The stem cell revolution: on the role of CD164 as a human stem cell marker

Accurately defining hierarchical relationships between human stem cells and their progeny, and using this knowledge for new cellular therapies, will undoubtedly lead to further successful treatments for life threatening and chronic diseases, which represent substantial burdens on patient quality of life and to healthcare systems globally. Clinical translation relies in part on appropriate biomarker, in vitro manipulation and transplantation strategies. CD164 has recently been cited as an important biomarker for enriching both human haematopoietic and skeletal stem cells, yet a thorough description of extant human CD164 monoclonal antibody (Mab) characteristics, which are critical for identifying and purifying these stem cells, was not discussed in these articles. Here, we highlight earlier but crucial research describing these relevant characteristics, including the differing human CD164 Mab avidities and their binding sites on the human CD164 sialomucin, which importantly may affect subsequent stem cell function and fate.


DISCOVERY OF HUMAN CD164 AS A HAEMATOPOIETIC AND SKELETAL STEM CELL BIOMARKER
Just over two decades ago, human (h) CD164 was identified as a functional biomarker on human haematopoietic precursors and their associated bone marrow stromal cells [1][2][3] . Importantly, a 1998 special focus report in 'Blood' 3,4 highlighted the findings that hCD164 was a sialomucin involved in human haematopoietic progenitor-stromal interactions, and that engagement of specific glycosylated hCD164 epitopes on quiescent human CD34+ haematopoietic stem/progenitor cells (HSPCs) could prevent their recruitment into the cell cycle 3,5 . Two decades on, by analysing human bone marrow Lineage (Lin) neg , CD34 + , CD34 low , and CD34 neg HSPCs with single cell RNAseq and improved surrogate transplantation models, Pellin and colleagues, in their article 'A comprehensive single cell transcriptional landscape of human hematopoietic progenitors' identified hCD164 as an important, reliable biomarker for defining the earliest branch points of hHSPC specification, which incorporates the basophil lineage and shows close similarities to murine lineage specification 6 . Interestingly, just prior to this observation, Chan and colleagues, in their paper entitled 'Identification of the human skeletal stem cell' meticulously purified, from hypertrophic zones of the growth plate in long bones, a subset of human PDPN + CD73 + CD164 + skeletal stem cells (hSSC), which were negative for CD45, CD146, CD235, Tie2 and CD31, and at the apex of a hierarchy of stem cells that could transition into an early bone-cartilage-stromal progenitor 7,8 . These cells were demonstrated to meet the rigorous definition of hSSCs 9,10 , as being locally restricted to the bone, with the ability to clonally self-renew, differentiate into multiple lineages and reconstitute a haematopoietic microenvironment both in vitro and in vivo in surrogate serial transplantation models 7,8,11 . Importantly, and reminiscent of initial studies 3,4 , regulatory cross-talk was shown to exist between CD164 + hHSCs and these CD164 + hSSCs 7,8,11 .
While it is gratifying to see hCD164 identified as a potentially improved biomarker for hHSPC and hSSC isolation, we wish to highlight a critical issue not addressed in the three recent manuscripts cited above [3][4][5] , namely the choice of hCD164 antibody used for these purposes. As described below, at least three epitopes defined by three distinct Classes of hCD164 antibodies (Mabs) exist on hCD164 and representative of each class do not stain cells equivalently.
The full length hCD164(E1-6) isoform is predicted to contain 9 Nlinked and 32 O-linked glycans, and a glycosaminoglycan (GAG) attachment site at the end of E5 and beginning of E6 3,5,13-17 . E1 encodes the first mucin domain, E2 and E3 encode the cysteine rich non-mucin domain, E4 to part of E6 encode the second mucin domain, and the remainder of E6 the transmembrane region, cytoplasmic domain containing a YHTL endocytic motif and 3′UTR. A cytokine binding pocket (reminiscent of TpoR, IL-4R, IL-6R, EpoR, G-CSFR, and GM-CSFR) is located in its non-mucin domain (Fig. 1a). Figure 1b shows a diagrammatic representation of hCD164 and showing putative intra-molecular disulphide bridges between four cysteines in the non-mucin domain 3,5,[13][14][15][16][17] . A total of eight cysteines are present and the remaining four may also contribute to intramolecular or inter-molecular disulphide bridges. The hCD164 glycoprotein has a molecular mass of 80-100 kD under reducing conditions and, under non-reducing conditions, can exist as a homodimer of 160-180 kD. A GAG attachment at the E5-E6 junction (TSGT), association with other components (e.g., the cytoskeleton, growth factors) or the existence of an hCD164 tetrameric form may increase the molecular weight to >220 kD 3,5,13-17 .  17 .

HCD164 MONOCLONAL ANTIBODIES AND EPITOPES
Further, we demonstrated that hCD164 resembled a subgroup of sialomucins, which include the hHSPC biomarker CD34. While this heavily glycosylated protein subset, like other sialomucins, is rich in serine and threonine residues, it is encoded by multiple exons 2,3,5,[13][14][15][16][17]32 . Earlier studies had also demonstrated that CD34 Mabs could be grouped into three Classes based on the resistance or susceptibility of their cognate epitopes to cleavage by the enzymes neuraminidase, chymopapain, and a glycoprotease from Pasteurella haemolytica [33][34][35] . For example, the CD34 epitope recognised by the My10 Class I CD34 Mab was sensitive to both C. perfringens sialidase and O-sialoglycoprotease treatments; the CD34 epitope identified by the QBEND 10 Class II Mab was sensitive to O-sialoglycoprotease, but not to C. perfringens sialidase; the CD34 epitope defined by the Tük3 Class III Mab was insensitive to both C. perfringens sialidase and O-sialoglycoprotease enzymes [33][34][35] .
Given this, we examined the glycosidase sensitivities of the cell surface hCD164 sialomucin and subsequently of recombinant soluble chimaeric domain-truncation hCD164-Fc/hCD164-His-tagged constructs 13,14,16,17 . We also analysed the global glycosylation patterns (lectin binding, HPLC, and mass spectrometry) of chimaeric recombinant domain-truncation constructs 17 . These latter studies revealed that hCD164 contains sialic acid moieties in both α2-3 and α2-6 linkages on O-glycans and N-glycans, but with O-glycans having a much higher degree of sialylation. Sialyl LewisX (sLeX) was not detected 17 . The N-glycans are more complex than the O-glycans, with high mannose, hybrid and mono-antennary, di-antennary, triantennary, and tetra-antennary complex forms (with or without bisecting GlcNAC). Most O-glycans are small core 1 or 2. Predominantly large neutral tri-antennary and tetra-antennary structures occur on N-glycans in mucin domain II, while smaller, biantennary structures are present in mucin domain I 17 .
Deglycosylation experiments followed by analyses with hCD164 Mabs were carried out on the hCD164(E1-3)-Fc soluble protein and hCD164 purified from KG1A cells using N-glycanase, O-glycosidase, sialidase, α-fucosidase, and O-sialoglycoprotease treatments, either separately or together (Fig. 2) [13][14][15][16][17] . The epitope recognised by the 105A5 Mab was sialic acid dependent, with sialic acid moieties situated on O-glycosylated chains attached to the exon 1-encoded peptide, but not on N-linked oligosaccharides. Oglycosidase and O-sialoglycoprotease significantly reduced 103B2/ 9E10 and 105A5 Mab binding to hCD164, but this was not seen with the other hCD164 Mabs. The hCD164 epitope recognised by the103B2/9E10 Mab was sensitive to N-glycanase treatment, and hence is dependent on the N-linked carbohydrates of exon. In contrast, the epitopes recognised by the remaining hCD164 MAbs were not removed by the deglycosylation procedures used. Thus, by comparing hCD164 Mabs with the CD34 Mab Classes, the hCD164 Mabs could be classified into three analogous categories [13][14][15][16][17] . Hence, the epitope recognised by the Class I 105A5 Mab is present ) also bind to a > 220 kD hCD164 protein species 13,14,16,17 . As indicated earlier, the latter may be an hCD164 tetramer, a hCD164 monomer or homodimer in association with other, possibly cytoskeletal or growth factor, components, or a GAG containing hCD164 species, an issue not yet resolved (Fig. 1) 6 . Importantly, hCD164 is also a biomarker for hMSCs/hSSCs, including CD271 hi CD45 lo/neg and CD56 + or CD56 neg primary bone marrow hMSC 7,8,25,26 . Notably, the Class IIIB Mab (67D2) has been used to identify and isolate PDPN + CD146 neg CD73 + CD164 + selfrenewing, serially transplantable hSSCs with multipotent (osteogenic, chondrogenic, and haematopoietic supporting stromal) potential from the human bone growth plate and diaphyseal zones 7 . Significantly, these hSSCs exist at the apex of the human skeletogenic differentiation hierarchy, are present in human foetal and adult bones, can be generated from BMP-2 stimulated human adipose MSCs and iPSCs, expand locally following acute skeletal injury and maintain human haematopoiesis 7,8,11 . While all the hCD164 epitopes described to date are present on hCD34 + and hCD133 + hHSPC subsets and on hBM MSCs, this is in direct contrast to their differential expression in other postnatal haematopoietic and non-haematopoietic tissues 2,3,5,13-17,23 . We observed that the epitopes detected by the Class I and II Mabs can be differentially expressed and distributed on reciprocal cell subsets in some tissues, while the Class III Mabs stained cells that bound both or either of the Class I and/or II hCD164 Mabs 23 . For example, the epitope recognised by the 103B2/9E10 Class II Mab, but not that by the 105A5 Class I Mab, is present on most vascular endothelium, on some high endothelial venules in lymphoid tissues, on the venous sinuses in spleen, on the thymic subcapsular epithelia which is associated with pre-T lymphoid cell trafficking and on the basal-layer epithelia of the tonsil and skin. In contrast, the epitope recognised by the 105A5 Class I Mab is present on lymphoid cells in tissues that are involved in lymphoid/blood cell recirculation, such as in the tonsil, spleen, and gut 23 . However, the CD1 + and cortical and medullary CD43 + thymic lymphoid cells stain weakly with the Class I, but not at all with the Class II, hCD164 Mabs, while thymic macrophages express both the hCD164 epitopes defined by the Class I and II Mabs 23 .

THE ROLE OF HCD164 IN REGULATING CELL FATE
The epitopes defined by the different Classes of hCD164 Mabs are implicated in possessing functional roles in haematopoiesis as illustrated by in vitro studies or based on hCD164 structural studies. Engaging the hCD164 molecule with specific hCD164 Mabs can modulate CD34 + or CD133+hHSPC proliferation, differentiation, adhesion to, and retention or migration in different microenvironmental niches. As a first exemplar, the hCD164 103B2/9E10 Class II Mab-defined epitope partially blocks adhesion of CD34 + and CD133 + hHSPCs to bone marrow MSCs 3,5,13,14,31 . Ligating this epitope with the 103B2/9E10 Class II Mab also inhibits cytokine-driven (IL-3, IL-6, SCF, and G-CSF) recruitment of single CD34 + CD38 lo/neg hHSPCs into the cell cycle in vitro 3,5,13,14 . Such effects are indicative of cognate ligands for hCD164 acting in cis or trans, albeit not yet fully defined. Our findings show that hCD164 lacks sLeX moieties required for P-selectin or E-selectin binding, while its sialylation in the absence of fucosylation suggests that hCD164 is a potential Siglec ligand, with preliminary studies indicating hSiglec-5 (but not hSiglec-3, hSiglec-7, hSiglec-8, hSiglec-9, or hSiglec-10) binding [13][14][15]17 . The second example demonstrates that culturing hBM CD34 + cells with IL-1β, IL-3, IL-6, G-CSF, GM-CSF, and SCF for 21 days in the presence of the Class I 105A5 or Class II 103B2/9E10 hCD164 Mabs reduces the absolute number of both total nucleated cells generated, and, when erythropoietin is added as well, decreases the development of clonogenic BFU-E and CFU-GM [13][14][15] . Furthermore, hCD164 is a key component of the CXCR4-VLA-4-VLA-5 complex, regulating CD133 + hHSPC CXCL12-mediated migration or retention in the hHSC niche and promoting CXCR4-mediated signalling, a process inhibited by the 103B2/9E10 Class II hCD164 Mab 31 . Although functional effects of engaging the hCD164 epitopes with Class III Mabs have not been fully assessed, the full length hCD164 contains a potential cytokine binding pocket located in its nonmucin domain, as well as a potential GAG attachment site at the E5-E6 junction of hCD164 2,3,5,13-17 , that could act in concert to modulate cytokine-mediated, chemokine-mediated, or adhesionmediated effects on the same or opposing cells. While this is purely hypothelical, it is important to keep in mind that these and other Mabs chosen to isolate hHSCs might inadvertently affect the function or fate of the hHSCs when assessed in vitro or in vivo.

BIOMARKERS AND LINEAGE HIERARCHIES
Biomarkers that identify definitive, but rare, repopulating human bone marrow stem cells have been the "holy grail" of stem cell research since the seminal discovery of monoclonal antibodies by Koehler and Milstein over 5 decades ago 36 . These, when coupled with the increasingly sophisticated technological developments [37][38][39][40][41][42][43][44] of cell sorting, surrogate in vivo models, gene editing, single cell barcoding, single cell deep sequencing, lineage tracing and fate mapping in vitro and in vivo, and human transplantation or regenerative medicine, have fuelled a series of debates on the hierarchical relationships of these human stem cells with their immediate progeny, and on selecting the best biomarkers and cell subsets for therapeutic use. To this end, biomarkers, lineage trees and lineage relationships have been constantly changing with progressive technological developments and research. Examples of these are indicated in the attached references 6,7, .
When defining lineage hierarchies and developing new cellular products with hCD164 as a biomarker, it is important to remember that, while Pellin et al. 6 and Chan et al. 7 used the 67D2 Class IIIB hCD164 antibody to identify hHSCs and hSSCs, different classes of hCD164 antibodies possess different avidities for this molecule and, by interfering with ligand binding sites, these may alter human hHSPC and hSSC function and hence fate. The information garnered from previous research should be carefully reviewed when considering the choice of hCD164 antibodies in future clinical studies. This said, Class II, and to a lesser extent, Class I hCD164 antibodies, with isotypes distinct from the Class III antibodies, may prove beneficial for confirming the purity of and hence establishing release criteria for the hCD164 Class IIIenriched hHSC and hSSC subsets for in vivo therapy.
In conclusion, using this knowledge in association with new cellular therapies will undoubtedly contribute to successful treatments for life threatening and chronic diseases, which are substantial burdens on both patient quality of life and healthcare systems globally 63,64 .