Engineering transplantable jejunal mucosal grafts using patient-derived organoids from children with intestinal failure

Abstract

Intestinal failure (IF), following extensive anatomical or functional loss of small intestine (SI), has debilitating long-term consequences on children¹. The priority of patient care is to increase the length of functional intestine, particularly the jejunum, to promote nutritional independence². Here we construct autologous jejunal mucosal grafts using pediatric patient biomaterials and show that patient-derived organoids (PDO) can be expanded efficiently in vitro. In parallel, we generate decellularized human intestinal matrix with intact nanotopography, that forms biological scaffolds. Proteomic and Raman spectroscopy analyses reveal highly analogous biochemical profiles of human SI and colon scaffolds, indicating that they can be used interchangeably as platforms for intestinal engineering. Indeed, seeding of jejunal organoids onto either type of scaffold reliably reconstructs grafts that exhibit several aspects of physiological jejunal function, and that survive to form luminal structures after transplantation into the kidney capsule or subcutaneous pockets for up to 2 weeks. Our findings provide proof-of-concept data for engineering patient-specific jejunal grafts for children with IF, ultimately aiding in the restoration of nutritional autonomy.

Intestinal failure (IF) occurs in children with a reduction in functional intestine below the minimal requirement to satisfy nutrient and fluid needs to sustain growth¹. Underlying etiologies include anatomical losses due to short bowel syndrome (SBS), dysmotility conditions such as Hirschsprung’s disease, or congenital epithelial defects such as microvillus inclusion disease². Ultimately, children with irreversible IF are referred for intestinal transplantation. However, due to a shortage of donor organs and high mortality rates³, there is an urgent unmet clinical need for innovative treatment strategies. Intestinal tissue engineering offers a personalized solution for IF, tailoring the cellular and scaffold composition of a graft to the individual patient’s disease etiology⁴. For example, patients with SBS caused by necrotizing enterocolitis or Crohn’s disease require reconstruction of a full thickness graft. Conversely, patients with intestinal dysmotility require reconstruction of a neuromuscular graft, whereas engineering of a mucosal graft is paramount for IF patients with purely epithelial defects. Whilst engineering of simpler tissues such as skin and cornea are well established in clinical practice^5,6, examples of successful clinical applications of more complex organs have only been demonstrated in few case reports of tracheal and bladder reconstruction^7,8. To date there have been no clinical studies of bioengineered intestine using IF patient-derived cells, thereby circumventing potential complications of immunosuppresion^3,4.

To maximize clinical relevance, PDOs were generated from 12 children who had IF or are at risk of developing IF due to co-morbidities or complex surgical backgrounds (Supplementary Table 1), where the starting material was limited to 2 endoscopic epithelial biopsies (~2mm size) (Fig. 1a). On average, 3 to 5 organoid units were established 4 weeks after crypt isolation, before expanding to over 10 million cells by week 8 (Fig. 1a-b). Once established, the organoid expansion efficiency was similar in all intestinal regions: duodenum, jejunum and ileum, and did not vary across different clinical backgrounds (Fig. 1c). We estimate that it would take 10 weeks from biopsy collection to expand sufficient numbers of PDOs for seeding a 20cm length of tubular scaffold that could achieve significant clinical benefit⁹. Quantitative reverse transcription-PCR (qRT-PCR) demonstrated that organoids expressed region-specific markers: apical brush border enzyme cytochrome-b reductase 1 (CYBRD1) and iron transporter solute carrier family 40 member 1 (SLC40A1) in duodenal organoids; brush border enzymes sucrase isomaltase (SI) and lactase (LCT) in jejunal organoids (as in native jejunal tissue (Extended Data Fig. 1)); and apical bile acid transporter (SLC10A2) and basolateral organic solute transporter (OSTB) in ileal organoids (Fig. 1d). Expression of the brush border enzymes SI and LCT in jejunal organoids was preserved after significant passaging time (P>25), indicating that regional identities of organoids are intrinsically programmed (Fig. 1e). Next, we tested expansion and differentiation potentials of jejunal PDOs since 90% of digestion and absorption occurs in the proximal 100-150cm of jejunum^9,10. PDOs were treated with a GSK3β inhibitor (CHIR99021) to boost Wnt signaling or with a gamma-secretase inhibitor (DAPT) to inhibit Notch signaling. qRT-PCR analysis showed that upon CHIR-treatment, stem (OLFM4 and LGR5) and Paneth cell (LYZ) genes were significantly upregulated while differentiation genes (MUC2, ALPI and CHGA) were downregulated (Fig. 1f). A strong induction of proliferation was simultaneously observed in CHIR-treated organoids, which was accompanied by protein expression of stem cell (LGR5, SOX9) and Paneth cell (LYZ) markers (Fig. 1g-h and Extended Data Fig. 1c-d). Conversely, DAPT-treated organoids displayed loss of stem cell and Paneth cell markers and gain of differentiation markers (Fig. 1f, i and Extended Data Fig. 1b-d). Together, these results show that PDOs expand rapidly in vitro, whilst maintaining their original intestinal region identity and differentiation potential.

Figure 1. Generation and characterization of primary intestinal organoids derived from targeted pediatric patient group.

a, Schematic overview demonstrating the expansion timeline after harvesting intestinal crypts endoscopically from pediatric patients. Each patient biopsy sample yields ~3-5 organoids by week 4, and over 10 million cells by week 8 after successive passaging as indicated by a representative phase contrast image of organoids in culture at week 8. b, Phase contrast images of human intestinal organoids [patient 6] established from isolation at the indicated time points. Original magnifications: X20 (days 0); X10 (days 3, 8); X5 (days 15, 18, 28). c, Representation phase contrast images of first passage expansion cultures of duodenal [patient 9] (left), jejunal [patient 2] (middle) and ileal [patient 3] organoids (right). Scale bars represent 200μm. d, Quantitative RT-qPCR analysis of human duodenal [patient 10], jejunal [patient 2] and ileal [patient 14] organoids for functional duodenal markers (CYBRD1; SLC40A1), jejunal markers (SI; LCT) and ileal markers (SLC10A2; OSTB). e, Quantitative RT-qPCR analysis of human jejunal organoids [patient 2, 7, 8] at passages 5, 15 and 25 for jejunal specific markers SI and LCT. f, Quantitative RT-qPCR analysis of human jejunal organoids [patients 2, 7, 8] cultured in basal media culture conditions as indicated in the method, treated with the GSK3β inhibitor CHIR99021 (CHIR), or the Notch inhibitor DAPT. g, Representative stainings for EdU and DAPI of human jejunal organoids [patient 2] in basal culture conditions and expansion conditions (+CHIR). Scale bars represent 30μm. h,i, Representative immunostaining of human jejunal organoids [patient 2] cultured in expansion conditions (+CHIR) (h) or differentiation conditions (+DAPT) (i) using the indicated antibodies to mark proliferating cells (Ki67), stem cells (LGR5 and SOX9), Paneth cells (LYZ), goblet cells (UEA-1), epithelial cells (E-cad), enterocytes (alkaline phosphatase) and enteroendocrine cells (CHGA). Scale bars represent 100μm. All images are representative of 3 experiments (b, c, h, i) or 2 experiments (g). Quantitative data shown represents mean ± s.e.m. of n = 3 experimental replicates (d) or biologically distinct replicates (e, f). Differences were analyzed by one-way ANOVA with Dunnett’s multiple comparisons test in (d, f) or with Tukey’s multiple comparisons (e). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 were considered significant; n.s., not significant. Detailed p values are stated in the Source Data.

Biological scaffolds were chosen for this study due to their inherent quality of retaining the natural microenvironment, which significantly impacts cell behavior and identity^11–21. We have previously described a decellularization protocol to generate rodent and piglet extracellular matrix (ECM) intestinal scaffolds^13,20. In this study, alongside characterizing piglet SI scaffolds (Extended Data Fig. 2a-d), we further increased the translational potential by fabricating human intestinal scaffolds from pediatric patients undergoing intestinal resections (Supplementary Table 1). Histological analysis confirmed the absence of cells and presence of ECM collagen in both SI and colon scaffolds (Fig. 2a-b). Scanning electron microscopy demonstrated remarkable preservation of mucosa, submucosa and muscularis ultrastructure (Fig. 2c). Importantly, intact crypt-villus axes of SI scaffolds and crypts of colon scaffolds were clearly identified (Fig. 2c), thereby providing the optimal nanotopography for intestinal graft reconstruction. We further evaluated the mechanical properties of human SI and colon scaffolds given the relevance of this to transplantation surgery. Quantification of the tensile stress and strain across the transverse axis of age-matched scaffolds showed no significant differences in Young’s modulus. However, SI scaffolds tended to have higher break points on the stress-strain curves (Extended Data Fig. 2e), which may be caused by the presence of villi in the scaffolds. This may be a surgical advantage in vivo but would require further testing in orthotopic transplantation models.

Figure 2. Decellularization and biomolecular characterization of human small intestinal and colon scaffolds.

a,b, Representative images of SI [patient 1] (a) and colon [patient 18] (b) samples before (native) and after decellularization. Top, macroscopic images; middle, H&E histological images; bottom, immunofluorescent staining using the indicated antibodies. Scale bars: top, 1cm; middle, 200μm; bottom, 100μm. c, Representative scanning electron micrographs of decellularized SI [patient 1] and colon [patient 18] scaffolds highlighting the microarchitecture of the mucosa (Mu) submucosa (S) and muscularis (M). Yellow arrowheads indicate intestinal crypts. The red arrow head indicates villi structure present on the SI scaffold. Scale bars: top, 100μm; bottom, 10μm. d, Raman spectral analysis of comparable histological regions (mucosa (Mu), submucosa (S) and muscularis (M)) of the native tissue (blue lines) and decellularized scaffolds (red lines) of SI [patient 1] (left) and colon [patient 18] (right) samples. Peaks at 726 and 780 cm^-1 were assigned to ring breathing vibrations of nucleic acids whilst peaks at 1078 and 1303 cm^-1 were assigned to δ(CH₂) and v(C-C)/v(C-O) modes of lipids. e, False colored heat maps representing direct classical least squares component analysis of Raman maps using previously acquired reference spectra of purified biomolecules. Distinct spatial distribution of Phenylalanine (PHE), Collagen (COL) and Glycosaminoglycans (GAGs) in SI [patient 1] and colon [patient 18] scaffolds is shown. Scale bars represent 50μm. Images representative of 2 experiments. f, Score plot from principal component analysis differentiates the distinct Raman biochemical spectral profiles of each distinct histological layer of the SI [patient 1] (blue) and colon [patient 18] scaffolds (red). Images are representative of at least 3 (a-c) or 2 (e) independent decellularization experiments. g, Venn diagrams showing total and extracellular proteins detected in the SI [patients 4, 11, 12, 13] and colon [patients 2, 15, 16, 17] scaffolds by mass spectrometry. Proteomics data represents samples from 4 biologically independent patient samples in each group.

Next, we investigated the biomolecular composition of human SI and colon scaffolds to explore the potential use of either scaffold type for jejunal reconstruction. Various biological scaffolds have been used in intestinal engineering to provide physiological growth signals that aid cell engraftment and self-organisation²². However, the characteristics and distribution of specific ECM cues in decellularized human scaffolds remains limited^23,24. We therefore employed Raman spectroscopy to compare the spectral profiles between native gut and scaffolds. The highly analogous profiles along with the loss of nucleic acid (wavenumbers 726 and 780cm^-1) and lipid (wavenumbers 1078 and 1303cm^-1) features confirmed that DET processing successfully removes all cellular material with preserved ECM components (Fig. 2d). Importantly, the spectra between SI and colon scaffolds were remarkably similar despite their structural and functional differences. Raman imaging further highlighted similar spatial distribution of specific ECM components in both scaffold types: collagen was predominantly localized in the submucosa; phenylalanine was most abundant in the muscularis; while glycosaminoglycans (GAGs), important in intestinal stem cell homeostasis^25,26, were enriched in the mucosa (Fig. 2e). Principal component analysis readily segregated the spectra into distinct clusters based on histological layers, where the mucosal spectra of both scaffold types tightly clustered together (Fig. 2f and Extended Data Fig. 2f). To validate the similarities observed in the Raman data, mass spectrometry was used to generate a global proteomic profile of human SI and colon scaffolds. Strikingly, the majority of proteins were detected in both scaffold types, including 17 collagen subtypes and 5 laminin subtypes (Fig 2g, Supplementary Table 2), whilst only 11/377 total proteins and 2/126 ECM proteins were detected in either SI or colon scaffolds alone (Supplementary Table 3 and Extended Data Fig. 2g-h). Altogether, our data support utilizing either scaffold types in subsequent graft engineering experiments.

To reconstruct jejunal grafts, we performed a series of organoid seedings in vitro before testing for the presence of jejunal-specific functions or in vivo transplantation (Fig. 3a). As proof-of-concept for creating a personalized patient-derived graft, subsequent experiments utilized jejunal organoids from a single IF patient (patient 2, Supplementary Table 1). In addition, jejunal fibroblasts were also isolated and seeded by injection into all human scaffolds intended for in vivo transplantation. Jejunal fibroblasts expressed fibronectin (F-NEC), vimentin (VIM), fibroblast surface protein marker-1 (FSP-1), laminin α5 (LAMA5) and scattered weaker co-expression of αSMA (Extended Data Fig. 3a), indicating a mixed population of jejunal fibroblasts and myofibroblasts. All scaffolds were individually mounted onto custom-built stages (Extended Data Fig. 3b). Scaffolds receiving fibroblasts were maintained in static culture for 3 days prior to organoid seeding onto the mucosal surface of all scaffolds. The seeded grafts were cultured in static conditions for a further 4 days before converting to dynamic conditions using a perfusion bioreactor system (Fig. 3a and Extended Data Fig. 3c). All grafts intended for in vivo transplantation were maintained in dynamic culture for 7 days (Fig. 3a). Histological analysis of these grafts showed full coverage of columnar epithelial cells on the scaffold surface with visible crypt units (Fig. 3b) and abundant fibroblasts located subepithelially prior to transplantation (Extended Data Fig. 3d). For comprehensive in vitro characterization, grafts were collected after 14 days of dynamic culture. Preliminary optimizations were performed using piglet SI scaffolds, the proteomic composition of which was recently published²⁰. First, micro-CT imaging was performed to provide an overall assessment of volume and distribution of epithelial cells across the scaffold (Extended Data Fig. 4a). Histology confirmed a polarized monolayer of columnar cells covering the scaffold surface and crypt compartments (Extended Data Fig. 4b-c). Immunofluorescent staining of collagen revealed a uniquely homogenous hyaline expression pattern in multiple regions beneath the epithelial cells, suggesting new matrix deposition by the epithelial cells (Extended Data Fig. 4d). Most intestinal cell types were readily detected, including lysozyme-positive Paneth cells, AB-PAS-positive goblet cells and alkaline phosphatase/ALPI-positive enterocytes alongside proliferating Ki67 positive cells (Extended Data Fig. 4e-h). Jejunal-specific sucrase isomaltase was also widely detected on the brush border (Extended Data Fig. 4i), confirming maintenance of region-specific identity. Next, human jejunal grafts were constructed using human SI and colon scaffolds. Immunostaining demonstrated that proliferation (indicated by PCNA) and differentiation of enterocytes (ALPI) and goblet cells (AB-PAS) was present (Fig. 3c-d). However, expression of chromogranin-A or lysozyme was not detectable, indicating a lack of terminally differentiated enteroendocrine or Paneth cells. We further performed in-depth electron microscopy analysis and confirmed the presence of microvilli, basement membrane and mucous vesicles in the grafts (Fig 3e and Extended Data Fig. 3e). Cells containing multiple secretory vesicles were also identified, which may represent early immature Paneth cells (Fig. 3e). Similar to the piglet scaffold seedings, regions of collagen-positive epithelial-secreted hyaline matrix were detected, indicative of ECM remodeling (Fig. 3c-d). Importantly, expression of ALPI was also detected on grafts constructed using colon scaffolds (Fig. 3d), supporting the notion that both scaffold types are effective for jejunal graft reconstruction. Tight junction (ZO1) and polarity (Na+/K+/ATPase) markers were also expressed on the reconstructed jejunal scaffold (Fig. 3f), which are essential elements for intestinal barrier function.

Figure 3. Bioengineering functional human jejunal mucosal grafts in vitro .

a, Schematic outline of the scaffold seeding strategies using a bioreactor circuit. The timeline shows seeding of each cellular component onto the scaffolds and the time periods in static and dynamic cultures (top). The bioreactor circuit design and all individual components are indicated (bottom). b, Representative H&E staining of a jejunal graft harvested at day 11, showing a monolayer of epithelial cells with invaginating crypt compartments marked by black arrowheads [colon scaffold - patient 2] (left, overview; right, close up). c,d, Representative histological and immunostaining images of jejunal grafts reconstructed using human SI scaffolds [patient 13] (c) or human colon scaffolds [patient 16] (d) at day 18. New matrix deposition is shown by newly synthesized collagen (white asterisks). Arrowheads indicate AB-PAS-positive goblet cells. Scale bars represent 50μm. e, Representative electron micrographs of a jejunal construct [SI scaffold - patient 5] showing microvilli (MV) (top left and right); basement membrane with basal lamina (BL) and reticular lamina (RL) at the scaffold (Sc) border (bottom left); Goblet cell (G) with mucous vesicles indicated by the orange arrow head (bottom middle) and Paneth cell (P) with secretory vesicles indicated by the green arrow head (bottom right). Scale bars represent 5μm. f, Representative immunofluorescent images of a jejunal graft seeded on human SI scaffolds [patient 5] using the indicated antibodies. Scale bars represent 50μm. g, Timeline indicating the experimental sampling (marked by each colored symbol) of jejunal grafts for functional analyses along the course of the in vitro culture period. h, Immunofluorescent staining showing β-AMCA peptide (red) uptake on a jejunal graft (piglet SI scaffold). Phalloidin staining (green) indicates epithelial cell boundaries. Scale bars represent 30μm. i-m, Functional analysis using jejunal grafts seeded on human colon scaffolds (green lines), human SI scaffolds (red lines), piglet scaffold (blue lines) or unseeded blank scaffolds (black lines). i, Dipeptidyl protease IV activity, as measured by nitroaniline release from the grafts [colon scaffolds - patients 19 & 20]. j, Disaccharidase enzyme activity, as measured by rising glucose concentrations following a sucrose challenge (solid lines) or PBS control (dashed lines) to the grafts [SI scaffolds - patients 11 & 12; colon scaffold - patient 2; piglet scaffold]. k, Barrier function, as measured by FITC-Dextran leakage percentage through jejunal grafts (piglet scaffold) from “luminal side” to “serosal side”. Blank scaffolds show an average of 61% baseline leakage as indicated by dashed line. l, Citrulline concentrations measurement in supernatant collected from the indicated graft cultures [SI scaffolds - patients 11 & 12; colon scaffold - patient 2; piglet scaffold]. All organoids used in this figure originate from patient 2. For all functional assays, data represents mean ± s.e.m. of 3 independently cultured jejunal grafts. Images are representative of 4 (b), 3 (c-f) and 2 (h) graft culture experiments.

To evaluate the functional capacity, we tested absorption, digestion and barrier competence of these grafts (Fig. 3g). First, peptide absorption was demonstrated by uptake of the fluorescently labelled peptide β-Ala-Lys-AMCA (Fig. 3h), indicating active peptide transporters. We then evaluated the digestive capacity by assessing both peptide hydrolysis and disaccharidase function. Dipeptidyl peptidase IV activity was detected, indicating that peptides absorbed by the graft were subsequently digested (Fig. 3i). High glucose production after sucrose challenges confirmed that the jejunal brush-border disaccharidase activity was present throughout cultures (Fig. 3j). In addition, assessment of intestinal barrier function indicated that 61% of FTIC-Dextran leakage was observed in the blank scaffold, while the leakage was reduced to 45% by day 18 of the graft culture (Fig 3k). Finally, we examined citrulline levels in our graft culture supernatants, since circulating citrulline is used as a clinical biomarker of IF and directly correlates with absorptive enterocyte mass²⁷. Indeed, increased citrulline concentration was observed over time (Fig. 3l), which represents a robust and non-destructive method for tracking enterocyte growth in vitro.

To examine jejunal graft survival and differentiation in vivo, we performed transplantation either under the kidney capsule or in subcutaneous pockets of immunodeficient mice. Owing to dimensional limitations, only grafts constructed on piglet scaffolds were feasible for the kidney capsule model. Grafts harvested 1-week post transplantation showed signs of neo-vascularization macroscopically (Fig. 4a). Serial histological sectioning of the graft demonstrated the presence of luminal structures populated with human nucleoli+ intestinal epithelium (Fig. 4b-c and Extended Data Fig. 5a). 3D volume reconstruction of the serial sections revealed continuous tubular structures throughout the graft (Fig. 4d and Supplementary Video 1). Unexpectedly, immunostaining of AB-PAS and ALPI was largely negative, suggesting a lack of goblet cell and enterocyte differentiation in this model (Extended Data Fig. 5b). On the other hand, a large population of vimentin+ and αSMA+ cells were found around the lumen, indicating infiltration of host myofibroblasts into the scaffold (Fig. 4e). Immunofluorescent staining showed high expression of stem cell markers OLFM4 and SOX9 (Fig. 4f, g). Together, the results suggest that high stromal infiltration drives the intestinal epithelium towards an undifferentiated state by recapitulating the stem cell niche of the intestinal crypts²⁵.

Figure 4. Characterization of the engineered jejunal graft following in vivo transplantation.

a-g, Kidney capsule transplantation model, 1-week. a, Macroscopic image of the kidney harvested after implantation of a jejunal graft in the kidney capsule. b,c, Histology of a transplanted jejunal graft as analyzed by H&E and human nucleoli staining. d, 3D volume rendered model of the jejunal graft structure after transplantation under the kidney capsule. e-g, Representative immunofluorescent images of transplanted jejunal grafts using the indicated antibodies. h-o, Subcutaneous transplantation model, 1 week (h-i,k-o) or 2 weeks (j). h, Top left, schematic representation of the luciferase-GFP reporter plasmid used to label jejunal organoids (bottom left). Scale bar showing bioluminescent signal intensity (middle). Seeded grafts using labelled organoids were detected using live in vivo bioluminescent imaging of mice in the subcutaneous transplantation model (right). i, Representative histology and immunostaining of transplanted jejunal grafts. AB-PAS, Alcian Blue - Periodic Acid Schiff. j, Quantification of the formation of lumens within jejunal grafts following subcutaneous implantation in Teduglutide-treated versus control mice, measured as the percentage of sections with a lumen per graft. Data represents mean ± s.e.m of n = 4 independently transplanted jejunal grafts at 2 weeks post-transplantation; Unpaired t-test, *P 0.0271. k, Representative histology of a transplanted human jejunal graft indicating the mucosa, submucosa and muscularis structure of the graft. Arrowhead indicates the epithelial layer of polarized columnar jejunal cells. l, Immunostaining analyses of cell proliferation on the scaffold, as indicated by PCNA (top) and EdU (bottom) staining. Arrowhead indicates the jejunal epithelium. m, Electron microscopy analysis identifying mucous granules of goblet cells in the native intestine (top left) and a jejunal construct (bottom left and right). n,o, Representative immunostaining of a jejunal graft formed using epithelial markers (pancytokeratin and E-cad), stromal marker (vimentin) and jejunal brush border enzyme (SI). Grafts in a-i were formed using piglet SI scaffold. Grafts in j were formed using human colon scaffolds from patient 2. Grafts in k-o were formed using human SI scaffolds from patient 1. All organoids and fibroblasts used in this figure originate from patient 2. Images shown are representative of 3 transplanted grafts (a-c, e-g), 6 transplanted grafts (h,i) and 2 transplanted grafts (k-o). All scale bars represent 50μm unless specified otherwise.

See Source Data.

Next, we investigated the subcutaneous transplantation model to accommodate larger grafts. Jejunal PDOs were labelled with a GFP-luciferase reporter prior to seeding, enabling epithelial growth monitoring with live bioluminescent imaging (Fig. 4h). Piglet scaffold grafts were tested first, for direct comparisons with the kidney transplantation model. Akin to the kidney capsule data, lumens of human nucleoli+ epithelial cells were detected 1 week after implantation (Fig. 4i). However, this time ALPI+ enterocytes and AB-PAS+ goblet cells were detected, which may be linked to lower levels of stromal cell infiltration (Fig. 4i). We subsequently transplanted thicker jejunal grafts constructed from human scaffolds. To enhance survival of larger dimension grafts, we explored the use of Teduglutide, a glucagon-like peptide-2 analogue licensed for use in patients with IF²⁸, since expression of the Teduglutide receptor (GLP2R) was observed in both human fibroblasts and organoids (Extended Data Fig. 5c). Two weeks after transplantation, serial sectioning showed intestinal lumens populated with epithelial cells in 10.4 ± 0.7% sections of all grafts receiving Teduglutide, compared to only 0.28 ± 0.6% sections of all grafts in the control group (p=0.0271) (Fig. 4j). Presence of CD31+ cells surrounding the lumens further implied neovascularization in the grafts (Extended Data Fig. 5d). These results suggest that Teduglutide treatment enhances epithelial cell survival on grafts in vivo. Distinct histological layers of the graft were also preserved and monolayers of human intestinal epithelial cells were identified on the mucosa (Fig. 4k). Proliferating cells were detected in both epithelial and pericryptal stromal cells of the scaffold (Fig. 4l). Although no clear AB PAS+ cells were detected, electron microscopy analysis identified cells with mucous granules, suggesting the presence of early goblet cell differentiation (Fig. 4m). Immunofluorescent staining for pan-cytokeratin, e-cadherin and sucrase isomaltase confirmed the epithelial and jejunal identity of cells (Fig. 4n,o). Vimentin+ fibroblasts were localized in proximity to epithelial cells recapitulating the native microenvironment (Fig. 4n). Together, these data indicate that human scaffolds support and maintain intestinal epithelial cell differentiation predominantly towards the enterocyte lineage in vivo.

Finally, we investigated the effect of vascularization on grafts by injecting human umbilical vein endothelial cells (HUVECs) into human scaffolds 1 day prior to subcutaneous transplantation. Among the 6 scaffolds injected with HUVECs, 2 contained vessels macro-and microscopically upon harvest, and these same 2 scaffolds had visible lumens on 33% of the serial sections (Extended Data Fig. 5e-f). Conversely, in the remaining 4 scaffolds without vessels, lumens were detected in only 1.4 ± 3% of the serial sections (Extended Data Fig. 5f). Blood vessels were often observed in close proximity to the lumens (Extended Data Fig. 5g), suggesting that vascularization enhances epithelial cell survival. Histological analysis confirmed the presence of goblet cells and sucrase isomaltase+ enterocytes in the epithelial lumens surrounded by fibroblasts and CD31+ vessels (Extended Data Fig. 5h), reiterating that jejunal identity is maintained in engineered patient-derived grafts in vivo.

In conclusion, we have established a clinically relevant protocol for timely intestinal graft reconstruction for IF using patient-derived materials. We have demonstrated that human jejunal organoids derived from minimal starting materials can be expanded efficiently in vitro, while maintaining their intrinsic region-specific functional identity and differentiation potential in vitro and in vivo. Our data support the concept of routinely banking PDOs as a new clinical standard at the point of intestinal resection in children at risk of IF to facilitate future options of personalized intestinal reconstruction. Importantly, we present the innovative concept of using colon scaffolds for SI graft reconstruction. This implies that colon derived from cadaveric donors or resected from children affected by conditions such as Hirschsprung’s disease could be decellularized, stored and donated for engineering functional SI mucosa. Furthermore, in IF etiologies such as midgut volvulus where colon is typically preserved, there is potential to convert the IF patient’s own colon to SI by replacing the colon mucosa with a jejunal mucosal graft engineered using the patient’s own jejunal organoids. Unlike previous publications that focused largely on iPSCs, synthetic scaffolds or rodent ECM matrix^{12,19,29–33}, the use of primary human materials (both cells and decellularized scaffolds) in this study is a highly relevant step towards clinical translation. Additionally, our results indicate that Teduglutide, which is licensed for use in IF³⁴, can be applied also in the context of tissue engineering to promote the survival of grafts in vivo. A follow-on study of human jejunal engineering using an orthotopic transplantation model would provide the most conclusive data for both structural and functional competence of our grafts. Whilst the current study provides important conceptual advances towards personalized human SI grafts for IF patients, the challenges of regaining neuromuscular function and full vascularization remain and are pivotal engineering steps for future full thickness intestinal reconstruction.

Methods

Animals and human tissue

Immune deficient NOD-SCID IL-2Rγnull (NSG) female mice, aged 9 - 14 weeks old, were used in all experiments (obtained from the Francis Crick Institute Biomedical Research Facility). All mice were housed in the animal facility at the Francis Crick Institute. The housing conditions comprised of IVC cages, automated water system, and 28% protein diet. A maximum of 5 adult mice weighing up to 20g are homed in a single cage, maintaining ambient temperature at 19-23°C, humidity at 55-/+10% and 12-hour light/dark cycles. All experiments were performed with ethical approval (Animal Welfare Ethical Review Body) under Home Office Project License PPLs PEF3478B3, 70/8904 and 70/8560.

Porcine (Sus scrofa domesticus) SI from the ‘Pietrain’ breed was used to derive piglet SI scaffolds. Piglets up to 3 kg in weight were euthanized following criteria outlined by the JSR veterinary advisors. Once sacrificed, the animals were transported to the lab via courier and the intestine was harvested immediately on arrival (within 6 hours of euthanasia).

Ethical approval for the use of human intestinal tissue was obtained from the Bloomsbury NRES committee (REC references 04-Q0508-79, and 18/EE/0150). The Committee was constituted in accordance with the Governance Arrangements for Research Ethics Committees and complied fully with the Standard Operating Procedures for Research Ethics Committees in the UK. Informed consent for the collection and use of human tissue was obtained from all patients, parents or legal guardians at Great Ormond Street Hospital, London. Patient demographics and clinical background information for all human tissue used in this study is listed in supplementary table 1.

Organoid and fibroblast culture conditions

Endoscopic biopsy specimens were cut into finer pieces and washed in cold PBS. For organoid isolation, the tissue fragments were incubated in 2mmol/L EDTA cold chelation buffer, consisting of distilled water with 5.6mmol/L Na2HPO4, 8mmol/L KH2PO4, 96.2mmol/L NaCl, 1.6mmol/L KCl, 43.4 mmol/L sucrose, 54.9 mmol/L D-sorbitol, 0.5mmol/L DL-dithiothreitol) for 30 minutes at 4°C as previously reported³⁵. Following this incubation period, crypts were released from the fragments by shaking vigorously. The supernatant was centrifuged at 800RPM for 5 minutes at 4°C, to form a pellet of intestinal crypts, which were washed in Advanced Dulbecco’s modified Eagle medium (DMEM) / F12 supplemented with 5% penicillin/streptomycin, 10mmol/L HEPES and 2mM of GlutaMAX. The crypts were then resuspended in Basement Membrane Extract® (BME) and seeded in a single droplet on pre-heated 48-well plates. The BME was polymerized for 20 minutes at 37°C, before adding 250μL/well of either human IntestiCult™ Organoid Growth Medium (STEMCELL Technology, #06010) or human organoid basal culture media consisting of conditioned media produced using stably transfected L cells (Wnt 50%; R-spondin 20%; Noggin 10%) and the following growth factors: B27 1X (Invitrogen), Nicotinamide 10mM (Sigma-Aldrich), N-acetyl cysteine 1mM (Sigma-Aldrich), TGF-β type I receptor inhibitor A83-01 500nM (Tocris), P38 inhibitor SB202190 10μM(Sigma-Aldrich), Gastrin I 10nM (Sigma-Aldrich), EGF 50ng/ml (Invitrogen). Rho-kinase inhibitor Y-27632 was added to the culture media for the first week in culture at a concentration of 10μM. The media of each well was changed every 2 days. Organoids in expansion were cultured in 3μM CHIR99021. Organoids in differentiation phase were cultured in 10μM DAPT for 48 hours. For the isolation of human intestinal fibroblasts, intestinal fragments left over from the chelation step above were washed in PBS and placed on the bottom of tissue culture dishes with DMEM supplemented with 10% heat inactivated Fetal Bovine Serum (FBS) (Sigma-Aldrich), 5% penicillin/streptomycin and 1x insulin-transferrin-selenium solution (both ThermoFisher). Fibroblasts grew from the fragments within 3-4 days. Cells used for seeding experiments were between passages 3-10.

HUVEC culture conditions

HUVECs (ATCC) were cultured in EGM-2 endothelial medium bullet kit (Lonza, cat.no. CC-3162) and were not used beyond passage 15.

mRNA isolation and quantitative PCR

RNA was extracted according to the manufacturer’s instructions (Qiagen RNAeasy). cDNA was prepared using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, #4368813). Quantitative PCR detection was performed using PowerUp™ SYBR® Green Master Mix (Applied Biosystems, A25742). Assays for each sample were run in triplicate and were normalized to the housekeeping gene β-actin. Primer sequences are listed in Supplementary Table 4.

Western blotting

BME droplets containing organoids in either basal, expansion (3μM CHIR) or differentiation (10μM DAPT) conditions were disrupted using ice cold Advanced Dulbecco’s modified Eagle medium (DMEM)/F12 supplemented with 5% penicillin/streptomycin, 10mmol/L HEPES and 2mM of GlutaMAX, pelleted at 1000RPM for 5 minutes and lysed in ice cold protein lysis buffer (150mmol/L NaCl, 30mmol/L Tris (pH 7.5), 1mmol/L EDTA, 1% Triton X-100, 10% glycerol, 0.1mmol/L phenylmethylslfonyl fluoride, 0.5mmol/L dithiothreitol, protease inhibitor cocktail tablets (EDTA-free) (Roche) and phosphatase inhibitor cocktail tablets (Roche)). Lysates were pelleted for 10 minutes at 13000PRM and supernatants stored at -80°C until analysis. Protein quantification was by Bradford assay. 40μg protein was loaded onto sodium dodecyl sulfate-polyacrylamide gels (SDS-PAGE) and transferred to membranes, which were blocked using 5% milk in Tris-buffered saline (50mM Tris, 150mmol/L NaCl, pH6) containing 0.1% Tween-20 (TBST) for 1 hour. Primary antibodies (as in Table S5) were added in 3% bovine serum albumin in TBST as follows: lysozyme (1:1000), Sox9 (1:1000), Olfm4 (1:1000), SI (1:200) overnight at 4°C. Membranes were washed in TBST before appropriate HRP-conjugated secondary antibodies were added at 1:5000 in 5% milk in TBST for 1 hour. Antibody binding was detected using chemiluminescence ECL Prime Western Blotting Substrate (GE Healthcare). Membranes were reprobed with anti-β actin peroxidase antibody (Sigma, A3854, 1:25000) for 10 minutes before ECL development. Band intensity was quantified in ImageJ and normalized to β actin.

Decellularization of SI and colon

The detergent-enzymatic treatment (DET) for decellularization, previously established on rat small bowel, was adapted for porcine and human intestine¹³. One cycle of DET consisted of the following steps: (i) 24 hours of deionized water at 4°C; (ii) 4 hours of 4% sodium deoxycholate (Sigma) at room temperature; (iii) 1 hour of deionized water at room temperature; (iv) 3 hours of 2000kU DNase-1 (Sigma) in 1M NaCl (for human tissue) or 0.15M NaCl and 10mM CaCl (for piglet tissue) at room temperature¹³. After harvesting piglet small intestine, the mesenteric artery, mesenteric vein and lumen were cannulated using a 29-gauge surgical cannula. The lumen of the intestine and the mesenteric artery were perfused with continuous fluid delivery of DET solutions using a Masterflex L/S variable speed roller pump at 3ml/min, for two cycles. Whole sections of human intestinal tissue collected from surgery were washed in cold PBS containing 5% penicillin/streptomycin to remove luminal contents, before starting decellularization by immersion in DET solutions and gentle agitation on a roller. For human scaffolds, two to three cycles of DET were performed, depending on both age of the patient and weight of the tissue received from the theatre. Specifically, age of patients: <6month: 1-2 cycles; 6 month-1 year: 2-3 cycles; >1year: 3 cycles; weight of tissue: <3g: 1-2 cycles; 3-20g: 2-3 cycles; >20g: 3 cycles. Gamma irradiation at a dose of 16000Gy for 15 hours was applied to sterilize the scaffolds and then preserved at 4°C, in PBS containing 1% penicillin/streptomycin prior to use in cell culture.

DNA and ECM Quantification

Tissue samples were taken at random immediately post-harvesting and after decellularization protocol for DNA and ECM components quantification. DNA was quantified using a PureLink Genomic DNA Mini Kit (Thermo Fisher). The final concentration of DNA in the samples was measured using a NanoDrop (model NanoDrop 1000 Spectrophotometer by Thermo Fisher). ECM components were quantified using a QuickZyme Collagen assay kit (QuickZyme Biosciences) to measure the collagen, a Blyscan Sulfated 18 Glycosaminoglycan Assay kit (Biocolor) for the glycoaminoglycans (GAGs) and a Fastin Elastin Assay kit (Biocolor) for elastin, according to manufacturers’ instructions.

Histology

Samples were fixed in 10% Neutralised Buffered Formalin (Sigma) at room temperature overnight, dehydrated in graded alcohols, paraffin embedded and sectioned at 4μm. Tissue slides were stained according to manufacturers’ instructions with Haematoxylin and Eosin (H&E) (Thermo Fisher), Alkaline Phosphatase (Vector Laboratories), Alcian-Blue Periodic Acid Schiff (Sigma). Picrosirius Red (PR), Elastic Van Gieson (EVG) and Alcian Blue (AB) (Thermo Fisher) staining was used to assess retention of collagen, elastin and glycosaminoglycans respectively.

Immunostaining

For immunofluorescence studies, paraffin embedded slides were rehydrated, permeabilized with 0.3% Triton X100 (Sigma, UK) for 30 minutes at room temperature. Heat mediated antigen retrieval was performed using a sodium citrate buffer (pH6). For whole mount immunostaining of intestinal organoids, cells were fixed with 4% PFA at room temperature for 20 minutes. Primary antibodies (Supplementary Table 5) were diluted in 1% BSA/PBS/0.01% Triton X-100 and applied overnight at 4°C. Samples were incubated with Alexa Fluor secondary antibodies (Invitrogen) (1:1000) for 45 min at room temperature, washed and mounted with ProLong™ Diamond Antifade Mountant with DAPI (ThermoFisher). EdU staining with the Click-iT EdU Alexa Fluor 568 Imaging kit (Life Technologies) followed the manufacturer’s protocol. DNA was stained with DAPI (Molecular Probes). Images were acquired using a Leica SP5 confocal microscope.

For immunohistochemistry, antigen retrieval was performed using a sodium citrate buffer. Slides were permeabilized using a 0.2% Triton X100 (Sigma, UK) for 30 minutes at room temperature, and blocked with 5% bovine serum albumin (BSA) for 30 minutes. Primary antibodies were detected using peroxidase conjugated secondary antibodies (1:1000) using standard protocols as described previously. Image analysis and capture was performed using a Leica stereomicroscope or an inverted Nikon microscope. All antibodies used are listed in Supplementary Table 5. Images were processed using ImageJ and Adobe Photoshop. Quantifications were performed on raw images however for presentational clarity, adjustments of brightness and contrast were applied equally to all panels of a given figure.

Mass spectrometry and data analysis

Four biological replicates of both decellularized colon and SI ECM scaffolds were used for mass spectrometry analysis. Samples were prepared for analysis as previously reported³⁶. Briefly, 1mg of lyophilized decellularized tissue was solubilized in 8M urea containing 10mM dithiothreitol. Iodoacetamide was added to a final concentration of 55mM and incubated for 30 minutes at room temperature protected from light. Proteins were treated with PNGaseF (25,000 units/mg) overnight. An initial protein digest using Lys-C (10 μg/mg for 4 hours) was followed up with two successive tryptic digests (20 μg/mg overnight; 10 μg/mg for 4 hours). All enzyme reactions were performed at 37ᵒC. Four biological replicates each of colon and intestine scaffolds were processed. Peptide material was cleaned up using C18 Sep-Pak columns (Waters 50 mg sorbent, WAT054955). Eluates were dried in a speedvac concentrator. Dried peptides were solubilized in 0.1 % trifluoroacetic acid (TFA) to a concentration of approx. 5 μg/μl, then diluted to 0.25 μg/μl in a glass auto-sampler vial immediately prior to analysis. Each of the eight samples were analyzed in technical triplicate (approx. 1 μg per injection) using a ThermoFisher Scientific QExactive mass spectrometer coupled to an UltiMate 3000 HPLC system for on-line liquid chromatographic separation. Each sample was initially loaded onto a C18 trap column (ThermoFisher Scientific Acclaim PepMap 100; 5 mm length, 300 μm inner diameter) then transferred onto a C18 reversed phase column (ThermoFisher Scientific Acclaim PepMap 100; 50 cm length, 75 μm inner diameter). Peptides were eluted at a flow rate of 250nL/min with a stepped gradient of 5-25% buffer B (80% acetonitrile, 0.1% formic acid, 5% DMSO) for 60 minutes followed by 25-40% for 20 minutes. Higher energy Collisional Dissociation (HCD) was used for MS/MS peptide fragmentation. Singly-charged and unknown charge state precursor ions were not analyzed. Full MS spectra were acquired in the orbitrap (m/z 300–1800; resolution 70k; AGC target value 1E6) with the MS/MS spectra of the ten most abundant precursors from the preceding MS survey scan then acquired (resolution 17.5k, AGC target value 1E5; normalized collision energy 28 eV; minimum AGC target 1E2). Selected precursors were dynamically-excluded for 15 s.

Raw data files were processed using MaxQuant software (version 1.6.0.13) for protein identification and quantification using intensity based absolute quantification (iBAQ). iBAQ values were calculated for colon and SI by combining technical and biological replicates. A SwissProt Homo sapiens protein database (downloaded July 2017 containing 26,389 reviewed sequences) was searched with a fixed carbamidomethylation of cysteine modification and variable oxidation of methionine and protein acetylation (N-term) modifications. Protein and peptide false discovery rates were set at 1 %. The MaxQuant protein groups output file was imported into Perseus software (version 1.4.0.2) for further statistical analysis and data visualization. Common contaminant proteins and reverse sequences were removed. Intensity values were log2 transformed and Gene Ontology cellular compartment (GOCC) descriptions were added by Perseus. Protein detection was called when it was detected in at least 3 out of the 4 biological replicates of either colon or SI scaffolds. This resulted in 377 total proteins detected in these intestinal scaffolds (Supplementary Table 2).

Raman Spectroscopy

Raman imaging was conducted using a Renishaw RA816 Biological Analyzer coupled to a 785 nm laser excitation source that is reshaped using cylindrical lenses to produce a line illumination (Renishaw plc, Wotton-under-edge, UK). A total laser intensity of approximately 158mW was focused onto the sample through a 50×/NA 0.8 objective. A 1500 l/mm grating was used to disperse the laser light providing a spectral range of 0 to 2100 cm^-1 in the low wavenumber range. The RA816 series undergoes a fully automated calibration and optimization sequence to ensure optimal performance, including calibration to the 520.5 cm^-1 silicon peak. Raman imaging was conducted on colonic and small intestine sections previously embedded in paraffin wax and cut at 8 μm. Sections were mounted onto stainless steel slides, deparaffinized in xylene and rehydrated in graded alcohol and distilled water prior to Raman analysis. A total of thirty single point spectra were acquired from each histological region of both the large and small intestine (mucosa, submucosa and muscularis propria) using a 15 seconds integration time. Large Raman maps were acquired using the Renishaw StreamLine™ mode using a 4.4 μm step size and 1.1 seconds integration time per pixel. A total of 8,544 spectra were acquired across the colon map and 22,825 spectra across the small intestine map.

Prior to any analysis all spectra were pre-processed to remove all non-chemical effects of the data acquisition process. All spectra were truncated between 400 – 1800 cm^-1 which encompasses the fingerprint region where the majority of all biological signal lies. Cosmic ray removal was conducted using the width of feature and nearest neighbor methods in Renishaw’s WiRE 5 software. Spectra were then imported into MATLAB R2017a (MathWorks, Natick, MA, USA) where baseline correction to a third order polynomial was conducted using the modified polyfit method³⁷. Spectra were vector normalized and then analyzed in both WiRE 5 and MATLAB R2017a.

Point spectra obtained from each distinct histological site in the colon and small intestine (mucosa, submucosa and muscularis propria) before and after decellularization were averaged and plotted on the same axis. The most obvious differences pre-and post-decellularization were highlighted. Principal component analysis (PCA) was then carried out to ascertain whether the biochemical difference between each distinct intestinal layer post-decellularization could be observed. This is an unsupervised multivariate analysis technique that allows an effective reduction in the dimensionality of the spectral dataset and hence facilitates the identification of combinations of highly correlated variables that best describe the variance in the data. Both of these procedures were carried out in MATLAB R2017a. The large high spatial resolution Raman maps were analyzed in WiRE 5 (Renishaw plc, Wotton-under-Edge, UK) using unsupervised Multivariate Curve Resolution – Alternating Least Squares (MCR-ALS) approach for initial exploratory analysis to determine the global composition of each specimen. Using the reconstructed component curves, we were able to identify some of the components abundant within each histological layer. Direct Classical Least Squares Component analysis was then used to acquire component images of a number of known biomolecules using previously acquired reference spectra. This enabled us to ascertain the spatial distribution of known biomolecules within the full thickness of the intestinal wall.

Mechanical testing

Age-matched SI or colon scaffolds were cut into transverse strips measuring approximately 10mm wide by 15mm long and blotted dry using tissue. The thickness of each sample was measured using electronic calipers. Tape was added to the ends of each sample before they were loaded into the grips of an Instron Tensile Tester 5565, equipped with a 500N load cell (Instron, High Wycombe, UK). A pressure of 3 bar was applied by the grips to hold both ends of the scaffold in place. Extension was set at 5mm/minute until yield. Any scaffolds that slipped before yield were excluded from analysis. BlueHill 3 software (Instron) was used to generate stress-strain curves and Young’s modulus was calculated from the 0-5mm extension range, with strain normalized to an initial length of 10mm (scaffold sample length ranged between 12-18mm). Samples (n=3) were measured in duplicate (with the exception of one colon sample since the dimensions of the tissue obtained was insufficient), and the values for Young’s modulus in Extended Data Fig. 2e are averages across 3 biological replicates per group calculated from the curves.

In vitro scaffold seeding using bioreactor system

Sterilized sections of acellular intestinal scaffolds were immobilized in custom made mini platforms and placed at the bottom of 12 well tissue culture plates. When seeding scaffolds were intended for transplantation, human jejunal fibroblasts were trypsinized and resuspended in DMEM before seeding by multiple microinjections via 26g cannulae (Terumo SKU:SR+DU2619PX) into the lateral walls of the scaffolds, at a density of 1x10⁶ cells/cm² and cultured in static conditions for 3 days. Human jejunal organoids were trypsinized and seeded onto the mucosal surface of the scaffolds at a density of 1x10⁶ cells/cm². The cells were allowed to engraft for a period of 30 minutes at 37°C, before covering the whole scaffold with basal human intestinal organoid culture media. The scaffolds were maintained for another 4 days in static culture conditions, before transferring the scaffolds into perfusion plates (Amsbio #AMS.AVP-KIT-5) and connecting this to a bioreactor circuit. The bioreactor circuit consisted of a media reservoir (custom made by Chem Glassware UK Manufacturers Ltd, London) with a 0.22μm air filter, inlet and outlet tubing (Cole Parmer cat.no. 224-2081), a peristaltic pump (Cole Parmer cat.no. 224-1505) and 3-way stopcocks (Becton Dickinson UK Ltd cat.no. 394601) for media sampling at both inlet and outlet points of the circuit. For in vitro functional and histology analyses, graft cultures were maintained for 14 days in dynamic culture conditions, with media circulating at a rate of 3ml/min. For in vivo transplantation, dynamic culture times were reduced to 7 days. For vascularization experiments, HUVECs were injected at a density of 1x10⁶ cells/cm² into the lateral walls of the grafts submucosally, 24 hours before transplantation into mice. Piglet scaffold seedings were performed in a similar manner, without pre-injections of fibroblasts or HUVECs due to the lack of accuracy of injection in the thin scaffold walls.

β-Ala-Lys-AMCA peptide uptake assay

Grafts in culture (or unseeded scaffolds as controls) were transferred from the bioreactor circuit into a 6 well plate and rinsed several times with PBS. They were then incubated with fresh human organoid culture media containing 25μM β-Ala-Lys-AMCA for 2 hours at 37ᵒC. After incubation the media was removed and the graft was rinsed in cold PBS three times. The grafts (or unseeded scaffolds as controls) were fixed in 4% paraformaldehyde for 30 minutes at room temperature. Whole mount immunostaining was performed as described above, with Alexa Fluor 568 phalloidin staining to mark cellular boundaries. The samples were then imaged using a Leica SP5 inverted confocal microscope. The fluorescence signal of β-Ala-Lys-AMCA was acquired using the UV laser.

Citrulline detection assay

Graft culture supernatants were collected and citrulline levels quantified by spectrophotometry according to methods reported previously³⁸. Briefly, 20μl of each test sample was added to 20μl of water and 10μl of working urease solution, then incubated at 37ᵒC for 30 minutes. 150μl of chromogenic reagent was then added to the solution and incubated for a further 60minutes at 100ᵒC to allow color development. Absorbance was read at 520nm in a 96 well microtitre plate using a Tecan microplate reader (Infinite® M1000 PRO). Concentration was determined by comparison to a citrulline standard curve.

Disaccharidase enzyme activity assay

Grafts in dynamic culture were transferred from the bioreactor circuit into 6 well plates. The scaffolds were washed in PBS three times then incubated with a solution of 56mM sucrose in PBS (or PBS alone in control wells) for 60 minutes. Aliquots of the solution were then sampled for glucose detection using the Amplex™ Red glucose/glucose oxidase assay kit (ThermoFisher cat.no. A22189) according to manufacturer’s protocol. Briefly, 50μl of the reaction working solution was added to 50μl of the test samples in a 96 well black flat bottom microtitre plate in triplicates, and incubated in the dark for 30 minutes at room temperature. The fluorescence (excitation 535nm, emission 590nm) was measured using a Tecan microplate reader (Infinite® M1000 PRO). Concentration was determined by comparison to a glucose standard curve.

Dipeptidyl peptidase IV assay

Grafts in culture (or unseeded controls) were removed from culture and washed in PBS three times. Gly-Pro p-nitroanilide hydrochloride (Sigma cat.no. G0513) was dissolved in PBS at a concentration of 1.5mM and added to the grafts for 1 hour at 37°C before supernatant was collected. Optical density (415nm) was measured in duplicate and concentration determined by comparison to a 4-nitroaniline (Sigma cat.no. 185310) standard curve.

Barrier function assay

Piglet scaffolds were mounted into CellCrowns (Sigma cat.no. Z742381) (instead of onto custom made mini-platforms) before seeding with organoids and culture as described above. Grafts in culture (or unseeded controls) were removed from culture and washed in PBS three times before 100μl of 500μg/ml FITC-Dextran (Sigma cat.no. 46944-100MG-F) in media was added to the apical chamber of the CellCrown. 900μl of media was added to the basal chamber. Samples were incubated for 2 hours at 37°C. Media was then collected from the basal chamber into individual wells of a flat-bottomed, black-walled 96-well plate in duplicate. Presence of FITC-Dextran was determined by measuring fluorescence (excitation 485nm, emission 535nm). Percentage leakage/permeability of FITC-Dextran through the grafts (from the “luminal side” to the “serosal side”) was calculated by comparison to empty well. Percentage leakage of blank scaffold was used as baseline leakage.

Electron microscopy and micro-CT imaging

For routine SEM imaging, decellularized human scaffold samples were fixed in 2% glutaraldehyde in 0.1 M phosphate buffer and kept at 4°C for 24 hours. Samples were then washed with 0.1 M phosphate buffer and cut into segments of approximately 1 cm in length and cryoprotected in 25% sucrose, 10% glycerol in 0.05 M PBS (pH 7.4) for 2 hours, then fast frozen in nitrogen slush and freeze fractured at -160°C. Samples were then returned to the cryoprotectant solution and allowed to thaw at room temperature. After washing in 0.1 M phosphate buffer (pH 7.4), the material was fixed in 1% OsO4 in 0.1 M phosphate buffer (pH 7.4). After rinsing with distilled water, specimens were dehydrated in a graded ethanol-water series to 100% ethanol, critical point dried from CO2 and mounted onto aluminum stubs using sticky carbon tabs. Samples were coated with a thin layer of Au/Pd using an ion beam coater (Gatan UK). Images were recorded using a Jeol 7401 field emission gun scanning electron microscope.

For routine TEM, MicroCT imaging, serial block face SEM imaging and montage SEM imaging, recellularized scaffold grafts were fixed overnight at room temperature in 10% neutralized buffered formalin (Sigma) followed by a second fixation step in 2.5% glutaraldehyde/4% paraformaldehyde in 0.1 M phosphate buffer (pH7.4). The sample was post-fixed in 2% reduced osmium (4% OSO4, 1.5% K3FE(CN)6) for 60 minutes on ice and then washed in ddH20.

To check the orientation of cells on recellularized scaffolds, the sample was embedded in CYGELTM (BioStatus, Leicestershire, UK) in an Eppendorf tube and a microCT scan was performed at 4kV/3W, with no filter, 1601 projections and a pixel size of 7.3379 μm using an Xradia 510 Versa (Zeiss). The data was automatically reconstructed using Scout-and-Scan™ Control System Reconstructor software (Zeiss) and viewed in TXM3DViewer software (Zeiss). With the presence of cells confirmed, the sample was immersed in ice, the CYGEL™ washed off the sample with ddH20 and the sample trimmed to approximately 1 mm³ blocks. The trimmed blocks were then incubated in 1% aqueous thiocarbohydrazide at room temperature for 20 minutes then washed in ddH20. The blocks were incubated in 2% aqueous osmium tetroxide for 30 minutes at room temperature and washed in ddH20. This was followed by a further incubation in 1% aqueous uranyl acetate at 4ᵒC overnight. The blocks were washed in ddH20 and then incubated in Walton’s lead aspartate for 30 minutes at 60ᵒC before being dehydrated through a graded series of ethanol, infiltrated with Durcupan resin (Sigma-Aldrich) and polymerized for 48 hours at 60ᵒC.

For routine TEM, 70 nm sections were cut on a UC6 ultramicrotome (Leica), picked up on formvar-coated slot grids and imaged in a Tecnai G2 Spirit Biotwin (ThermoFisher) with an Orius CCD camera (Gatan UK). For Serial Block Face SEM, Samples were trimmed to the region of interest, mounted on an aluminum pin (Leica Microsystems) and sputter coated with 2 nm of platinum. See Supplementary Table 6 for serial block face SEM imaging conditions. For SEM montage images, after the orientation of the cells within the block was determined by microCT, 200 nm sections were cut on a Powertome ultramicrotome (RMC) and picked up on ITO-coated coverslips. The coverslips were mounted on SEM pin stubs (Agar Scientific) using a sticky carbon tab and sputter coated with 0.5 nm platinum. Sections were viewed in a Quanta FEG scanning electron microscope (ThermoFisher) using a backscattered electron (BSE) and large montage images acquired using MAPS 1.1 software (ThermoFisher). See Supplementary Table 7 for SEM montage imaging conditions. The montaged image in Fig. 3e and S3e was generated by stitching together individual images using TrakEM2, a plug-in of the FIJI framework³⁹. The montaged image in Fig. 4m were generated by stitching together individual images using MAPS 1.1 software (ThermoFisher).

Lentivirus preparation for human organoid labelling

The lentiviral vector pHIV-LUC-ZsGreen (Addgene Inc. MA, USA, Plasmid #39196, kind gift from Dr Bryan Welm, Department of Surgery, University of Utah) was used to generate a lentivirus containing both ZsGreen fluorescent protein and firefly luciferase from an EF1-alpha promoter. Human jejunal organoids were labelled by lentiviral transduction as previously reported39. Briefly, LUC-ZsGreen lentivirus was produced by co-transfecting 293T cells (ATCC) with the above plasmid along with packaging vectors PAX2 and VSV-G envelope plasmid (kind gifts from Dr Hans Clevers). Transfection was performed according to manufacturer’s instructions for 8 hours at 37°C. The medium (Opti-MEM®) was exchanged for virus collection. After 24 hours, the virus-containing medium was purified by centrifugation at 2500 rpm (4 °C) and filtered through a 0.45μm membrane and ¼ volume of PEG-itTM was added to the filtered supernatant before ultracentrifugation at 2300 x g for 30 mins at 4°C (SW28 rotor, Optima LE80K Ultracentrifuge, Beckham). The viral pellet was resuspended in 1ml of pre-cooled Opti-MEM® (Gibco), aliquoted and stored at −80 °C. Human organoids were dissociated into single cells and cultured in the presence of 250μl 2x organoid culture media and 250μl viral particles and 2.5μl of TransDuxTM. Transduction efficacy was determined measured as the proportion of cells expressing the fluorescent protein ZsGreen 72 h after transduction.

Bioluminescence imaging (BLI)

BLI was performed using an IVIS Spectrum in vivo imaging system (PerkinElmer, Waltham, MA, USA) and Living Image 4.3.1 software (PerkinElmer). Mice were injected intraperitoneal with 150 mg/kg D-luciferin (PerkinElmer) twenty minutes prior to imaging. All images were taken at field of view C or D, with automatic exposure time, pixel binning set to 8, f-stop 1 and open emission filter. This generated pseudo-colored scaled images overlaid on grey scale photographs, providing 2-dimensional localization of the source of light emission. All images were analyzed using Living Image 4.3.1 software (PerkinElmer). Regions of interest (ROI) were drawn manually and the light emission was quantified in photons s−1. ROI shapes were kept constant between images within each experiment.

In vivo transplantation models

NSG mice were anaesthetized with a 2-5% isoflurane: oxygen gas mix for induction and maintenance. The dorsum of each animal was shaved and the skin cleansed with 70% ethanol and povidone-iodine solution. For kidney capsule transplantation (n=3), each seeded graft (cultured as 1cm² patches) was carefully cut in half and gently folded to maintain the epithelial surface internally, before immediately inserting the graft under the capsule of the kidney. For subcutaneous transplantation (n=12), closed blunt scissors were used to create subcutaneous pockets bilaterally in the dorsum of each mice and one folded graft segment was inserted in each pocket. Subcutaneous Teduglutide or vehicle (PBS) was administered at a dose of 0.2 mg/kg/daily after implantation. The mice were sacrificed at 7 or 14 days for analysis. 2 hours prior to culling, each mouse was administered a dose of EdU (3ul/g of a stock solution of 10mg/ml).

3D volume rendering of kidney capsule transplant data

Serial sections were cut of the paraffin embedded sample. Odd numbered slides were stained with H&E while even numbered slides were kept unstained for further immunostaining analyses. Odd numbered H&E slides were then serially scanned (Olympus VS120 slide scanner) and the region of interest was aligned manually using Amira Software (ThermoFisher). Using the aligned slices, the kidney, scaffold, epithelial ring and lumen were segmented manually and saved as four separate label fields before generating 3D surfaces. A movie was created using Amira Animation Director.

Quantitation of epithelial lumen and vessels from in vivo grafts

Odd numbered H&E slides were scanned and images opened in QuPath software (University of Edinburgh)⁴⁰. Each slide was checked for presence of epithelial lumen and blood vessels and this was quantified as a percentage across all sections.

Statistics and Reproducibility

Statistical analysis was conducted on data from three or more biologically independent experimental or biological replicates wherever possible, as stated in the figure legends. Data distribution was assumed to be normal. Results are expressed as mean ± standard error of the mean (s.e.m.). Statistical significance was analyzed using unpaired student’s t test (two sided) between two different groups (Fig. 4j; Extended Data Fig. 2e and 5f). For analysis between more than two groups, statistical significance was analyzed using One-Way ANOVA with Dunnett’s (Fig. 1d, 1f; Extended Data 1a, 1d and 2b) or Tukey’s (Figure 1e) post hoc multiple comparison test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 were considered significant. Exact p values are stated in each relevant figure legend where appropriate. All attempts of replication were successful. Statistical analysis was carried out using Prism 8 (GraphPad Software).

Extended Data

Reporting summary.

Further information on the research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The data that support the findings of this study are available within the paper and its supplementary information files. All unprocessed western blot images and statistics are supplied as Source Data. All source data for Figs 1-4 and Extended Data Figs 1-5 are supplied in the Supplementary Dataset. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via PRoteomics IDEntifications (PRIDE) partner repository with the dataset identifier PXD019816, available for download at publication.