TNF-α Released by Comigrating Monocytes Promotes Transendothelial Migration of Activated Lymphocytes¹

Migration of leukocytes across vascular barriers and their accumulation in soft tissues are stimulated by chemotaxins (1) and facilitated by agents that induce cell surface display of complementary leukocytic and endothelial adhesion molecules (2, 3, 4). Agents capable of inducing these processes include products of infectious microorganisms (5), activated complement components (6), reagents generated by the clotting cascade (7, 8), as well as cytokines and chemokines produced by activated endothelial cells, smooth muscle cells, and other vascular appendages (9, 10, 11) (12, 13). Recent investigations using unstimulated mononuclear leukocytes (MNLs)³ suggest that lymphocytes follow the migratory path of monocytes that crossed through an endothelial monolayer 4 h or more before (14). In a model of routine immunosurveillance, unstimulated monocytes, migrating spontaneously in the absence of defined chemotactic stimuli, release sufficient quantities of both TNF-α and IL-1 to induce the adjacent endothelial cells to express CD54 (ICAM-1) and CD106 (VCAM-1) (14, 15). These endothelial adhesion molecules create attachment sites for additional leukocytes. TNF-α is also a chemotaxin for lymphocytes (14) so that monocytes leave a chemotactic trail as they migrate. When lymphocytes or monocytes were activated intravascularly, we postulated that paracrine signaling between these cells stimulates even more rapid transendothelial migration of lymphocytes. Specifically, we postulated that lymphocytes adjacent to activated blood monocytes may become motile, increasing the likelihood that they will adhere to and migrate across nearby endothelial barriers. To evaluate these hypotheses, we used immobilized anti-CD3 to activate freshly isolated leukocytes and tested whether monocyte-derived cytokines increased the motility of lymphocytes and their tendency to migrate across confluent endothelial barriers.

Protocols to obtain informed consent from blood donors were approved by Institutional Review Boards of Baylor College of Medicine (Houston, TX) and Houston Veterans Affairs Medical Center, TX. Blood was collected into preservative-free heparin, and the MNLs were isolated by density-gradient centrifugation over Ficoll-Hypaque (Organon Teknika, Durham, NC). Isolated MNLs were washed with Ca²⁺Mg²⁺-free HBSS (Life Technologies, Gaithersburg, MD), and resuspended in RPMI 1640 with 10% FCS (14). All reagents, including blocking Abs used in these experiments, were unreactive for endotoxin by the Limulus amebocyte assay (Associates of Cape Cod, Woods Hole, MA) that will detect as little as 0.03 E.U./ml. MNLs are not activated during the isolation protocol, as they have the same expression of L-selectin, CD11b/CD18, and CD11a/CD18 as unseparated MNLs in whole blood. MNLs were cultured in polypropylene tubes or Teflon jars to minimize subsequent monocyte activation and/or loss through adherence to the culture vessel walls.

To purify lymphocytes, MNLs were incubated twice with iron beads (Lymphocyte Separation Reagent; Technicon, Tarrytown, NY) at a bead:cell ratio of 4:1, for 30 min at 37°C. Monocytes that had ingested iron beads were removed with a magnet. Residual monocytes were removed by adherence to plastic for 30 min at 37°C. Purified lymphocytes contained <0.1% monocytes, as assessed by flow-cytometric analysis with anti-CD14- and anti-CD33-specific mAbs. Purified monocytes for reconstitution experiments were isolated by adherence to plastic and were recovered by vigorous pipetting and scraping.

To activate them, MNLs or purified lymphocytes were incubated with immobilized anti-CD3 for 2 h at 37°C. Up to 5 × 10⁶ MNLs were added to T25 flasks previously coated with anti-CD3 (OKT3, gift of Dr. Frank Orson, Baylor College of Medicine) or, for control purposes, with Ag nonspecific isotype-matched mouse IgG (= sham-treated MNLs). Flasks were coated with 2.5 ml of Ab at 10 μg/ml for 1 h at 37°C and washed before use. In other experiments, MNLs were cultured with monensin (3 mM; from Quillaja bark; Sigma, St. Louis, MO), LPS (1 μg/ml; from Escherichia coli LPS, serotype O127:B8; Sigma), actinomycin D (5 μg/ml; Sigma), cycloheximide (100 μg/ml; Sigma), or rTNF-α (in varying doses; Genzyme, Piscataway, NY).

Purified lymphocytes were stained with the intracytoplasmic fluorescent dye Cell Tracker Green (Molecular Probes, Eugene, OR) during their incubation with immobilized anti-CD3 or IgG. Purified monocytes were allowed to ingest latex beads conjugated with red fluorescent dye (Molecular Probes). Lymphocytes and monocytes were remixed at a 10:1 ratio and incubated at 37°C for 45 min. Cells were gently loaded into the counting chamber of a hemocytometer to minimize disruption of clusters. Rosettes of green lymphocytes and red monocytes were enumerated by epifluorescent microscopy.

MNLs were resuspended in fresh medium for the migration assay. One-half million MNLs were transferred to 24-well plates containing a confluent monolayer of HUVECs growing on a 3-mm-thick pad of hydrated collagen, as previously described (14). After 2 h at 37°C, nonadherent MNLs were gently removed with four washes using HBSS. Lightly adherent MNLs were removed following a brief exposure to 1× trypsin. MNLs that had migrated into or below the endothelial monolayer were recovered by digesting the whole pad with collagenase (Sigma). Migratory MNLs were stained with lymphocyte- and monocyte-specific mAbs and enumerated with the flow cytometer. Migration into collagen was evaluated using the same methods, except that the collagen pads did not have monolayers of endothelium and the trypsin step was omitted. In certain experiments, in which we sought to nullify the effects of TNF-α, we added 1 μg/ml monoclonal mouse anti-human TNF-α (clone 35G10F3 from Genzyme), 2 μg/ml rat monoclonal anti-TNF R-II (p80, CD120b, clone M1 from Genzyme), 2 μg/ml mouse anti-TNF R-I (p60, clone 16803.1 from R&D Systems, Minneapolis, MN), or 2 μg/ml soluble TNF R-I (R&D Systems). Blocking reagents were present as the MNLs were incubated with the anti-CD3; MNLs were then washed to remove any soluble anti-CD3, and the same blocking reagents were readded and remained present during the subsequent migration assay. Blocking reagents were used in doses capable of neutralizing at least 100-fold more TNF-α than was present in the culture supernatants (see Table I). Because of concern that Abs or soluble receptors may not be able to diffuse through a confluent endothelial monolayer, we also tested these inhibitors in a configuration in which the endothelium was growing on a collagen pad in a tissue culture insert and inhibitor could be added to the basal side of the monolayer. However, we found that inhibitors added to the basal and apical surfaces of the endothelium were no more effective than the same reagents added to the apical surface alone (data not shown).

MNLs were stained with mAbs to specific adhesion markers, as previously described (16). The percentage of cells expressing each marker was identified with two-color fluorescence with monocytes defined as the CD14- or CD33-positive population, and lymphocytes as the CD14- or CD33-negative population within the bitmap for mononuclear leukocytes. Ab to L16, an epitope expressed on LFA-1 when it assumes an activated conformation (clone NKI-L16), was a gift from Dr. C. G. Figdor, University Hospital (Nijmegen, The Netherlands) (17). We also used anti-Mac-1 (CD11b; clone 94) from Coulter (Hialeah, FL), and anti-CD4 (clone 13B8.2), anti-CD8 (clone B9.11), and anti-CD3 (clone UCHT1) from Immunotech (Westbrook, ME).

TNF-α in culture supernatants and in the freeze-thaw lysate of freshly isolated MNLs was measured with the low level ELISA (R&D Systems). To quantify TNF-α message, 9 μg of total RNA was isolated from 5 million MNLs with RNAzol B (Biotex, Friendswood, TX) and reverse transcribed to create a cDNA library (first-strand cDNA synthesis kit; Clontech, Palo Alto, CA). TNF-α cDNA was amplified from this library and quantified by competitive PCR using the reverse-transcriptase PCR Amplimer set and the MIMIC TNF-α kits from Clontech. This assay uses a template (MIMIC) that competes for the same TNF-α primers as the cDNA, but generates a different size PCR product. Constant quantities of cell-derived cDNA are mixed with decreasing quantities of the MIMIC, the mixtures are amplified, and the intensity of the two PCR product bands is compared. When the amplimer generated from the native TNF-α cDNA is of equal concentration to the amplimer generated from the MIMIC, the quantity of TNF-α cDNA and MIMIC is known to be equivalent in that tube.

One hundred million freshly isolated MNLs were pelleted and lysed with three freeze-thaw cycles, followed by incubation in lysis buffer (50 mM Tris, 300 mM NaCl, 0.5% Triton X-100, 10 mg/ml leupeptin, 1 mM PMSF, 10 mg/ml aprotinin, and 1.8 mg/ml iodoacetamide). The lysate was electrophoresed under reducing conditions through 12% SDS-polyacrylamide gels and transferred to nitrocellulose paper. Strips from the nitrocellulose blot were blocked with 100% FCS (Life Technologies) for 1 h, and washed with TTBS (Triton X-100, Tris borate). The strips were incubated with Abs to TNF-α (2 mg/ml in FCS) for 1 h and washed with TTBS. Bound anti-TNF-α Abs were detected with biotinylated F(ab′)₂ goat anti-mouse IgG (Cappel, Durham, NC), followed by streptavidin-alkaline phosphatase and NBT/BCIP (Pierce, Rockford, IL).

Heparinized whole blood samples or isolated MNLs were aliquoted into conical 12 × 75-mm polypropylene tubes, and incubated for 5 min with 50 μl of normal donor plasma to block the cell receptors for Fc. Temperature was maintained at 0 to 4°C throughout the assay by keeping all tubes and reagents immersed in crushed ice. Three microliters of anti-CD14 (clone TUK4; Dako, Carpenteria, CA) were added to identify monocytes. When studying LPS-stimulated cells, the anti-CD14 was replaced with anti-CD33 (mAb clone D3HL60.251; Immunotech, Westbrook, ME), since exposure to LPS decreases monocyte expression of CD14, but does not affect expression of CD33. After 30 min, the cells were washed once with 4 ml of cold Dulbecco’s PBS (DPBS). Cells were then incubated for 30 min with 100 μl of goat anti-mouse IgG (1:32) to coat and cover murine IgG-specific Ags displayed by either the primary anti-CD14 or anti-CD33 mAbs. We verified that this blocking step prevented reaction between the anti-CD14 or anti-CD33 mAbs and the FITC sheep anti-mouse IgG that is added in a later step to detect cytokine-specific Abs.

The externally stained cells were washed with DPBS, fixed with 200 μl of 1% buffered paraformaldehyde in DPBS for exactly 10 min, and washed once more in DPBS. When studying whole blood samples, the erythrocytes were lysed by addition of 1 ml of saponin buffer (0.1% saponin in DPBS with 10 mM HEPES), followed by vortexing for 15 s. The cells were then pelleted, and the supernatant was discarded before proceeding.

At this point, MNLs were simultaneously permeabilized and probed for intracytoplasmic cytokine by the addition of 200 ng of cytokine-specific mAb in 100 μl of saponin buffer. This quantity of buffer, added to MNLs in about 100 μl of DPBS, gave a final saponin concentration of 0.05%. In control tubes, isotype-matched murine IgG (Sigma) was used in lieu of cytokine-specific mAb to establish how much IgG would be incorporated nonspecifically into the cytoplasmic compartment. After 30 min, the cells were washed once with the saponin wash buffer (0.01% saponin in DPBS with 10 mM HEPES) and counterstained with FITC-conjugated sheep anti-mouse IgG diluted 1/120 in saponin wash buffer. The cells were then washed once with saponin wash buffer, and once with DPBS before fixation with 1% paraformaldehyde.

In two-color flow-cytometric analyses, monocytes were identified by the anti-CD14 or anti-CD33 counterstained with phycoerythrin-conjugated anti-mouse IgG; lymphocytes were identified as the phycoerythrin-negative population. Uptake of cytokine-specific Abs was detected by the presence of FITC anti-mouse IgG. Delimiters for the FITC signal were set so that ≤5% of the cells incubated with nonspecific IgG (in lieu of cytokine-specific mAbs) were positive. Results are presented as the net percentage of cells above this threshold that displayed intracytoplasmic (FITC) fluorescence.

To identify a mAb that was highly sensitive to and specific for intracellular TNF-α, we tested seven commercial mAbs purportedly specific for this cytokine. Two of these did not stain saponin-permeabilized MNLs at all, even when used at concentrations as high as 2000 ng/100 μl. The remaining anti-TNF-α mAbs, when tested at saturating concentrations of 200 ng/100 μl, caused intracytoplasmic staining of 90.6 to 92.3% of the monocytes in freshly isolated MNLs. To further evaluate these five mAbs, we analyzed their ability to react with TNF-α, and only TNF-α, in cytoplasmic extracts of freshly isolated MNLs fractionated by PAGE and blotted to nitrocellulose membranes, as described above. We also evaluated the ability of rTNF-α to inhibit the uptake of these mAbs by saponin-permeabilized monocytes. Finally, we use Scatchard analyses to measure the avidity of each of the mAbs for TNF-α.

A total of 5 μg of purified rTNF-α (Genzyme) was conjugated to 1 g of CN-Br Sepharose (Pharmacia, Piscataway, NJ) so that we have a known amount of TNF-α per milliliter of Sepharose slurry. To estimate nonspecific uptake of mAbs to Sepharose beads, we also conjugated known amounts of BSA to Sepharose using the same conditions. Each of the five mAbs to TNF-α was labeled with ¹²⁵I using iodobeads (Pierce); unbound iodine was removed by gel filtration, followed by dialysis against PBS. Aliquots containing 0.5 × 10⁻¹² mol of radiolabeled mAb were incubated with serial dilutions of TNF-α-Sepharose containing between 3 × 10⁻¹² and 0.19 × 10⁻¹² mol of bound TNF-α. Control tubes received equal amounts of BSA-Sepharose. Incubations were performed in triplicate, at room temperature (25°C), for 1 h on a rocking platform in polypropylene tubes precoated with BSA. The Sepharose was then removed by centrifugation and washed twice with PBS containing 1% BSA. We subtracted the quantity of mAb bound to BSA-Sepharose from that bound to TNF-Sepharose to arrive at the quantity of Ab bound specifically to TNF-α on the Sepharose. Molar uptake of mAb was calculated using the known m.w. and sp. act. of the radiolabeled mAb.

The mAb chosen for intracellular identification of TNF-α, clone 35G10F3 from Genzyme, had the highest avidity for this cytokine among the five mAbs that reacted with intracytoplasmic TNF-α. One milligram of purified rTNF-α blocked 98% of the uptake of this Ab by saponin-permeabilized MNLs. When saponin was omitted from all steps of the assay, the anti-TNF-α did not bind to the MNLs, demonstrating that this assay is identifying intracellular, and not surface-bound, TNF-α. Clone 35G10F3 reacted with a single 17-kDa protein in the lysate of freshly isolated MNLs, and the migration of that protein exactly matched that of 17-kDa rTNF-α electrophoresed in on the same gel. Using the same assay protocol, we identified clone B1 (Genzyme) as a reagent to detect intracellular IL-1β, and clone M10 (Genzyme) as a reagent to detect intracellular IL-6. To be able to detect intracellular IL-6, we had to include 3 mM monensin in the media to prevent secretion of the IL-6 as it was synthesized. There were sufficient quantities of TNF-α and IL-1 in the cells for detection, even without monensin. Addition of monensin did not affect the numbers of positive cells or the intensity of the staining for TNF-α or IL-1.

Saponin-treated MNLs, stained for the presence of intracellular TNF-α, were examined by immunofluorescent microscopy with a Nikon Eclipse 800 microscope. Images were captured with an AT 200 CCD camera (Photometrics, Tucson, AZ) using Cellview high resolution fluorescence imaging software (Computer Signal Processing Inc. (CSPI), Billerica, MA). To further clarify the site of the intracellular TNF-α accumulations, images were deconvoluted by exhaustive photon reassignment using the EPR software from Scanalytics (CSPI).

Activation of MNLs by incubation with immobilized anti-CD3 for 2 h induced a 4.80 ± 0.50-fold increase in the number of lymphocytes that migrated across confluent monolayers of endothelium (Fig. 1,A, mean ± SEM of 14 donors, p < 0.01 Wilcoxon) compared with the migration of MNLs incubated with immobilized nonspecific IgG. Anti-CD3 treatment significantly increased transendothelial migration of both CD4 and CD8 T cells, and of both CD45RO and CD45RA T cells (data not shown). Exposure to anti-CD3 also increased locomotor activity, leading to a 1.99 ± 0.25-fold increase in the number of lymphocytes that migrated into matrices of hydrated collagen (Fig. 1 B, mean ± SEM of 14 donors; p < 0.01, Wilcoxon). The effect of anti-CD3 treatment on monocyte migration was relatively modest, with typically less than a 0.1-fold increase in the numbers of migrating monocytes (data not shown).

Purified lymphocytes, activated with anti-CD3, formed rosettes when recombined with monocytes. Anti-CD3-treated lymphocytes formed 4.4 ± 1.2-fold (mean ± SD; p < 0.05, Wilcoxon) more clusters with monocytes than did control lymphocytes incubated with nonspecific IgG. The increased quantity of adhesion molecules expressed on anti-CD3-treated MNLs may have promoted the formation of rosettes. Lymphocyte CD11b (Mac-1) increased 12.6 ± 3.5-fold after stimulation with anti-CD3 (mean ± SD, n = 8 donors; p < 0.01 by Mann-Whitney U) compared with the baseline of 1.7 ± 1.4% positive cells after IgG stimulation. Activated LFA-1 (the L16 epitope on CD11a/CD18) increased by 3.1 ± 0.7-fold following stimulation with anti-CD3 (mean ± SD, n = 9 donors; p < 0.01 by Mann-Whitney U) from a baseline expression on 19.5 ± 6.3% sham-stimulated lymphocytes. Anti-CD3 treatment did not affect the expression of CD62L (L-selectin), CD49d (VLA-4), CD44, or CD31 (platelet endothelial adhesion molecule; PECAM) on lymphocytes (data not shown).

We had shown previously that monocytes migrating through endothelial monolayers promote the migration of resting unstimulated lymphocytes placed above the same endothelial barrier 4 h or more later. We postulated that activated lymphocytes would respond to the same monocyte-derived signals, and would do so more quickly. We found that the increased lymphocyte migration induced by anti-CD3 treatment was totally dependent on the presence of comigrating monocytes. In the absence of monocytes, anti-CD3-stimulated lymphocytes migrated in the same numbers as did sham-stimulated MNLs incubated with nonspecific murine IgG. Readdition of autologous monocytes increased transendothelial migration of the anti-CD3-stimulated lymphocytes in a dose-dependent manner (Fig. 2), and to a much greater degree than was seen with sham-stimulated lymphocytes.

Anti-CD3 stimulation increases the quantity of TNF-α released by MNLs (p < 0.05, Mann-Whitney U test) (Table I). We therefore postulated that TNF-α might participate in the signaling process whereby monocytes enhance the migration of anti-CD3-stimulated lymphocytes. We used three strategies to inhibit the interaction of TNF-α with leukocyte TNF-α receptors: neutralizing Abs to TNF-α, neutralizing Abs to TNF-α receptors (R-I and R-II), and soluble TNF R-I to compete for TNF-α. Blocking agents were present while the MNLs were incubated with anti-CD3, and again during the migration into collagen or across endothelial monolayers. Soluble TNF R-I had the greatest impact on migration, blocking 92 ± 1% of the lymphocyte migration into collagen and 40 ± 3% of the migration across HUVECs (p < 0.05, Mann-Whitney U test). Lymphocytes in the presence of soluble TNF R-I migrated in no greater numbers than did sham-stimulated MNLs (data not shown). Abs to TNF R-I (p60) and R-II (p80) also inhibited migration into collagen (by 84 ± 1% and 79 ± 2%, respectively) and across HUVECs (by 29 ± 8% and 31 ± 8%, respectively). Abs to TNF-α also inhibited lymphocyte migration into collagen and across endothelial cells (Fig. 3).

To further define the role of TNF-α, we removed monocytes from the lymphocyte population before stimulation with anti-CD3 (or IgG). We then added increasing quantities of rTNF-α to the lymphocytes at the onset of the subsequent migration assay to evaluate whether this would reproduce the migration-enhancing effect seen with monocytes. TNF-α, in concentrations between 1 and 20 pg/ml, increased the migration of anti-CD3-stimulated lymphocytes, but not sham-stimulated lymphocytes, into collagen matrices (Fig. 4,A); concentrations of TNF-α above 60 pg/ml did not increase lymphocyte migration. In a different experiment, exogenous TNF-α also increased the migration of anti-CD3-stimulated lymphocytes across endothelial monolayers (Fig. 4 B, p < 0.05, Mann-Whitney U test).

To explore why these concentrations of TNF-α induced anti-CD3-stimulated, but not sham-stimulated lymphocytes to migrate, we looked for differences in the expression of TNF R-I (p60) and TNF R-II (p80) on these two cell populations. Few cells displayed TNF R-II even after 2 h of incubation with anti-CD3. However, after anti-CD3, the cell surface density of TNF R-I (p60) increased fivefold on anti-CD3-stimulated lymphocytes compared with sham-stimulated lymphocytes (Fig. 5 A, p < 0.05, Mann-Whitney U test). When the incubation with anti-CD3 was extended to 4 and 6 h, the expression of TNF R-I on lymphocytes returned to baseline.

TNF R-I was shed from the cells after stimulation with anti-CD3, and could be detected in culture supernatants. After 2 h, the supernatants of anti-CD3-stimulated cells contained 19 ± 1 pg/ml of TNF R-I, whereas supernatants of sham-stimulated leukocytes contained only 7 ± 1 pg/ml of TNF R-I (p < 0.05, Mann-Whitney U test). Four hours after stimulation, the levels of soluble TNF R-I in the supernatants of anti-CD3- and nonspecific IgG-stimulated cells were not significantly different.

The migratory behavior of the lymphocytes paralleled the changes in TNF R-I expression. There was a sixfold increase in the numbers of lymphocytes that migrated into collagen after 2 h of stimulation with anti-CD3 relative to nonspecific IgG-stimulated lymphocytes (Fig. 5 B; p < 0.05, Mann-Whitney U test). However, after 4- and 6-h anti-CD3 stimulation, the numbers of migrating lymphocytes dropped.

Our results suggest that, for a brief period following anti-CD3 stimulation, lymphocytes become highly responsive to the migration-enhancing effects of TNF-α. We postulated that circulating monocytes contain intracytoplasmic stores of TNF-α and synthetically active TNF-α mRNA that allow them to rapidly synthesize and release TNF-α upon stimulation. We found 68 copies of TNF-α message per nanogram of RNA in MNLs isolated within 20 min of venipuncture. This is equivalent to one copy of TNF-α message per eight MNLs (data not shown).

Circulating MNLs also contain small, but detectable stores of TNF-α protein. In two experiments, we found that lysates of well-washed freshly isolated MNLs contained, respectively, 0.05 ± 0.01 and 0.07 ± 0.02 pg of TNF-α protein/10⁶ MNLs. In addition to their preformed stores, MNLs can produce TNF-α very rapidly upon stimulation. One million freshly isolated MNLs secreted 25 ± 11 pg of TNF-α after only 1 h of stimulation with LPS in vitro (Table II). The rapid production of TNF-α depended, at least in part, on the preexisting TNF-α mRNA. Despite addition of actinomycin D to block synthesis of new message, MNLs could still release 6.5 ± 0.1 pg of TNF-α after 1 h in culture, and 15.4 ± 1.3 pg after 2 h.

To identify which MNLs contained TNF-α, we used avid anti-TNF-α mAbs to probe saponin-permeabilized MNLs. We found TNF-α in 88 ± 14% of monocytes in whole blood, and approximately the same frequency in freshly isolated CD14⁺ monocytes (Fig. 6). If not permeabilized with saponin before addition of anti-TNF-α, less than 1% of the MNLs were positive, which indicated that we were detecting intracytoplasmic and not membrane-bound TNF-α (data not shown). Further confirmation was obtained with fluorescent microscopic images of these cells that showed the TNF-α in cytoplasmic vesicles (Fig. 7).

When placed in culture, freshly isolated monocytes initially released much of their preformed TNF-α. This release was accelerated by stimulation with immobilized anti-CD3 (Fig. 8,A) or LPS (Fig. 8, B and C). However, if the cells were kept in culture for an additional 2 or 3 h, the TNF-α-specific staining returned to initial levels, suggesting that the MNLs replenished their intracellular stores. The recovery was abrogated by addition of cycloheximide to block new protein synthesis, but not by actinomycin D that blocks synthesis of mRNA (Fig. 8 C).

The kinetics of TNF-α production are very different from those for IL-1 or IL-6. Very few circulating monocytes contained IL-1 (4 ± 1% of monocytes; mean ± SD of 20 donors), and none contained IL-6. When placed in tissue culture, unstimulated monocytes gradually begin to produce IL-1, so that by 5 h in vitro, 12% of the cells were reactive (Fig. 9,A). LPS stimulation greatly increased the number of monocytes that produced IL-1 (Fig. 9,A), but IL-1 production could be completely blocked by either cycloheximide or actinomycin D (Fig. 9 B). The kinetics of IL-6 production were even more delayed. IL-6-positive monocytes were not detectable until 6 h or more after stimulation with LPS. Despite maximal LPS stimulation for 12 h or more, only a subset of monocytes could be induced to produce IL-1 or IL-6 (data not shown).

These studies were undertaken to investigate potential paracrine interactions between circulating monocytes and lymphocytes that may stimulate transendothelial migration of activated T lymphocytes and monocytes. Lymphocytes incubated with immobilized anti-CD3 increased their expression of CD11b/CD18 (MAC-1) and changed their membrane CD11a/CD18 (LFA-1) to a higher affinity conformation (L16⁺). Monocytes in the same sample concomitantly increased their display of CD54 (ICAM-1), a ligand for both MAC-1 and LFA-1. Possibly as a result, anti-CD3-stimulated lymphocytes clustered in rosettes with monocytes. This created a microenvironment likely to facilitate signaling between lymphocytes and monocytes. One result of this signaling was a prompt increase in lymphocyte locomotor activity that enhanced the ability of these cells to penetrate collagenous tissue matrices and confluent endothelial barriers.

The increased migratory behavior of anti-CD3-stimulated lymphocytes appeared to depend on TNF-α released into the microenvironment shared with these lymphocytes. Indeed, the ability of the monocytes to up-regulate migration of activated lymphocytes could be replaced by concentrations of rTNF-α between 1 and 20 pg/ml. Further support for the notion that TNF-α drives migration of anti-CD3-activated lymphocytes came from experiments that showed that addition of Abs to TNF-α or TNF-α receptors opposed the migration-enhancing effects of immobilized anti-CD3. The most effective inhibitor in our assay was soluble TNF R-I. This is consistent with its known high affinity for soluble TNF-α. Furthermore, soluble TNF R-I is almost one-tenth the size of IgG, and may be better able to diffuse into the microenvironment between clustering monocytes and lymphocytes and block TNF-mediated signals between these cells.

It is not clear which lymphocyte receptor for TNF is involved in this process. Blocking Abs specific for TNF R-I and Abs specific for TNF R-II inhibited the migration of anti-CD3-activated lymphocytes. Some data suggest that TNF R-I may have a more dominant role. Anti-TNF R-I was able to inhibit the migration of activated lymphocytes when used at doses 100-fold more dilute than those shown in Figure 3; anti-TNF R-II, used at this dilute concentration, had no inhibitory effect (data not shown). Second, anti-CD3 stimulation up-regulated the expression of TNF R-I on lymphocytes, but had no effect on TNF R-II expression. For this reason, we postulated that the increased migration seen after anti-CD3 stimulation may be related to the increased expression of TNF R-I on the activated cells. On the other hand, TNF R-II is particularly able to interact with transmembrane TNF-α and may be a critical participant in juxtacrine interactions between clustered lymphocytes and monocytes (18).

Several lines of evidence suggested that circulating mononuclear leukocytes actively produce TNF-α, can release stores of this cytokine, and will promptly produce more when appropriately stimulated. Perhaps the most compelling evidence for this conclusion came from flow-cytometric analyses of saponin-permeabilized monocytes in freshly drawn whole blood samples (Fig. 6), and our studies showing that stimulation with immobilized anti-CD3, like stimulation with LPS, transiently reduces monocyte intracellular stores of TNF-α (Fig. 8). More than 80% of circulating monocytes in healthy donors contain intracytoplasmic stores of TNF-α. Considering that a microliter of blood normally has at least 100 monocytes, there is a high likelihood that there will be sufficient numbers of TNF-α-positive monocytes to ensure timely release of this cytokine when effective agonists are encountered within the microvasculature. In contrast, less than 5% of circulating monocytes contain intracellular stores of IL-1, and none contain IL-6. Following maximal stimulation with LPS, it took 3 h or more in culture for 60 to 80% of the monocytes to produce IL-1 and more than 6 h before any IL-6-producing cells could be detected. This provides further support for the idea that TNF-α is more likely than IL-1 or IL-6 to be responsible for the increase in lymphocyte locomotor activity observed in these studies.

We stimulated MNLs with immobilized anti-CD3 and, in control experiments, with bacterial LPS, because these agonists mimic stimuli likely to be encountered by lymphocytes and monocytes, respectively, under physiologic conditions. Both stimuli cause monocytes to release TNF-α. Activation of monocytes by LPS is understandably achieved by signaling through CD14 (19). The use of LPS allows us to evaluate the immediate kinetics of TNF-α release after direct monocyte stimulation. It was not readily apparent, however, why immobilized anti-CD3 might stimulate monocytes to release TNF-α. We postulate that solubilized anti-CD3 molecules may adhere to the lymphocyte surface and serve as a bridge to link lymphocyte CD3 and monocyte FcR. In addition, stimulation of T cells via the CD3-TCR complex activates LFA-1 and up-regulates Mac-1 (CD11b), facilitating adhesive interactions with monocytes displaying the counterligand CD54 (ICAM-1). These adhesive interactions are likely to enhance monocyte receipt of stimuli from anti-CD3-stimulated T cells that cause shedding of membrane-bound TNF-α (20, 21) and other signals that stimulate biosynthesis and continued release of this cytokine into the microenvironment shared by these T cells.

Several studies indicate that, when activated, lymphocytes move from the circulation into tissues in significantly greater numbers as compared with resting lymphocytes (22, 23). Activation of monocytes also promotes their transendothelial migration, as shown by studies in which an injection of LPS in vivo induces accumulation of monocytes into all organs (24). Monocyte products are likely to influence the migration of activated lymphocytes in several ways: as they migrate, monocytes release agents that stimulate expression of endothelial cell adhesion molecules (14, 15). Activated lymphocytes, expressing increased quantities of the reciprocal adhesion molecules, will adhere to and migrate through the activated endothelium (25, 26). Migrating monocytes may also release chemokines and cytokines such as TNF-α, monocyte-chemotactic protein-1, macrophage-inflammatory protein-1α, and macrophage-inflammatory protein-1β (13, 14, 27, 28, 29) that direct the migration of activated lymphocytes (30, 31). We show in this study that lymphocytes are particularly likely to be responsive to TNF-α signaling due to the transient increase in cell surface display of TNF R-1 induced by anti-CD3 stimulation.

We recognize that monocytes, as they differentiate into tissue macrophages, may also begin to produce metalloproteinases that can disrupt endothelial monolayers and the cross-linkages between tissue matrix proteins (32, 33). However, it is unlikely that metalloproteinase-mediated events contribute to our results since our migration assay is completed in 6 h from the time that MNLs are isolated from blood. Indeed, we have studied the effects of inhibitors of monocyte/macrophage-derived metalloproteinases (TIMP (tissue inhibitor of metalloproteinase)-1 and TIMP-2, at 1 μg/ml) in our model, and find that use of these agents does not suppress migration of resting or anti-CD3-activated lymphocytes (data not shown).

The present investigations were prompted by our recent studies of HIV-1-infected patients who have a high percentage of activated blood leukocytes (16). In these patients, the fraction of lymphocytes that migrate through confluent endothelial barriers correlated significantly with the quantity of TNF-α released by the migrating MNLs (34). We postulate that the TNF-α-driven migration of activated T lymphocytes, demonstrated in this study, may be a mechanism used not only in HIV-1 infections, but also in many systemic vasculitides and in organ-specific inflammatory diseases to stimulate lymphocytes to migrate across endothelial barriers to form perivascular and soft tissue leukocyte infiltrates.