Within the pancreatic islet, the β-cell represents the ultimate biosensor. Its central function is to accurately sense glucose levels in the blood and consequently release appropriate amounts of insulin. As the only cell type capable of insulin production, the β-cell must balance this crucial workload with self-preservation and, when required, regeneration. Evidence suggests that the β-cell has an important ally in intraislet endothelial cells (ECs). As well as providing a conduit for delivery of the primary input stimulus (glucose) and dissemination of its most important effector (insulin), intraislet blood vessels deliver oxygen to these dense clusters of metabolically active cells. Furthermore, it appears that ECs directly impact insulin gene expression and secretion and β-cell survival. This review discusses the molecules and pathways involved in the crosstalk between β-cells and intraislet ECs. The evidence supporting the intraislet EC as an important partner for β-cell function is examined to highlight the relevance of this axis in the context of type 1 and type 2 diabetes. Recent work that has established the potential of ECs or their progenitors to enhance the re-establishment of glycemic control following pancreatic islet transplantation in animal models is discussed.
The Islets of Langerhans
The Pancreatic Islet: The Ultimate Biosensor
Distributed throughout the exocrine pancreas, the islets of Langerhans house the central regulator of glucose homeostasis—the β-cell. β-Cells are the major cellular component of islets, making up approximately 80% of islet cellularity in mice and 50–60% in humans. The remainder of the islet comprises other endocrine cells (including glucagon-secreting α-cells, somatostatin-secreting δ-cells, pancreatic polypeptide-secreting γ-cells, and ghrelin-producing ε-cells), as well as endothelial cells (ECs) and supportive cells, including pericytes.
The body’s response to glycemic load is extremely rapid—circulating plasma insulin levels increase within 1 min of ingestion of food as β-cells release preformed insulin from granules. In addition to this rapid first-phase response, islets continue to release insulin in a pulsatile fashion over a longer period (second-phase response) in the presence of a persisting glucose stimulus. Pulsatile insulin release is coordinated within the islet and throughout the pancreas by intercellular gap junctions and ATP spreading. Pancreatic hormone secretion is additionally regulated by the autonomic nervous system, and endocrine cells in murine islets are heavily innervated by parasympathetic and sympathetic neurons. Human islets, on the other hand, are only sparsely innervated, with most neurons contacting intraislet smooth muscle cells (1). Thus neuronal control of human β-cells may be indirect and mediated via the vasculature. The exquisite regulation of blood glucose is vital for the proper function of multiple systems—including neuronal, cardiovascular, and renal. Microvascular beds throughout the body are sensitive to glucose-mediated toxicity. In diabetes, hyperglycemia leads progressively to associated macrovascular and microvascular complications, including retinopathy, nephropathy, neuropathy, and cardiovascular disease (2).
Pancreatic islets are highly vascularized and receive 10% of pancreatic blood flow despite comprising only 1–2% of tissue mass. Small islets (<100 µm) are incorporated into the microcapillary beds of the exocrine pancreas, while larger islets are supplied by up to three dedicated arterioles. The vasculature is crucial for pancreatic development, largely via vascular endothelial growth factor (VEGF)-A-mediated signals. Intraislet ECs are thin and highly fenestrated, allowing for sensitive detection of blood glucose levels and rapid dissemination of secreted insulin (Fig. 1), and it has been postulated that every β-cell is in contact with a vascular EC. Islet cellular arrangement differs among species. While rodent islets classically consist of a β-cell core surrounded by a non-β-cell mantle, the structure of human islets is more irregular and varies with islet size and perfusion. One in-depth study of human islets proposed a trilaminar arrangement comprising a layer of β-cells sandwiched between two α-cell layers with vessels lining both faces, and this structure folded upon itself to form an islet (3). This model suggests the existence of more heterologous intercellular contacts compared with rodent islets.
In common with other microvascular cells, intraislet ECs express CD31 and von Willebrand factor, internalize acetylated LDL, upregulate endothelial markers upon activation (including E-selectin and L-selectin), contain Weibel–Palade bodies in the cytoplasm, and form tight junctions (4). In humans, intraislet ECs are the only microvascular cell type found to express high levels of α-1 proteinase inhibitor at cell junctions (4), which appears to maintain them in a nonproliferative state.
In addition to providing a conduit for blood flow, intraislet ECs directly enhance insulin transcription and secretion and stimulate β-cell proliferation (5). This may be through the secretion of humoral factors, through the production of basement membrane components, or via cell-contact-dependent mechanisms.
The β-Cell/EC Axis
Within the pancreatic islet, the intimate association between β-cells and ECs has implications for both cell types, and it is more appropriate to consider the components within the multicellular islet as a mosaic unit, rather than distinct cellular entities. Multiple mechanisms for driving bidirectional communication between β-cells and ECs exist—including those mediated by soluble, extracellular matrix (ECM), and cell-bound molecules (Fig. 2).
Within the pancreatic islet, soluble factors are an important component of the crosstalk between β-cells and ECs. While multiple soluble islet-derived factors have been implicated in β-cell survival and insulin secretion, it is only recently that their cellular origin has been more accurately investigated.
As the major secreted β-cell product, insulin acts in an autocrine and paracrine manner to promote β-cell survival. Additionally, insulin causes the upregulation of endothelial nitric oxide synthase in ECs (6), which is likely to promote intraislet blood flow and enhance its own dissemination. β-Cells secrete VEGF-A in large amounts early in development and throughout adult life (7). VEGF-A expression is further upregulated in islets by hypoxia and glucose (8) and is important for the establishment of native intraislet vasculature (9), maintenance of β-cell mass (10), and the revascularization of islets following transplantation (11). In ECs, VEGF-A induces cell migration and proliferation and maintains fenestrations. The potential of VEGF-A to enhance the transplantation of pancreatic islets has been thoroughly investigated in animal models (12,13). However, additional VEGF-A is not necessarily beneficial—while VEGF-A overexpression may increase the number of intraislet vascular ECs, the resultant vasculature is dysfunctional and islet function is impaired (14). Within islets, the abundant production of proangiogenic VEGF-A is balanced by angiostatic and antiangiogenic factors, including thrombospondin-1 (15). Thus it appears that in the steady-state, VEGF-A production is strictly controlled to maintain the intraislet vasculature at an appropriate density and with functional architecture.
Vascular ECs produce multiple factors that modulate gene expression, proliferation, and cell survival in β-cells. In islets, thrombospondin-1 is almost exclusively expressed by the intraislet endothelium. Thromobspondin-1 is upregulated in human islets by high glucose (16), and while knockout mice have increased β-cell mass and improved revascularization of transplanted islets (17), they are glucose intolerant and display impaired islet function (15). In addition to its antiangiogenic properties, thrombospondin-1 activates latent transforming growth factor-β (a potent regulator of adult β-cell insulin gene transcription and islet function ). Endothelin-1, a vasoconstrictive protein, is produced by ECs and may directly stimulate insulin secretion in β-cells (19). Hepatocyte growth factor (HGF) is a proangiogenic factor expressed by intraislet ECs (20). HGF signals by high affinity binding to its receptor c-Met, which is expressed by β-cells. The overexpression of HGF in islets reduces β-cell death in islet transplantation models (21) and may be responsible for β-cell proliferation during pregnancy (20). Most recently, conditional knockout of HGF signaling in the pancreas increased the susceptibility of mice to streptozotocin-induced hyperglycemia (22). While these mice had normal glucose homeostasis and β-cell mass, they were more vulnerable to β-cell damage mediated via the nuclear factor κ-light-chain-enhancer of activated B-cells (NK-κB) pathway. Thus, HGF/c-Met signaling is likely to be of importance for islet survival, particularly during cytokine-mediated damage, as occurs following islet transplantation and during the development of diabetes.
ECM proteins form depots for growth factors to support cellular proliferation and function. The majority of ECM proteins exist within basement membranes (underlying vessels and other cellular structures) or in the interstitial matrix. Pancreatic islets are encased within a peri-islet basement membrane and associated interstitial matrix, containing multiple ECM components, including laminins, perlecan, collagen type IV, and nidogens (23). In addition, intraislet blood vessels have their own basement membrane (5). In the mouse islet, endocrine cells are either associated with peri-islet or vascular basement membranes, while in human islets, intraislet blood vessels contain a double basement membrane (24).
Within pancreatic islets, the ECM plays a multifaceted role and is implicated in islet development (25), function (5), and survival (26). The peri-islet membrane is an important physical barrier for immune cell infiltration during diabetes development (23,27). β-Cells do not contribute to the islet ECM directly and instead depend on intraislet ECs to synthesize their basement membrane (5). ECM components increase β-cell survival and proliferation within pancreatic islets (26) and promote insulin gene expression (5), mainly via β1-integrins on the surface of β-cells. This has been demonstrated in experiments where the improved insulin secretion found in β-cells cultured with intraislet EC-derived factors (28) or attached to ECM (29) is prevented by anti-β1-integrin blocking antibodies.
Cell Surface Proteins
There are multiple cell surface proteins that are likely to be involved with β-cell/EC crosstalk. Cellular contacts within pancreatic islets are crucial for calcium flux and insulin secretion following glucose sensing, and glucose-stimulated insulin release is enhanced in whole islets compared with isolated β-cells (30). These contacts allow the islet to respond to stimuli as an entire unit via coordinated calcium oscillations that amplify insulin secretion and quickly return to the resting state when appropriate. Connexins, ephrins, and cadherins are expressed in pancreatic islets and have all been implicated in β-cell function. Here we will concentrate on the connexin molecules—a family of proteins well known for their gap junction properties that may play a newly appreciated role in paracrine signaling (31).
Connexins cluster on the cell surface at gap junctions, oligomerizing to form hemichannels that dock with identical or different hemichannels on neighboring cells. The major β-cell connexin is Cx36 (this nomenclature indicates a connexin with subunits of 36 kDa), which was initially thought to be restricted to neurons. Cx36 junctions allow the passage of ions [in particular, Ca2+ and cAMP (32)] between β-cells during synchronized glucose-induced Ca2+ oscillations (30) and work to dampen Ca2+ elevations following membrane depolarizations (33). Thus Cx36 within islets allow less excitable β-cells to act as a buffer, suppressing electrical activity in their neighbors. Islets deficient in Cx36 have increased basal insulin release with diminished glucose-stimulated insulin responses (30), and in humans, Cx36 is expressed on a genomic region that mediates increased risk of type 2 diabetes (34), further suggesting an important role for this molecule in islet function.
Cx43 and Cx45 have also been identified in pancreatic islets and are specifically expressed on intraislet endothelium rather than on β-cells (35). Interestingly, islet-specific overexpression of Cx43 increases islet size and insulin content (36). A new islet-associated connexin, Cx30.2, has been very recently described and is expressed at cell–cell contacts in both ECs and β-cells (37). Thus it is tantalizing to suggest that β-cells and ECs may communicate directly via heterotypic connexin junctions, as has been postulated to occur in the retina (38) and kidney (39). Alternatively, recent evidence that uncoupled hemichannels sample small molecules (e.g., ATP) from the extracellular environment (31) suggests that the role of connexins within the islet may not be limited to physical cellular contact, as they may also participate in paracrine signaling.
Implications of the β-Cell/EC Axis in Diabetes
Type 1 Diabetes
Type 1 diabetes is caused by the specific autoimmune destruction of β-cells within pancreatic islets. Cells infiltrating pancreatic islets during the early stages of type 1 diabetes include macrophages and cytotoxic T cells. The intraislet vasculature represents a barrier for infiltrating autoreactive cells in type 1 diabetes, and thus ECs have been implicated as an important target in the diabetogenic process. During islet damage, such as after streptozotocin treatment in mice, lymphocytes adhere to the intraislet endothelium more readily (40). In addition to permitting the entry of infiltrating immune cells, there is evidence that cytokine-activated intraislet ECs mediate β-cell apoptosis via the release of nitric oxide, which activates apoptotic caspase and DNA damage pathways (41,42). Human intraislet ECs become activated and upregulate their expression of adhesion molecules and proinflammatory cytokines following infection with coxsackie B virus (43)—a postulated initiator of type 1 diabetes.
Type 2 Diabetes
The pathogenesis of type 2 diabetes includes genetic and environmental factors that result in elevated circulatory free fatty acid levels and insulin resistance, followed by β-cell dysfunction. Mechanistically, glucotoxicity, lipotoxicity, proinsulin biosynthesis, and advanced glycation end products act in concert to inhibit β-cell function and increase apoptosis (44). In the hyperglycemic, hyperlipidemic type 2 diabetes patient, vascular ECs are in constant contact with these potentially cytotoxic substances. While hyperglycemia may directly induce apoptosis in intraislet ECs by upregulating the expression of reactive nitric oxygen species via a Jun NH2-terminal-kinase-mediated (JNK) pathway (45), an earlier vascular-associated pathogenic mechanism has been postulated to exist. In models of type 2 diabetes, early microvasculature changes are detectable within islets and precede the onset of hyperglycemia, overt EC destruction, and β-cell degeneration. In the Zucker fatty rat, endothelial thickening and loss of fenestrations were accompanied by increases in VEGF-A mRNA and protein levels (46)—contrary to what might be predicted by VEGF-A knockouts that have a similar islet vascular phenotype (47). However, VEGF-A causes the overproduction of ECM proteins and in this way may contribute to inflammation and fibrosis in islets via the recruitment of macrophages and other leukocytes (48,49). Thus onset of type 2 diabetes may be preceded by VEGF-A-driven endothelial dysfunction, overproduction of ECM, and inflammation, which combine to bring about the ultimate destruction of the β-cell.
Implications of the β-Cell/EC Axis in Pancreatic Islet Transplantation
Despite improved insulin delivery techniques and drug design, approximately 90% of diabetic patients develop severe cardiovascular complications over time. Pancreatic islet transplantation represents a promising therapy for type 1 diabetes that is effective in reducing HbA1c and hypoglycemia (50,51) and is beginning to yield improved protection from diabetes-associated complications (52). Upon organ procurement, islets are separated from their dense capillary network and rich blood supply and are dependent upon the diffusion of oxygen and nutrients from the islet periphery until revascularization can occur. During enzymatic isolation from the donor pancreas, islets come under a myriad of cellular assaults, including ischemia, physical stress, and loss of contact with key basement membrane proteins and signaling molecules (53), resulting in a substantial loss of viability before transplantation. Clinically, islets are cultured prior to transplantation, and intraislet ECs reduce rapidly in this period (54,55). Therefore, not only do islet transplants face multiple cytotoxic challenges, but they are depleted of the building blocks required for rapid revascularization. In the first few days post-transplantation, islets undergo a dramatic remodeling process accompanied by changes in insulin content, and hypoxic and inflammatory insults have a detrimental effect on islet survival and function. A major limitation for islet transplantation is the substantial amount of cell death seen in this early post-transplant period, which accounts for the loss of up to 60–80% of transplanted islet mass (56). This necessitates the use of multiple donors for each recipient—a major limitation for the roll out of this therapeutic option into the wider type 1 diabetes community.
Revascularization of Transplanted Islets
Many groups have examined revascularization events following islet transplantation. One difficulty has been correlating results from the renal subcapsular mouse model in which an aggregate of pristine rodent islets are transplanted into a syngeneic avascular site, to the human situation, where islets of variable purity and quality are dispersed throughout the portal circulation of an often hyperglycemic, immunosuppressed patient. Nevertheless, the field has gained important insight into the mechanisms underlying islet revascularization. This has been further enabled by powerful models such as the transplantation of islets into the rodent anterior chamber of the eye to allow real-time microscopic analysis of cellular events (57).
Unlike solid organs, transplanted islets are not directly reconnected to the blood supply. Following intraportal transfusion, approximately half of surviving transplanted islets remain within the portal vein tributaries, while the remainder migrate further into the vessel wall (58). Vascular sprouting, angiogenesis, and revascularization occur within the first few days post-transplantation. These processes involve recruited bone-marrow-derived cells, recipient local vascular cells, and donor “passenger” ECs derived from the islet transplant itself (54,59,60). The contribution from donor cells may explain the superior function of freshly isolated islets (61), which are endowed with additional donor intraislet ECs, and small islets (62), which are less dependent on revascularization. Multiple bone marrow lineages, including hematopoietic, mesenchymal, and endothelial, all participate in the early stages of islet engraftment via the production of soluble factors, recruitment of accessory cells, and/or incorporation into newly formed vasculature. Ultimately, the functional intraislet vasculature in a transplanted islet is a mosaic of donor and recipient-derived cells, the majority of these being generated from existing local vessels, with bone-marrow-derived cells comprising less than 10% of long-term intraislet vasculature (58). Islet engraftment is a slow process, and while islet blood flow is reestablished in 7–14 days (63), the maturation of these vessels is likely to take several months. In addition, immunosuppressive drugs, including mammalian target of rapamycin inhibitors (e.g., rapamycin), potentially compound the problem, as they have been found to inhibit angiogenesis in human intraislet ECs (64). Although some studies, including postmortem analysis of an islet transplant recipient (65), show little difference in the ultimate vascular density of transplanted islets, most studies suggest that revascularized islets display a decreased vascular density and lower oxygen tension compared with native islets, regardless of the transplantation site (66). Recent work using an accurate microsphere technique showed that the perfusion of intraportally transplanted islets was 5% that of native islets at 1 month post-transplantation, with 24% the vascular density (58).
Future Directions to Improve Islet Transplantation
Considering the importance of the vasculature for islet function, it is of little surprise that research has been directed at enhancing the revascularization response post-transplantation. While a number of molecules and pathways have been targeted, the approach showing most potential to date is the overexpression of VEGF-A by transplanted islets. However, as discussed above, the question of optimal VEGF-A dose for islet vasculature is not a simple one. While VEGF-A is crucial for the vascularization of islets during development and following transplantation (7,47), continued β-cell overexpression of VEGF-A can impair islet morphology and result in poorly functioning islets with an inflammatory phenotype (14,49).
Another approach to improve outcomes for islet transplantation aims to promote β-cell survival (67,68). While multiple studies have shown potential for maintaining viable β-cells in the early post-transplant period, it is known that even long-term-surviving islets have impaired function, and this may be due to the paucity of functional intraislet vasculature. Human embryonic stem cells may now be differentiated into insulin-producing glucose-responsive cells that express markers of islet cell development, such as pancreatic duodenal homeobox-1 (69), and cure diabetes in rodent models (70). While stem cell-derived β-cells hold enormous promise for alleviating the problem of islet donor shortage, it is likely that these cells will need to be cotransplanted with cells or factors capable of directing their appropriate differentiation and supporting β-cell survival. Thus, assuming we are able to overcome the acute phase of immunological and cytotoxic attack of the islet mass, a parallel challenge is to establish appropriate vasculature, sufficient to support an optimally functioning pancreatic islet in the longer term.
Restoration of the β-Cell/EC Axis With Endothelial Progenitor Cell Therapy
A cellular cotherapy at the time of transplantation is one strategy to improve islet revascularization and/or enhance β-cell function. The ideal candidate should be able to migrate to the required site, deliver supportive factors, and persist for the appropriate time. Mesenchymal stem cells stimulate vessel ingrowth into islets in vitro (71), and their cotransplantation improves islet transplant outcomes in rats (72). While somewhat efficacious, concerns linger regarding the potential invasiveness and pluripotency of mesenchymal stem cells, especially in the immune-compromised transplant population.
Endothelial progenitor cells (EPCs) are a circulating bone-marrow-derived cell population first described by Asahara et al. (73). EPCs are able to home to sites of tissue damage or ischemia and can participate in wound healing, postnatal vasculogenesis, and re-endothelialization of blood vessels (74). Over 200 clinical trials involving EPCs are currently underway, and significant progress has been made in the utilization of these cells as a therapeutic tool. Given their complex interactions with multiple cell types, augmentation of EPC function is as important as increasing gross peripheral numbers. Human autologous cell therapies using EPC-containing products (such as bone marrow or mobilized peripheral blood) are feasible and effective in the treatment of coronary and peripheral ischemic syndromes (75). Most recently, human clinical cell therapy trials have applied bone-marrow-derived mononuclear cells or CD34+ or CD133+ isolated hematopoietic progenitor cells because of their easy accessibility and safety [reviewed in (76)].
EPCs represent an ideal candidate to provide an engraftment niche for transplanted islets. EPCs may enhance islet engraftment in multiple ways—by providing not only the building blocks for revascularization, but also the angiogenic and cell survival factors to augment the process. EPCs secrete multiple factors, including HGF, which enhances the survival of transplanted islets and stimulates β-cell proliferation. While mature ECs lack the survival characteristics and plasticity to re-establish intraislet vasculature, EPCs possess these properties. In addition, EPCs isolated from peripheral blood or bone marrow permit the use of autologous EPCs, lessening the risk of sensitization to multiple alloantigens in the transplant setting. Considering the ability of EPCs to migrate to distant sites, alternate forms of delivery, including intravenous infusion of exogenous EPCs or upregulation of endogenous EPCs from the bone marrow, may be considered as an adjunct to pancreatic islet transplantation in the future. Another alternate mode of delivery is the construction of mosaic structures in vitro (77), which may comprise combinations of islet cells, β-cell progenitors, EPCs, mature ECs, and/or mesenchymal stem cells, prior to transplantation. This approach has the advantage of delivering a minimal number of adjunct cells placed in the optimal anatomical location to exert their supportive roles. Consideration should also be given to additional proangiogenic supportive cells such as microvascular pericytes, which encircle capillaries and microvessels and regulate microvascular physiology (78), and proangiogenic myeloid cells such as Gr1+CD11b+ cells, which are also mobilized from the bone marrow in response to granulocyte colony-stimulating factor, migrate to sites of neovascularization, and promote angiogenesis (79).
In 2012, Kang et al. (80) demonstrated the potential of human EPCs for enhancing the engraftment of transplanted porcine islets in an immunodeficient mouse xenograft model. After cotransplanting EPCs with islets under the kidney capsule, they saw an increase in vascular cell density within the first 2 weeks of transplantation. Despite showing no difference in vessel density at 1 month, there was improved glycemic control in cotransplanted animals due to an increase in engrafted β-cell mass, highlighting the impact of hastening early revascularization events. More recently, researchers have used syngeneic models to show that cotransplanted murine EPCs improve the cure rate, function, and vascularization of a marginal transplanted islet mass (81,55). Observations that the EPCs localized within and around the islet mass and that VEGF was produced by these cells, suggest that EPCs may act in a paracrine and autocrine manner to improve islet engraftment. Taken together, these studies support the notion that EPCs, or the factors they produce, may be of benefit in enhancing the survival and engraftment of transplanted pancreatic islets.
The Intraislet EC: A Key Player in Diabetes and Islet Transplantation?
While the destruction of otherwise healthy β-cells is a hallmark of type 1 diabetes, type 2 diabetes is characterized early by β-cell dysfunction. Regardless of these differences, it is clear that the vasculature, in particular within the pancreatic islet, is likely to play an important role in both diseases. Likewise, the successful reestablishment of intraislet vasculature following pancreatic islet transplantation will be important both for β-cell survival in the early post-transplant period and optimal islet function in the longer term.
As such, the intraislet EC may represent an important common breaking point—whereby failure of this multifunctional cell type could result in multiple downstream events with detrimental effects on glycemic control. Dysfunctional intraislet endothelium is likely to enhance the recruitment and infiltration of autoreactive cells and ultimately further stimulate the autoimmune process. Disruption of vascular tone and suboptimal intraislet blood flow will result in altered dissemination of insulin, interrupted clearance of toxic metabolites and reduced provision of blood-borne oxygen and nutrients. Considering the vascular ECs themselves, alteration in phenotype during islet isolation, culture, and transplantation would perturb the production of ECM proteins, secreted β-cell-supportive factors, and direct cellular signals that are so crucial for the proper operation of the pancreatic islet unit.
In short, a dysfunctional intraislet vascular endothelium may contribute to the progression of type 1 diabetes, the worsening of type 2 diabetes, and the failure of transplanted pancreatic islets. Therapies that prevent breakdown of the intricate β-cell/EC axis within the pancreatic islet, or restore this crosstalk once it has been interrupted, have the potential to improve outcomes for diabetic patients in the future.
Acknowledgments. The authors thank David Heinrich (Flinders Medical Centre, Australia) for assistance in preparing the original artwork, Al Powers (Vanderbilt Diabetes Center, U.S.) for reviewing the manuscript in draft form, and Daniel Nyqvist (Karolinska Institute, Sweden) for the provision of Fig. 1, reproduced with permission from Nyqvist et al. (82).
Funding. Research in the authors’ laboratories is supported by grants from the Australian Research Council, the National Health and Medical Research Council, and Diabetes Australia. C.S.B. holds a South Australian Cardiovascular Research Development Program Fellowship. D.J.K. holds an ARC Future Fellowship. C.F.J. is supported by a fellowship from The Flinders University of South Australia.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. H.P., C.S.B., P.T.H.C., and D.J.K. contributed to discussion and reviewed and edited the manuscript. C.F.J. wrote the original draft of the manuscript.