(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Single-cell atlas of the human neonatal small intestine affected by necrotizing enterocolitis [1]

['Adi Egozi', 'Department Of Molecular Cell Biology', 'Weizmann Institute Of Science', 'Rehovot', 'Oluwabunmi Olaloye', 'Department Of Pediatrics', 'Yale School Of Medicine', 'New Haven', 'Connecticut', 'United States Of America']

Date: 2023-05

Necrotizing enterocolitis (NEC) is a gastrointestinal complication of premature infants with high rates of morbidity and mortality. A comprehensive view of the cellular changes and aberrant interactions that underlie NEC is lacking. This study aimed at filling in this gap. We combine single-cell RNA sequencing (scRNAseq), T-cell receptor beta (TCRβ) analysis, bulk transcriptomics, and imaging to characterize cell identities, interactions, and zonal changes in NEC. We find an abundance of proinflammatory macrophages, fibroblasts, endothelial cells as well as T cells that exhibit increased TCRβ clonal expansion. Villus tip epithelial cells are reduced in NEC and the remaining epithelial cells up-regulate proinflammatory genes. We establish a detailed map of aberrant epithelial–mesenchymal–immune interactions that are associated with inflammation in NEC mucosa. Our analyses highlight the cellular dysregulations of NEC-associated intestinal tissue and identify potential targets for biomarker discovery and therapeutics.

Funding: O.O. is supported by Yale University start-up funds, Patterson Mentored Trust Research award and the CTSA Grant Number KL2TR001862 from the National Center for Advancing Translational Science (NCATS), a component of the NIH. L.K. is supported by Yale University start-up funds, Yale Program for the Promotion of Interdisciplinary Science, Binational Science Foundation award number 2019075 and NIH grants R21TR002639, R21HD102565, and R01AI171980. S.I. is supported by the Wolfson Family Charitable Trust, the Edmond de Rothschild Foundations, the Fannie Sherr Fund, the Dr. Beth Rom-Rymer Stem Cell Research Fund, the Helen and Martin Kimmel Institute for Stem Cell Research, a research grant from the Richard F. Goodman Yale/Weizmann Exchange Program, the Minerva Stiftung grant, the Israel Science Foundation grant no. 1486/16, the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme grant no. 768956 and the Chan–Zuckerberg Initiative grant no. CZF2019‐002434. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Copyright: © 2023 Egozi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

With the goal of developing better treatments for premature infants with NEC, we reconstruct a single-cell atlas of neonatal and NEC small intestine (SI) tissue with an emphasis upon cellular localizations and interactions. Using single-cell RNA sequencing (scRNAseq) data, we define transcriptional signatures and ligand–receptor interactions associated with NEC. This is combined with deconvolution of bulk RNA sequencing (RNAseq) data for cell type abundance validation and imaging mass cytometry (IMC) to define cellular interactions. Our data demonstrate that aberrant cellular interactions are associated with NEC intestinal inflammation. Specifically, we describe an increase in inflammatory macrophages and inflammatory changes in conventional and regulatory T cells accompanied by increased T-cell receptor beta (TCRβ) clonality. There is profound remodeling of NEC mucosa characterized by a decrease in top villus epithelial cells and an increase in proportion of endothelial cells and fibroblasts. All 3 cell types exhibit an increase in the transcription of inflammatory genes. Our NEC single-cell atlas identifies networks controlling intestinal homeostasis and inflammation thereby improving our understanding of NEC pathogenesis and identifying biomarkers and targets for therapeutics discovery.

NEC is a multifactorial disease involving environmental, microbial, host, and immune factors. However, despite intense research over the past several decades, the precise etiology of NEC continues to be elusive [ 6 ] and effective prevention methods or treatment options are unavailable. Dysregulation in both the mucosal immune system and the epithelial barrier are hypothesized to be associated with NEC, yet the mechanism of how these cells contribute to disease onset or progression is not clear. Previous studies have focused on specific cellular populations [ 7 , 8 ] and a comprehensive systems biology analysis is lacking.

Each year in the United States more than half a million infants are born prematurely. Necrotizing enterocolitis (NEC) is a devastating gastrointestinal (GI) complication that is associated with the degree of prematurity and with high rates of mortality and morbidity. NEC most often affects infants born at <32 weeks’ gestation and the onset of symptoms occurs 2 to 8 weeks after delivery [ 1 ]. The current incidence of NEC is 1% to 7% of all the infants admitted to the Neonatal Intensive Care Unit with prevalence rising up to 15% for the most premature infants [ 2 ]. A recent analysis in infants born prior to 29 weeks’ gestation showed a decline in all-cause mortality; however, mortality related to NEC has increased [ 3 ]. In addition to short-term complications, NEC is associated with high rates of long-term morbidity that include GI strictures, feeding intolerance, and short gut syndrome, but also significant systemic consequences such as microcephaly and neurodevelopmental delays [ 4 , 5 ].

To identify the chemokines and cytokines that could be associated with recruitment of inflammatory macrophages and lymphocytes or could be responsible for altered signaling within the NEC small intestine, we performed a ligand–receptor analysis between all pairs of cell types in our atlas ( S7 Table , Methods). To this end, we parsed a database of ligands and matching receptors [ 51 ] and defined an interaction potential between each pair of cell types as the product of the ligand expression in the sender cell type and the expression of matching receptor in receiving cell type. This potential was computed separately for the NEC cells and the neonatal cells, and the ratios of interaction potentials between NEC and neonatal samples were statistically assessed via random re-assignment of cells to the 2 groups (Methods). Consistent with increased activation of endothelial cells in our scRNAseq data, we found that NEC vascular endothelial cells up-regulated signaling to leukocytes via integrin receptors and cytokines including: macrophages via VCAM1-ITGB1 and IL6-F3 ( Fig 6B ), dendtritic cells via VCAM1-ITGB1/ITGB7, and IL6-IL6ST ( Fig 6C ), and naïve T cells via MADCAM1-ITGB7, VCAM1-ITGB1/ITGB7/ITGB2/ITGA4, and IL6-IL6R/IL6ST ( Fig 6D ). Similarly, interactions of vascular endothelial cells through VCAM1-ITGB1/ITGB7/ITGA4/ITGB2 were up-regulated in all T-cell subsets including activated and regulatory T cells ( S5C Fig ) and MADCAM1-ITGB7 were additionally up-regulated in Tregs ( S5C Fig ). NEC lymphatic endothelial cells, primarily interacted with leukocytes via cytokines and chemokines including: to macrophages via CCL3L1-CCR1 ( Fig 6E ), dendtritic cells via CXCL9/CXCL10/CXCL11-CXCR3 and CCL20-CCR6 ( Fig 6F ), and naïve T cells via CXCL9/CXCL10/CXCL11-CXCR3 ( Fig 6G ). Our ligand–receptor analysis indicated that NEC epithelial cells interact with macrophages through elevated levels of TNFSF9 ( Fig 6H ), and NK cells signals to NEC epithelial cells via IL1B-IL1R1/IL1R2 and GZMB-PGRMC1 ( Fig 6I ). Fibroblasts interact with macrophages through CXCL10-SDC4 and IL15-IL2RA/IL15RA ( Fig 6J ). Finally, regulatory T cells interacted with both macrophages and fibroblasts through IL1B-IL1R1 and TNFSF14-TNFRSF14 ( S5D Fig ). Numerous other interactions between cell types were identified ( S5E–S5G Fig ), including interaction between Treg IL1B and epitelial IL1R1 and IL1R2 and activated T cells and epithelial cells via GZMB-CHRM3 and ICAM1-EGFR ( S5E Fig ). Overall, epithelial–mesenchymal–endothelial–immune interaction were altered in NEC with increased ligand–receptor interactions between endothelial cells and immune cells, fibroblasts and immune cells, and epithelial cells and macropages and T cells and regulatory T cells.

( A ) Dot plot showing the 20 interaction types that have the highest increase (red) or decrease (blue) interaction values between neonatal (n = 3) and NEC (n = 6) samples ( Methods ). Interactions values ( S6 Table ) were computed by IMC analysis using Histocat 1.7.6.1 and 999 permutations and a p-value <0.01 [ 50 ]. Dot size corresponds to the interaction values in NEC. ( B–D ) Significantly elevated molecular interactions between vascular endothelial cells and macrophages (B), dendritic cells (C), naïve T cells (D). ( E–G ) Significantly elevated interactions between lymphatic endothelial cells and macrophages (E), dendritic cells (F), and naïve T cells (G). ( H ) Interactions between epithelial cells and macrophages. ( I ) Interactions between NK cells and epithelial cells. ( J ) Interactions between fibroblasts and macrophages. Shown are 16–25 significant interactions (q-value <0.01) with highest fold change ( Methods ). In all interaction, maps sender population is on the y-axis, receiver population is on the x-axis. (B–J) Neo = Neonatal. The data underlying this figure is available at the Zenodo repository under the following: https://doi.org/10.5281/zenodo.5813397 and in S6 and S7 Tables . IMC, imaging mass cytometry; NEC, necrotizing enterocolitis.

To define the cellular interactions in NEC small intestine, we performed a nearest neighbor analysis of IMC data using histoCAT, a tool that identifies statistically significant interactions/avoidances between cellular clusters [ 50 ]. Numerous cell type interactions were altered in NEC ( Fig 6A and S6 Table ). Consistent with up-regulation of genes associated with leukocyte recruitment on vascular endothelial cells such as SELE ( Fig 4D and 4K ), NEC tissue had increased interactions of vascular endothelial cells with DCs. Additionally, fibroblasts had increased interactions with several cellular populations including epithelial cells and monocytes/macrophages and B cells. Consistent with increase in T-cell clonality, NEC small intestine showed significant interactions between memory T cells and antigen presenting cells including DCs and macrophages. Finally, Tregs had significantly altered interactions with several other cell types including decreased interactions with CD16+ macrophages ( Fig 6A ).

To understand if cellular interactions are altered in NEC in the native state, we applied IMC [ 49 ]. In this method, formalin-fixed paraffin-embedded sections of small intestine (6 NEC and 3 neonatal samples) are incubated with a cocktail of heavy metal chelated antibodies (Methods, S5 Table ), ablated and analyzed. Clustering analysis of 20,819 cells from IMC data revealed numerous clusters including epithelial cells, vascular endothelial cells, fibroblasts, B cells, T cells, macrophages, DCs, and clusters of cells that could not be identified with the markers used (other, S5A and S5B Fig ).

Clustering of the fibroblasts demonstrated a small neuronal population and 319 fibroblasts ( S4A and S4B Fig ). The low cell number of fibroblasts was likely an artifact of the processing of the tissue for the scRNAseq analysis. Using computational deconvolution, we identified an increase in the overall fibroblasts in NEC tissue (fold change = 11.7, q = 0.04, Figs 1E and S1C ). NEC fibroblasts exhibited an increase in several inflammatory genes including IL1B, CSF2, CSF3, EREG, and CCL20 ( S4C Fig and S3 Table ).

( A ) Re-clustered atlas of the enterocyte cluster colored by crypt-villus zone. ( B ) Single-cell atlas annotated by condition. ( C ) UMAPs colored by top villus marker–APOA4, proliferation marker–MKI67, and stem cell marker–LGR5. Color bar is log10 (normalized expression). ( D ) smFISH of epithelial cells demonstrates increase in APOA4 + (white) epithelial cells towards the top of the villus. ( E ) smFISH of crypt cells, magenta dots are individual mRNAs of MKI67, green dots are individual mRNAs of LGR5 in the crypts. Blue are DAPI-stained nuclei, scale bars: 20 μm. ( F ) Estimates of the proportions of villus-crypt zones subsets based on computational deconvolution of bulk sequencing data. Each dot is a sample, proportions were renormalized over all villus-crypt zones, q-values are computed based on FDR correction for enterocytes only ( Methods ). Only samples with Spearman correlations >0.3 between the mixture data and the synthetic mixtures are shown (neonatal: n = 4, NEC: n = 6). ( G ) Representative images from Histocat 1.7.6.1 showing villus blunting in NEC compared to neonatal tissue. DNA– 191/193-intercolator (blue), SMA- smooth muscle actin (red), panCK- pancytokeratin (green). ( H ) Differential gene expression between NEC and neonatal cells for the mid-bottom villus zone. Included are all genes with sum-normalized expression above 5 × 10 −5 . Red dots are the top 20 most differentially expressed genes among the genes with q-value <0.02 and fold change above 2 or below 1/2. ( I ) TLR4 gene signature in NEC and neonatal samples. P-value calculated using two-sided Wilcoxon rank-sum test. The data underlying this figure is available at the Zenodo repository under the following: https://doi.org/10.5281/zenodo.5813397 and in S3 and S10 Tables . FDR, false discovery rate; NEC, necrotizing enterocolitis; smFISH, single-molecule fluorescence in situ hybridization.

( A ) Re-clustered atlas of the lymphatic and vascular endothelial cluster. ( B ) Single-cell atlas annotated by condition. ( C ) Top 8 markers of the markers for the cell types in A . ( D, E ) Differential gene expression between NEC and neonatal cells for vascular endothelial cells (D) and lymphatic endothelial cells (E). Included are all genes with sum-normalized expression above 10 −4 . Red genes are selected differentially expressed genes among the genes with q-value <0.02 and fold change above 2 or below 1/2. ( F, G ) GSEA of pathways enriched (red) or depleted (blue) in NEC samples compared to neonatal samples for vascular endothelial (F) and lymphatic endothelial cells (G) with q-value <0.1. (K) = Kegg pathways, (H) = Hallmark pathways. ( H ) Representative immunofluorescence images from neonatal and NEC samples stained with LYVE-1 (red) and TUNEL staining (white). Gray asterisk (*) represents apoptotic endothelial cells (Tunel + LYVE-1 + ). Scale bar: 90 μm. ( I ) Quantification of Tunel + LYVE-1 + cells. Each dot represents 1 image, 2 images/sample (neonatal: n = 4, NEC: n = 4). ( J, K ) smFISH demonstrating increase in SELE + endothelial cells in NEC. Red dots are individual mRNAs of SELE; cyan dots are individual mRNAs of CLDN5, a marker of vascular/lymphatic endothelial cells. Blue are DAPI-stained nuclei; scale bar: 10 μm. J and K are representative images from n = 2 subjects per group. The data underlying this figure is available at the Zenodo repository under the following: https://doi.org/10.5281/zenodo.5813397 and in S10 Table . GSEA, gene set enrichment analysis; NEC, necrotizing enterocolitis; smFISH, single-molecule fluorescence in situ hybridization.

Tregs have an important role in regulating mucosal homeostasis. Various reports have hypothesized different roles of Tregs in the pathogenesis of NEC [ 29 , 36 , 37 ]. Our computational deconvolution showed no difference in the proportion of Tregs between NEC and neonatal tissue ( S2E Fig ). However, DGE analysis of the Tregs revealed an increase in SELL (CD62L), CCR7, as well as SOCS1. CCR7 expression suggests that there are more recent thymic emigrants or natural Tregs in NEC mucosa. Furthermore, SOCS1 is essential for maintaining T reg suppressive function [ 38 ]( Fig 3E and S3 Table ). Interestingly, there was a reduction in the expression genes classically associated with Treg identity or supression function such as CTLA-4, TIGIT, iCOS, TNFRSF4 (encoding the Ox40 receptor found on intestinal Tregs and effector memory T cells), and IL2RA ( Fig 3E and S3 Table ). Our analysis indicates no changes in the proportion of Tregs, but they exhibit a phenotype of reduced suppressive activity and could represent impaired T-cell function in NEC.

Computational deconvolution analysis of the bulk RNAseq data indicated a nonsignificant increase in activated T cells, and a significant increase in Th1 cells and proliferating T cells and a decrease in naïve T cells ( S2E Fig , using FDR <0.25 as measure of significance). To investigate if T cells, ILCs, and NK cells were transcriptionally altered in NEC, we compared the transcriptomes of all T-cell clusters between NEC and neonatal samples. DGE showed an increase in an inflammatory signature in all subtypes except the naïve T cells that were transcriptionally similar between the groups ( Fig 3E ). The ILC cluster had an increase in CCR7, a gene associated with trafficking ( Fig 3E ), while the NK cluster had an increase in CCL3 [ 33 ], CD83, XCL, GZM, and TNF-related genes ( Fig 3E ). Similarly, the activated T-cell cluster showed an up-regulation of inflammatory and cytotoxic genes such as GZMA/B, GNLY, KLRC1, CSF2, IL23A, CXCL8; genes indicative of activation such as IL2RA; and genes downstream of IFN-γ signaling such as IRF1, GBP5, and STAT [ 34 ] ( Fig 3E ). We could not detect IL17A expression in our atlas and IL17F and IL22, cytokines also produced by Th17 cells, were only up-regulated in 1 NEC case and did not meet the threshold to be included in the DGE analysis. However, a number of other IL17 signature genes were up-regulated in activated T cells associated with NEC including CCL20, TIMP1, and BATF [ 35 ] ( Fig 3E ). We could not perform DGE for the Th1 cluster as there were very few contributing neonatal T cells. In summary, T cells in NEC show increased clonality and expression of proinflammatory genes compared to the non-NEC mucosa.

To identify if the clones expanded in NEC bind to known epitopes, we performed a search of the public clones database. Public clones have a unique amino acid or nucleotide sequence, are present across individuals, and can be easily identified in a published database ( https://vdjdb.cdr3.net/ ). The majority of the top clones observed in NEC were previously reported public clones. Our search revealed only 3 public clones that were significantly enriched in NEC. Interestingly, the amino acid sequences of 2 of these clones had an identical sequence to clones known to bind to cytomegalovirus (CMV) (CMV-1, CMV-2) ( S2D Fig and S4 Table ).

The T-cell receptor repertoire is crucial in adaptive immunity and unique sequences constitute diversity. To interogate the TCR repertoire and determine clone-size distribution and gene usage, we performed NGS of the TCRβ in NEC and neonatal SI and large intestinal (LI) samples. Overall, TCRβ clonality was increased in NEC compared to neonatal samples ( Fig 3C ). Additionally, analysis of gene usage revealed that the use of variable (V), diversity (D), and joining (J) segments differed between NEC and neonatal cases ( Fig 3D ) with an increased frequency of TRBV10 and decreased use of TRBV15, TRBJ1-4, and TRBJ2-1 in NEC compared to neonatal cases ( S2B Fig ). NEC patients had shorter CDR3β length with fewer deletions and fewer insertions than neonatal controls ( S2C Fig ), a phenomenon previously reported in IBD [ 32 ].

( A ) Re-clustered atlas of the T/NK cluster. ( B ) Single-cell atlas annotated by condition. ( C ) NGS of TCRβ candy plots where each small square represents 1 clone with the squares proportional to the number of T cells with a particular clone with quantification on the right. Each dot represents 1 tissue sample (neonatal: n = 7, NEC: n = 6, from 7 neonatal and 4 NEC patients). ( D ) PCA plot of variable (V), differential (D), and joining (J) regions use in NEC and neonatal cases (neonatal: n = 7, NEC: n = 6, from 7 neonatal and 4 NEC patients). Shaded areas are 95% confidence intervals. ( E ) Differential gene expression between NEC and neonatal T-cell populations. Red dots are selected differentially expressed genes among the genes with q-value <0.02 and fold change above 2 or below 1/2. Included are all genes with normalized expression above 10 −4 . The data underlying this figure is available at the Zenodo repository under the following: https://doi.org/10.5281/zenodo.5813397 and in S3 Table . ILC, innate lymphoid cell; NEC, necrotizing enterocolitis; NGS, next-generation sequencing; PCA, principal component analysis; TCRβ, T-cell receptor beta.

The contribution of lymphocytes to the pathogenesis or progression of NEC has been controversial, some studies have reported an increase in the abundance of T cells while others have shown a decrease [ 27 – 29 ]. Moreover, work from the Hackam group has suggested that Th17 lymphocytes are critical to NEC pathogenesis [ 30 , 31 ]. Our atlas included 6,400 T/NK cells, with 7 distinct T and NK cell subsets inculding: NK cells, innate lymphoid cells (ILCs), naïve T cells, regulatory T cells (Tregs), T helper cell 1 (Th1), proliferating T cells, and activated T cells (Figs 3A and 3B and S2A ).

( A ) Re-clustered atlas of myeloid lineages. Mϕ–macrophages. ( B ) Single-cell atlas annotated by condition. ( C ) Top 8 markers of the myeloid cell subtypes. ( D ) Differential gene expression between NEC and neonatal macrophages. ( E ) Differential gene expression between NEC and neonatal DCs. Red dots (D, E) are selected differentially expressed genes among the genes with q-value <0.02 and fold change above 3 or below 1/3. Included are all genes with sum-normalized expression above 10 −4 . ( F ) GSEA of pathways enriched (red) or depleted (blue) in NEC samples compared to neonatal samples for macrophages and DCs (q-value <0.2). (K) = Kegg pathways, (H) = Hallmark pathways. ( G ) Estimates of the proportions of distinct myeloid cell subsets based on computational deconvolution of bulk sequencing data. Each dot is a sample, proportions were renormalized over all myeloid cells, q-values are computed based on FDR correction for myeloid cells only ( Methods ). Gray lines are medians, black/pink boxes are 25–75 percentiles. Only samples with Spearman correlations >0.3 between the mixture data and the synthetic mixtures are shown (neonatal: n = 4, NEC: n = 6). The data underlying this figure is available at the Zenodo repository under the following: https://doi.org/10.5281/zenodo.5813397 and in S3 and S10 Tables . DC, dendritic cell; FDR, false discovery rate; GSEA, gene set enrichment analysis; NEC, necrotizing enterocolitis.

Our atlas included 11,308 cells and revealed 8 main cell clusters representing macrophages, dendritic cells (DCs), B cells, T/Natural Killer (NK) cells, vascular/lymphatic endothelial cells, fibroblasts, enteroendocrine, and other enterocytes populations ( Fig 1B–1D ). Each cluster exhibited distinct gene expression markers ( Fig 1D and S2 Table ). We verified the stability of the expression signatures by reconstructing them based on subsampled patients and cells ( S1A and S1B Fig ). Our atlas enabled exploring the detailed gene expression changes in distinct cell subsets. Furthermore, bulkseq analysis followed by computational deconvolution facilitated determination of major cluster abundances (Figs 1E and S1C ) that revealed a substantial increase in the proportions of fibroblasts and a decrease in the proportions of enterocytes. Among the remaining cells, we identified a significant increase in the proportions of T/NK cells, vascular/lymphatic endothelial cells and a decrease in the proportions of DCs and B cells (Figs 1E and S1C , using false discovery rate (FDR) <0.25 as measure of significance).

We performed scRNAseq on 11 subjects (6 NEC and 5 neonatal samples, Fig 1B and 1C ). This analysis enabled the extraction of gene expression signatures of the diverse epithelial, stromal, and immune cell populations. While single-cell atlases are powerful for extraction of expression signatures, estimation of population proportions from such datasets can be skewed due to differential viability of extracted cells [ 10 ]. To enable precise estimation of population proportions, we performed computational deconvolution of bulk RNAseq (6 NEC and 4 neonatal samples that passed threshold, Methods) data based on the clusters identified in the scRNAseq dataset ( Fig 1A ). We implemented next-generation sequencing (NGS) of TCRβ to identify clonality changes associated with NEC (4 NEC and 7 neonatal samples). Finally, we utilized IMC of 6 NEC and 3 neonatal samples to define niche cellular interactions enriched in NEC ( Fig 1A ). Overall, we generate a resource atlas that can enable the exploration of ligand–receptor interactions between any pairs of cell types.

( A ) Experimental layout—human small intestinal tissues from neonates and NEC patients were harvested and used for scRNAseq, bulk RNAseq, smFISH, IMC, and NGS of the TCRβ. Samples used in various experiments listed on the right-hand side ( S1 Table ). ( B ) Single-cell atlas annotated by cell type. ( C ) Single-cell atlas annotated by condition. ( D ) Top 6 markers of the markers of the cell types in B . ( E ) Estimates of the proportion of enterocytes, fibroblasts, dendritic cells, macrophages, B cells, T-NK cells, enteroendocrine cells, and vascular/lymphatic endothelial cells based on computational deconvolution of the bulk RNAseq using the atlas single-cell populations ( Methods ). Each dot is a sample, fractions of enterocytes and fibroblasts normalized to the sum of cell fractions, remaining fractions normalized to the sum of all cells after excluding fibroblasts and enterocytes; q-values are computed based on FDR correction for all cell populations in the full atlas ( Methods ). Gray lines are medians, black/pink boxes are 25–75 percentiles. Only samples with Spearman correlations >0.3 between the mixture data and the synthetic mixtures are shown (neonatal: n = 4, NEC: n = 6). The data underlying this figure is available at the Zenodo repository under the following: https://doi.org/10.5281/zenodo.5813397 and in S10 Table . FDR, false discovery rate; IMC, imaging mass cytometry; NEC, necrotizing enterocolitis; NGS, next-generation sequencing; RNAseq, RNA sequencing; scRNAseq, single-cell RNA sequencing; smFISH, single-molecule fluorescence in situ hybridization; TCRβ, T-cell receptor beta.

To characterize the cell states associated with NEC, we analyzed intestinal tissues from 19 patients with NEC and from 13 neonates that underwent intestinal surgery for non-NEC-related conditions. The samples were used in various assays listed in S1 Table and Fig 1A . The infants with NEC were significantly younger than the neonatal subjects (median gestational age at birth of 28 weeks in NEC versus 37 weeks in the comparison group, p < 0.001). However, the postnatal age of patients at the time of surgery was similar between the 2 groups (median age of 24.5 days in NEC versus 14 days in neonatal group, p = 0.089). Finding appropriate patients to serve as controls for comparison to samples from patients with NEC is challenging because healthy premature infants do not undergo intestinal surgery. Other controls beyond non-NEC-related neonatal tissue that have been historically used include fetal tissue, tissue from infants with spontaneous ileal perforation (SIP), and tissue obtained from re-connection surgery in infants recovered from NEC. Fetal tissue has not been exposed to an abundant microbiome and is likely immunologically different from that of postnatal samples. Infants with SIP, although gestationally age matched to those with NEC, usually undergo surgery in the first week of life and may have potential SIP-specific immune dysregulation [ 9 ]. Samples obtained from patients who have recovered from NEC (post-NEC surgeries) are obtained after prolonged periods of parenteral nutrition and antibiotic use both of which can drastically alter the mucosal tissue. As such, we opted to use non-NEC-related neonatal samples as the comparison group as the best chronologically matched tissue without known immune disturbances.

Discussion

The pathogenesis of NEC remains poorly understood and to date, few studies utilize single-cell approaches to describe the altered interactions among various intestinal cells [9]. Here, we systematically explored the changes in cell proportions and expression signatures in NEC using single-cell approaches. We present a single-cell atlas with spatial resolution of the small intestine from neonates with and without NEC that identified 8 distinct cellular populations: macrophages, DCs, B cells, T/NK cells, vascular/lymphatic endothelial cells, fibroblasts, enteroendocrine cells, and enterocytes. Our data exposes substantial changes in both abundances and transcriptional profiles of immune and non-immune populations in NEC small intestine.

NEC mucosa was marked with inflammatory changes in numerous innate immune populations. We observed an increased proportion of inflammatory macrophages that expressed inflammatory cytokines associated with lymphocyte recruitment including IL1A and IL1B and TNFA [52]. Furthermore, pathway analysis was suggestive of inflammatory responses including TNFa signaling via NFkB, cytokine receptor interactions, NLR and TLR signaling. NLR and TLR activation have been implicated in NEC pathogenesis both in human disease and in murine models, with the current study providing a context that macrophages are responding to these signals [23–25] and may be interacting with both immune and non-immune cells within the small intestine.

From the adaptive immune side, consistent with previous work that implicated Th17 T cells in the pathogenesis of NEC [30,31,53], we observed an up-regulation of genes classically expressed by Th17 cells (BATF and CCL20 [54,55]) and additionally identified increased cytotoxic activity in NEC-activated T cells. Tregs are critical for maintaining homeostasis and play a major role in intestinal inflammation through cell–cell interactions and secreted factors. Previous studies showed conflicting results regarding the change in proportions of Tregs in NEC [36,37,53,56]. Similar to a recent study demonstrating heterogeneity between NEC samples [11], we observed large variability in the proportion of Tregs in NEC tissue. Our findings were all indicative of a decrease in Treg suppressive abilities including down-regulation of genes associated with Treg identity or function (CTLA4, ICOS, TIGIT, and TNFRSF4, Fig 3E), overall alterations in the interactions of Tregs with other cell types including a reduction in the interactions with CD16+ macrophages (Fig 6A and S6 Table), and up-regulation of IL1B-IL1R interactions (S5D and S5E Fig) previously shown to render Tregs less suppressive [57]. As such, altered Treg function likely contributes to the progression of inflammation in NEC.

Surprisingly, major differences in TCRβ clonality, VDJ use, CDR3β length, and presence of shared clones were evident in NEC compared to controls. Specifically, we noted an increase in the frequency of public clones known to bind CMV antigens. It is unlikely related to the presence of active intestinal CMV infection, as these clones were present in samples from all patients with NEC and in majority of healthy fetuses [58]. Public clones are known to be promiscuous [59] and these could represent clones against bacterial or self-antigens that are similar to peptides found on CMV. We also noted shorter CDR3β length with fewer deletions and fewer insertions in NEC compared to the neonatal non-NEC comparison group, similar to what has been reported in IBD [32]. In the future, analysis of paired alpha/beta TCR chain analysis could further define specific clone/antigen interaction to characterize the role of clonal expansion in NEC pathogenesis.

Our investigation also revealed inflammatory signaling in non-immune cell types. Villus-top enterocytes were significantly reduced in NEC. A similar phenotype was observed in a “gut on a chip” model of NEC [60,61], as well as in a recent study of samples from patients with NEC [62]. Enterocytes that were present in NEC exhibited an inflammatory phenotype, with up-regulation of chemokines involved in the recruitment of immune cells, and enrichment of ligand–receptor interactions between epithelial cells and immune cells. Furthermore, this was associated with an increased epithelial STAT3 phosphorylation and up-regulation of STAT3-dependent genes including REG1A [63], REG1B, LCN2 [64], and PLA2G2A, antimicrobial peptides that are up-regulated upon epithelial damage [65] and shown to be up-regulated in IBD [43–46]. Similarly, fibroblasts demonstrated an increase in inflammatory genes (IL1B and CSF3) and extensive interactions with immune cells both by ligand–receptor interactions and by IMC neighborhood analysis. Previous work had identified inflammatory fibroblasts as potential drivers of IBD [66]. It is intriguing to postulate that a similar mechanism might be at play in NEC.

The proportion of inflammatory endothelial cells was also elevated in NEC-affected small intestine. Vascular endothelial cells up-regulated genes involved in vasoconstriction (EDN1), clotting (F3, SERPINE1) angiogenesis (CXCL1, CXCL8, and ALPN), proliferation [67], leukocyte recruitment (SELE), leukocyte adhesion (ICAM1, Fig 4D), and inflammatory pathways (TNFa and IFNγ signaling, Fig 4F). Our interaction analyses further revealed elevated ligands–receptor pairs involving the vascular adhesion molecules VCAM1 and MADCAM1 (Fig 6B–6D and S7 Table). Integrins interacting with VCAM1 and MADCAM1 include the α4β7(ITGA4/ITGB7) complex that is up-regulated on activated lymphocytes, and innate cells, leading to leukocyte extravasation into intestinal high endothelial venules [68–71]. Up-regulation of α4β7 and α4β1 is thought to be pathogenic in IBD [72] and these ligand receptor interactions were enriched for in NEC samples. Blockade of α4β7 with Vedolizumab, a humanized monoclonal antibody against α4β7, is effective for induction and maintenance of remission in IBD in multiple studies [73–77]. It remains to be seen whether the long latency of action of this drug would still render it effective to the prevention and treatment of NEC. Overwhelming inflammation in endothelial cells may lead to dysfunction or death and contribute to intestinal inflammation. Indeed, pathway analysis of DGE in NEC endothelial cells showed apoptosis as one of the up-regulated pathways, a finding which we confirmed by TUNEL staining.

NEC is a progressive disease and our atlas captured dysregulation in the small intestine in the subset of infants that required surgery. One limitation of our study is that it did not include infants who recovered from NEC with medical therapy. Validation that some of the markers identified here can also be detected in the blood of infants with NEC and their identification in infants with various stages of NEC would be interesting. Additionally, understanding the role of the dysbyosis in NEC pathogenesis by including biopsy-associated microbiome data could be the focus of future studies.

In summary, we provide a comprehensive atlas of cellular dysregulation in NEC, accompanied by localization and ligand–receptor interaction analysis. Our study demonstrates profound inflammatory changes in NEC small intestine with increase in IL1β and TNFa producing macrophages, inflammatory signature in T cells with decreased suppressive signatures in Tregs accompanied by inflammatory changes in endothelial, epithelial, and fibrobast cells. We also identify a number of potential interactions such as MADCAM1-α4β7 between endothelial and T cells that could represent future therapeutic targets for NEC treatment. Our data provides a resource for future biomarker and therapeutic development in NEC.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002124

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.