Learn more: PMC Disclaimer | PMC Copyright Notice
Engineering cellular trafficking via glycosyltransferase-programmed stereosubstitution
Abstract
The proximate hurdle for cell trafficking to any anatomic site is the initial attachment of circulating cells to target tissue endothelium with sufficient strength to overcome prevailing forces of blood flow. E-selectin, an endothelial molecule that is inducibly-expressed at all sites of inflammation, is a potent effector of this primary braking process. This molecule is a member of a family of C-type lectins, known as selectins, that bind sialofucosylated glycans displayed on either a protein (i.e., glycoprotein) or lipid (i.e., glycolipid) scaffold. On human cells, the predominant E-selectin ligand is a specialized glycoform of CD44 known as hematopoietic cell E-/L-selectin ligand (HCELL). This review focuses on the biology of HCELL/E-selectin interactions in cell migration, and discusses the utility and applicability of glycosyltransferase-programmed stereosubstitution (GPS) for glycoengineering HCELL expression. Without compromising cell viability or native phenotype, this exoglycosylation technology literally “sweetens” CD44, licensing E–selectin dependent vascular delivery for all cell-based therapeutics.
Introduction
Degenerative and inflammatory diseases such as osteoporosis, chronic obstructive pulmonary disease, arthritis, atherosclerosis, inflammatory bowel disease, and multiple sclerosis affect billions of people worldwide. The clinical manifestations of these chronic diseases, and also of acute inflammatory conditions such as myocardial infarction and stroke, reflect a balance between tissue destruction and repair. Most current therapies for such diseases target the inflammatory component without acting on the repair processes. Stem cell-based therapeutics offers the real opportunity to achieve tissue regeneration, either by directly contributing to formation of new cells and/or by environmental trophic effects leading to recruitment/support of other cells necessary for tissue regrowth. Depending on stem cell type (e.g., mesenchymal stem cells), administered stem cells may concomitantly blunt inflammation via potent immunomodulatory effects.1 Although the exact mechanism(s) by which various stem cells may exert tissue repair are unsettled at present and may differ depending on both the stem cell type and the pathologic entity, the clinical realization of this potentially curative therapeutic approach depends, at the outset, on getting the pertinent cells to the sites where they are needed. Moreover, the capability to achieve patient- and disease-specific treatments through adoptive cell therapeutics employing immune effector cells (e.g., for cancer immunotherapy) or regulatory cells (e.g., for treatment of autoimmune diseases) similarly depends on delivery of appropriate cells to sites of disease. Thus, a fundamental precondition for successful implementation of all cell-based therapies is to achieve adequate tissue colonization at requisite anatomic locations.
Delivery of cells for clinical indications can be achieved by direct (local) injection into involved tissue(s), or by intravascular (i.e., systemic) administration. At first glance, direct delivery might appear to be the most efficient approach, especially considering that a concentrated bolus of cells could be applied to an affected area. However, local injection may actually be counterproductive to intended tissue repair and is also limited in scope: (1) By introducing pertinent cells in media suspension under hydrostatic pressure, the injection procedure could harm the delivered cells and, furthermore, could further compromise tissue integrity and disrupt incipient repair processes, thereby exacerbating the inflammatory condition in situ. (2) By virtue of being an invasive method, the injection needle/device (and the suspension solution) could induce target tissue damage and/or instigate collateral tissue damage; (3) Direct injection is feasible only for organs/tissues with well-defined anatomic boundaries (e.g., the heart, but not the lung); (4) The injection procedure could be technologically demanding and labor-intensive, requiring use of sophisticated delivery systems with substantial imaging support, especially for relatively inaccessible and/or fragile organs/tissues (e.g., the central nervous system); (5) Most importantly, many degenerative and inflammatory conditions are widely distributed and multifocal in nature (e.g., osteoporosis, inflammatory bowel disease, multiple sclerosis, etc), and thus direct injection is neither practical nor effective. The vascular route of administration is mandated for these and all generalized “systemic” disorders, as well as for any tissues with problematic access and/or anatomy not amenable to local injection (e.g., the pancreas in diabetes, the lung in chronic obstructive pulmonary disease). The capacity to administer cells repeatedly with minimal effort is another important practical advantage of systemic infusion. Therefore, creation of methodologies to optimize the expression/activity of molecular effectors directing the physiologic migration of intravascularly administered cells is key to achieving the tremendous promise of all cell-based therapeutics.
The molecular basis of cell migration
Recruitment of circulating cells to any anatomic site involves a coordinated sequence of events, conventionally described as comprising four overlapping steps. Under both steady-state and pathologic conditions, extravasation typically occurs in post-capillary venules under hemodynamic shear conditions of 1–4 dynes/cm2.2 This process is initiated by decelerative adhesive interactions between cells in flow and vascular endothelium at the target tissue(s), characteristically involving tethering of blood-borne cells on the endothelial surface, followed by sustained cellular rolling at velocities below that of blood flow (step 1). The molecules that mediate these adhesive contacts are called “homing receptors”, because, as defined historically, these structures were thought to direct tropism of circulating cells to the respective target tissue(s). From a biophysical perspective, a homing receptor functions as a molecular brake, displaying fast on-off binding kinetics with its pertinent endothelial counterreceptor(s), which is translated into cellular torque under the action of fluid shear forces.3 Thereafter, a cascade of events ensues, characteristically precipitated by chemokines engaging their cognate ligand(s) on the surface of the blood-borne cells, resulting in G-protein –coupled activation of integrin adhesiveness (step 2). The consequent integrin attachment to endothelial coreceptors results in firm adherence (step 3), followed by endothelial transmigration (step 4).4 This “multi-step paradigm” holds that cell migration to any tissue is regulated by a discrete combination of homing receptor and chemokine receptor expression on a given circulating cell, allowing for recognition of a pertinent “traffic signal” displayed by the relevant vascular adhesive ligands and chemokines expressed within target endothelium. Though step 1 interactions are reversible, homing receptor expression is obligatory to achieve tissue infiltration, as tethering and rolling adhesive interactions are prerequisite for elaboration of all downstream events, including extravasation. Thus, maintaining and/or enforcing expression of homing receptors is required to achieve cell delivery via the vasculature to any predetermined anatomic site.
Selectins and their ligands: the sweet homing receptors
Several molecules (including CD44 and members of the integrin family (e.g., α4β7)) are effectors of step 1 tethering and rolling interactions, but the selectins and their ligands are the most potent mediators of these adhesive events (reviewed in 4). This family of C-type (i.e., Ca++-dependent) lectins is comprised of three glycoproteins: L-selectin expressed on mature leukocytes and hematopoietic stem cells, P-selectin expressed platelets and endothelium, and E-selectin, which is expressed only on endothelium. By definition, “homing receptors” are displayed on circulating cells. Thus, L-selectin is a homing receptor, whereas the ligands for P- and E-selectin serve as homing receptors. L-selectin expression is regulated by cell surface proteolytic cleavage (shedding) mediated by membrane-associated metalloproteases, one of which is ADAM17.5
Though shedding of P- and E-selectin has been described, the predominant mechanism regulating expression of these molecules is transcriptional induction. Notably, E- and P-selectin are constitutively expressed in two endothelial beds: the microvasculature of skin and marrow.6, 7 Also, P-selectin (but not E-selectin) is located in alpha granules of platelets and in Weibel-Palade bodies of endothelial cells, and its surface expression can be induced via granular translocation in response to inflammatory agents such as histamine and thrombin. However, more typically, de novo synthesis and surface expression of these selectins is induced by inflammatory cytokines in all microvessels (including skin and marrow). Though P-selectin and E-selectin are prominently expressed at sites of tissue injury and inflammation, it is important to draw distinction between the cytokine induction of expression of these selectins among rodents and primates. Importantly, although IL-1 and TNF each induce transcription of mRNA encoding P-selectin and E-selectin in rodents, the primate P-selectin promoter lacks the relevant response elements for these cytokines and only E-selectin is transcriptionally induced.8 In fact, experiments in transgenic mice bearing the human P-selectin gene on a murine P-selectin knock-out background have shown that TNF administration actually decreases human P-selectin expression (i.e., decreased P-selectin mRNA levels are observed), whereas it increases murine P-selectin expression in wild type animals9; a similar pattern of decreased human P-selectin mRNA was observed in skin of the transgenic mice undergoing contact hypersensitivity reactions. This key distinction in regulation of P-selectin expression has profound implications, indicating that results obtained from both steady-state and inflammatory models in mice may over-emphasize the contribution(s) of P-selectin, and under-emphasize the contribution(s) of E-selectin, compared to clinical reality. Thus, in humans, optimizing the expression/activity of E-selectin ligands, not P-selectin ligands, would yield the most efficient trafficking of cells to inflammatory sites.
All three selectins display Ca++-dependent binding (i.e., are C-type lectins) to sialofucosylated glycans, the prototype of which is the tetrasaccharide known as sialyl Lewis X (sLex) (see Fig. 1). This structure can be displayed on glycoproteins or glycolipids; on native cell membranes, glycoprotein E-selectin ligands generally possess greater binding activity than do glycolipids under hemodynamic flow conditions (i.e., mediate slower rolling and higher resistance to detachment under shear). Both L- and P-selectin readily bind to sulfated forms of sLex, and binding of both L- and P-selectin to the glycoprotein known as P-selectin glycoprotein ligand-1 (PSGL-1) requires display of a sulfated tyrosine adjacent to an O-linked sLex structure located within the N-terminus of the protein scaffold. However, E-selectin adherence to any of its ligands is not dependent on sulfation, and, importantly, E-selectin preferentially binds to unsulfated sLex.2, 10 Under shear conditions, binding interactions mediated by E-selectin display greater resistance to detachment and, also, slower rolling velocities than those of P-selectin and L-selectin;11–14 these observations suggest that E-selectin receptor/ligand interactions provide optimal endothelial contact time for elaboration of steps 2 and 3 events (chemokine recognition and integrin activation, respectively).
Structure of sialylated Lewis X (sLex). Schematic depiction of the canonical selectin binding determinant, sLex (see box), showing linkages of relevant monosaccharide units. Color key figures correspond to respective monosaccharides. The disaccharide unit consisting of galactose and N-acetylglucosamine is known as a “lactosamine” unit. In sLex, the lactosamine is capped with a sialic acid in a(2,3)-linkage to galactose; this structure is known as a “sialyllactosamine” unit. By definition, theβ(1,4)-linkage between galactose and N-acetylglucosamine classifies the lactosamine unit as “type 2”. The sLex structure is a terminal type 2 sialyllactosamine unit containing fucose in α(1,3)-linkage to N-acetylglucosamine.
Criteria for identification of E-selecin ligands
The identity and function of E-selectin ligands can be elucidated using both in vivo and in vitro approaches. Regarding the former, various murine models have been employed to assess the capacity of leukocyte E-selectin ligands to engage vascular E-selectin. While these studies can offer insights under physiologic settings, rolling adhesive events attributable to E-selectin receptor/ligand interactions vary depending on numerous factors, including the animal strain employed, the inflammatory model and/or stimulus, vessel type, vessel diameter/dimension(s), and blood flow velocity/fluid shear stress. It is important to recognize that many of these “physiologic” models assess properties of non-physiologic cells, i.e., products of genetically modified mice with either altered expression of glycosyltransferases directing synthesis of glycan determinants of E-selectin ligands or deletion of one or more scaffold proteins (e.g., PSGL-1) that may display the relevant glycan determinants. Compensatory changes frequently accompany genetic alteration(s), which could confound the identification of E-selectin ligands by unveiling and/or inducing aberrant post-translational modifications and/or altered cell surface distribution of pertinent molecules. Thus, the data derived from these genetically altered mice do not accurately reflect the identity and function of native E-selectin ligands under authentic physiologic conditions, and, in general, these models of manipulated E-selectin ligands say more about the cellular context(s) than about the biology of cell migration. To avoid shortfalls inherent in studies of genetic mutants, one could utilize function-blocking mAb to dissect relevant contributions of (reputed) E-selectin ligands. However, to date, no such antibodies to any E-selectin ligand have been described. An alternative approach would be to utilize RNAi to dampen expression of relevant targets. However, gene silencing also has the potential to induce compensatory changes and, additionally, could alter cell biology in ways that could indirectly impair cellular engagement to E-selectin.
A variety of in vitro approaches are available to assess the ability of membrane molecules to serve as E-selectin ligands. The availability of E-selectin-immunoglobulin chimeric constructs as a probe greatly facilitates the qualitative and quantitative evaluation of E-selectin ligand activity on intact cells (by flow cytometry) or on resolved cell glycoproteins or glycolipids (by western blots and/or eastern blots, respectively).15 However, since all selectins natively engage their cognate ligands under hemodynamic shear conditions, assays performed under non-static conditions (e.g., parallel plate flow chamber assays under fluid shear conditions3 or Stamper-Woodruff assays16, 17 are more reliable indicators of physiologic function than those under static conditions. In this regard, we have developed a novel assay system called the “blot rolling assay” which allows for identification of selectin ligand activity, under appropriate hydrodynamic fluid shear conditions, of glycoproteins resolved by western blot.18–20 Though biochemical studies of isolated molecules are informative, membrane topography is a critical factor in promoting selectin receptor/ligand interactions,21, 22 and, therefore, isolation of a putative selectin ligand from the native membrane and clustering of that molecule on a support surface could either dampen or augment pertinent binding activity displayed on relevant cells. Therefore, in general, studies of ligand activity on intact cells, preferably engaging E-selectin expressed on a physiologic cell type (i.e., endothelial cells), are preferred over studies of isolated ligands displayed as non-physiologic clusters on artificial support surfaces.
Given these considerations, the identity of a putative E-selectin ligand ––or of any selectin ligand, for that matter ––requires a combination of biochemical and operational criteria. First, the molecule must display the relevant glycan motif(s) that serve as binding determinants (e.g., sLex). Secondly, the molecule must be displayed on a native (primary) cell of interest. Third, the molecule must be capable of engaging E-selectin, in particular, under relevant physiologic fluid shear conditions. Fourth, and most importantly, expression of that ligand on a pertinent cell, ideally in absence of any other ligand(s), should endow that cell with the capacity to engage E-selectin as displayed on endothelial cells; alternatively, in settings where a cell may express additional E-selectin ligands, the contribution of the examined ligand to overall E-selectin binding should be readily apparent.
E-selectin ligands of human and murine hematopoietic stem/progenitors: gatekeepers of marrow migration
The recovery of hematopoiesis following hematopoietic stem cell transplantation depends, at the outset, on the ability of infused hematopoietic stem/progenitor cells (HSPCs) to home to bone marrow. As noted above, human and murine marrow microvascular endothelial cells constitutively express E-selectin,6, 23 and numerous studies over the past two decades have shown that HSPC migration to marrow involves expression of E-selectin ligands, in combination with the chemokine receptor CXCR4 and the β1 integrin VLA-4 (for review, see 24). Intravital microscopy studies in mice have revealed that E-selectin and the chemokine that binds to CXCR4, CXCL12 (otherwise known as SDF-1), are co-expressed in a restricted fashion, expressly at those microvessels that recruit HSPCs.23 In addition, marrow microvasculature endothelial cells constitutively express VCAM-1,6 a ligand for VLA-4, but in a broader distribution than that of E-selectin.23 Altogether, these studies have established that homing to marrow involves a multistep cascade, whereby E-selectin receptor/ligand interactions mediate Step 1 tethering and rolling, followed by CXCL12 binding to CXCR4 yielding activation of VLA-4 (step 2), resulting in VLA-4 firm adhesion to VCAM-1 (step 3), and subsequent extravasation (step 4).
It is well-known that human HSPCs express robust E-selectin ligand activity, and that the highest E-selectin ligand activity is found on primitive HSPCs (i.e., CD34+ lineage – cells).25, 26 To elucidate the molecular effectors of E-selectin binding, we performed a comprehensive assessment of all glycoprotein E-selectin ligands expressed on human and mouse HSPCs.15 These studies showed that native human HSPCs express three glycoprotein E-selectin ligands: (1) hematopoietic cell E-/L-selectin ligand (HCELL), an sLex-decorated glycoform of CD44 that binds E-selectin and L-selectin; (2) CLA (cutaneous lymphocyte antigen), a glycoform of PSGL-1 bearing abundant sLex determinants displayed on Ser/Thr-linked glycans (i.e., O-glycans) that binds E-selectin (in addition to L- and P-selectin); and (3) an sLex-bearing glycoform of CD43 that binds E-selectin (called here “CD43E”). Consistent with results of previous studies on human hematopoietic cell lines,19, 27 studies of native human HSPCs showed that HCELL is the most potent (i.e., confers the slowest rolling velocity and the highest resistance to detachment under fluid shear stress) of all the E-selectin ligands expressed on human cells, and that abrogation of HCELL expression by siRNA targeting of CD44 mRNA in hematopoietic cells resulted in a profoundly reduced ability to engage E-selectin.15 Mouse HSPCs express only CLA and CD43E, and, consistent with absence of HCELL expression, mouse HSPCs have markedly lower E-selectin ligand activity compared to human HSPCs.15
GPS to enforce HCELL expression: glycoengineering cell migration
The CD44 molecule is best known for serving as the principal receptor for hyaluronic acid (HA) (for review, see 28). Though CD44 is found on the surface of almost all human cells, expression of the HCELL glycoform in healthy individuals is restricted to HSPCs. This tight regulation of expression, coupled with this glycoprotein’s high E-selectin binding activity, raised the possibility that HCELL serves an important role in directing osteotropism. To examine this issue, we developed a platform technology to custom-modify glycans of cell surface CD44 to create the HCELL glycoform. This effort involved specific formulation of glycosyltransferases and reaction buffer conditions to avoid effects on cell viability and phenotype (with exception of intended effects on CD44 structural biology) (for details, see 2, 29). Notably, cellular HCELL expression resulting from external glycosylation (i.e., “exoglycosylation”) of membrane CD44 is transient, with surface turn-over to the native CD44 molecule complete within 48 hours30. Accordingly, following infiltration within target tissue(s), there is rapid reversion of native CD44 expression on extravasated cells.
The role of HCELL as a “bone marrow homing receptor” has been directly examined using two complementary approaches employing GPS. In the first case, we utilized a target cell devoid of Step 1 effectors, human mesenchymal stem cells (MSCs). The human MSCs used in our studies expressed CD44 and VLA-4, but did not express CXCR430; biochemical studies revealed that the cells expressed a CD44 glycoform decorated with N-linked terminal type 2 sialyllactosamines, indicating that the glycoprotein was missing only α(1,3)-fucosylation at the terminal N-acetylglucosamine to complete the sLex determinant (see Fig. 1).30 Accordingly, we treated the MSCs in vitro with the α(1,3)-fucosyltransferase known as fucosyltransferase VI (FTVI) together with the nucleotide sugar donor, GDP-fucose. Follow exofucosylation, western blot showed that the only glycoprotein displaying sLex and reactive with E-selectin was CD44 (i.e., by definition, HCELL), indicating that CD44 was the predominant target of glycoengineering (see Fig. 2). When injected into immunocompromised mice, HCELL+ MSC migrated robustly to marrow within 1 hour of injection, whereas native (HCELL−) MSC showed minimal osteotropism. Most importantly, HCELL+ MSC infiltrated the marrow parenchyma, lodged within endosteal surfaces, and created human osteoid within mouse bone. Thus, enforced HCELL expression piloted colonization of human MSC within marrow, with preservation of viability and phenotype, yielding human osteoblasts that contributed to bone formation. In a complementary approach, we isolated murine HSPCs (lineage−Sca-1+ cKit+ cells; “LSK” cells) and enforced HCELL expression by GPS, again, via exofucosylation with FTVI. Notably, exofucosylation of mouse LSK cells only yielded expression of HCELL (i.e., not other E-selectin ligands), and, just as in native human HSPCs (and in MSCs expressing HCELL following exofucosylation), the relevant E-selectin binding determinants on the CD44 scaffold were expressed on N-linked glycans.15 In short-term homing studies, enforced HCELL expression conferred ~3.5-fold increased osteotropism of LSK cells. Collectively, these findings establish that expression of HCELL itself confers high efficiency cell trafficking to marrow, supporting the designation of this structure as the human “bone marrow homing receptor”.
Application of glycosyltransferase-programmed stereosubstitution (GPS) to enforce HCELL expression. Cell surface CD44 can be converted to the HCELL glycoform by glycan engineering via GPS. In both MSC and HSPC, native CD44 displays type 2 sialyllactactosaminyl glycans; color key figures correspond to respective monosaccharides (for details, see Fig. 1). Ex vivo treatment of cells with fucosyltransferase VI (FTVI) drives α(1,3)-fucosylation of CD44 glycans, thereby generating sLex and, accordingly, HCELL.
The finding that human MSC expressing HCELL were capable of colonizing marrow in absence of CXCR4 expression prompted us to investigate the molecular basis of HCELL-mediated transendothelial migration. Studies under both hydrodynamic flow and static conditions showed that HCELL engagement of endothelial E-selectin, or of CD44 engagement with HA, in each case triggered VLA-4 adhesion to its ligands VCAM-1 or fibronectin in absence of chemokine input.31 This cross-talk between HCELL/CD44 and VLA-4 was mediated by signaling through engagement of a CD44-dependent G-protein coupled Rac1/Rap1-signalling cascade, resulting in transendothelial migration (without chemokine effects) on human endothelial cells expressing both E-selectin and VCAM-1.31 These results thus refine the conventional multistep paradigm, defining the components of a novel mechanosignalling cascade that couples CD44 ligation with VLA-4 activation, yielding a “step 2-bypass” pathway of integrin activation, with commensurate firm adherence and transendothelial migration, without dependence on chemokines and other chemotactic agents. Though enhanced trafficking of HCELL+ cells to endothelial beds expressing E-selectin has been formally demonstrated thus far only for marrow (i.e., yielding robust osteotropism of both HCELL+ MSC and HCELL+ HSPC), it is likely that enforced HCELL expression would steer cell migration to other endothelial sites that express E-selectin. In all inflammatory conditions, involved tissue microvascular endothelial beds upregulate expression of both E-selectin and VCAM-1, under induction of cytokines such as TNF-αand IL-1. Since essentially all stem cells and leukocytes express CD44 as well as VLA-4, this novel regulatory pathway of cell migration has important implications for systemic cell delivery to sites of tissue injury/inflammation, for all indications of adoptive cellular therapeutics.
Regardless of whether cellular trafficking occurs through the step 2-bypass pathway or the conventional multistep cascade, glycoengineering to create HCELL will program highly efficient step 1 interactions on E-selectin, a key prerequisite for any blood-borne cell to migrate to any endothelial bed where this molecule is expressed. Notably, in all studies to date, α(1,3)-fucosylation has been the only carbohydrate substitution necessary to create the HCELL glycoform on cells that express CD44, and it is rather remarkable that a single monosaccharide substitution of this target glycoprotein can have such profound effects on cell migration. Future studies will explore how “sweetening” CD44 into HCELL via GPS may enable not only regenerative medicine, but also cell-based immunotherapeutics employing effector lymphocytes (e.g., for cancer or infectious disease) or regulatory lymphocytes (e.g., for autoimmune conditions).
Acknowledgments
This review summarizes many years of effort to elucidate the structure and biology of HCELL, and I wish to thank all my talented and devoted co-workers for their invaluable assistance in this formidable endeavor. This work was supported by the National Institutes of Health, in particular, the National Heart Lung Blood Institute (PO1 HL107146, RO1 HL60528, and RO1 HL73714) and the National Cancer Institute (RO1 CA121335). According to National Institutes of Health policies and procedures, the Brigham & Women’s Hospital has assigned intellectual property rights regarding HCELL and GPS to the inventor (RS), who may benefit financially if the technology is licensed.
Footnotes
RS’s ownership interests were reviewed and are managed by the Brigham & Women’s Hospital and Partners HealthCare in accordance with their conflict of interest policy.