Caco-2 cells (ECACC, Porton Down, UK) and neonatal human dermal fibroblasts (HDFn, Thermo Fisher Scientific, Loughborough, UK) were maintained in complete DMEM supplemented with 10% foetal bovine serum (FBS), 1% penicillin/streptomycin, 2 mM L-glutamine, and 1% non-essential amino acids (NEAA – all Thermo Fisher Scientific). Cultures were maintained at 37°C, 5% CO₂ in a humidified incubator and passaged following manufacturer guidelines using 0.25% Trypsin EDTA (Thermo Fisher Scientific UK). THP1-Blue™ NF-κB Cells (Invivogen, UK), hereafter named THP-1-NF-κB, were maintained in Roswell Park Memorial Institute (RPMI) 1640 media (Thermo Fisher Scientific) supplemented with 10% heat-inactivated FBS, 1% penicillin/streptomycin, 2 mM L-glutamine, 1% NEAA in a humidified incubator at 37°C, 5% CO₂. THP-1-NF-κB were derived from THP-1 monocyte cell line by stable integration of an NF-κB inducible secreted alkaline phosphatase (SEAP) reporter construct that allows monitoring of NF-κB activation by determining SEAP activity. Maintenance medium was supplemented with 10 µg/mL blasticidin (Invivogen) every alternate passage to maintain selection pressure. For THP-1-NF-κB cell differentiation, cells were seeded into T175 flasks (Sarstedt, Nümbrecht, Germany) at 0.8x10⁶ cells/mL and incubated for 48 hours in RPMI-1640 supplemented with 100 ng/ml phorbol 12-myristate 13-acetate (PMA, Thermo Fisher Scientific). Differentiated macrophages were detached using 1X TrypLE™ Express Enzyme (Sigma Aldrich).

In vitro IBD mucosal constructs were generated using Alvetex^® Scaffold inserts (Reprocell Europe Ltd, UK). Scaffolds were prepared following manufacturer instructions before cell seeding. IBD mucosal constructs were adapted from our previously published intestinal models by incorporation of a viable immune component (26). Firstly, a fibroblast tissue layer was generated within the Alvetex^® Scaffold. For this, 0.5x10⁶ HDFn cells were seeded onto inserts and cultured for 12 days in complete DMEM supplemented with 5 ng/mL TGF-β1 (Peprotech, London, UK) and 100 μg/mL ascorbic acid (Sigma Aldrich, UK); medium was replaced every 3-4 days. On day 12, immune-competent lamina propria compartments were established through addition of 0.5x10⁶ pre-differentiated THP-1-NF-κB macrophages onto the fibroblast tissue equivalent and cultured in complete DMEM for a further 2 days. Caco-2 cells were then added to the 3D cultures on day 14 at a density of 0.4x10⁶ cells/insert and were cultured in complete DMEM for an additional 21 days, during which the medium was replaced every 3-4 days.

To stimulate an inflammatory microenvironment, IBD mucosal constructs were stimulated with 1 μg/mL lipopolysaccharide (LPS), derived from Escherichia coli (E. coli), serotype O55:B5 (Sigma Aldrich) in complete DMEM for 24-96 hours. LPS supplemented DMEM was changed every 24 hours. For cytokine stimulation of models, 10 ng/mL recombinant human TNF-α (R&D systems, Minneapolis, USA), or 10 IU/mL recombinant human IFN- γ (Peprotech) was added to complete DMEM every 24 hours for 96 hours.

To determine NF- κB activity in THP-1-NF-κB macrophages within IBD models, QUANTI-Blue™ (QB - Invivogen) assays were performed. Upon stimulation with LPS, NF-κB activation results in secreted embryonic alkaline phosphatase (SEAP) release into culture medium. QB solution was prepared as per manufacturer’s instructions and 180 μL was added per well of a flat-bottom 96-well plate (Sarstedt). 20 μL of culture medium was taken from models post-stimulation and added to each well of the 96-well plate and incubated at 37°C for 10 minutes. SEAP levels were subsequently determined using a spectrophotometer at 655 nm.

Post-stimulation cell culture supernatants were collected and analysed in triplicate via bead-based multiplex technology (Eve Technology, Calgary, AB). Inflammatory cytokines and chemokines were quantified by Human Cytokine/Chemokine 48-Plex Discovery Assay^® array, and MMPs/TIMPs were quantified by Human MMP and TIMP Discovery Assay^® array.

Epithelial integrity was determined by measuring transepithelial electrical resistance (TEER) with an Epithelial Volt Ohm Meter (EVOM2) equipped with STX2 chopstick electrodes. Final TEER values, represented as Ω*cm², were determined by subtraction of the blank insert TEER value and multiplying by cell-surface area of the Alvetex^® scaffold (Equation 1):

TEER values are represented as percentage normalised to unstimulated control values.

Paracellular permeability assays were performed to assess the permeation of FITC-labelled dextran-20 (FD20) across the epithelial layer of the IBD models in the apical-basal direction. Transport assays were conducted in Hanks Buffered Saline Solution (HBSS) (Sigma Aldrich). Cell culture medium was removed from models before washing twice in PBS. Models were then placed in 12-well plates and supplemented with 1.5 mL HBSS basally and 0.4 mL apically with HBSS containing FD20. Cultures were incubated for 1 hour at 37°C. Samples were then taken from the basal compartment and values determined using a plate reader. Apparent permeability coefficient was calculated as follows:

VR is the volume of the receiver compartment; dCR/dt is the change in concentration of the analyte over time; A is the transport interface surface area and CD0 is the analyte concentration of the donor compartment at the start of the experiment.

IBD mucosal constructs were PBS washed and fixed in 4% paraformaldehyde (Thermofisher Scientific) for 2 hours. Samples were dehydrated in a graded ethanol series and incubated in Histoclear II (National Diagnostics, United States), before they were embedded in paraffin wax (Thermofisher scientific) using embedding moulds (CellPath, Newton, UK). 5 μm transverse sections were mounted for haematoxylin & eosin (H&E) staining. For this, slides were deparaffinised in Histoclear II prior to gradual rehydration in ethanol. Slides were incubated in Mayer’s haematoxylin (Sigma-Aldrich) for 5 minutes, washed in deionised water then incubated in alkaline ethanol to blue the nuclei. Slides were then dehydrated to 95% ethanol, counterstained with eosin (Sigma Aldrich) before being dehydrated with dry ethanol and mounted using Omni-mount (National diagnostics). All histological samples were imaged using a Leica microscope.

For immunofluorescence analysis, paraffin-embedded sections were first deparaffinised in Histoclear II and then rehydrated through an ethanol series. Heat-mediated antigen retrieval was achieved by incubation in citrate buffer (Sigma Aldrich) for 20 minutes. Samples were then permeabilised in 0.4% Triton X-100 (Sigma Aldrich) and blocked in 20% normal calf serum (NCS - Thermofisher Scientific) for 1 hour. Primary antibodies were diluted in blocking buffer and applied overnight at 4°C ( Supplementary Table 1 ). Following this, samples were washed 3 times in PBS before incubation in Alexa Fluor-conjugated secondary antibodies and Hoechst nuclei stain (Thermofisher Scientific) diluted 1:1000 in blocking buffer before mounting in Vectashield (Vector Labs, Peterborough, UK). Images were acquired using Zeiss 880 confocal microscope.

To calculate proliferation, paraffin-embedded sections were immunolabelled with Ki67 and counterstained with Hoechst (1:1000). To detect apoptotic cells, terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL assays (Promega) were performed on paraffin-embedded sections, and counterstained with Hoechst (1:10,000). 5 non-overlapping images were acquired per model to calculate mean percentage per model. Labelled sections were imaged using a Zeiss 880 confocal microscope. Ki67/TUNEL⁺ cells were counted and calculated as a percentage of total epithelial cells per image.

For TEM analysis, samples were fixed in Karnovsky’s Fixative consisting of 8% PFA, 25% glutaraldehyde (Agar Scientific, Stansted, UK), and 0.2 M cacodylate buffer (Agar Scientific) for 1 hour before washing in 0.1 M cacodylate buffer. Samples were further fixed in 1% osmium tetroxide (Agar Scientific) for 1 hour, washed in 0.1 M cacodylate buffer before being dehydrated through an ethanol series before embedding in Epon resin (Agar Scientific) containing dodecenylsuccinic anhydride (DDSA), methyl nadic anhydride (MNA) and benzyldimethylamine (BDMA). Ultra-thin (70 nm) sections were obtained using a diamond knife (Agar Scientific) and loaded onto formvar-coated slot grids (Agar Scientific). Grids were then stained with uranyl acetate (Agar Scientific) and lead citrate (Sigma Aldrich) prior to imaging using a Hitachi H7600 electron microscope.

IBD mucosal constructs were washed in PBS before being lysed in ice-cold Mammalian Protein Extraction Reagent (MPER – Thermofisher Scientific) containing 1% HALT™ protease and phosphatase inhibitor cocktail (Thermofisher Scientific). Lysates were homogenised and pelleted for 20 minutes at 12,000 rpm before subsequent harvesting of protein supernatant. To quantify the protein content of the supernatant, Bradford assays (Bio-Rad, Oxford, UK) were performed following manufacturer instructions. 30 μg of sample protein, diluted 4:1 with 90% Laemmli buffer (Bio-Rad): 10% 2-meraptoethanol (Sigma Aldrich) was heated to 95°C for 5 minutes before loading into SDS-PAGE gels, electrophoresed and transferred onto a 0.45 μM nitrocellulose membrane (GE Healthcare Life Sciences, Buckinghamshire, UK) via wet-transfer. Membranes were blocked in 5% milk in Tris-buffered saline (TBS-T) before incubation in primary antibody diluted in 5% milk-TBS-T overnight at 4°C ( Supplementary Table 1 ). Membranes were washed in TBS-T before incubation in appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies (Sigma Aldrich) diluted 1:10,000 in 5% milk-TBS-T for 1 hour at room temperature. Clarity Enhanced Chemiluminescence (ECL, Bio-Rad) was used to detect antibody binding.

RNA was extracted from IBD mucosal constructs using Reliaprep™ RNA Tissue Miniprep System (Promega, Southampton, UK) following manufacturer instructions. 1 μg of complimentary-DNA was generated using High Capacity cDNA Reverse Transcription (RT) kit (Applied Biosystems, Thermofisher Scientific). RT-qPCR detection was performed based on SsoAdvanced SYBR green (Bio-Rad) using a CFX Connect Real-Time PCR System. Assays were run in triplicate. Quantification was performed using ΔΔCt method to obtain fold change of the desired gene in stimulated samples relative to unstimulated controls, and signal was normalised to housekeeping gene GAPDH. KiCqStart Primers (Sigma Aldrich) were used and are listed in Supplementary Table 2 .

Fixed healthy and diseased human intestinal tissues were obtained by Reprocell (Glasgow, United Kingdom) following appropriate ethical and consent protocols, ensuring compliance with local laws and regulations. Tissues were received at Durham University under a formal MTA agreement and were processed following relevant UK HTA rules and guidelines at the time of publication.

Statistical analysis was performed on data from 3 independent biological experimental repeats unless otherwise stated in the figure legend. All analyses were carried out using Graphpad Prism 9 and data are expressed as mean ± SEM. Normality of data was tested using Shapiro-Wilk test. All data containing multiple groups were analysed using a one-way ANOVA with Tukey’s post-hoc multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 were accepted as significant.

To effectively model IBD in vitro, a physiologically relevant microenvironment must be recapitulated. To achieve this, we used Alvetex^® scaffold technology – an inert, porous polystyrene scaffold - to establish three-dimensional (3D) co-cultures representative of the IBD mucosa. A lamina propria-like compartment was created through co-culture of fibroblasts and immune cells within and on the surface of the scaffold, supporting the formation of an overlying epithelium, as outlined in Figure 1A . Macrophage-like cells were included as they are a predominant cell type involved in both innate and adaptive mucosal immunology. Histological analysis of IBD mucosal constructs revealed a confluent 3D fibroblast layer populated with macrophages, subjacent to a columnar epithelium ( Figure 1B ). Electron micrographs confirmed polarisation of the epithelium containing basal nuclei ( Figure 1C ). A confluent microvilli brush-border covered the epithelium with visible rootlets projecting below the cell membrane. Moreover, apical-junction complexes consisting of tight junctions, adherens junctions and desmosomes formed between adjacent epithelial cells. Electron dense layers at the epithelial-stromal show clear tri-lamina basement membrane formation and vesicular crosstalk between the tissue compartments. Within the lamina propria, fibroblasts produce a dense ECM comprising longitudinal (lc) and transverse collagen (tc) fibres.

Immunofluorescence characterisation of the lamina propria compartment ( Figure 1D ) revealed a heterogeneous population of vimentin⁺ and αSMA⁺ fibroblasts, reflective of a fibrotic IBD myofibroblast-like phenotype. CD68⁺ macrophages also populated the protein-rich ECM compartment. Abundant endogenous ECM was secreted by resident fibroblasts, creating an instructive tissue-like microenvironment in which cells reside and migrate. This ECM network contained essential intestinal ECM proteins including collagen-I, -III, IV and fibronectin. Epithelial proteins, including differentiation marker villin, junctional proteins including E-cadherin, claudin-1, and ZO1 as well as functional polarisation marker Na⁺K⁺ATPase ( Figure 1E ) demonstrate the presence of essential barrier components. In addition, localised collagen-IV staining indicated the presence of a basement membrane (BM).

To evaluate the potential of the IBD mucosal constructs to be used as an in vitro tool for disease research, we aimed to induce an inflammatory phenotype that mimicked hallmarks of disease. Models were stimulated daily with LPS for 24-96h, and histological findings were compared to healthy and diseased human tissue ( Figure 2A ). Common histopathological changes were observed between healthy and diseased human tissue. Similarly, we observed evidence of IBD histopathology in inflamed IBD models including time-dependent epithelial disruption and loss of epithelial polarisation and organisation. Moreover, epithelial breaches became apparent following 48-hours of stimulation (black arrows) which resulted in areas of cellular invasion into the lamina propria compartment similar to the epithelial distortion observed at atrophic crypts in the diseased tissue. Areas of epithelial pseudostratification were also identified following stimulation, matching observations in UC tissue (27). Incorporation of macrophages recapitulated the cellular abnormality of diseased tissue following application of a proinflammatory stimulus.

The NF-κB transcription pathway is largely upregulated in IBD and is a major regulatory component responsible for many inflammatory responses (28). NF-κB activation was significantly higher in macrophages within stimulated IBD models than unstimulated controls ( Figure 2B ). Increased expression of CD68 and CD86 were indicative of M1 macrophage polarisation in stimulated models ( Figure 2C ). Inflammatory markers associated with IBD were then investigated and compared to IBD tissue ( Figure 2D ). Comparable CD14⁺ macrophage staining was observed in diseased tissue and stimulated models. Inflamed IBD models also exhibited similar expression of nitric-oxide synthase (iNOS) to diseased tissue. iNOS expression was localised to the epithelial layer and infiltrating leukocytes in both CD, UC and inflamed IBD mucosal models. iNOS⁺ fibroblasts were also evident within IBD mucosal models. Cyclooxygenase-2 (COX-2) is a pro-inflammatory enzyme upregulated in IBD tissue (29). The expression of COX-2 was particularly evident in epithelial and immune cells in IBD tissue and inflamed IBD models. Western blot analysis revealed a time-dependent increase in COX-2 in stimulated models ( Figure 2E ), suggesting induction of chronic inflammation.

Next, we investigated the cytokine secretome of IBD mucosal constructs following LPS stimulation ( Figure 3 ). When grouped according to inflammatory signature into CD-typical TH1/TH17, UC-atypical TH2 ( Figure 3B ), and cytokines common to both pathologies ( Figure 3A ), it was clear that the stimulated secretome exhibited features of both UC and CD, though IL-5 expression was negligible throughout.

Significantly upregulated cytokines included pro-inflammatory TNF-α, TNF-β, IL-1β, IL-6, IFN-γ, IL-12p40, IL-15, IL-17, IL-18 and IL-22 ( Figure 3A ). Anti-inflammatory cytokines including IL-4, IL-10, IL-13 and IL-27 were also upregulated ( Figure 3B ). Differential trends in the temporal regulation of cytokines were evident. Some cytokines were most significantly upregulated following 24 hours of stimulation and decreased in a time-dependent manner. In contrast, TNF-β, IL-6 and IL-15 exhibited a time-dependent increase. Cytokines associated with epithelial dysfunction, including TNF-α, IL-1β, and IL-6 were significantly heightened in inflamed models (30). Surprisingly, IL-6 levels exhibited a profound increase following 96 hours stimulation, suggesting initiation of chronic inflammation as IL-6 functions in orchestrating the chronic inflammatory response (31).

Chemokines and growth factors are important immune regulators in IBD where they potentiate the inflammatory response through binding cognate receptors on nearby cells such as immune and stromal cells. Thus, we measured the temporal release of IBD-associated chemokines and growth factors in the secretome of inflamed IBD models ( Figure 4 ).

CXCL8 (IL-8) is a major proinflammatory chemokine released by immune cells, fibroblasts and epithelial cells in IBD (32). Stimulated models elicited a potent CXCL8 response throughout the stimulation period ( Figure 4B ). CXCL1 (GROα), a disease biomarker, was also significantly increased following 24 hours of stimulation, and remained elevated across the stimulation period (33). Although almost absent in controls, CXCL9 secretion was significantly induced upon stimulation, simulating upregulated levels in the IBD mucosa, where it is induced by IFN-γ, and correlates to disease activity in UC (34). Whilst negligible in controls, CXCL10 exhibited a moderate increase during 24-72 hours of stimulation. Of note however, was the substantial increase in CXCL10 observed only at 96 hours recapitulating the significant elevation in disease (35). Mirroring this expression pattern was CCL2 (MCP-1) and IL-6. Downstream angiogenic molecules including VEGF were also upregulated in a time-dependent manner ( Figure 4C ) (36). Similarly, PDGF-AA and PDGF-AB/BB, known to be profibrogenic, were significantly higher in inflamed models than controls, all of which are reflective of IBD tissue/serum (37, 38).

CCL7 (MCP-3), which correlates to the extent of epithelial destruction in IBD patient biopsies, was also significantly upregulated following stimulation (39). Secretion of CX3CL1 (fractalkine) was significantly increased, particularly at 96 hours. Also of note was the increase in G-CSF and GM-CSF, both of which displayed comparable temporal trends. Finally, CCL5 (RANTES), which was found to be abundantly expressed in granulomatous IBD tissue found in CD, was highest following 48-hours of stimulation (40).

Next, we assessed how the inflammatory response affected the epithelial barrier in inflamed IBD models. A significant reduction in TEER occurred following repetitive stimulation, indicating loss of barrier integrity ( Figure 5A ). Assessment of paracellular barrier function using FD20 demonstrated an increase in paracellular permeability ( Figure 5B ), suggesting a leaky-gut phenotype was induced by stimulation.

To elucidate potential mechanisms of epithelial dysfunction, epithelial structure was subsequently investigated. TEM analysis revealed microvilli loss/damage at the epithelial surface of inflamed models, coupled with loss of rootlet visibility ( Figure 5C ). This contrasted the confluent brush border with clear rootlet structure observed in controls, suggesting aberrations to the actin-cytoskeleton at these surface specialisations. To further investigate this, transcript and protein levels of villin were examined, which were significantly depleted in inflamed IBD mucosal models, confirming the loss of an integral cytoskeletal protein that may contribute to epithelial dysfunction. Additionally, the overall epithelial architecture appeared less columnar and polarised in inflamed mucosal models than in controls ( Figure 5D ). Instead, epithelial height was reduced in stimulated models, which exhibited a cuboidal, flattened phenotype.

Areas of pseudostratified epithelium were apparent in inflamed models ( Figure 2A ). To further investigate this, Ki67⁺ quantification was employed to quantify epithelial proliferation. Increased epithelial proliferation was observed in stimulated models ( Figure 5E ). Notably, Ki67⁺ cells were most abundant in pseudostratified areas of the epithelium. TUNEL assays also revealed a time-dependent increase in apoptotic epithelial cells in stimulated models ( Figure 5F ). Together, these data demonstrate increased epithelial turnover in inflamed IBD models, matching observations previously made in UC tissue (41).

Paracellular permeability across the intestinal epithelium is largely regulated by tight junctions and aberrations in tight junction composition are well documented in IBD (42). Herein, differential expression in CLDN2, CLDN4 and CLDN7 mRNA levels were observed, but not OCLN mRNA in inflamed IBD mucosal models compared to controls ( Figure 5G ). Repetitive stimulation reduced CLDN4 and CLDN7 mRNA, which encodes a sealing claudin, and a claudin implicated in regulating paracellular flux, respectively. In contrast, CLDN2 was increased, which encodes a pore-forming claudin.

Chronic inflammation disrupts stromal homeostasis in IBD, generating a proinflammatory mucosal state and extensive ECM remodelling. As such, we next investigated stromal inflammation in IBD models. Inflamed IBD models secreted higher levels of all MMPs quantified than controls ( Figure 6A ). MMP-2, -3, -9, and -13 were the most prevalent MMPs in inflamed models, with MMP-1, -9, and -10 exhibiting significant upregulation following stimulation. Of particular interest, MMP-9 levels were found to be markedly higher than any other MMP in inflamed IBD constructs. ( Figure 6B ). This was of particular importance as MMP-9 is a potential disease biomarker and has been reported to be the most abundant MMP in IBD patient mucosal tissue (3). Immunofluorescent staining confirmed more abundant MMP-9 expression in UC and CD tissue, as well as inflamed IBD constructs compared to healthy tissue and control models. Strikingly, accumulation of MMP-9 was observed subjacent to the epithelial barrier in UC tissues, which to our knowledge is a unique finding of this study. This MMP-9 accumulation was restricted to the upper epithelium and not towards the base of the crypts. Similar observations were made in inflamed IBD models where accumulation was evident in close proximity to the BM.

The TIMP family were also quantified to identify changes in the MMP : TIMP balance ( Figure 6C ). A significant reduction in TIMP-2 and TIMP-3 following stimulation, taken together with an increase in MMPs, confirmed alteration to the stromal niche in inflamed IBD models. LAMA1 and LAMA5 upregulation ( Figure 6D ) further indicated remodelling of the extracellular matrix, particularly at the BM. In line with this observation, COL4A1 mRNA was also increased in inflamed models.

We then sought to investigate the mechanisms underpinning the upregulation of MMPs within inflamed models. To do this, models were stimulated with recombinant cytokines implicated in the pathogenesis of CD that were also significantly upregulated by LPS herein. Both cytokines produced a significant increase in MMP-1, matching concentrations achieved through LPS stimulation ( Figure 6E ). Similar to LPS stimulation, we found that TNF-α, but not IFN‐γ, resulted in highest secretion of MMP-3 and MMP-9 of all MMPs measured. In addition to this, TNF-α but not IFN‐γ, significantly increased production of MMP-10 and MMP-13. TNF-α stimulation elicited higher MMP production across all MMPs quantified than IFN‐γ stimulation, further evidencing the role of TNF- α in mucosal tissue remodelling. Conversely, with the exception of MMP-1, these results indicate IFN-γ alone is insufficient to directly cause MMP-mediated tissue destruction. MMP-3 upregulation is also of clinical importance given its upregulation in CD and UC tissue (43). The MMP-3 concentration observed following TNF-α stimulation exceeded that of LPS stimulation, suggesting TNF-α is strongly implicated in MMP-3 overexpression. Finally, MMP-9 production was considerably increased by TNF-α. However, the MMP-9 response was higher in LPS stimulated models, suggesting multifactorial upregulation.

TNF-α, but not IFN‐γ stimulation also reduced TIMP release in stimulated models ( Figure 6F ). In particular, TNF-α induced a strong reduction in the level of TIMP-2 and TIMP-3, suggestive of increased ECM degradation.