|Home | About | Journals | Submit | Contact Us | Français|
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/us/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
The kinetics and regulatory mechanisms of T cell migration through the endothelium have not been fully defined. In experimental, filter-based assays in vitro, transmigration of lymphocytes takes hours, compared with minutes, in vivo. We cultured endothelial cell (EC) monolayers on filters, solid substrates, or collagen gels and treated them with TNF-α, IFN-γ, or both prior to analysis of lymphocyte migration in the presence or absence of flow. PBL, CD4+ cells, or CD8+ cells took many hours to migrate through EC-filter constructs for all cytokine treatments. However, direct microscopic observations of EC filters, which had been mounted in a flow chamber, showed that PBL crossed the endothelial monolayer in minutes and were highly motile in the subendothelial space. Migration through EC was also observed on clear plastic, with or without flow. After a brief settling without flow, PBL and isolated CD3+ or CD4+ cells crossed EC in minutes, but the numbers of migrated cells varied little with time. Close observation revealed that lymphocytes migrated back and forth continuously across endothelium. Under flow, migration kinetics and the proportions migrating back and forth were altered little. On collagen gels, PBL again crossed EC in minutes and migrated back and forth but showed little penetration of the gel over hours. In contrast, neutrophils migrated efficiently through EC and into gels. These observations suggest a novel model for lymphoid migration in which EC support migration but retain lymphocytes (as opposed to neutrophils), and additional signal(s) are required for onward migration.
Lymphocytes of all classes must migrate through the endothelium to home to lymph nodes or to enter inflamed or infected tissue. In the context of inflammation, flowing cells are captured by specialized, fast-acting adhesion receptors (such as VCAM-1 and E- or P-selectin), presented by venular endothelial cells (EC) responding to cytokines such as TNF-α (TNF), IL-1β (IL-1), or IFN-γ (IFN) . Initial capture is followed by activation of the lymphocytes by surface-presented chemokine(s), which induce integrin activation and stabilization of adhesion, followed by migration over and through the endothelial monolayer. In vitro, flow-based assays have shown that depending on the stimulus applied to the EC, T cell capture is possible through VCAM-1, E-selectin, or P-selectin (although efficiency for each may depend on the T cell subset) and that stable adhesion is mediated through binding of activated α4β1-integrin to VCAM-1 and αLβ2-integrin to ICAM-1 [2,3,4]. Transendothelial migration of T cells has been observed within minutes of adhesion in such flow systems for TNF-treated EC (where blockade of β2-integrins was inhibitory) , for EC stimulated with TNF plus IFN , and for EC that had been stimulated with TNF and had stromal-derived factor-1α (SDF; CXCL12) or CCL19 (EBV-induced molecule 1 ligand chemokine) added to their surface . In the last study, little migration was seen without an added chemokine (CXCL12 or CCL9), and even then, migration was much more effective in the presence of flow than if flow were stopped.
In contrast, most studies about the regulation of lymphocyte migration through endothelium have used static assays, in which EC have been grown on porous filters. These necessarily test migration away from the subendothelial space as well as through the EC, and periods from 2 h to 36 h are required for migration through the construct (e.g., refs. [7,8,9]). In such systems, T cells migrated spontaneously through unstimulated endothelium/filters over time [7, 10, 11], although the proportion migrating only reached ~10% of those added. Perhaps surprisingly, treatment of the EC with a range of cytokines (IL-1, TNF, IFN, or TNF+IFN) caused modest [12, 13] or negligible [8, 10, 14] increases in T cell transmigration compared with unstimulated EC, again over prolonged periods. Addition of inflammatory chemokines, such as CCL2 (MCP-1), CCL3 (MIP-1α), or CCL5 (RANTES) below the filter increased migration of memory T cells slightly, although CXCL12 (a homeostatic chemokine) caused a marked increase in the migration for naïve and memory T cells . Transendothelial migration studies have also been carried out for EC grown directly on collagen gels, where quite a small proportion of added T cells (~10%) migrated into the gel over 2–4 h, and cytokine stimulation of the EC again had little effect [15, 16].
Considering the T cell phenotype, memory CD4+ T cells migrated more efficiently than naive cells through resting or cytokine-treated EC (e.g., refs. [14, 16]) and in the presence of a chemotactic gradient . Transmigration of T cells tends to be increased by stimulation prior to assay, for instance, incubation with phorbol dibutyrate or IL-2, culture for 2–24 h (which up-regulates expression of the αLβ2-integrin), or differentiation over weeks [7,8,9,10]. Nevertheless, studies using freshly isolated lymphocytes without manipulation reported migration through resting or IL-1β-stimulated EC at 24 h [11, 14] comparable with that observed in others studies when T cells were cultured overnight (e.g., ref. ). It should be noted that in the studies described above, where transendothelial migration occurred in minutes under flow, the T cells had first been subjected to prolonged incubation in presence  or absence of IL-2 [3, 6].
The studies outlined above leave some uncertainties about the rate at which migration occurs, first through the endothelial monolayer and then away from it and the requirement for flow or for preactivation of T cells or EC for efficient transendothelial migration. Hours are required for T cells to cross EC/filter constructs or enter collagen gels below EC, as opposed to minutes to cross EC alone, suggesting that assays are strongly influenced by migration through the subendothelial matrix and/or filter itself. In vivo, it is likely that naive as well as memory T cells are recruited to peripheral tissue , and PBL are evidently recruited without prior activation. Although direct studies of the kinetics of lymphocyte recruitment across inflamed endothelium are lacking, studies in the rat have shown that T cells are recruited across the vessel wall in the peripheral lymph nodes and Peyer’s patch within 30–40 min of their infusion [18, 19]. To investigate these problems, we studied kinetics of lymphocyte migration through EC cultured on filters, plastic wells, or collagen gels and treated with different cytokines and compared results in static and flow-based assays. Previously, we used similar approaches to analyze kinetics of neutrophil migration through endothelial monolayers and filters [20,21,22]. Here, we found that lymphocytes could cross endothelial monolayers in minutes with little evidence of a requirement for flow. The migrated cells were highly motile, and the prolonged periods required to transit filter systems could not be attributed to transendothelial migration itself. However, we made the novel observations that some lymphocytes underwent multiple transits back and forth across endothelial monolayers and that few motile, transmigrated lymphocytes entered collagen gels (e.g., compared with neutrophils). We suggest that EC tend to retain migrating lymphocytes and that a separate signal is required to overcome this and allow lymphocytes to move into the stroma.
Venous blood from healthy individuals was collected in EDTA tubes (Sarstedt, Leicester, UK). PBMC were isolated by centrifugation of blood on histopaque 1077, and PBL were prepared by panning PBMC on culture plastic to remove monocytes . In some experiments, to obtain activated T cells, PBMC were cultured for 7 days in the presence of 10 μg/ml PHA (Sigma-Aldrich, Poole, UK), which initially induced clumping and proliferation of lymphocytes; however, by Day 7, cells had dispersed and were spherical and smaller than freshly isolated PBL (judged by Coulter Counter volume distribution). In other experiments, T cells (CD3+) or CD4+ T cells were purified from PBMC by negative selection using magnetic Dynabeads (Dynal, Wirral, UK) and a cocktail of mAb to remove cells bearing CD19, CD11b (HIB19 and ICRF44, respectively, both from BD PharMingen, Oxford, UK), CD14, CD16 (RM052 and 3G8, respectively, Beckman Coulter, UK), and in the case of CD4+ T cell selection, CD8 (OKT8, eBioscience, UK) . Isolated cells were washed, counted, and adjusted to a final concentration of 2 × 106/ml in PBS containing Ca2+ and Mg2+ or Medium 199 (Gibco Invitrogen Compounds, Paisley, Scotland), supplemented with 0.15% BSA (PBSA or M199BSA, respectively; Sigma-Aldrich). On a few occasions, neutrophils were isolated using a two-step density gradient as described [20,21,22] and suspended at 2 × 106/ml in M199BSA.
HUVEC were isolated from umbilical cords as described previously  and cultured in M199 supplemented with 20% FCS, 10 ng/ml epidermal growth factor, 35 μg/ml gentamycin, 1 μg/ml hydrocortisone (all from Sigma-Aldrich), and 2.5 μg/ml amphotericin B (Gibco Invitrogen Compounds). Primary HUVEC were dissociated using trypsin/EDTA (Sigma-Aldrich) and seeded on six-well tissue-culture plates (Falcon, Becton Dickinson Labware, Franklin Lakes, NJ, USA), uncoated, low-density, 3.0 μm pore polycarbonate Transwell filters (which were placed in matching plates; BD PharMingen), glass chamber slides (Lab-tek, Nalge Nunc International, Naperville, IL, USA), or collagen gels (see below). Seeding density was chosen to yield confluent monolayers within 24 h. TNF (100 U/ml, Sigma-Aldrich) and/or IFN (10 ng/ml, Peprotech Inc., London, UK) was added to confluent monolayers for 4 h or 24 h before the assay with neutrophils or lymphocytes, respectively.
To form collagen gels, type 1 collagen, dissolved in 0.6% acetic acid (2.15 mg/ml; First Link Ltd., West Midlands, UK), was mixed with 10× M199 and FCS (1.66 ml, 0.1 ml, and 0.34 ml, respectively). The pH was neutralized by addition of 0.15 ml 1 N NaOH, and 1 ml was dispensed into a six-well plate and allowed to gel at 37°C. The gel was then equilibrated with HUVEC culture medium for 48 h before seeding and culture with HUVEC as above. In some experiments, CXCL10 (IFN-inducible protein-10) or CXCL12 (SDF; 80 or 800 ng/ml, Peprotech Inc.) was added to the collagen after neutralization with NaOH. The gel was polymerized as above and equilibrated with HUVEC culture medium containing chemokine at the same concentration.
Lymphocyte migration was assessed using 24-well format Transwell filters. HUVEC were washed to remove residual cytokines, fresh M199 + BSA was placed in the lower chamber, and PBL were added to the upper chamber. The lymphocytes were allowed to settle, adhere, and migrate through HUVEC at 37°C in a CO2 incubator for the desired period. Migration was stopped at the chosen time by transferring the filter into a fresh well, leaving the transmigrated cells in the original lower chamber. The lymphocytes suspended in the upper chamber were removed and pooled with cells obtained when the filter was washed twice. These cells were taken to represent nonadherent lymphocytes. The nonadherent and transmigrated cells were counted using a Coulter Multisizer II (Coulter Electronics Ltd., Essex, UK). From the known number of added lymphocytes, the percentage of lymphocytes that adhered and the percentage of lymphocytes that transmigrated were calculated.
In some experiments, the surface phenotypes of adherent and transmigrated lymphocytes were assessed by flow cytometry. Freshly isolated, nonadherent (upper chamber) or transmigrated (lower chamber) lymphocytes were labeled with anti-CD4-PE or anti-CD8-FITC (Becton Dickinson, Oxford, UK) for 30 min on ice. Fixed volume counts for positively labeled cells were made using a Coulter XL flow cytometer and analyzed using WinMDI. In this way, we calculated the percentage adhesion and transmigration for CD4+ and CD8+ T cell subsets.
Adhesion and transmigration were assessed by direct microscopic observation as described previously . HUVEC in six-well plates were washed with PBSA to remove residual cytokines, and purified PBL or CD3+ or CD4+ T cells were added for 5 min. Nonadherent cells were removed from the HUVEC by gentle washing with PBSA (which took 2 min), and video recordings of the endothelial surface were made using phase-contrast videomicroscopy as follows: Five fields were briefly recorded immediately after washing to analyze the number of adherent cells and their position above or below the monolayer (see below); a single field was recorded for 15 min to follow lymphocyte behavior; a further five fields were recorded briefly to analyze the position of lymphocytes above or below the monolayer. Manipulations and microscopy were carried out inside a Perspex box held at 37°C.
The video recordings were digitized and analyzed off-line using Image-Pro Plus software (DataCell Ltd., Finchampstead, UK). The numbers of adherent cells were counted in the video fields, averaged, and then converted to cells per mm2 using the calibrated microscope field dimensions and multiplied by the known surface area of the HUVEC to calculate the total number adherent. This number was divided by the known total number of lymphocytes added to obtain the percentage of the lymphocytes that had adhered. Each lymphocyte was classified as phase-bright with a round or distorted shape and adherent to the surface of the HUVEC and phase-dark and spread and migrating below the HUVEC. The percentage of adherent lymphocytes that had transmigrated was calculated at each time, with time zero taken as the end of the settling period. The migration velocities of phase-dark lymphocytes underneath the HUVEC were measured by digitizing a sequence of images 1 min apart for 6 min. In each digitized image, cells were outlined, and the position of their centroid was determined. Migration velocity (μm/min) was the average distance moved by the centroid per minute.
HUVEC on collagen gels were washed with M199 + BSA to remove residual cytokines, and purified PBL, PHA-stimulated PBL, or neutrophils were added for 10 min. Nonadherent cells were removed from the HUVEC by gentle washing with PBSA, and phase-contrast video-microscope recording were made 0.25 h, 1 h, 3 h, and 24 h after the original addition of leukocytes. Five video fields were recorded. In each field, images were first recorded at the endothelial surface, and then recordings were made as the microscope was focused gradually down in 50 μm steps. Cells visible with the endothelial monolayer were counted and divided into those that were phase-bright (above EC) and those that were phase-dark (just below EC). Cells within each 50 μm step were counted as they came into focus; these cells were typically irregular in shape and phase-bright. The focal depth of the gels was ~300 μm. After averaging counts in the five fields, data were expressed as the percentage of the adherent cells in each vertical region. On some occasions, a single field at the endothelial surface was recorded for 5 min to analyze the migratory behavior of the leukocytes. All manipulations of gels and microscopy were carried out at 37°C.
Filters coated with HUVEC were cut from the Transwell holders and placed on a 75 × 25-mm coverslip. Alternatively, the bases of chamber slides coated with HUVEC were freed from the fluid reservoirs attached to their surface. The coverslips or slides were incorporated into a parallel plate flow chamber and attached to a perfusion system mounted on the stage of a phase-contrast videomicroscope enclosed in a Perspex chamber at 37°C, as described [21, 26]. The flow channel dimensions were 20 × 4 × 0.13 mm (length×width×depth) for the filters and 50 × 10 × 0.25 mm for the chamber slides. At one end, they were connected to a Harvard withdrawal syringe pump, which delivered flow at a rate equivalent to a wall shear stress of 0.1 Pa. At the other end, they were connected to an electronic switching valve (Lee Products, Gerards Cross, UK), which selected flow from two reservoirs containing PBL in PBSA or cell-free PBSA. A 4-min bolus of PBL was perfused over the HUVEC followed by cell-free wash buffer. Video recordings were made of a series of microscope fields along the centerline of the flow channel after 2 and 11 min of washout, and between these times, a single field was recorded.
Video recordings were analyzed essentially as above, except that lymphocytes adherent to the surface of HUVEC could be classified as rolling-adherent (spherical cells moving over the surface much slower than free-flowing cells) or stationary-adherent (typically with distorted shape and actually migrating slowly on the surface). Phase-dark, transmigrated cells could be counted for either substrate, but when observing filters in the flow chamber, recordings were also made of the underside of the filter, where a few cells might be found. This was achieved by focusing the microscope stage up and down 10 μm (i.e., the filter thickness). The sum of the numbers adherent in all categories was divided by the number perfused during the bolus to obtain total PBL adhesion as percent of cells perfused.
Effects of multiple treatments were tested using ANOVA, followed by comparison with control by Dunnett test. Single treatments were compared with controls by paired t-test.
Settling of lymphocytes onto EC filter constructs and quantification of the number collected from the back have been widely used to assess “transendothelial” migration. In initial experiments with unstimulated or TNF-treated HUVEC, we found that few PBL migrated through the filter after 2 h or 4 h (e.g., 2.4±1.0% of added cells migrated through TNF-stimulated HUVEC at 4 h; mean±sem; n=3). The proportion increased by 24 h (e.g., 11.0±3.2% of added cells that migrated through TNF-stimulated HUVEC; mean±sem; n=4), and so, we made comparisons between variously treated HUVEC at this time. Figure 1 shows that EC treated with cytokines (TNF or IFN alone or together) tended to support greater lymphocyte transmigration compared with unstimulated HUVEC, although there was no consistent difference between the cytokine treatments. This trend was also evident when CD4+ or CD8+ T cells were analyzed separately (Fig. 1), and the two types of T cells behaved similarly to each other. The proportion of lymphocytes that was adherent was high after 24 h (~50%) and not affected significantly by cytokine treatments (data not shown). Such long contact times are not physiological and presumably increase nonspecific background adhesion. We thus reduced the initial contact time by washing off nonadherent lymphocytes after 10 min and maintained the 24-h migration endpoint. This decreased lymphocyte adhesion, and there was now a tendency toward greater adhesion for EC stimulated with TNF + IFN compared with untreated cells (21.8±3.7% vs. 14.8±6.0% of added PBL adherent, respectively; mean±SEM from six experiments). However, transmigration was also much lower (3.4±0.6% vs. 1.6±0.5% of added PBL transmigrated, respectively).
These findings showed that PBL and the major T cell subclasses took hours to migrate through endothelial/filter constructs, a distance of only ~10 μm, and that cytokine stimulation of the EC only increased recruitment about twofold. Lymphocyte recruitment across the wall of inflamed vessels is expected to be more rapid and stimulus-specific. Thus, we observed lymphocyte interactions with HUVEC treated with TNF plus IFN (the treatment inducing most efficient transmigration) on filters under flow conditions. Few flowing PBL adhered to unstimulated HUVEC cultured on filters, but many more were adherent when the EC had been stimulated with TNF plus IFN (0.6±0.3% vs. 5.8±2.0% of PBL perfused, respectively; mean±sem; n=4). The cells adherent to the cytokine-treated EC were attached firmly with only a small percentage rolling, and within 11 min, ~30% had migrated through the endothelial monolayer (Fig. 2). At this time, few lymphocytes had migrated through the filter itself (Fig. 2). The migration velocities of the phase-dark cells under the HUVEC averaged ~5 μm/min (4.9±0.6 μm/min; mean±sem of means from three experiments).
Thus, lymphocytes adhered and migrated quickly through cytokine-stimulated monolayers in the presence of flow. They then migrated freely under the monolayer but did not appear below the filter within minutes. Indeed, it took hours to negotiate the filter in the static assay. As direct observation of kinetics of migration through the endothelium or the filter was not possible in the “standard” static filter assay, it is not possible to conclude at this stage the exact step at which the hold-up occurred. It is possible that transendothelial migration was slower in the absence, compared with the presence, of flow  or that lymphocytes quickly crossed the endothelial monolayer but were held up by the filter in either case. To clarify this point, we compared microscopic observations of migration kinetics through HUVEC under static or flow conditions using clear, solid substrates.
When PBL were allowed to settle for 5 min, few (~5%) adhered to unstimulated HUVEC cultured in multiwell plates, but cytokine-stimulated EC supported much higher levels of attachment (Fig. 3A). We were surprised to find that 2 min after washing, a significant proportion of the adherent cells had transmigrated and that after a further 15 min, this proportion remained essentially the same (Fig. 3B). Transmigration was higher for cytokine-treated monolayers, particularly in the presence of IFN. Although some adherent cells did transmigrate through unstimulated HUVEC in minutes, the absolute number observed was small as a result of the low level of adhesion. Examining migration through cytokine-treated EC in more detail, we recorded individual fields and repeatedly assessed transmigration. Although there were minor fluctuations in the proportion of PBL transmigrated, there were no significant upward trends (Fig. 3C). We also noted the velocity of migrated cells, which averaged ~8 μm/min and tended to be faster in the presence of IFN (Table 1A). Thus, in a static assay with short initial contact times, we detected cytokine-specific induction of lymphocyte adhesion and could observe transendothelial migration within minutes.
The detailed analysis of individual fields revealed another unexpected phenomenon. PBL were seen to continue transmigrating (going from phase-bright to phase-dark) throughout the entire 15-min period, but other cells were migrating in the opposite (basal-to-apical) direction. The continual forward and reverse migration explained the nearly constant level of transmigration observed at any time. Some cells made several transits back and forth within the observation period, and some stayed in the same compartment throughout. To quantify this behavior, we followed individual cells second-by-second over a period of 6 min and recorded if and when they moved between the basal and apical surfaces of the HUVEC. A sequence of pictures of a multiply migrating cell is shown in Figure 4A, and some typical behaviors are illustrated schematically in Figure 4B. For cytokine-treated HUVEC, ~20% of PBL made at least one transit during the observation period, and about one-third of these made multiple transits (Table 1A). By measuring the number of transits during the 6-min period, we calculated the average interval between transits for those cells that moved between compartments (mean ~4 min; Table 1A). The values varied little between the cytokine treatments (TNF, IFN, or TNF+IFN). Transits were seen with unstimulated HUVEC (data not shown), but again, the number of cells observed was small. On average, half of the transits were forward and half in the reverse direction, which was consistent with the observation noted above—that overall levels of transmigration were constant over the total observation period.
We wondered whether specific lymphocyte subpopulations might be more efficient in migration or prone to reverse migration and so, carried out a series of experiments comparing PBL with purified CD3+ and CD4+ T cells from the same donors. We observed no significant difference in the proportion of these lymphocyte populations adhering or transmigrating through EC stimulated with TNF + IFN (Fig. 5). In addition, the multiple transit behavior of purified T cell subpopulations and the velocities of migrated cells were not significantly different from PBL (Table 1B).
Unstimulated EC cultured in chamber slides consistently failed to recruit flowing lymphocytes (Fig. 6A). However, after cytokine stimulation, the EC efficiently captured much greater numbers of flowing lymphocytes, with greater adhesion observed on EC stimulated with TNF or TNF + IFN compared with IFN alone (Fig. 6A). As with HUVEC cultured on filters, few captured lymphocytes rolled, and most adhered firmly. Nearly one-half of the adherent PBL transmigrated through the cytokine-stimulated monolayers within 11 min, and IFN was the most effective inducer of migration (Fig. 4B). Velocity of migrated cells averaged ~8 μm/min, which is similar to that observed in the static assay, and again, velocity tended to be faster in the presence of IFN (Table 1C).
We checked whether multiple transits were observed under flow as well as in the static assay. In fact, the proportions of cells undergoing at least one transit or undergoing multiple transits in a 6-min period were similar, if a little higher, in the presence of flow (Table 1C). Overall, the average intervals between transits were nearly identical to the values obtained in the static assay (Table 1). Again, values for these variables were similar for the different cytokine treatments.
The above findings suggested that under static or flow conditions, lymphocytes crossed endothelial monolayers quickly but were reluctant to move on from the subendothelial space and that this might have been linked to repeated migration back and forth. However, even the filters represent a solid barrier to migration over most of their surface, and so, we decided to observe migration out of the subendothelium into collagen gels over prolonged periods. First, we analyzed neutrophil migration through TNF-treated HUVEC into gels, as we had characterized neutrophils previously in all of the other models used here (e.g., refs. [21, 22, 26]). Figure 7A shows changes in the distribution of neutrophils above and just below the EC and in the gel over time. After 15 min, a high proportion of adherent neutrophils had penetrated the monolayer and were visible, phase-dark just below it. By 1 h, a few more cells had migrated under the EC, but nearly all of the transmigrated cells were now found in the gel. The cells moved further into the gel by 3 h, and by 24 h, the neutrophils were essentially evenly distributed throughout the depth of the gel. The lymphocytes were much less efficient in entering the gels (Fig. 7B). A significant proportion of adherent cells had migrated through the endothelial monolayer within 15 min, as in the other models. This proportion increased to ~50% after 1 h, but <5% were found in the gel at this time. By 3 h, ~10% were in the gel, but these had not penetrated far compared with the neutrophils. Even at 24 h, only 10% were in gel, and most of these were still in the first 100 μm.
We also recorded the behavior of the lymphocytes at the endothelial surface after 15 min to allow comparison with the observations on solid substrate. Again, we observed cells undergoing forward and backward migration, with some making multiple transits through the endothelial monolayer, at frequencies comparable with those seen on the clear plastic (Table 1D). The phase-dark cells were highly motile and had velocity averaging 5–6 μm/min (Table 1D).
We considered whether activated T cells would migrate more efficiently into gels. However, although PHA activation increased lymphocyte migration significantly through the endothelial monolayer (64.6±3.5% vs. 45.2±6.4% of adherent cells migrated at 3 h following PHA treatment vs. freshly isolated PBL, respectively; mean±sem; n=3; P<0.05 by paired t-test), migration into the gel was not altered significantly (9.1±3.9% vs. 5.1±4.9% migrated at 24 h with or without PHA, respectively; mean±sem; n=3). Furthermore, we still observed multiple transits back and forth across EC by PHA-activated lymphocytes. In fact, the proportions of cells undergoing at least one transit (~14%) or undergoing multiple transits (~5%) in a 6-min period were similar to freshly isolated lymphocytes. Thus, activation was insufficient to induce migration away from the subendothelial space, suggesting the need for a second signal, presumably from within the tissue.
Chemokines CXCL10 and CXCL12 were added to gels at a concentration (80 ng/ml) that would be expected to increase chemotaxis through filters (unpublished observations). However, in two experiments with each, PBL migration through cytokine-treated endothelium or into the gel were not increased, and the tendency to migrate back and forth across the endothelial monolayer remained unaltered (data not shown). The chemokines most likely diffused across the EC and into the surrounding medium during endothelial culture, washing, and the adhesion assay itself, diluting any gradient. We thus increased the concentration of CXCL10 added to the gel tenfold and observed a small but consistent increase in the migration of lymphocytes into the gel when compared with untreated gels (Fig. 7C). Of note, the total level of transmigration across the endothelial monolayer was not increased (averaging 51.2±9.8% or 54.3±5.6% for gels with or without chemokine, respectively; mean±sem from three experiments measured at 24 h), but the presence of CXCL10 tended to reduce lymphocyte migration back and forth across the endothelial monolayer. The proportion of adherent cells undergoing one or more transit over 6 min was reduced from 14.2 ± 4.8% to 6.9 ± 2.2%, and the proportion undergoing more than one transit was reduced from 6.9 ± 2.3% to 3.9 ± 1.8% (mean±sem from three experiments), although these trends did not reach statistical significance. Thus, the presence of a chemokine, such as CXCL10, could promote lymphocyte migration away from the endothelium into the underlying matrix.
Using direct microscopic observation of endothelial monolayers treated with different cytokines, we found that freshly isolated PBL could migrate across endothelial monolayers in minutes in the presence or absence of flow. Although the lymphocytes migrated at 5–10 μm/min underneath the EC, they did not move quickly through 10 μm-thick porous filters or into collagen gels. Interestingly, a significant proportion of PBL could be seen to migrate back and forth across the endothelial monolayer, sometimes repeatedly. In consequence, the proportion under the endothelium did not vary much over ~15 min. The above phenomena could be observed for CD3+ T cells (which would be expected to make up the great majority of PBL) and the CD4+ T cell subset, as well as PHA-activated cells. In nonvisual, static, filter-based assays, little transmigration was detected within hours, and 24 h were needed to obtain a proportion of lymphocytes under the filter comparable with that seen in minutes during direct observations of transendothelial migration. After 24 h, penetration of gels by lymphocytes was inefficient (e.g., compared with neutrophils). Given the speed at which lymphocytes were seen to migrate under EC, it seems that EC tended to retain lymphocytes in their vicinity and that a signal to migrate through stroma and/or across the filters was lacking. Studies in which exogenous chemokine was added to collagen gels, and penetration of the matrix was increased supported this concept.
Efficient migration of T cells through EC treated with TNF and IFN (reaching ~40% of those adherent) has been described previously under flow conditions similar to those used here . In studies using TNF alone, Luscinskas et al.  observed that migration occurred in minutes but did not quantify the proportion of cells migrating. Cinamon et al.  found that only ~5% of adherent T cells migrated through TNF-treated EC under flow, but this proportion increased greatly when SDF was added to the endothelial surface. Most striking was their observation that if flow were stopped, there was negligible transmigration. In each of these flow-based studies, isolated T cells were preincubated for prolonged periods, presumably to induce activation and greater migration, although Piali et al.  only stated this explicitly.
We did not observe marked differences in migration behavior in the presence or absence of flow, which is required to model the capture processes operating in the vasculature, and here, as expected, cytokine-treated EC supported much greater adhesion than unstimulated EC in the presence of flow. However, we found that the proportion of adherent cells transmigrating and the frequency with which cells migrated in and out of the monolayer were similar, with or without flow for EC treated with TNF, IFN, or both. IFN tended to induce the highest level of migration and to induce lymphocytes to migrate more rapidly, but the trends were similar with or without flow. Throughout the study, treatment of EC with cytokines increased the efficiency of transmigration, as well as capture in flow-based assays, but migration did not require addition of an exogenous chemokine. Previously, the presence of flow and binding of SDF to the surface of TNF-treated HUVEC were found necessary to obtain efficient transendothelial migration of T cells . However, the definition of transmigration was different from that used here (only cells that were observed to change once from phase-bright to phase-dark were counted), and the T cells had been cultured for 15–18 h before analysis.
Lymphocytes were not purposefully activated in most of our studies on the basis that PBL are recruited directly from the circulation during inflammation in vivo. Although this recruitment may be more efficient for memory cells, naive cells are also recruited to nonlymphoid tissue . Here, migration behavior was broadly similar for PBL, T cells, and CD4+ and CD8+ subsets. We have found preferential migration of CD45RA-negative (i.e., memory phenotype), CD4+, or CD8+ cells through Transwell filters (assessed using flow cytometry; unpublished observations) in line with reports by others [15, 16, 28]. Thus, it is likely that PBL migrating through endothelial monolayers were enriched in memory cells, but preseparation of naive and memory populations would be required to define their relative migration efficiencies, e.g., in flow models. In comparing results here with previous work with activated T cells, it may be worth noting that there is considerable inter-donor variation in migration of T cells (e.g., compared with neutrophils) in our experience. Gathering data from several studies in our laboratory, we have observed transendothelial migration through EC treated with TNF and IFN in direct microscopic flow assays between 2% and 50% of adherent PBL (mean=17%; n=26). When we did study activated lymphocytes, they migrated more efficiently through cytokine-stimulated endothelial monolayers than resting lymphocytes in agreement with previous reports [7,8,9,10]. Despite this, the activated lymphocytes did not migrate efficiently into an underlying collagen gel, indicating that activation per se was insufficient to generate the necessary signals for migration away from the subendothelial space. Nevertheless, the possibility that antigen presentation by EC specifically facilitates migration of cognate T cells cannot be discounted .
The data presented here reveal some problems relating to interpretation of widely used, filter-based migration assays. Using prolonged incubations under static conditions, there is a high level of lymphocyte adhesion and significant transmigration through filters, even without cytokine stimulation of EC. With cytokine treatment, lymphocytes still take far longer to cross filters than to cross endothelial monolayers and far longer than is required to enter tissue in vivo. When we reduced the initial contact time between PBL and EC before wash-off of nonadherent cells, fewer PBL adhered, but the proportion of the adherent cells that migrated through the filter was still small after many hours. Practically, this suggests that mechanisms found to support or regulate migration in such assays might not be specific to the transendothelial (as opposed to the trans-filter) stage.
In our experience, neutrophils migrate through endothelial monolayers and migrate underneath the monolayers at similar rates to the lymphocytes observed here, but a larger proportion can be observed directly to migrate through filters in minutes in a flow assay . In static trans-filter assays, neutrophils can be collected as early as 15–30 min after addition to the upper surface. Here, we found that neutrophils that had migrated through EC quickly moved into collagen gels and became uniformly dispersed with time. This suggests random diffusion into the gel. Lymphocytes did not “diffuse” in this way, and indeed, it seems that most were actually retained in or near the EC layer. The inefficiency of lymphocyte migration away from the subendothelial space was linked with the tendency of many migrated cells to move back through the monolayer to the luminal surface in minutes. We have not observed such behavior with neutrophils. Taken together, these observations suggest that lymphocytes continually interacted with EC and that a critical signal was lacking in the in vitro models, which was required to drive migration of lymphocytes away from endothelium into tissue. Here, the addition of CXCL10 to the collagen under EC moderately enhanced lymphocyte migration into the gel and reduced the tendency to move back and forth across the endothelial monolayer. In general, stromal chemoattractants may increase efficiency of migration into tissues for all leukocytes, but it seems that neutrophils, as opposed to lymphocytes, do not need such a signal to free themselves from the endothelium. Given the heterogeneity in the chemokine receptors expressed by different lymphocytes, more than one subendothelial chemokine might be required to induce efficient migration of the entire population. Conversely, stromal production of a restricted chemokine subset could provide a further “subendothelial” level of control within the recruitment cascade.
Reverse migration has been described for monocytes over hours after migration though HUVEC cultured on amniotic tissue  or over tens of minutes under flow after migration through TNF-treated HUVEC cultured in glass capillaries . We described reverse migration of neutrophils over hours in a similar flow model . Reverse migration of neutrophils has since been observed directly in an inflamed microvessel of the zebra fish . Here, some PBL and T cells shuttled back and forth between these compartments over minutes. Repeat migration of lymphocytes between the lumen and subendothelial spaces has not been described previously. Nearly 10% of all adherent cells made more than one transit in a 6-min period, and of cells that started underneath the endothelial monolayer, over 40% underwent reverse migration in the same time. Whether comparable behavior occurs in vivo is unknown, and we are not aware of any published, real-time observations of migration of lymphocyte (as opposed to neutrophils) in inflamed vessels.
In conclusion, lymphocytes (and specifically T cells) migrated across EC in minutes and were highly motile thereafter, and some at least migrated back and forth across the endothelium repeatedly. The presence of flow improved the specificity of the assays (in the sense that binding to unstimulated EC was negligible and effects of cytokines more clear-cut than in static assays) but did not strongly influence the initial migration or the back and forth movement in the models used here. It appears that lymphocytes may be actively retained by EC and require a signal from another source to induce their migration onward from the subendothelial space. Consequently, the process of lymphocyte recruitment appears to involve an additional, regulated stage, in all probability to minimize nonspecific, subset recruitment.
This work was supported by The Wellcome Trust (Grant #077828). Umbilical cords were collected with the assistance of the Birmingham Women’s Health Care National Health Service Trust.