|Home | About | Journals | Submit | Contact Us | Français|
Parkinson disease (PD) is the most common movement disorder. It is characterized by bradykinesia, postural instability, resting tremor, and rigidity associated with the progressive loss of dopaminergic neurons in the substantia nigra pars compacta. Another pathological hallmark of PD is the presence of α-synuclein proteiniacous inclusions, known as Lewy bodies and Lewy neurites, in some of the remaining dopaminergic neurons. Mounting evidence indicates that both genetic and environmental factors contribute to the etiology of PD. For example, genetic mutations (duplications, triplications or missense mutations) in the α-synuclein gene can lead to PD, but even in these patients age-dependent physiological changes or environmental exposures appear to be involved in disease presentation. Several additional alterations in many other genes have been established to either cause or increase the risk of Parkinson disease. More specifically, autosomal dominant missense mutations in the gene for leucine-rich repeat kinase 2 (LRRK2/PARK8) are the most common known cause of PD. Recently it was shown that G2019S, the most common diseasing-causing mutant of LRRK2, has dramatic effects on the kinase activity of LRRK2: while activity of wild-type LRRK2 is inhibited by manganese, the G2019S mutation abrogates this inhibition. Based on the in vitro kinetic properties of LRRK2 in the presence of manganese, we proposed that LRRK2 may be a sensor of cytoplasmic manganese levels and that the G2019S mutant has lost this function. This finding, alongside a growing number of studies demonstrating an interaction between PD-associated proteins and manganese, suggest that dysregulation of neuronal manganese homeostasis over a lifetime can play an important role in the etiology of PD.
Parkinson disease (PD) is the most common movement disorder, affecting over 6 million people worldwide. PD can present with a juvenile or early onset, but it predominantly afflicts individuals over the age of 55 and the incidence of disease sharply rises after the age of 65 (Gelb et al., 1999; Moghal et al., 1994; Simuni and Hurtig, 2000). The clinical features of PD include bradykinesia, postural instability, resting tremor, and rigidity which are predominantly associated with the progressive loss of dopaminergic neurons in the substantia nigra (SN) pars compacta (Cronford et al., 1995; Forman et al., 2005; Forno, 1996). It is believed that during normal aging approximately 0.1–0.2% of the dopaminergic neurons in this area are lost per year, but this rate is greatly accelerated in patients with PD and symptoms manifests when ~70–80% of these neurons have been lost (Calne and Peppard, 1987; Damier et al., 1999; Pakkenberg et al., 1991; Uversky, 2004). Another pathological hallmark of PD is the presence of α-synuclein proteiniacous inclusions, known as Lewy bodies (LBs) and Lewy neurites (LNs), in some of the remaining dopaminergic neurons (Cronford et al., 1995; Forman et al., 2005; Forno, 1996). Importantly, the loss of neurons in PD is not exclusive to nigral dopaminergic neurons as several other neuronal populations are also affected (Braak et al., 2006; Cronford et al., 1995; Forman et al., 2005; Forno, 1996).
Although the majority of PD cases are idiopathic, ~10% of cases report with a family history, and a growing number of mutations have been associated with familial and sporadic forms of the disease (see Table 1) (Lesage and Brice, 2009; Westerlund et al., 2010). In addition to genetic defects, a range of environmental and occupational factors, such as pesticides like paraquat to more ubiquitous metals such as manganese, have been implicated as risk factors in PD (Dick et al., 2007; Elbaz et al., 2009; Lai et al., 2002). While there are some cases where a single environmental or monogenetic factor may lead to PD (MPTP poisoning or triplication of α-synuclein, respectively) (Farrer et al., 2004; Przedborski et al., 2001; Singleton et al., 2003), it is more likely that a subtle yet complex interplay exists between genetic and environmental factors in the etiology of disease. For example, the autosomal dominant G2019S mutation in leucine-rich repeat kinase 2 (LRRK2) is the most common known cause of familial and sporadic patients with PD, yet its penetrance is age-dependent, and some individuals may never be afflicted (Goldwurm et al., 2007; Kachergus et al., 2005). The manifestation of such a predominant mutation as a late onset disorder, where it is still not fully penetrant, suggests that genetic defects may serve to predispose individuals to certain environmental challenges.
Although LBs and LNs were originally observed ~ 100 years ago in the brain of PD patients, it was not until 1997 following the discovery of a PD kindred with a point mutation in the gene for α-synuclein (SNCA) that its presence in these proteinaceous inclusions was examined (Polymeropoulos et al., 1997; Spillantini et al., 1997). It is now accepted that α-synuclein filaments are the major ultrastructural component of these cytoplasmic pathological inclusions, which can be observed in a spectrum of neurodegenerative diseases termed “synucleinopathies” (Forman et al., 2005; Goedert, 2001; Spillantini et al., 1997).
α-Synuclein is a highly charged 140-amino acid heat stable protein that is soluble and natively “unfolded” (Davidson et al., 1998; El-Agnaf et al., 1998; Weinreb et al., 1996). It is predominantly expressed in neurons of the central nervous system (CNS), where it localizes to presynaptic terminals in close proximity to synaptic vesicles (George et al., 1995; Jakes et al., 1994; Withers et al., 1997). Although the function of α-synuclein is still poorly understood, several studies suggest that it is involved in modulating synaptic transmission and neuronal plasticity (Abeliovich et al., 2000; Cabin et al., 2002; Iwai et al., 1995; Murphy et al., 2000; Withers et al., 1997; Greten-Harrison et al., 2010), as well as providing support in the assembly and folding/refolding of SNARE proteins critical for neurotransmitter release, vesicle recycling, and synaptic integrity (Chandra et al., 2005; Burre et al., 2010).
In addition to the first missense mutation (A53T) that was identified in α-synuclein, two additional disease-causing missense mutations were found: an A30P mutation in a German family (Kruger et al., 1998), and an E46K mutation in a Spanish family (Zarranz et al., 2004). In vitro, α- synuclein can readily form fibrils similar to those seen in LBs (Conway et al., 1998; Giasson et al., 1999; Wood et al., 1999) and the A53T and the E46K mutations can both increase this rate of fibril formation (Conway et al., 1998; Giasson et al., 1999; Greenbaum et al., 2005) suggesting a link between α-synuclein aggregation and disease. This link was strengthened with the identification of several PD kindreds with triplication (Farrer et al., 2004; Singleton et al., 2003) or duplication of the α-synuclein gene (Chartier-Harlin et al., 2004). In vitro, increased concentrations of α-synuclein have been shown to promote the polymerization of α-synuclein into fibrils (Wood et al., 1999), and patients with a gene triplication have greater disease severity and younger age of onset than those with gene duplication, suggesting a possible “SNCA gene dosage effect” leading to PD (Singleton and Gwinn-Hardy, 2004). Furthermore, these findings indicate that a 50% increase in the expression of α-synuclein due to gene duplication is sufficient to cause disease and is consistent with the aggregation of α -synuclein contributing to disease.
The polymerization of α-synuclein from unstructured monomer to mature amyloid fibrils proceeds through the formation of partially folded intermediates and several altered-sized oligomers (Conway et al., 2000; Uversky et al., 2001a). Several of these intermediates (as well as products that may not culminant into fibrils) have been described as spheres (2–6 nm in size), chains of spheres (also termed protofibrils) and rings resembling circular protofibrils (also termed annular protofibrils) (Conway et al., 2000; Ding et al., 2002; Goldberg and Lansbury, Jr., 2000). Some investigations suggest that protofibrils or some form of α-synuclein oligomers may increase membrane permeability leading directly to cell death [reviewed in (Waxman and Giasson, 2009)], or indirectly injuring cells by α -synuclein extracellular activation of microglia, and subsequent induction of proinflammatory responses and generation of ROS (Gao et al., 2008; Lee et al., 2010b; Su et al., 2009; Theodore et al., 2008). However, there is also substantial evidence, especially using transgenic α-synuclein mice, supporting the toxic nature of mature α-synuclein inclusions that can behave as "sieves", which can trap other macromolecules and perturb cellular homeostasis, axonal transport, and synaptic transmission [reviewed in (Waxman and Giasson, 2009)]. Aside from the sequestering of other proteins, the loss of synuclein function in and of itself may have relevance towards disease progression, as studies of genetically altered mice null for α-, β, and γ-synuclein demonstrated age-dependent neuronal dysfunction and alterations in synaptic structure and transmission, and revealed that continuous presynaptic SNARE-complex assembly required a nonclassical chaperone activity mediated by synuclein proteins (Burre et al., 2010; Greten-Harrison et al., 2010). Importantly, the proposed mechanisms of α-synuclein toxicity are not mutually exclusive, and several forms of aberrant α-synuclein aggregates may lead to neuronal demise.
Recently, LRRK2 has generated the most attention in PD research due to the identification of numerous mutations that are associated with disease. In 1997, Hasegawa and Kowa reported a Japanese kindred with an autosomal dominant mode of PD inheritance (Hasegawa and Kowa, 1997), and in 2004, two separate groups simultaneously identified the LRRK2 gene in a handful of other families as the gene responsible for disease (Paisan-Ruiz et al., 2004; Zimprich et al., 2004). Since then, over 75 sequence variations in the LRRK2 gene have been found, and it is regarded as the most common known cause of familial and sporadic cases of PD (Biskup and West, 2009; Dachsel and Farrer, 2010; Giasson and Van Deerlin, 2008). The LRRK2 gene spans ~7.5mb and contains 51 exons. It encodes a large 2,527 amino acid protein with multiple complex domains, including N-terminal leucine-rich repeats, a GTPase ROC (Ras of complex proteins) domain followed by COR (C-terminal of Roc) domain, a mitogen-activated protein kinase kinase kinase (MAPKKK) catalytic domain, and C-terminal WD40 repeats (Figure 1) (Mata et al., 2006). At least 5 missense mutations in LRRK2 are considered definitely pathogenic (R1441C/G, Y1699C, G2019S, and I2020T), and 2 others are considered increased risk factors for disease (R1628P, G2385R) (Dachsel and Farrer, 2010; Giasson and Van Deerlin, 2008).
LRRK2 is predominantly a cytoplasmic protein, but it can be associated with various organelles such as the golgi, mitochondria, and ER, as well as various membranes such as lipid rafts and synaptic vesicles (Biskup et al., 2006; Giasson et al., 2006; Gloeckner et al., 2006; Greggio et al., 2006; West et al., 2005). LRRK2 is found in numerous tissues including the lungs, heart, liver, kidney, spleen, testes, and brain (Giasson et al., 2006; Li et al., 2007; Maekawa et al., 2010). In human and rodent brain, LRRK2 is expressed and localized to a variety of neuronal populations, but it is highly expressed in the dopaminoceptive regions, such as the caudate-putamen, and frontal cerebral cortex (Biskup et al., 2006; Galter et al., 2006; (Higashi et al., 2007a; Higashi et al., 2007b; Melrose et al., 2006; Melrose et al., 2007; Simon-Sanchez et al., 2006; Taymans et al., 2006). Some studies showed negligible levels of LRRK2 mRNA in human SN dopaminergic neurons (Galter et al., 2006), but others have reported clearly detectable expression in these neurons (Higashi et al., 2007a). Similarly, in rodent brains, several studies have reported the absence or extremely low levels of LRRK2 mRNA in the SN (Galter et al., 2006; Higashi et al., 2007b; Melrose et al., 2006), while others have reported detectable levels (Simon-Sanchez et al., 2006; Taymans et al., 2006). Nevertheless, using LRRK2 antibodies, the protein was detected at moderate levels in nigral dopaminergic neurons (Higashi et al., 2007b; Taymans et al., 2006).
The majority of patients with LRRK2 mutations clinically present with typical features of PD, but some additional clinical symptoms have also been described (Biskup and West, 2009; Giasson and Van Deerlin, 2008). To date, only a relatively small number of cases have come to autopsy, but the majority of these patients present with classic degeneration of the SN with traditional Lewy pathology. Despite most cases sharing similarity to idiopathic PD, a significant number of individuals possess atypical features such as the presence of tau predominant pathological inclusions or α-synuclein negative, tau negative, ubiquitin positive inclusions (Biskup and West, 2009; Giasson and Van Deerlin, 2008). Cases with LRRK2 mutations may also present as PD with no abnormalities in α -synuclein pathology; for example, 2 patients with the G2019S mutation have been reported to lack Lewy pathology (Gaig et al., 2007; Giasson et al., 2006), and 6 of the 8 autopsied patients with the I2020T mutation also present without α-synuclein inclusions (Funayama et al., 2005; Hasegawa and Kowa, 1997). Currently, it is unclear how patients even carrying the same mutation in LRRK2, manifest with varying pathology associated with PD.
The G2019S mutation is by far the most common LRRK2 mutation. A large case-control study from 2008 found that G2019S accounts for more than 85% of the patients carrying a LRRK2 mutation (Healy et al., 2008), which accounts for 0.6–1.6 % of sporadic PD (Deng et al., 2005; Gilks et al., 2005; Kachergus et al., 2005) and 2–8 % of familial PD cases (Correia et al., 2010; Deng et al., 2005; Hernandez et al., 2005; Kachergus et al., 2005; Nichols et al., 2005; Paisan-Ruiz et al., 2005). In certain ethnicities, such as North African Arabs and Ashkenazi Jews, the G2019S mutation is an even greater factor contributing to PD, as it is present in 22–41% of individuals with disease (Lesage et al., 2006; Lesage et al., 2005; Ozelius et al., 2006). Given that an active kinase domain is required for LRRK2-mediated neurotoxicity in cellular models (Greggio et al., 2006; Ho et al., 2009; Iaccarino et al., 2007; MacLeod et al., 2006; Smith et al., 2005), and that this highly prevalent pathological mutation lies within the activation loop of the kinase domain (see Figure 1), much attention has been directed towards understanding how aberration of kinase activity leads to disease.
Protein kinases require the formation of an ATP-divalent metal cation (DMC) complex for the phosphoryl transfer of the γ-phosphate of ATP to a protein substrate (Knowles, 1980). Typically, Mg2+ serves as the essential metal ion for catalysis, however Mn2+ and other divalent cations may support nucleotide binding and subsequent phosphoryl transfer (Courtneidge, 1985; Elberg et al., 1995; Koland and Cerione, 1990; Sun and Budde, 1997; Toru-Delbauffe et al., 1986). G2019 is part of the highly conserved DYG motif in subdomain VII of the kinase region (see Figure 1) (Hanks et al., 1988; Kannan and Neuwald, 2005), and the D residue in this motif is involved in Mg2+ binding and proper coordination of ATP-Mg2+ in the active site (De Bondt et al., 1993; Karlsson et al., 1993; Levinson et al., 2006; Taylor and Radzio-Andzelm, 1994). The DFG motif is located at the N-terminal hinge region of the activation loop that switches from an open and extended conformation in the active state to a more closed conformation in the inactive state (De Bondt et al., 1993; Kannan and Neuwald, 2005; Taylor and Radzio-Andzelm, 1994). In the process of inactivation, the G residue usually performs an extreme twist, thereby facilitating this D residue to turn away from the catalytic site. The G2019S mutation may disrupt the ability of this movement, thereby keeping the D residue positioned in the active site for longer activation periods that may contribute to a greater catalytic rate. Given this model, initial in-silico studies suggested that the G2019S mutation could cause an increase in LRRK2 kinase activity (Albrecht, 2005). Indeed, the overwhelming majority of studies on LRRK2 kinase activity show that the G2019S mutation increases activity around 2–3 fold over wild-type (Greggio and Cookson, 2009), and that this increase may be responsible for the toxic nature of this protein. Recently, small molecule LRRK2 kinase inhibitors were shown to protect against LRRK2-mediated neurodegeneration in vitro and in vivo (Lee et al., 2010a), supporting the notion that overactive kinase activity may be a potential target for therapeutic intervention.
Manganese is an essential trace mineral necessary for normal development and biological function (Roth, 2006). It is an essential co-factor for many enzymes including arginase, glutamine synthetase, and superoxide dismutase-2. The homeostatic level of free manganese in tissue and cells is maintained by various transporters and by binding to various proteins (Aschner et al., 2007; Au et al., 2008). Despite its necessity for proper metabolic function, excessive exposure to manganese is a well-recognized occupational and environmental hazard (Aschner et al., 2007; Au et al., 2008) which can lead to an extrapyramidal syndrome, referred to as manganism (Aschner et al., 2007; Perl and Olanow, 2007). Although this condition has motor symptoms that resemble PD, it also has several distinguishing features separate from PD, such as dystonia and the lack of response to dopamine replacement therapy (Aschner et al., 2007; Perl and Olanow, 2007). Additionally, neuropathological assessments demonstrate preferential cell loss in the globus pallidus and to a lesser degree the caudate, putamen, and subthalamic nucleus, with relative sparing of the SN (Aschner et al., 2007; Perl and Olanow, 2007). This has brought many authors to question the role of manganese in a number of PD-relevant pathways. However, despite these clinical and pathological differences between idiopathic PD and individuals developing motor disability due to acute manganese exposure, recent findings suggest that lifetime exposure to lower levels of manganese in some individuals may be a risk for typical PD (Squitti et al., 2009). This point of divergence is noteworthy, as it is possible that the degenerative mechanism for acute overexposure is different from its toxicity in a lower, more physiologically relevant form of exposure.
At the cellular level, manganese toxicity occurs through several mechanisms including mitochondrial dysfunction and subsequent depletion of ATP, protein aggregation, oxidative stress, and disruption of neurotransmitter synthesis [reviewed in (Benedetto et al., 2009)]. Specific to dopaminergic cells, studies in rat striatal tissue have shown that manganese can block dopamine synthesis at the level of tyrosine hydroxylation (Hirata et al., 2001) as well as induce the release of dopamine from intracellular stores (Daniels et al., 1981). This extracellular release may add to the deleterious effects of manganese, as recent evidence from C. elegans suggests that manganese can oxidize extracellular dopamine before it is recovered at the synapse (Benedetto et al., 2010). Oxidative stress may also result from the activation of microglia by manganese. Multiple studies have shown that sub-toxic concentrations of manganese in neuronal cultures result in cell death in the presence of microglia (Kitazawa et al., 2005; Latchoumycandane et al., 2005; Zhang et al., 2009), and that manganese may act by itself, or in combination with proinflammatory factors to induce neuronal cell death (Zhang et al., 2010).
Throughout the literature, there are a number of reports establishing a link between manganese exposure and PD or PD-related disorders, however there is some realm of conflict between certain epidemiological studies. Separate from manganese exposure alone, it is more likely that an interaction exists between genetic and environmental factors that may shape the etiology of disease and account for differences in susceptibilities.
Manganese and α-synuclein can combine to form deleterious effects on cell survival, as neuronal cell lines overexpressing α-synuclein show increased sensitivity to manganese challenge (Cai et al., 2010; Li et al., 2010; Pifl et al., 2004). Furthermore, manganese stress has been shown to induce the expression of α-synuclein in cultured cells (Cai et al., 2010; Li et al., 2010), and manganese can accelerate the rate of α synuclein fibrillization in vitro (Uversky et al., 2001b). The importance of the increased α -synuclein expression is underscored by the genetic findings that indicate that a 50% increase in α-synuclein expression is sufficient to cause PD. Therefore, both of these effects of manganese on α-synuclein would promote the formation of α-synuclein aggregates in vivo.
Recent in vitro studies characterizing the kinase activity of LRRK2 have shown that of several DMCs assayed, only Mg2+ or Mn2+ can be used as an ATP cofactor to support the kinase activity of wild-type LRRK2; however Mg2+ is a much better cofactor at promoting this activity (Covy and Giasson, 2010; Lovitt et al., 2010). In sharp contrast, both Mn2+and Mg2+ are equally effective at promoting the activity of G2019S LRRK2 (Covy and Giasson, 2010; Lovitt et al., 2010). It was shown that this difference was a result of dramatic changes in the kinetic properties for the ATP-DMC substrate. While wild-type LRRK2 showed much slower catalytic rates in the presence of Mn2+ compared to Mg2+, G2019S LRRK2 demonstrates similar ATP turnover rates in the presence of Mn2+ and Mg2+. More importantly, both wild-type and G2019S LRRK2 demonstrated significantly higher affinities for Mn-ATP compared to Mg-ATP (Covy and Giasson, 2010; Lovitt et al., 2010). Consequently, sub-stoichiometric concentrations of Mn2+ can inhibit the kinase activity of wild-type LRRK2 driven by ATP-Mg2+, but under similar conditions, the kinase activity of the G2019S mutant remains highly active.
It was proposed that these enzymatic kinetic properties of LRRK2 may reflect important characteristics of this enzyme that were designed for biological functionality. As such it was suggested that LRRK2 may act as a sensor for increased cytoplasmic manganese levels, which results in decreased kinase activity that would modulate downstream counteractive measures (Covy and Giasson, 2010; Lovitt et al, 2010). As a consequence of its alterations in kinetic properties, it is predicted that the G2019S mutant would remain largely active at physiologically elevated manganese levels and therefore this putative signaling pathway should be compromised. Interestingly, in a recent study analyzing the effects of mutant LRRK2 on α-synuclein, it was shown that the transgenic expression of G2019S LRRK2 in mice that express A53T human α-synuclein caused a dramatic acceleration in the progression of neurodegeneration (Lin et al., 2009). Although the mechanism(s) for this synergistic effects were not clearly established, altered manganese metabolism is a possibility.
In 2006, mutations in ATP13A2/PARK9 were identified as the cause of Kufor-Rakeb syndrome, a juvenile recessive multisystemic neurodegenerative disorder with prominent parkinsonism (Ramirez et al., 2006). Splicing, short duplication or deletion mutations resulting in truncated forms of the protein were found in the original family, as well as a family from Chile (Ramirez et al., 2006). More recently a homozygous missense mutation (F182L) in ATP13A2 was found in a Japanese family with Kufor-Rakeb syndrome (Ning et al., 2008), while another homozygous missense mutation (G504R) was identified in a individual with juvenile PD (Di Fonzo et al., 2007). ATP13A2 is a large, 1,180 amino acid protein belonging to the lysosomal type 5 P-type ATPase family of transporters (Ramirez et al., 2006). Recently, ATP13A2 has been shown to be protective against α -synuclein induced toxicity in yeast, C. elegans, and primary neurons (Gitler et al., 2009). In addition, ATP13A2 may play a role in sequestering heavy metal ions possibly by acting as a lysosomal transporter, as it exhibits protective affects in yeast against a number of metals including manganese, cadmium, nickel, and selenium (Gitler et al., 2009; Schmidt et al., 2009).
Interestingly, some studies in cultured cells show that the expression of parkin (PARK2) can protect against manganese-induced toxicity in cultured cells (Higashi et al., 2004; Roth et al., 2010). Recessive mutations in the parkin/PARK2 gene are the most common cause for early-onset parkinsonism (Giasson and Lee, 2001; Kitada et al., 1998). Parkin is an E3 ligase that can function to modify proteins via poly-ubiquitination, which can then be recognized as a target for degradation by the proteasome (Giasson and Lee, 2001; Giasson and Lee, 2003), and the majority of parkin mutations have been shown to cause a loss of this function. 1B–DMT1, the major isoform of the divalent metal transporter DMT1, is a substrate for parkin (Roth et al., 2010). DMT1 has broad substrate specificity, but plays an important role in the transport of manganese (Aschner et al., 2007; Au et al., 2008), suggesting that parkin may play a significant role in maintaining the steady state levels of this transporter. Therefore, the loss of parkin activity may lead to a disruption in the regulation of cellular manganese metabolism, rendering cells more vulnerable to increased uptake of manganese.
Although direct evidence linking manganese and the demise of neurons in PD is still limited, some of the studies summarized here seem to build circumstantial evidence that dysregulation of manganese homeostasis at the cellular level may play an important role in the etiology of PD, at least for some patients. Notwithstanding the controversy surrounding some of the epidemiological studies, and the differences in the pathological findings in patients with acute exposure to manganese, it is difficult to simply overlook the effects of manganese on two of the most prominent factors in PD: α-synuclein and LRRK2. While there are still many questions about the exact mechanisms linking manganese and PD, there is evidence suggesting a direct effect of this metal cation on α-synuclein expression and aggregation. More recently, the most common mutation associated with PD, the G2019S kinase mutation in LRRK2, was also shown to have a unique and provocative affect linked to manganese. Based on the effects of this mutation and the kinetic properties of LRRK2 in the presence of Mn2+, we proposed that LRRK2 may be a cytoplasmic sensor of manganese and that the G2019S mutation abrogates this function. Perhaps the relative low levels of LRRK2 in dopaminergic neurons render these cells more vulnerable to the effects associated with the G2019S mutation. This hypothesis will have to be validated in vivo. Nevertheless, the mounting evidence linking manganese to PD suggests that it is possible that a dysregulation of manganese homeostasis over a lifetime can play an important role in the etiology of PD.
This work was supported by grants from the National Institute of Neurological Disorders and Stroke (NS053488) and The Ellison Medical Foundation (AG-NS-0331-06).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of Interest
The authors declare that there are no conflicts of interest.