Microfluidic chest cavities reveal that transmural pressure controls the rate of lung development

Branching morphogenesis is a common developmental program used to create complex ramified networks of epithelial tubes that support the flow of liquid or air (Warburton et al., 2010). During development of the mammalian lung, dynamic and recursive branching leads to a stereotyped airway epithelial architecture that is crucial for survival after birth (Warburton et al., 2010). Several clinical conditions result in fetal pulmonary hypoplasia, an underbranched lung, which is a major cause of respiratory insufficiency and mortality in newborns (Jobe and Ikegami, 2000; Smith et al., 2005). Pulmonary hypoplasia often co-presents with mechanical defects in the thoracic cavity, suggesting that abnormal airway branching can result from altered physical loads on the lung, even in the absence of an obvious genetic defect (Smith et al., 2005).

Much of our current understanding of the mechanisms that regulate branching morphogenesis is related to signaling pathways elucidated via genetic manipulation in small animal models (Herriges and Morrisey, 2014). Unfortunately, the small size of developing mouse lungs has rendered quantitative physical experiments intractable and, consequently, the interaction of physical mechanisms with genetic programs has not been defined. Clinical observations and studies in vivo of embryonic lung development in large animal models suggest a crucial role for transmural pressure, the difference between the pressure within the airway lumen and that of the pleural cavity. Experimentally decreasing transmural pressure (Fewell et al., 1983) or reducing fluid movement (Jesudason et al., 2005; Miller et al., 1993) results in underbranched airways. However, the inability to simultaneously alter the mechanical loads on the lung and observe the resulting branching dynamics has represented a critical barrier to our understanding of the mechanisms underlying relatively common fetal disorders (Warburton et al., 2010).

Here, we present a microfluidic approach to precisely control the mechanical environment of the fetal lung by dynamically and quantitatively controlling the transmural pressure across intact murine embryonic lung explants (Fig. 1A). Our ‘microfluidic chest cavities’ consist of two separate chambers connected by a fabricated glass microneedle (Fig. 1B). Lungs dissected on embryonic day (E) 12 were intubated through the trachea with the glass microneedle and secured using a modified finger-trap suture (Fig. 1C), and then the chambers were sealed between two glass coverslips. The pleural chamber (with fluid pressure P_pleural) enclosed the explant, while the lumenal chamber (with fluid pressure P_lumen) accessed the airways via the intubated trachea. The pressure difference between the two chambers defined the transmural pressure (ΔP=P_lumen–P_pleural) across the developing lung, morphogenesis of which was followed via time-lapse imaging for up to 78 h. In contrast to the flattened architecture observed in standard explants (Fig. S1A,B), lungs cultured within the microfluidic chest cavities developed a 3D branched architecture with distinct lobation patterns (Fig. 1D) – one lobe on the left and four lobes on the right – similar to the intact lung in utero (Fig. S1C-E). To characterize the resulting airway morphology, we used confocal sections of immunostained explants to create 3D reconstructions of the branched epithelium (Fig. 1E; Movie 1), followed by a skeletonization algorithm to quantify morphological parameters and the branching pattern of the airways.

As the lung develops, the airway epithelium secretes fluid into the lumen that generates a positive pressure against the closed fetal larynx (Harding and Hooper, 1996), thus producing a static or ‘tonic’ ΔP of ∼200-400 Pa (1.5-3 mm Hg) (Olver et al., 2004; Schittny et al., 2000). Surgically occluding the trachea, which is thought to increase ΔP, has been shown to enhance branching in murine lung explants (Blewett et al., 1996; Unbekandt et al., 2008). To quantify precisely the effects of ΔP on airway architecture, we cultured explants for 48 h in the microfluidic chest cavities over a range of ΔP spanning those observed during normal (Harding and Hooper, 1996) and pathologic (Kitterman, 1996) development in vivo, and under standard culture ex vivo (Fig. 2A). We found that lung development occurred in a pressure-dependent manner, with higher ΔP leading to increased growth of the epithelium and mesenchyme, as well as commensurate increases in the overall size of the organ and architectural complexity of the airways. The number of terminal branches increased as ΔP increased (Fig. 2B), consistent with qualitative observations from tracheal occlusion studies.

These pressure-mediated changes in airway architecture could arise from two different mechanisms. Transmural pressure could increase the number of branches by stochastically altering the topology of the epithelial tree (i.e. changing the pattern of branching), or it could increase the rate at which a conserved airway topology elaborates over time. To distinguish between these two possibilities, we quantified the topology of the airway epithelial tree after 48 h at each ΔP by constructing lineage diagrams from our skeletonized 3D reconstructions, using anatomical nomenclature reported previously (Metzger et al., 2008) (Fig. 2C; Fig. S2). This analysis revealed that the branching pattern of the airways remained remarkably stereotyped, consistent with that of lungs isolated from embryos later in gestation. Lungs cultured at lower ΔP were less mature overall and generated a subset of the branches that formed in explants at higher pressures. These data show that ΔP controls the rate, but not the pattern, of airway epithelial branching.

Gene expression profiling supported this conclusion. We conducted RNA sequencing (RNA-Seq) analysis on lungs held for 48 h at low (20 Pa) or high (300 Pa) ΔP, as well as those isolated directly from embryos on E12.5, E13.5 and E14.5. We then analyzed genes that were differentially expressed (P<0.05) between E14.5 and E12.5 (2095 genes), and between high and low ΔP (2237 genes), using the NextBio platform (Kupershmidt et al., 2010), and found significant overlap between the two (P=5.5×10^–72; Fig. 2D); of these overlapping genes, the majority changed in the same direction in both data sets (P=3.3×10^–90 increased, P=1.1×10^–78 decreased; Fig. 2E). However, because many genes did not overlap, we performed a principal component analysis (PCA) to better understand these differences (Fig. S3A). Not surprisingly, PCA revealed that key differences in gene expression resulted from comparing lungs held in culture to those isolated directly from the embryo (Fig. S3B,C). Nonetheless, PCA of both the full data set and the 500 most-variable genes showed that the high ΔP and E14.5 lungs clustered separately from the remaining samples (Fig. 2F; Fig. S3D-F), consistent with our topology analysis.

These data suggest that ΔP drives expression of lung developmental gene programs. Gene Ontology analysis showed regulation of transcripts associated with morphogenesis, development, cell motility and migration, and extracellular matrix in both groups (Fig. S3G). RT-PCR analysis confirmed that high ΔP enhances the expression of genes that drive airway branching morphogenesis, such as fibroblast growth factors (FGFs) (Fig. S4A-D). High ΔP also enhances the expression of genes involved with airway maturation, including surfactant proteins and mucin (Fig. S4E-J). Collectively, our data suggest that lungs held at higher ΔP are more developmentally mature than those held at lower pressures.

To characterize how ΔP affects the kinematics of morphogenesis, we used time-lapse imaging of lungs cultured within the microfluidic chest cavities to quantify branching dynamics. Tracing the projected area of the epithelium over time (Fig. 3A) allowed us to annotate lineage diagrams with the time at which branches formed (Fig. 3B). These temporal lineage studies surprisingly revealed that branching occurred synchronously approximately every 10 h for lungs held at a ΔP of 300 Pa (Fig. 3C); that is, branches formed simultaneously (within ±10 min of each other) at distant locations throughout the epithelial tree (Fig. 3D). The timing between these synchronized bursts of branching was pressure dependent (Fig. 3E), with higher ΔP decreasing the interval (Δt) between bursts.

This demonstration of periodic synchronized branching emphasizes the fact that morphogenesis of the airway epithelium is regulated globally within the lung. As the lung develops, smooth muscle encircles the airways and helps guide branch sites (Kim et al., 2015) (Fig. S5A,B). In cultured explants, nonuniform peristalsis-like contractions of the proximal airway smooth muscle have been observed to move fluid throughout the lumen, causing the tips of distal branches to cyclically dilate and relax (Schittny et al., 2000). To visualize the interaction of the airway epithelium and surrounding smooth muscle, we explanted lungs from transgenic mouse embryos expressing red fluorescent protein under the control of the α-smooth muscle actin promoter (αSMA-RFP) (Magness et al., 2004) (Fig. S5C). In addition to periodic high-frequency (∼1 min⁻¹) peristalsis-like contractions (Fig. S5D), we also observed longer timescale systemic smooth muscle contractions, which caused a net flux of fluid into the extending branches (Fig. 3F; Movie 2), and gradually decreased the diameter of the proximal airways throughout the epithelial tree (Fig. 3G). These systemic contractions (dashed red lines in Fig. 3B) began 6-12 min before new branches appeared, continued as the branches extended, and were regulated by ΔP (Fig. 3E). Pharmacologically inhibiting smooth muscle contraction with nifedipine (Movie 3) (Kim et al., 2015), an L-type calcium channel blocker, disrupted systemic contractions and arrested branch extension (Fig. 3H). These results suggest that ΔP regulates both the timing of airway branching and of systemic smooth muscle contractions, which may be coupled globally across the lung. Although systemic contractions occurred periodically, and branches formed with each contraction, not every generation branched with each contraction.

In congenital diaphragmatic hernia, the herniation of abdominal organs is often confined to one side of the chest cavity, resulting in airway hypoplasia localized to that side. These observations suggest that physical boundary conditions might play a role in locally defining ΔP. To distinguish between effects caused by lumenal pressure and those caused by spatial constraints, we held P_lumen constant while constraining one lobe of the lung under hydrogel (Fig. 4A), decreasing ΔP. As a comparison, we ligated the primary bronchi of lungs cultured within our microfluidic chest cavities to asymmetrically increase ΔP in the occluded bronchus (Fig. 4B). In both cases, physically constraining ΔP on one side of the developing lung resulted in different extents of branching within the same explant (Fig. 4C). At later stages of development, the individual lobes of the lung interlock (Fig. S1C) and the organ conforms to the contours of the thoracic cavity, despite the absence of structural connections between the lung and the chest wall. In the relatively spacious microfluidic chest cavity, the explants developed the stereotyped lobation pattern, but the gross geometry of the lobes was rounded and noninterlocking. To determine whether ΔP and physical boundary conditions might regulate the global architecture of the lung, we cultured explants in microfluidic chest cavities of defined shapes. The final overall geometry of these explants conformed to the shape of the microfluidic chamber, generating cuboidal (Fig. 4D), cylindrical, triangular or rectangular-prism-shaped organs (Fig. S6), with branches (Fig. 4E) that formed interlocking lobes. Branching morphogenesis of each lobe halted once the daughter branches contacted the walls of the microfluidic chest cavity (Fig. 4F). For cuboidal chest cavities, lungs formed 19±1 terminal branches before morphogenesis stopped (Fig. 4G), consistent with a space-filling program.

These data reveal that development of the airway epithelium and its surrounding mesenchyme are controlled by the relative pressure of the fluid contained within the lumen of the lung (Fig. 4H). Although other studies have identified key molecular and genetic mediators of airway stereotypy (Metzger et al., 2008; Morrisey and Hogan, 2010), our data suggest that airway branching morphogenesis is not driven solely by a closed genetic program but is also controlled by mechanical cues, consistent with an increasing appreciation for mechanical forces in the morphogenesis of a wide variety of tissues (Desprat et al., 2008; Hutson et al., 2003; Martin et al., 2009; Nelson and Gleghorn, 2012; Nerurkar et al., 2006; Varner et al., 2010). We have shown that ΔP regulates the rate and timing of epithelial branching, and that branching events are synchronized throughout the developing airway tree. Importantly, ΔP regulates both the timing of branching morphogenesis and the frequency of systemic contractions of airway smooth muscle. These results reveal the existence of a pressure-dependent mechanical ‘clock’ that coordinates morphogenesis at distant sites within the developing organ to control and synchronize branching events within the airway epithelium. Thus, the functional behavior of one developing tissue (the contraction of smooth muscle) is connected to the morphogenesis of an adjacent tissue (the branching of the epithelium). It will be interesting to determine the molecular and physical connections between systemic smooth muscle contraction and synchronized epithelial branching in the future.

The function of airway smooth muscle has been controversial (Fredberg, 2007; Mead, 2007; Mitzner, 2004) and most often investigated in the context of postnatal diseases such as asthma (Janssen and Killian, 2006). Our findings suggest the possibility of a specific and unappreciated instructional role for airway smooth muscle during early lung development. The timing of systemic contractions of airway smooth muscle and formation of epithelial branches are linked, and both are regulated by ΔP. Although the underlying molecular mechanism is unclear, mechanical stretch has been found to regulate the contractile phenotype of vascular smooth muscle cells (Stegemann et al., 2005), and similar signaling could be at work in the developing lung. Indeed, overexpressing Cftr was found to both increase luminal fluid secretion and enhance the expression of smooth muscle-specific proteins in the embryonic rat lung (Cohen and Larson, 2006). Our data suggest that increasing the frequency of smooth muscle contraction might normalize airway development in fetuses suffering from conditions that usually result in airway hypoplasia.

Organ development is currently considered to be a genetically driven process, in part because the experimental tools used to investigate morphogenesis have been genetic in nature. Our study shows that mechanical tools, such as those presented here, can reveal the nongenetic physical mechanisms that also regulate development. Here, we have shown that a microenvironmental factor (fluid pressure) can quantitatively regulate the rate of morphogenesis. It is possible that pressure-mediated control of morphogenesis is a conserved regulatory mechanism across different organs and species, as other studies have suggested that fluid pressure might drive expansion of the chicken neural tube (Desmond and Jacobson, 1977), Drosophila trachea (Tsarouhas et al., 2007) and zebrafish Kupffer's vesicle (Navis et al., 2013). The ability to regulate the mechanical microenvironment of developing organs offers a powerful vertebrate model to study morphogenesis, mimic common fetal diseases, and parse the interactions between molecular and physical mechanisms of development.

Embryos were isolated from euthanized timed-pregnant mice (Mus musculus, CD-1) at E12.5 and stored in chilled phosphate-buffered saline (PBS) prior to dissection. Transgenic embryos expressing αSMA-RFP (Kim et al., 2015) were similarly isolated and stored. All experiments complied with ethical regulations for the care and use of animals, as approved by the Princeton University Institutional Animal Care and Use Committee. Sample size was determined by estimating the variance in preliminary experiments, with a power analysis for a power of 0.8. Lungs were dissected following standard protocols in chilled PBS supplemented with 10 U/ml penicillin and 10 µg/ml streptomycin (Sigma-Aldrich), taking care to extract the entire length of the trachea. Explants were either cultured in the microfluidic chest cavities or at the air-fluid interface on floating rafts (Schittny et al., 2000) in DMEM:F12 medium (HyClone) supplemented with 5% fetal bovine serum (Atlanta Biologicals), 10 U/ml penicillin and 10 µg/ml streptomycin. To pharmacologically inhibit airway smooth muscle contraction, nifedipine (10 µM; Sigma-Aldrich) was added to the culture medium for the duration of the experiment. For time-lapse analysis, explants were cultured within the microfluidic chest cavities at 37°C and 5% CO₂ in a humidified stage-top incubator. Brightfield and/or fluorescence images were acquired every 6 min for the duration of the culture period.

Microfluidic devices consisted of a layer of micromolded polydimethylsiloxane (PDMS) bonded between two glass coverslips for structural support and optical clarity. Standard photolithographic techniques were used to generate a 100 µm thick SU-8 silicon master consisting of the two chambers and fluidic channels. PDMS was replica molded from these masters and a circular biopsy punch (8 mm diameter) was used to create the chambers in the 4 mm thick PDMS. A second biopsy punch (1 mm diameter) was used to connect the two chambers. A pulled glass catheter was inserted through the hole between the two chambers and the catheter was trimmed to length. A glass coverslip and the PDMS/catheter assembly were bonded together following treatment in a plasma cleaner. The fluidic channels, chambers and glass catheter were wetted with culture medium and bubbles were cleared from the system. A lung explant was inserted into the pleural chamber and intubated with the catheter, and the trachea was then affixed to the catheter with a modified finger-trap suture. A second plasma-cleaned glass coverslip was bonded to the roof of the chambers. The fluidic channels enabled dynamic control of the pressure within the two chambers, thus permitting application of either a constant or variable ΔP. Pressure was controlled hydrostatically by altering the height of the column of fluid above each chamber. Every 18 h, two thirds of the culture medium was replaced with fresh medium.

To alter the shape of the chest cavity, a square, circular, triangular or rectangular hole was punched into a 200 µm thick sheet of PDMS. The punched sheet was inserted into the pleural chamber of the microfluidic chest cavities prior to insertion and intubation of the lung explants, which were cultured at ΔP=200 Pa for 48 h.

Explants were fixed in 4% paraformaldehyde in PBS for 15 min at room temperature. To visualize the airway epithelium, samples were labeled using anti-E-cadherin monoclonal antibody (Invitrogen, 1:200) and Alexa Fluor 488-conjugated secondary antibody (Invitrogen, 1:200). High-resolution confocal stacks (z=2 µm) were acquired of the immunostained explants, and these data were reconstructed to create 3D volume models of the epithelial tree using Amira (FEI Visualization Sciences Group). Briefly, the 3D image data were segmented and fit to an isosurface to define a 3D object. A skeletonization algorithm was then applied to the model of the airways to define the centerlines and connection points of all branches. We used this framework to quantify and compare the branching pattern, hierarchy and aspect ratio of the airways. To compare the effects of different ΔP, we quantified the branching hierarchy of at least four explants per pressure condition per experiment; experiments were conducted independently using embryos from at least three different mothers to permit conclusions about stereotypy.

For time-lapse experiments, morphometric analysis was conducted using 2D maximum intensity projections of brightfield and fluorescence images. To quantify branch contraction, we measured the diameter of the airway halfway along the length of the proximal branches as a function of time. Measurements of ≥20 branching events from seven different explants were used to construct lineage diagrams (Fig. 3B-E). Because the pattern of branching in the mouse lung is stereotyped, the locations of daughter branches are known in advance; the time at which they emerged was determined by measuring when the epithelium at the site of a new branch deformed by ≥15 µm (∼20% of the mean luminal diameter) relative to the surrounding airway.

Fluorescent polystyrene microspheres (0.5 µm diameter, Invitrogen) were microinjected into the lumens of five freshly isolated lung explants. Explants were cultured in the microfluidic chest cavities and imaged over time to track the movements of the beads within the luminal fluid. Particle tracking was performed using ImageJ (https://imagej.nih.gov/ij/).

Gene expression analysis was carried out on lungs at stages E12.5, E13.5 and E14.5, as well as lung explants cultured at either 20 Pa (low ΔP) or 300 Pa (high ΔP) for 48 h. Tissue from several lungs was homogenized using QIAshredder (Qiagen) columns. RNA from the resulting sample was purified using the RNEasy Fibrous Tissue Mini Kit (Qiagen) and cDNA was synthesized using the Verso cDNA synthesis kit (Thermo Fisher Scientific). For RNA-Seq, the quality of total RNA was assessed by RNA 6000 Nanochip using an Agilent Bioanalyzer. The cDNA libraries were constructed using polyA-enriched RNA according to the manufacturer's instructions (Illumina). Sequencing was performed at the Lewis Sigler Institute for Integrative Genomics sequencing core facility using the Illumina HiSeq 2000 Platform. Reads were mapped to transcripts using TopHat (Trapnell et al., 2009) and gene counts were identified by HTSeq-count (Anders et al., 2015). Differentially expressed genes were identified using DEseq2 (Love et al., 2014), which employs a statistical model to handle low-expressing genes by shrinking their estimated log-fold changes to avoid false-positive P values. PCA was completed on these statistically normalized data using Python. Instead of choosing a cutoff for magnitude of relative expression, genes were considered to be differentially expressed if their P value, adjusted for multiple testing by the Benjamini–Hochberg procedure, was <0.05. A total of 2113 genes were selected as differentially expressed between E14.5 and E12.5 lung samples, and 2259 genes were selected as differentially expressed between explants held at high and low ΔP; these matched 2095 and 2337 features, respectively, in the Nextbio database (Kupershmidt et al., 2010), which was used for gene set overlap and Gene Ontology analysis as described previously (Cichon et al., 2015). The Nextbio server uses the 2012-10-10 data version of Gene Ontology annotations.

For quantitative PCR analysis, cDNA was synthesized using the Verso cDNA synthesis kit (Thermo Fisher Scientific). Quantitative RT-PCR was performed on a StepOnePlus real-time PCR system (Applied Biosystems) using iTaq SYBR green supermix (BioRad) and the primers indicated in Table S1. Primer pairs were designed using Primer-BLAST (NCBI) software (Ye et al., 2012). Expression levels for each transcript were normalized to that of 18S rRNA in each sample.

We thank the Lewis-Sigler Institute for Integrative Genomics Sequencing Core Facility for RNA sequencing and Lance Parsons for assistance with data analysis.