High endothelial venules associated with T cell subsets in the inflamed gut of newly diagnosed inflammatory bowel disease patients

Summary

Naive and central memory T lymphocytes (T_N and T_CM) can infiltrate the inflamed gut mucosa in inflammatory bowel disease (IBD) patients. Homing of these subsets to the gut might be explained by ectopic formation of tertiary lymphoid organs (TLOs), containing high endothelial venules (HEVs). We aimed to evaluate the presence of HEVs and TLOs in inflamed intestinal mucosa of newly diagnosed, untreated IBD patients in relation to the presence of T_N and T_CM lymphocytes. IBD patients (n = 39) and healthy controls (n = 8) were included prospectively. Biopsy samples of inflamed and normal intestine, respectively, were analysed by immunohistochemistry for lymphocytes (CD3/CD20), blood vessels (CD31) and peripheral lymph node addressin (PNAd) expression (MECA‐79). T_N and T_CM lymphocyte subsets were identified by flow cytometric immunophenotyping. A higher number of HEVs was found in the inflamed colon of patients with ulcerative colitis [median 3·05 HEV/mm²; interquartile range (IQR) = 0–6·39] and ileum of Crohn's disease patients (1·40; 0‐4·34) compared to healthy controls (both 0; P = 0·033). A high density of colonic HEVs (HEV^high) was associated with increased infiltration of T_N and T_CM in the inflamed gut (median 87%; IQR = 82–93% of T cell population), compared to HEV^low patients (58%; 38–81%; P = 0·003). The number of colonic follicles was higher in HEV^high patients (median 0·54/mm²; IQR 0·28–0·84) compared to HEV^low patients (0·25/mm²; 0·08–0·45; P = 0·031) and controls (0·31/mm²; 0·23–0·45; P = 0·043). Increased homing of T_N and T_CM lymphocytes to inflamed gut tissue in IBD patients might be facilitated by ectopic formation of extrafollicular HEVs and TLOs in a subgroup of patients.

Introduction

Lymphocyte activation and homing of lymphocytes to target organs are key phenomena in chronic inflammation, including inflammatory bowel diseases (IBD). Lymphocyte activation is triggered by interaction of antigen receptors with cognate antigens. The initial priming of naive T lymphocytes (T_N) occurs in secondary lymphoid organs [(SLOs); spleen, lymph nodes, gut‐associated lymphoid tissue (GALT); Peyer's patches, isolated lymphoid follicles and appendix)]. SLOs are genetically encoded, specialized environments for optimal antigen presentation and lymphocyte activation. Upon antigen priming in the SLO, activated T lymphocytes migrate to non‐lymphoid tissue sites, where they may exercise effector (proinflammatory or regulatory) functions upon renewed antigen encounter. It is well known that activated effector memory T lymphocytes (T_EM) travel to the gut mucosa following α4β7/mucosal vascular addressin cell adhesion molecule 1 (MAdCAM‐1) and CC chemokine receptor‐9/chemokine (C‐C motif) ligand 25 (CCR9/CCL25) interaction at the level of post‐capillary venules 1, 2.

In physiological conditions, T_N and central memory T lymphocytes (T_CM) are thought to migrate exclusively to SLOs 1, 3. Specialized vessels, so‐called high endothelial venules (HEVs), direct tissue migration of circulating T_N and T_CM into SLOs. This process is guided by different vascular addressins, such as MAdCAM‐1 and peripheral lymph node addressin (PNAd). PNAd on endothelial cells is the ligand of L‐selectin, which is expressed on T_N and T_CM lymphocytes, while practically absent on the majority of T_EM or effector memory T cells re‐expressing CD45RA (T_EMRA). PNAd can be identified by the monoclonal antibody MECA‐79 4, 5. Distribution of these addressins differs between peripheral lymph node, mesenteric lymph node and Peyer's patches. MAdCAM‐1 is displayed constitutively on HEVs in Peyer's patches, on mesenteric lymph nodes and on flattened venular endothelial cells in the intestinal lamina propria 6. PNAd is present predominantly on HEVs in peripheral lymph nodes, mesenteric lymph nodes and at very low levels on the abluminal side of HEVs in Peyer's patches 4.

We and others have demonstrated the presence of T_N and T_CM lymphocytes in non‐lymphoid gut tissue 7, 8, 9, 10, 11. In our previous work, we identified three subgroups of newly diagnosed IBD patients with either increased percentage of mucosal T_N, T_CM or T_EM 7. It is still unclear how homing of these T_N and T_CM lymphocytes to non‐lymphoid tissue is established.

Processes occurring during chronic inflammation and formation of solid tumours were shown to induce ectopic formation of HEV‐like vessels, in inflamed non‐lymphoid and tumour tissue, constituting the so‐called tertiary lymphoid organs (TLOs) 12, 13, 14, 15. Little is known about PNAd⁺ HEV‐like vessels in IBD pathogenesis (reviewed in 16), as published data is limited to patients with long‐standing disease under anti‐inflammatory treatment 5, 17.

We aimed to investigate the presence of HEVs and their relation to lymphoid follicles in the inflamed intestinal mucosa of newly diagnosed, untreated IBD patients. Furthermore, as T_N and T_CM lymphocytes are known to express L‐selectin 7, 18, 19, we aimed to determine whether the infiltration of these cells is correlated with the presence of HEVs and TLOs in the inflamed intestinal mucosa.

Material and methods

Patients and samples

At the IBD centre of the Rijnstate Hospital, Arnhem, the Netherlands, 39 newly diagnosed adult IBD patients and eight healthy controls were included prospectively into the study. The diagnosis of IBD was based on a combination of clinical, endoscopic, histological and radiological internationally accepted criteria. Intestinal biopsy specimens of the macroscopically most inflamed ileal and colonic mucosa (when ulcerations were present these biopsy specimens were taken from the edge of this ulceration) were obtained at the moment of diagnosis, before the initiation of any medical treatment. Thus, when disease was located in the ileum, only ileal biopsies were analysed: when disease was located in the colon, only colonic biopsies were analysed, and when the disease was located in ileum and colon both ileal and colonic biopsies were analysed. As well as regular histopathological analysis, four samples were used to immunophenotype the mucosal lymphocyte population. Patient characteristics were documented and disease phenotype was assessed.

Ileal and colonic biopsy specimens were obtained from eight control subjects undergoing ileocolonoscopy for iron deficiency anaemia or polyp surveillance. Biopsy specimens of controls were taken from macroscopically non‐affected, non‐inflamed ileal and colonic areas. Regular histopathological examination confirmed that these areas had no signs of inflammation. As well as regular histopathological analysis, four samples were used to immunophenotype the mucosal lymphocyte population. Parallel venous blood samples were drawn following endoscopic evaluation at primary diagnosis. Serum samples were coded and stored at −20°C until analysis.

Immunohistochemistry

Immunohistochemistry was carried out on 3 µm thick sections from formalin‐fixed, paraffin‐embedded archived blocks of biopsied specimen of ileal and colonic mucosa. The following monoclonal antibodies were used for immunostaining: MECA‐79 [identifies 6‐sulpho‐sialyl Lewis on core‐1 branched O‐linked sugars (PNAd) which binds L‐selectin; Sanbio, Uden, the Netherlands], CD31 (monoclonal, identifies human CD31 to demonstrate endothelial cells; Sanbio), CD3 (clone LN10, identifies human CD3 to demonstrate T cells; Leica Microsystems BV, Eindhoven, the Netherlands) and CD20 (clone L26, identifies human CD20 to demonstrate B cells, Leica Microsystems BV). Slides were incubated with the primary antibody in an automatic immunostainer (Bond, Leica Biosystems Nussloch GmbH 2016, Eindhoven, the Netherlands).

Stained slides were scanned with an Intellisite high‐resolution scanner (Philips ultra‐fast scanner 1.6 RA; Philips Digital Pathology, Best, the Netherlands) and analysed with the Intellisite IMS image analyser. Quantification was performed by two independent observers who were blinded to the clinical outcome and flow cytometric analyses. The total surface area of each slide was measured in square millimetres (mm²). Absolute numbers of CD31⁺ vessels and MECA‐79⁺ HEVs were determined by optical counting of vessel numbers on all slides. The location of HEVs relative to lymphoid follicles was determined on CD3 and CD20 immunostained sections and absolute numbers of intra‐ and extrafollicular HEVs were quantified. The density of extrafollicular HEVs (MECA‐79⁺ vessels/mm² excluding the surface of follicles) and the ratio of HEVs among all vessels (MECA‐79⁺ vessels/CD31⁺ vessels) were calculated. In malignancy, a cut‐off point for increased density of HEVs has been described in patients with breast cancer 20. However, no such approach has been undertaken in chronic inflammation. We discriminated patients with a high and a low density of HEVs using a cut‐off point set at the median density of extrafollicular HEVs/mm² in the inflamed colon of ulcerative colitis (UC) patients.

Presence of inflammation was established for all sections based on known architectural changes and inflammatory features such as crypt distortion, decreased crypt density, erosions, ulcers, increased lamina propria cellularity, crypt abscesses and epithelioid granuloma. Lymphoid follicles were defined as dense nodular lymphocyte aggregates containing B and T cell compartments. Their presence was assessed on CD3 and CD20 immunostained sections. The number of follicles was counted. In the absence of a clear‐cut definition of TLOs in the gut, we defined TLO as additional follicles in patients compared with the number of follicles in controls. Furthermore, it is our interpretation that the presence of extrafollicular HEVs represent the early phase of lymphoid neogenesis (TLOs).

Immunophenotyping the mucosal lymphocytes

Intestinal biopsy specimen were sampled and analysed as described earlier 7. Briefly, as well as the standard histopathological analysis of biopsy specimens, four samples of the most inflamed regions of colon and ileum were used for flow cytometric immunophenotyping. Biopsy specimens were kept in phosphate‐buffered saline solution at 2–8°C and processed within 8 h. Specimens were minced finely in Hanks's/1% bovine serum albumin using a 70‐mm gaze and spatula followed by Ficoll density gradient centrifugation.

The homogenate was resuspended, after washing, in 0·5 ml Hanks's/1% bovine serum albumin. The cell number was estimated by microscopic counting with a KOVA glasstic slide (Hycor Biomedical Ltd., Penicuik, UK). For flow cytometric analysis, 200 ml of the total cell suspension was used, disregarding absolute cell number. Single‐cell suspensions of the biopsies were analysed using a FACS Canto (BD Biosciences, San Jose, CA, USA). Mucosal lymphocyte subpopulations were differentiated: B cells (CD19⁺), T cells (CD3⁺), regulatory T cells [CD3⁺ CD25^highforkhead box protein 3 (FoxP3⁺)], mucosal T cells (CD3⁺CD103⁺), T helper cells (CD3⁺CD4⁺) and cytotoxic T cells (CD3⁺CD8⁺) and expressed as percentages of the whole lymphocyte population (CD45⁺/low side‐scatter). The maturation state of T cells was assessed using CD45RA and CD27 21, defining naive T cells (T_N: CD45RA⁺CD27⁺), central memory T cells (T_CM: CD45RA^–CD27⁺), effector memory T cells (T_EM: CD45RA^–CD27^–) and T effector memory cells re‐expressing CD45RA (T_EMRA: CD45RA⁺CD27^–). We also analysed groups of T cells according to their known migration phenotypes. T_N and T_CM lymphocytes are known to enter HEVs through the ligand PNAd, as they are L‐selectin⁺ 18, 19, and were analysed both independently as well as taken together. The same analysis was performed for T_EM and T_EMRA lymphocytes, which are known to be L‐selectin^– 18, 19

Multiplex immunoassay

In a subgroup of the current patient cohort we analysed serum chemokines using Multiplex immunoassays, including CXCL‐13 and CCL‐19, which are known to be required for lymphoid organogenesis 22. These assays were performed at the MultiPlex Core Facility of the Laboratory of Translational Immunology (UMC Utrecht, the Netherlands) using an in‐house validated platform (ISO9001). Briefly, colour‐coded magnetic beads (MagPlex Microsperes, Luminex, Austin, TX, USA) were conjugated to analyte‐specific antibodies and incubated with standard dilutions or sample for 1 h during continuous shaking in the dark. Pretreatment of samples with HeteroBlock (Omega Biologicals, Inc., Bozeman, MT, USA) was performed, to prevent interference by binding of heterophilic antibodies. Plates were washed (Bio‐Plex Pro II Wash Station; Bio‐Rad, Hercules, CA, USA) and a corresponding cocktail of biotinylated detection antibodies was added followed by streptavidin–phycoerythrin (PE) incubation. Fluorescence intensity of PE was measured with a Flexmap 3D system (Luminex) and analysed by BioPlex Manager Software (version 6.1; Bio‐Rad) using five‐parameter curve fitting.

Statistical analysis

Data of ileal biopsy specimens of Crohn's disease (CD) patients were compared to ileal biopsy specimens of healthy controls. The data of colonic biopsy specimens of UC and CD patients were compared to the colonic biopsy specimens of healthy controls. Ileum and colon were not mutually compared because of anatomical and physiological differences, including the influence of the microbiome. Data was collected and analysed with spss statistics version 21.0.0.0 (IBM Corp., Armonk, NY, USA). Differences between groups were tested using non‐parametric tests. Continuous variables were described as median with interquartile range [(IQR), 1st–3rd quartiles]. The Mann‐Whitney U‐test was used to identify differences in continuous variables. Categorical variables were described as absolute frequencies. Kruskal–Wallis analysis was used followed by the Dunn test when applicable. Spearman's rank correlation was performed to measure the degree of association between variables. Statistical significance was defined as a P‐value less than 0·05.

Ethical considerations

Written informed consent was obtained from each participating patient. The procedures were performed in accordance with the Declaration of Helsinki. The regional medical ethics committee approved the protocol, according to Dutch Law (NL28761.091.09).

Results

Ten newly diagnosed patients with UC and 29 newly diagnosed IBD patients with CD were included in this study. In addition, eight healthy controls were studied. Baseline characteristics of patients are outlined in Table 1.

Table 1.

Baseline characteristics

	CD (n = 29)	CU (n = 10)
Age at diagnosis in years, median (IQR)	28 (20–44)	28 (20–35)
Female, n (%)	21 (72%)	7 (70%)
Symptoms before diagnosis in months
0–3	10 (35%)	7 (70%)
3–6	9 (31%)	1 (10%)
> 6	10 (35%)	2 (20%)
Follow‐up period in months, median (IQR)	55 (23–62)	44 (22–48)
Family with IBD, n (%)	4 (14%)	4 (40%)
Smoking status
Never	17 (59%)	8 (89%)
Current	10 (34%)	1 (11%)
Ceased	2 (7%)	–
Unknown	–	–
HBI score, median (IQR)	9 (6–13)	–
CD disease location (Montreal)		–
• Ileal, n (%)	8 (28%)
• Colonic, n (%)	5 (17%)
• Ileocolonic, n (%)	16 (55%)*
• + Upper GI involved, n (%)	10 (39%)
• + Perianal disease, n (%)	5 (17%)
CD disease behaviour (Montreal)		–
• Inflammatory, n (%)	21 (72%)
• Stricturing, n (%)	7 (24%)
• Penetrating, n (%)	1 (4%)
SES‐CD score, median (IQR)	14 (8–22)	–
SES‐CD severity		–
• Mild (4–10), n (%)	11 (38%)
• Moderate (11–19), n (%)	9 (31%)
• Severe (> 19), n (%)	9 (31%)
UC clinical score (Montreal)	–
• Remission, n (%)		–
• Mild, n (%)		1 (10%)
• Moderate, n (%)		4 (40%)
• Severe, n (%)		5 (50%)
UC disease location (Montreal)	–	–
• Proctitis, n (%)		–
• Left‐sided, n (%)		–
• Pancolitis, n (%)		10 (100%)
UC endoscopic severity (Mayo)	–
• Normal, n (%)		–
• Mild, n (%)		2 (20%)
• Moderate, n (%)		4 (40%)
• Severe, n (%)		4 (40%)

*One patient was excluded from the analyses of colon biopsies because of technical failure of MECA‐79 staining. CD = Crohn's disease; UC = ulcerative colitis; IQR = nterquartile range; IBD = inflammatory bowel disease; HBI = Harvey–Bradshaw Index; SES‐CD = simple endoscopic score for Crohn's disease.

Number of blood vessels and follicles was comparable in patients and controls

In each sample, we determined the absolute number of blood vessels (CD31⁺ vessels and MECA‐79⁺ vessels, denoted hereafter as HEVs) and follicles by immunohistochemical staining. HEVs were determined by optical counting of MECA‐79⁺ vessel numbers on all slides. The location of these HEVs relative to lymphoid follicles was determined on CD3 and CD20 immunostained sections and absolute numbers of intra‐ and extrafollicular HEVs were quantified. The density of extrafollicular HEVs (MECA‐79⁺ vessels/mm² excluding the surface of follicles) and the ratio of HEVs among all vessels (MECA‐79⁺ vessels/CD31⁺ vessels) were calculated. Overall, we observed high heterogeneity in patient results. No statistical significant difference was found when comparing the entire patient group and controls for the number of vessels or follicles in biopsy specimen (see Table 2). In addition, the number of HEVs within these follicles was comparable. However, in UC there was a trend towards more HEVs in the inflamed colon [4·80/mm² (0.45–6·14)] than in the colon of healthy controls [0·71/mm² (0–1·85), P = 0·109].

Table 2.

Blood vessels and follicles in gut mucosa of inflammatory bowel disease (IBD) patients and controls

	CD (inflamed)	UC (inflamed)	Healthy controls (non‐inflamed)	P‐value
Number of CD31+ vessels/mm2 colon, median (IQR)	84 (60–107)	125 (63–139)	71 (43–89)	n.s.
Number of CD31+ vessels/mm2 ileum, median (IQR)	93 (70–114)	–	87 (69–102)	n.s.
Number of HEVs/mm2 colon, median (IQR)	0·35 (0–5·17)	4·80 (0·45–6·14)	0·71 (0–1·85)	n.s.
Number of HEVs/mm2 ileum, median (IQR)	1·58 (0·34–6·69)	–	1·23 (0–5·19)	n.s.
% MECA‐79 expressing CD31+ vessels colon, median (IQR)	0·85 (0–4·80)	4·63 (0·38–12·24)	0·79 (0–2·59)	n.s.
% MECA‐79 expressing CD31+ vessels ileum, median (IQR)	2·02 (0·39–6·82)	–	1·02 (0–8·99)	n.s.
Colonic follicles/mm2 tissue, median (IQR)	0·27 (0·12–0·51)	0·51 (0–0·82)	0·31 (0·23–0·45)	n.s.
Ileal follicles/mm2 tissue, median (IQR)	0·35 (0·15–0·57)	–	0·37 (0–0·55)	n.s.

CD = Crohn's disease; UC, = ulcerative colitis; IQR = interquartile range; HEV = high endothelial venules; n.s. = not significant.

Increased extrafollicular HEV expression in IBD patients compared to controls

Next, we calculated the percentage and density of extrafollicular HEVs and found that colon samples of healthy controls did not show any, whereas the inflamed colon of UC patients showed a high percentage [median 3·71% (IQR = 0–6·23%), P = 0·033, Fig. 1] and a high median density [3·05/mm² (IQR = 0‐6·39), P = 0·033] of extrafollicular HEVs. In the inflamed colon of CD patients, we observed a trend towards a higher number of extrafollicular HEVs [median 0·10 (IQR = 0–2·98)], compared to healthy controls [median 0 (IQR = 0–0), P = 0·092].

Figure 1.

High endothelial venules (HEVs) in inflamed colonic tissue of patients with inflammatory bowel disease. Density of extrafollicular HEVs in colonic tissue (a), determined in the inflamed colonic tissue of patients with Crohn's disease (CD) and ulcerative colitis (UC) as well as in healthy controls. The density is higher in UC patients when compared to controls (P = 0·033) and there is a trend for higher densities in CD patients compared to controls (P = 0·092). Immunohistochemical staining (B1–B5). Representative photomicrographs, with a magnification of ×200, of an immunostained colonic biopsy sample from a Crohn's disease patient with MECA‐79 staining (b1), CD20 staining showing B lymphocytes (b2) and CD3 staining showing T lymphocytes (b3). HEVs are shown in closer detail in b4 (extrafollicular) and b5 (inside a follicle). Arrows in Fig. b1, b4 and b5 indicate HEVs. Red arrows indicate HEVs inside a lymphoid follicle and black arrows indicate extrafollicular HEVs. [Colour figure can be viewed at wileyonlinelibrary.com]

In the ileum of CD patients, the percentage of extrafollicular HEVs was higher in inflamed ileum [median 1·17% (IQR = 0–3·77%)] compared to ileum of healthy controls [median 0% (IQR = 0–0·4%), P = 0·028]. These samples showed a median density of 1·40 extrafollicular HEVs/mm² (IQR = 0–4·34), whereas in healthy controls only a few HEVs/mm² (0%, IQR = 0–0·50) were observed (P = 0·033). The presence of extrafollicular HEVs in the inflamed ileum and colon of IBD patients might represent an early phase of TLO formation.

HEV density is not associated with neovascularization

To examine whether the difference in density of PNAd‐expressing HEV‐like vessels in patients could be attributed to neovascularization, we quantified the number of HEVs/mm² and correlated this with the total number of CD31⁺ vessels in patients and controls. We also compared colonic UC with colonic CD biopsy specimens. We did not find a correlation, and no statistical difference was found between UC and CD patients (P = 0·198), indicating that the differences in HEV density were not related to or associated with neovascularization.

To assess a potential correlation with several confounders, as listed in Table 1, we found that HEV density in inflamed colon was correlated negatively with smoking (Spearman's rho = −0·543, P = 0·002) when analysing all IBD patients. However, when subdivided into CD and UC, it was correlated significantly only in CD (Spearman's rho = −0·683, P = 0·001) and not in UC (Spearman's rho = −0·139, P = 0·721). Smoking was not correlated with HEV density of inflamed ileum in CD patients. We did not find any correlations with other baseline characteristics (data not shown).

HEV^high IBD patients have increased colonic lymphoid follicles and infiltration of T_N ± T_CM cells

Immunohistochemical findings were correlated with the maturation state of T cell infiltrates, assessed by flow cytometric immunophenotyping of biopsied specimens taken from the same area of inflamed tissue and harvested during the same ileocolonoscopy. The extrafollicular HEV density in inflamed colon of IBD patients (n = 30, 10 UC patients, 20 CD patients) was correlated positively with the percentage of mucosal T_N + T_CM lymphocytes (Spearman's rho = +0·544, P = 0·002) and negatively with the percentage of mucosal T_EM + T_EMRA lymphocytes (Spearman's rho = −0·545, P = 0·002). Comparable percentages of lymphocyte subsets were found for UC and CD patients [T_N + T_CM lymphocytes (P = 0·406), T_EM + T_EMRA lymphocytes (P = 0·451)].

In malignancy, a cut‐off point for increased density of HEVs has been described in patients with breast cancer 20. However, in chronic inflammation, no such approach has yet been undertaken. We discriminated patients with a high and a low density of HEVs using a cut‐off point set at the median density of extrafollicular HEVs/mm² in the inflamed colon of UC patients. According to this cut off‐value, we identified 10 HEV^high patients and 20 HEV^low patients (Table 3). HEV^high patients displayed fewer cytotoxic T cells (P = 0·00008), a higher CD4 : CD8 ratio (P = 0·0008), fewer mucosal T cells (P = 0·015), more T_CM cells (P = 0·0009), more T_N + T_CM cells (P = 0·003), fewer T_EM cells (P = 0·022), fewer T_EMRA cells (P = 0·008) and fewer T_EM + T_EMRA cells (P = 0·004) in the inflamed colonic mucosa than HEV^low patients (Table 3 and Fig. 2).

Table 3.

Characteristics and fluorescence activated cell sorter (FACS) results of high endothelial venules (HEV) (HEV^high) and HEV^low IBD patients

	HEVhigh (n = 10)	HEVlow (n = 20)	P‐value
CD/UC	5/5	15/5	0.169
Age at diagnosis in years, median (IQR)		28 (19–33)	n.s.
Female, n (%)	9 (90%)	14 (70%)	n.s.
Symptoms before diagnosis in months			n.s.
0–3	4 (40%)	12 (60%)
3–6	5 (50%)	3 (15%)
> 6	1 (10%)	5 (25%)
Follow‐up period in months, median (IQR)	34 (22–56)	50 (28–65)	n.s.
Family with IBD, n (%)	3 (30%)	5 (25%)	n.s.
Smoking status
Never	9 (90%)	11 (55%)	0·029
Current	–	8 (40%)
Ceased	–	1 (5%)
Unknown	1 (10%)	–
HBI score, median (IQR)	7 (5–14) (n = 5)	10 (7–13) (n = 15)	n.s.
CD disease location (Montreal)			n.s.
Ileal, n (%)	–	–
Colonic, n (%)	2 (40%)	3 (20%
Ileocolonic, n (%)	3 (60%)	12 (80%)
+ Upper GI involved, n (%)	1 (10%)	7 (44%)
+ Perianal disease, n (%)	2 (20%)	3 (17%)
CD disease behaviour (Montreal)			n.s.
Inflammatory, n (%)	5 (100%)	11 (73%)
Stricturing, n (%)	–	3 (20%)
Penetrating, n (%)	–	1 (7%)
SES‐CD score, median (IQR)	14 (9–19) (n = 5)	20 (12–34) (n = 15)	n.s.
SES‐CD severity			n.s.
• Mild (4–10), n (%)	2 (40%)	2 (13%)
• Moderate (11–19), n (%)	3 (60%)	5 (33%)
• Severe (> 19), n (%)	0 (0%)	8 (53%
UC clinical score (Montreal)			n.s.
• Remission, n (%)	–	–
• Mild, n (%)	1 (20%)	–
• Moderate, n (%)	2 (40%)	2 (40%)
• Severe, n (%)	2 (40%)	3 (60%)
UC disease location (Montreal)			n.s.
• Proctitis, n (%)	–	–
• Left‐sided, n (%)	–	–
• Pancolitis, n (%)	5 (100%)	5 (100%)
UC endoscopic severity (Mayo)			n.s.
• Normal, n (%)	–	–
• Mild, n (%)	2 (40%)	–
• Moderate, n (%)	1 (20%)	3 (60%)
• Severe, (%)	2 (40%)	2 (40%)
FACS analysis of inflamed colon*
• CD3, median (IQR)	50 (45–63)	59 (52–72)	0·267
• CD19, median (IQR)	39 (29–49)	23 (18–42)	0·108
• CD4, median (IQR)	74 (38–83)	61 (36–75)	0·169
• CD8, median (IQR)	15 (11–19)	29 (22–37)	0·00008
• CD4/CD8 ratio, median (IQR)	4·3 (3·0–6·1)	2·2 (1·2–3·2)	0·0008
• CD103, median (IQR)	7 (4–12)	15 (8–43)	0·015
• TN, median (IQR)	23 (20–40)	23 (9–54)	0·779
• TCM, median (IQR)	58 (41–72)	33 (20–45)	0·001
• TEM, median (IQR)	12 (6–18)	26 (12–46)	0·024
• TEMRA, median (IQR)	2 (1–3)	6 (2–12)	0·010
• Treg, median (IQR)	12 (8–18)	14 (12–17)	0·379

*Expressed as percentage of the whole lymphocyte population. CD = Crohn's disease; UC = ulcerative colitis; HEV = high endothelial venule; IQR = interquartile range; IBD = inflammatory bowel disease; HBI = Harvey–Bradshaw Index; SES‐CD = simple endoscopic score for Crohn's disease; T_N = naive T lymphocyte; T_CM = central memory T lymphocyte; T_{EM =} effector memory T lymphocyte; T_EMRA = effector memory T lymphocyte re‐expressing RA; T_reg = regulatory T lymphocyte; n.s. = not significant.

Figure 2.

Correlation between high endothelial venules (HEVs) and T cell subpopulations. Lymphocyte subsets as well as maturation state of T lymphocytes was assessed with flow cytometric immunophenotyping of biopsy specimens taken from the same inflamed colonic tissues during the same ileocolonoscopy. Patients are divided into HEV^high (n = 10) or HEV^low (n = 20) groups using the median density of extrafollicular peripheral lymph node addressin (PNAd) expressing HEV‐like vessels in inflamed colon of ulcerative colitis (UC) patients as cut‐off value. Statistically significant results are indicated with an asterisk (*). (a) Shows more central memory T cells (T_CM, CD45RA^–CD27⁺, P = 0·0009), fewer effector memory T cells (T_EM, CD45RA^–CD27^–, P = 0·022) and fewer effector memory T cells re‐expressing RA (T_EMRA, CD45RA⁺CD27^–, P = 0·008) in HEV^high patients when compared with HEV^low patients. There was no statistically significant difference for naive T cells (T_N CD45RA⁺CD27⁺). (b) Shows fewer cytotoxic T cells (CD3⁺CD8⁺, P = 0·00008), a higher CD4 : CD8 ratio (P = 0·0008) and fewer mucosal T cells (CD3⁺CD103⁺, P = 0·015) in HEV^high patients. There was no statistically significant difference for B cells (CD19⁺), T cells (CD3⁺), T helper cells (CD3⁺CD4⁺) and regulatory T cells [CD3⁺CD25^highforkhead box protein 3 (FoxP3⁺)]. (c) Representative flow cytometric dot‐plots from inflamed colonic biopsy specimens of inflammatory bowel disease (IBD) patients which show the distribution of the maturation of (CD3⁺) T lymphocytes. (c1) A HEV^high patient with mainly T_N (CD45RA⁺CD27⁺, 54%) and T_CM (CD45RA^–CD27⁺, 38%) lymphocytes. (c2) A HEV^low patient with mainly (91%) T_EM (CD45RA^–CD27^–) lymphocytes. [Colour figure can be viewed at wileyonlinelibrary.com]

The number of colonic follicles was higher in HEV^high IBD patients [median 0·54/mm² (IQR = 0·28–0·84)] compared to HEV^low patients [median 0·25/mm² (IQR = 0·08–0·45), P = 0·031] and to healthy controls [0·31/mm² (IQR = 0·23–0·45), P = 0·043]. The number of colonic follicles in the inflamed colon of patients was comparable between CD and UC (P = 0·348).

In a subgroup of the current patient cohort we analysed a series of serum cytokines and chemokines, of which CXCL‐13 and CCL‐19 (both associated with lymphoid organogenesis 22) were present at higher concentrations in the HEV^high patient group. Patients in this group (n = 4) had higher baseline levels of CXCL‐13 [median 67 pg/ml (IQR = 52–742 pg/ml)] compared to patients in the HEV^low group [n = 8, median 36 pg/ml (IQR = 27–47 pg/ml), P = 0·008] and to healthy controls [n = 4, median 24 pg/ml (IQR = 18–28 pg/ml), P = 0·029, Fig. 3]. Furthermore, there was a trend towards higher CCL‐19 levels in HEV^high patients (median 176 pg/ml, IQR = 126–199 pg/ml) compared to HEV^low patients (median 125 pg/ml, IQR = 90–141 pg/ml, P = 0·073).

Figure 3.

Serum CXCL‐13 concentrations in inflammatory bowel disease (IBD) patients and healthy controls. Serum of a subgroup of patients and healthy controls was analysed by multiplex immunoassay for several cytokines and chemokines. Patients are divided into high endothelial venule (HEV)^high or HEV^low groups using the median density of extrafollicular HEVs in inflamed colon of ulcerative colitis patients as cut‐off value. HEV^high patients had higher CXCL‐13 concentrations compared to HEV^low patients (P = 0·008) and compared to healthy controls (P = 0·029). [Colour figure can be viewed at wileyonlinelibrary.com]

Discussion

In the present study, increased percentages of extrafollicular HEVs were observed in inflamed colon and ileum of newly diagnosed IBD patients compared to healthy controls. In controls, these HEVs were restricted to lymphoid follicles (SLOs). A high heterogeneity was observed in the number of extrafollicular HEVs in IBD patients. We identified two subgroups of patients according to the density of the extrafollicular HEVs in the inflamed colonic mucosa: HEV^high and HEV^low. In HEV^high patients, a higher percentage of T_N + T_CM lymphocytes was observed in the inflamed colonic mucosa compared to HEV^low patients. Patients in the HEV^high group also displayed a higher number of lymphoid follicles (TLOs) in the inflamed colonic mucosa compared to patients in the HEV^low group and controls. Furthermore, chemokines associated with TLO formation (CXCL‐13 and CCL‐19) were upregulated in the serum of HEV^high patients

In physiological conditions, HEVs are known to be present exclusively within SLOs, whereas in the present study we demonstrated their presence outside follicles in the inflamed intestine of IBD patients. It is our interpretation that extrafollicular HEVs represent an early phase of lymphoid neogenesis (TLO formation), as these vessels were virtually absent in healthy controls. In addition to Peyer's patches, GALT also contains other forms of SLOs, such as isolated lymphoid follicles and submucosal lymphocyte aggregations. Isolated lymphoid follicles are seen mainly within the mucosa, while submucosal lymphocyte aggregations lie deeper within the muscularis mucosa and submucosa 23, 24. Peyer's patches are present mainly in jejunum and ileum, while isolated lymphoid follicles have also been found in the human colon 24. In accordance with this, we found HEVs within lymphoid follicles of the normal gut mucosa of both ileum and colon of healthy controls; however, they did not display such vessels outside these follicles.

In IBD, HEVs were investigated previously in one study using MECA‐79 immunostaining of biopsy specimens from UC patients and surgical specimens of CD patients, presumably with long‐standing disease and under immunomodulatory treatment 5. Increased HEVs were found in inflamed colonic mucosa of UC patients and (to a lesser extent) in CD patients, compared to non‐inflamed colonic mucosa of other IBD patients. However, no healthy controls were included, T cell maturation in the gut mucosa was not analysed and no serial biopsy samples at different time‐points and phases of disease within the same patients were analysed. Numbers of HEVs in patients with non‐inflamed colonic mucosa have not been investigated during active disease, and therefore it might concern a subgroup of patients without ectopic HEV formation, similar to our HEV^low group. Our results expand upon this previous study, as we demonstrated differences in HEV density and TLOs in newly diagnosed, untreated IBD patients with active disease in correlation with different T cell subsets in the gut. Direct comparison with our presented results is difficult, as we included healthy controls and studied mucosal T cell maturation status and the presence of HEVs in relation to lymph follicles. We demonstrated increased extrafollicular HEVs formation in a subgroup of patients (HEV^high) in the early phase of disease. The lack of extrafollicular HEVs in the other subgroup (HEV^low) might be explained by a different disease mechanism in these patients or the early phase of disease, assuming that formation of HEVs in inflamed mucosa is a characteristic of chronicity. Increased percentages of T_EM cells in HEV^low patients suggest predominant recruitment of T_EM cells into the inflamed gut. These T_EM cells, known to be α4β7⁺, might enter the inflamed gut by binding to MAdCAM‐1 present on endothelial cells, in the absence of extrafollicular HEVs.

Lymphoid follicles in the gut were studied previously in established CD patients. In line with our findings in colonic samples of early IBD patients, they demonstrated the presence of submucosal de‐novo lymphoid follicles in surgical ileal resection specimens of CD patients 17. The authors interpreted these lymphoid follicles as being TLOs, based mainly on their location in the submucosa, muscularis propria and subserosa, but theoretically these follicles might still represent SLOs. However, no MECA‐79 immunostaining with HEV quantification was performed 17. Furthermore, the use of ileal resection specimen suggests a highly refractory patient group with long‐standing disease. Potential differences between patients and associated mucosal T cell subsets were not investigated.

The mechanism of TLO development has not yet been clarified. Nevertheless, chemokines such as CXCL‐13, CCL‐19 and CCL‐21, together with lymphotoxin α₁β₂ [LTα₁β₂, tumour necrosis factor (TNF) family member], expressed on lymphoid tissue inducer cells (LTi), were shown to be required for lymphoid organogenesis 22. LTα₁β₂ is implicated in the generation of follicular dendritic cells and HEVs out of stromal cells, while CCL‐19, CCL‐21 and CXCL‐13 control the organization of the T and B cell follicular regions 12, 22, 25. Previously, increased CXCL‐13 expression has been observed in inflamed gut of UC patients 26, while research in transgenic mice demonstrated that CXCL‐13 expression promotes B cell recruitment and formation of lymphoid follicles 27. Remarkably, we found increased serum CCL‐19 and CXCL‐13 in HEV^high patients. This suggests a systemic reflection of gut‐localized formation of TLOs in these patients.

Recently, research in mice lacking LTi, it was demonstrated that lymphoid organogenesis was linked to the activation of stromal cells by TNF‐α. This suggests that increased TNF‐α expression during chronic inflammation might play an important role in TLO development in IBD 28. Development of TLOs containing HEVs has been reported in other diseases with chronic inflammation and in patients with solid tumours. In patients with rheumatoid arthritis, TLOs were associated with an inferior therapeutic response to anti‐TNF treatment and reversal of TLOs was a good marker of beneficial therapeutic response 29.

In our study, the higher number of colonic follicles found in the HEV^high group indicated that, in this subgroup of IBD patients, de‐novo lymphoid follicles were being formed, representing TLOs. The patients in the HEV^high group not only had more follicles, they also exhibited more HEVs outside these follicles, and this was correlated with a higher percentage of mucosal T_N + T_CM lymphocytes (known to be L‐selectin⁺ 18, 19) and a lower percentage of mucosal T_EM + T_EMRA lymphocytes (known to be L‐selectin^– 18, 19). Remarkably, the patients in the HEV^low group were comparable to healthy controls regarding these findings. HEVs express different functional ligands (PNAd, MAdCAM‐1) for L‐selectin, facilitating extravasation and entry of T_N + T_CM lymphocytes into lymphoid organs 30. Our findings suggest differences between IBD patients, both UC and CD, regarding TLO formation and recruitment of different lymphocyte subsets to the inflamed gut mucosa. These results expand further on our previous paper in newly diagnosed IBD patients with increased numbers of T_N and T_CM cells in the inflamed gut of a subgroup of patients 7. Taken together, our results point out that increased homing of T_N and T_CM lymphocytes to non‐lymphoid gut tissue in a subgroup of IBD patients might be facilitated by de‐novo formation of extrafollicular HEVs and TLOs.

A limitation of the present study was that mucosal biopsy specimens precluded full examination of the deeper layers of the intestine wall (submucosa, muscularis propria and subserosa), which potentially contain HEVs and TLOs in IBD patients, particularly in CD. However, the strengths of our study were the prospective design, the sampling of untreated patients early in the course of IBD and the use of a standardized description of TLOs and HEVs in inflamed mucosa of patients.

The clinical and therapeutic relevance of these findings in IBD patients remain to be elucidated; the presence of lymphoid follicles in initial biopsy specimens may, however, be predictive for colectomy due to drug refractory UC 31. Furthermore, novel treatment strategies targeting T cell migration to the inflamed gut mucosa of IBD patients, such as anti‐α4β7 [vedolizumab (Entyvio^®)] and anti‐MAdCAM‐1 monoclonal antibodies (PF‐00547659), are emerging 32, 33. HEV^low patients with a higher percentage of T_EM lymphocytes, suggesting increased recruitment of α4β7⁺ T cells, might theoretically benefit more and thus respond better to vedolizumab treatment than HEV^high patients. Moreover, these findings may lead to further development of new treatment targets in IBD, designed to influence HEV and TLO formation by targeting the lymphotoxin pathway 34.

Conclusion

In conclusion, higher numbers of extrafollicular HEVs were demonstrated in IBD patients compared to healthy controls. An increased density of extrafollicular HEVs and TLOs in IBD patients was associated strongly with mucosal infiltration of T_N + T_CM lymphocytes, while a low density of HEVs correlated with higher numbers of mucosal T_EM cells. This suggests that the homing of T_N and T_CM lymphocytes to non‐lymphoid gut tissue in IBD patients might be facilitated by extrafollicular HEVs and de‐novo TLO formation within inflamed mucosa. Different ‘T cell migration phenotypes’ based on TLO formation in the early phase of IBD may allow for risk‐stratification of patients and enable individualized treatment.

Disclosure

The authors have no disclosure to declare.

Contributors

All authors participated in the conception and design of the study. C. H., C. S., M. G., P. W. and E. v. L. participated in patient recruitment and material collection. J. M and M. R. were responsible for immunohistochemical staining and quantification. E. v. L. was responsible for immunophenotyping of the mucosal lymphocytes. C. S. and C. H. were responsible for statistical analyses. All authors were members of the writing group and participated in the drafting and revision of the manuscript. All authors approved the final version of the manuscript.