development contribute to all lineages, includingneuroendocrine cells (16).Our CGRPCreER mice provide a tool to follow the fate of

CGRP+ cells during lung development. We injected a singledose of TM into pregnant CGRPCreER/+;ROSA26mTmG/+ mice atdifferent gestational ages and collected the lungs of progeny atpostnatal day 30 or 60. This allowed us to label cells that expressCreER by eGFP expression at a particular stage of lung de-velopment and follow the cell types they eventually generate.The identity of lung cells labeled by eGFP was determined byimmunostaining using markers specific for distinct lung celltypes (Fig. 1 G and H and Fig. S1 A–F). When the embryoswere exposed to TM between 12.5 and 14.5 dpc, PNECs werelabeled (Fig. 1G). Surprisingly, a small fraction of alveolar type I(stained with anti-T1α) and type II (stained with anti-SP-C)pneumocytes were also labeled (Fig. 1 H and I). Clusters oflineage-labeled type I cells are located within the same alveolus(arrows in Fig. 1I), suggesting that the cluster is derived froma common progenitor. By contrast, secretory Clara cells [stainedwith anti-Clara cell 10-kDa protein (CC10)] or ciliated cells(stained with anti-acetylated tubulin) were never labeled. These

results suggest that PNECs may share a common origin with al-veolar but not conducting cell types during early lung development.Alternatively, during development, CGRPCreER may label separateprogenitor populations that give rise to PNECs and alveolar cells,respectively. Interestingly, when embryos were exposed to TM at

Fig. 1. Lineage tracing of PNECs in embryonic and postnatal lungs. (A) β-gal(LacZ) staining of lung sections from Shh-Cre/+;R26R/+ adult mice. (B and C)Immunostaining of lung sections from Nkx2.1-Cre/+;ROSA26mTmG/+ or Sox9-Cre/+;ROSA26mTmG/+ adult mice. Extensive (nearly ubiquitous) labeling ofepithelial cells by β-gal (blue) or eGFP (green) was observed. Labeled epi-thelial cells include clustered PNECs (NEBs; arrows), which were identifiedby anti-SYP (brown), anti-CGRP (red), or additional antibodies. (D) Immu-nostaining of lung sections from adult wild-type (WT) mice. PNECs (CGRP+)are found in clusters and in close association with Clara cells (CC10+). (E)Immunostaining of lung sections from adult WT mice. CGRP immunoreac-tivity coincides with that of Ascl1 (Mash1). (Inset) Solitary PNEC. (F) Immu-nostaining of lung sections from adult CGRPCreER/+;ROSA26mTmG/+ micewithout TM injection. No eGFP-labeled cells were detected. (G–I) Immunos-taining of lung sections from adult CGRPCreER/+;ROSA26mTmG/+ mice [post-natal day 60 (P60)] exposed to tamoxifen at 12.5 dpc. Activation of CreERby tamoxifen during embryogenesis led to eGFP labeling of PNECs (G) aswell as alveolar cells, including type I (T1α+) and type II (SP-C+) pneumo-cytes (H and I). (H Inset) Individual images of eGFP and CGRP immunos-taining of the same cells. Arrows in I point to labeled type I cells. No Claracells or other cell types were labeled by eGFP. (J) Immunostaining of lungsections from adult CGRPCreER/+;ROSA26mTmG/+ mice (P60) exposed to ta-moxifen at 15.5 dpc. Only PNECs were labeled. Arrow points to unlabeledClara cells. (Inset) Separate images of eGFP and CGRP immunostaining ofthe same cells. (K–O) Immunostaining of lung sections from adult CGRPCreER/+;ROSA26mTmG/+ mice injected with tamoxifen at 1 mo of age and analyzedat different time points post-TM injection. (M–O Insets) Individual imagesof eGFP and CGRP immunostaining of the same cells. (P and Q) A model ofPNEC specification during lung homeostasis. (Scale bars: 25 μm for panelsin each row.)

17532 | www.pnas.org/cgi/doi/10.1073/pnas.1207238109 Song et al.

11.5 dpc, only a small number of type I and type II cells werelineage labeled, and no PNECs were labeled. If TM injection isperformed at or after 15.5 dpc, only PNECs were labeled (Fig. 1J),suggesting that by ∼15.5 dpc the PNEC lineage has segregatedfrom other lineages. In our lineage-tracing studies (Table S1), noother lung epithelial cells or lung mesenchymal cells, such asfibroblasts or blood vessels, were labeled. Interestingly, a smallnumber of muscle cells in the pulmonary veins that likely originatefrom cardiomyocytes (17) were labeled when TM was injectedbefore 15.5 dpc (Fig. S3). CGRPCreER activity can also be detectedin several mouse tissues other than the lung, consistent with thepresence of a diffuse neuroendocrine system (Fig. S4).We then asked whether the committed PNEC fate is stable

during normal homeostasis in adult lungs. We injected CGRPCreER/+;ROSA26mTmG/+ mice with TM at 1–2 mo of age and analyzedcell types labeled by eGFP at 0, 2, 4, 6, 8, and 12 mo after in-jection (Fig. 1 K–O). In mice studied even at 1 y after TM in-jection, only PNECs remained labeled by eGFP (Fig. 1O),though the percentage of lineage-labeled PNECs within NEBswas slightly decreased, consistent with reduced PNEC numberwith age. No other cell types were labeled, which suggests thatthe fate of PNECs is steady in adult lungs during homeostasis(Table S2).

Pulmonary Neuroendocrine Cells Contribute to Clara Cells and CiliatedCells During Lung Injury. Although the PNEC fate is stable in adultlungs during homeostasis, we wondered whether PNECs in theadult lung could serve as progenitors for other cell types duringtissue injury. To test this, we used the well-established naphtha-lene-induced lung injury model to ablate Clara cells. Weinjected CGRPCreER/+;ROSA26mTmG/+ mice at 1–2 mo of agewith TM to label PNECs, and this was followed by naphthaleneinjection. Exposure to naphthalene led to extensive death of Claracells, followed by proliferation of the surviving Clara cells andsubsequent restoration of lung epithelium by 1 mo (Fig. 2 A–D).The surviving Clara cells do not express a cytochrome P450enzyme that metabolizes naphthalene to toxic epoxide.In contrast with a very low rate of PNEC proliferation during

homeostasis, the number of PNECs labeled by CGRP and Ki67significantly increased after naphthalene injection (Fig. 2 E–H),indicating increased PNEC proliferation after lung damage. In-terestingly, we also detected eGFP-labeled Clara cells and ciliatedcells (Fig. 2 I–L), which suggests that unlike in development andhomeostasis, following lung injury PNECs possess the capa-bility of producing Clara cells and ciliated cells. We speculatethat the labeled ciliated cells were derived from labeled Claracells (produced from PNECs) because lineage tracing of Claracells has revealed their potential to generate ciliated cells (18).No other cell types, including type I and type II pneumocytes,were labeled by eGFP in CGRPCreER/+;ROSA26mTmG/+ micethat received tamoxifen and naphthalene. These results arguea lineage relationship between PNECs and Clara cells during lunginjury. Interestingly, lineage tracing of CC10-expressing cells (in-cluding variant Clara cells) actually shows no contributions ofthese cells to PNECs (18).

Pulmonary Neuroendocrine Cells Have Very Limited Potential toRegenerate After Ablation. Because lung epithelium can be replen-ished after lung injury, we investigated whether PNECs canbe regenerated after ablation. We took advantage of efficientcell killing by diphtheria toxin (DTA) to genetically eliminatePNECs in mice, and performed crosses to obtain CGRPCreER/+;ROSA26DTA/+ mice. Induction of Cre expression in PNECsby TM injection should result in DTA expression from theROSA26DTA/+ allele by removing floxed sequences that blockDTA expression. We tested the efficiency of cell killing by TMinjection to adult CGRPCreER/+;ROSA26DTA/+ mice and collectedlungs a few days after the last injection. Virtually no PNECs weredetected when a battery of PNEC markers was used, indicatingnearly complete eradication of PNECs (Fig. S5). By contrast, otherlung cell types, such as Clara cells, ciliated cells, and type I and

type II cells, were unaffected in CGRPCreER/+;ROSA26DTA/+

mice (Fig. S5).We collected lungs fromCGRPCreER/+;ROSA26DTA/+

mice at 1, 6, 10, and 12 mo post-TM injection, and at no time pointscould we detect regenerated PNECs (Fig. 3 A and B). We concludethat once ablated PNECs cannot be replaced by other cell typesduring homeostasis.We also tested whether PNECs, once eliminated, can be regen-

erated during lung injury in CGRPCreER/+;ROSA26DTA/+ mice. Inthis study, we ablated PNECs in CGRPCreER/+;ROSA26DTA/+

mice by TM injection and then induced lung injury with naph-thalene, which can stimulate PNEC proliferation as describedabove. No PNEC regeneration was found 1 mo after naphthaleneinjection (Fig. 3F), by which time Clara cells had fully regen-erated. Furthermore, even with repeated naphthalene-inducedlung injury, we observed no evidence of PNEC recovery. Takentogether, these results indicate the inability of other lung cellstypes to replenish PNECs during homeostasis or following tissueinjury, turnover of PNECs seems to be the only mechanism oftheir renewal.

Selective Ablation of Pulmonary Neuroendocrine Cells Does Not AffectClara Cell Regeneration Following Naphthalene-Induced Lung Injury.PNECs are found in close association with Clara cells, in par-ticular, the naphthalene-resistant Clara cells, which led to the

Fig. 2. CGRP+ cells can give rise to Clara cells and ciliated cells during Claracell regeneration induced by naphthalene administration. (A–G) Immunos-taining of lung sections from adult CGRPCreER/+;ROSA26mTmG/+ mice. CreER inPNECs (CGRP+) of adult lungs was activated by TM injection and followed bynaphthalene injection to ablate Clara cells (CC10+) in these animals. Claracells (A) were almost completely eliminated in this process (e.g., day 3 afternaphthalene injection shown in B). Occasionally, cells that express both CGRPand CC10 can be found (arrow in B). Regeneration of surviving Clara cellswas apparent by day 6 after naphthalene treatment (D). Naphthalene alsoinduced PNEC proliferation as indicated by increased Ki67+ PNECs (F and G).A time course of changes in the proliferation rate of PNECs after naphtha-lene treatment is shown in H. (F Inset) Individual images of CGRP and Ki67immunostaining of the same cells. (I–L) Immunostaining of lung sections fromadult CGRPCreER/+;ROSA26mTmG/+ mice 1 mo after naphthalene injection, whichwas performed following tamoxifen administration. The combined treatmentof TM and naphthalene resulted in labeling of PNECs as well as Clara cells(CC10+) and ciliated cells [acetylated (Ac)-tubulin+] (I–L). No other cell typeswere labeled by eGFP, which is in contrast to the stable lineage of PNECsduring homeostasis. Note that L is a merged image of I–K. (Scale bars: 25 μmfor panels in each column.)

Song et al. PNAS | October 23, 2012 | vol. 109 | no. 43 | 17533

GEN

ETICS

speculation that PNECs maintain a stem cell niche essential forClara cell regeneration during lung injury. We used CGRPCreER/+;ROSA26DTA/+ mice in which PNECs have been ablated to testthis hypothesis. We injected CGRPCreER/+;ROSA26DTA/+ mice withTM, followed by naphthalene administration. PNECs inCGRPCreER/+;ROSA26DTA/+ lungs were almost completely elimi-nated as revealed by immunostaining (Fig. 3 D–F and Fig. S5)compared with intact PNECs in ROSA26DTA/+ controls (Fig. 3J–L). At 3 d after naphthalene treatment, the majority ofClara cells were already killed and released from the airway

basement membrane in both CGRPCreER/+;ROSA26DTA/+ (Fig.3G) and ROSA26DTA/+ (Fig. 3M) lungs. By day 6, nascentregenerating Clara cells comprised 40% of the total airwayepithelium in PNEC-ablated (Fig. 3H) and control (Fig. 3N) lungs.After 1 mo, similar to ROSA26DTA/+ lungs (Fig. 3O), Clara cellshad almost completely regenerated in CGRPCreER/+;ROSA26DTA/+ lungs (Fig. 3I), whereas PNECs remained absent(Fig. 3F). To rule out the possibility that Clara cells in contact withPNECs before PNEC ablation had already received inductivesignals and acquired the potential to regenerate, we repeatednaphthalene-induced lung injury and still observed equallyefficient Clara cell regeneration. In all of these studies, therewas no difference in the kinetics of Clara cell regenerationbetween CGRPCreER/+;ROSA26DTA/+ and ROSA26DTA/+ mice,which suggests that Clara cell regeneration can occur in the ab-sence of PNECs. Previous studies suggest that a small populationof CGRP+;CC10+ cells within NEBs may give rise to regeneratedClara and ciliated cells during naphthalene-induced lung in-jury. We found very few CGRP+;CC10+ cells in naphthalene-injured lungs (Fig. 2B) and did not detect these doubly positivecells in uninjured or PNEC-ablated lungs. This population is thusunlikely to account for the speedy and extensive Clara cell re-generation after naphthalene-induced lung injury and is also in-capable of replenishing ablated PNECs.

Inactivation of p53, Rb, and Pten in Pulmonary Neuroendocrine CellsLeads to SCLC. Because up to 90% of human SCLCs have sus-tained mutations in the tumor suppressors p53 and Rb, we usedthe CGRPCreER mouse line to inactivate both p53 and Rb inPNECs. We bred mice to produce CGRPCreER/+;p53f/f;Rbf/f mice(f: floxed) and delivered TM to these animals. We found thatselective removal of p53 and Rb in PNECs resulted in PNEChyperplasia and tumor development a few months after TMadministration (e.g., 5–6 mo post-TM; Fig. 4B). In our study,among 32 TM-injected CGRPCreER/+;p53f/f;Rbf/f adult mice at5–6 mo of age, 4 developed PNEC hyperplasias and small lungtumors (Fig. 4B), 12 developed PNEC hyperplasia, and 16 had noobvious morphological defects. Because Pten mutations are alsofound in a significant number of human SCLCs (19, 20), wespeculated that mutations in Pten could cooperate with p53/Rbmutations to promote SCLC development. Indeed, loss of Pten, inaddition to p53 and Rb, significantly accelerated tumor initiationand progression. Analysis of these triple-mutant mice at 2–4 wkafter TM administration revealed the presence of small hyper-plastic lesions that are likely the precursors for SCLC (Fig. 4C).Compared with CGRPCreER/+;p53f/f;Rbf/f mice at the same stage,not only more and larger PNEC hyperplasias and SCLC tumorswere present in the sections of CGRPCreER/+;p53f/f;Rbf/f; Ptenf/f

lungs (Fig. 4 C and F), but unlike double-mutant mice, theselesions in triple-mutant mice at 2.5 mo post-TM showed signs oftumor invasiveness. For instance, malignant tumor cells (Fig. 4D)can be found surrounding and inside blood vessels (Fig. 4E). Atotal of 17/17 TM-injected CGRPCreER/+;p53f/f;Rbf/f;Ptenf/f adultmice developed PNEC hyperplasias (Fig. 4C) and large lungtumors (Fig. 4F) at 2–3 mo of age. All double- and triple-mutantmice analyzed developed thyroid tumors (Fig. S6). Most double-mutant mice died within 6–7 mo, and triple-mutant mice diedwithin 3 mo after TM injection, likely due to lesions in the lungand thyroid (Fig. S6). These results are consistent with a modelin which loss of tumor suppressors leads to progressive acquisi-tion of tumor properties.Tumors developed in p53/Rb double-knockouts or p53/Rb/Pten

triple-knockout mice are comprised of Ki67+ proliferating cells(Fig. 4G) that express neuroendocrine markers (Fig. 4 H and I).The proliferation rate increased considerably with time (Fig. 4J–O), and lesions in p53/Rb/Pten mice had a higher rate ofproliferation than those in p53/Rb mice at any time point post-TM (Fig. 4L). In fact, significant PNEC proliferation was apparenteven 1 wk post-TM in p53/Rb/Pten mice (Fig. 4K). These obser-vations confirm and extend previous reports (12–15) suggestingthat PNECs can be the cells of origin for SCLC, offering insight

Fig. 3. PNECs are dispensable for Clara cell regeneration during naphthalene-induced lung injury. (A–C ) Immunostaining of lung sections from adultCGRPCreER/+;ROSA26DTA/+ (experimental) and ROSA26DTA/+ (control) mice.Activation of CreER by TM induced DTA expression in CGRPCreER/+;ROSA26DTA/+

adult lungs, resulting in efficient killing of PNECs. PNECs (SYP+ or CGRP+)or NEBs could not be detected in these animals (A and B), whereas othercell types were unaffected (Fig. S5). Long-term tracing of PNECs in adultCGRPCreER/+;ROSA26DTA/+ mice showed no sign of PNEC replenishment. (D–I)Immunostaining of lung sections from adult CGRPCreER/+;ROSA26DTA/+ mice.Activation of CreER by TM in adult lungs induced DTA expression, resultingin efficient PNEC ablation. These mice were subsequently treated withnaphthalene to ablate Clara cells (CC10+). Despite the absence of PNECs(D–F ), Clara cell regeneration followed a time course (G–I) similar to thatin control mice (M–O). Extensive Clara cell proliferation was observed byday 6 after naphthalene injection (H), and Clara cells had fully regeneratedby 1 mo after the initial insult (I). (J–O) Immunostaining of lung sections fromadult ROSA26DTA/+ control mice that received an identical regimen of TMand naphthalene treatment. PNECs (arrows in J–L) were unaffected, andthe distribution of lung cell types and the kinetics of Clara cell regeneration(M–O) are similar to those in WT mice. (P) A model of PNEC specification andfunction during lung injury. (Scale bars: 100 μm for panels in each row)

17534 | www.pnas.org/cgi/doi/10.1073/pnas.1207238109 Song et al.

into the complex interactions between tumor suppressors intumor development.

SCLC Can Originate from Differentiated Pulmonary NeuroendocrineCells. Our CGRPCreER/+ mice offer an opportunity to examinetumor initiation and progression by labeling PNECs and fol-lowing their fate over time; this was achieved by producingCGRPCreER/+;ROSA26mTmG/+;p53f/f;Rbf/f mice. In differentiatedPNECs where CGRPCreER is expressed and activated by TM toremove p53 and Rb, these cells are also permanently labeled witheGFP (from ROSA26mTmG; Fig. 4 P and S), allowing us tofollow their fate during subsequent steps of cell transformationand tumor development. We examined labeled PNEC cells inCGRPCreER/+;ROSA26mTmG/+;p53f/f;Rbf/f lungs at various timepoints after TM injection. We found that almost all CGRP+ cellswithin the early hyperplastic lesions (i.e., containing a couple ofdozen cells; Fig. 4T) were also labeled by eGFP (Fig. 4 P and S), andmany of the labeled PNECs were actively dividing (Fig.4Q). Similarly, TM injection into CGRPCreER/+;ROSA26mTmG/+;p53f/f;Rbf/f;Ptenf/fmice labeled proliferating PNECs, which likely ledto tumor development (Fig. 4 V–X). Thus, themajority of, if not all,labeled proliferating PNECs in the hyperplasic lesions and tumorsare derived from PNECs that express multiple differentiationmarkers. These results support the idea that SCLC can originatefrom differentiated PNECs within NEBs.

DiscussionCompared with other lung cell types, in vivo analysis of PNECsis underexplored. In this study, we report a genetic tool to in-vestigate PNEC lineage and function. Our data suggest thatPNECs may share a common lineage with alveolar cells eitherfrom common or separate progenitors during lung developmentbut can contribute to Clara cells and ciliated cells in lung injury(Figs. 1 P and Q and 3P). The distinct cell fates adopted byPNECs during homeostasis and injury clearly demonstrate theplasticity of PNECs, and places PNECs in the same category asbasal cells, Clara cells, and type II cells, which can contributeto other cell types under different conditions. Different types ofinsults to lung epithelium induce distinct tissue responses. Forinstance, bleomycin administration eliminates alveolar cells andcauses fibrosis without apparent regeneration. By contrast, in-fluenza viral infection also kills alveolar cells, but these cells arereplenished through progenitor cell proliferation (21). It is pos-sible that with other forms of injury not tested in this study,PNECs can be induced to give rise to additional cell types. In-terestingly, PNECs cannot be replenished once ablated duringhomeostasis or naphthalene-induced lung injury. Similarly, a dif-ferent type of lung injury may induce PNEC regeneration. In allof these settings, the signals that trigger cell proliferation andfate change remain to be further investigated (22). Our resultsare consistent with the model in which homeostasis and tissueinjury provide distinct microenvironments for maintaining anintact epithelium, because each context directs PNECs togenerate different cell types.PNECs have been implicated in multiple aspects of lung func-

tions, including airway oxygen sensing, regulation of pulmonaryblood flow, control of bronchial tonus, modulation of immuneresponses, and maintenance of a stem cell niche. These diverseactivities are likely mediated by the amines and peptides releasedfrom PNECs, but their in vivo relevance is largely unconfirmed.Our CGRPCreER mice offer a unique opportunity to test thesehypotheses. Mice in which PNECs are ablated peri- or

Fig. 4. SCLC can originate from differentiated PNECs. (A–I) Histology andimmunostaining of lung sections from WT, CGRPCreER/+;p53f/f;Rbf/f [doubleknockout (DKO)], and CGRPCreER/+;p53f/f;Rbf/f;Ptenf/f [triple knockout (TKO)]adult mice. Cancer cells in triple-mutant mice displayed aberrant mitosis(arrow in D) and also showed signs of tumor invasiveness (arrow in E).Tumors are comprised of proliferating Ki67+ cells (G) and also expressedneuroendocrine markers such as Ascl1 (Mash1) (H) and SYP (I). (J–O) Immu-nostaining of lung sections from WT, CGRPCreER/+;p53f/f;Rbf/f, and CGRPCreER/+;p53f/f;Rbf/f;Ptenf/f adult mice. PNECs lacking p53 and Rb displayed extensivePNEC proliferation as judged by Ki67 staining (J). Proliferating PNECs couldbe found 1 wk after TM injection (K) in triple-mutant mice, and the pro-liferation rate increased substantially with time (L–O). (P–X ) Immunos-taining of lung sections from WT, CGRPCreER/+;ROSA26mTmG/+;p53f/f;Rbf/f,and CGRPCreER/+;ROSA26mTmG/+;p53f/f;Rbf/f;Ptenf/f adult mice. TM injection se-lectively ablated p53/Rb or p53/Rb/Pten in PNECs, which were simultaneously

labeled by eGFP. Inactivating p53 and Rb in PNECs (CGRP+) (T) led to ex-tensive PNEC proliferation as observed by Ki67 staining (Q). Most, if not allPNECs, were also labeled by eGFP (P and S), suggesting that differentiatedPNECs proliferated and subsequently led to tumor growth. Similarly, TMinjection into CGRPCreER/+;ROSA26mTmG/+;p53f/f;Rbf/f;Ptenf/f mice labeledproliferating PNECs (V and W ), which likely led to tumor development.(Scale bars: A–C and F–I, 100 μm; D, E, J, K, and M–X, 20 μm.)

Song et al. PNAS | October 23, 2012 | vol. 109 | no. 43 | 17535

GEN

ETICS

postnatally provide an ideal setting to explore various physio-logical functions of PNECs.Contrary to the current model, our results demonstrate that

PNEC ablation has no apparent effect on Clara cell regenerationin naphthalene-induced lung injury. Though we cannot formallyexclude the possibility that a very small number of residual PNECsmaintain a stem cell niche required for Clara cell recovery, thespeedy and extensive regeneration of Clara cells even after re-peated injury renders such a possibility remote, which suggeststhat signals derived from cells other than PNECs are responsiblefor Clara cell proliferation. Nonetheless, PNECs may play a re-dundant role or be required for Clara cell regeneration in otherforms of injury.Mouse models of SCLC have been reported in which tumor

suppressors p53 and Rb are inactivated in the lung epithelium viaintrabronchial injection of adenoviral Cre (12) or an adenovirus(Ade-CGRP-Cre) in which a ∼2-kb rat CGRP promoter wasused to drive Cre expression (13). Ade-CGRP-Cre is expressedin PNECs as well as in a small number of Clara cells and ciliatedcells (13). Importantly, these mouse models recapitulate criticalaspects of human SCLC. Our CreER-based genetic systemcomplements studies using an adenovirus-based approach. Inparticular, our genetic system allows us to manipulate geneactivity and label PNECs to follow their progression during tu-mor development. We observed earlier lethality associated withsevere defects in the lung and/or other neuroendocrine organswhen using the murine CreER system activated by full doses ofTM injection. Though earlier lethality may limit the study oftumor metastasis, it has no impact on investigations of tumorinitiation and progression. Furthermore, we expect to overcomethis problem by injecting a lower dose of TM or by trans-plantation of tumors into a host mouse to study tumor behaviorover an extended time course. Our results suggest that majortumor suppressors display complex interactions in SCLC de-velopment. Though removal of Pten in the background of p53/Rbmutations accelerates tumor development and confers tumorinvasiveness, loss of p53 and Pten does not appear to exert ob-vious effects within the same time frame. Identifying genes andpathways perturbed in tumors at various stages of progressionwould be a critical step toward understanding how these tumorsuppressors individually and in combination control tumordevelopment.Our investigation of the relationship between PNECs and

SCLC offers insight into the early stages of SCLC development(Fig. S6). Labeled PNECs within NEBs proliferate in early hy-perplastic lesions. These cells, characterized by expression of all

known PNEC markers, constitute the majority of dividing cellswithin the hyperplastic lesions upon removal of tumor suppressors.Tumors most likely are derived from proliferating PNECs. Thesefindings provide strong evidence to support the notion that SCLCcan originate from terminally differentiated PNECs that seldomproliferate during homeostasis. We cannot rule out the possi-bility that a subpopulation of PNECs that expresses low levels ofCGRP and possesses special properties can constitute the cell oforigin of SCLC. However, a very high percentage of labeledPNECs at early stages of tumor development after removal ofp53/Rb or p53/Rb/Pten would suggest that if such a PNEC sub-population exists, it is either prevalent or highly proliferative foran extended period to generate tumors. In this scenario, in con-trast to the high proliferative potential of this special PNECsubpopulation, the rest of differentiated PNECs lacking p53/Rbor p53/Rb/Pten can only undergo limited division and do notcontribute significantly to tumor formation. Without additionalmarkers, distinguishing distinct subpopulations of differenti-ated PNECs is not currently feasible. In this regard, ourCGRPCreER system provides a means to isolate a pure popula-tion of PNECs through fluorescence-activated cell sorting formolecular characterization.Though PNECs can give rise to SCLC, ablation of p53 and Rb

in surfactant associated protein C-positive (SP-C+) cells using anadenoviral approach also leads to SCLC, albeit at a much lowerrate (13). It is unclear how the loss of tumor suppressors in SP-C+ cells allows these cells to acquire neuroendocrine proper-ties. Whether other cell types can also give rise to SCLC hasnot been extensively examined. We surmise that during tumordevelopment, cells are converted to relatively immature (oruncommitted) states that reflect their developmental origin. Inthis case, it is interesting to note that PNECs and alveolar cellsmay share a common lineage during lung development. Genomicstudies would be instrumental in identifying genes and pathwaysthat underlie early stages of transformation in PNECs or otherlung cell types—some of them would be ideal candidates forearly diagnosis of SCLC or driver mutations that can be tailoredfor targeted therapies.

Materials and MethodsMouse strains and standard procedures used in this study are described in SIMaterials and Methods.

ACKNOWLEDGMENTS. We thank Peter Baluk for discussion on cardiomyo-cytes in the pulmonary vein and Ming-Tseh Lin for advice on tumor histology.

1. Morrisey EE, Hogan BL (2010) Preparing for the first breath: Genetic and cellularmechanisms in lung development. Dev Cell 18(1):8–23.

2. Domyan ET, Sun X (2011) Patterning and plasticity in development of the respiratorylineage. Dev Dyn 240(3):477–485.

3. Warburton D, et al. (2010) Lung organogenesis. Curr Top Dev Biol 90:73–158.4. Cardoso WV, Whitsett JA (2008) Resident cellular components of the lung: Developmental

aspects. Proc Am Thorac Soc 5(7):767–771.5. Linnoila RI (2006) Functional facets of the pulmonary neuroendocrine system. Lab

Invest 86(5):425–444.6. Domnik NJ, Cutz E (2011) Pulmonary neuroepithelial bodies as airway sensors: Puta-

tive role in the generation of dyspnea. Curr Opin Pharmacol 11(3):211–217.7. Mahvi D, Bank H, Harley R (1977) Morphology of a naphthalene-induced bronchiolar

lesion. Am J Pathol 86(3):558–572.8. Stripp BR, Maxson K, Mera R, Singh G (1995) Plasticity of airway cell proliferation

and gene expression after acute naphthalene injury. Am J Physiol 269(6 Pt 1):L791–L799.

9. Reynolds SD, Giangreco A, Power JH, Stripp BR (2000) Neuroepithelial bodies ofpulmonary airways serve as a reservoir of progenitor cells capable of epithelial re-generation. Am J Pathol 156(1):269–278.

10. Reynolds SD, et al. (2000) Conditional Clara cell ablation reveals a self-renewingprogenitor function of pulmonary neuroendocrine cells. Am J Physiol Lung Cell MolPhysiol 278(6):L1256–L1263.

11. van Meerbeeck JP, Fennell DA, De Ruysscher DK (2011) Small-cell lung cancer. Lancet378(9804):1741–1755.

12. Meuwissen R, et al. (2003) Induction of small cell lung cancer by somatic inactivationof both Trp53 and Rb1 in a conditional mouse model. Cancer Cell 4(3):181–189.

13. Sutherland KD, et al. (2011) Cell of origin of small cell lung cancer: Inactivationof Trp53 and Rb1 in distinct cell types of adult mouse lung. Cancer Cell 19(6):754–764.

14. Park KS, et al. (2011) Characterization of the cell of origin for small cell lung cancer.Cell Cycle 10(16):2806–2815.

15. Schaffer BE, et al. (2010) Loss of p130 accelerates tumor development in a mousemodel for human small-cell lung carcinoma. Cancer Res 70(10):3877–3883.

16. Rawlins EL, Clark CP, Xue Y, Hogan BL (2009) The Id2+ distal tip lung epitheliumcontains individual multipotent embryonic progenitor cells. Development 136(22):3741–3745.

17. Mueller-Hoecker J, et al. (2008) Of rodents and humans: A light microscopic and ul-trastructural study on cardiomyocytes in pulmonary veins. Int J Med Sci 5(3):152–158.

18. Rawlins EL, et al. (2009) The role of Scgb1a1+ Clara cells in the long-termmaintenanceand repair of lung airway, but not alveolar, epithelium. Cell Stem Cell 4(6):525–534.

19. Yokomizo A, et al. (1998) PTEN/MMAC1 mutations identified in small cell, but not innon-small cell lung cancers. Oncogene 17(4):475–479.

20. Kim SK, et al. (1998) Alterations of PTEN/MMAC1, a candidate tumor suppressor gene,and its homologue, PTH2, in small cell lung cancer cell lines. Oncogene 16(1):89–93.

21. Kumar PA, et al. (2011) Distal airway stem cells yield alveoli in vitro and during lungregeneration following H1N1 influenza infection. Cell 147(3):525–538.

22. Xing Y, Li A, Borok Z, Li C, Minoo P (2012) NOTCH1 is required for regeneration ofClara cells during repair of airway injury. Stem Cells 30(5):946–955.

17536 | www.pnas.org/cgi/doi/10.1073/pnas.1207238109 Song et al.

Supporting InformationSong et al. 10.1073/pnas.1207238109SI Materials and MethodsMice.Toproduce theCGRPCreERmouse line, the translational startATG in the mouse CGRP genomic locus was replaced witha CreER-FRT-PGK-Neo-FRT cassette through gene targeting inE14 ES cells. The FRT-PGK-Neo-FRT cassette was subsequentlyremoved by crosses with FLPe mice. p53 floxed [FVB.129-Trp53tm1Brn] and Rb floxed [FVB;129-Rb1tm2Brn] mice wereobtained from the National Cancer Institute Mouse Repository.Shh-Cre [B6.Cg-Shhtm1(EGFP/cre)Cjt/J], Nkx2.1-Cre [C57BL/6J-Tg(Nkx2-1-cre)2Sand/J], Pten floxed [C;129S4-Ptentm1Hwu/J], RO-SA26mTmG [STOCK Gt(ROSA)26Sortm4(ACTB-tdTomato-EGFP)Luo/J],R26R [B6;129S4-Gt(ROSA)26Sortm1Sor/J], andROSA26eGFP-DTA (ab-breviated as ROSA26DTA in this study) [STOCK Gt(ROSA)26Sortm1(DTA)Jpmb/J] mice were obtained from Jackson Labora-tory. The Institutional Animal Care and Use Committee at theUniversity of California, San Francisco, approved all experimentsperformed in this study.

Tamoxifen and Naphthalene Administration. Tamoxifen (TM; Sigma)was dissolved inMazola corn oil to make a stock solution of 50 mg/mL. For embryonic lineage tracing, pregnant female mice wereinjected intraperitoneally (i.p.) with a single dose of TM (0.25 mg/gbody weight). For lineage tracing in adults, TM was injected intoadult mice daily for 4 d. Lungs were collected for analysis at in-dicated time points. For naphthalene-induced lung injury, naph-thalene (Sigma) was dissolved in corn oil at 50 mg/mL. Adult micewere first injected with TM every other day for four times. The nextday following the last dose of TM, mice were injected i.p. withnaphthalene (275 mg/kg body weight). Lungs were collected foranalysis at indicated time points. At least three animals were an-alyzed for each genotype.

Histology, Immunohistochemistry, and β-Gal Staining. Mouse lungswere inflated with 1 mL of 2% paraformaldehyde (PFA) and fixedfor 4 h at 4 °C in 2% PFA. After dehydration and processing,lungs were embedded in paraffin for sectioning and immuno-histochemistry. All of the tissues collected were sectioned at 6 μmthickness for histological analysis. Paraffin sections were stainedwith the following primary antibodies: chick anti-GFP (1:200;Abcam), rabbit anti-calcitonin gene-related peptide (CGRP, 1:400;Sigma), goat anti-Clara cell 10-kDa protein (CC10, 1:200; SantaCruz), rabbit anti-prosurfactant protein C (SP-C, 1:400; Millipore),rabbit anti- synaptophysin (SYP, prediluted; Life Technologies),mouse anti-Ascl1 (Mash1) (1:200; BD), mouse anti-Ki67 (1:200;BD), mouse anti-acetylated (Ac)-tubulin (1:1,000; Sigma), andhamster anti-T1α (1:200; Developmental Studies HybridomaBank). Secondary antibodies and conjugates used includedonkey anti-chick DyLight 488 (1:1,000; Jackson ImmunoR-esearch Laboratories), donkey anti-rabbit Alexa Fluor 594(1:1,000; Life Technologies), donkey anti-goat Alexa Fluor 594or 647 (1:1,000; Life Technologies), donkey anti-mouse AlexaFluor 594 or 647 (1:1,000; Life Technologies), and DAPI(1:10,000; Sigma). For biotinylated secondary antibodies (goat

anti-rabbit, 1:1,000; donkey anti-goat, 1:1,000; and horse anti-mouse, 1:1,000; Jackson ImmunoResearch Laboratories), thesignal was detected using HRP-conjugated streptavidin (1:1,000;Vector Laboratories) in combination with either the chromogenicsubstrate DAB (Vector Laboratories) or fluorogenic substrateAlexa Fluor 594 tyramide (1:200, TSA kit; Perkin-Elmer). Anti-bodies against Ascl1 (Mash1), Ki67, and SYP may require thebiotin-streptavidin amplification step for optimal signal detection.Fluorescent images were acquired using a SPOT 2.3 CCD cameraconnected to a Nikon E1000 epifluorescence microscope. Ad-justment of red/green/blue histograms and channel merges wereperformed using SPOT Advanced software or ImageJ. Confocalimages were captured on a Leica laser-scanning confocal mi-croscope. Multiple optical sections were obtained to distinguishcell boundaries. For X-gal staining, lungs were inflated with 1 mLof 1% PFA and 0.2% gluteraldehyde and fixed 1 h at 4 °C in 1%PFA and 0.2% gluteraldehyde. To detect β-gal activity, wholelungs or frozen sections (10 μm) were stained in X-gal stainingsolution following standard procedures.

Analysis of Cell Proliferation. In Fig. 2H, in which proliferatingCGRP+ cells were quantified after naphthalene-induced lunginjury, three mice were analyzed at each time point. A total of326 CGRP+ cells at day 0, 182 CGRP+ cells at day 3, and 245CGRP+ cells at day 6 were characterized. The data (ratio ofKi67+CGRP+/CGRP+) were presented as mean ± SEM. P <0.0005 by two-tailed t test.In Fig. 4L, in which proliferating CGRP+ cells were quantified

after removal of tumor suppressors, three mice were analyzed ateach time point for a given genotype. CGRP+ cells characterizedin this study included 200 double-knockout (DKO) and 200 tri-ple-knockout (TKO) cells at 1 wk; 312 DKO and 534 TKO cellsat 1 mo; and 356 DKO and 1,023 TKO cells at 2 mo. The data(ratio of Ki67+CGRP+/CGRP+) were presented as mean ± SEM.P < 0.0009 by two-tailed t test.

Model of Pulmonary Neuroendocrine Cell Specification and FunctionDuring Lung Homeostasis and Injury. Lineage tracing during em-bryogenesis suggests that pulmonary neuroendocrine cells (PNECs)share a common lineage with alveolar cell types (type I and type IIpneumocytes) from either common or separate progenitors (Fig.1 P and Q). Clonal analysis will provide further insight into thisprocess. During lung injury, PNECs proliferate and can give riseto Clara cells and ciliated cells in addition to new PNECs (Fig.3P), whereas loss of PNECs has no impact on Clara cell re-generation (Fig. 3P). Accumulation of mutations in PNECs leadsto progressive acquisition of carcinoid or small-cell lung cancerproperties (Fig. S6).In a recent study, Li and Linnoila (1) reported that Ascl1+ cells

can contribute to airway and alveolar cells during embryogenesisand produce Clara, ciliated, and alveolar cells following naph-thalene-induced injury. Further study will shed light on the re-lationship between the CGRP+ and Ascl1+ populations.

1. Li Y, Linnoila I (2012) Multi-directional differentiation of Ascl1-defined progenitors inlung development and injury repair. Am J Respir Cell Mol Biol, 10.1165/rcmb.2012-0027OC.

Song et al. www.pnas.org/cgi/content/short/1207238109 1 of 5

A

NEB

B C

D

NEB

Ciliated cell

Clara cellTII cell

TI cell

E F

Ac-tubulin SYP SYP

CC10 SP-C T1

Fig. S1. Major epithelial cell types in the mammalian lung. (A–F) Immunostaining of lung sections from wild-type (WT) adult mice to label distinct epithelialcell types, including ciliated cells (Ac- tubulin), PNECs in neuroepithelial bodies (NEBs; SYP), Clara cells (CC10), type II cells (SP-C), and type I cells (T1α). Arrows ineach panel points to the indicated cell type, which displays characteristic properties. (Scale bar: 75 μm.)

eGFP (mTmG) anti-CGRP merged

+ tamoxifen

+ tamoxifen

tamoxifen

CreERT2tPA

ATG

ATG

1 2 3 4 5 6

Common exons Calcitonin-specific exon CGRP-specific exons

4 5 6

Wt mouse CGRP locus

Targeted CGRPCreER locus31

A

B C D

E F G

H I J

CGRPCreER/+

; ROSA26mTmG/+

Fig. S2. CGRPCreER mice confer spatial and temporal control of gene activities in PNECs. (A) Schematic diagram depicting gene targeting of a CreERT2tPAcassette at the translational start (ATG) of the mouse Calcitonin/CGRP genomic locus. The resulting allele is designated CGRPCreER in this study. CGRPCreER

insertion disrupts the production of Calcitonin/CGRP mRNA, but homozygous CGRPCreER mice are viable and exhibit no obvious phenotypes similar to previousreports of Calcitonin/CGRP knockout mice (1). (B–J) Immunostaining of lung sections from adult CGRPCreER/+;ROSA26mTmG/+ mice. Cre activation upon TMadministration resulted in eGFP expression (detected by anti-GFP antibodies) from the ROSA26mTmG allele by removing sequences that block its expression.Efficient activation of CreER from the CGRP locus was observed as judged by the exclusive presence of eGFP signal (green) in PNECs after TM administration toadult mice. Nearly every PNEC in the NEB was labeled by eGFP (B and E). PNECs were identified by anti-CGRP (blue; C, F, and I) as well as other antibodies (notshown). No eGFP signal was detected in mice without TM injection (H–J). (Scale bar: 35 μm.)

1. Hoff AO, et al. (2002) Increased bone mass is an unexpected phenotype associated with deletion of the calcitonin gene. J Clin Invest 110(12):1849–1857.

Song et al. www.pnas.org/cgi/content/short/1207238109 2 of 5

eGFP anti-SMA DAPI mergedA B C D

TM 13.5 dpc

TM 13.5 dpc

P1

P30

E F G H

I J K L

TM 13.5 dpc P1

Fig. S3. Some muscle cells (likely cardiomyocytes) in the pulmonary veins were labeled in CGRPCreER mice. (A–L) Immunostaining of lung sections fromCGRPCreER/+;ROSA26mTmG/+ mice at postnatal day 1 (P1) (A–H) or P30 (I–L). TM was injected at 13.5 days postcoitum (dpc). Some muscle cells [smooth muscleactin (SMA+)] (A–H) labeled by eGFP at P1 were subsequently found in the pulmonary veins (I–L). Similar results were obtained when tamoxifen was ad-ministered before 15.5 dpc. These eGFP-labeled cells (I) in adults no longer express SMA; instead, they display characteristic histological features, includingstriations reminiscent of cardiomyocytes. Cardiomyocytes are well documented to be present in pulmonary veins between heart and lung (but not in intra-pulmonary veins). (Scale bar: 75 μm.)

B

Brain Thyroid

Skin Mammary gland Kidney Adrenal medulla

C

E F GIntestine

D

H

A

Brain

eGFPeGFP

Fig. S4. Detection of CGRPCreER activity in various mouse tissues. (A–H) Immunostaining of sections from adult CGRPCreER/+;ROSA26mTmG/+ mice injected withTM postnatally. eGFP signals can be detected in several mouse tissues other than the lung. For instance, several neurons and fibers in the brain were labeled.Neurons and fibers in A and B are located near the cerebral cortex, as indicated by the boxed area in A Inset. In addition, parafollicular C cells in the thyroid,and scattered cells in the intestine, skin, mammary gland, kidney, and adrenal medulla, also displayed CGRPCreER activity. Ascl1CreER is more broadly expressed,at least in the nervous system, than CGRPCreER. It would be interesting to compare results of lineage and functional studies on PNECs using Ascl1CreER andCGRPCreER. (Scale bar: 25 μm.)

Song et al. www.pnas.org/cgi/content/short/1207238109 3 of 5

CGRPCC10

Ac-TubulinCC10

Ac-TubulinCC10

CGRPCC10

CGRPCreER/+

; ROSA26DTA/+

ROSA26DTA/+

NEB

CGRPCreER/+

; ROSA26DTA/+

ROSA26DTA/+

CGRPCreER/+

; ROSA26DTA/+

ROSA26DTA/+

NEB

SYPSYP

A B

C D

E F

2.5 mon post-DTA activation

Fig. S5. Efficient and selective ablation of PNECs by diphtheria toxin (DTA) expression. (A–F) Immunostaining of lung sections from adult CGRPCreER/+;ROSA26DTA/+ mice. ROSA26DTA/+ mice serve as controls. Activation of CreER by TM in adult lungs induced DTA expression, resulting in efficient cell killing. PNECsor NEBs could not be detected in these animals (A), whereas other cell types were unaffected (C and E). We examined over 100 sections of each lung to coverdifferent regions, and it was very difficult to find any CGRP+ or SYP+ cells. We did find occasional solitary PNECs in CGRPCreER/+;ROSA26DTA/+ lungs, likely due toincomplete ablation and not a result of PNEC regeneration because the occurrence of residual PNECs is similar (1 PNEC per 20 sections) at any given time pointafter ablation. We failed to detect any PNEC clusters 1 y after PNECs’ ablation. Distribution of cell types in ROSA26DTA/+ mice (B, D, and F) is similar to that in WTlungs. The epithelial cells were identified by a battery of cell type-specific markers, such as SYP (PNEC), CGRP (PNEC), CC10 (Clara cell), and Ac-tubulin (ciliatedcell) shown here. (Scale bar: 75 μm.)

AWild-type thyroid

B

C

D

E

F

p53 ; Rb p53 ; Rb ; Pten/2.5 mon post-TM 2 mon post-TM

PNECHyperplasiaTransformation

PNEC (carcinoid or SCLC)

ParafollicularC cells

HyperplasiaTransformation

Medullary thyroid tumor

G

Fig. S6. Development of thyroid tumors from CGRP+ cells in the thyroid. (A–F) Histology of thyroid sections from WT, CGRPCreER/+;p53f/f;Rbf/f, and CGRPCreER/+;p53f/f;Rbf/f;Ptenf/f adult mice injected with TM postnatally. A total of 32 CGRPCreER/+;p53f/f;Rbf/f adult mice and 17 CGRPCreER/+;p53f/f;Rbf/f;Ptenf/f adult mice wereused in this analysis. All mice developed thyroid tumors. Normal thyroid architecture is destroyed, which could lead to thyroid dysfunction. TM-injectedCGRPCreER;p53f/f;Rbf/f and CGRPCreER;p53f/f;Rbf/f;Ptenf/f mice developed lung and thyroid tumors but not other tumors at the time of analysis. No tumors werefound in the lymph nodes adjacent to the thyroid or tissues adjacent to the thyroid. These observations suggest that lung and thyroid tumors represent primarytumors, and there is no distant metastasis from thyroid or lung tumors at the time of our analysis. These mice died most likely from a combination of re-spiratory failure and thyroid dysfunction, including lung and thyroid tissue destruction, obstruction of the trachea by the thyroid tumor, and adverse effectsfrom secreted amines and peptides from the tumors. (Scale bars: A/C/E and B/D/F, 100 μm.) (G) A model of tumor development by disruption of tumor sup-pressors in CGRP+ cells in different tissues.

Song et al. www.pnas.org/cgi/content/short/1207238109 4 of 5

Table S1. Percentage of lineage-labeled lung epithelial cell types after TM injections at indicated time during embryonic development

Time of TMadministration

Labeled neuroendocrinecells, CGRP+

Labeled Claracells, CC10+

Labeled ciliatedcells, Ac-tubulin+

Labeled type IIcells, SP-C+

Labeled type Icells, T1α+

E11.5 0% 0% 0% 0.2 ± 0.1% (7 ± 3%) 1.2 ± 0.4% (92 ± 3%)n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

E12.5 30 ± 10% (15 ± 6%) 0% 0% 0.2 ± 0.07% (6 ± 4%) 1.1 ± 0.3% (77 ± 4%)n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

E13.5 43 ± 11% (21 ± 3%) 0% 0% 0.12 ± 0.3% (7 ± 2%) 0.5 ± 0.1% (70 ± 4%)n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

E14.5 49 ± 7% (33 ± 4%) 0% 0% 0.09 ± 0.04% (5 ± 2%) 0.34 ± 0.2% (61 ± 5%)n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

E15.5 61 ± 9% (100%) 0% 0% 0% 0%n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

P0 66 ± 13% (100%) 0% 0% 0% 0%n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

Lungs were collected at postnatal day 30 for analysis. The numbers in parentheses indicate the percent of eGFP-labeled cells that generated a given cell type.

Table S2. Percentage of lineage-labeled lung epithelial cell types at different time points after four doses of TM injection in adult mice

Post-TMadministration

Labeled neuroendocrinecells, CGRP+

Labeled Claracells, CC10+

Labeled ciliatedcells, Ac-tubulin+

Labeled type IIcells, SP-C+

Labeled type Icells, T1α+

3 d 75 ± 7% 0% 0% 0% 0%n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

6 mo 70 ± 6% 0% 0% 0% 0%n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

1 y 61 ± 9% 0% 0% 0% 0%n = 3 mice n = 3 mice n = 3 mice n = 3 mice n = 3 mice

Song et al. www.pnas.org/cgi/content/short/1207238109 5 of 5