Impaired Barrier Function and Autoantibody Generation in Malnutrition Enteropathy in Zambia

2.1. Clinical Procedures - Adults

In adult participants, following an overnight fast, blood and urine samples were collected and then 100 ml of test solution A was given by mouth. After exactly 3 h, further urine and blood samples were collected. 1 ml of 20% chlorhexidine was added to urine samples. Blood samples were collected into EDTA, plain and trace element-free lithium heparin tubes (Vacutainer, Becton Dickinson; and TekLab respectively). Blood samples were kept in the dark for 20 min then centrifuged (537g, 15 min). Test solution A contained 1 g l-rhamnose (Sigma, Poole, UK), 5 g lactulose (as lactulose syrup USP) per dose, diluted in water to 100 ml.

The next day, after an overnight fast, urine and blood samples were collected as above. After 2–3 min of observation and oxygen saturation recording, intravenous sedation (2.5–10 mg diazepam, 50–100 mg pethidine) was given and endoscopy performed using Pentax EG2990i gastroscopes (Pentax, Slough, UK). 2 ml of gastric contents were aspirated from the stomach for pH measurement, and 0.5–3 ml intestinal fluid was aspirated slowly from the duodenal lumen. Three biopsies were collected into normal saline, orientated under the dissecting microscope (Swift Optical Instruments, Schertz, TX at 10 × magnification) in the endoscopy unit, the villus configuration noted, and the biopsies fixed in formal saline. After 3 h, further blood and urine samples were collected, and the exact time noted.

2.2. Endoscopy (Children)

Children were evaluated on the day of endoscopy by an experienced paediatrician (BCA) to confirm fitness for endoscopy, and then fasted from 0600 after a feed of F100, until 1000, the time of the endoscopy. Endoscopy was carried out using a Pentax 2490k paediatric gastroscope (8 mm external diameter) under sedation by an anaesthetist using ketamine-based protocols. Two biopsies were orientated and fixed as above. Lactase deficiency was tested in an additional biopsy for using the Biohit Lactose Intolerance Quick Test kit (Biohit, Ellesmere Port, UK). Test solution C, for children, contained 0.2 g rhamnose and 1 g lactulose per dose, made up to 50 ml, and was instilled down the endoscope as previously described (Kelly et al., 2004).

2.3. Clinical Procedures (Children in the Community)

Absorption and permeability testing was carried out using a very similar protocol to that used in adults, except that the period of fasting was limited to 1 h and that test solution C was taken orally. HIV testing was not carried out on these children.

2.4. Morphometry

Biopsies were processed, paraffin embedded, sectioned into 3 μm sections and stained with haematoxylin and eosin. Morphometry was carried out using NanoZoomer Digital Pathology (Hamamatsu Photonics, Hamamatsu, Japan) as previously described (Kelly et al., 2004, Kelly et al., 2016) (Supplementary Fig. S1). The following measurements were made: villus height (VH) and width (VW), crypt depth (CD), villus mucosal perimeter (VP) as a measure of epithelial surface area and related to length of muscularis mucosae, and villus cross-sectional area (VA) as a measure of villus volume. VH/CD and villus surface area/volume ratio (SAVR) (Kelly et al., 2016) were derived from these. Claudin-4 immunostaining was performed on 3 μm sections (see Supplementary material).

2.5. Sugar Analysis

Urine samples were collected at 3 h in adults, and 60 min in children. Lactulose and rhamnose were assayed by mass spectrometry in urine samples as previously described (Faubion et al., 2016).

2.6. Biomarkers

To estimate the severity of translocation, we measured plasma lipopolysaccharide (LPS) and carried out PCR for bacterial 16S ribosomal DNA as direct markers of translocation (see Supplementary data), while serum lipopolysaccharide binding protein (LBP), soluble CD14 (sCD14), CD163 and C-reactive protein (CRP) were used as indirect markers of the host response to translocated bacterial molecules (Table 2). Quantification of LPS was by the Pyrochrome Limulus Amoebocyte Lysate (LAL) assay (Associates of Cape Cod, Liverpool, UK). Serum LPS binding protein, CRP, sCD14, CD163, IGF-1 and IGFBP3 were measured by ELISA (R&D systems, Abingdon, UK: RRID:SCR_006140). DNA was isolated from plasma using the QIAamp DNA Mini Kit (Qiagen: RRID:SCR_008539), according to the manufacturer's instructions, then analyzed by PCR (see Supplementary material). Total immunoreactive GLP-2 was measured in fasted serum by ELISA (Millipore, St Charles, MO: RRID:SCR_008983). Intestinal FABP was measured by ELISA (Hycult, Cambridge Bioscience, Cambridge, UK: RRID: SCR_002245). Protein was extracted from duodenal aspirates using a lysis buffer made up of 2-mercaptoethanol and Laemmli buffer, then Western blots were probed for TFF3 using mouse monoclonal antibody to TFF3 (Abcam, Cambridge, UK: RRID: SCR_012931) and detected using VECTASTAIN anti-mouse IgG Biotinylated antibody (Vector laboratories, Peterborough, UK: RRID:SCR_000821). All assays were conducted in duplicate.

2.7. Coeliac Serology

Tissue transglutaminase IgA antibodies were measured in serum using the Orgentech ELISA kit (Launch Diagnostics, Longfield, UK) and the Quanta Lite ELISA kit (Inova Diagnostics, San Diego, USA). Antibodies to deamidated gliadin peptides were measured using the Quanta Lite Gliadin IgGII kit (Inova). The Orgentech ELISAs were run in both the Lusaka and Mayo Clinic laboratories (ρ = 0.88; p < 0.0001); the Inova TTG and DGP ELISAs were run only in the Mayo Clinic.

2.8. Data Analysis

Data analysis was carried out using Stata 13 (Stata Corp, College Station, TX). Spearman's rank correlation coefficient and the Kruskal-Wallis test were used for hypothesis testing as most variables were not normally distributed.

2.9. Role of the Funding Source

The funding sources played no role in the decision to publish, data analysis, or in the drafting of the manuscript. The corresponding author had free access to all data.

3.1. Morphological Severity of Enteropathy

Children hospitalised with SAM, and all adults, underwent endoscopy to collect duodenal biopsies. Children from the community did not have endoscopy. All biopsies were examined immediately after endoscopy using a binocular microscope and all demonstrated evidence of villus morphological change, with ridges and convolutions and absence of finger-like villi. This corresponded to histological evidence of villus blunting and infiltration of the mucosa by chronic inflammatory cells (Fig. 1). Villus height (VH), crypt depth (CD), villus width (VW), and VH:CD ratio were measured satisfactorily in biopsies from 41 adults and 22 children; they did not differ significantly between adults with EE and children with SAM, irrespective of HIV status (Table 1). However, two biopsies from children with SAM showed total villus atrophy (Fig. 1); both these children died during their admission. One other child died on the ward; this child had partial villus atrophy with a measured villus height (VH) of 198 μm and a villus height:crypt depth ratio of 1.2. Overall, the three children who died in hospital had lower VH (median 83 μm; IQR 74–198) than those who survived (216 μm; 175–246; P = 0.04) but crypt depth (CD) did not differ (170 vs 156 μm; P = 0.32). SAVR was reduced in HIV-infected children with SAM, but otherwise HIV status had little impact on mucosal morphology (Table 1). Epithelial breaks (Fig. 1, Fig. 2) were seen in all biopsies from children with SAM and in 53/54 (98%) of 54 (98%) adults evaluated. The frequency of breaks was unrelated to villus height, crypt depth or other morphometric parameters (data not shown).

Of 17 children with SAM from whom a biopsy was available for lactase testing, 7 (41%) were lactase deficient (Table 1). Villus height was markedly reduced in 3 children with lactase deficiency whose morphometry data were available (median 83 μm; IQR 74–143) compared to 8 children without lactase deficiency (median 201 μm; IQR 170–230; P = 0.02).

3.2. Claudin 4 and E-Cadherin Immunostaining

To clarify the significance of these breaks, we performed immunostaining for the tight junction protein claudin-4 and the adherens junction protein E-cadherin (Fig. 2). Overall, claudin-4 expression was reduced in all biopsies, compared to previous experience in healthy tissue (Turner, 2009, Gunzel and Yu, 2013), and E-cadherin immunostaining was disrupted at the sites of epithelial breaks. However, claudin-4 immunostaining was markedly increased and disorganised at the villus tip and at sites of epithelial breaks (Fig. 2).

3.3. Biomarkers of Mucosal Integrity

Lactulose and rhamnose, measured as percentage excretion of the oral dose given and as a urinary ratio, were used to measure intestinal permeability (Faubion et al., 2016). Failed tests occurred in 65% of the children with SAM and 30% of the children in the community; this was largely due to inability to pass urine in the 60 min specified in the protocol. Lactulose recovery and the lactulose:rhamnose (LR) recovery ratio were greater in malnourished children than in adults or community children (Table 2). The LR ratio was inversely correlated with villus surface area in adults and children in whom both morphometric measurements and satisfactory LR measurements were obtained (ρ = − 0.35, P = 0.02; n = 47), but with no other morphometric measurement. Rhamnose recovery was positively associated with villus surface area:volume ratio (ρ = − 0.40, P = 0.005; n = 47) but not with other morphometric parameters. Intestinal type fatty acid binding protein (FABP) was increased in all three groups of children, especially children with SAM, compared to adults (Table 2).

3.4. Markers of Microbial Translocation

Lipopolysaccharide (LPS) concentrations were higher in plasma which was also positive for bacterial 16S rRNA gene DNA (median 137, IQR 61–257 EU/ml) compared to plasma in which no bacterial DNA was detected (median 79, IQR 0–191 EU/ml; P = 0.01). Direct markers of translocation (16S DNA and LPS) were markedly higher in adults and malnourished children than in healthy children (Table 2). Indirect markers of bacterial translocation (LBP and CD163, markers of the host response to pathogen associated molecular patterns) were elevated in malnourished children compared to other groups (Table 2). In all children combined, irrespective of nutritional status, IGF-1 was modestly and inversely correlated with LPS (ρ = − 0.23, P = 0.02; n = 100), consistent with translocation of microbes or their components being inhibitory for growth hormone axis activity. No such relationships were detected between IGF-1 and inflammatory markers (CRP, sCD14, CD163).

3.5. Mucosal Repair Peptides

We measured two proteins which promote epithelial restitution: the hormone glucagon-like peptide 2 (GLP-2) (Mutanen and Pakarinen, 2016) in blood, and trefoil factor 3 (TFF3) (Hoffmann, 2009) in intestinal secretions. Serum GLP-2 concentrations were higher in children than adults (Table 2), and lower than the reference range (Mutanen and Pakarinen, 2016) in all our participants. LPS was inversely correlated with GLP-2 (ρ = − 0.35, P = 0.01; n = 54) and this association was stronger in HIV negative adults (ρ = − 0.51; P = 0.001; n = 37; Fig. 3). In children with SAM, translocation was associated with reduced TFF3: LPS was undetectable in 3 of 4 children with detectable TFF3 dimer in duodenal aspirates (P = 0.03; Fig. 3).

3.6. Anti-Tissue Transglutaminase (TTG) and Related Antibodies

In view of the extensive structural change seen in some biopsies, and most notably in two children with subtotal villus atrophy who expired before discharge, testing for coeliac disease serology was explored in children with SAM and in adults, even though the therapeutic as well as the staple diets of children with SAM contain minimal gluten. Tissue transglutaminase (TTG) IgA antibody concentrations, by two different ELISAs, were normal in sera from all adults and children, and the results of both ELISAs correlated closely (ρ = 0.81, P < 0.0001; n = 20). However, anti-TTG antibodies, even within the normal range, were higher in HIV infected children (median 2.4 U/ml, IQR 2.3–2.7) than in HIV-uninfected children (2.1 U/ml, 2.0–2.2; P = 0.01). This difference was not seen in adults. IgG anti-deamidated gliadin peptide (DGP) antibodies were negative in adults, but positive in 3 children, one of whom died in hospital and one of whom died after discharge. Although TTG IgA concentrations were within the normal range, the two children who died in hospital had higher TTG IgA (Inova Quant) concentrations (0.72 and 1.09 U/ml) than children who survived (median 0.55; IQR 0.52–0.6); this was statistically significant (P = 0.04) even though the number of events was very small. Anti-DGP IgG concentrations were higher in children with SAM (median 2.7 U/ml, IQR 1.5–8.6) than in adults (1.6, 1.4–2.1; P = 0.005), and were inversely correlated with villus height (ρ = − 0.79, n = 13, P = 0.001; Fig. 4). Anti-DGP IgG concentrations were also higher in children who died (median 55 U/ml; IQR 8.6–102) than in those who did not (2.7; 1.5–5.1; P = 0.04). All three serological tests were strongly correlated with LBP (ρ = 0.71, P = 0.006 for TTG IgA (Orgentech); ρ = 0.83, P < 0.0001 for TTG IgA (Inova; Fig. 4); and ρ = 0.57, P = 0.006 for anti-DGP IgG).