Connexin37 Regulates Cell Cycle in the Vasculature

Gap junctions have been classically described as tumor-suppressive cell growth regulators due to the early observation that cancer cells lack gap junction-mediated intercellular communication [1], and that forced expression of connexins – the constituent proteins of gap junctions – limits cancer cell proliferation [2]. Although gap junctions have been classically studied with regard to their role as intercellular channels that electrochemically couple adjacent cells (channel-dependent signaling), it is also possible for cytoplasmic domains of the connexin (Cx) protein to serve as a scaffold and/or binding partner for intracellular signaling cascade effectors to regulate signaling independent of its gap junction channel function (channel-independent signaling). Thus, an open question within the gap junction field is whether connexin regulation of cell growth and proliferation relies on gap junction channel-dependent exchange of signaling molecules, or on channel-independent binding of connexins to cell cycle regulators – or possibly, on both. An improved understanding of this mechanism is essential, as connexin protein family members are diversely expressed in nearly all tissues of the body where their capacity to limit cell proliferation is likely not restricted solely to the pathological setting of cancer. Instead, connexins and gap junctions likely play a central role in the regulation of tissue growth and organization under a broad array of physiological (e.g., development) and pathological (e.g., wound healing, disease) contexts.

Endothelial-lined blood and lymphatic vessels critically interconnect all tissues of the body to circulate nutrient- and oxygen-rich blood, and to remove metabolic waste. The blood vasculature forms during embryonic development from primordial endothelial cells that coalesce into a primitive vascular plexus (vasculogenesis). This plexus expands by endothelial cell proliferation (sprouting angiogenesis); equally important, however, is the subsequent wave of antiproliferative signals that then suppress endothelial cell proliferation to support reorganization of the plexus (vessel remodeling) into a mature, quiescent, and stable vasculature. Indeed, endothelial cells that line the inner luminal surface of mature blood vessels are characteristically quiescent, but can be induced by injury or disease to reenter an “activated,” proliferative state that supports de novo vessel growth and vessel wall repair [3]. Thus, cell cycle control in endothelial cells is central to their function, both during initial vessel formation as well as in response to injury or disease. Four connexin isotypes are commonly expressed in the blood vasculature – Cx37, Cx40, Cx43, and Cx45. All have been shown to play critical roles in vascular development, maintenance of vascular tone, and responses to vessel injury and/or disease [4‒12]. Of these, Cx37 serves a distinct and important role to regulate endothelial cell cycle during vessel development and in disease. Here, we review Cx37’s function in the vasculature and further address the mechanism by which Cx37 likely controls vascular cell cycling.

Although Cx37 is expressed in multiple tissues of mature animals, including lung [13, 14], ovary [15], immune cells [16], and bone [17], it is a principle connexin in the endothelium of blood and lymphatic vessels [18, 19]. In blood vessels, Cx37 forms gap junction channels primarily between endothelial cells [20, 21]; Cx37 also participates in myoendothelial junctions that electrochemically couple endothelial and smooth muscle cells [20‒22] in some (but not all) vessel beds (e.g., in mesenteric arteries [22, 23] but not in coronary vessels [24, 25]). In the resistance vessels of postnatal [10, 12] and adult animals [20, 24], Cx37 (and Cx40) is expressed at high levels in the endothelium in areas of nonturbulent laminar blood flow. However, in regions of flow disturbance and high shear (such as at vessel branch points [20], where endothelial cell turnover rates are increased nearly 1,000-fold [3]), Cx37 (but not Cx40) is profoundly downregulated. These data served as early evidence to suggest that Cx37 expression may be sensitive to luminal blood flow and might link hemodynamic flow to cell turnover rate.

Fluid shear stress – that is, the frictional force of blood as it flows over the inner luminal surface of endothelial-lined blood vessels – plays an important role in the blood and lymphatic vasculature and is a key determinant of endothelial cell orientation, migration, proliferation status, and proinflammatory signaling. Indeed, the function and expression of many signaling effectors in endothelial cells – including cell surface ligand-activated receptors, ion channels, intracellular kinases, small signaling molecules, and microRNAs – are flow-sensitive [26]. Upstream of these flow-sensitive intermediaries are several proposed mechanosensors, including the junctional mechanosensory complex composed of PECAM-1, VE-Cadherin, and VEGFR2 (as well as VEGFR3, when expressed) [27‒30]. Notch signaling is also flow-sensitive [31] and the Notch1 receptor has recently been proposed to function as a direct mechanosensor of fluid shear stress in established arteries [32]. Together, changes in hemodynamic flow sensed by these mechanosensors can induce widespread changes in endothelial cell gene expression (and function) via activation of several mechanotransductory transcription factors that induce broad changes in gene expression in response to flow. Classically, physiological laminar flow induces endothelial cell elongation, alignment, quiescence, and anti-inflammatory signaling via activation of Krüppel-like master transcription factors KLF2 and KLF4 [33]. By contrast, disturbed, oscillatory, and turbulent flow (as occurs in certain areas of the vascular tree such as at vessel branchpoints, as well as in developing or remodeling vessels or during wound healing) suppresses KLF2/4 activation and induces endothelial cell proliferation and proinflammatory signaling via activation of master transcription factors such as NFΚB [27, 34‒36].

Several groups have now confirmed that Cx37 expression is also highly sensitive to flow. Indeed, the Cx37 gene (Gja4) is directly bound by core mechanosensitive transcriptional regulators Klf2 [37] and Notch-activated RBPJ [10], and is also a downstream target of the BMP/Alk1 signaling cascade [38]. Although the specific mechanosensors that directly sense fluid shear stress upstream of flow-activated Cx37 signaling are unclear, Notch is a prime candidate in endothelial cells since it has recently been found to be a mechanosensor in arteries [32]. Fang et al. [10] reported that in isolated primary human endothelial cells, flow-sensitive upregulation of Cx37 via Notch is maximal at arterial levels of shear (i.e., 18 dynes/cm²). Interestingly, Notch-Cx37 activity is reduced at higher or lower levels of fluid shear stress, suggesting that this signaling axis may be a downstream effector of the fluid shear stress setpoint signaling cascade (reviewed in Baeyens et al. [29, 36]) – that is, a homeostatic setpoint hypothesized to be distinct to endothelial cells based on their position in the vascular tree, such that changes in fluid shear stress away from their setpoint induces (inward or outward) vessel remodeling to restore homeostatic fluid flow. Consistent with this idea, Peacock et al. [38] recently found that flow-activated Cx37 is abolished by Alk1-mediated Smad1/5 signaling via BMP9 (but not BMP10) activation, and that changes in Cx37 expression are associated with outward remodeling of developing embryonic blood vessels. The Notch-Cx37 axis also appears to be intact in lymphatic endothelial cells [39], which have a lower fluid shear stress set point compared to endothelial cells of the blood vasculature [29], and where Cx37 may additionally be regulated by the action of FOXC2, NFATc1, and others [40]. Thus, mechanosensitive Cx37 transcriptional expression appears to be conserved across endothelial cells of distinct identities and vascular systems as well as targeted by multiple master signaling cascades, suggesting that Cx37 is a key regulator of endothelial cell function.

Endothelial Cx37 is first detected beginning at embryonic day 8.5 in mice – prior to the establishment of systemic blood flow – when it is broadly expressed in developing cardiac [41] and yolk sac [38] blood vessels, as well as in the endothelial cells of the developing endocardium [41]. It remains unknown what signals govern this early upregulation of Cx37 in the primitive vasculature; however, as the vasculature matures and blood flow is established, flow-sensitive signaling appears to assume primary control of Cx37 to limit its expression to arteriolar blood vessels [10, 38], as well as in some veins [42] and in the downstream-facing side of venous valve leaflets [43], where it persists in the adult vasculature. In developing yolk sac vasculature, for example, Cx37 expression is initially widespread but becomes increasingly heterogeneous corresponding with the onset of blood flow, such that it is absent in enlarging microvessels but then restricted to the large arteries by embryonic day 9.5 [38]. Similarly, Cx37 is broadly detected in lymphatic endothelial cells beginning at embryonic day 13.5 coincident with early stages of lymphatic vessel development, but becomes increasingly restricted to the lymphatic valves as the vasculature matures [8]. Thus, changes in Cx37 expression coincide with reorganization and remodeling of immature blood vessels into arteries, veins, and valves, suggesting a role for Cx37 in that process. Consistent with this idea, the developing blood and lymphatic vasculature in Cx37^−/− animals is profoundly disorganized, with reported malformation of arteries [10, 12], valvular veins [43], and lymphatic vessels [8].

Unique among the vascular-expressed connexins, Cx37 profoundly arrests cells in the G1 phase of the cell cycle (as observed both by fluorescence-activated cell sorting – an approach highly sensitive to changes in cell cycle distribution – as well as by in situ techniques) [10, 44]. Work by Fang et al. [10] showed that this occurs via upregulation of the cell cycle regulator p27/Kip1 (by an as-yet-unknown mechanism) [10], which limits passage through the G1/S cell cycle checkpoint (Fig. 1). Using the Cdt1-mOrange cell cycle reporter mouse, Fang found that endothelial cells in developing arteries of wild-type (Cx37-intact) animals are predominantly in G1 phase of the cell cycle; however, in the developing neonatal retinal vasculature of Cx37^−/− mice [10], the number of arterial endothelial cells in G1 is greatly reduced (with concomitant increase in number of cells in S, G2, and M phases of the cell cycle). This effect is replicated in vitro in primary human endothelial cells using silencing RNA to knockdown endogenous Cx37 expression, which significantly reduces the distribution of cells in G1 [10]. In vivo ablation of Cx37 drives endothelial cell overproliferation in developing blood and lymphatic vasculature [8, 10], leading to disruption of specialized vessel structures such as arteries (including loss of arteriovenous specification and failure of mural cell investment [10, 12]), as well as malformation of lymphatic and venous valves [8, 43]. These phenotypes appear to be a direct consequence of the loss of Cx37-mediated cell cycle control, since pharmacological induction of G1 arrest supports upregulation of arteriovenous identity markers even without flow-activated increases in Cx37 [10]. Thus, Fang et al. [10] proposed that flow-dependent Notch-Cx37-p27 signaling is not an arteriovenous specification signal. Rather, Cx37-dependent signaling may serve as a rheostat for G1 arrest, such that moment-to-moment changes in Cx37 rapidly sensitize cells to local vessel remodeling cues and allow them to generate context-appropriate responses. For example, flow-sensitive BMP9-activated Alk1-Smad1/5/9 signaling suppresses Cx37 to support endothelial cell migration and outward remodeling [38], whereas flow-activated Notch and Klf2 pathways drive Cx37 transcription to stabilize resulting vessels and permit vessel specification and other remodeling events [10, 45].

Given that Cx37 is required for organized vessel growth and remodeling during development, it is perhaps unsurprising that dysregulated Cx37 is also associated with diseases involving pathological vessel growth and disorganization. The mature endothelium in vessels with laminar flow is highly stable, and endothelial cells only rarely undergo cell division except following injury or in disease [3]. It is possible, therefore, that the characteristic stability and quiescence of mature blood vessels depend, at least in part, upon the sustained presence of Cx37, whereas transient downregulation of Cx37 expression – or perhaps posttranslational modifications that briefly shift Cx37 out of a growth arresting conformation (as proposed in Fig. 1) – temporarily alleviates this growth suppression to support new vessel growth. Thus, pathological dysregulation of Cx37 could lead to insensitivity of endothelial cells to these signals, resulting in an unstable vasculature prone to disorganized and overaggressive vessel growth responses.

To test this hypothesis, Fang et al. [10] assessed the impact of germline knockout of Cx37 on vascular development in the superficial retinal vasculature, a classic model of vessel development, as well as following surgically induced ischemia and vessel remodeling in the hindlimb of adult mice [7, 9]. Fang et al. [10] found that the neonatal retinal vasculature of early postnatal Cx37^−/− mice develops in a hyperdense, plaque-like manner with disrupted arteriogenesis, an observation that was subsequently confirmed by Hamard et al. [12] in these mice. Specifically, loss of Cx37 leads to reduced vascular plexus outgrowth, increased vascular density, aberrant tip cell positioning, and increased endothelial cell proliferation of angiogenic stalk cells, as well as decreased mural cell coverage of developing arteries [10, 12]. With regard to the latter observation, Fang et al. [10] reported that Notch-Cx37-mediated growth arrest supports arterial identity marker expression, whereas Hamard et al. [12] found that Cx37 is required for endothelial expression of soluble activators of mural cell recruitment, PDGF-β and Angiopoietin 2; both observations suggest that loss of Cx37 in developing arteries disrupts arterial specification. In both studies, the authors confirmed that Cx37 expression in vivo is Notch signaling dependent, and that pharmacological inhibition of Notch inhibitor (via DAPT) mimics the hyperdense vascular phenotype observed in Cx37^−/− mice [10, 12].

Although this developmental vascular phenotype largely resolves during later vessel remodeling and reorganization [10, 12], the adult vasculature of Cx37^−/− mice retains hallmarks of early disorganization and overgrowth, such as an excess of native collateral vessels in the hindlimb and pial circulations [7]. Furthermore, adult Cx37^−/− mice exhibit greater angiogenesis and collateral remodeling in response to surgically induced vascular resection and ischemia [7, 9], suggesting that Cx37-deficient blood vessels respond more readily and more aggressively to proangiogenic signaling. While in the postischemic hindlimb this was almost completely protective from ischemic tissue damage [7, 9], loss of Cx37 cannot be considered wholly beneficial. For example, Cx37 downregulation (via application of TNFα) exacerbates the risk of vascular malformation in mice with Smad1/5 haploinsufficiency [38] (a model of the rare genetic disease Hereditary Hemorrhagic Telangiectasia in which Alk1 signaling is dysregulated), and Cx37 polymorphism in humans is associated with increased risk for breast cancer [46], a highly vascularized cancer type [47], which may be due, at least in part, to more aggressive tumor angiogenesis. In humans, somatic mutations that impact the first transmembrane domain of Cx37 (and likely alter Cx37 gap junction channel function) have been linked to the development of hepatic hemangiomas, which are vascular malformations characterized by enlarged or excessive blood vessels [48, 49]. Surprisingly, however, Sathiyandan et al. [50] found that in Cx37^−/− mice, tumor angiogenesis is impaired – not overaggressive, as would be predicted based on its developmental angiogenesis phenotype [10, 12] – leading to reduced tumor burden. This may be an indication that other antiproliferative pathways may have compensated for the loss of Cx37 in these mice with regard to preventing pathological tumor angiogenesis; certainly, more studies are needed to address this possibility.

Nonetheless, despite evidence that Cx37 regulates growth in the vasculature, Cx37 is dispensable for conducted vasomotor responses [6] – a function that instead relies primarily on Cx40-comprised endothelial gap junctions [51‒53]. We therefore propose that Cx37 gap junction channels are typically present in a closed state in healthy, established, and quiescent vessels, and that it is this closed Cx37 channel conformation that induces cell cycle arrest in remodeling and established endothelial cells under laminar flow (Fig. 1). By contrast, we propose that disturbed flow and other proangiogenic signals activate flow-dependent kinases [27] (e.g., upregulation of PKC [54], etc.) that switch Cx37 from this dominant closed conformation to open conformations that are no longer capable of inducing growth arrest, but instead support gap junction channel-mediated intercellular communication. Thus, these proposed open conformations permit endothelial cell proliferation (Fig. 1) to support angiogenesis and other forms of vessel growth. This proposed model of Cx37-mediated cell cycle control in endothelial cells (Fig. 1) is further discussed below.

Despite the evidence that flow-induced changes in Cx37 expression regulate endothelial cell cycle and that this is necessary (but likely not sufficient) for normal vascular development and remodeling, the specific mechanisms underlying cell cycle regulation by Cx37 remain uncertain. In mouse and human endothelial cells, Cx37 induces G1 arrest, while its knockdown or germline knockout leads to increased endothelial cell cycling [10]. Induced or constitutive Cx37 expression similarly arrests connexin-deficient cancer cells (e.g., rat insulinoma, Rin cells, HeLa cells) in the G1 phase of the cell cycle [44, 55], suggesting that Cx37-dependent signaling events that mediate cell cycle control remain intact in these cells. Thus, using Rin “model” cell systems, we conducted a series of studies to identify the specific Cx37 channel conformations (which we determine are closed, fully open, or partially open using electrophysiological approaches) and amino acid residue(s) (which we study using site-directed mutagenesis approaches) are necessary for Cx37-mediated growth suppression.

Gap junction channels form at sites of cell-cell contact following the docking of individual hemichannels – one from each of two contacting cells. Both gap junction channels and undocked hemichannels can then serve as a conduit for either intercellular (gap junction channels) or transmembrane (hemichannel) passage of current carrying ions such as K⁺ and Na⁺. Some gap junction channels (depending on connexin isotype composition) also support intercellular diffusion of larger molecules such as ATP, cAMP, IP₃, and miRNAs. Each hemichannel complex is itself composed of six individual connexin proteins, each with four membrane-spanning domains with one connecting intracellular and two extracellular loops that together form the pore (pore-forming domain), the two extracellular loop domains also mediate hemichannel docking, and intracellular N- and C-terminal domains that include several regulatory moieties. Gap junction channels (and undocked hemichannels) adopt multiple conformations to include at least one closed conformation, one fully open conformation, and multiple partially open conformations. The channels transition rapidly between these conformations (referred to as gating), with the relative time in open conformations quantified as open probability. Interactions of the channel’s pore-forming domain with its N- and C-terminal (CT) domains, respectively, determine the channel’s open probability. These interactions are affected by “gating agents” such as voltage, pH, volatile anesthetics, and phosphorylation state of the 12 connexins in a gap junction channel (or 6 in a hemichannel). The CT also interacts in a phosphorylation-dependent manner with protein components of various signaling cascades, an interaction that can also influence the overall conformation of the channel.

Human (hCx37) and mouse (mCx37) Cx37 vary considerably in their amino acid sequences (particularly in their CT domains), but both form high conductance (∼350 pS) gap junction channels with similar permselective and gating properties, and both display multiple stable open conformations (subconductance states) and a closed conformation [44, 56‒59]. The hemichannels formed by mouse and human Cx37 also share similar channel behaviors [16, 60‒62]. To determine whether a specific open conformation of either Cx37 gap junction channels or hemichannels is necessary for Cx37 to control cell cycle, Good et al. [61, 63] created Cx37 mutants that blocked activity of both types of channel activity, as well as a mutant that blocked activity of gap junction channels, but not hemichannels [64]. Those studies showed that Cx37 must be able to form active gap junction channels to be growth-suppressive – that is, hemichannel activity in the absence of gap junction channel activity is not sufficient. These studies did not, however, determine whether a specific gap junction channel conformation (e.g., fully open, partially open, closed) was necessary for growth suppression by Cx37.

Nelson et al. [62] discovered that wild-type mCx37 gap junction channels – which are potently growth-suppressive in Rin cells – overwhelmingly prefer the closed conformation, suggesting that this may be the growth-suppressive conformation. Consistent with this idea are two additional observations. First, Cx37 is growth suppressive at cell densities where gap junction channels cannot form due to lack of cell-cell contact [44]; and second, conducted vasomotor responses in intact blood vessels are largely unaffected by Cx37 deletion [6], suggesting that although Cx37 is expressed and capable of forming active gap junction channels in endothelial cells, these channels are closed in mature vessels with laminar flow.

Nelson et al. [62] further showed that the pore-forming domain without its CT domain (Cx37-273tr*V5, where the last 50 amino acids have been truncated) forms an active gap junction channel but is not growth-suppressive. Instead, C-terminally truncated Cx37 channels display a preference for a partially open conformation (rather than the growth-suppressive closed conformation of channels formed by full-length, wild-type Cx37). This indicates that – as with other connexin isotypes [65‒68] – the Cx37 CT interacts with the pore-forming domain to regulate channel conformation. Nelson et al. [62] also showed that expressing Cx43-CT37, a chimera formed by the joining of the Cx43 pore-forming domain with the Cx37 CT, supports formation of active gap junction channels, but the chimera is not growth-suppressive. Together, these studies collectively show that interaction between the Cx37 CT and a Cx37 pore-forming domain able to form active gap junction channels is critical for growth suppression. They further suggest that regulatory proteins targeting the Cx37 CT (e.g., growth-related kinases and phosphatases, including those activated in the vasculature by flow-dependent signaling) may alter the CT’s interaction with the pore-forming domain to enable rapid and context-appropriate switching between open (growth permissive) conformations and a closed (growth arresting) conformation (Fig. 1).

Connexins have long been recognized as phosphoproteins, and several studies have shown that the presence (or absence) of even a single phosphoryl group can profoundly alter connexin conformation and function [68‒70]. The Cx37 CT has seven serine residues (serines 275, 285, 302, 319, 321, 325, and 328) and one tyrosine residue (Y332) with a high likelihood (>90% probability) of being phosphorylated by growth regulatory kinases, including MAPK8 (JNK2), CK2b, PKCε, GSK, GRK-3, and STE-Unique [71]. To identify the requirement for these high-probability phosphorylation targets for Cx37 growth suppression, Jacobsen et al. [71] mutated all seven serine residues – or just serines 275, 302, and 328 – into either (phospho-incapable) alanine or (phospho-mimetic) aspartate. The results showed that mCx37 could not only support or suppress proliferation but could also trigger cell death, and that the shift between these phenotypic outcomes (i.e., arrest vs. proliferation vs. death) depends upon Cx37 phosphorylation state. Jacobsen et al. [72] subsequently showed that five of these high probability serine targets are natively phosphorylated (serines 275, 319, 321, 325, and 328) in Rin cells. Although none of the tyrosines in the CT (Y259, Y266, Y281, and Y332) were detected as phosphorylated in this study, Y332 was not recovered in mass spectrometry samples, suggesting that there could be undetected phosphorylation at this site.

Kinases known to target Cx37 in vitro include GSK-3β [55] and protein kinase C (PKC) ([71] and Lampe, unpublished), and other cyclin-dependent and mitogen-activated kinases are additionally predicted to bind to and phosphorylate the Cx37 CT. Indeed, Fang et al. [10] found that Cx37-dependent phospho-activation of p27 requires intact MAPK/ERK signaling, although whether this involved direct phosphorylation of Cx37 by MAPK/ERK is still unknown. Furthermore, Hamard et al. [12] found that phospho-ERK levels are increased in the disorganized, hyperproliferative retinal microvessels of Cx37^−/− mice. Jacobsen et al. [72] found evidence that the S319 site of the mCx37 CT is often endogenously phosphorylated – this site was phosphorylated 204 of the 559 (36%) times it was detected (36%) by mass spectrometry of the mCx37 protein. Several kinases are predicted (GPS 3.0; Group-based Prediction System: http://gps.biocuckoo.org/online.php) to target this site with high probability including MAPK (JNK2, Erk5), CDK (5, 1, CDC28, CDC2), PKC (PRKCB), and more. Of these, JNK2 and Erk5 are of particular interest because both are predicted to target S319 with high probability. Further, JNK is a downstream target of ERK [73], which both Fang et al. [10] and Hamard et al. [12] identify as involved in Cx37-mediated growth suppression in developing vessels. JNK expression is also mechano-sensitive [74] and implicated in endothelial cell cycle arrest [75]. Thus, we propose that JNK and/or ERK may phosphorylate Cx37 at S319 to enable transition between its growth-supporting and growth-suppressive conformations (Fig. 1).

Underscoring the likely importance of the S319 site in Cx37 growth regulation in the vasculature, Taylor et al. [76] found that expression of Cx37 with S319 mutated to mimic constitutive phosphorylation (mCx37-S319D) is sufficient to eliminate cell proliferation, whereas expression of the phospho-incapable mCx37-S319A mutant alleviates Cx37-mediated growth arrest. Importantly, growth-suppressive mCx37-S319D mutant gap junction channels and hemichannels (like wild-type Cx37) prefer the closed conformation, suggesting that a common gap junction channel conformation – which we propose is the closed channel (Fig. 1) – supports growth arrest. These results further suggest that the downstream signaling effector of Cx37-mediated growth arrest is not a channel permeant, but rather phosphorylation- and conformation-dependent change in the interactions of the CT with cell cycle regulators (e.g., transcription regulator) that result in inhibition of cell cycle progression (i.e., a channel-independent mechanism).

Interestingly, the Cx37 319 residue is polymorphic in humans with either proline (Cx37-1019C) or serine (Cx37-1019T) residues present at high frequency within the general population [77‒81]. The Cx37-1019C (Cx37-P319) variant – which presumably cannot be phosphorylated at the 319 site (but could be phosphorylated at nearby sites) – is strongly associated with increased risk for cardiovascular disease [78, 80, 81] and breast cancer [46]. Yet, in vitro studies in HeLa or SK-Hep-1 cancer cells find that Cx37-P319 – not Cx37-S319 – is growth suppressive [55]. These data would appear to contradict the clinical correlation of Cx37-P319 with cardiovascular disease, as well as the results of studies by Taylor et al. [76], showing that phosphorylation at S319 is both necessary and sufficient for growth suppression. Several pieces of information that could resolve the basis for this apparent discrepancy are missing. First, the phosphorylation status of Cx37 in quiescent endothelium in mice and humans (with either polymorphism) is unknown and could differ for the polymorphic forms. Second, the phosphorylation status of hCx37-P319 and hCx37-S319 as expressed in HeLa and SK-Hep-1 cells is unknown and could differ from what occurs in Rin cells, as both HeLa and SK-Hep-1 are metastatic cancer cell lines with likely aberrant growth regulatory signaling [82, 83]. And third, the downstream mediators of Cx37-induced growth arrest in Rin and endothelial cells are unknown and could be dysregulated in HeLa and SK-Hep-1 cells.

Nonetheless, studies implicating S319 in endothelial cell proliferation and cardiovascular disease point to the likely pivotal role played by (phosphorylation at) this site. Jacobsen et al. [72] suggested that S319 serves as a crucial hinge point (in a phosphorylation-dependent manner) for intradomain interactions (i.e., within the CT, between residues 273–318 and 320–333) necessary for growth phenotype control. Here, we expand upon this idea to propose that phosphorylation of Cx37 at S319 (possibly by flow-activated JNK and/or ERK) serves as a critical “switch” between open (growth-supporting) channel conformations and a closed (growth-arresting) conformation in remodeling vessels (Fig. 1).

To also address the role of phosphorylation at serines 275, 321, and 328, Jacobsen et al. [71, 72] mutated these residues to phospho-mimetic (aspartate) or phospho-incapable (alanine) amino acids and assessed the impact of the mutated forms on cell cycle control at low cell density (where gap junction channels are not able to form) and at high cell density (where they do form active gap junction channels). As summarized in Table 1, mutation at these sites reveals a complex relationship between Cx37 phosphorylation state and cell cycle control. Unexpectedly, Jacobsen et al. [72] found that some Cx37 phospho-mutants trigger cell death, suggesting that in addition to (or perhaps as an extension of) its growth-suppressive signaling, Cx37 can also be involved in regulating cell survival (and apoptosis). Of all the Cx37 death-inducing mutants, perhaps most interesting are the Cx37 S328 mutants, in part because S328 in Cx37 aligns with S368 in Cx43, which is a PKC-targeted site [84]. Expression of Cx37 with either an alanine or an aspartate at S328 induces cell death, suggesting that both the phosphorylated and dephosphorylated states are critical for cell survival, with each channel conformation perhaps serving as part of a critical checkpoint during the cell cycle to prevent apoptosis. Each of the death-inducing mutants of Cx37 forms active channels, but they do not share a preference for a specific conductance state or open probability characteristic. Yet, the capacity to open is necessary for the death phenotype as suggested by the Cx37-S321D,T154A mutant, which does not form active channels and does not trigger cell death [72]. Taken together, we propose that in addition to its primary (growth-suppressive) closed and (growth-permissive) open conformations, some open conformations can be further modified by phosphorylation at S321 (and possibly S275 or S328), which triggers cell death (Fig. 1) by a still unknown downstream mechanism. What physiological role (if any) this death-inducing mechanism might play in blood vessels in vivo is unclear, but interestingly, S321 and S275 are also predicted to be phosphorylated by ERK, which promotes endothelial cell apoptosis following vessel injury and disease [85, 86].

Altogether, the studies summarized above indicate that intradomain, phosphorylation-dependent interactions between residues 320–333 and residues 273–318 of the CT – hinged at serine 319 – dictate cell cycle progression versus arrest, as well as regulate cell survival and apoptosis, and that Cx37 dynamically moves between these distinct phospho-states in a context-dependent manner to support vessel growth (i.e., angiogenesis) and maturation (remodeling) (Fig. 1).

In our proposed model, Cx37 phosphorylation profile determines intradomain interactions within the CT, as well as between the CT and the pore-forming domain, to induce distinct channel conformations including a closed conformation that arrests cell cycling (Fig. 1). Fang et al. [10] showed that p27 is both necessary and sufficient for Cx37-mediated G1 arrest, with other downstream targets also possible. How Cx37’s closed conformation might regulate the expression and/or activity of downstream cell cycle regulators such as p27 remains unclear. There are several possibilities: Cx37 may both induce expression and limit intercellular and transmembrane passage of a cell cycle inhibitor signal through Cx37 gap junction channels or hemichannels, which would allow that signal to concentrate in the cell to threshold levels; or, Cx37 may bind to and promote activation of kinases that target cell cycle regulators; or, Cx37 may serve as a scaffold that sequesters cell cycle regulators in an inactive state. There is possible precedence for this last possibility with regard to the modulation of vessel tone.

Both hCx37 (1019C>T) polymorphic variants and mCx37 bind endothelial nitric oxide synthase (eNOS) [87]. When bound to Cx37, eNOS activity is inhibited; and when a recombinant peptide that mimics the reductase domain of eNOS is bound to Cx37, gating of Cx37 gap junction channels is altered (i.e., increased incidence of transitions between closed and fully open conformations, with an unknown impact on open probability). Thus, interaction of eNOS with Cx37 affects the activity of both proteins. Cx37-eNOS interaction in vivo could support the growth-arrested state (i.e., closed gap junction channel conformation) in mature vessels with laminar flow and limit NO production by sequestering (some) eNOS in an inactive state. With stimulation (e.g., turbulent flow, acute injury, disease), changes in Cx37 phosphorylation (perhaps at Y332) might release eNOS making it (along with caveolin-released eNOS) now available to augment NO production and thereby contribute to control of vascular tone (Fig. 1). NO is reported to be inhibitory to Cx37 gap junction channel activity, such that increases in NO cause reduced permeation of these channels to small dye molecules [88]. This may reflect an inhibitory effect of NO on SHP-2 activity, a phosphatase known to bind to the phosphorylated Y332 motif on the Cx37 CT [89]. Thus, we speculate that disturbances of flow, injury, or disease may lead to changes in Cx37 phosphorylation (possibly at Y332) that induce release of bound eNOS, increase channel open probability, and support a proliferative state. If so, these changes would be reversed as NO levels fall and SHP-2 dephosphorylates Y332, returning the gap junction channel to the original closed, growth-arresting conformation (Fig. 1). Hyperpolarizing current and NO could directly pass through Cx37 gap junction channels while they are open, both facilitating vasodilation [90]. Thus, the interplay between Cx37, eNOS, and SHP-2, could have profound effects on vascular tone and reactivity via a mechanism that requires both active Cx37 gap junction channels and protein-protein interactions at the Cx37 CT.

In addition to its expression in vascular endothelium, Cx37 is also expressed in other perivascular cells during development and in the adult vasculature, where in some cell types it also seems to regulate cell proliferation. Cx37 is ordinarily found at low levels in vascular smooth muscle of some arteries [91‒94], but not in others [24], and its expression in the media is significantly upregulated during normal [92] and injury-induced remodeling [95]. Specifically, Allagnat et al. [95] showed that in a carotid artery ligation model of intimal hyperplasia – which drives abnormal dedifferentiation, proliferation, and migration of vascular smooth muscle cells – there is an acute induction of Cx43 and vascular smooth muscle cell proliferation, but this is followed by an eventual upregulation of Cx37 in the medial layer along with localized reduction in cell proliferation and Akt-dependent phosphorylation. This suggests that Cx37 in the media functions as a negative feedback control mechanism to limit additional medial thickening. By contrast, neointimal formation and medial thickening are significantly enhanced in Cx37^−/− mice in association with a twofold increase in smooth muscle cell proliferation [95]. Thus, although Cx37 is only expressed at low levels in vascular smooth muscle compared to its expression in endothelial cells, it may also function in this cell type to maintain vessel wall stability and control the growth response following injury.

Separate from its growth-suppressive role in the vessel wall, Cx37 also regulates cardiovascular health through its role in immune cells. Cx37 is expressed in circulating monocytes, where instead of regulating cell proliferation, it controls monocyte adhesion and extravasation – a critical early step in the pathogenesis of atherosclerotic lesions [16]. Wong et al. [16] found that atheroprone Cx37-deficient mice (ApoE^−/−;Cx37^−/−) have significantly increased atherosclerotic lesion formation (compared to Cx37-expressing ApoE^−/− mice), and further showed that this was a cell autonomous effect of circulating Cx37-deficient leukocytes. Further work in vitro revealed that Cx37 hemichannel-mediated ATP release by monocytes limits their adhesion to vascular endothelium, and that this inhibitory mechanism appears to be impaired in patients with the Cx37-C1019 polymorphic variant (P319), which has been shown in several clinical studies to confer greater patient risk for atherosclerosis and myocardial infarction [16, 78], as well as other cardiovascular disease such as atrial fibrillation [96].

Although Cx37 is frequently co-expressed with other vascular connexins (i.e., Cx40 and Cx43 [20]) in the vessel wall, this review has focused on the structure and function of Cx37 as a homomeric/homotypic gap junction channel. This is because although Cx37 can form mixed composition (heteromeric) channels with Cx43 (and likely Cx40) in vitro in overexpression settings, it is unlikely that Cx37 mixes with these connexins in vivo since Cx37 oligomerizes into hemichannels in the endoplasmic reticulum, whereas both Cx43 and Cx40 oligomerize in the Golgi apparatus [97, 98]. Thus, it remains unclear whether heteromeric channels (in which Cx37 intermixes with Cx40 or Cx43) are formed in the vasculature in vivo, and if so, what impact (if any) the presence of other connexins in a Cx37-containing gap junction channel might have on Cx37’s cell cycle control properties [98]. More studies are needed to address these questions.

In summary, based on several studies in which site-directed mutagenesis was used to examine gap junction channel activity and cell cycle impact, we propose that contextual cues (e.g., changes in hemodynamic flow) activate selective signaling cascades that modify the phosphorylation profile of Cx37 to induce specific channel conformations (e.g., closed vs. multiple open channel conformations) (Fig. 1). We further propose that these changes in Cx37 conformation regulate endothelial (and smooth muscle) cell cycle status to drive proliferation, cell cycle arrest, or cell death – likely via changes in gene expression of cell cycle regulators (Fig. 1). Although published data are best explained by this proposed mechanism, several points remain outstanding: (1) although we propose that changes in hemodynamic flow alter Cx37 phosphorylation state in vivo, this has not yet been directly demonstrated; (2) although studies of phospho-mimetic and phospho-incapable Cx37 mutants show phosphorylation-dependent changes in Cx37 channel conformation, it has not yet been directly demonstrated that hemodynamic flow or associated activation of flow-dependent kinases induces shifts between the open and closed conformations of Cx37 channels in vivo; (3) because Cx37 is most strongly expressed on the arterial side of the vascular tree, as well as in the leaflets of venous and lymphatic valves, it is unclear whether Cx37-mediated cell cycle control is limited to these areas of the vasculature; (4) although some phospho-mimetic Cx37 mutants induce cell death in vitro, it is unknown whether this reflects a physiological function of Cx37 in vivo; (5) the specific phosphorylation status of connexins that comprise open and closed gap junction channels is unknown; and (6) although we propose candidate kinases that phosphorylate the Cx37 CT to switch it between growth-suppressive versus -permissive conformations (Fig. 1), further studies are needed to directly identify the kinases that directly target Cx37, an endeavor likely to be complicated by cross-talk between these regulatory pathways. Altogether, more work is needed to address these and other outstanding questions with regard to our proposed mechanism of Cx37 cell cycle control in the endothelium (Fig. 1).

Cx37 is increasingly understood to be a potent regulator of vessel stability and growth [7, 10, 12], and its dysfunction is implicated in cardiovascular diseases involving vessel malformation [38, 49], vessel wall hyperplasia [95], as well as immune cell responses in atherosclerosis [16]. Polymorphic variation in the Cx37 gene that alters the protein’s amino acid sequence at site 319 in the Cx37 C-terminus is well established to influence risk for cardiovascular disease in healthy and diabetic patients [78, 80, 81, 96]. Thus, Cx37 is a potentially attractive target in the pursuit of novel therapies to address blood vessel disorganization and overgrowth, as well as other forms of cardiovascular disease. In particular, Cx37’s proposed role in determining vessel wall stability and responsiveness to growth (or injury) stimuli suggest that Cx37 may be an important (if still largely unrecognized) disease modifier in a variety of cardiovascular diseases. Studies to better understand how Cx37 status (with regard to polymorphic sequence, phosphorylation, and/or overall expression and function in blood vessels) might influence the severity of patient clinical outcomes are clearly warranted.

In conclusion, Cx37 expression regulates endothelial cell cycle status, serving as an apparent switch between proliferative and remodeling vasculature (Fig. 1). It also appears to regulate several other aspects of cardiovascular health such as vessel tone and immune response. Strategies to selectively target Cx37 in abnormal blood vessels have yet to be developed, and will likely require a better understanding of the mechanism by which Cx37 regulates cell cycle status in vascular cells, and how it is regulated by upstream signaling effectors to do so. Based on comprehensive site-directed mutagenesis studies of Cx37 intracellular domains, we propose that in mature vessels with laminar flow, Cx37 is phosphorylated (possibly at S319) to induce adoption of a distinct growth-suppressive closed conformation that requires interaction within the CT and between the pore-forming and C-terminal domains (Fig. 1). It is possible that this distinct gap junction channel conformation (or other growth-permissive conformations) could be directly and specifically targeted for therapeutic intervention, leaving other Cx37 conformations largely unaffected. Certainly, an improved understanding of how the closed conformation regulates downstream cell cycle regulatory proteins – likely through creation of protein-protein interaction sites distinct to this conformation – would also be valuable in the development of novel therapies for the treatment of a wide array of cardiovascular diseases. The development of such therapies therefore requires a better understanding of Cx37’s functions and their upstream and downstream roles in endothelial cell regulation, and thus demands extensive further study.

The authors wish to acknowledge with gratitude the many insightful discussions with (and experimental contributions of) colleagues over the years – these form the foundation of this review. Without their input, this body of work would not have been possible.

The authors have no conflicts of interest to disclose.

This work was funded, in part, by the Department of Defense (W81XWH2110108, Jennifer S. Fang).

Jennifer S. Fang and Janis M. Burt contributed equally to the writing and editing of this review and to the creation of the figures and tables.