CD44 serves as a cell surface receptor for various extracellular matrix molecules, mainly hyaluronan, and messenger molecules, such as growth factors, and has important functions in normal and disease states, the predominant one being cancer. CD44 coordinates both structural and signaling events through its highly conserved intracellular domain. Although short and devoid of any enzymatic activity, the CD44 intracellular domain possesses structural motifs that promote the interactions with cytoplasmic effectors involved in important cellular pathways, including cell trafficking, transcription, and metabolism, which regulate cellular functions like growth, survival, differentiation, stemness, and therapeutic resistance.
1. Introduction
The main function of cell surface transmembrane receptors is to sense extracellular signals and transduce them into an intracellular response. Most often, they bind to their specific ligand(s), leading to their conformational change and activation, which in turn alters their association with the cytoskeletal network and/or downstream signal transduction pathways to enable cells to respond to changes in their microenvironment. As such, CD44 is a single-chain transmembrane glycoprotein/part-time proteoglycan that belongs to the class of cell adhesion molecules (CAMs). The human gene encoding CD44 contains 19 exons and is located at chromosome 11p13. Exons 1 to 16 give rise to the extracellular domain (ectodomain) of CD44: the N-terminal signal sequence (exon 1), the link-homology hyaluronan-binding module (exons 2 and 3), and the stem region (exons 4–16). Exon 17 encodes the hydrophobic single-pass transmembrane domain, and exon 19 encodes the 73-amino-acid intracellular domain (ICD) or cytoplasmic tail of CD44
[1]. The inclusion of exon 18, normally absent in most CD44 transcripts, results in a short five-amino-acid cytoplasmic tail, which is generated by the use of an alternative translation stop codon
[2]. Exons 1–5, 16, 17, and 19 are constant and common to all CD44 isoforms. The standard form of CD44 (CD44s) is encoded by these constant exons, while CD44 variant forms (CD44v) arise from the combination of constant exons with variable exons 7–15 (identified as v2–v10) into the stem region as a result of alternative mRNA splicing, giving CD44 proteins the potential for vast structural and functional diversity. Further post-translational modifications such as glycosylation with O-glycans, N-glycans, and glycosaminoglycans in specific CD44 isoforms add to the complexity and diversity of CD44 proteins. Importantly, the link-homology hyaluronan-binding domain, the transmembrane domain, and the cytoplasmic tail of CD44 show high similarity between species
[3][4]. This remarkable sequence conservation indicates the important role of these domains in CD44 functions.
As a cell surface receptor, CD44 interacts with various extracellular matrix (ECM) molecules, such as hyaluronan (HA), osteopontin, collagen, fibronectin, and laminin, as well as growth factors and cytokines, MMPs, and serglycin proteoglycan
[5]. Many of these interactions are cell-type-dependent and contribute to the plethora of CD44 functions. For example, CD44 can exist in “inactive” or “active” states according to the binding affinity to its main ligand HA
[6]. In its “active” state, CD44 binds to HA, resulting in the promotion of cell migration, chemotaxis, rolling, and adhesion, as well as the organization of an HA-rich pericellular milieu and also the internalization and metabolism of HA. As a consequence, these CD44–HA interactions regulate several aspects of cellular behavior such as cell proliferation, growth, survival, differentiation, and pericellular matrix remodeling. However, the long list of functions of CD44 also include HA-independent activities. Importantly, CD44 is an established cancer stem cell (CSC) marker in several tumor types
[7]; for example, the CD44 and CD271 in human head and neck squamous cancer cells
[8] and the CD44 and CD133 in colorectal cancer cells
[9]. These observations imply a central functional role in tumor biology when considering that CD44 is able to promote epithelial-to-mesenchymal transition (EMT) and therefore invasiveness and metastasis
[10][11][12].
Although modifications of the extracellular domain of CD44 (such as O-/N-glycosylation, glycosaminoglycan substitution, and alternative splicing) regulate important functions of CD44, they cannot fully explain its critical involvement in such diverse cellular processes. Intracellular events also regulate CD44 actions, and this requires a functional intracellular domain. Given that CD44 lacks any intrinsic enzymatic (e.g., kinase) activity, it must be able to organize both membrane and cytosolic components and coordinate their functions. It is now evident that the short cytoplasmic tail of CD44 contains several structural motifs with the potential to selectively interact with cytoskeletal proteins and signaling effectors, often in cooperation with adjacent plasma membrane receptors. The assembly and coordination of such structural and signaling events can be regulated by post-translational modifications of the transmembrane and intracellular domains of CD44, such as palmitoylation, phosphorylation. and proteolytic processing.
2. Structural Features of CD44 ICD
The transmembrane and intracellular domains of CD44 are required for its proper membrane localization, ligand binding, and functions such as cell adhesion and migration [4]. For instance, mutation or deletion of the ICD results in the aberrant localization of CD44 within cellular membranes, an inability to bind HA, and impaired HA-mediated cell migration and tumor development [13][14][15][16]. Therefore, these highly conserved domains are essential for ligand binding and subsequent downstream intracellular events (i.e., outside-in signaling), probably by promoting the stabilization and clustering of CD44 receptors at the plasma membrane [4][13][17]. In contrast, their precise amino acid sequence is not a prerequisite for CD44 functions, since the replacement of either domain with equivalent domains from different adhesion receptors does not impair binding to HA, cell adhesion to HA matrices, or rolling interaction of lymphoid cells with HA pericellular coats [13][18], suggesting that the shape and conformation of these domains do not affect the inside-out signaling of CD44.
The 72-amino-acid-residue cytoplasmic tail contains two positively charged amino acid clusters in the juxtamembrane domain (
292RRRCGQ
KKK300), which constitute the FERM-binding domain that mediates the interaction of CD44 with ERM (ezrin/radixin/moesin) cytoskeletal proteins
[19][20]. This sequence also contains the putative acylation site Cys
295, suggesting that partition of CD44 into lipid rafts may regulate CD44 association with ERM proteins (see Section 4, “Regulation of cytoskeletal organization and cell phenotype”). The FERM-binding domain is followed by the ankyrin-binding domain (
304NSGNGAVEDRKPSGL
318)
[16], an additional cytoskeleton association site, the dihydrophobic basolateral targeting motif
331LV
332 [21], and four C-terminal amino acids (
358KIGV
361) that represent the PDZ (PSD-95/Dlg/ZO-1)-domain-binding peptide
[4] (
Figure 1).
Figure 1. Structural motifs of CD44 ICD and their interactions with cytoplasmic proteins. Ser residues (Ser291, Ser316, and Ser325) that are subject to phosphorylation are highlighted. Cytoskeletal proteins (blue), kinases (yellow), and proteins involved in cell-trafficking machinery, metabolism, and transcription are shown (for details, see text). Proteins with yet uncharacterized interactions with specific domains/sites of CD44 ICD are indicated in black boxes. The functional importance of CD44 ICD interactions in (patho)physiological processes are also shown.
3. Hyaluronan Internalization and Interactions with the Cell-Trafficking Machinery
The dihydrophobic motif
331LV
332 within the intracellular tail of CD44 targets the receptor to the basolateral membrane of polarized epithelial cells
[21], similarly to other proteins such as the major histocompatibility complex (MHC) class II invariant chain, the lysosomal integral membrane protein (LIMP)-II, and others
[22]. In contrast to these proteins, which undergo clathrin-dependent endocytosis mediated by the LV motif, CD44 has been shown to interact with the non-clathrin-dependent machinery whereby it internalizes its main ligand (HA), suggesting that the intracellular domain of CD44 allows for additional specific interactions that dictate the exclusion of CD44 from clathrin-coated pits. In chondrocytes, for example, CD44 mediates HA internalization through pathways independent from pinocytosis, caveolin, and clathrin
[23][24][25].
The critical role of CD44 ICD for HA endocytosis and turnover is evident, since tail-less CD44 mutants have a shortened half-life, reduced stability at the cell surface, and weakened binding to HA and are unable to internalize HA
[26]. There are several pathways suggested for CD44-mediated HA internalization. CD44–HA complexes either follow a pathway of trafficking to lysosomes for degradation or follow a pathway in which they are routed directly to the recycling endosomes for return to the plasma membrane. Ubiquitination of CD44 ICD by membrane-associated RING-CH (MARCH)-VIII ubiquitin ligase results in CD44–HA endocytosis and translocation of CD44 into EAA1-containing compartments and late endosomes/lysosomes for degradation
[27], while HA is gradually degraded by the coordinated actions of β-endoglycosidase HYAL-1, β-exoglycosidases, β-glucuronidase, and β-N-acetylglucosaminidase
[28].
Alternatively, CD44 can be recycled back to the plasma membrane by clathrin-independent sorting proteins such as Hook 1 and EHD proteins. Hook 1, a microtubule- and cargo-tethering protein, promotes microtubule-dependent sorting of CD44 away from the EAA1+ endosomes and into recycling tubules to end up in the plasma membrane
[29]. In addition, the proteomic approach identified Eps15 homology domain protein 2 (EHD2), a lipid-raft-residing protein that couples endocytosis to the actin cytoskeleton, as a potential CD44-interacting protein that binds in vitro to the last 15 amino acids of the CD44 C-terminal tail containing the PDZ-binding motif
358KIGV
361 [30], suggesting that CD44 harbors internalization signals in its cytoplasmic tail that are recognized by endocytic adaptors. This is in agreement with the observation that the acylation of CD44 at Cys
286/Cys
295 and its resulting localization into lipid rafts is crucial for HA uptake
[25][26][31].
4. Regulation of Cytoskeletal Organization and Cell Phenotype
In migrating cells, CD44 shows a preferential distribution in actin polymerization regions such as the lamellipodia, filopodia, and apical microvilli, suggesting a role of CD44 in the regulation of actin cytoskeleton reorganization. This regulation is not direct, since CD44 ICD does not contain any binding sites for actin filaments but instead structural motifs such as the FERM (4.1 protein, Ezrin, Radixin, Moesin)-binding domain and the ankyrin-binding domain, which allow CD44 to connect and interact with the cytoskeleton. Specific clusters of basic residues (
292RRRCGQ
KKK300) in the cytoplasmic tail of CD44 constitute the FERM-binding domain that interacts with ERM proteins and merlin/NF2 followed by a structural motif containing Ser
316 (
304NSGNGAVEDRKP
SGL
318) that binds the ankyrin cytoskeletal protein
[4] (
Figure 1).
The ERM proteins and merlin are binding partners for a number of transmembrane receptors (like ICAM, syndecans, L-selectin, and integrins) acting as cross-linkers between the plasma membrane and the cortical actin filaments
[32][33][34][35]. The neurofibromatosis type 2 (NF2) gene encodes the merlin protein, which has tumor-suppressing functions through regulating Hippo signaling, as well as receptor tyrosine kinases and downstream signal transduction pathways
[36][37][38]. ERM proteins and merlin show a similar domain organization, with the highest homology in the conserved three-lobe N-terminal FERM domain (head).
Phosphorylation of ERM proteins promotes their transition from the closed “inactive” conformation (head-to-tail self-association) to the open “active” conformation (dissociation of the three-lobed structure from the C-terminal domain) and their recruitment to the plasma membrane, where they bind membrane phospholipids such as phosphatidylinositol 4,5-biphosphate (PIP
2). These interactions stabilize ERM proteins close to membrane-associated molecules such as CD44, which adopts a more open conformation at its ICD upon Ser
325 phosphorylation
[39]. Activated ERM binds to the FERM-binding domain of CD44 ICD phosphorylated at Ser
325 (a constitutive phosphorylation site operated by CaMKII), forming a dynamic CD44-cytoskeleton association. This association can be disrupted upon PKC activation, which triggers the complete dephosphorylation of Ser
325 followed by the phosphorylation of other Ser residues such as Ser
291 and Ser
316, leading to dissociation of the ERM proteins from CD44 and its disengagement from the actin cytoskeleton
[4][40][41].
Therefore, the phosphorylation status of CD44 at specific Ser residues within the ICD is crucial for the dynamic association of the receptor with the cytoskeletal network and CD44-mediated chemotaxis, motility, and cell phenotypic changes. These events are not random but instead fine-tuned by the coordinated action of kinases and phosphatases. These enzymes could bind directly or indirectly to CD44. For example, ezrin interacts with PKC, thereby acting as a scaffolding protein for other signaling molecules to regulate phosphorylation of CD44 and associated proteins/receptors
[42]. On the other hand, the existence of a PDZ-binding motif in CD44 ICD raises the possibility of a direct interaction of CD44 with PDZ domain-containing phosphatases or indirectly through PDZ domain adaptor proteins, which in turn bind protein phosphatases, thereby regulating the phosphorylation level of CD44 itself as well as other adjacent signaling molecules/receptors. For example, binding of syntenin, a syndecan-binding PDZ protein
[43], to the C-terminus of protein tyrosine phosphatase (PTP) eta could result in the recruitment of this transmembrane PTP into syndecan-containing complexes
[44][45].
The interaction of CD44 with the actin cytoskeleton through binding to ERM proteins and ankyrin is further stabilized by the partitioning of CD44 into detergent-insoluble, cholesterol-rich nano-domains (lipid rafts) of the plasma membrane
[4][46][47] (
Figure 1). The highly conserved intramembrane domain of CD44 can be reversibly palmitoylated at Cys
286 and Cys
295, which enhances localization of the receptor in cellular membranes as well as its interactions with cell surface and cytoplasmic proteins, such as several RTKs, innate receptors (TLRs), ABC transporters, EMMPRIN, PI3K, Src kinases, ezrin, and hyaluronidase 2, which preferentially localize into lipid rafts
[48][49][50][51]. Interestingly, the amount of CD44 in lipid rafts can vary in different cell types and can be displaced from these membrane domains upon E-cadherin expression, which negatively regulates HA–CD44 interactions and CD44-dependent tumor invasion and branching morphogenesis
[52]. In immune cells, CD44 is directed into lipid rafts via palmitoylation, which impairs the CD3-mediated signaling
[53]. The strong association between CD44 and the actin filament network causes CD44 to act like a transmembrane picket that connects the cytoskeleton to pericellular milieu, forming barriers against diffusion of molecules through the plasma membrane
[54].
Of particular interest is the cross-linking of CD44 receptors, which has been associated with the metastatic potential of several human cancers. Receptor aggregation can be drastically induced by both extracellular (pericellular HMW HA) and intracellular (phosphorylated moesin/ezrin) cues
[55]. The association of CD44 with ezrin and actin promotes HA binding to cells
[56][57].
5. Regulation of Cell–Cell Contact Inhibition and Cell Growth
CD44-mediated contact inhibition of cell growth is modulated by counteracting CD44 ICD interacting proteins; e.g., the tumor suppressor protein merlin and ERM proteins. Their reversible dynamic association with CD44 ICD regulates the effects of the receptor on cell growth and motility since it provides an ON (CD44–ERM)/OFF (CD44–merlin) mechanism in the CD44–actin cytoskeleton association and signal transduction pathways (
Figure 1). The importance of these interactions in cell physiology is evidenced by the finding that long-lived naked mole rats, which synthesize HA of exceptionally high molecular size, show a remarkable resistance to cancer through regulation of cell–cell contact inhibition via the HA–CD44–merlin signaling axis
[58][59][60]. Mechanistically, ERM proteins and merlin are both hypophosphorylated in high-cell-density conditions
[61]. Under these conditions, ERM proteins are inactive, while merlin adopts the active unphosphorylated closed conformation and interacts with HA-bound CD44 and E-cadherin, thus stabilizing the homophilic E-cadherin interactions and cell–cell adhesion complexes and rendering the cells stationary. Activated merlin prevents proteolytic processing of the CD44 extracellular domain, which in turn preserves cell density signaling and inhibits cell proliferation and migration
[62].
In addition to merlin, PAR1b (partitioning defective 1b) also links CD44 and the Hippo pathway
[63]. Ooki and colleagues found that HMW–HA-mediated CD44 clustering induces the interaction of CD44 ICD with the PAR1b/microtubule affinity-regulating kinase 2 (MARK2) complex, which normally inactivates MST1 and MST2 Ser/Thr kinases through binding and subsequent inhibitory phosphorylation, thus triggering Hippo signaling activation and contributing to contact inhibition of cell growth.
6. Cleavage and Intracellular Release of CD44 ICD: A Master Regulator of the Transcriptome
The proteolytic cleavage of the extracellular domain or shedding of membrane-associated proteins is an irreversible post-translational modification that regulates cell–cell communication, intercellular signaling, and biological functions by releasing growth factors, enzymes, and soluble receptors
[64]. In some instances, the remaining protein in the membrane is further cleaved, generating additional products that include a cytoplasmic fragment that can function in intracellular signaling, revealing an additional level of regulation of cellular functions. In line with this, CD44 can undergo successive proteolytic processing by a number of proteases, resulting in the generation of both extra- and intracellular bioactive fragments. These proteases are recruited and act at the cleavage site under specific prerequisites. The formation of CD44 dimers has been suggested to be critical
[65].
The sequential cleavage of CD44 begins with the ectodomain shedding by transmembrane MMPs (i.e., MT1-MMP and MT3-MMP) as well as other CD44 sheddases (i.e., disintegrin or ADAM 10, ADAM 17, and meprin β), resulting in the generation of a membrane-associated fragment of CD44 while its extracellular N-terminal region is released to the ECM
[5][66]. ERM proteins are involved in CD44 shedding, since they mediate the colocalization of MT1-MMP and CD44. In particular, the radixin FERM domain simultaneously binds the cytoplasmic tails of MT1-MMP and CD44, forming a complex that is stabilized through anchoring to filamentous actin. This ternary complex facilitates the recognition of the CD44 stem region by the PEX domain of the MT1-MMP ectodomain, resulting in CD44 shedding
[67][68] (
Figure 2).
Figure 2. Schematic illustration of the roles of CD44 ICD in the regulation of transcriptome and cell metabolism. CD44 (via its ICD) dictates the metabolic shift from mitochondrial oxidative phosphorylation to aerobic glycolysis (the Warburg effect) through regulation of important metabolic enzymes such as PKM2, PFKFB4, and LDH. Enhanced pentose phosphate flux results in NADPH production that promotes GSH synthesis, which in turn suppresses ROS accumulation. Increased lactate production due to enhanced aerobic glycolysis results in the generation of an acidified pericellular microenvironment, which favors MMP activation. MMPs can be also activated by the TGFβ/TβR/TRAF6/RAC1 axis, thus triggering CD44 cleavage and the translocation of CD44 ICD to the nucleus, where it regulates the transcription of target genes, including CD44 itself (for details, see text).
7. Regulation of Cell Metabolism
One of the proteins that were identified by the proteomic approach to interact with the C-terminus of CD44 was the M2 splice isoform of pyruvate kinase (PKM2), a glycolytic enzyme that catalyzes the generation of pyruvate and ATP from phosphoenolpyruvate and ADP
[30]. While this finding was considered to possibly be non-specific since PKM2 is an abundant protein that is often detected in biotin-based pull-down experiments, two years later, Tamada and colleagues confirmed the interaction of CD44 with PKM2
[69]. This finding suggested a role of CD44 in cancer cell metabolism, since PKM2 is a key enzyme that regulates aerobic glycolysis or the Warburg effect, a metabolic pathway utilized by tumor cells regardless of the local availability of molecular oxygen to produce energy (ATP) and promote tumor growth
[70]. The low or high enzymatic activity of the PKM2 isoform dictates the conversion of pyruvate to lactate, which leads to the Warburg effect or to acetyl-CoA for mitochondrial oxidative phosphorylation, respectively
[70][71][72][73]. Interestingly, the interaction of CD44 with PKM2 reduces the enzymatic activity of PKM2 via tyrosine phosphorylation by RTKs, which triggers glycolysis and the flux to the pentose phosphate pathway, a major source of NADPH (
Figure 2). This mainly occurs in hypoxic or p53-mutated glycolytic cancer cells, since p53 is known to positively regulating mitochondrial oxidative phosphorylation through cytochrome c oxidase 2 induction
[69][74]. Knock-down of CD44 enhanced oxidative phosphorylation, which led to GLUT1 suppression and a subsequent reduction in glucose uptake and pentose phosphate pathway flux. These changes resulted in reduced NADPH production, down-regulation of reduced glutathione (GSH) synthesis, and increased accumulation of ROS
[69].
Further, CD44 ICD enhanced aerobic glycolysis through induction of 6-phosphofructo-2-kinase/fructose-2, 6-biphosphate 4 (PFKFB4), a key enzyme that catalyzes 6-phosphofructose (F6P) to fructose-2,6-biphosphate (F-2,6-BP) via bidirectional conversion during glycolysis
[75][76].
Additional important CD44 ICD-induced genes encoding enzymes involved in aerobic glycolysis include
ALDOC (which encodes aldolase c, fructose biphosphate), and
PDK1 (which encodes pyruvate dehydrogenase kinase-1)
[75], highlighting CD44 as a gatekeeper of the Warburg effect (
Figure 2).
8. Conclusions
In conclusion, the association of CD44 ICD with cytoskeletal effectors (ERM, merlin, IQGAP1, and ankyrin) drives cytoskeleton rearrangements and affects the distribution of organelles and transport of molecules. In addition, through specific interactions of its cytoplasmic tail, CD44 regulates the cell-trafficking machinery, the transcriptome, and major cell metabolic pathways, with substantial impact on cell functional properties like survival, proliferation, adhesion, differentiation, therapeutic resistance, stemness properties, and EMT .