Modelling bronchial epithelial-fibroblast cross-talk in idiopathic pulmonary fibrosis (IPF) using a human-derived in vitro air liquid interface (ALI) culture

Idiopathic Pulmonary Fibrosis (IPF) is a devastating form of respiratory disease with a life expectancy of 3–4 years. Inflammation, epithelial injury and myofibroblast proliferation have been implicated in disease initiation and, recently, epithelial-fibroblastic crosstalk has been identified as a central driver. However, the ability to interrogate this crosstalk is limited due to the absence of in vitro models that mimic physiological conditions. To investigate IPF dysregulated cross-talk, primary normal human bronchial epithelial (NHBE) cells and primary normal human lung fibroblasts (NHLF) or diseased human lung fibroblasts (DHLF) from IPF patients, were co-cultured in direct contact at the air–liquid interface (ALI). Intercellular crosstalk was assessed by comparing cellular phenotypes of co-cultures to respective monocultures, through optical, biomolecular and electrical methods. A co-culture-dependent decrease in epithelium thickness, basal cell mRNA (P63, KRT5) and an increase in transepithelial electrical resistance (TEER) was observed. This effect was significantly enhanced in DHLF co-cultures and lead to the induction of epithelial to mesenchymal transition (EMT) and increased mRNA expression of TGFβ-2, ZO-1 and DN12. When stimulated with exogenous TGFβ, NHBE and NHLF monocultures showed a significant upregulation of EMT (COL1A1, FN1, VIM, ASMA) and senescence (P21) markers, respectively. In contrast, direct NHLF/NHBE co-culture indicated a protective role of epithelial-fibroblastic cross-talk against TGFβ-induced EMT, fibroblast-to-myofibroblast transition (FMT) and inflammatory cytokine release (IL-6, IL-8, IL-13, IL-1β, TNF-α). DHLF co-cultures showed no significant phenotypic transition upon stimulation, likely due to the constitutively high expression of TGFβ isoforms prior to any exogenous stimulation. The model developed provides an alternative method to generate IPF-related bronchial epithelial phenotypes in vitro, through the direct co-culture of human lung fibroblasts with NHBEs. These findings highlight the importance of fibroblast TGFβ signaling in EMT but that monocultures give rise to differential responses compared to co-cultures, when exposed to this pro-inflammatory stimulus. This holds implications for any translation conclusions drawn from monoculture studies and is an important step in development of more biomimetic models of IPF. In summary, we believe this in vitro system to study fibroblast-epithelial crosstalk, within the context of IPF, provides a platform which will aid in the identification and validation of novel targets.

Human Lung Fibroblasts (NHLF; From healthy non-smokers; ID 85620; ID 90350; ID 92283) or Diseased Lung Fibroblasts (DHLF; From IPF patients, ID 85609; ID 95650; ID 95554) were initiated into FGM™-2 fibroblasts growth medium-2 BulletKit™ (Lonza, Switzerland).Once confluent, cells were washed with PBS × 1, harvested with TrypLE express (ThermoFisher, USA) and neutralized with neutralizing solution (HBSS containing 5% FBS; ThermoFisher, USA).Cells were seeded at a density of 3 × 10 4 cm 2 on the underside of a Transwell ® insert as a monoculture or as a co-culture (NHBEs on apical surface of insert).Once Co-culture was initiated, media was changed in all wells to co-culture media, which contained an 80:20 media ratio (NHBE media: fibroblast media).This ratio was chosen based on cell proliferation (Incucyte ®; Sartorius, Germany) and area under the curve (AUC) analysis (Figure S1).Media was replaced every 2-3 days and all cultures were grown for a further 7 Experimental setup of in vitro cell cultures showing the various configurations (B).Primary human fibroblasts (healthy or diseased) and epithelial cells were grown in monoculture (left) or co-culture (right).Optimised experimental protocol (C) where epithelial cells were seeded (day 0), air-lifted (day 7) and differentiated (days 7-28) before fibroblasts were added to the underside of the membrane inserts (day 28).After one week of fibroblast culture (either in mono-or co-culture), cell groups were stimulated with TGFβ-1 for a further week before end point assays were performed.NHBE = Normal Human Bronchial Epithelial cell; NHLF = Normal Human Lung Fibroblast; DHLF = Diseased Human Lung Fibroblasts.Schematic created with Biorender.com.

TGFβ treatment
To induce a proinflammatory environment, cell cultures were treated with recombinant human TGFβ-1 protein (240-B; R&D Systems, USA) at a concentration of 1 ng/ml, in the basal compartment of cultures, for 7 days.Media with fresh TGFβ-1 was replaced every 2-3 days.

TEER
TEER measurements were taken using the EVOM 3 and STX-2 plus electrode (World Precision Instrument, USA).Prior to recordings with cells at ALI, media was replaced on apical and basal sides of the Transwell ® insert and allowed to equilibrate for 30 min.Cells were removed from the incubator and allowed to reach room temperature.TEER values are reported as Ω cm 2 where TEER = (Raw value -blank, no cell, value) * 0.33 cm 2 (growth surface area of insert) and averaged as 3 readings per well.Values are presented as the mean ± SD for each cell group (n = 8-32).

Cell lysis, RNA extraction and cDNA synthesis
After treatment and final TEER measurements, Transwell ® inserts containing either monocultures or co-cultures were placed into a tube containing lysis buffer (Beckman Coulter©, Switzerland) and incubated at 37 °C for 25 min.RNA was extracted through multiple wash and magnetic bead binding steps, using the RNAdvance tissue kit (Beckman Coulter ©, Switzerland).RNA samples were eluted in 20 µl dH 2 O, added to reverse transcriptase mix (High-capacity RNA-to-cDNA™ kit; ThermoFisher; USA), sealed and briefly centrifuged to remove any bubbles.Samples were transferred to a 96 well thermal cycler (Eppendorf Mastercycler ® personal, Germany) and incubated at 37 °C for 60 min.The reaction was stopped by heating to 95 °C for 5 min and held at 4 °C.cDNA samples were diluted 1:3 in dH 2 O and stored at − 80 °C until needed.

Real time q-PCR
cDNA samples were thawed and transferred to a 384 well q-PCR plate (MicroAmp™ EnduraPlate™; ThermoFisher, USA).TaqMan™ probes (ThermoFisher; USA) were multiplexed with housekeeper gene and gene of interest (Table 1).Reaction mix was prepared by adding MasterMix (TaqMan™; ThermoFisher, USA), housekeeping probe and gene of interest probe to each cDNA sample using the ECHO525 acoustic handler (Beckman Coulter ©, Switzerland).Plates were briefly centrifuged to remove any bubbles then run on a Thermo QuantStudio™ 7 Flex real-time PCR system (40 cycles).Data for real time q-PCR are presented as fold change (Log(2)) from control group.The 2-(ΔΔCt) method was used for data analysis, where cycle threshold (CT) represents the PCR cycle number where the fluorescence single becomes distinguishable from background noise where: The ΔΔCT values were then log transformed to give the relative fold gene expression to untreated controls.Values are presented as the mean ± SD for each cell group (n = 6-11).

ELISA
After treatment, and prior to final TEER measurements, the supernatant from the basal side of cultures were aspirated and frozen at -80 °C until needed.Samples were then washed and prepared using the MULTI_SPOT assay system proinflammatory panel 1 human kit (MSD ®, USA) which contains detection antibodies for the inflammatory analytes listed below (Table 2).Samples were read using plate reader (MESO SECTOR S 600; MSD ®, USA) and analyte concentrations (pg/ml) calculated by fitting signals to a calibration curve using DISOVERY WORKBENCH ® version 4.0 (MSD ®, USA).Data is presented as pg/ml or as fold change compared to untreated controls.Values are presented as the mean ± SD for each cell group (n = 4-8).

Surface contrast and cell morphology analysis using ImageJ
For surface contrast quantification, images of the apical surface of cell cultures were captured using brightfield microscopy (EVOS M7000; ThermoFisher, USA), at the center of each well.Magnification (× 10) and brightness of the lens was kept consistent throughout.Images were converted to 32-bit greyscale and analyzed in ImageJ-Win64 using the auto-threshold, analyze and measure functions to give a histogram plot of grey scale value distribution.In this context, surface contrast is given as the mean grey value, as a percentage of the image area.
For each cell condition 3-4 images were averaged and presented as the mean ± SD.For cell/nucleus morphology analysis of H&E stained histology slices, images were converted to 8-bit greyscale, thresholded and analyzed using the particle plugin function.Nucleus size is defined as area in µm 2 .Nucleus roundness is defined as how closely the shape fits a circle, where a value of 1 is a perfect circle, and is calculated as the ratio of the nucleus area by the square of its parameter.Nucleus aspect ratio is defined as the ratio between the major and minor axis of the nucleus shape, where 1 is a perfect circle and infinity is completely linear.For histology 1-2 images were used and nucleus number per image was n = 75-162.For epithelial height, 50 lines per image were drawn (from the apical to basal section of the epithelium), measured and the average taken.Values were presented as the mean ± SD.

Statistical analysis
Statistical analysis was performed in GraphPad Prism (version 8.0).For pre-treatment conditions an ordinary one-way ANOVA with Tukey's multiple comparisons test was performed.For post-treatment conditions, comparing untreated to treated, multiple T tests per row with Holm-Sidak multiple comparison test was used.Data

Development of the in vitro ALI co-culture model
To generate the model, the two main functional cell types of the respiratory airways, fibroblasts and epithelial cells, were chosen.Primary NHLFs, DHLFs or NHBEs were purchased from commercial sources and grown in either a monoculture or co-culture configuration (Fig. 1B).Although the use of IPF-derived epithelial cells would better reflect the diseased environment, and would provide further insights into IPF dysregulated cross-talk, they are extremely difficult to access.Thus, this model provides an alternative method to generate IPF-related epithelial phenotypes in vitro with mixed donors.For co-cultures, NHBEs were grown on the apical side of a collagen coated Transwell ® insert at the Air Liquid Interface (ALI), with fibroblasts grown on the underside.In this way, the in vivo 3D tissue structure of the upper airway epithelium was mimicked.Following optimization of a media recipe for co-cultures (further details in methods and in Additional file 1, Figure S1), a five week protocol was established (Fig. 1C); NHBEs were cultured in ALI conditions (epithelial monocultures and co-cultures) for three weeks before fibroblasts were seeded on the underside of the Transwell insert (fibroblast monocultures and co-cultures) and treatments began.

Characterization of cellular morphology
Respiratory epithelial cells, when cultured at ALI in vitro, differentiate and give rise to multiple cell types found in vivo 27 .This includes basal stem cells, mucus-secreting goblet cells and ciliated cells.The synergistic functions of mucus and cilia aid in the capture and clearance of noxious stimuli, known as muco-ciliary clearance, and is a defining feature of a well-developed healthy airway epithelium in vivo.Changes in cell number/function, or mucus amount/viscosity are implicated in numerous lung diseases 28,29 , and muco-ciliary clearance often quantified in vitro using fluorescent tracers and/or ciliary beat assays 30,31 .However, it is also possible to observe apical topography and cilia beating in vitro, in real-time, using reductionist approaches such as brightfield microscopy and image analysis.For example, patterns may be visualized on the apical surfaces of cultures, which form as a product of mucus secretion, cilia beating and the dimension constraints of the circular culturing well.This can be seen with both epithelial monocultures (Fig. 2A) and NHLF co-cultures (Fig. 2B), where uniform apical secretion patterns were often observed.In contrast, with DHLF co-cultures, the topography of the apical surface secretions appears disordered (Fig. 2C).In order to quantify this visual phenomena, brightfield images were converted, thresholded and analyzed using ImageJ (Additional file 1, Figure S2), where surface contrast is given as a proxy for apical muco-ciliary uniformity.This is possible due to the contrast in mean grey values between apical secretions and cells, as a percentage of the whole image (Fig. 2D); For epithelial monocultures and NHLF co-cultures, surface contrast is comparable (58.8 ± 2.6% and 60.8 ± 2.3% respectively), whereas DHLF co-cultures have a significantly lower surface contrast (36.6 ± 4.2%).Detailed differences in epithelial cell morphology are visualized in Hematoxylin and eosin (H&E) and immunostained histology sections (Fig. 2E-G) and morphological changes were quantified via image segmentation analysis using ImageJ (Figure S3).NHBE monocultures displayed a characteristic pseudostratified columnar epithelial structure and expression of specific epithelial markers including Mucin5B, Mucin 5AC and E-cadherin (Fig. 2E).Interestingly, co-culture with fibroblasts significantly reduced the height of the pseudostratified epithelium, and altered the ratio of Mucin5B/5AC expression, regardless of whether fibroblasts were diseased or healthy (Fig. 2E-K).DHLF co-cultures (Fig. 2H) displayed a significant change in epithelial cell morphology that resembled a squamous phenotype (Fig. 2G), including a reduction in nucleus size, roundness and an increase in nucleus aspect ratio (Fig. 2I-K).

Characterization of barrier properties and gene expression
Transepithelial Electrical Resistance (TEER) is a commonly used method to non-invasively measure in vitro barrier systems.TEER is dependent on the resistance to ion flow through paracellular pathways, including cadherin, adherens and tight-junction proteins, where a higher resistance indicates a stronger barrier 32 .TEER measurements of cultures at day 35 are shown in Fig. 3A; Of note, is that fibroblast monocultures display minimal TEER values, as their physiological role is not of a barrier cell origin 33 .NHBE monocultures display a TEER value of 694.4 ± 90 Ω cm 2 which is in line with previous reported values of between 600 and 1200 Ω cm 2 (depending on donor variability) 27,34,35 .However, when grown in co-culture, fibroblasts had a synergistic effect on TEER, reaching significantly higher values than NHBE monoculture in both NHLF (766.9 ± 71.9; **, P = 0.003) and DHLF (914.3 ± 120.4; ****, P = < 0.001) co-cultures.Interestingly, DHLF co-cultures displayed a small, but significant, increase in TEER value compared to NHLF co-cultures (****, P = < 0.001).Similarly, mRNA levels of proteins which regulate paracellular transport, namely epithelial Cadherin (CDH1) and the tight junction protein, ZO-1, increase in a culture condition-dependent manner and reach significance in DHLF co-cultures (Fig. 3B; *, P = 0.026; **, P = 0.0061).Ciliated cell mRNA levels (DN12 and FOXJ1) also showed a co-culture dependent increase, reaching significance in DHLF co-cultures, with over a threefold increase compared to NHBE monocultures (****, P = < 0.001).Mucin mRNA (Fig. 3B) and protein levels (Additional file 1, Figure S4) did not significantly differ between any groups.However, a co-culture dependent reduction in basal cell mRNA levels (P63 and KRT5) were observed, reaching significance in KRT5 expression for both healthy (**, P = 0.0062) and diseased (***, P = 0.009) co-cultures.Next, to assess the effect of epithelial co-culture on fibroblast phenotypes, gene expression is presented as normalized to respective fibroblast monocultures (Fig. 3C).A significant downregulation in mRNA levels of vimentin (****, P = < 0.001; **, P = 0.0036) and α-smooth muscle actin (α-SMA; *, P = 0.033), were observed in both NHLF and DHLF co-cultures.Additionally, a co-culture dependent increase in Overall, the above data demonstrated a significant effect of fibroblast co-culture on epithelial cell phenotypes.This included a reduction in mRNA levels of basal cell markers, epithelium thickness and an increase in barrier integrity (TEER, tight junction and cadherin protein expression).When cultured with healthy fibroblasts, no significant changes in apical topology, cell nucleus morphology or cell density was observed compared to NHBE monoculture.However, when cultured with IPF fibroblasts, highly significant changes were seen in barrier function, cilia mRNA expression and epithelial cell morphology.This included significant alternations in apical topology, epithelium thickness, cell nucleus size/shape and cell density which indicated transition to a squamous phenotype.DHLF co-cultures showed a significant upregulation of TGFβ-2 mRNA levels compared to any other group.Taken together, this data may implicate the role of fibroblast-induced barrier alterations and markers of EMT in the IPF phenotype.

Platform validation with a proinflammatory stimulus
The transforming growth factor beta (TGFβ) family (isoforms 1-3) are potent regulators of fibrotic processes in mammals and is the main cytokine family upregulated in lung fibrosis and in IPF patient biopsies [24][25][26] .TGFβ-1 is chosen specifically as a proinflammatory stimulus due to it being the most physiologically abundant and ubiquitously expressed isoform, is readily available and is a well authenticated as a stimulus to induce fibrosis in vitro.However, the effects of TGFβ-1 stimulation, have mostly been characterized in monocultures of fibroblasts or epithelial cells and few studies have looked to evaluate the effect in co-cultures on epithelial barrier properties 7,8,[36][37][38][39] .Thus, we looked to validate and directly compare, the effect of TGFβ-1 stimulation on cocultures of healthy epithelial cells grown with either healthy or IPF-derived fibroblasts.Details of primers and probes used for real time q-PCR can be found in Table 1 of the methods section.

Discussion
With respect to clinical relevance and translatability, the in vitro model used here mimics the in vivo conditions of the airway epithelial-stromal layer more accurately than routine monocultures of fibroblasts or bronchial epithelial cells.Previously, the effect of direct epithelial-fibroblastic co-culture on the structure, function and barrier integrity of bronchial epithelial cell in the context of IPF models, has been poorly described.Conventionally co-cultures grown fibroblasts indirectly on culture plastic, and their temporary effects on epithelial cells inserted above on hanging inserts are measured.However, our model addresses a significant gap in the in vitro literature, where the effects of direct contact/cross-talk may be investigated.Although the use of IPF-derived epithelial cells would better reflect the diseased environment, and would provide further insights into IPF dysregulated cross-talk, they are extremely difficult to access.Thus, this model provides an alternative method to generate IPF-related epithelial phenotype with mixed donors and is an is an important first step in generating a more biomimetic model.
The model was successively used to study interactions between bronchial epithelial cells from healthy donors and fibroblasts from either healthy or IPF donors.Contrary to what was expected, the effect of co-culture decreased epithelial thickness and stratification, irrespective of whether fibroblasts were from healthy of IPF donors.Additionally, immunostaining revealed that the ratio of MUC5AC to MUC5B expression was altered in co-cultures, skewed in the direction of MUC5B.A reduction in epithelial height, thickness and stratification is a phenotype often observed in lung diseases such as COPD and cigarette-smoke epithelial injury 17,22,23 .Furthermore, overexpression and genetic polymorphisms of MUC5B are known risk factors associated with IPF 49,50 .In DHLF co-cultures, a reduction in epithelial height was the greatest and EMT-related morphological changes (nucleus elongation and size reduction) reached significance compared to any other group.This phenotype is a likely consequence of DHLF-induced EMT, as reported previously 30,31 .For the phenotype observed with NHLF co-cultures, a possible explanation is that fibroblasts induce the de-differentiation of epithelial tissue and that NHLF co-cultures actually represent an intermediate EMT phenotype.Another possibility may lie within how cells are routinely cultured on cell culture plastic, prior to seeding.It has been shown previously, for example, that culturing fibroblast on hard surfaces 45 or at a high densities 6 induce phenotypic transitions towards FMT and IPF-related cell populations.Thus, whether healthy fibroblasts retain their true in vivo phenotype when cultured over time, remains a controversial topic.To minimize this effect however, all cells were seeded onto collagen coated membrane at no higher than passage one and epithelial cells were differentiated at ALI for 28 days prior to seeding fibroblasts.Additionally, three healthy and three IPF donors were used which increases our level of confidence that any differences observed are due to differences in disease state, rather than simply donor to donor variability.Techniques such as single cell RNA sequencing should be utilized in future studies to elucidate the causal mechanisms behind this co-culture dependent alteration in epithelial phenotype.
Another characteristic feature of other lung diseases, such as asthma 20,21 and COPD 22,23 , is epithelial barrier dysfunction.However, the pathogenic implications of barrier (dys)function in the progression of IPF is ill defined.TEER is a commonly used method to determine the quality of differentiated epithelial airway development and barrier integrity 32 .Contrary to what one might expect, given the lower stratification of epithelial layers, our study demonstrated that fibroblast co-culture had a synergistic effect on TEER, reaching the highest values when cultured with DHLFs.As TEER is dependent on the resistance to ion flow through paracellular pathways, the co-culture dependent increase in TEER may be influenced by the expression of proteins that regulate paracellular barrier permeability.Supportive of this hypothesis, it the increase in mRNA levels of the tight junction proteins CDH1 and ZO-1 with co-culture.Previously, reported data also demonstrate an increased ZO-1 expression in ex-vivo IPF bronchiolar biopsies 51 and in vitro bronchial co-culture models 37,52,53 .Additionally, a decrease in epithelial height/stratification corresponds to an increase in epithelial tissue density, when co-cultured with fibroblasts.Thus, an increase in tissue density may correlate to the increase in resistance (TEER), a relationship previously described 54 .
Functional muco-ciliary clearance mechanisms are another defining feature of a well-developed healthy airway epithelium with changes in cell number/function, or mucus amount/viscosity implicated in numerous lung diseases 28,29 .Qualitatively, the synergistic functions of mucus secretion, cilia beats and the circular dimensions of the culture surface gave rise to apical surface patterns which could be observed clearly using brightfield microscopy.Epithelial cell surface topography was uniform and consistent between NHBE monocultures and NHLF cocultures.However, when grown with DHLF the apical topography was disrupted and heterogenous.Using image processing, this visual phenomena was quantified using surface contrast and given as a proxy for muco-ciliary uniformity.Together with H&E stains, taken of the same samples, a significant reduction in epithelial muco-ciliary uniformity and cilia projections were observed in the DHLF co-culture condition.Additionally, over a threefold increase in ciliated cell mRNA markers were apparent, while mucin mRNA and protein levels did not differ between any groups.Changes in cilia, rather than mucin function, therefore, may contribute to the aberrant surface topology observed.It is worth noting however that cilia function is determined here through imaging processing of fixed samples, brightfield captures and mRNA analysis.Complementary functional assays may be used in future studies to assess if changes in gene/protein expression correlate to a cell functionality e.g.cilia beat/motility assays or mucin viscosity.Interestingly, basal cell markers were downregulated in cocultures, irrespective of fibroblast origin.As stem cells of the lung epithelium, basal cells may be associated with normal wound healing processes, proliferation, injury or EMT 28 .Future studies could look to explore specific downstream pathways and other basal cell markers e.g.KRT14, SOX2, TP63 to shed light on any pathogenic implications.In terms of fibroblast specific markers, vimentin and α-SMA mRNA levels were downregulated.Additionally, mRNA levels of TGFβ-2, were significantly upregulated.Although vimentin and α-SMA are markers of FMT 52 , and TGFβ-2 plays a pathological role in lung fibrosis 33,55,56 , analysis at the protein and functional level should be conducted in future studies, to define the implications of these changes.Nonetheless, the significant quantitative and qualitive differences observed between each co-culture condition, with respect to epithelial and fibroblast monocultures, highlights the utility of this platform to be used for interrogating the effects of direct epithelial-fibroblastic crosstalk in the development of IPF phenotypes.We hope this platform can therefore be used for identifying and validating novel targets, that would otherwise be missed in routine monoculture or indirect co-culture studies.
Next, we looked to validate and directly compare, the effect of TGFβ-1 stimulation on co-cultures.TGFβ-1 is the most widely authenticated stimulus used to induce fibrosis in in vitro fibroblast and epithelial monocultures, however, few studies have looked to evaluate if direct cross-talk between the cell populations influence the TGFβ-1 response.Additionally, the majority of studies assess the effect of TGFβ-1 by means of cytokine, protein and gene expression but the effect on epithelial barrier properties in co-cultures is poorly described 7,8,33,[37][38][39] .In this study, imaging techniques revealed a significant disruption of epithelial muco-ciliary uniformity in TGFβ-1 treated NHBE monocultures and NHLF co-cultures, that was not apparent in DHLF co-cultures.Similarly, TGFβ-1 induced the formation of fibrotic-like foci in NHLF co-cultures and monocultures that was not observed in DHLF co-cultures.This suggests that exogenous TGFβ-1 may be producing less of a stimulatory effect on observational phenotypes in IPF fibroblasts, potentially due to the enhanced pro-inflammatory environment intrinsically produced by diseased fibroblasts.This is supported by the significantly higher levels of TGFβ-1 and -2 mRNA in untreated DHLF compared to NHLF monocultures (Additional file 1, Figure S6) and, when stimulated, release of all cytokines tested are significantly lower in DHLF (Additional file 1, Figure S6).Additionally, as shown in Fig. 3C, levels of TGFβ isoforms are higher in DHLF than NHLF co-cultures prior to stimulation.
TGFβ-1 significantly increased TEER values in NHBE monocultures and NHLF co-cultures but not DHLF co-cultures.Although in vitro models of other lung disease, such as asthma 20,21 and COPD 22,23 , report reductions in epithelial integrity compared to healthy controls, actual values of TEER are wildly variable.In addition, the majority of studies fail to match pre-post treatment TEER values, meaning that baseline well-to-well and donor-to-donor variability is not accounted for.In a primary in vitro model of asthma, which directly compared the effect of TGFβ-1 stimulation on co-cultures of bronchial epithelial and fibroblast cells, a protective effect of fibroblasts against TGFβ-1-induced EMT and barrier integrity was observed 52 .Additionally, healthy and DHLF co-cultures gave differential pro-fibrotic responses in response to TGFβ-1 52 .However, few studies have looked to characterize the pathogenic implications of epithelial barrier (dys)function specifically in the progression of IPF.This is partly due to the traditional use of non-barrier forming cells such as fibroblast or alveoli (type 2) cells and/or the inability to source IPF derived epithelial cells.In the context of this study, which specifically looks at the effect of IPF fibroblasts on NHBEs phenotypes, TEER values not defined in literature.Future work, beyond the scope of this study, should look to source IPF derived epithelial cells (if possible).In this way, the direct coculture of IPF epithelial and fibroblast cells can then be compared to the in vitro IPF epithelial monocultures 9,45,57 to fully elucidate the cell-and disease-specific contribution to epithelial barrier resistance.
In terms of epithelial thickness, stratification, apical topology and nucleus shape, H&E images revealed substantial disruptions in TGFβ-1 treated NHBE monocultures.Additionally, mRNA markers of EMT were highly and significantly upregulated, which is in line with previous reports of TGFβ-1 treated bronchial epithelial cell lines and primary cells 7,8,36 .In contrast, epithelial cell morphology and mRNA levels of EMT markers, in NHLF co-cultures, remained analogous to pretreatment conditions.This may suggest a protective role of healthy fibroblast against TGFβ-1-induced EMT, which is a finding previously reported in a primary co-culture model of the epithelial-mesenchymal trophic unit 52 .Furthermore, quantification of TGFβ-1-induced pro-inflammatory cytokine release conferred a protective effect of NHLF co-cultures against TGFβ-1-induced inflammation; When stimulated and compared to either monoculture alone, NHLF cocultures released significantly less IL-6, IL-8, IL-13, IL1-β and TNF-α.Additionally, the significantly enhanced TGFβ-1 induced inflammatory response of healthy vs diseased fibroblasts monocultures (Additional file 1, Figure S7) was nullified when co-cultured with NHBEs.However it is also worth considering the length and administration of treatment.For example, since TGFβ-1 is given basally, stimulation is limited by diffusional distance and thickness of cell layers.In the case of co-cultures, where there is an additional layer of cells, extra time and/or a higher concentration of treatment may be needed in order to exert the same effect as in monocultures.It may be that here, NHLF co-cultures present an intermediate phenotype, prior the EMT.Indeed, there were some significant alterations in mRNA expression of club cell and goblet cell markers which may indicate the initiation of a fibrotic phenotype 58 .Future studies should investigate longer treatment times and any time-dependent changes in transitional markers.There was no significant effect of TGFβ-1 stimulation on phenotypic transitions in DHLF monocultures or co-cultures, compared to their respective untreated condition.However, DHLF co-cultures displayed an agonistic effect on TGFβ-1-induced inflammation, with increases in IL-6, IL-8 , IL-1β.IL-1β TNF-α compared to DHLF or NHBE monoculture.Interestingly, NHLF monocultures showed a treatment induced upregulation of senescent mRNA markers.Although cellular senescent pathways have been implicated in IPF pathology and in fibroblasts isolated from IPF patients 59,60 , other studies report TGFβ-induced FMT alterations in fibroblast monocultures [7][8][9] .One explanation for the lack of reproducibility between experimental studies is changes to environmental conditions or configurations.As aforementioned, whether healthy fibroblasts retain their true in vivo phenotype when cultured over time, also remains a controversial topic.This highlights the need for models which better reflect the native tissue ECM environment, which we hope this study will provide a starting point for.Nonetheless, TGFβ-1 stimulation gave rise to differential responses between each culture condition and holds implications for any translation conclusions drawn from studies were cells are grown in monoculture or in indirect contact.
In summary, analysis at the gene and secreted protein level may indicate a protective effect of healthy fibroblasts and NHBEs when they were co-cultured.TGFβ-induced EMT, FMT and inflammatory responses were significantly blunted compared to either cell type grown as a monoculture.In contrast, DHLF co-cultures demonstrated an enhanced inflammatory response without any significant phenotypic transitions.To elucidate casual mechanisms of co-culture however, future studies may wish to evaluate protein levels and/or single cell RNA sequencing.The differential responses between culture groups highlight the importance of epithelial-fibroblastic cross-talk in the initiation of inflammatory pathways and should be taken into consideration when developing in vitro models of IPF, especially for therapeutic and drug discovery applications.Based on the overall results reported, co-culture with either NHLFs or DHLFs may both be useful models in generating diseased epithelial phenotypes while overcoming the scarcity in obtaining primary IPF epithelial cells.
Next steps, beyond the scope of this initial study, should look to build upon the complexity of the model by including additional cell types implicated in IPF pathogenesis.For example, it is known that both innate and regulatory immune cells play an important role in IPF progression [46][47][48] .Thus, future studies could look to improve the clinical relevance of the model by including direct and/or indirect immune cells cross-talk.Additionally, the incorporation of small airway or alveoli cells e.g.Human Type II Alveolar Epithelial Cells (PAEpiC2) could be compared and contrasted to the bronchial epithelial model generated here.Indeed, Type II Alveolar Epithelial Cell injury and bronchiolization of the alveoli epithelium are known to occur in the initial stages of IPF development [43][44][45] .Thus, comparisons of co-cultures generated with anatomically distinct epithelial cell populations (bronchial or type II alveolar epithelial cells) would provide a more comprehensive insight into the involvement of fibroblasts in the development of IPF.Finally, since the pathogenesis of IPF is complex and mechanisms extend beyond EMT/inflammation 1,10,47,61 , future studies may wish to stimulate cultures with a range of compounds e.g.viruses, particles or chemicals to comprehensively evaluate the model for their desired research question.

Conclusions
The majority of previous in vitro IPF studies have looked to characterize cellular changes involved in disease progression using reductionist models such as monocultures of fibroblasts or epithelial cells.Although this step is essential in defining the unique role each cell plays, the effect of the physiological (in vivo) environment e.g.multiple cell types, 3D structure is nullified.We have shown that the effect of co-culturing epithelial cells in a quasi-3D model, with healthy or diseased fibroblasts, had a significant effect on barrier properties and gene expression of markers associated with a healthy epithelial phenotype.Additionally, both co-culture and monoculture groups gave differential responses to TGFβ-1, a well authenticated pro-inflammatory stimulus used model IPF in vitro.This finding is important to consider when interpreting monoculture studies for therapeutic discovery and is an important step in development of more biomimetic models of IPF.Future studies could look to improve the biomimicry of the model further with the use of additional cell types e.g.immune cells, alveolar type II cells, IPF-derived HBEs in response to inflammatory stimuli.Culturing cells in a more physiological 3D and biomechanical environment, such as hydrogels, scaffolds or microfluidics will also facilitate developmental processes and response to drug challenge [11][12][13][14] .Additionally, the development of automated in-line sensors to monitor the development and progression of disease phenotypes will greatly improve our understanding of, and ability to interrogate, these 3D systems.Indeed, future work aims to adapt our previously published electronic transmembrane device 62 for lung-on-chip applications.

Figure 1 .
Figure 1.Schematic representation of the human respiratory environment in health and fibrotic disease (A).Experimental setup of in vitro cell cultures showing the various configurations (B).Primary human fibroblasts (healthy or diseased) and epithelial cells were grown in monoculture (left) or co-culture (right).Optimised experimental protocol (C) where epithelial cells were seeded (day 0), air-lifted (day 7) and differentiated (days 7-28) before fibroblasts were added to the underside of the membrane inserts (day 28).After one week of fibroblast culture (either in mono-or co-culture), cell groups were stimulated with TGFβ-1 for a further week before end point assays were performed.NHBE = Normal Human Bronchial Epithelial cell; NHLF = Normal Human Lung Fibroblast; DHLF = Diseased Human Lung Fibroblasts.Schematic created with Biorender.com.

Table 1 .
List of TaqMan™ primers and probes used for real time q-PCR.

Table 2 .
List of analytes and typical detection limits used for ELISA.*Taken from MSD ® MULTI_SPOT assay system proinflammatory panel 1 human kit manual.