How Endothelial Cells Regulate Transmigration of Leukocytes in the Inflammatory Response

The Point of No Return in the Inflammatory Response

The inflammatory response, the body’s rather stereotyped response to tissue damage of any kind, evolved to heal wounds and fight infections. However, recent improvements in human health mean that we are living long enough to experience the other side of the double-edged sword of inflammation. There has not been time enough on an evolutionary scale to select against the baggage that comes with our ability to heal wounds and kill foreign microorganisms. Inflammation is at the root of virtually all pathology. Thus, there has been great interest in understanding how inflammation is regulated and so to design pharmaceutical and other interventions to control it.

Acute inflammation involves soluble and cellular mediators that act over the first minutes to hours of the inflammatory response. At the site of inflammation, generally in postcapillary venules, inflammatory mediators such as histamine secreted by mast cells cause local endothelial permeability, allowing preformed agents such as antibody and complement to cross endothelial junctions into the tissue. It is a common misconception that the increase in vascular permeability is causally related to increased leukocyte extravasation. Although these phenomena are both part of the acute inflammatory response, peak vascular leakage occurs well before leukocyte extravasation begins.¹ Furthermore, the gaps between endothelial cells created by these agents measure only a few hundred angstroms (tens of nanometers), not nearly large enough to allow white blood cells to pass through. Other mechanisms are therefore required to allow leukocytes into the area. Recent in vivo studies demonstrate conclusively that vascular permeability and transendothelial migration (TEM) are separate phenomena. Mice made genetically deficient in the actin-bundling protein cortactin have vessels that are constitutively leaky and have an exaggerated vascular leak response to inflammation. However, leukocyte emigration in these mice is actually reduced, compared with wild-type littermates, because of reduced intercellular adhesion molecule 1 (ICAM-1) clustering and therefore reduced leukocyte adhesion and transmigration.²

The Multistep Extravasation Paradigm

The process of leukocyte extravasation has been broken down into a series of adhesive and signaling interactions between leukocytes and endothelial cells. Over the past two decades, we have learned much about the cellular and molecular interactions that take place during this process.^3–5 Histamine, thrombin, and other mediators of acute inflammation stimulate exocytosis of Weibel–Palade bodies from endothelial cells, bringing P-selectin to the luminal surface. The same increased vascular permeability that allows plasma to leave the bloodstream causes local hemoconcentration, which decreases the rate of blood flow, allowing selectin ligands expressed on the surface of the slowly moving leukocytes to contact P-selectin and thus causing the leukocytes to roll along the luminal surface. Rolling brings the leukocytes close enough for their chemokine receptors to be activated by chemokines attached to the endothelial surface via glycosaminoglycans. Under certain conditions, other cell-surface receptors (eg, platelet-activating factor receptor) bind to and are activated by nonchemokine ligands expressed on the endothelial surface. The activated receptors, through an inside-out signaling cascade, activate leukocyte integrins for tighter adhesion to endothelial surface. This leads first to slow rolling on the endothelium and then to firm arrest. Attached leukocytes then locomote over the endothelial surface, using primarily the β₂ integrin CD11b/CD18 bound to ICAM-1^6–8 to reach the junction. When viewed in vivo by intravital microscopy, leukocytes are sometimes seen to migrate over several endothelial cells (occasionally migrating against the direction of blood flow) before they cross the endothelial cells, usually passing through the endothelial junctions.

Rolling, activation, adhesion, and locomotion (intraluminal crawling) are all critical for leukocyte extravasation, but these are reversible. In fact, most leukocytes that enter a venule at the site of inflammation do not roll, most leukocytes that roll do not adhere, and most that adhere do not extravasate.⁹ However, once leukocytes make the commitment to cross the endothelium into the tissue, with notable exceptions,¹⁰ they never go back—at least not as the same type of cell. And all of the good, the bad, and the ugly of inflammation takes place once leukocytes cross blood vessels. Thus, diapedesis or TEM is a logical process to study if one hopes to control an ongoing inflammatory response.^5,11

Diapedesis Is a Distinct Step in TEM

Electron microscopy studies of leukocytes in the process of TEM show that the leukocytes squeeze in amoeboid fashion between tightly apposed endothelial cells.^12,13 Neutrophils have the pharmacological armamentarium to blow a hole in the vessel wall, but instead of barging in like a SWAT team, they sneak in like spies, minimizing leakage of vascular contents into the interstitial tissues. It was this tight apposition of leukocyte and endothelial cell membranes that fascinated me. Considering that it takes only 1 to 2 minutes for the leukocyte to completely traverse the endothelial border in vitro⁶ or in vivo,¹⁰ this suggests that very specialized membrane–membrane interactions are taking place. In the 1980s, when I started my training, leukocytes were assumed to direct the inflammatory response. They were known to have receptors for specific chemoattractants as well as adhesion molecules (later defined as integrins) promoting their migration. Endothelial cells were assumed to serve only as a relatively inert Teflon-like lining to blood vessels, preventing blood from clotting. However, the ability to culture endothelial cells in vitro, which was pioneered at the Gimbrone and Jaffe laboratories in the early 1970s,^14,15 catapulted vascular biology into the molecular era. It soon became apparent that endothelial cells play a major role in the regulation of most metabolic and physiological processes, including blood pressure, glucose metabolism, thrombosis, hemostasis, and inflammation. I wondered whether there were specific endothelial molecules at the junctions important for intercellular adhesion that regulate diapedesis, in the same way that the selectins and vascular cell adhesion molecule 1 (VCAM-1) and ICAM-1 on the apical surface are important for leukocyte rolling and adhesion, respectively.^16–19

Adhesion Molecules at the Junction with Unique Roles in TEM

Monoclonal antibodies (mAbs) were made against endothelial cells and selected for those that bound to antigens concentrated at endothelial cell borders. This strategy yielded mAbs that recognize platelet endothelial cell adhesion molecule (PECAM),²⁰ vascular endothelial cadherin (VE-cadherin),²¹ and CD99.²² Blocking antibodies against the homophilic interaction domain of PECAM, when added to either leukocytes or endothelial cells in in vitro transmigration assays, block TEM as completely as adding the antibodies to both cell types at the same time, demonstrating the homophilic nature of this interaction.²³ Moreover, blocking PECAM–PECAM interactions has no effect on adhesion of leukocytes to²³ or locomotion on⁶ the endothelium. Removing the PECAM blockade allows leukocytes to resume transmigration.²³ These experiments demonstrate that TEM is a distinct step in leukocyte emigration that is molecularly dissectable from the other steps. This finding has been replicated in many models of acute inflammation in vivo, and antibodies or soluble PECAM-Fc chimeras that interfere with the homophilic interaction of mouse PECAM (or genetic deletion of PECAM) block inflammation in peritonitis,^24,25 dermatitis,²⁶ arthritis,²⁷ experimental autoimmune encephalitis,²⁸ and antigen-specific T-cell trafficking in the central nervous system.²⁹ Anti-PECAM antibodies also block inflammation in various cat and rat models (reviewed in ref. 9).

This role for PECAM in extravasation has been borne out in all mouse strains tested except for C57BL/6 mice,^30,31 which have a way to bypass the requirement for PECAM for leukocyte extravasation. In the thioglycollate peritonitis model, this trait maps to a small region on the short arm of mouse chromosome 2.³¹ It is important to note that PECAM has a separate immunosuppressive role in T cells,^32,33 and knocking out PECAM in C57BL/6 mice may in some cases lead to hyperactive immune responses.^34,35 Thus, animal models using PECAM-deficient mice of the C57BL/6 strain that are measuring both the afferent and efferent limbs of the immune response may show enhanced disease.³⁶ Nevertheless, observations with PECAM led the way for discovery of other endothelial cell molecules that regulate TEM. Some of these, such as CD99 and poliovirus receptor (CD155), appear to function only in the process of diapedesis. Others, such as ICAM-1, ICAM-2, VCAM-1, and a number of the junctional adhesion molecule (JAM) family members, have functions in earlier steps in the pathway, as well as potential roles in diapedesis.^5,11

The LBRC Is Critical for Paracellular Transmigration

During TEM, the membrane from the LBRC is redirected. Instead of returning evenly along the cell junction, it traffics to the site where the leukocyte is transmigrating. The membrane surrounding the transmigrating leukocyte is enriched with membrane from the LBRC for the entire duration of its passage (Figure 2). We call this process targeted recycling of the LBRC. It is not just an epiphenomenon—it appears to be the sine qua non for transmigration.

Everything that we can do to block targeted recycling blocks TEM. Interfering with leukocyte and endothelial cell PECAM–PECAM interactions blocks targeted recycling and TEM.⁴⁴ The LBRC is trafficked along cortical microtubules, so depolymerizing microtubules or blocking kinesin molecular motors under proper experimental conditions selectively blocks targeted recycling of the LBRC and blocks TEM.⁴⁷ The appearance of the blocked leukocytes in all cases is the same. There is no effect on adhesion or ability to migrate to the endothelial cell border; leukocytes are arrested over the junctions and are unable to initiate TEM.⁴⁷ Mutating tyrosine 663 on the cytoplasmic tail of PECAM blocks the ability of PECAM to enter and exit the LBRC and to engage in targeted recycling. Cells in which native PECAM is replaced with PECAM Y663F are unable to support TEM, but normal levels of TEM can be rescued by re-expressing wild-type PECAM.⁴⁸

Antibodies against endothelial cell PECAM, CD155, and CD99 block TEM, but only when added at 37°C, conditions under which they enter the LBRC.^22,23,38 If the same antibodies are added to endothelial cells at 4°C, conditions under which they saturably bind to all of their antigens present at the junction but do not enter the LBRC, there is no blockade of TEM when the unbound antibody is washed away and leukocytes are added at 37°C.^22,23,38 Thus, it is possible that only the fraction of junctional adhesion molecules within the LBRC is required for TEM.

The LBRC Is Critical for Transcellular Transmigration

Targeted recycling of the LBRC seems to be necessary for TEM of neutrophils, monocytes, and lymphocytes under all conditions tested—even conditions in which we could not block TEM by anti-PECAM antibodies. However, what highlights the importance of the LBRC in TEM is the exception that proves the rule. Greater than 90% of leukocyte TEM takes place at endothelial cell borders, but under certain conditions leukocytes can migrate through the endothelial cell cytoplasm in what is called transcellular migration. Under certain in vitro conditions, up to 30% of neutrophils could be made to migrate transcellularly.^49,50 Generally, these are conditions in which the leukocyte is activated directly by a chemokine or other agent exposed to it on the apical surface of cytokine-activated endothelium. Zahara Mamdouh and Alexei Mikhailov⁴⁶ found that, during transcellular migration, the LBRC migrates from the endothelial border to surround the leukocyte for the duration of its passage. Microtubule depolymerization, which prevents movement of the LBRC, blocks both transcellular and paracellular transmigration. Furthermore, anti-PECAM or an anti-CD99 mAb can block transcellular migration once it has begun, even though these molecules are not present on the apical surface of the endothelial cells.⁴⁶ Thus, the LBRC traffics to wherever the leukocyte is transmigrating, even if it is far from the endothelial border.

ICAM-1 and VCAM-1 Clustering

As a prelude to TEM, ICAM-1 and VCAM-1 cluster under the leukocyte as it approaches the junction.^51–53 Clustering of ICAM-1 requires c-Src-dependent phosphorylation of cortactin to start the remodeling of actin^54,55 and to perpetuate ICAM-1 clustering,⁵ along with joining of tetraspanin molecules to these clusters.⁵⁶ Although it is not universally observed, under certain in vitro conditions the apical surface membrane appears to extend upward, to partially envelop the leukocyte during TEM.^50,57 These so-called transmigratory cups are enriched in ICAM-1, VCAM-1, and tetraspanins overlying a core of actin filaments.^49,58

Clustering of ICAM-1 induces Src phosphorylation,⁵⁹ which is required for several steps in TEM^38,54,60 (in addition to the role in activating cortactin). ICAM-1 clustering also activates RhoA.^61,62 Interestingly, just like the self-enforcing interactions between cortactin phosphorylation and ICAM-1 clustering, active RhoA enhances ICAM-1 clustering.⁶³ RhoA in turn activates Rho kinase (ROCK), which inactivates protein phosphatase 1c and inhibits the inactivation of myosin light chain kinase (MLCK). The net result is promotion of the activity of MLCK, which tugs on actin filaments and effectively loosens the junctions.

Calcium Signaling

Elevation of endothelial cytosolic free calcium ion concentration ([Ca⁺²]_i) is known to be required for transmigration to proceed. Blocking the ability of endothelial cells to flux calcium using a cell-permeable calcium chelator leads to the same phenotype as blocking PECAM, in which leukocytes arrest on the apical surface of the endothelium, tightly adherent but unable to transmigrate.⁶⁴ It has been demonstrated that increased [Ca⁺²]_i leads to activation of MLCK through calmodulin (CaM).⁶⁵ The source of the increased [Ca⁺²]_i is not clear, although VCAM-1 clustering and cross-linking of P- or E-selectin⁶⁶ have been implicated. MLCK activates nonmuscle myosin in endothelial cells for interaction with actin filaments, leading to tension development within the cell, which likely promotes loosening of the junctions.

Movement of VE-Cadherin

Although we have been focusing on the recruitment of the LBRC to the site of transmigration, there is extensive evidence that certain membrane proteins, in particular VE-cadherin and its associated catenins, transiently leave the site of TEM.^67,68 Under resting conditions, vascular endothelial cell protein tyrosine phosphatase (VE-PTP) associates with VE-cadherin, the main adhesion molecule of endothelial cell adherens junctions, and maintains it in a state of reduced phosphorylation at the endothelial junctions.⁶⁹ Leukocyte adhesion to activated endothelial cells triggers dissociation of VE-PTP from VE-cadherin, allowing phosphorylation of the latter.^69,70 A recent study⁷¹ suggests that this occurs through a signaling pathway in which VCAM-1 activates Rac1, which activates endothelial NADPH oxidase, the reactive oxygen metabolites of which activate Pyk2. Pyk2 is hypothesized to phosphorylate an alternative substrate of VE-PTP, which then displaces binding to VE-cadherin.

Displacement of VE-PTP from VE-cadherin is the first step in its dissociation from the membrane. This allows access to VE-cadherin by protein kinases. VE-cadherin is tethered to the cortical actin cytoskeleton by α-catenin, to which it binds via p120 and β-catenin.⁷² Phosphorylation of VE-cadherin on tyrosine 658 by Src and on tyrosine 731 by Pyk2 inhibits binding of p120 and β-catenin, respectively. Dissociation of VE-cadherin from p120 and β-catenin facilitate its removal from the adherens junction. These interactions not only regulate endothelial cell permeability, but also regulate TEM. Mutation of tyrosine residues 658 or 731 to phenylalanine inhibits the release of p120- and β-catenin, maintains more VE-cadherin at the cell border, and reduces TEM.⁷² This demonstrates that removal of VE-cadherin from the endothelial cell border at the site of TEM is not just an epiphenomenon—it is a requirement for transmigration. The relationship between the influx of membrane from the LBRC and the apparent removal of VE-cadherin from the site of TEM is the subject of ongoing study. In Figure 3, I summarize and try to integrate some of the major adhesion and signaling events that occur during the transmigration process.

Other Molecules Involved in TEM

Many molecules are reported to play a role in TEM, because antagonizing these molecules blocks leukocyte migration in transmigration assays. Many of these studies do not distinguish between a true block in diapedesis and a block in leukocyte adhesion, chemotaxis, ability to migrate per se, detachment from the underside of the endothelial cell, migration across the extracellular matrix, or (in some assays) migration through the pores of a filter support.⁷³ However, even discounting such findings, several other endothelial cell molecules have been reported to play a role in mediating TEM, in that blockade by antibodies or knockdown of these molecules reduces TEM. In addition to PECAM, CD155, CD99, and the recently reported CD99L2,^74–76 these other molecules include JAM-A,^77,78 JAM-C,^10,79 ICAM-2,^43,80,81 CD47,⁸² MUC18 (CD146),⁸³ activated leukocyte cell adhesion molecule (ALCAM),⁸⁴ nepmucin,⁸⁵ and even the purinergic receptor P2Y₂.⁸⁶ Restoration of TEM by re-expressing a candidate molecule after it has been knocked down is a more reliable way to establish that the molecule plays a role in transmigration than by simply knocking it down or blocking its function. This has now been demonstrated with a number of molecules.⁵ Nevertheless, to our knowledge, PECAM is the only molecule demonstrated to play a role in TEM by a true gain of function. Expression of PECAM in PECAM-deficient ECV-304 cells imparts the ability to support transmigration in a manner that could be blocked by anti-PECAM.⁴⁸ Interestingly, at baseline these cells express ICAM-1 and CD99; however, in the absence of PECAM, this is not sufficient to support transmigration.⁴⁸

Where these molecules exert their control over TEM and how they relate to PECAM, CD99, and the LBRC are questions for future research. In fact, how all of the mechanisms discussed in this section relate to each other is not known. Do they occur sequentially? Is there a hierarchy? One consideration is that inflammation, although a stereotyped response, is not identical at all times and in all places. Depending on the stimulus, the vascular bed, and the timing, different leukocyte types and mechanisms may predominate. However, considering that during diapedesis the leading edge of the leukocyte is under the endothelial cell interacting with the basement membrane while the trailing uropod is still on the apical surface, it is likely that many of these molecular interactions and mechanisms occur simultaneously in different parts of the endothelial cell and leukocyte.

Isolating the LBRC

Significant progress has been achieved in our understanding of the TEM process. However, as with all important scientific problems, the answers to one question lead to even more questions. How is the LBRC formed, and what is it composed of? We are taking two approaches to answering these questions. Homogenization and subcellular fractionation of endothelial cells on sucrose density gradients is yielding some important information. David Sullivan in my research group (unpublished data) took advantage of the fact that the LBRC is accessible by antibodies at 37°C but not at 4°C to bind HRP–conjugated antibody to PECAM on the surface of the cell border but not in the LBRC. After homogenization and flotation away from contaminating organelles, membrane vesicles were briefly exposed to 3,3′-diaminobenzidine (DAB) and H₂O₂. In vesicles containing membrane from the endothelial cell surface, the HRP oxidizes DAB to insoluble multimers that precipitate on the membrane and increase the density of the vesicles containing them. On the other hand, vesicles of LBRC do not contain HRP–labeled antibody and consequently are considerably less dense than the vesicles containing surface membrane. The spectrum of proteins seen is compared to that seen when both surface and LBRC are labeled. This variation of the DAB density shift^87,88 is being used to purify LBRC membrane and analyze it for novel LBRC-associated proteins. Dr. Sullivan has found an unsuspected cytosolic protein associated with the LBRC that appears to play a major role in its targeted recycling (unpublished data).

What Directs Molecules Into and Out of the LBRC?

We have previously demonstrated that mutation of tyrosine 663 on the cytoplasmic tail of PECAM blocks TEM without affecting classic PECAM signaling.⁶⁰ We found that this mutation interfered with the ability of PECAM to enter and leave the LBRC. This suggests that Y663 might be a part of a localization signal that allows PECAM to traffic into the LBRC. Dr. Gong Feng in my research group has been making chimeric molecules in which the extracellular and cytoplasmic domains of molecules known to be included in the LBRC (eg, PECAM) and those known to be excluded from the LBRC (eg, VE-cadherin) are swapped to determine the rules for sorting of molecules into the LBRC (unpublished data).

The Minimal Machinery Necessary for TEM

Interest in transcellular migration of leukocytes has been far greater than its physiological presence might suggest. Direct in vivo measurements show that polymorphonuclear leukocytes migrate transcellularly less than 10% of the time.¹⁰ Nonetheless, we are interested in this phenomenon because it could allow us to understand the mechanisms necessary for transmigration itself, without the potential confounding factors related to junctional integrity. Drs. Annette Gonzalez and Alexei Mikhailov have been studying this process using human umbilical vein endothelial cells (HUVECs) in which the endogenous PECAM has been replaced by PECAM-eGFP. Studying monocyte–endothelial cell interactions by live-cell imaging, they have found that leukocytes tightly adherent to the apical surface of the endothelial cells attract a cloud of fluorescent PECAM-bearing membrane (presumably the LBRC, as in ref. 45) that follows them as they locomote along the apical surface. When a monocyte transmigrates, the cloud coalesces into a ring that tightly surrounds the leukocyte as it migrates through the endothelial cell, contracts as the uropod vanishes beneath the endothelial cell, and dissipates after transcellular migration. Presumably, the membrane reverts to interconnected tubulovesicular structures and relocalizes to the perijunctional zone, but optical limitations preclude direct observation.

Which molecular motors and adaptor proteins are responsible for moving the LBRC toward the leukocyte and recovering it after TEM? Bita Cyrus in my research group is studying these questions using a combination of biochemical, microinjection, and knockdown strategies.

The Plasticity of Endothelial Junctions

One of the difficulties limiting study of the mechanisms of transcellular migration is that even under the most conducive circumstances only approximately 30% of polymorphonuclear leukocytes and approximately 10% of monocytes migrate transcellularly. It has been postulated that transcellular migration occurs more frequently at the blood–brain barrier, where true tight junctions exist and paracellular migration would be expected to be particularly difficult.^58,89 To study the mechanisms of transcellular TEM in a robust system, Ryan Winger in my research group has developed ways to make the endothelial cell borders tighter. Using a combination of methods reported in the literature to decrease junctional permeability, he has been able to routinely grow endothelial cell monolayers with transendothelial electrical resistances of 250 Ω·cm², which is as high as has been reported in vitro with human cells.^34,35,90 In preliminary experiments, however, transmigration rates of monocytes are not diminished, transmigration is still PECAM and CD99 dependent, and, surprisingly, 99% of the transmigration events are paracellular. Data show the remarkable plasticity of the endothelial junctions, because during TEM the electrical resistance does not decrease.

The Relationship of LBRC and VE-Cadherin Movements

How are the movements of the LBRC and VE-cadherin coordinated? During TEM, VE-cadherin is removed from the cell border at the site of transmigration, whereas the LBRC moves in.^67,68 VE-cadherin gap formation can be prevented by mutations in the cytoplasmic tail of VE-cadherin that prevent it from interacting with β-catenin and p120.⁷² Just like targeted recycling of the LBRC, VE-cadherin gap formation is required for TEM, and manipulations that inhibit gap formation also reduce TEM. Does VE-cadherin leave first and then the LBRC moves in to fill this vacuum? Does the LBRC move in and push VE-cadherin out of the way? Are these two movements of membrane somehow coordinated or cooperative? Or are they totally independent? Dr. Annette Gonzalez is approaching this problem using a series of experiments in which she selectively blocks either targeted recycling or VE-cadherin gap formation and determines the effect on the other membrane compartment. The results of these experiments should unambiguously determine the relationship of these events, both of which are essential for TEM (Figure 4).

Other Challenges

We have known for more than 20 years that a transient increase in cytosolic free calcium ion concentration⁶⁴ is required for TEM. However, we still do not know which step or steps in transmigration are regulated by the calcium flux. Evan Weber, a student in my research group, is working on this problem using an in vitro TEM assay that can dissect the contributions of adhesion, locomotion, diapedesis, and migration across the basement membrane⁹¹ to the leukocyte extravasation process. The role of CD99 in TEM was first demonstrated more than a decade ago, but CD99 is a unique molecule with no homologs in the genome (other than the somewhat distantly related paralog CD99L2⁹²) and no known signaling motifs on its cytoplasmic tail.²² Understanding its mechanism of action has therefore been a challenge. Richard Watson, another student in my research group, has been studying signaling events that can overcome a CD99 blockade to transmigration and is tracing these pathways back to CD99.

Several other unanswered questions remain. How is it that the molecules that regulate TEM are all present together in the LBRC, yet PECAM, CD155, and CD99 act sequentially? Why are there so many molecules on leukocytes and endothelial cells that, when blocked with antibodies or knocked down or out, have such a large effect on TEM? If interfering with the function of any one of them blocks TEM by more than 50%, then the others should act in a common step (otherwise, it would easily be possible to block 100% of TEM, which is never seen). What is that common step? Why don’t the other molecules compensate for the one molecule that has been deleted or inhibited in these studies? When we fully understand the mechanisms of TEM, we will have the answers to such questions and will, we hope, have a much better handle on controlling the inflammatory response.