Intestinal neutrophil extracellular traps promote gut barrier damage exacerbating endotoxaemia, systemic inflammation and progression of diabetic retinopathy in type 2 diabetes

Abstract

Aims/hypothesis: Within the small intestine, neutrophils play an integral role in preventing bacterial infection. Upon interaction with bacteria or bacteria-derived antigens, neutrophils initiate a multi-staged response of which the terminal stage is NETosis, formation of protease-decorated nuclear DNA into extracellular traps. NETosis has a great propensity to elicit ocular damage and has been associated with diabetic retinopathy and diabetic macular oedema (DME) progression. Here, we interrogate the relationship between gut barrier dysfunction, endotoxaemia and systemic and intestinal neutrophilia in diabetic retinopathy.

Methods: In a cohort of individuals with type 2 diabetes (n=58) with varying severity of diabetic retinopathy and DME, we characterised the abundance of circulating neutrophils by flow cytometry and markers of gut permeability and endotoxaemia by plasma ELISA. In a mouse model of type 2 diabetes, we examined the effects of diabetes on abundance and function of intestinal, blood and bone marrow neutrophils, gut barrier integrity, endotoxaemia and diabetic retinopathy severity. Pharmacological inhibition of NETosis was achieved by i.p. injection of the peptidyl arginine deiminase 4 inhibitor (PAD4i) GSK484 daily for 4 weeks between 6 and 7 months of type 2 diabetes.

Results: In human participants, neutrophilia was unique to individuals with type 2 diabetes with diabetic retinopathy and DME and was accompanied by heightened circulating markers of gut permeability. At late-stage diabetes, neutrophilia and gut barrier dysfunction were seen in db/db mice. The db/db mice exhibited an increase in stem-like pre-neutrophils in the intestine and bone marrow and a decrease in haematopoietic vascular reparative cells. In the db/db mouse intestine, enhanced loss of gut barrier integrity was associated with elevated intestinal NETosis. Inhibition of NETosis by the PAD4i GSK484 resulted in decreased abundance of premature neutrophils in the intestine and blood and resulted in neutrophil retention in the bone marrow compared with vehicle-treated db/db mice. Additionally, the PAD4i decreased senescence within the gut epithelium and yielded a slowing of diabetic retinopathy progression.

Conclusions/interpretation: Severity of diabetic retinopathy and DME were associated with peripheral neutrophilia, gut barrier dysfunction and endotoxaemia in the human participants. db/db mice exhibited intestinal neutrophilia, specifically stem-like pre-neutrophils, which was associated with elevated NETosis and decreased levels of vascular reparative cells. Chronic inhibition of NETosis in db/db mice reduced intestinal senescence and NETs in the retina. These changes were associated with reduced endotoxaemia and an anti-inflammatory bone marrow milieu with retention of pre-neutrophils in the bone marrow and increased gut infiltration of myeloid angiogenic cells. Collectively, PAD-4i treatment decreased gut barrier dysfunction, restoring physiological haematopoiesis and levels of haematopoietic vascular reparative cells.

Keywords: Diabetes; Diabetic retinopathy; Gut permeability; NETosis; Neutrophil; PAD4.

Fig. 1

Individuals with type 2 diabetes with diabetic retinopathy and DME exhibit peripheral neutrophilia and elevated NLR associated with elevated circulatory markers of gut permeability and endotoxaemia. In a cohort of individuals with type 2 diabetes with varying severity of diabetic retinopathy/DME and age- and sex-matched healthy control individuals, we assessed the abundance of circulating neutrophils (CD45⁺CD15⁺SSC^hi) by flow cytometry. (a) Quantification of neutrophil abundance in individuals with type 2 diabetes, irrespective of eye disease severity, compared with healthy control individuals. (b, c) Quantification of neutrophils (b) and NLR (c) in individuals with type 2 diabetes stratified by severity of eye disease. (d–f) Quantification of FABP2 (d), LBP (e) and PGN concentrations (f) as determined by ELISA. (g) Bivariate correlations of human neutrophil abundance and plasma FABP2, LBP and PGN. Each data point is indicative of one independent measure from one participant (HC, n=59; T2D_{No DR}, n=27; T2D_{NPDR without DME,} n=4; T2D_{PDR without DME,} n=6; T2D_{NPDR with DME,} n=12; T2D_{PDR with DME,} n=9). *p<0.05, **p<0.01, ***p<0.001. DR, diabetic retinopathy; HC, healthy control; Neu, neutrophils; T2D, type 2 diabetes

Fig. 2

Expansion of the central and small intestine neutrophil compartments occurs in mice with type 2 diabetes but not in those with type 1 diabetes. (a) Schematic detailing timeline of studies comparing the mouse models of type 1 diabetes (Akita) and type 2 diabetes (db/db) with 10 months of respective diabetes duration along with secondary and tertiary cohorts of db/db mice euthanised at 4 and 7 months of type 2 diabetes duration. We used flow cytometry to assess the abundance of neutrophils (CD45⁺CD11b⁺Ly6G⁺) in intestinal and central compartments. (b–d) Quantification of neutrophil abundance in the peripheral blood (b), bone marrow (c) and small intestine (d) of Akita (type 1 diabetes) and db/db (type 2 diabetes) mice with 10 months of respective diabetes duration and age-matched WT mice. (e–g) Quantification of neutrophil abundance in the peripheral blood (e), bone marrow (f) and small intestine (g) of db/db (type 2 diabetes) mice with 4, 7 and 10 months of diabetes duration and age-matched WT mice. Each data point represents one independent observation from one experimental animal (n=4–18 per group). *p<0.05, **p<0.01, ***p<0.001. T1D, type 1 diabetes; T2D, type 2 diabetes

Fig. 3

db/db mice exhibit heightened intestinal NETosis throughout diabetes duration. To assess whether mice with type 2 diabetes mice experience elevated NETosis in the small intestine, we performed confocal IF microscopy of FFPE tissue sections from db/db mice with 4 and 7 months duration of diabetes. We identified intestinal neutrophils as being Ly6G⁺NIMP-R14⁺ cells (a). We found neutrophils (yellow) localised in the intra-epithelial space (a; scale bar, 10 µm) and between epithelial and endothelial basement membranes, and between intestinal crypts as indicated by collagen IV staining (b; scale bar, 100 µm). Variations in NET structure were demonstrated and found to be comprised of one or more cells, each white arrow indicates a cell body (c; scale bar, 10 µm). (d–g) Representative images of Cit-H3 co-staining with the neutrophil markers MPO and NIMP-R14 at 4 months are shown (d; scale bar, 100 μm [low power] and 25 μm [high power]), together with quantification of Cit-H3 MFI per total cells (e), ratio of NETotic neutrophils (CitH3⁺MPO⁺ cells / CitH3⁻MPO⁺ cells) (f) and NIMP-R14 MFI per total cells (g). (h–k) Representative images of Cit-H3 co-staining with the neutrophil markers MPO and NIMP-R14 at 7 months are shown (h; scale bar, 100 μm [low power] and 25 μm [high power]), together with quantification of Cit-H3 MFI per total cells (i), ratio of NETotic neutrophils (CitH3⁺MPO⁺ cells / CitH3⁻MPO⁺ cells) (j) and NIMP-R14 MFI per total cells (k). Each data point represents the average of three or four randomly dispersed images taken from the tissue of one experimental animal (n=4 mice per group). *p<0.05, **p<0.01, ***p<0.001. BM, basement membrane; Cr, crypt

Fig. 4

db/db mice exhibit loss of gut barrier integrity at 10 months of type 2 diabetes. Using confocal microscopy and western blot, we assessed the expression of epithelial- and endothelial-specific gut barrier junction proteins that are known to decrease (p120-catenin, ZO-1 and VE-cadherin) or increase (PV-1) with gut barrier permeability. (a–c) Representative images (a) of epithelial p120-catenin and ZO-1 staining, together with quantification of p120-catenin (b) and ZO-1 (c) MFI. (d–f) Representative images of endothelial VE-cadherin and PV-1 staining (d), together with quantification of VE-cadherin (e) and PV-1 (f) MFI. (g–i) Images of western blots indicating expression of E-cadherin, PV-1 and β-actin (g), and quantification of E-cadherin (h) and PV-1 (i) expression, each normalised to β-actin loading control. (j) Representative images of retinal VE-cadherin and acellular capillary staining. In VE-cadherin images, white arrows indicate gaps in staining. In acellular capillary images, black arrows indicate individual acellular capillaries. (k) Enumeration of retinal acellular capillaries. (l) Correlation between intestinal neutrophil abundance and number of acellular capillaries. In IF experiments, each data point represents the average of three randomly dispersed images taken from the tissue of one experimental mouse (n=3 mice per group). In western blot analysis, each data point represents one independent observation from one experimental mouse (n=3 mice per group). *p<0.05, **p<0.01, ***p<0.001. Scale bars, 100 µm. ROI, region of interest

Fig. 5

Inhibition of PAD4 reduces retinal NETosis and slows progression of diabetic retinopathy in conjunction with decreasing gut leakage and neutrophilic inflammation. (a) To assess the impact of global NETosis inhibition on diabetic retinopathy pathogenesis, gut permeability and systemic inflammation, male db/db mice with 6 months of type 2 diabetes duration (7 months age) were randomly divided into two groups (n=5 each); age-matched WT mice served as non-diabetic, non-injected controls. Over the course of 4 weeks, starting at 6 months of type 2 diabetes duration, db/db mice received daily i.p. injections (100 μl) of either vehicle (db/db-Veh, 0.9% normal saline) or the selective, reversible PAD4i GSK484 (db/db-PAD4i, 4 mg/kg in 0.9% normal saline). (b) Representative images are shown: retinal Cit-H3 and MPO; retinal acellular capillaries, the hallmark histopathological characteristic of murine diabetic retinopathy; and IDO1, a marker of intestinal inflammation. (c–f) Quantification of retinal Cit-H3 expression (c), retinal acellular capillaries (d), intestinal IDO1 (e) and senescence-associated β-galactosidase (f). (g–i) Quantification of plasma FABP2 (g), PGN (h) and neutrophil elastase (i) as determined by ELISA. (j) Quantification of plasma 4 kDa FITC–dextran concentration. Each data point represents one independent observation from one experimental mouse (n=3–5 mice per group). *p<0.05, **p<0.01, ***p<0.001. Scale bars, 50 µm. mo, months; NE, neutrophil elastase; SA-β-gal, senescence-associated β-galactosidase; T2D, type 2 diabetes; Veh., vehicle

Fig. 6

PAD4 inhibition decreased the infiltration of mature neutrophils into the small intestine and retention of pre-neutrophils in the bone marrow. We used flow cytometry to assess the impact of PAD4 inhibition on the abundance of neutrophil subtypes and vascular reparative cells within the peripheral blood, bone marrow and small intestine. (a–d) Quantification of the abundance of total neutrophils (a), pre-neutrophils (b), immature neutrophils (c) and mature neutrophils (d) within the peripheral blood. (e–l) The quantification was replicated for the bone marrow (e–h) and the small intestine (i–l). Each data point represents one independent observation from one experimental mouse (n=5 per group). *p<0.05, **p<0.01, ***p<0.001

Fig. 7

The effect of PAD4 inhibition on functionality of diabetic neutrophils. Flow cytometric analyses displaying community dynamics and function with t-distributed stochastic neighbour embedding (tSNE) plots and histograms. Data are presented for peripheral blood, bone marrow and small intestine neutrophil populations. tSNE plots were created with all markers in the neutrophil gating strategy (CD45, CD11b, Gr-1, cKit, Ly6-G, CXCR2 and CXCR4), markers of activation/degranulation (CD63, CD107a) and a marker of migration (CD62L). Histograms display the geometric MFI of functionality-related surface markers. (a) In the peripheral blood, pre-neutrophils exhibited reduced CXCR2 (black arrow), CD63 (blue arrow), CD107a (red arrow) and CD62L (magenta arrow) in db/db-PAD4i mice. Mature neutrophils of db/db-PAD4i mice exhibited decreased CXCR2 (black arrow) and elevated CD62L (magenta arrow) expression. (b) In the bone marrow, mature neutrophils of db/db-PAD4i mice exhibited reduced CXCR2 (black arrow) and CD62L (magenta arrow) expression. (c) In the small intestine, pre-neutrophils of db/db-PAD4i mice exhibited reduced CXCR2 (black arrow), CD62L (magenta arrow) and CD107a (red arrow) expression. Immature and mature neutrophils of db/db-PAD4i mice exhibited reduced CD63 (blue arrow) and CD107a (red arrow) expression