Developmental Programs of Lung Epithelial Progenitors: a Balanced Progenitor Model

Abstract

The daunting task of lung epithelium development is to transform a cluster of foregut progenitors into a three-dimensional tubular network with distinct cell types distributed at their appropriate locations. A complete understanding of lung development needs to address not only how, but also where, different cell types form. We propose that the lung epithelium forms through regulated deployment of three developmental programs: branching morphogenesis to expand progenitors and build a tree-like tubular network, airway differentiation to specify cells for the proximal conducting airways, and alveolar differentiation to specify cells for the peripheral gas exchange region. Each developmental program has its unique morphological features and molecular control mechanisms; their spatiotemporal coordination can be accounted for in a balanced progenitor model where progenitors balance between alternative developmental programs in response to spatiotemporal cues. This model integrates progenitor morphogenesis and differentiation, and provides new insights to lung immaturity in preterm birth and lung evolution. Advanced gene targeting and three-dimensional imaging tools are needed to achieve a comprehensive understanding of lung epithelial progenitors on molecular, cellular and morphological levels.

Introduction

The epithelium in the mammalian lung forms a tree-like tubular network consisting of two functional compartments: the proximal conducting airways and the peripheral gas exchange region. Air flows through the conducting airways, and is filtered, warmed and moistened before reaching the peripheral compartment where O₂ and CO₂ exchange with the blood occurs. Each compartment has distinct cell types with basal, secretory, ciliated and neuroendocrine cells found in the conducting airways and alveolar type 1 and 2 cells found in the gas exchange region. Therefore, the lung epithelium is far from an aggregate of different cell types and lung development is about not only how cell types are specified, but also how such specification is coupled with building a tubular network.

Significant progress has been made in our understanding of lung development in particular through phenotypic analysis of mouse mutants and has been summarized in several excellent reviews^1–4. Here we focus mainly on data from mouse models and attempt to formulate an integrated model of morphogenesis and differentiation with the goal of explaining both how and where in the three-dimensional (3D) respiratory tree different cell types form. We focus on developmental processes, instead of genes, to systematically identify the steps that are required to transform a cluster of progenitors into an organized and functional lung, including those steps that are essential but unexplored. Specifically, three developmental programs must be used to build a functional lung: branching morphogenesis to build a tree-like tubular network, airway differentiation to specify cells for the proximal conducting airways, and alveolar differentiation to specify cells for the peripheral gas exchange region. In this review, we first review the origin of lung epithelial progenitors, then the morphological features and molecular control mechanisms of individual developmental programs separately and lastly discuss their precise spatiotemporal deployment in a balanced progenitor model.

Two concepts in the review need to be defined to avoid potential confusion. First, the epithelial progenitors are defined as cells at the distal end of the growing respiratory tree that give rise to both airway and alveolar cells. The “proximal progenitors” for the airway cells in other reviews^{1, 2} are the descendants of, rather than the proximal equivalent of, the progenitors in this review. Second, the branching program includes both morphological (new branch formation) and cellular (self-renewal) aspects of the progenitors. We consider these two aspects inseparable because normally only progenitors form new branches and branching expands progenitors; abnormal progenitor maintenance can affect branching; and branching is not a separate development process that is completed before airway and alveolar differentiation.

I. Origin of lung epithelial progenitors

The lung originates as the result of specification and separation of a group of progenitors from the ventral region of the anterior foregut (Fig. 1). This process requires generation and interpretation of positional information along both the anterior-posterior and dorsal-ventral axes and also a morphogenetic event separating a single foregut tube into the dorsally located esophagus and ventrally located trachea tubes. On the molecular level, lung progenitor specification requires activation of lung lineage transcription factor Nkx2.1 in the ventral region and temporary suppression of Sox2 and p63 transcription factors that are enriched in the adjacent dorsal region⁵. Since Sox2 and p63 are also expressed in the lung later in development^{6, 7}, identification of esophagus specific lineage markers should be useful for future study of foregut specification.

Figure 1.

OPT imaging of foregut specification.

(A) OPT images of an E9.5 mouse embryo immunostained for E-Cadherin (a pan-epithelium marker), SOX2 (foregut and future conducting airway marker) and SOX9 (lung progenitor marker) showing the formation of the lung (Lu), liver (Li), dorsal and ventral pancreas (dP and vP). The left (filled arrowhead) and right (open arrowhead) lung buds are viewed from different angles (0°, 45° and 90°). Cross sections at four levels of the anterior foregut (I, II, III and IV) show that SOX9-expressing lung buds (cross section III) occupy a subregion of SOX2 low-expressing foregut (cross sections II and III), which is presumably also NKX2.1 positive. Scale: 100 um. (B) Schematics showing the specification of the trachea and lung from the ventral anterior foregut between E9.5 and E10.5. The side view is drawn based on (A). The contribution of SOX9 expressing cells at E9.5 to the trachea and lung is unknown (question mark). A cross section of the trachea region at E9.5 and a longitudinal section of the lung bud at E10.5 are shown in the lower panel to illustrate the similarity with respect to epithelial gene expression and mesenchymal signals.

Positional information is often encoded by localized production and secretion of signaling molecules. Although it is unclear whether and how positional information during earlier embryogenesis is utilized to establish their expression patterns, Wnt2, Wnt2b and Bmp4 are expressed in the ventral foregut mesenchyme adjacent to the future lung progenitors^8–10. Loss-of-function experiments show that dorsal-ventral patterning is disrupted in the absence of either Bmp or Wnt ligands in the mesenchyme or the Bmp receptors or Wnt signal mediator (Ctnnb1) in the foregut epithelium^{8, 10–12}. Gain-of-function experiments show that ectopic activation of canonical Wnt signaling is sufficient to expand the lung progenitor domain posteriorly to the stomach epithelium^{8, 11}. Epistasis experiments show that Wnt signaling activation is insufficient to restore lung progenitors in the Bmp receptor mutant, suggesting Wnt signaling is parallel to or upstream of Bmp signaling¹². Interestingly, Bmp4 and Bmp receptor mutants initiate Nkx2.1 expression normally^{10, 12}. Although this may reflect the timing of gene inactivation, it remains possible that Bmp signaling drives the ventral outgrowth and consequently separation of the trachea, which is necessary to avoid a putative dorsal inhibitory signal such as Noggin⁵ and maintain Nkx2.1 expression. This possibility can be tested by examining Nkx2.1 expression in mosaic Bmp receptor mutants without a tube separation defect.

Foregut patterning along the anterior-posterior axis is less understood. Despite its well-established role in other developmental contexts, there is no good evidence that combinatorial Hox gene expression is used to pattern the foregut along this axis¹³. Instead, the lung progenitor field is adjacent to the heart and organ culture experiments show that this spatial proximity exposes the lung progenitor field to a high concentration of heart-derived Fgfs, which appears to favor the lung cell fate over the more posterior liver cell fate¹⁴. Although this finding needs in vivo validation, crosstalk between the lung and the heart may be used to ensure their co-evolution¹⁵.

The separation of the foregut tube into the esophagus and the trachea is a striking case of tubulogenesis, involving fusion of opposing epithelial walls. Successful tube separation depends on correct specification of both the esophagus and the trachea as demonstrated in Nkx2.1, Sox2 and aforementioned Bmp and Wnt signaling mutants^{8, 10–12, 16, 17}. Little is known about the underlying cell biology; it appears analogous to the formation of the semicircular canals via fusion in the center of an epithelial pouch, although loss of an apoptotic factor affects inner ear, but not foregut, morphogenesis^18–20 Live imaging of the separation process in culture is likely to shed light on this unique tubulogenesis process and its failure in human tracheal atresia and tracheoesophageal fistula²¹.

How about the more posteriorly-located lung buds that give rise to the lung itself? Since both the trachea and lung buds express and require Nkx2.1 for their normal development¹⁶, it is conceivable that the lung buds are specified and separated in a similar process as the trachea and that the left-right buds reflect the bilateral body plan instead of a branching process reviewed in details later. This argument is supported by the presence of left-right lung buds even when subsequent branching is completely blocked in the Fgfr2 mutant²². In this context, it is worth noting the similarity between the dorsal-ventral axis of the esophagus/trachea and the proximal-distal axis of the lung buds with respect to polarized expression of Sox2 in the epithelium and inductive signals in the mesenchyme (Fig. 1B). However, loss of the Bmp receptors affects Nkx2.1 expression in the trachea but not the lung buds whereas the Fgf10 mutant appears to have the trachea but not lung buds^{23, 24}, suggesting distinct mechanisms for trachea versus lung bud specification¹². We have used a 3D imaging technology, optical projection tomography (OPT)²⁵, to study foregut specification and show that SOX9, a transcription factor for lung epithelial progenitors, is expressed in a subset of SOX2 low-expressing cells in the ventral anterior foregut from the earliest stage of lung bud formation (Fig. 1). Future lineage tracing experiments will determine the relative contributions of Sox9 and Nkx2.1 expressing cells to the trachea and lung.

II. The branching morphogenesis program

After specification and separation from the foregut, lung epithelial progenitors undergo an elaborate process of branch formation termed branching morphogenesis (Fig. 2A). This developmental program not only builds a tree-like structure for efficient space utilization, but also expands the progenitors at the distal end and leaves behind their differentiating descendants, ensuring that the lung is not a random aggregate of different cell types. We have used a combination of OPT and confocal microscopy to track the complete developmental history of lung epithelial progenitors and shown that new branches are added as late as embryonic day (E) 18.5 in mice, after which progenitors cluster around expanded air lumens and are eventually depleted (Fig. 2A).

Figure 2.

A complete developmental history of lung epithelial progenitors.

(A) OPT (top panels, scale: 250 um) and confocal (bottom panels, scale: 20 um) projection images of mouse lungs at indicated embryonic stages immunostained for E-Cadherin and SOX9. The confocal images show the growing edges of the lungs, allowing comparison over developmental time. The E-Cadherin staining is shown as a sectional view in grey scale. The branching program expands progenitors (filled arrowhead) and leaves behind differentiating descendants (open arrowhead) for the conducting airways (top panels) or the gas exchange region (bottom panels). Branches with a bud-stalk structure form as late as E18.5 (dashed line in bottom panels), after which progenitors cluster around an expanded lumen (postnatal day (P)2, P7) and eventually are depleted (~P14) and replaced by juxtaposed AT1 and AT2 cells tiling the gas exchange surface. Alveolar walls can form between (early) or within (late) branches. The accessary lobe grows from the right main bronchus to the left side (dashed line in top panels). SOX9 is also expressed in the cartilages (square bracket). (B) Two modes of branch formation: bifurcation via splitting an existing branch bud, and lateral branching via de novo bud outgrowth from an existing branch stalk. (C) Types of abnormal branching: overgrown with cystic branches and hypoplastic with or without cystic branches.

Morphological features

Tissue level

Branching is directional in response to organ level patterning information. Such global patterning is exemplified by the growth of the accessary lobe in mice, which initiates from the right main bronchus and crosses the midline to reach the left lobe (Fig. 2A). A higher level of Fgf10 expression has been found to cap lobar branch buds including that of the accessary lobe and possibly contribute to their extensive outgrowth. The mechanisms establishing such Fgf10 expression are unknown and may be relevant to species specific lobation patterns^26–28.

Early branching events are nearly identical among inbred mouse strains²⁹. Each lung lobe has a characteristic shape made of edges and surfaces corresponding to and possibly driven by the three branching subroutines with planar and orthogonal bifurcation forming edges and surfaces respectively and domain branching determining their locations²⁹. At some level, branching must be influenced by physical constraints including branch self-avoidance and thoracic space filling. Computational modeling based on 3D images comparing normal and mutant lungs as well as other branched organs is likely to shed light on the branching process that has fascinated theoretical and experimental biologists since Aristotle^30–32.

Cell level

The cellular behavior underlying branching morphogenesis is largely unclear, although progress has been made in examining branching-associated proliferation, cell division orientation and cell movement. A detailed live imaging analysis of cultured lungs has followed the two modes of new branch formation: bifurcation, which splits existing branch buds, and lateral (or domain) branching, which forms buds de novo on the side of existing branch stalks (Fig. 2B)³³. Proliferation appears to occur via interkinetic nuclear migration and occurs at a higher rate in the branch buds than stalks³³, although the contribution of this differential proliferation to branching morphogenesis is still unclear^{34, 35}. Biases in cell division orientation match the changes in tissue morphology in an expected manner with cell division longitudinal and perpendicular to the branch axis associated with branch elongation and expansion respectively³³. Epithelial cells move as a group rather than individually and can change their positions between branch buds and stalks³³. These observations suggest that epithelial progenitors constitute a “branching zone” within which dynamic cell movement and proliferation are constrained by local mechanical forces and drive new branch formation^{33, 34}. Future live imaging experiments of lungs with the epithelium mosaically labeled for multiple cellular compartments, such as the nucleus and plasma membrane, and in combination with pharmacological and genetic manipulations are likely to further dissect the branching process on a single cell level. It will also be important to determine how such a local branching process is coordinated among branches to form the organ level branching pattern described above.

Molecular control mechanisms

Progenitor markers

A number of genes have been found to be preferentially expressed in the branching region of the lung epithelium^{36, 37}. Among these, Id2 and Sox9 are commonly used as markers of epithelial progenitors. Additional progenitor markers can be identified through targeted expression screen³⁸ and analysis of existing expression databases, such as Eurexpress containing in situ hybridization data of ~20,000 genes on E14.5 mouse embryo sections³⁹ (Fig. 3). A comprehensive list of progenitor markers will not only identify candidate regulators, but also allow for a molecular analysis of genetic mutants and eventually an assembly of a gene regulatory network of the progenitors.

Figure 3.

Marker discovery using Eurexpress.

Left panel: an example in situ hybridization image from Eurexpress³⁹ (Cldn7, euxassay_002232) showing that major organs can be recognized at E14.5. Right panels: our screen to identify markers for the progenitors (top: Sox9, euxassay_018446; Lin7a, euxassay_011082; Lama3, euxassay_011044) and conducting airways (bottom: Sox2, euxassay_019525; Mycl1, euxassay_019486; Acsl1 (Acyl-CoA synthetase long chain family member 1, euxassay_001151). The expression pattern is identified based on the presence (solid line) and absence (dashed line) of in situ signals in the branching (green) and non-branching (red) regions.

Genetic mutants

Genetic mutants have been identified with a wide range of branching defects (Fig. 2C). The induction of ectopic branches from the trachea epithelium by the lung mesenchyme in classic tissue graft experiments demonstrates the presence of instructive branching signals in the mesenchyme⁴⁰. Loss of Fgf10 from the mesenchyme or its receptor Fgfr2 in the epithelium leads to lung agenesis or no branching respectively, suggesting that the Fgf signaling is the major branching signal^22–24. Surprisingly, the precise expression pattern of Fgf10 appears not necessary to rescue some of the early branching events in the Fgf10 mutant⁴¹. Components of Wnt, Bmp, Shh and Tgf signalings are also present near the branching region and can modulate the Fgf signaling¹.

Less severe disruption of branching morphogenesis can lead to a hypoplastic lung with fewer but normal-looking branches, as exemplified in Wnt7b and Hhip mutants^{42, 43} (Fig. 2C). Counter-intuitively, another group of hypoplastic lung mutants have fewer and yet cystic branches, as exemplified in Dicer1, Sox9, HDAC1/2 mutants^44–46, whose phenotypes beyond early branching should be interpreted with caution as discussed later. Conversely, excessive Fgf branching signal leads to an overgrown lung with also cystic branches^{45, 47, 48}. No mutants have been found to have excessive but normal looking branches as seen among species with different number of branches (e.g. mice versus humans), possibly because maximal branching efficiency has been achieved for the available thoracic cavity in a given species. A better understanding of the underlying cell biology is necessary to categorize branching mutants beyond these morphological categories. Further molecular analysis with a comprehensive panel of progenitor markers and epistasis analysis of multiple pathways are likely to reveal the gene regulatory network of the progenitors.

Recent studies have also identified additional branching mutants with abnormal epithelial cell behaviors. Alterations to the Ras pathway change branch shape and are associated with a change in the relative frequency of random versus directional cell division⁴⁹. Loss of Integrin beta 1 (Itgb1) completely blocks branching and converts the monolayered lung epithelium to a multilayered epithelium associated with abnormal apical-basal cell division⁵⁰. Genes implicated in planar cell polarity and cytoskeleton organization have also been shown to be involved in branching^51–53. In some cases, the observed cell biology is shown to be causal to the morphological change by mathematical modeling⁴⁹, although such causality is difficult to establish experimentally without direct and possibly physical manipulation of cellular behaviors⁵⁴.

III. The airway differentiation program

Early in development, branching morphogenesis leaves behind tubes that become the conducting airways. Although there may be a subpopulation of progenitors with restricted differentiation potential⁵⁵, Id2^CreER56 and Sox9^CreER (our unpublished data) lineage tracing experiments show that most if not all conducting airway cells are descendants of Id2/Sox9 expressing progenitors. Therefore, early progenitors undergo airway differentiation, during which their descendants first express a pan-airway marker Sox2⁶ and subsequently become specialized conducting airway cells.

Morphological features

Tissue level

Branching morphogenesis only generates tubes, but the length and diameter of such tubes must be subsequently controlled to become the normal, mature airways. In addition, there must be some coordination among parent-daughter branches to achieve a nearly universal tapering pattern. Indeed, classic morphometric studies of a human lung cast reveal a log-linear relation between branch diameter and generation⁵⁷. Such an equation follows Murray’s Law, which is considered the most efficient network design⁵⁸. Little is known about how airway length, diameter and tapering are controlled, although oriented cell division and apical membrane biogenesis have been shown to regulate the diameters of renal tubules and Drosophila tracheal tubes respectively⁵⁹.

Cell level

During airway differentiation, cells must proliferate to fuel the continuous growth of airway tubes and meanwhile make choices leading to specialized cell fates. Progenitors at branch buds have a higher rate of proliferation than their descendants more proximally, suggesting separate control mechanisms or an attenuation of proliferation as progenitors are left behind. The mouse airways contain four major cell types with distinct function, morphology and distribution. Club (Clara) cells are the main secretory cells, are dome shaped and secret mucus to trap microbes and foreign particles^{60, 61}. Ciliated cells have apical motile cilia to move mucus out of the lung. To ensure functional coupling during mucociliary clearance, club and ciliated cells are intermingled, suggesting a lateral inhibition mechanism of cell fate decision as described later. In mice, basal cells are mostly limited to the pseudostratified epithelium of the trachea and main stem bronchi, a proximal-distal distribution mirroring that of airway cartilages. Neuroendocrine cells are either solitary or form clusters near branching points and are considered cells of origin for small cell lung cancer, although their physiological function is unclear^{62, 63}. Little is known about what controls the proximal-distal and junction-specific distributions of basal and neuroendocrine cells respectively.

In the adult lung, there is a cellular hierarchy within which ciliated cells are terminally differentiated, club cells are capable of self-renewal and regenerating ciliated cells and basal cells are capable of self-renewal and regenerating both ciliated and club cells, although club cells may dedifferentiate to basal cells under certain conditions^64–67. It is unclear if a similar cellular hierarchy exists during development or different cell types are specified directly from Sox2-expressing cells. Future clonal analysis using Sox2^CreER in whole-mount airways⁶⁸ is likely to shed light on the proliferation, differentiation and cellular hierarchy during airway differentiation.

Molecular control mechanisms

Cell type specific transcription factors

The SOX2 expression domain starts proximal to new branch formation, making SOX2 the earliest known marker of airway differentiation⁶. Sox2 is also required for the differentiation and maintenance of several airway cell types^{69, 70}. Cell type specific transcription factors have been identified for ciliated (Foxj1), basal (p63) and neuroendocrine (Ascl1) cells based on studies of analogous cell types in other organs and shown to function as positive regulators of their respective lineages^{7, 71–74}. In contrast, no cell type specific transcription factor is known for club cells, which are arguably unique to the lung. Additional cell type specific transcription factors must exist and can be identified through systematic expression screen as demonstrated by recent identification of Myb as a novel regulator of ciliated cell differentiation⁷⁵. Similar to branching morphogenesis, Eurexpress should be a useful source for gene discovery (Fig. 3).

Cell fate choice

One major advance in the understanding of airway differentiation is the discovery of Notch pathway as a control mechanism for club, ciliated and neuroendocrine cell fate choice^76–80. Although many details remain to be elucidated including club cell heterogeneity such as differential sensitivity to the cytotoxicant naphthalene⁸¹, genetic analyses of multiple Notch receptors and their ligands support a model where Notch signaling is used during two rounds of cell fate decision: neuroendocrine versus club cells and ciliated versus club cells. In both cases, club cells form as a result of a high level of Notch signaling in response to its ligand from cells adopting the alternative fates. Based on such a model, it is tempting to hypothesize that the missing club cell specific transcription factor may be a combination of NOTCH ICD (intracellular domain, high Notch activity) and a lung specific transcription factor. An intriguing candidate is Nkx2.1, which is shown to regulate the promoter of club cell specific protein (Scgb1a1 or Ccsp)⁸² and together with Sox2 may confer airway specificity.

Notch signaling is also important for the differentiation of goblet cells, another type of secretory cells that are present in humans but are very rare in mice under baseline conditions and associated with mucus hyper-production. In addition to such interspecies difference, it is worth noting that abnormal mucus production may be caused by reversible conversion of club cells to goblet cells in response to inflammation, de novo differentiation of club/goblet cells during development and homeostasis, or gastric transformation of lung cells^{60, 83–86}. Although further studies are needed to elucidate how Notch signaling is related to other goblet cell regulators such as Spdef^{87, 88} and the similarity of mouse bronchioles to human airways, pathways controlling goblet cell differentiation are attractive therapeutic targets of lung diseases with excessive mucus production including chronic obstructive pulmonary disease (COPD).

IV. The alveolar differentiation program

Late in development, the epithelial progenitors stop producing conducting airway descendants and start an alveolar differentiation program that leads to the formation of alveolar type 1 (AT1) and type 2 (AT2) cells, as demonstrated by Id2^CreER lineage tracing experiments⁵⁶.

Morphological features

Tissue level

The acinus is the basic gas exchange unit attached to the end of each terminal bronchiole, the last branch generation of the conducting airways. The acinus is organized to ensure directional airflow and consists of the alveolar duct, alveolar sac and alveolus along the proximal-distal axis. Due to the 3D complex nature of the late developing lung, the change in tissue morphology during alveolar differentiation is not well understood^{89, 90}. We have used OPT and confocal microscopy to follow morphogenesis and distribution of progenitors during late development and provided evidence that a branch bud-stalk structure with distally located SOX9 progenitors persist and continue to generate new branch buds at least up to E18.5 in mice, a time when alveolar differentiation has occurred along the alveolar duct (Fig. 2A and⁴⁵). Based on these developmental intermediates and other data, we propose the following model of acinus formation (Fig. 4): progenitors continue to branch at the distal end of the respiratory tree to generate tubes for the future alveolar ducts; their descendants left behind become AT1 and AT2 cells and AT1 cell flattening expands the air lumen; eventually, there is an insufficient number of progenitors to support further branching and the remaining progenitors are depleted via differentiation into AT1 and AT2 cells to form the primary alveolar sacs, which are further subdivided into mature alveoli via secondary septation.

Figure 4.

A model for acinus morphogenesis in the mouse lung.

Branching continues beyond the terminal bronchiole to generate tubes that will become the alveolar ducts (duct). Eventually, there is insufficient number of progenitors to support further branching and the remaining branch buds expand to form alveolar sacs (sac), which is subsequently divided into alveoli (a) via secondary septation. Air space expansion is accompanied and possibly driven by AT1 cell flattening. Primary and secondary septa may correspond to early and late forming alveolar walls (Fig. 2A). Side top panel: overlay of the left diagrams showing transformation of a tube to a clefted bubble. Note that cleft formation may involve inward growth and/or expansion of neighboring epithelia. Side bottom panel: the alveolar ducts are surrounded by alveolar sacs/alveoli possibly derived from short tubular branches perpendicular to the ducts and therefore may appear to have outpouchings on sections.

One prediction of this model is that the signaling pathways required for early branching that gives rise to the conducting airways also control late branching that gives rise to the acini. If true, it would argue that despite the functional and cellular differences between the conducting airway and gas exchange compartments in the mature lung, they are generated via a continuous branching process. Further examination of this model will require a detailed perinatal time course analysis of multiple epithelial, mesenchymal and vascular markers in the whole lung or acinus. Such 3D images of whole-mount tissues should also be useful to complement and confirm section-based stereological measurements of acinus morphology⁹¹. Ultimately, studies of acinus morphogenesis needs to elucidate how tubes become “clefted bubbles” on the tissue level and how clustered progenitors become juxtaposed AT1 and AT2 cells on the cellular level (Fig. 2A, 4 and below).

Cell level

The gas exchange surface is covered with two major cell types, AT1 and AT2 cells. AT1 cells cover >90% of the gas exchange surface and are extremely thin (~0.1 um) such that O₂ and CO₂ can passively diffuse across. Electron microscopy studies show that AT1 cells can be highly branched and span several alveoli, making accurate quantification of cell number and 3D morphology difficult if not impossible⁹². AT2 cells are cuboidal and secret surfactants that are vital by reducing surface tension to keep the alveoli open. A recent study shows that AT1 and AT2 cells come from bipotential progenitors that express markers of both cell types⁹³. Future lineage tracing experiments are necessary to determine if these bipotential progenitors are SOX9 cells undergoing alveolar differentiation at the very distal end of the epithelial tree (Fig. 2A) and when and how they are committed to AT1 versus AT2 cell fate.

After commitment, AT1 cells must flatten and fold to achieve their elaborate morphology⁹². Novel cell biology concepts may emerge by studying AT1 cell differentiation. Similarly, AT2 cells must form specialized organelles, lamellar bodies, to process and store surfactants, a demanding protein synthesis task susceptible to pathogenesis⁹⁴.

Molecular control mechanisms

Cell type markers

AT1 and AT2 cells are originally identified by their morphological and cellular features under electron microscopy, which remain the gold standard of cell type definition. Recent time course transcriptome analyses of whole lung or isolated distal epithelium have identified a number of genes upregulated during alveolar differentiation, which include many known alveolar cell markers and therefore should be useful to identify novel alveolar specific genes or regulators^{45, 95, 96}. In particular, unlike Sox2 for conducting airway differentiation, no transcription factors have been identified that are specific for alveolar cells in general or AT1 and AT2 cells specifically, although an atypical homeobox protein HOPX appears specific to mature AT1 cells⁹⁷.

Besides finding new markers, it is important to more thoroughly characterize existing markers during development. For example, Sftpc is highly specific for AT2 cells in the mature lung, but its promoter is active as early as lung specification⁵⁵. A recent study has systematically examined existing alveolar markers and grouped them based on the onset and restriction of their expression to AT1 and AT2 cells⁹³. Similar analysis extended to additional markers and postnatal periods, for example during secondary septation, should be useful in defining a molecular path of alveolar differentiation.

Genetic mutants

Two classes of genetic mutants have been reported with alveolar differentiation defects. In the first group, lung development proceeds normally up to E16.5 but subsequent alveolar differentiation is blocked morphologically and molecularly, often in association with persistent progenitors^98–104. Future experiments need to determine whether these mutants disrupt a single or multiple signaling pathways and the contribution of the mesenchyme to epithelial differentiation^105–109.

A second group includes many mutants with defective branching in early development, which have also been examined for alveolar differentiation in late development. Caution should be taken because the observed alveolar differentiation phenotype may be secondary to a smaller and/or defective progenitor pool resulting from abnormal early branching (Fig. 5D). Future experiments using temporally controlled genetic drivers, such as Id2^CreER and Sox9^CreER, are likely to establish a direct and cell autonomous function of these genes in alveolar differentiation. In addition, cystic branching defects may appear as congenital cystic adenomatoid malformation, bronchopulmonary dysplasia or emphysema, which all have abnormally dilated airspace (Fig. 5D)^{110, 111}.

Figure 5.

Balanced progenitor model in the mouse lung.

(A) Lung epithelial progenitors (SOX9, green nucleus) balance among three developmental programs: branching (green arrow), which expands progenitors and builds a tree-like structure; airway differentiation (red arrow), which activates SOX2 (red nucleus) and then forms specialized airway cells (for example, club and ciliated cells); and alveolar differentiation (blue arrow), which forms AT1 and AT2 cells and the former needs to fold extensively. (B) This balance shifts over time: early in development, progenitors branch/expand and undergo airway differentiation; late in development, progenitors continue to branch/expand and undergo alveolar differentiation; in perinatal periods, progenitors stop branching, are depleted (dashed green arrow) via alveolar differentiation (thicker blue arrow) and transition from development to homeostasis. (C) Regulated deployment of the three developmental programs following one branch. The left side is color-coded for the progenitors (green line) and their early (red line, conducting airways) versus late (blue line, gas exchange region; wavy line, deformation of tubular structures by air space expansion, see also Fig. 2A and 4) descendants, and the developmental programs (arrows, see (B) for details). The right side shows the morphology and is labeled for the three generations of conducting airways. (D) Three possible outcomes of a cystic branch resulting from an early branching defect. The left side is colored coded as in (C). Depending on whether a cystic branch bud becomes a normal-looking branch stalk (corrected) and whether it undergoes alveolar or airway differentiation, the location of the cyst and ratio between the gas exchange and airway compartments differ. Therefore caution should be taken in interpreting phenotypes beyond early branching.

V. Balanced progenitor model

We propose that branching morphogenesis, airway differentiation and alveolar differentiation do not exist in isolation, but instead are three alternative developmental programs that lung epithelial progenitors need to choose from (Fig. 5A). The progenitors balance among the three programs and such a balance shift over developmental time (Fig. 5B). This balanced progenitor model intuitively explains the separation of proximal conducting airway and peripheral gas exchange region in the respiratory tree, and better defines proximal-distal patterning, a term commonly used in describing the spatial difference in the lung (Fig. 5C). This model attempts to provide a conceptual framework of lung development and makes testable predictions.

Branching versus airway differentiation — progenitors and their early descendants

Early in development, progenitors expand via branching morphogenesis and leave behind descendants that become the conducting airways. In this context, proximal-distal patterning refers to non-branching Sox2-expressing airways and branching Id2/Sox9-expressing progenitors. The progenitors may either stay Sox9 positive and continue to branch, or exit the branching region to activate Sox2 and undergo airway differentiation. It appears plausible that multiple signaling pathways and morphogen gradients in the branching regions will maintain Sox9 expression and inhibit Sox2 expression. The key pathway(s) controlling such a Sox9 to Sox2 switch need to be tested in a mosaic experiment where isolated pathway mutant cells in an otherwise normal branch bud lose Sox9 expression and ectopically express Sox2.

Branching versus alveolar differentiation — progenitors and their late descendants

Late in development, progenitors continue to branch and leave behind descendants that become AT1 and AT2 cells of the gas exchange region. By analogy to branching in early development, proximal-distal patterning in this context refers to differentiating alveolar cells and Sox9 expressing branching progenitors. Different from early branching, the progenitor pool size decreases via alveolar differentiation in late development to a level insufficient to support further branching and finally progenitors are depleted and the lung transitions from development to homeostasis.

This balance between progenitor branching/self-renewal and alveolar differentiation must be precisely controlled to ensure sufficient but not excessive number of cells for gas exchange and should be taken into consideration in treatment of lung immaturity in preterm birth. For example, antenatal glucocorticoids are often used to promote lung maturation in women at risk of preterm birth, but their potentially adverse effects on accelerated progenitor depletion need to be studied experimentally and monitored clinically. In addition, other life supporting measures including O₂ supplement and mechanical ventilation may impair progenitor self-renewal and differentiation and contribute to bronchopulmonary dysplasia, a frequent long-term complication in preterm birth. Mechanisms of lung progenitor depletion may be general to how development ends and how organ sizes are controlled and coordinated within an organism. In addition, reversal of the depletion process may be relevant to regeneration and tumorigenesis.

The balance between branching and alveolar differentiation appears to occur even during early development. Recent studies show that Sox9 deficient progenitors branch ineffectively and also precociously express some, but not all, alveolar markers, suggesting that branching promoting genes may have a second function in suppressing precocious alveolar differentiation^{45, 112}. Additional such dual function genes may be identified by examining alveolar differentiation markers during early development of known branching mutants.

Furthermore, such dual function genes may be important during lung evolution (Fig. 6). The Xenopus lung does not branch, initiates alveolar differentiation immediately after lung specification and forms a primitive gas exchange organ with two sacs. Sox9 is present in the Xenopus genome but not expressed in the lung, raising the intriguing possibility that the branching program has evolved to meet an increasing demand for oxygen and thus lung complexity by temporarily suppressing alveolar differentiation until late development. Indeed, some AT1 and AT2 cell markers, such as Pdpn and Sftpc^{113, 114}, are expressed at a low level early in development, possibly reflecting incomplete suppression of an ancestral program that initiates alveolar differentiation right after lung specification (Fig. 6B).

Figure 6.

Branching versus alveolar differentiation in evolution.

(A) Top: OPT images of E11.5 control and epithelial Sox9 mutant mouse lungs and a stage 42 Xenopus embryo immunostained for SOX9. Bottom: Surfactant protein B (Sftpb) whole mount in situ hybridization of E10.5 control and epithelial Sox9 mutant mouse lungs and a stage 42 Xenopus embryo. When Sox9 is absence in the lung buds (dashed line) in the mouse mutant and Xenopus, Sftpb is expressed. Modified with permission from⁴⁵. (B) The branching program may be an evolutionary addition that temporarily delays alveolar differentiation to allow an increase in lung complexity by expanding the progenitors and building the conducting airways. Genes promoting branching (for example Sox9) have a second function to suppress (blunt arrow) the ancestral program that starts alveolar differentiation immediately after lung specification (gray dashed line in Xenopus).

Airway versus alveolar differentiation — a switch in differentiation choice/potential

Progenitors undergo airway differentiation early in development, but alveolar differentiation late in development. Therefore the proximal-distal patterning in the mature lung (i.e. the conducting airways versus gas exchange region) may be in essence a temporal switch in the differentiation choice/potential of progenitors. Although the exact timing of this switch is unclear⁵⁶, it must precede birth in such a manner that ensures a sufficient functional gas exchange compartment at birth. Mechanisms controlling the switch can be either intrinsic, via alteration and/or restriction of progenitor differentiation potential; or extrinsic, via differential response of the same progenitors to a temporal change in environmental factors; or a combination of both. According to our model, gain or loss of the signal(s) promoting the switch (i.e. in favor of alveolar differentiation) should precociously activate alveolar differentiation early in development or prolong airway differentiation late in development, respectively. Such a temporal switch establishing the airway and gas exchange compartments appears independent of branching and may involve pathways distinct from those extensively studied during early branching morphogenesis.

Regulation of Sox2 expression

The three developmental programs partially converge on the transcriptional regulation of the conducting airway marker Sox2. A putative Sox2 promoter should contain a positive element (binding site) that activates Sox2 as progenitors initiate airway differentiation, and negative elements that suppress Sox2 either during branching or alveolar differentiation (Fig. 7A). Figure 7A shows the predicted Sox2 expression when one of the regulatory elements is disrupted. Our preliminary data show that at least the putative positive element is at a distant site from the Sox2 coding region (Fig. 7B).

Figure 7.

Sox2 promoter analysis.

(A) Top: a putative Sox2 promoter contains a positive regulatory element (binding site) for the airway differentiation program and two negative regulatory elements for branching and alveolar differentiation. For simplicity, an alternative scenario is not illustrated but does not affect the discussion: each developmental program may not have its own element, but instead interfere with alternative elements. For example, the branching program may compete with and block the positive element for airway differentiation, therefore does not require its own regulatory element. Bottom: predicted Sox2 expression patterns in early and late development when each element is individually disrupted. See Figure 5C for color scheme. Without the airway differentiation element, Sox2 expression is lost and Sox2-independent mechanisms may prevent expansion of the alternative programs (dashed line). Without the branching element, Sox2 expression expands into the branching region and such expansion inhibits branching⁶ and likely interferes with subsequent alveolar differentiation. Without the alveolar differentiation element, Sox2 expression is normal in early development but continues to expand in late development such that all progenitors’ descendants express Sox2. (B) Stereomicroscope images of an E14.5 Sox2-BAC (bacterial artificial chromosome, RP23-4D1) transgenic embryo expressing a membrane Tomato fluorescent protein and its littermate control (Chen J. and Krasnow M.A.). The promoter is active in the neural tube but not the lung, suggesting that the airway differentiation element (A) is at a distant site from the coding region.

Comparison with the five stages of lung development

Lung development is historically divided into five stages characterized by distinct morphological features occurring in subregions of the lung¹¹⁵. Our balanced progenitor model is based on 3D images of whole lungs, emphasizing the continuity of the branching program and its temporal and spatial coordination with airway and alveolar differentiation. Several experimental observations not easily explained by the five-stage model are the predicted outcome of our model: new branches are still generated when the gas exchange region undergoes differentiation during the canalicular and saccular stages; and part of the future gas exchange region is already present during the pseudoglandular stage^116–120. In addition, our model integrates morphogenesis, which builds the respiratory tree, and differentiation, which generates the two distinct compartments; views each developmental program not in isolation but rather as alternative balanced choices of progenitors; and suggests new gene function(s) in controlling such a balance.

Such a model may be extended to the development of non-epithelial cells. The epithelial tree co-aligns with the vasculature and neurons and also coordinates with mesenchymal cell differentiation. The branching program may communicate with or receive input from signals regulating the development of vascular, neuronal and contractile mesenchymal cells. In addition, the concept of a temporal switch of differentiation programs may be applicable to other cell lineages. For example, Fgf10-expressing mesenchymal progenitors differentiate into airway smooth muscle early in development and lipofibroblasts late in development¹²¹. Similar lineage tracing experiments are necessary to determine if endothelial progenitors differentiate into arteries and veins early in development and capillaries late in development.

Conclusion

This review focuses on three developmental programs of lung epithelial progenitors: branching morphogenesis, airway differentiation and alveolar differentiation, and also summarizes our current understanding in a balanced progenitor model where the progenitors balance the three programs in response to spatiotemporal cues. Such a model emphasizes not just the initiation and continuation of a process, but also its termination and transition to an alternative process. This review also highlights the importance of developing advanced imaging methods and gene targeting tools with temporal and cellular precision to tackle the 3D complexities of the lung.