Human Leukocyte Subpopulations from Inflamed Gut Bind to Joint Vasculature Using Distinct Sets of Adhesion Molecules¹

About 25% of inflammatory bowel disease (IBD)³ patients suffer from extraintestinal inflammatory complications. Among the most common nonintestinal lesions observed in these patients are episodes of reactive arthritis, which vary from mild self-limiting joint inflammations to treatment-resistant destructive arthritis. The etiopathogenesis of IBD and its complications remain unknown (1, 2), but a putative link between gut and joint in leukocyte trafficking patterns (3, 4, 5, 6, 7) may help in understanding the extraintestinal inflammatory response.

Normally naive lymphocytes continuously recirculate between different lymphoid tissues of the body (8, 9, 10). When a lymphocyte finds its cognate Ag in Peyer’s patches, which represent the organized lymphoid tissue of the gut, the cell starts to divide and differentiate. The progeny of activated immunoblasts is ultimately transferred back to the circulation via the efferent lymphatics. In contrast to the random migration pattern of naive cells, these effector cells now have the remarkable capacity to selectively re-enter the intestinal tissues, especially lamina propria, to exert their immune functions (11, 12).

Lymphocytes as well as other types of leukocytes use multiple adhesion and activation molecules to exit from the circulation (8, 9, 10). Normally L-, E-, and P-selectins and their carbohydrate-containing mucin-like counterparts (like P-selectin glycoprotein ligand-1 (PSGL-1) for P- and E-selectin, and peripheral lymph node addressins (PNAd) for L-selectin) mediate the initial phases of extravasation (13). However, CD44 and vascular adhesion protein-1 (VAP-1) also contribute to the rolling of blood-borne cells (14, 15). Chemokines are thought to be the key players in converting loosely rolling cells into stably bound cells (16). This takes place by activation of leukocyte CD18 integrins and α₄β₇ and α₄β₁ (17), which then bind to their endothelial counter-receptors belonging to the Ig superfamily (ICAM-1, ICAM-2, and VCAM-1). Finally, the leukocyte leaves the blood by transmigrating between the endothelial cells into the tissue stroma by poorly characterized mechanisms (18).

Binding of lamina propria lymphocytes from normal gut to different vascular beds and the adhesion molecules involved have been described (6, 19). However, gut leukocytes in IBD include an inflammation-associated population, and binding of IBD-originating mucosal leukocytes to mucosal and lymph node vessels is different from that of normal mucosal leukocytes (20, 21). Therefore, we studied here whether mononuclear leukocytes isolated from gut specimens of patients with IBD would bind to vessels in joints and which adhesion molecules mediate the interactions between IBD gut leukocytes and synovial vasculature. Our results indicate that different subsets of mucosal immune cells use profoundly distinct adhesion molecules to attach to vessels in synovial membrane. These leukocyte-endothelial interactions may regulate homing of activated, gut-originating immunoblasts and macrophages into distant joints and hence help to explain the onset of reactive arthritis following IBD.

Isolation of lamina propria leukocytes was performed according to the method described by MacDermott et al. (19, 22). In brief, surgical specimens of the affected gut from 17 IBD patients (10 Crohn’s disease and 7 ulcerative colitis) were obtained over 5 years and were analyzed freshly each time. Lamina propria was dissected free, epithelial cells were detached by EDTA treatments, and leukocytes were released from the remaining lamina propria pieces by overnight digestion with type I collagenase. Mononuclear cells were collected by Ficoll centrifugation.

Synovial samples from chronically inflamed synovectomy specimens were collected from eight patients. Normal synovial tissue was removed from a cadaver donating bone transplants. All tissue collection procedures were approved by local ethical committees.

The mAbs against leukocyte homing receptors and endothelial adhesion Ags used in this study are listed in Table I.

Isolated mononuclear cells were stained with the primary mAbs and appropriate FITC-conjugated second stage reagents for flow cytometry as previously described (19). At least 10,000 cells were analyzed using a FACScan instrument and CellQuest software (Becton Dickinson, San Jose, CA).

The nonstatic Stamper-Woodruff assay that has been successfully used for studying the functions of different classes of cell adhesion molecules was used for the binding experiments (6, 19). To that end, 8-μm frozen sections were cut from synovial specimens onto microscopic slides with predrawn wax-pen circles and incubated with saturating concentrations of mAbs against endothelial adhesion molecules for 30 min at 7°C. Lamina propria leukocytes in suspension were pretreated with an irrelevant negative control mAbs, added to sections (2 × 10⁶ cells in 50 μl of RPMI 1640 medium supplemented with 10% FCS) under constant rotation (60 rpm), and allowed to bind for another 30 min. Thereafter, the nonadherent cells were tilted off, and the adherent cells were fixed to the sections with glutaraldehyde. In other experiments aliquots of the lamina propria cells were preincubated with anti-homing receptor mAbs before adding the cells on top of synovial sections. Thereafter, the assay was continued as described above.

The numbers of small lymphocytes, large immunoblasts, and macrophages bound to synovial vessels were counted under darkfield microscopy. The characteristic size and light-scattering patterns allow clear distinction among these three leukocyte subtypes in this assay (6, 19). Large immunoblasts were defined as clear cells with a diameter at least 1.3 times (and two-dimensional surface area >1.75 times) larger than that of small lymphocytes. Macrophages were large cells with ruffled membrane, which appeared white in the darkfield microscopy. At least 100 vessels were counted from each sample in each individual experiment. Altogether the total numbers of vessel-adherent small lymphocytes, immunoblasts, and macrophages counted in the negative controls were 1079, 670, and 1415 cells, respectively, from 1125 vessels. The effect of mAbs was expressed as the percentage of bound cells in the presence of the given mAb compared with the number of bound cells in the presence of appropriate noninhibiting control mAb. The numbers of independent assays are shown in the figures. The proportions of small lymphocytes, immunoblasts, and macrophages in the input population were also counted microscopically to assess the relative adhesiveness of the three leukocyte subclasses.

Adhesion of lamina propria cells to endothelial cells (Ax) transfected with an expression plasmid encoding VAP-1 or with a vector only was analyzed using a modification of a previously described protocol (23). In brief, Ax cells are a spontaneously immortalized cell line originating from endothelial cells of rat high endothelial vessels. The parental cells have lost their capacity to bind lymphocytes in vitro, but VAP-1 transfection reconstitutes the adhesion cascade (23, 24). The VAP-1 or mock (=control) transfectants were plated on microscopic slides within wax-pen circles. When the cells were confluent, 2 × 10⁶ leukocytes were added to the circle and allowed to bind under constant rotation (60 rpm) at 7°C for 30 min. The nonbound cells were gently washed off by two dippings in cold RPMI 1640, and the adherent leukocytes were fixed to endothelial monolayers in 1% glutaraldehyde. The slides were evaluated under normal light microscope, and the number of adherent leukocytes in predefined area of 6.25 mm² was counted using an ocular grid. The relative adhesion ratio of 1.0 was arbitrarily assigned to describe the number of PBL bound to the Ax VAP-1 transfectants. The relative adhesion ratios of other leukocyte populations were compared with that by taking into account the percentage of each leukocyte subclass in the input population.

The mean ± SEM of the binding experiments are shown. The inhibitory effect of mAbs was compared with negative control treatments using paired, two-tailed Student’s t test.

Because homing of mucosal leukocytes to joints might be important in the pathogenesis of reactive arthritis following gastrointestinal inflammation or infection, we analyzed whether leukocytes isolated from lamina propria of gut specimens obtained from IBD patients can adhere to synovial vessels. Using an in vitro adhesion assay, it was evident that three easily identifiable leukocyte subclasses from IBD gut (small lymphocytes, large immunoblasts, and macrophages) all specifically interacted with chronically inflamed synovial vessels (Fig. 1 A). In contrast, a completely normal synovial membrane is very thin and contains few vessels; consequently, very few leukocytes bound to these vessels, precluding any meaningful further experimentation (data not shown). In the inflamed membrane the relative adhesiveness of the three gut-derived leukocyte types was evaluated as follows. The numbers of different cell types bound per one vessel were counted. The percentage of each cell type in the input population was also microscopically counted. Then, value 1.0 was assigned to the number of small lymphocytes bound per one vessel, and the relative adhesion ratios (assuming that there would be as many (absolute cell number) immunoblasts and macrophages in the starting population as there were small lymphocytes) for the two other cell populations were calculated. These results showed that immunoblasts bound 2.5 ± 0.7 times better (n = 13) and macrophages 14.8 ± 4.3 times better (n = 4) to synovial HEV than did small lamina propria lymphocytes. Taken together, all subpopulations of gut leukocytes bound well to activated joint vessels, and especially the synovia-binding capacity of activated gut immunoblasts and macrophages was remarkable.

To study which lymphocyte-homing receptors mediate the interaction between small lymphocytes and synovial vessels, we took advantage of function-blocking mAbs and the nonstatic in vitro frozen section assay. Inhibition of lymphocyte L-selectin, α₄ integrins, and CD44 resulted in a statistically significant inhibition (Fig. 2). Because better inhibition was obtained by blocking α₄ integrins than by blocking β₇ integrins, both α₄β₁ and α₄β₇ heterodimers probably contribute to synovial binding, especially because α₄β₁ binds better to VCAM-1 than does α₄β₇ (25, 26) (although possible differences in the affinities of mAbs should be taken into account as well). Marginal inhibition was also observed with function-blocking mAbs against PSGL-1 and CD18. On the endothelial side, preincubation of synovial sections with mAbs against ICAM-1, VCAM-1, PNAd, and VAP-1 yielded significant inhibition of binding of small lymphocytes from IBD guts to joint vessels. In contrast, blocking of E- or P-selectin or ICAM-2 did not significantly alter this interaction. Thus, small lamina propria lymphocytes isolated from IBD gut use VAP-1, CD44, and interactions between CD18 integrins and ICAM-1, α₄ integrins and VCAM-1, and L-selectin and PNAd to bind to synovial vessels.

Large lymphocytes activated in the IBD gut used a completely different molecular repertoire to adhere to synovial vessels (Fig. 3). On lymphocyte side blocking of PSGL-1, L-selectin, CD44, α₄ integrins, β₇ integrins, and CD18 were all inefficient in significantly perturbing the interaction. Analyses of the contribution of endothelial adhesion molecules revealed that only VAP-1 (by 46%) and ICAM-1 (by 23%) markedly inhibited this interaction, whereas the other vascular molecules played no significant role. In fact, pretreatment of synovial sections with anti-ICAM-2 mAb actually increased the number of synovia-adherent immunoblasts. Hence, activated mucosal immunoblasts from IBD gut mainly use VAP-1 in synovial adherence, but other, molecularly undefined mechanisms apparently contribute significantly to this interaction.

Because mucosal macrophages also avidly adhered to synovial vessels and might play an important role in carrying gut-derived Ags to joints, the molecular nature of their synovial adherence was determined. This interaction was clearly selectin dependent, because blocking of PSGL-1 on macrophages or its counter-receptor P-selectin on synovial vessels practically abolished binding of this leukocyte subset (Figs. 1,B and 4). Also E-selectin and VAP-1 on the endothelial site contributed significantly to this interaction (Fig. 4). Thus, blocking of the PSGL-1-P-selectin interaction almost totally prevents synovial adherence of mucosal macrophages originated from IBD bowel, whereas it has no significant effect on the same adhesive event in the case of lamina propria lymphocytes.

Because the synthesis of PSGL-1 on mucosal leukocytes has not been analyzed previously, we subjected the leukocytes isolated from lamina propria of IBD patients to immunofluorescence and FACS analyses. These studies revealed that significant numbers of the cells from all subsets of mucosal leukocytes stained with an anti-PSGL-1 mAb, although the PSGL-1 expression was weaker than in blood lymphocytes or monocytes (Fig. 5). The expression was brightest in macrophages and immunoblasts, but ∼30% of small lymphocytes also displayed PSGL-1 positivity. Because it was the macrophages, that showed by far the strongest PSGL-1 dependence in their synovial binding, these data suggest that, like on PBL (27, 28), the expression of PSGL-1 on different leukocyte subpopulations in the lamina propria does not directly correlate to its functional activity.

Because VAP-1 was the only endothelial adhesion molecule involved in binding of all mucosal leukocyte subsets to synovial vessels, we further examined these interactions using a transfectant model. Ax endothelial cells stably transfected with VAP-1 are >95% VAP-1 positive, whereas VAP-1 is completely absent from the mock transfectants (data not shown) (23). These assays showed that small lymphocytes, immunoblasts, and macrophages all adhered to nonstimulated Ax cells transfected with VAP-1 (Fig. 6 and Table II). Only negligible binding was detectable to mock transfectants. When the binding to mock transfectants was subtracted from the binding to VAP-1 transfectants, and the frequency of these three mucosal leukocyte subsets in the input population was determined, we could estimate the relative avidity of binding (Table II). These data from four independent experiments revealed that all gut-derived leukocyte types bound better to Ax VAP-1 transfectants than did lymphocytes isolated from peripheral blood. Moreover, among gut cells immunoblasts bound best to VAP-1 transfectants (relative adhesion ratio, 4.2), followed by small lymphocytes and macrophages (relative adhesion ratios, 3.4 and 2.9, respectively). These results are in line with the data from frozen section assays, because the contributions of VAP-1 to the binding of these leukocyte subsets to synovial vasculature were also almost equal. In contrast, P-selectin, which accounts for the superior adhesion of mucosal macrophages to synovial venules, is lacking from these transfectants (24). Thus, VAP-1 can indeed support binding of all mononuclear leukocyte classes isolated from IBD gut.

We show here that IBD patients have in their bowel mucosal leukocytes that are capable of binding to vessels in joints. We further show that the molecular mechanisms directing synovial adherence of different populations of gut leukocytes in IBD are remarkably different. In IBD small mucosal lymphocytes use multiple adhesion molecules in synovial adherence, whereas binding of activated immunoblasts and macrophages mostly relies on VAP-1- and PSGL-1-P-selectin-mediated interactions, respectively. The recirculation pattern of gut leukocytes from IBD patients may be involved in the pathogenesis or aggravation of reactive arthritis, which is not an unusual complication of these mucosal disorders.

This study is the first one in which synovial binding of leukocytes isolated from the gut of patients suffering from IBD (hereafter referred to as IBD lymphocytes/immunoblasts/macrophages) to synovial vasculature is analyzed. Previously we have analyzed the binding of leukocytes isolated from normal, noninflamed gut (hereafter called normal gut lymphocytes/immunoblasts/macrophages) to vessels in gut, peripheral lymph nodes, and synovium. We have also analyzed binding of IBD cells to mucosal and lymph node venules. From those large materials we know that our synovial samples always contain E-selectin-, P-selectin-, VAP-1-, VCAM-1-, ICAM-1-, and ICAM-2-positive vessels and completely lack MAdCAM-1. Because PNAd is absent from 20% of synovial samples, care was taken to confirm its presence in the samples used for adhesion assays. Earlier flow cytometric analyses by us and others have revealed that IBD leukocytes have the LFA-1^int, CD44^high, β₇^int, α₄ integrin^{low to int}, L-selectin^low phenotype (21, 29). Here we further showed that the cells are modestly positive for PSGL-1.

Comparison of the present results with those from studies with normal gut cells reveals significant alterations in the behavior of IBD leukocytes (Table III). Direct comparison of these data can be performed, because the protocols for cell isolation and adhesion assays have remained exactly the same, and they were all performed in the same laboratory and analyzed by the same persons. In fact, due to the long collection period needed for human gut resectants, the assays with IBD and normal cells have been performed at an overlapping time period, and hence even many of the same synovial samples have been used for both cell types. Our published results show that normal gut lymphocytes can also bind to synovial vessels (6, 19). Of lymphocyte-homing receptors, only α₄ integrin and CD44 blocked this interaction by 25–30%, and on endothelial cells only VAP-1 supported the binding. Interestingly, β₇ integrin or VCAM-1 had no effect on binding of normal gut lymphocytes to synovial vessels. Thus, small IBD lymphocytes displayed a much more promiscuous usage of adhesion molecules in their synovial adherence. In regard to α₄ integrins the observed difference indicates that normal gut lymphocytes may only use α₄β₁ for synovial adhesion, while both α₄β₁ and α₄β₇ may contribute to the synovial adherence of small IBD cells. The finding that multiple adhesion molecules contribute to binding of small IBD cells is in agreement with the concept that the selectivity of gut homing is lost in a severely inflamed gut (21, 30, 31). Under normal conditions small lymphocytes in lamina propria are thought to be mainly memory cells that gain entrance into this nonorganized lymphoid organ primarily by binding of α₄β₇ integrin to mucosal addressin cell adhesion molecule-1 (MAdCAM-1). Under chronic inflammation, many other inflammation-inducible endothelial adhesion molecules appear in lamina propria (21, 32), which help to recruit cells belonging to nonmucosal pools into the intestine as well. Our present findings indicate that isolated IBD small lymphocytes are still equipped with the capacity to bind to these other endothelial determinants. In addition, the residence of IBD leukocytes in an inflammatory microenvironment may increase the functional activity of other homing receptors as well (33). Comparison of the present results to earlier data also shows that small IBD lymphocytes from gut bind to synovial vessels 4 times and IBD blasts 10 times better than small lymphocytes from peripheral blood of IBD patients, showing selectivity of synovial adherence of gut-originating lymphocytes.

Activated mucosal immunoblasts have met their cognate Ag in the intestine. These cells display remarkable dual homing specificity. Both in vitro and in vivo analyses suggest that immunoblasts originating from intestine can bind and home to synovium (6, 7, 19). Notably, normal gut immunoblasts do not adhere to vessels in peripheral lymph node even when it is inflamed (19), indicating that the increased adhesiveness is not just a reflection of general stickiness of activated cells. Normal gut blasts bind to synovial vessels primarily via CD44 interacting with a nonhyaluronate ligand and via ICAM-1-LFA-1- and VAP-1-dependent pathways (6, 19). Synovial adherence of IBD immunoblasts was also partially dependent on the same molecules (Table III). These data suggest that resident and additional IBD gut-infiltrating immunoblasts mainly use the same endothelial determinants for recognizing synovial vasculature.

IBD macrophages bound very well to joint vessels. Although these leukocytes are thought to be mostly sessile cells, there is evidence that tissue macrophages can leave the stroma by reverse transmigration and are indeed found in efferent lymph (34, 35). Therefore, these cells can be envisioned to play an important role in carrying mucosal Ags into distant tissues such as synovium. In joints gut-originated memory cells and immunoblasts, which have already become activated in the intestine against the same Ag, may then very rapidly trigger a reactive inflammation at this nonintestinal location. We are aware that our hypothesis faces two approximations. Firstly, due to the unavailability of surgical synovial samples from joints that are healthy or at the early stages of arthritis progression, the use of chronically inflamed samples was mandatory for us. Secondly, we have measured leukocyte-endothelial cell interactions in vitro using an assay that well reflects but does not necessarily fully reproduce the in vivo migration pathways. Nevertheless, it is the best method available to study leukocyte adhesion to physiologically relevant endothelial cells expressing natural adhesion molecules in humans. Bearing these limitations in mind, we believe that our findings are indicative of a possible connection between mucosal and synovial homing during the pathogenesis and/or aggravation of reactive arthritis in IBD.

Gut-derived macrophages were exquisitely sensitive in their synovial recognition to blockade of the P-selectin-PSGL-1 pathway. Interestingly, Ag-fed macrophages are able to elicit brisk induction of endothelial P-selectin when they are incubated on endothelial cells (36). Thus, these Ag-processing cells might be able to trigger the onset of inflammatory changes on synovial vasculature, which would then lead to increased immigration of this leukocyte class. Moreover, the contribution of VAP-1 to synovial targeting of macrophages derived from IBD bowel deserves special mention. VAP-1 is not involved in blood monocyte adhesion to inflamed vessels in lymph nodes, nor does it support synovial adherence of macrophages isolated from noninflamed gut (6, 15). Hence, in the inflamed gut some nonphysiological stimuli involved in the synthesis of the as yet uncharacterized leukocyte ligand of VAP-1 apparently takes place, possibly on those macrophage subtypes that appear to be unique for IBD intestine (37).

We show here that lamina propria leukocytes from IBD and normal gut use distinct adhesion mechanisms to recognize synovial vessels. Collectively, our analyses of IBD leukocyte binding to mucosal and synovial vasculature suggest that selective interference of either adhesive interaction would be possible. The basal leukocyte trafficking to gut takes place mainly via MAdCAM-1, which, in contrast, is completely absent from synovial vessels (38). Thus, interference with MAdCAM-1 or its α₄β₇ integrin receptor is effective in ameliorating IBD-like inflammations in the gut (39, 40), but probably contributes very little or not at all to leukocyte trafficking to inflamed joints. Blockade of VAP-1, in contrast, would severely affect leukocyte binding to joints, whereas it only minimally influences lymphocyte trafficking into normal gut (41). Moreover, when selective blockade of gut-originating macrophages to joint would be desirable, interference with the PSGL-1-P-selectin pathway would be most feasible. Future trials with antiadhesive therapy will eventually show whether the modulation of cell trafficking will be useful as an adjunctive therapy to combat the disabling joint complications of Crohn’s disease and ulcerative colitis.

We thank Drs. D. Haskard, E. Butcher, and R. McEver for donating mAbs. The expert help of Riikka Lehvonen and Anne Sovikoski-Georgieva is gratefully acknowledged.