Biomaterial-based 3D human lung models replicate pathological characteristics of early pulmonary fibrosis

Abstract

Idiopathic pulmonary fibrosis (IPF) is a progressive and incurable lung disease characterized by tissue scarring that disrupts gas exchange. Epithelial cell dysfunction, fibroblast activation, and excessive extracellular matrix deposition drive this pathology that ultimately leads to respiratory failure. Mechanistic studies have shown that repeated injury to alveolar epithelial cells initiates an aberrant wound-healing response by surrounding fibroblasts through secretion of mediators like transforming growth factor beta (TGF- β), yet the precise biological pathways contributing to disease progression are not fully understood. To better study these interactions there is a critical need for lung models that replicate the cellular heterogeneity, geometry, and biomechanics of the distal lung microenvironment. In this study, induced pluripotent stem cell-derived alveolar epithelial type II (iATII) cells and human pulmonary fibroblasts were arranged to replicate key features of human lung micro-architecture and embedded in soft or stiff poly(ethylene glycol) norbornene (PEG-NB) hydrogels that recapitulated the mechanical properties of healthy and fibrotic lung tissue, respectively. The co-cultured cells were then exposed to pro-fibrotic cytokines and growth factors. iATIIs and fibroblasts exhibited differentiation pathways and gene expression patterns consistent with trends observed during IPF progression in vivo. A design of experiments statistical analysis identified stiff hydrogels combined with pro-fibrotic biochemical cue exposure as the most effective condition tested in this study for modeling fibrosis in vitro. Finally, treatment with Nintedanib, one of only two Food and Drug Administration (FDA)-approved drugs for IPF, was assessed. Treatment reduced fibroblast activation, as indicated by downregulation of key activation genes, and upregulated several epithelial genes involved in alveolar repair. These findings demonstrate that human 3D co-culture models hold are a promising tool for advancing our understanding of IPF and identifying new therapeutic targets.

Graphical Abstract

Introduction

Idiopathic pulmonary fibrosis (IPF) is an incurable respiratory disease that results in lung tissue scarring and progressive respiratory failure [1]. Hallmarks of the disease include epithelial cell dysfunction, fibroblast activation, excessive extracellular matrix (ECM) deposition, and thus disrupted gas exchange [2–5]. The ECM provides structural support to the lungs and undergoes continuous remodeling, creating a dynamic milieu rich in biophysical and biochemical cues. Comprised of over 150 different types of proteins, enzymes, growth factors, and proteoglycans [5], the ECM plays a crucial role in lung homeostasis and disease pathogenesis. Due to the complexity of the lung extracellular microenvironment, fully replicating in vivo conditions remains a significant challenge. Therefore, many researchers use reductionist models to investigate specific cellular and molecular interactions within a more controlled setting [6–10]. In healthy lung tissue, stiffness typically ranges between 1–5 kPa, whereas fibrotic lung tissue stiffness exceeds 10 kPa [10, 11]. Strong evidence indicates that cell-matrix interactions are key drivers of fibrosis progression, yet the mechanisms underlying these responses are not fully elucidated [5, 6, 10–14]. While IPF remains idiopathic, there is growing evidence that genetic and environmental risk factors [15, 16], including older age [17, 18], history of smoking or exposure to airborne hazards [19, 20], and male sex [17, 21], predispose individuals to the disease.

The alveolar region of the lungs—the primary site of gas exchange—is especially vulnerable to damage in chronic respiratory diseases. Within this region, alveolar epithelial type II (ATII) cells play a central role as progenitor cells. These surfactant protein C (SFTPC)-producing cells proliferate, differentiate, and replenish alveolar type I (ATI) cells, which are specialized for gas exchange and are critical to effective epithelial repair [22–24]. Repeated alveolar injury can trigger dysregulated wound-healing responses in both epithelial progenitors and adjacent fibroblasts. Previous studies have shown that coculturing fibroblasts and ATII cells is essential for investigating the cell–cell interactions that drive IPF pathogenesis [6, 25, 26].

Bidirectional epithelial–mesenchymal crosstalk and paracrine signaling profoundly influence fibroblast recruitment, differentiation, and fibrosis progression [27–30]. However, many in vitro models still rely on single-cell cultures of epithelial cells or fibroblasts in isolation. As a result, these systems cannot fully capture the complex intercellular communication between these key subpopulations [31, 32]. There remains a clear need to engineer dynamic three-dimensional distal lung models that support the co-culture of alveolar epithelial cells and fibroblasts. Such models will enable direct study of cell-cell and cell-matrix interactions within a physiologically relevant microenvironment and advance understanding of fibrotic disease mechanisms.

Specifically in IPF, increased impairment of epithelial regeneration results in accumulation of cells stuck in the transition from ATII to ATI characterized by markers of cell-cycle arrest, downregulation of ATII markers, upregulation of ATI markers, and high expression of unique genes including keratins, claudin-4, stratifin, and genes in the transforming growth factor-β (TGF-β) pathway [33, 34]. The cell-cycle arrest of cells in transition from ATII to ATI may result in secretion of chemokines and cytokines such as TGF-β that activate nearby fibroblasts and recruit profibrotic macrophages [33–36]. Persistence and accumulation of transitional alveolar epithelial cells have been strongly linked to disease initiation and progression, highlighting dysregulated epithelial repair as a critical area of IPF research [34, 37–39].

Primary human cells are widely used in lung models to better replicate human-specific cellular and molecular processes. However, primary ATII cells rapidly differentiate to ATI cells within days, which leads to heterogenous cell populations after approximately one week [40, 41]. To overcome this limitation and reduce confounding cellular variables, researchers have increasingly used induced pluripotent stem cell (iPSC)-derived ATII (iATII) cells that can retain a progenitor cell phenotype for months in culture [42, 43].

Many in vitro models in pulmonary regenerative medicine rely on culturing cells on substrates with supraphysiological stiffnesses that do not match lung tissue (e.g., tissue culture plastic), or neglect to investigate the three-dimensional (3D) interaction between cells and the microenvironment. Extensive evidence demonstrates that 3D culture systems more accurately mimic in vivo conditions, preserving cellular physiology and molecular characteristics while enhancing translational relevance [40, 42–45]. When experiments do maintain a 3D microenvironment, most protocols for iPSC differentiation and organoid culture rely almost exclusively on materials (e.g., Matrigel) that do not provide control over mechanical properties [11, 46], geometric cues [6], or biochemical composition [47] – factors that all profoundly impact stem cell fate in vivo [48, 49].

Natural or biologically derived biomaterials—including Matrigel, collagen, alginate, hyaluronic acid, decellularized ECM, and gelatin—remain among the most common materials for pulmonary cell culture applications [12, 23, 42, 43, 50–57]. These systems offer a rich milieu of biochemical cues that promote cell adhesion, proliferation, and differentiation. However, several critical drawbacks limit utility, including poor tunability of mechanical properties, pronounced batch-to-batch variability, and rapid or uncontrolled degradation. Such limitations constrain the capacity to recapitulate the mechanically dynamic microenvironment characteristic of fibrotic lung [58–60].

In contrast, engineered hydrogel systems have emerged as powerful tools enabling precise manipulation of key microenvironmental parameters, including stiffness, degradation kinetics, temporal remodeling, and presentation of bioactive ligands [12, 61]. Although synthetic hydrogels are extensively used in other tissue engineering and mechanobiology contexts, adoption in pulmonary research remains comparatively limited [10, 12, 62, 63]. Among these platforms, poly(ethylene glycol) norbornene (PEG-NB) hydrogels are particularly versatile and cytocompatible. Although PEG-NB hydrogels provide well-defined 3D cell culture environments with finely tunable mechanical and biochemical properties, supporting more rigorous investigation of cell–matrix interactions and disease-relevant phenotypes in the distal lung, [12, 64] only a few studies have used PEG-based hydrogels to model the lung microenvironment [6, 65, 66].

Here, we present an engineered 3D lung model that recapitulates key features of early pulmonary fibrosis while controlling for confounding variables such as sex, an important consideration given that nearly 70% of IPF patients are male [21]. iATII spheroids were magnetically aggregated together and embedded within PEG-NB hydrogels containing pulmonary fibroblasts, replicating key features of the cellular spatial arrangement found within the alveoli [65, 67]. This co-culture platform provided an environment that facilitated epithelial-fibroblast interactions [6]. Soft (elastic modulus (E) = 5.06 ± 0.13 kPa) and stiff (E = 18.99 ± 1.30 kPa) hydrogel microenvironments supported cell viability while effectively recapitulating the mechanical properties of healthy and fibrotic lung tissue, respectively. Beyond mechanical stiffness, pro-fibrotic biochemical cues, previously described as a fibrosis cocktail, were supplemented into the culture medium to induce epithelial injury and subsequent fibroblast activation [8, 68, 69]. Gene expression analyses revealed that epithelial and fibroblast responses within stiff hydrogels exposed to the fibrosis cocktail closely matched trends measured in fibrotic lung tissues. To further validate the model as a viable platform for pre-clinical therapeutic drug screening, Nintedanib, an FDA-approved anti-fibrotic drug was tested. This treatment downregulated multiple fibroblast markers and upregulated transitional and ATI markers, indicating a possible recovery in epithelial repair and a decrease in fibrotic phenotypes. This work tested the hypothesis that both a fibrotic microenvironment and epithelial-mesenchymal interactions were necessary to support screening of anti-fibrotic therapies. These results show that 3D human co-culture models are a promising for tool for advancing our ability to study cell-cell and cell-matrix interactions in pulmonary fibrosis and supporting drug discovery and validation studies.

Materials and methods

2.1. PEG-NB synthesis

As previously published, terminal residue conjugation of an eight-arm, 10 kg/mol PEG-hydroxyl macromer (PEG-OH) produced norbornene functionalized end groups [6, 70]. In brief, PEG-OH (5 g, JenKem Technology) was lyophilized (~ 0.1 mBar, ~ −80°C) and subsequently dissolved in ~35 mL anhydrous dichloromethane (DCM; Sigma-Aldrich, cat. #270997–1L) under moisture-free conditions in a flame-dried Schlenk flask. 4-Dimethylaminopyridine (DMAP; 0.24 g, .002 mol, Acros Organics, cat. #148270050) was added to the flask and pyridine (1.61 mL, 0.02 mol, Sigma Aldrich 494410) was injected dropwise into the reaction mixture. Separately in a second flame-dried Schlenk flask, N,N’-Dicyclohexylcarbodiimide (DCC; 4.13 g, 0.02 mol, Fisher Scientific, cat. #AC113901000) was dissolved in anhydrous DCM, again under moisture-free conditions. To this flask, norbornene-2-carboxylic acid (4.9 mL, 0.04 mol, Acros Organics, cat. #453300250) was added in a dropwise manner. After 30 minutes of stirring at room temperature, the reaction mixture was filtered through Celite 545 (EMD Millipore, cat. #CX0574–1). Then, the filtrate was added to the first flask and left to react for 48 h (while protected from light). A series of wash steps using 5% sodium bicarbonate, saturated brine (~40 grams of sodium chloride in 100 mL deionized water), and deionized water removed undesired byproducts. Each time the reacted polymer was mixed and left to separate out into two phases for approximately 5 minutes using a separatory funnel. Anhydrous magnesium sulfate (Fisher Scientific, cat. #M65–500) was added to the organic elute to remove excess water and then filtered out using filter paper (Cytiva, cat. #1002–150). The organic product was precipitated with cold diethyl ether (Fisher Scientific, cat. #E1384) and then concentrated with a rotary evaporator. Following a 4°C overnight incubation, the diethyl ether was removed using vacuum filtration. The precipitate was vacuum dried at room temperature in a desiccator overnight, again protected from light. Dialysis with the precipitate occurred at room temperature over 72 h, where the 3.5 L of deionized water was changed four times daily. After dialysis, the product was collected and lyophilized (~ 0.1 mBar, ~ −80°C) to obtain a solid white powder.

Nuclear magnetic resonance (NMR) spectroscopy confirmed the end-group functionalization and purity of the PEG-NB. A Bruker DPX-400 FT NMR spectrometer was used to collect the ¹H NMR spectrum of the product using 184 scans and a 2.5 s relaxation time. Only synthesis products above 90% functionalization were used throughout these experiments (Supplemental Fig. 1), and chemical shifts for protons (¹H) were recorded relative to deuterochloroform as parts per million (ppm).

2.2. iATII cell culture and magnetic labeling

iATIIs (Table 1), generously provided by the Kotton Laboratory (Center for Regenerative Medicine, Boston University) and commercially known as BU3 NGST cells (RRID: CVCL_WN82), containing thyroid transcription factor NKX2 homeobox 1 Green fluorescent protein (NKX2–1^GFP) and Surfactant protein C tdTomato (SFTPC^tdTomato) reporters, were maintained in 40 μL of 8 mg/mL growth factor reduced Matrigel (Corning, cat. #356231) and medium containing CHIR, KGF, Dexamethasone, 8-Bromoadenosine 3’, 5’- cyclic monophosphate sodium salt, and 3-Isobutyl-1-methylxanthine (CK + DCI) (Table S1) [42, 43]. During routine passaging, 0.05% Trypsin-EDTA (~15 min, Gibco, cat. #25–300-062) was used to dissociate iATIIs back into a single cell state [42, 43]. Nanoshuttle (1μL per every 10,000 cells, Greiner Bio-One, cat. #657846) was then added to a proportion of the iATIIs so that the cells could be magnetically aggregated a few days later. This was done by pipetting the cells and Nanoshuttle up and down gently until visibly homogenous (1–2x) and then centrifuging at 300 × g for 5 minutes at 4°C. This process was repeated an additional two times before the iATII pellet was resuspended in 40 μL of 8 mg/mL Matrigel and standard passaging protocols were resumed [42, 43]. Nanoshuttle is a nanoparticle assembly that consists of gold, iron oxide, and poly-L-lysine. This mixture enables the Nanoshuttle beads to attach to the cell membranes electrostatically. iATIIs with Nanoshuttle were left to grow into small alveolospheres for 4–5 days prior to use in experiments.

Table 1.

Human cell and serum information.

Material	Identifier	Donor Information
Fibroblasts	hNLF01	71-year-old male
Fibroblasts	hNLF40A	69-year-old male
Fibroblasts	hNLF15A	60-year-old male
iATIIs	BU3 NGST	32-year-old male
Serum	1146887	65-year-old male

2.3. Fibroblast cell culture

Patient-specific fibroblasts from non-diseased, non-smokers were isolated using a previously established protocol [71]. Briefly, freshly harvested lung explants were minced into 1–2 mm³ sections and cultured in growth medium (Dulbecco’s Modified Eagle Medium (DMEM) and 10% v/v heat-inactivated fetal calf serum (FBS)). After 10 – 14 days, the fibroblasts were trypsinized and maintained in culture for at least one passage before freezing. Prior to use in experiments, these frozen vials of non-diseased, non-smoker patient-specific human lung fibroblasts (Table 1) were thawed and expanded at 37°C and 5% CO₂ in T75 flasks containing growth medium (Dulbecco’s Modified Eagle Medium (DMEM), 10% v/v charcoal-stripped fetal bovine serum (CS-FBS), and 1% v/v penicillin/streptomycin). All fibroblasts used in experiments were seeded between passages two and seven.

2.4. Preparation of the embedding hydrogel

The initial PEG-NB weight percent determined whether the embedding hydrogel corresponded to a soft (5.25 wt%) or stiff (7.75 wt%) formulation, with a 0.7 ratio of thiols to norbornenes. A matrix metalloproteinase-9 (MMP9)-degradable peptide (Ac-GCRD-VPLSLYSG-DRCG-NH2, GL Biochem) was used as a crosslinker and both fibronectin (CGRGDS, 2 mM, GL Biochem) and laminin (CGYIGSR, 2 mM, GL Biochem) mimetic peptides, as well as 2 mg/ml Laminin/Entactin (Corning, cat. #354259) were incorporated into the formulation to enhance cell adhesion. Lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP, 1.1 mM) acted as the photoinitiator. Cell culture medium (CK + DCI) was used to reconstitute the PEG-NB, MMP9-degradable peptide crosslinker, CGRGDS, and CGYIGSR. Individual volumes of each component were mixed to produce the final hydrogel precursor solutions (soft: 5.25 wt% PEG-NB, 13.29 mM MMP9, 2 mM CGRGDS, 2 mM CGYIGSR, 2 mg/mL laminin/entactin, and 1.1 mM LAP; stiff: 7.75 wt% PEG-NB, 19.62 mM MMP9, 2 mM CGRGDS, 2 mM CGYIGSR, 2 mg/mL laminin/entactin, and 1.1 mM LAP). Next, 2 uL of each hydrogel precursor solution was pipetted onto a pH indictor strip to confirm that the pH of the hydrogel precursor solution fell between 7 and 8, prior to any cell exposure. One lot of the MMP9-degradable peptide was received from the supplier in a form that created an overall pH outside of this range. For hydrogels made using that crosslinker, 1 uL of 1 M sodium hydroxide was added to the final hydrogel precursor solution to neutralize the overall pH prior to embedding cells.

2.5. Rheological characterization of acellular hydrogels

Hydrogels for rheological evaluation were prepared by pipetting 40 μL of final hydrogel precursor solution between two glass slides covered in parafilm and separated by a 1 mm gasket. After ultraviolet (UV) light exposure at 365 nm with 10 mW cm⁻² intensity (Omnicure, Lumen Dynamics) for 5 min, these samples polymerized. Hydrogels were then swollen in phosphate buffered saline (PBS) overnight prior to characterization. The elastic modulus of the hydrogels (e.g., stiffness) was measured with an 8-mm parallel plate geometry on a Discovery HR2 rheometer (TA Instruments) as previously described [9, 70]. In brief, a hydrogel was placed onto the Peltier plate (37°C) and the geometry was lowered until it was in contact with the hydrogel surface and an axial force of 0.03 N was registered. The storage modulus (G’) plateau was determined by measuring the storage modulus at different 5% increments of compression until a maximum was reached [6, 70]. The storage modulus plateau for the soft hydrogels occurred at 25% compression, and 30% compression for the stiff hydrogels. The hydrogel samples then underwent a frequency oscillation with logarithmic sweep of frequencies (1–100 rad s⁻¹) and 1% strain. From here, the elastic modulus (E) was calculated under the assumption that the hydrogels were incompressible and exhibited bulk-elastic characteristics with a Poisson ratio of 0.5 [10, 72–74]. There were no differences in elastic modulus measurements whether these hydrogels were swollen in cell culture medium or PBS (Fig. S2).

2.6. Formation of 3D alveolosphere aggregates

Both enzymatic (2 mg/mL dispase; Thermo Fisher Scientific, cat. #17105–041) and manual pipetting were used to free the iATII alveolospheres from Matrigel constructs over 30 minutes. The alveolospheres were then rinsed in DMEM and pelleted, using centrifugation at 300 × g for 5 minutes and 4°C. This washing step was completed a total of 3x to ensure the enzyme was completely removed. Following the last spin, the alveolospheres were resuspended in CK + DCI medium (Supplemental Table S1; [42, 43]) supplemented with 10 μm Y-27632 (Tocris, cat. #1254), counted, and then transferred into 24-well cell-repellent plates (Greiner Bio-One, cat. #662970) at a concentration of 400 spheres and 250 μL of CK + DCI + 10 μm Y-27632 medium per well. Magnetic levitation drives (Bio-Assembler, Greiner Bio-One, cat. # 662840) were added to the plates and then transferred onto an orbital shaker (~60 rpm) within a cell culture incubator (37°C, 5% CO₂) to facilitate aggregation of alveolospheres into 3D aggregates. After 3 hours, the magnetic levitation drives were removed and replaced with magnetic concentrating drives (Greiner Bio-One, cat. # 662840) and the epithelial cells were allowed to settle over approximately 5 minutes. At this point, medium was removed manually from each well and a hydrophobic pen (Vector Laboratories, cat. #H-4000) was used to draw a circle around the cells. This barrier ensured that the alveolospheres were completely encapsulated in the hydrogel precursor solution prepared in the subsequent steps. It was important to minimize the amount of time the iATIIs were left without medium to ensure high viability.

2.7. Hydrogel embedding of 3D alveolosphere aggregates

Just prior to embedding, fibroblasts were trypsinized, assessed for viability with Trypan Blue, and counted on a hemocytometer. Fibroblasts were pelleted and then resuspended in the precursor hydrogel solution, so the final concentration was 1,000 fibroblasts/μL. A total of 40 μL of embedding hydrogel precursor solution containing fibroblasts (40,000 fibroblasts total per sample) was added directly on top of the exposed alveolosphere aggregate. Five minutes of UV light exposure (365 nm, 10 mW cm⁻², Omnicure, Lumen Dynamics) polymerized the hydrogels. Prior work has established that there are no significant differences in cell viability and transcriptome when cells are exposed to 365 nm light for this length of time [70, 75, 76]. Afterwards, these samples were carefully transferred into new 24 well plate wells that contained CK + DCI medium [42, 43] that was supplemented with 10 μm Y-27632 and 1% serum from a 65-year-old male donor (Table 1) for 48 h (37°C and 5% CO₂).

2.8. Fibrosis cocktail exposure

Samples were maintained at 37°C with 5% CO₂ in CK + DCI medium [42, 43] supplemented with 1% serum from a 65-year-old male donor (Table 1). For fibrotic activation experiments, samples were either exposed to a fibrosis cocktail (FC) or vehicle control (VC). The fibrosis cocktail contained 5 ng/ml recombinant transforming growth factor beta (TGF-β; PeproTech, cat. #100–21), 10 ng/ml platelet-derived growth factor AB (PDGF-AB; Thermo Scientific, cat. #PHG0134), and 5 μM 1-Oleoyl Lysophosphatidic Acid (LPA; Cayman Chemical Company, cat. #62215) [8, 68, 77, 78]. Dosing began on day 2 and continued until day 8, where each well was replenished with medium (1 mL/well) containing the FC or VC (PBS supplemented with 0.1% bovine serum albumin (BSA)) every 48 h.

2.9. Morphology, Live-dead and immunofluorescence staining

To assess SFTPC expression prior to aggregation, alveolospheres were pipetted onto a glass slide and visualized using an inverted fluorescence microscope. The number of alveolospheres expressing tdTomato was manually counted across five regions and compared to the total number of alveolospheres visible under phase contrast. To visually assess where fibroblasts were within the hydrogel relative to the iATIIs, CellTracker Green CMFDA (10 μM, Thermo Fisher, cat. #C7025) was used for non-viability related samples. This fluorescently tagged the fibroblasts green after incubating the CellTracker Green CMFDA on the cells for 45 min in serum free medium. Following, the fibroblasts were dissociated with trypsin, embedded as described in section 2.7, and imaged while in culture. The constructs were imaged on a Leica Spinning Disk confocal microscope with both 10x and 40x objectives. Diameter of both alveolospheres and aggregates was quantified in Fiji (ImageJ) by using the line tool to draw along the major and minor axes and then the measure function to provide numerical values.

Commercially available human pulmonary fibroblasts (HPFs) were used for all viability studies (passage 2–7). A ReadyProbes Cell Viability Imaging Kit (Thermo Fischer Scientific, R37609) quantified the number of live cells in each construct. Samples treated with FC or VC were collected for imaging after 2, 4, 6, or 8 days. Prior to imaging, 1 drop of NucBlue (nuclei) and 1 drop of NucGreen (dead) was added to each 1 mL of cell culture medium to make a staining medium. Samples were transferred into 24-well plate wells containing 300 μL of staining medium and on an orbital shaker for 1 h (37°C, 5% CO₂). Afterwards, samples were transferred onto a glass slide and covered in PBS to maintain hydration during imaging. A hydrophobic pen was used to confine the PBS to the sample area.

Fluorescently stained samples were imaged on either an Olympus CKX53 upright microscope adapted for fluorescent capabilities with DAPI, FITC, and TRITC filters or a 4 channel Olympus BX63 epifluorescent microscope. For image quantification, six random points in the construct were imaged at 4x using 100 ms exposures for the DAPI (nuclei) channel, 10–20 ms for the FITC (dead) channel, and 100 ms for the TRITC (SFTPC) channel. Exposures were kept constant each day of imaging and between samples that were directly compared. Images were post-processed and analyzed with ImageJ software (NIH). Total cell viability was quantified using Equation 1.

$$\mathit{\text{Percentage of Live Cells}}=1-\left(\frac{\mathit{\text{NucGreen Area}}}{\mathit{\text{NucBlue Area}}}\right)*100$$ Equation 1:

2.10. Magnetic-activated cell sorting (MACS)

Epithelial cells and fibroblasts were purified out of co-culture using magnetic column isolation based on the expression of epithelial cell adhesion molecule (EpCAM; CD326), which is a cell surface marker that is found on ATII, transitional epithelial, and ATI cells [6, 79, 80]. On days 6 and 8, an enzymatic digestion solution containing 5 mg/mL dispase and 1 mg/mL elastase (Worthington Biochemical, cat. #LS002292) was prepared fresh in DMEM. Old medium was removed from each well and replaced with 700 μL of the enzymatic digestion solution. Both enzymatic and manual pipetting were used to free the cells from hydrogel constructs, which took up to 1 h. The dispase and elastase enzymes targeted the MMP9-degradable crosslinker sequence and helped facilitate degradation. Four to six samples of the same experimental group were pooled together to form each technical replicate. Once hydrogel degradation occurred, the solution of iATIIs and fibroblasts was transferred into 15 mL test tubes, diluted 1:1 with DMEM, and then centrifuged at 300 × g for 5 minutes and 4°C to create a cell pellet.

Trypsin-EDTA (0.05%; Gibco, cat. #25–300-062) was added to the cell pellets and then transferred into 6-well plates to allow for the alveolospheres to dissociate into single cells for approximately 16 min, with manual pipetting (~2x) at the halfway timepoint. To deactivate the trypsin, a 10% v/v CS-FBS in DMEM medium was then added to the samples and then centrifuged for 5 mins at 300 × g for 5 minutes and 4°C. Next, the supernatant was manually discarded, and the cell pellets were resuspended in 10 μL of anti-CD326 (epithelial cell adhesion molecule (EpCAM)) Microbeads (Miltenyi Biotec, cat. #130–061-101) and 70 μL of buffer that consisted of 2 mM ethylenediaminetetraacetic acid (EDTA; ThermoFisher, cat. #AM9260G) and 0.5% BSA (Sigma-Aldrich) in PBS (PEB buffer). These samples were incubated at 4°C for 15 min to allow bead binding to cells before an additional 1 mL of PEB buffer was added to each and the test tubes were centrifuged at 300 × g for 5 minutes and 4°C. From each cell pellet, the supernatant was removed, and the cells were resuspended in 500 μL of PEB buffer, ensuring there were no cell clumps.

iATIIs were positively selected using a MiniMACS Separator and Starting Kit (Miltenyi Biotec, cat. #130-090-312) set up and MS columns (Miltenyi Biotec, cat. #130-042-201) according to the manufacturer’s protocol. Briefly, the MS columns were primed by passing 500 μL of PEB buffer through the columns. Then, a test tube was placed beneath the column and the cells in 500 μL of PEB buffer was passed through to allow the EpCAM⁺ cells bound to beads to be trapped in the magnetic field. An additional two rinse steps of 500 μL of PEB were done to ensure all EpCAM⁻ fibroblasts were collected in the test tube below. Lastly, a new test tube was placed beneath the column at this point, and 1 mL of PEB buffer was added to the column. The column was then removed from the magnet holder, and the plunger was compressed to elute the positively selected EpCAM⁺ cells into the second test tube. Each respective test tube containing cells was centrifuged one final time at 300 × g for 5 minutes and 4°C.

2.11. RNA isolation and cDNA synthesis

Immediately following MACS isolations, the supernatant was removed from the iATII and fibroblast cell pellets, and 300 μL of TRIzol Reagent (Fisher Scientific, cat. #15-596-026) was added to each test tube. The samples were pipetted and briefly vortexed before being stored at −20°C for up to one month. After samples were thawed, the sample volume was transferred into a 1.5 mL Eppendorf tube. 100 μL of 1-bromo-3-chloropropane (BCP; Fisher Scientific, cat. #NC9551474) was added to each Eppendorf tube. Next, each Eppendorf tube was briefly vortexed, incubated at room temperature for 5 min, and incubated on ice for an additional 5 min. At this time, the Eppendorf tubes were centrifuged at 12,000 × g for 15 min, so the clear aqueous layer of the sample could be transferred into a separate RNAse-free 1.5 mL microcentrifuge tube. An equal amount of 100% ethanol (EtOH) was added to the clear RNA layer volume and the two were briefly vortexed. Up to 700 μL of total volume was transferred into a RNeasy Plus Micro Kit column (Qiagen, cat. #74034) and purified according to the manufacturer’s instructions. RNA quantity and purity, as assessed by the ratio of 260 nm and 280 nm (A₂₆₀/A₂₈₀) absorbance readings, were measured using a BioTek plate reader and a Take3 Micro-Volume Plate. The isolated RNA was then converted into cDNA using a high-capacity cDNA Reverse Transcription Kit (Applied Biosystems, cat. #4368814) according to the manufacturer’s protocol.

2.12. Assessment of cell specific gene expression

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) assessed gene expression for a variety of different ATII (SFTPC, LAMP3), ATII-ATI (KRT17, CLDN4), ATI (PDPN, AQP5), aberrant basaloid cell (FN1), fibroblast activation (CTGF, CTHRC1, LTBP2), and ECM remodeling (COL1A1, FN1, MMP2, MMP9) markers. iTaq Universal SYBR Green Supermix (Bio-Rad, cat. #1725121) and a CFX Opus 96 (Bio-Rad) were used for all experiments. Fibroblast gene expression was normalized to ribosomal protein L30 (RPL30), which served as the housekeeping gene, whereas epithelial gene expression was normalized to ribosomal protein S18 (RPS18). All human primers were acquired from Integrated DNA Technologies (Supplemental Table S2). All Ct values were natural log transformed to normalize the data for statistical analyses [81], and then relative gene expression was calculated using a 2^−ΔCt approach. After statistical analyses were computed, values were untransformed (by taking the exponent of the natural logged value) and presented in figures.

2.13. Validating response to anti-fibrotic drug treatment

To narrow down conditions for therapeutic drug testing, the influence of the age of donor-derived fibroblasts, substrate elastic modulus, exposure, and time on either epithelial gene expression or fibroblast activation gene expression were investigated with a design of experiment (DOE) approach using JMP software (Pro 18 Version, SAS). The resulting least-squares regression model identified the stiff hydrogel formulation as the most fibrotic microenvironment. Therefore, all subsequent drug studies were only conducted with donor specific fibroblast cells in stiff hydrogels and gene expression was assessed at the day 8 timepoint. For drug treatment, 10 μM of Nintedanib (Tocris, cat. #7049) in dimethyl sulfoxide (DMSO) was supplemented into the cell culture medium (CK + DCI + 1% male serum) simultaneously with the FC components starting on day 4 and replenished every 48 h at the subsequent medium changes. Meanwhile, samples that were kept in cell culture medium (CK + DCI + 1% male serum) with only the FC components served as controls.

2.14. Statistical methods

Data sets were assessed for normality using Shapiro-Wilks tests. For parametric datasets, either two-way ANOVAs followed by Tukey’s honest statistical difference (HSD) tests (GraphPad Prism) or unpaired t-tests (GraphPad Prism) were applied. Mann Whitney tests were applied to determine statistical significance for nonparametric datasets. For each viability timepoint, images were acquired from six individual hydrogels (n=6 technical replicates per experimental condition). Two-way ANOVAs followed by Tukey’s honest statistical difference (HSD) tests (GraphPad Prism) were applied to compare VC and FC results on days 2, 4, 6, and 8. Similarly, for rheological measurements, separate hydrogels (n=3–6 technical replicates per stiffness) were measured and Mann Whitney tests calculated statistical significance (GraphPad Prism).

For all MACS isolations, 4–6 technical replicates of the same experimental group were pooled together to form each biological replicate. In total, the RT-qPCR results presented in Figure 4 represents the 3 biological replicate averages (from the 3 separate donor specific fibroblast lines). For all gene expression results, the raw values were natural log transformed to normalize the data for statistical analyses [81]. After statistical analyses were computed, values were untransformed (by taking the exponent of the natural logged value) and presented in figures. These transformed relative gene expression values were entered into the DOE and a standard least squares model was applied to fit the model and identify best fit lines. The JMP software identified a multi-factorial design and approached the statistical analysis similarly to a three-way ANOVA. For the Nintedanib studies, 6 technical replicates were pooled together to form each biological replicate. Figure 7 presents the 3 biological replicate averages (from the 3 separate donor specific fibroblast lines). Statistical significance was determined with unpaired t-tests for these results. All statistical tests applied are individually listed within each figure caption.

Fig. 4. iATII-fibroblast co-culture recapitulated gene expression patterns characteristic of fibrotic tissue.

Heat map showing relative gene expression of ATII, transitional epithelial, and ATI markers assessed in the epithelial cell fraction of co-culture models, and markers of activation and ECM remodeling assessed in fibroblasts. Gene expression patterns in 3D lung models matched trends measured in fibrotic lung tissue (n=3 biological replicates, with 4–6 technical replicates pooled prior to RNA isolation).

Fig. 7. 3D distal lung models were responsive to anti-fibrotic treatment.

(A) Schematic representation of the therapeutic treatment experimental timeline and outputs. (B) Relative gene expression for EpCAM⁺ cells after Nintedanib treatment, normalized to FC-only samples. Data presented as mean ± SEM with symbols representing biological replicates (n=3 biological replicates, with 4–6 technical replicates). All gene expression data were normalized using a natural log transformation for statistical analyses. Statistical significance was determined by unpaired t-tests between Nintedanib-treated and untreated conditions for each cell line. (C) Relative gene expression for fibroblasts after Nintedanib treatment, normalized to FC-only samples. Data presented as mean ± SEM with symbols representing biological replicates (n=3 biological replicates, with 4–6 technical replicates). All gene expression data were normalized using a natural log transformation for statistical analyses. Statistical significance was determined by unpaired t-tests between Nintedanib-treated and untreated conditions for each cell line.

Results

3.1. PEG-NB hydrogels recapitulated key aspects of fibrotic tissue

To engineer a cellular microenvironment that recreated fibrosis progression in vitro, iATIIs and fibroblasts were embedded in well-defined, tunable stiffness 3D PEG-NB hydrogels (Fig. 1A). These hydrogel formulations consisted of a 93% functionalized PEG-NB macromer (Supplemental Fig. S1), an MMP9-degradable peptide crosslinker, a fibronectin mimetic peptide (CGRGDS), a laminin mimetic peptide (CGYIGSR), and entrapped laminin/entactin protein complex (Fig. 1A). Both the pendant peptides and the laminin/entactin protein complex facilitated increased cellular adhesion to hydrogels. Modification of the PEG weight percent varied the elastic modulus (e.g., stiffness) to produce hydrogels with discrete stiffnesses. Soft hydrogels demonstrated an elastic modulus of 5.06 ± 0.13 kPa, while stiff hydrogels exhibited an elastic modulus of 18.99 ± 1.30 kPa, effectively matching the mechanical properties of healthy and fibrotic lung tissue [10, 11], respectively (Fig. 1B). iPSC-derived iATIIs and fibroblasts were arranged within these hydrogels to replicate key features of 3D lung micro-architecture.

Fig. 1. Engineered hydrogels supported iATII-fibroblast co-cultures with stiffnesses tuned to match healthy and fibrotic lung tissue.

(A) Cells were embedded in 3D arrangements mimicking key features of alveolar micro-architecture within soft or stiff hydrogel formulations that contained PEG-NB, an MMP9-degradable crosslinker, CGRGDS, CGYIGSR, and entrapped laminin/entactin. (B) Rheological measurements for the average elastic modulus (E) of soft (E = 5.06 ± 0.13 kPa, n=6 technical replicates) and stiff hydrogels (E = 18.99 ± 1.30 kPa, n=6 technical replicates). Soft and stiff hydrogel formulations fall within the ranges for healthy (green region) and fibrotic (red region) lung tissue stiffness. Columns represent mean ± SEM. Symbols represent technical replicates. Normality was assessed with a Shapiro-Wilks test, and statistical significance was determined by a Mann Whitney test. (C) Schematic representation of the experimental timeline and outputs.

iATII alveolospheres were magnetically aggregated together using Nanoshuttle and bioassembler magnetic drives (Greiner) to form alveolosphere aggregates. Then, fibroblasts were distributed throughout the embedding hydrogel precursor solution and added on top of the iATII aggregates. As a result, the samples consisted of an epithelial core surrounded by fibroblasts, which enabled epithelial-fibroblast crosstalk within 3D lung models. Fig. 1C describes the different experimental outputs and mediums that were used over the course of the 8-day experimental timeline. Samples were kept in CK+DCI medium supplemented with 1% human serum and a rock inhibitor for the first 48 hours. On day 2, the medium was then switched to CK+DCI medium supplemented with 1% human serum and either the fibrosis cocktail (FC) or vehicle control (VC) components. This medium was replenished every 48 h until day 8. Cell viability was assessed on days 2, 4, 6, and 8 while magnetic column isolations to separate epithelial and fibroblast cellular subpopulations occurred on days 6 and 8 (Fig. 1C).

3.2. Cells arranged in 3D lung models maintained high viability

3D lung models successfully replicated important cellular arrangements characteristic of distal lung tissue. Magnetic aggregation of iATII alveolospheres formed an arrangement of epithelial spheroids clustered together to replicate aspects of alveolar architecture (Fig. 2A). To visualize the spatial arrangement of fibroblasts relative to the alveolosphere aggregates fibroblasts were labeled with CellTracker (green) (Fig. 2A). The proximity of fibroblasts to epithelial aggregates within these 3D lung models enables cell-cell interaction as described in our previous studies [6]. Epithelial aggregates maintained SFTPC-tdTomato reporter fluorescence (red) over time. Quantification showed that 86% of alveolospheres were positively expressing the SFTPC reporter at the time of embedding (Fig. 2B). The mean area of the alveolospheres at the time of seeding was approximately 30,000 μm² (Fig. 2C), within the range reported for human alveoli (20,000 to 40,000 μm²) [82, 83]. The average diameter of alveolosphere aggregates after embedding was 827 ± 65 μm (Fig. 2D).

Fig. 2. 3D lung models replicated key features of distal lung architecture.

(A) Representative maximum intensity projection of confocal images of 3D lung models in PEG-NB hydrogel showing the spatial arrangement of iATIIs expressing SFTPC (red) near alveolar fibroblasts (green), counterstained for nuclei (blue). White box represents region of interest (ROI) that is called out within higher magnification image of that area. Scale bars = 100 μm. (B) Percentage of alveolospheres expressing SFTPC prior to embedding within 3D lung models. (C) Average alveolosphere area prior to embedding represented as mean ± SEM (n=35). (D) Average aggregate diameter within 3D lung models presented as mean ± SEM (n=18). Symbols represent technical replicates.

Total cellular viability in 3D distal lung models was measured with a ReadyProbes Cell Viability Imaging Kit at days 4, 6, and 8 for each of the four conditions combining two different stiffness (soft or stiff) hydrogels and two different exposures (VC or FC) and day 2 in soft or stiff hydrogels before FC exposure. Representative images showed all nuclei marked by blue fluorescence and dead cells marked by green fluorescence (Fig. 3A). After the embedding process on day 2, approximately 92% of cells remained alive within the soft and stiff hydrogels (Fig. 3B–C). Cell viability within the soft hydrogels measured approximately 98% at day 4 and 96% by day 8, with no differences in viability between VC and FC conditions (Fig. 3B). Stiff hydrogel cell viability was approximately 85% at day 4 and increased slightly to 92% by day 8 (Fig. 3C), indicating that even if the stiff microenvironment induced some initial cell death post-embedding, overall, these models promoted high cell viability. The results also showed that there were no statistically significant differences in cell viability between the VC and FC exposures for both soft and stiff hydrogels across all timepoints (Fig. 3B–C).

Fig. 3. Cells maintained viability in 3D hydrogels over time.

(A) Representative images of cells within soft and stiff 3D lung models on days 2 (VC only) and 8 (VC and FC). Cell nuclei were stained blue and dead cells were labeled green. Scale bars = 100 μm. (B) Quantification of cell viability in soft hydrogels either treated with the vehicle control (VC) or fibrosis cocktail (FC). Columns represent mean ± SEM, n=6 technical replicates. Symbols represent technical replicates. Normality was assessed with a Shapiro-Wilks test, and statistical significance was determined by a two-way ANOVA with Tukey’s multiple comparisons test, ns = no significance. (C) Quantification of cell viability in stiff hydrogels either treated with VC or FC. Columns represent mean ± SEM, n=6 technical replicates. Symbols represent technical replicates. Normality was assessed with a Shapiro-Wilks test, and statistical significance was determined by a two-way ANOVA with Tukey’s multiple comparisons test, ns = no significance.

3.3. Pro-fibrotic biochemical cues induced epithelial damage and fibroblast activation

After samples were in culture, FC was supplemented into the medium for half of the samples to provide pro-fibrotic cues and further induce fibrotic activation independent of microenvironmental stiffness. The FC (5 ng/ml TGF-β, 10 ng/ml PDGF-AB, and 5 μM LPA) was initially added on day 2 and replenished every 48 hours until day 8. Magnetic column isolations positively selected for EpCAM⁺ cells on days 6 and 8, enabling cell-specific relative gene expression. This approach ensured ATII, ATI, and alveolar epithelial transitional cells were separated from the fibroblasts. Relative gene expression results determined by RT-qPCR were visualized within a heat map (Fig. 4). Surfactant protein C (SFTPC) and lysosome-associated membrane protein 3 (LAMP3) served as ATII markers, keratin 17 (KRT17) and claudin-4 (CLDN4) served as ATII-ATI transitional epithelial cell markers, podoplanin (PDPN) and aquaporin-5 (AQP5) served as ATI markers. Fibroblast activation markers were connective tissue growth factor (CTGF), collagen triple helix repeat containing 1 (CTHRC1), and latent TGF-β binding protein 2 (LTBP2), while ECM remodeling by fibroblasts was measured with collagen 1 alpha chain 1 (COL1A1), fibronectin 1 (FN1), matrix metalloproteinase 2 (MMP2) and matrix metalloproteinase 9 (MMP9). Increased microenvironmental stiffness and FC exposure were expected to downregulate ATII gene expression, and increase ATII-ATI, ATI, and fibroblast activation gene expression. Relative gene expression values followed these expected results across nearly all genes, with a notable deviation occurring with the AQP5 gene expression (Fig. 4). Overall, the heat map results supported the hypothesis that epithelial damage from FC exposure led to epithelial differentiation and contributed to fibroblast activation and ECM remodeling. Numeric relative gene expression values are available in the Supplementary Material (Fig. S3). Immunostaining also confirmed that fibroblasts were producing intracellular collagen at the protein level in all conditions (Fig. S4.).

3.4. Statistical analyses revealed factors that created the pronounced fibrotic features in 3D distal lung models

Two individual designs of experiments further analyzed statistically how the input variables of fibroblast donor age, hydrogel elastic modulus, exposure (VC or FC), and day of collection for analysis influenced either epithelial or fibroblast gene expression. Primary fibroblasts were isolated from non-diseased male donors (Table 1) aged 60, 69, and 71 years old. The average elastic modulus measurements for the soft and stiff hydrogels were 5 kPa and 18.99 kPa. Additionally, gene expression was assessed on days 6 and 8. In response to epithelial injury, ATII cells can differentiate into ATI cells to repair damaged tissues. In pulmonary fibrosis dysregulated healing may result in the accumulation of transitional epithelial cells [33, 34]. Following in vivo observations, the statistical model was directed to maximize all response variables except SFTPC gene expression, which was minimized to replicate a fibrotic healing response. Trend lines for each output variable plotted in response to each input variable provided a visual depiction of how epithelial cells responded to each input (Fig 5A).

Fig. 5. Results showed that fibroblasts donor age, elastic modulus, exposure, and time all significantly influenced epithelial cell gene expression.

(A) Trend lines from experimental data showed how SFTPC, LAMP3, KRT17, CLDN4, PDPN, and AQP5 gene expression changed in relation to the age of fibroblast donors, elastic modulus of the hydrogel, exposure, and time. Data presented as mean ± SEM (n=3 biological replicates, with 4–6 technical replicates). All gene expression data were normalized using a natural log transformation for statistical analyses. (B) The effect magnitude analysis identified all factors and interactions that significantly influenced epithelial gene expression. These results also predicted that the most fibrotic microenvironment for epithelial cells within the conditions tested would be achieved by day 6 using the 69-year-old donor-derived fibroblasts embedded in stiff (18.99 kPa) hydrogels and the fibrosis cocktail (FC).

Epithelial gene expression was upregulated within the stiff hydrogels (Fig 5A). The combination of input variables tested in this work that created the most fibrotic condition for epithelial cells was achieved by day 6 when the 69-year-old donor-derived fibroblasts were embedded within stiff (18.99 kPa) hydrogels and exposed to the FC (Fig. 5A). The input variable of hydrogel elastic modulus had the greatest influence on the epithelial gene expression (p = 0.001) (Fig. 5B), based on the models and factors that were tested. The next most influential factors were the interaction between the age of donor-derived fibroblasts and the elastic modulus (p = 0.002), exposure (p = 0.006), the age of donor-derived fibroblasts alone (p = 0.015), the interaction between the age of donor-derived fibroblasts with time and the elastic modulus (p = 0.019), the interaction between the time and elastic modulus (p = 0.030), time (p = 0.036), and the interaction between the age of the donor-derived fibroblasts and exposure (p = 0.047) (Fig. 5B).

In contrast to the epithelial results, fibroblast activation was more responsive to FC exposure than microenvironmental modulus. Relative fibroblast activation gene expression was upregulated within the FC condition compared to VC (Fig. 6A). The fibroblast results identified that the most fibrotic activation within the conditions tested here would occur at day 8 with the 60-year-old donor-derived fibroblasts embedded in stiff (18.99 kPa) hydrogels with FC exposure (Fig. 6A). The input variable of exposure had the greatest influence on the fibroblast gene expression (p = 0.005) (Fig. 6B). The next most influential factors tested here were the hydrogel elastic modulus (p = 0.008), time (p = 0.009), the interaction between the age of the fibroblast donor and the elastic modulus (p = 0.010), the interaction between the age of fibroblast donor and exposure (p = 0.011), donor age (p = 0.020), and the interaction between donor age and time (p = 0.046) (Fig. 6B).

Fig. 6. Statistical analysis results showed that fibroblast donor age, elastic modulus, exposure, and time all significantly influenced fibroblast activation gene expression.

(A) Trend lines from experimental data showed changes in COL1A1, FN1, CTGF, CTHRC1, and LTBP2 gene expression relative to the age of fibroblast donors, hydrogel elastic modulus, exposure, and time. Data are presented as mean ± SEM (n=3 biological replicates, with 4–6 technical). All gene expression data were normalized using a natural log transformation for statistical analyses. (B) The effect magnitude analysis identified all factors and interactions that were significant in influencing fibroblast activation gene expression. These results also predicted that assessment of 3D lung models containing the 60-year-old donor-derived fibroblasts embedded in stiff (18.99 kPa) hydrogels with FC exposure on day 8 would produce the highest levels of fibroblast activation for the conditions examined in these experiments.

3.5. Human 3D lung models were responsive to therapeutic drug treatment

To further evaluate 3D distal lung models, responsiveness to Nintedanib, an FDA-approved IPF treatment, was tested on co-cultured iATII cells and fibroblasts in stiff hydrogels, which represented the most fibrotic microenvironment available in this parameter space (Fig. 5A and Fig. 6A). Cells were exposed to the FC from days 2 to 6 to induce epithelial injury and fibroblast activation, followed by Nintedanib treatment during the remaining two days (Fig. 7A). On day 8, epithelial cells (EpCAM⁺) and fibroblasts were separated using MACS, and subsequent gene expression was assessed (Fig. 7A). Samples either received a combination of the FC exposure and Nintedanib treatment, or FC exposure alone, which served as a control. Nintedanib treatment led to the upregulation of transitional epithelial markers KRT17 (p = 0.0009) and CLDN4 (p = 0.0064) relative to the FC-only samples (Fig. 7B). The alveolar epithelial type I marker PDPN was also upregulated (p = 0.0062), while the aberrant epithelial marker FN1 [8, 84] was downregulated (p = 0.0013) in Nintedanib-treated samples (Fig. 7B). Meanwhile, expression levels of SFTPC, CTGF, LAMP3, and AQP5 remained relatively unchanged between conditions (Fig. 7B).

Similarly, several genes showed statistically significant differences when comparing Nintedanib-treated fibroblasts to FC-only fibroblasts. No fibroblast genes were upregulated in the Nintedanib-treated samples; however, LTBP2 (p = 0.0165), FN1 (p = 0.0009), and COL1A1 (p = 0.0019) were all downregulated (Fig. 7C). Furthermore, expression levels of CTHRC1 and CTGF remained relatively unchanged between conditions (Fig. 7C).

Discussion

Here, we present 3D human lung models designed to support iATII-fibroblast co-culture within biomaterial systems that integrate synthetic, tunable stiffness hydrogels with donor-specific cells and serum as a model for pulmonary fibrosis [6]. By controlling the microenvironmental stiffness of the embedding hydrogel and delivering pro-fibrotic biochemical cues, we assessed how these factors synergistically influenced epithelial cell gene expression and fibroblast activation. Given that nearly 70% of IPF cases occur in males [21], this study also aimed to establish a male-specific IPF model with male-derived iATIIs, fibroblasts, and serum to minimize sex as a confounding variable. Prior studies highlighted significant differences in fibroblast activation response based on the sex and age of the serum used to supplement the cell culture medium [85]. These results underscored the importance of using sex- and age-matched serum in disease modeling [85]. Expanding the availability of female iATIIs in the future and replicating this work with female cells and serum will be crucial for investigating potential sex-specific disease mechanisms and better understand the large dimorphism in this disease [86]. Likewise, the average age of onset for IPF is typically between 60 and 70 years old. It is rare in individuals under 50, and the risk increases with age [87]. The age of human serum also impacts relative hormone levels and thus cellular activation [85], so all human serum and fibroblasts used in this study were sourced from older (≥60 years old) donors to match the demographics of the most at-risk population for IPF.

Our prior work demonstrated that the inclusion of an MMP9-degradable crosslinker supported iPSC-derived lung progenitor spheroid growth, a precursor cell type to the iATIIs better than other MMP-degradable peptide sequences [63]. Additionally, studies by Loebel et al. showed that the laminin/entactin full proteins incorporated into the hydrogel formulations presented here promotes iATII growth in hydrogel micro-wells. Laminin, a key protein in the alveolar basal lamina, plays a crucial role in lung morphogenesis and supports alveolar growth [88–90]. Laminin/entactin entrapped in hyaluronic acid hydrogels successfully supported iATII growth and self-assembly into spheroids without the need for Matrigel [51]. One limitation of this work and the study presented here is that the laminin/entactin entrapped protein within the hydrogels was derived from mice. Identifying a suitable human laminin protein alternative would be more translationally relevant to ensure the model consists entirely of human-derived materials. During lung development, MMP secretion predominantly shifts from MMP2 to MMP3 to MMP9 [91, 92], which reflects dynamic ECM remodeling and cellular behavior. MMP9 has also been widely implicated in early fibrotic tissue as critical ECM regulator, where it activates latent TGF-β1 and contributes to fibrosis progression [93–95]. Thus, the MMP9-degradable peptide crosslinker, laminin peptide mimic, and laminin/entactin protein complex were strategically included within the hydrogel formulations to facilitate cellular adhesion and remodeling in 3D [47].

Building on previously published work, a key distinction of this approach was that iATIIs naturally form spheres, or alveolospheres when cultured in 3D [42, 43, 51]. By leveraging this inherent geometry, iATIIs were aggregated successfully into a larger epithelial structure without the need for microsphere templates [6, 65]. Photodegradable microsphere templates have been used to generate cyst structures mimicking a single alveolus, a model which can be used to study crosstalk between templated epithelial cells and fibroblasts but which limits the ability of epithelial cells to provide autocrine support across a larger structure that replicates features of lung geometry and micro-architecture, and also risks spontaneous and uncontrolled differentiation of primary cells [65, 66]. Self-assembling alveolospheres have also been generated in a 2.5D microwell system, where the dimensions of each well determine the ultimate spheroid size, but without full embedding in a supportive material [51].

Our 3D study allows for alveolospheres to be surrounded by tunable biomaterial and neighboring fibroblasts in a more physiologically relevant geometry. These engineered models also supported high total cell viability (>75%) throughout the 8-day culture period. While not statistically significant, a dip in cell viability to 78% ± 6.2% was observed on day 6 within the stiff hydrogels that were exposed to the fibrosis cocktail. This trend suggests that the combined effects of increased microenvironmental stiffness and pro-fibrotic biochemical cues may have contributed to higher cell death. However, viability returned to above 80% by day 8, indicating that cell clearance may have occurred since fibroblasts can act as non-professional phagocytes [6, 96–98]. This corresponds to observed effects of FC treatment on alveolospheres in Matrigel, where exposure caused a trend towards increase in cellular damage markers without overall loss in number of alveolospheres [8], indicating that FC can cause a certain amount of stress without overt toxicity across a 3D culture.

Successful culture of iATII cells in engineered hydrogels is particularly advantageous given the widespread reliance of iATII growth in Matrigel [22, 39, 42, 43, 51, 84, 99], which presents significant regulatory challenges, as no Matrigel-derived products have been approved by the FDA or deemed safe for clinical applications such as autologous cell therapies [50]. Additionally, by incorporating a dynamic stiffening mechanism with the soft hydrogels, it would enhance the model’s physiological relevance and better recapitulate disease progression. Leveraging the tyrosine residues within the laminin, peptide mimics, and degradable crosslinker could enable ruthenium-based crosslinking, as demonstrated by Nizamoglu et al. [53–55]. Future work could consist of including additional cells types, such as immune cells in the embedding hydrogel formulation to increased model complexity, or exploring the use of hybrid-hydrogels, which have gained significant traction for harnessing the advantages of both synthetic and natural hydrogels [9, 10, 100]. Enhancing the hydrogel formulation with additional ECM proteins could improve viscoelasticity, increase cell remodeling, and further refine the model for studying fibrosis [54]. This design consideration will be particularly important if the goal is to seed single iATII cells within the hydrogels and allow them to self-assemble into alveolospheres over extended culture durations, such as several weeks, which is not currently feasible in fully elastic hydrogels. Extended culture could result in the development of characteristics representative of later stages of pulmonary fibrosis, including excessive aberrant ECM deposition.

In this study, the temporal expression of several ATII, transitional epithelial, ATI, and fibroblast activation genes were monitored to assess whether gene expression profiles characteristic of early events in pulmonary fibrosis could be effectively recapitulated over eight days in culture. Treatment of alveolospheres alone in Matrigel with FC has been observed to decrease expression of the ATII marker SFTPC while increasing the transitional marker KRT17 and promoting aberrant ECM gene expression by epithelial cells, including expression of FN1 [8]. Similarly, in the co-culture model presented here, SFTPC [42, 43, 99] and LAMP3 [23, 101] served as ATII-specific markers, while KRT17 [23, 34, 37] and CLDN4 [34, 102] identified transitional epithelial states. It is important to note that robust red fluorescence from the SFTPC^tdTomato reporter was observed at the time of seeding the iATIIs, indicating high activity of the SFTPC promoter [42, 43]. PDPN [8, 84, 99] and AQP5 [22, 103] were used as ATI markers, while fibroblast activation was assessed through the expression of COL1A1 [6, 104], FN1 [105, 106], CTGF [107, 108], CTHRC1 [104, 109], and LTBP2 [110, 111]. Interestingly, CTHRC1⁺ cells have been shown to express pathologic ECM genes in fibrotic lungs and exhibit high mobility, often accumulating within fibrotic foci [104, 109, 112]. The upregulation of this gene within the FC-exposed samples suggests that early fibrosis may be occurring. Compared to the VC samples, exposure to the FC resulted in a clear and expected downregulation of ATII markers and upregulation in transitional epithelial, ATI, and fibroblast activation markers. However, AQP5 exhibited a notable deviation from this trend, potentially suggesting that iATIIs may not have been able to fully differentiate into ATI cells or underwent apoptosis. It should also be noted that many of the relative gene expression values, like COL1A1 and FN1, remained high across all conditions. We believe these findings suggest that expanding fibroblasts on tissue culture plastic activates the cells and results in lower sensitivity and changes amongst the experimental conditions once in 3D microenvironments. To eliminate this confounding variable, it could be beneficial to repeat these studies using primary fibroblasts that were never cultured on tissue culture plastic and instead, are immediately embedded in 3D.

Hydrogel elastic modulus, the interaction between fibroblast donor age and elastic modulus, and FC exposure were the top three factors to influence epithelial cell gene expression. In contrast, the fibroblast activation was most influenced by FC exposure, hydrogel elastic modulus, and time. Notably, since all the input variables (age from which the donor fibroblasts were derived, elastic modulus, exposure, and time) were statistically significant within both analyses, these results provide strong evidence that these factors should be considered and integrated into experimental designs for models of pulmonary fibrosis.

To further validate the potential of human 3D long models, responsiveness to anti-fibrotic treatment, with Nintedanib was assessed. Nintedanib inhibits the receptor kinases of PDGF, fibroblast growth factor (FGF), and vascular endothelial growth factor (VEGF) [113]. This multi-tyrosine kinase inhibition has been shown to inhibit ECM production, fibroblast proliferation, and fibroblast-to-myofibroblast differentiation [113–116]. Our findings contribute to the limited body of research that examines the specific effects of FC on alveolospheres, since the impact of these pro-fibrotic biochemical cues on lung epithelial cells remains largely unexplored [117]. Alveolospheres in Matrigel treated with FC show an acquisition of mesenchymal-type markers including FN1 that can be partially rescued by treatment with Nintendanib, but Nintendanib treatment in this model system failed to rescue SFTPC expression, indicating an incomplete reversion of epithelial cell injury [8].

In human 3D lung models, after two days of treatment, KRT17, CLDN4, and PDPN were significantly upregulated relative to the FC-only control samples. This finding suggests that Nintedanib treatment in the context of fibroblast co-culture may have supported healthy ATII-to-ATI differentiation and preserved an ATI cellular subpopulation. Additionally, FN1 expression was evaluated in EpCAM⁺ cells, as this marker has been linked to the emergence of an aberrant basaloid cell population [8, 84]. Interestingly, Nintedanib downregulated FN1 expression, which suggests that the epithelial cells were more likely to maintain normal functionality with treatment. In line with these findings, Nintedanib treatment also significantly downregulated LTBP2, FN1, and COL1A1, which are highly expressed in fibrotic tissue. These results are consistent with findings from a previous study in which lung fibroblasts were pretreated with Nintedanib prior to TGF-β stimulation, demonstrating that Nintedanib reduced the expression of fibrotic markers and phenotypes [118]. In contrast, the current study applies Nintedanib after fibrotic stimulation, a timeline that more closely reflects clinical treatment scenarios. Given that stiffness was identified as a key driver of fibrotic activation in these human 3D lung models and studies featuring hybrid-hydrogel models [9,10], future work should explore mechanosensitive drugs for potential anti-fibrotic effects.

Conclusion

In summary, we have engineered an iATII-fibroblast model of human pulmonary fibrosis that integrates key factors that contribute to fibrogenesis and evaluates respective impacts on gene expression. Given the significant gap in our understanding of the events associated with fibrotic progression, human 3D lung models could address crucial mechanistic questions through additional systematic testing. This biomaterial platform enables independent interrogation of biomechanical and biochemical cues by leveraging tunable synthetic hydrogels to reproduce important aspects of the spatial cellular arrangement and, biochemical microenvironment, and mechanical properties of native distal lung tissue. While designed to simplify the complexity of fibrosis to the most critical parameters, our results demonstrate that fibroblast donor age, hydrogel stiffness, pro-fibrotic biochemical cues, and time all significantly influenced fibrotic gene expression. Our findings also identified that the combination of stiff hydrogels and fibrosis cocktail exposure resulted in a more fibrotic cellular microenvironment than other combinations. Synergistically, these factors captured epithelial injury and fibroblast activation, which aligns these results with fibrosis initiation in vivo. Nintedanib treatment reduced fibrotic epithelial and fibroblast outcomes, highlighting that potential for 3D human lung models to be valuable in drug discovery and validation.

Statement of significance.

This study leverages advanced biomaterials and biofabrication techniques to engineer physiologically relevant, donor-specific, and sex-matched models of pulmonary fibrosis, addressing the critical need for pre-clinical therapeutic drug screening platforms. These human 3D lung models successfully replicated key features of fibrotic lung tissue. Tuning microenvironmental stiffness of 3D PEG-NB hydrogels to match fibrotic lung values and exposing human iATII cells and fibroblasts to pro-fibrotic biochemical cues recreated hallmark characteristics of in vivo fibrosis pathogenesis, including epithelial differentiation and loss, as well as fibroblast activation. The utility of these models was further validated by demonstrating responsiveness to Nintedanib, a clinically available treatment for IPF. These findings highlight the transformative potential of well-defined biomaterial-based 3D models for elucidating complex disease mechanisms and accelerating therapeutic drug discovery for chronic pulmonary diseases like idiopathic pulmonary fibrosis.

Data availability

The data that support the findings of this study are openly available in Mendeley Data at doi: 10.17632/3jchc2mkbf.2