|Home | About | Journals | Submit | Contact Us | Français|
Infantile spasms are an age-dependent epilepsy that are highly associated with cognitive impairment, autism, and movement disorders. Previous classification systems have focused on a distinction between symptomatic and cryptogenic etiologies, and have not kept pace with the recent discoveries of mutations in genes in key pathways of central nervous system development in patients with infantile spasms. Children with certain genetic syndromes are much more likely to have infantile spasms, and we review the literature to propose a genetic classification of these disorders. Children with these genetic associations with infantile spasms also have phenotypes beyond epilepsy that may be explained by recent advances in the understanding of underlying biological mechanisms. We therefore also propose a biologic classification of the genes highly associated with infantile spasms, and articulate models for infantile spasms pathogenesis based on that data. The two best described pathways of pathogenesis are abnormalities in the gene regulatory network of GABAergic forebrain development, and abnormalities in molecules expressed at the synapse. We intend for these genetic and biologic classifications to be flexible, and hope that they will encourage much needed progress in syndrome recognition, clinical genetic testing, and ultimately the development of new therapies that target specific pathways of pathogenesis.
Infantile spasms (ISS) are characterized by clusters of epileptic spasms with ictal electrodecrement, usually occurring before the age of 1 year . The incidence of ISS is 0.25–0.4 per 1,000 live births , and they are important because of their frequent association with severe developmental outcome, including autism .
There have been many variations in terminology describing ISS over the years, with the eponym “West syndrome” generally referring to the triad of spasms, hypsarrhythmia, and mental retardation or regression – despite the difficulty of detecting the latter in a young infant presenting with ISS. Hypsarrhythmia is regarded by some as a key feature in the diagnosis of ISS, but can be so variable that it remains surprisingly troublesome to define, and evolves over time . As a consequence, the literature contains references to “atypical” and “modified” hypsarrhythmia [5, 6], and there is little consensus on the meaning of these terms, or on how much of the electroencephalographic record should be abnormal [4, 7, 8]. In fact, the international West Delphi Group consensus statement did not regard the presence of hypsarrhythmia as essential for the diagnosis of ISS , and correctly refocused the core of the electroclinical disorder on the ictal event associated with an electrodecrement on EEG in an appropriately aged child. ISS without hysparrhythmia are well documented . We will continue to refer to the disorder as infantile spasms, while recognizing that in some children they may not occur during infancy, and may not manifest as classic flexor or extensor spasms, and hypsarrhythmia may only be an associated feature of the interictal EEG.
In the past, the terms symptomatic, cryptogenic, and idiopathic were introduced to categorize ISS, and epilepsy in general . The limitations of this approach to classification are well noted within the epilepsy community [11, 12], and these terms become even more untenable with growing evidence that ISS result from disturbances in key genetic pathways of brain development. Recent studies show that patients with ISS may have mutations in several genes including ARX, CDKL5, FOXG1, GRIN1, GRIN2A, MAGI2, MEF2C, SLC25A22, SPTAN1, and STXBP1 [13–24]. While not all patients with mutations in these genes have ISS, the phenotype is consistent enough to speak of them as ISS-associated genes, modified by incomplete penetrance and variable expressivity. Taken together, these single genes, as well as other candidate genes identified from pathogenic copy number variants, suggest that abnormalities in ventral forebrain development and synaptic functional pathways play critical roles in ISS pathogenesis . We caution that the pathogenesis of ISS may mirror autism in its genetic complexity, where each genetic abnormality probably accounts for about 1% of the disorder [26, 27]. Therefore, many more ISS-associated genes are likely to be discovered, and the evolving genetic data demand a reassessment of this disorder.
New molecular insights from the recent genetic data imply that all forms of ISS may be “symptomatic”, and that a diagnosis of “cryptogenic” ISS should no longer be accepted. For this reason, the International League Against Epilepsy (ILAE) recently recommended replacing the symptomatic, cryptogenic, and idiopathic classification system of epilepsy syndromes . We strongly agree with their recommendation, but argue that it is also possible to take the first steps toward genetic and biologic classifications of ISS.
Many clinical studies of ISS have focused on the subsequent intractable epilepsy that develops in 40–60% of affected children [28, 29]. Although important, this approach has underemphasized the broader phenotypes to which ISS may belong. In contrast, classification systems based on the emerging genetic and biologic data would incorporate the historical connections between ISS and phenotypes other than intractable epilepsy. These observations are not new, and the spectrum of neurologic associations with ISS has always been much broader than intractable epilepsy, as it was for William James West’s son.
West syndrome is an unusual eponym because it is named after the father of the proband. The English physician, William James West, is credited with writing the first description of ISS in English by reporting the history of his son, James . Most authors since have focused only on West’s descriptions of the seizures. However, one report highlighted James’ later history, which included lack of speech and “frequent fits of idiotic laughter, and rollings of the head... delighted by music and gay colours”. This description leads “to the suspicion that James had features of autism” . Langdon-Down, who cared for him in later life, reported that James and other children he observed with ISS had “a great tendency to automatism and rhythmical actions” including “salaams, horizontal swayings, and rotations of the head and body” . These observations may be compatible with the stereotypies often seen with autism.
Later reports continued to show that autism could follow ISS [33–36]. This includes reports of tuberous sclerosis complex (TSC), ISS, and autism , with a similar correlation noted in patients with Down syndrome as well [38, 39]. Some argue that the subsequent autistic spectrum disorder is a consequence of severe epileptic encephalopathy [40, 41]. While this remains a compelling hypothesis, the most recent data suggest a primary biologic link between ISS and autism .
Several early reports of children with ISS described hyperkinesis, but the nature and significance of the finding remained unclear [36, 42]. The use of the term “cerebral palsy” as a putative “predisposing cause” may obscure recognition of a movement disorder as a specific feature associated with ISS [43, 44]. The discovery of ARX as the first gene clearly associated with (previously “cryptogenic”) ISS led to a wider appreciation that involuntary movements may be a key clinical feature of certain syndromes that include ISS . The description of MEF2C deletions and mutations in patients with ISS and dyskinesia strengthens this association [20, 21, 46]. The identification of MEF2C as another ISS-associated gene is interesting as MEF2C may be a transcriptional target of ARX during ventral forebrain development when GABAergic interneurons proliferate and migrate . It is perhaps not surprising that children with mutations in ARX and MEF2C have similar phenotypes. Additionally, dyskinetic movements have been reported in children with CDKL5 deletions  and STXBP1 mutations as well .
We agree with the spirit of the recent ILAE recommendation that epilepsies should be divided into genetic, structural/metabolic, and unknown categories . However, the distinction between genetic and structural/metabolic is artificial and should be modified, as most epilepsies associated with structural brain malformations or inborn errors of metabolism are also genetic. Those that are not primarily genetic, such as epilepsy due to brain injury and infection, are distinct and deserve a place of their own.
A genetic classification has practical benefits. When ISS are recognized as a symptom found in many developmental disorders, rather than seen in isolation, the many disorders associated with ISS can be divided into two main groups: those with the predisposing genotype known and those with the predisposing genotype unknown. Of course, we recognize several important subdivisions, as listed in Table 1. We think that this approach will facilitate future progress in several areas including genetics, pathogenesis, and treatment. Figure 1 illustrates the evaluation of ISS using the classification system presented in Table 1.
This group provides the core data supporting our emerging understanding of ISS pathogenesis, and from this group the later biologic classification of ISS can also be built. These are the newly described genes in which mutations, intragenic expansions, deletions and/or duplications frequently are associated with spasms. Some children with ARX and STXBP1 mutations initially presented with early infantile epileptic encephalopathy (EIEE, Ohtahara syndrome) [17, 50–52] and illustrate a spectrum of infancy-onset epileptic encephalopathies. EIEE, early myoclonic encephalopathy (EME), and ISS remain enigmatic in their relationship to one another, and patient series of these infantile epilepsies overlap in genotype [49, 53–55], suggesting factors of incomplete penetrance and variable expressivity. The type of early-onset epilepsy may be related to severity of mutation, although more data are needed to show this relationship. Therefore, the term prototypic developmental epilepsies acknowledges the variability but likely biologic kinship of these phenotypes.
The two genes most commonly associated with classical lissencephaly, PAFAH1B1/LIS1 and DCX, as well as the less common TUBA1A are strongly associated with ISS  with a prevalence as high as 80% in children with PAFAH1B1/LIS1 deletions or mutations . Severe mutations of ARX associated with X-linked lissencephaly with abnormal genitalia (XLAG) result in severe early-onset epilepsy that overlaps with EIEE. This pattern can evolve into ISS, but more commonly progresses to severe intractable epilepsy [58, 59]. ARX, DCX, and PAFAH1B1/LIS1 are all expressed in GABAergic interneurons [60–62], and we hypothesize that spasms are linked to deficits in this neuronal cell type, rather than the obvious cortical malformation.
This category also includes TSC, as mutations in TSC1 and TSC2 are associated with cortical tubers, subependymal nodules and other TSC-related brain malformations. Around 40% of patients with TSC may have ISS , although the prevalence varies and precise phenotype-genotype data are not available. Recent molecular and pathology evidence suggests tuber formation may be the result of a combination of germline and somatic TSC1/2 mutations . The presence, location, and type of tubers correlate with ISS in some studies [65, 66]. A strong correlation has been established between TSC, ISS, and the subsequent development of autism [67–69], although the mechanism of this relationship remains unclear. Dysregulation of the mammalian target of rapamycin (mTOR) pathway in TSC is a key event leading to abnormal brain morphogenesis . Given the phenotypic heterogeneity of TSC, ISS may occur when the mTOR pathway is perturbed in a specific neuronal population at a key neurodevelopmental stage.
An association between ISS and certain inborn errors of metabolism has long been recognized. The best known examples include the aminoacidopathies phenylketonuria (PKU) and nonketotic hyperglycinemia (NKH); the organic acidemias methylmalonic acidemia (MMA), proprionic acidemia (PA) and maple-syrup urine disease (MSUD); and a disorder of copper metabolism, Menkes disease [71–77]. Mutations of KCNJ11 cause DEND syndrome (developmental delay, epilepsy, neonatal diabetes), which commonly features ISS and is a rare example of an ion channel gene associated with spasms [78–80]. The genetically-characterized mitochondrial disorders appear to be infrequently associated with ISS, with very few clinical reports available . Mutations in SUCLA2 produce a specific mitochondrial encephalopathy, but only one reported patient had ISS .
Metabolic diseases presumably cause epilepsy by disturbing neuronal energetics or by toxicity from the offending metabolite . However, PKU may have secondary effects on synapse function , while NKH and a few other metabolic disorders are associated with agenesis of the corpus callosum and cortical malformations . It is possible that abnormalities in neuronal energetics, synapse function, and brain morphogenesis may all be interrelated in metabolic patients with ISS, and their symptoms reflect the susceptibility of specific neuronal populations to the offending metabolite at a key point in neurodevelopment. Not surprisingly, ISS are usually not the only epilepsy type associated with these metabolic disorders, and overlap exists with other early infant-onset epilepsies, such as EME.
At least five well-known syndromes of chromosomal copy number variation are predisposing genotypes for ISS. These are deletion 1p36.3, Pallister-Killian syndrome (tetrasomy 12p), duplication of maternal 15q11q13, Miller-Dieker syndrome (deletion 17p13.3), and Down syndrome (trisomy 21) [25, 86–95]. The biological mechanism(s) of ISS in these syndromes are not known, except for Miller-Dieker syndrome in which the deletion of PAFAH1B1/LIS1 causes both abnormal primary glutamatergic and GABAergic interneuron migration [89, 96, 97]. Similarly, deletion of MAGI2 at 7q11.2 may be associated with ISS in Williams syndrome , although there are also reports of ISS in Williams syndrome patients with deletions that do not include MAGI2 [25, 98, 99]. Overall this suggests that copy number changes in key genes, in the right developmental context, influence the occurrence of ISS in these disorders. Further, “multiple hit” genomic events may account for some of the phenotypic variability seen in diseases mediated by copy number variants [100–102].
ISS have occasionally been reported with several other well-characterized predisposing genotypes, often in reports predating molecular characterization of the disorder. For example, we found several reports of ISS in patients with Freeman-Sheldon, Schinzel-Giedieon, Smith-Lemli-Opitz, and Sotos syndrome [103–109]. There are also reports of ISS in patients with mutations in FLNA and ARFGEF2, genetic causes of periventricular nodular heterotopia [110, 111]. However, other seizure types are more common than ISS in most of these disorders, especially patients with heterotopia due to FLNA mutations, making it difficult to argue that these genes are reliably causative.
Several reports suggest an association between ISS and rasopathies (neurofibromatosis type I and cardio-facio-cutaneous syndrome), but not commonly enough to suggest a consistent link [112–114]. A single report each of ISS with Smith-Magenis and Pallister-Hall syndromes (the latter with hypothalamic hamartoma) suggest a rare association [115, 116].
This is a novel category introduced to refer to patients who present with ISS, but who lack features specific for any unifying diagnosis and accordingly have an unknown predisposing genotype. When reviewed at later ages, these children have features suggesting a global developmental disorder such as autistic features and dyskinesias that may not have been manifest at the time of presentation with ISS. Brain imaging studies may be normal, or may have findings that do not define a specific syndrome – such as cortical gray or white matter volume loss, asymmetric ventricles, simplified gyral pattern, hypogenesis of the corpus callosum, polymicrogyria, and/or cerebellar malformations. The subsequent epilepsy course after ISS is variable, with some children developing intractable epilepsy and others few seizures. Due to the spectrum of imaging and neurobehavioral findings in this group, this is not a replacement for the previous “cryptogenic” designation. Rather, we think that all or most of these patients will ultimately be found to have predisposing genotypes, particularly in early CNS developmental genes.
This group contains well-described developmental disorders with unknown genotype but unifying phenotypes that have ISS as a core finding. The two best examples are Aicardi and progressive encephalopathy with edema, hypsarrhythmia and optic atrophy (PEHO) syndromes. Aicardi syndrome consists of agenesis of the corpus callosum, polymicrogyria, chorioretinal lacunes, other more variable anomalies, with ISS being a prominent feature . It affects only females, suggesting a gene located on the X chromosome. PEHO syndrome is less common, appears to overlap with a PEHO-like phenotype, and most resembles a neurodegenerative disorder with marked brain and optic nerve atrophy [118–120].
Several more heterogeneous conditions have also been associated with a high incidence of ISS. Around 7% of patients with focal cortical dysplasia (FCD) may develop ISS , and the cortical abnormalities may be below the resolution of MRI . Interestingly, persistence of immature GABAergic networks appear to play a role in focal cortical dysplasia , and dysfunction in the mTOR signaling pathway is a shared feature between FCD and tuberous sclerosis complex . Finally, isolated hemimegalencephaly (HMEG) is included in this category because without cutaneous or systemic findings it is a clinical entity distinct from the many syndromes that also feature this malformation [125, 126].
As a clear relationship between ISS and inborn errors of metabolism has been established, we also recognize a category of diseases where a biochemical defect is demonstrable, but the genetic etiology remains unclear. Patients with pyridoxine-responsive ISS have been reported , but none with mutations in the one known pyridoxine-dependent epilepsy gene ALDHA7 [128, 129]. Several reports link mitochondrial respiratory chain dysfunction and ISS, but without a clear genetic mutation identified, making it unclear if the mitochondrial dysfunction is primary or secondary [130–133]. Finally, neonatal hypoglycemia has long been offered as a cause of ISS [44, 134], sometimes in the setting of hyperinsulinism . The evaluations reported in these patients are often unclear or incomplete, so it is difficult to make a clear causal connection between the two.
Extrinsic brain injuries such as hypoxic-ischemic encephalopathy (HIE), near-miss sudden infant death syndrome, stroke, and infection are commonly accepted as causes of ISS. These infants rarely undergo further evaluation. A closer survey of the patients reported, however, leads to some interesting observations. One is that brain imaging patterns such as atrophy, calcifications, and white matter hyperintensities previously interpreted as due to ischemic injury or infection can overlap with predisposing genotypes such as CDKL5 and SPTAN1 mutations [15, 22, 136, 137]. Clinicians should therefore be aware of the conditions that may present “HIE or infection look-alikes” on imaging.
Reports remain in which ischemic events were well documented with no other predisposing condition identified, and it would appear that 5–10% of infants with ISS in these series have a presumed ischemic or infectious etiology [138–142]. Our concern is that this literature is retrospective, and may not accurately reflect the prevalence of ISS among children with acquired brain injuries. About 25% of children with perinatal hypoxia will have subsequent cerebral palsy, cognitive impairment, and epilepsy , but ISS do not appear to be a significant seizure type described in this population . Despite the increased survival of very-low birth weight infants at risk for HIE, the rate of ISS has not increased , suggesting that other factors contribute to the development of spasms in this population. Perhaps hypoxia represents a “second hit” in a population made vulnerable to spasms by yet-undiscovered predisposing genotypes. While some may comment that there is no scientific data for such an assertion, we argue that the data are insufficient in either case. The study that prospectively follows a large cohort of patients with perinatal hypoxia and documents the prevalence of ISS has not yet been done.
Similar observations may be made about the many case series documenting infection as the cause of ISS [146–148]. While epilepsy can certainly be a long-term complication of CNS bacterial or viral infection during infancy, ISS as a specific consequence are not captured in outcome data [149–153]. If ISS are rare sequelae of CNS infection, predisposing genotype(s) may exist that lead to the expression of spasms in infants who have experienced CNS infection. It is also possible that ischemic or infectious events at a critical period in perinatal development may selectively damage specific neuronal populations – for example emerging GABAergic interneuron synaptic networks – and result in the later emergence of ISS. Therefore, children with extrinsic injury patterns are prime candidates for further study.
This group of disorders contains syndromic phenotypes with unknown causative genotypes, in which ISS have been mentioned only in occasional clinical reports. In this category are frontal-perisylvian periventricular nodular heterotopia-polymicrogyria (PNH-PMG), hypomelanosis of Ito, macrocephaly polymicrogyria polydactyly hydrocephalus syndrome (MPPH) with the overlapping disorder macrocephaly capillary malformation syndrome (MCAP), and sebaceus nevus syndrome [154–157]. Also noted here are various nonsyndromic CNS neoplasms that have been reported with ISS, including choroid plexus papilloma, hypothalamic hamartoma (also described in Pallister-Killian syndrome), and ganglioglioma [158–160]. The association of these conditions and syndromes with ISS comes from a few patient reports, and requires further study to understand the exact relationship.
Several developmental disorders with epilepsy are not associated with ISS, or a surprisingly small number of patients have been reported, suggesting some biological phenomena that may protect against the development of spasms. These disorders include Rett syndrome due to MECP2 mutations , and given that CDKL5 probably interacts with MECP2 , it is not clear why the epilepsy phenotypes in these two developmental disorders are so different.
The 15q11q13 locus is also intriguing in this regard, as duplications of maternal 15q11q13 are a common cause of both epilepsy – including ISS – and autism, while deletion of maternal 15q11q13 causes Angelman syndrome and paternal deletion causes Prader-Willi syndrome. Very few examples of ISS have been reported in Angelman syndrome patients . ISS has not been reported in Prader-Willi syndrome, though epilepsy is 16 times more prevalent than in the general population .
Dravet syndrome due to SCN1A mutations poses another interesting situation. Infants with Dravet syndrome typically present with fever-induced status epilepticus followed by intractable multifocal epilepsy . Dravet syndrome has a severe behavioral phenotype, as well as an ataxia/movement disorder . Despite the fact that GABAergic interneurons express SCN1A at high levels, and the mouse model of severe myoclonic epilepsy of infancy demonstrates abnormal interneuron excitability, there is only a single patient report of ISS in this population [167–169]. Further inquiry into these exceptions may yield important insights into the biology of ISS.
A biologic classification system for ISS begins with the recognition that this phenotype in many, if not most, disorders results from specific disturbances of gene regulation of brain development and function. The known ISS associated genes can be classified in a biologic model drawing from Gene Ontology (GO) , as well as gene expression data combined with CNS developmental stages. This biologic model can help formulate hypotheses to explain the varied phenotypes seen with ISS, and hopefully identify new treatment targets.
Single genes highly associated with ISS can be classified biologically if data exist regarding the gene expression pattern in specific cell types across the spectrum of CNS development. As these data are incomplete in humans, they must sometimes be extrapolated from animal models, usually the mouse. When these data are combined with the GO molecular function of the gene product, a role for the gene product in neurodevelopment can be described. This allows for classification systems built on biological relationships between genes and gene products.
One of the benefits of a biologic classification system is that hypotheses to explain phenotype differences among patients with mutations in ISS associated genes can be made. It is hoped that grouping by biologic mechanism will also aid identification of future disease-specific therapies. Table 2 summarizes the ISS phenotype subgroups after this biologic classification. The molecular function of each gene highly associated with ISS was determined from the GO database (www.geneontology.org), and the primary literature is referenced for evidence of role(s) in or effect(s) on specific stages of brain development. Information from the mouse gene expression databases at BGEM (www.stjudebgem.org), GENSAT (www.gensat.org), and MGI (www.informatics.jax.org) was accessed as well.
The groups are organized so that Group A comprises transcription factors, Group B comprises genes important in proliferation and cell migration, and later groups reflect the terminal differentiation stages of synapse development. The two best characterized groups are the gene regulatory network of ventral forebrain development and the pre- and post-synaptic protein networks, illustrated in Figure 2.
Group A comprises patients who have abnormalities in transcription factors that are essential for ventral CNS development. These include duplications of FOXG1, mutations of ARX, and deletions of MEF2C. Abnormalities in FOXG1 affect development of major telencephalic structures, including the neocortex, hippocampus, and the lateral and medial ganglionic eminences . Patients have mental retardation with an autistic phenotype and stereotypic movements [18, 172]. ARX is an important regulator of GABAergic interneuron differentiation and migration . More severe mutations in ARX lead to lissencephaly, intractable epilepsy, and a phenotype overlap with Group B1. While MEF2C expression is altered in the absence of ARX , its exact position in the gene regulatory network of forebrain development remains unclear. Both ARX and MEF2C patients can have ISS associated with prominent movement disorders.
Group B1 comprises patients with mutations in DCX, PAFAH1B1/LIS1, and TUBA1A, which encode binding proteins that are expressed throughout much of forebrain development, in both glutamatergic and GABAergic cell types. Synaptogenesis appears to be disrupted by a secondary effect of abnormal neuronogenesis and migration. These genes also play direct roles in axonogenesis and synapse formation. Group B1 patients have severe brain malformations that include lissencephaly and variable cerebellar hypoplasia. Group B2 is related conceptually to Group B1 because genes in both groups are involved in neuronal proliferation and migration. Similar to Group B1 patients, TSC patients also have characteristic brain malformations, and can have severe cognitive impairment, and intractable epilepsy, but the neurobehavioral phenotype in TSC often distinctly involves autism.
Group C1 patients have mutations in genes such as GLDC and PAH that play roles in neuronal and non-neuronal cell metabolism but not directly in brain morphogenesis. Mutations in these genes appear to have secondary effects on synapse function, best documented in PKU. These patients have metabolic diseases that, if controlled, as a general rule (but not always) have less severe cognitive impairment and less epilepsy. Group C2 comprises patients with metabolic disorders such as DEND and Menkes disease where the causative gene also has a role in neurodevelopment and the phenotype is characterized by a severe encephalopathy. The genes that cause these disorders have ion transport and/or catalytic activity and play primary roles in synapse development.
Group D comprises patients with abnormalities of MAGI2, SPTAN1, and STXBP1, which encode proteins with direct roles in synaptic development and function. Group D patients tend to have intractable epilepsy, with the MAGI2 deletion patients so far reported also having Williams syndrome. The recent description of patients with mutations of GRIN1  and GRIN2A  associated with ISS illustrates that further discoveries in synapse development and function likely lay ahead.
ISS have long posed classification difficulties, and may be most accurately conceptualized as a phenotype “at the tip of the iceberg” of a broader group of developmental disorders that overlap with autism, intractable epilepsy, and movement disorders. This review does not claim to resolve all of these issues, but seeks to open a discussion in new directions. Knowledge is accumulating rapidly about the genetic associations of ISS, and ISS appear to be a genetically heterogeneous condition that can result from abnormalities in key brain development pathways. Like autism, each associated gene may account for only 1% of the condition. Taken together, however, these are not “rare causes”, as is currently commonly perceived. Instead, animal models of the genetic syndromes highly associated with ISS will likely prove helpful in improving our understanding of pathogenesis of ISS, and hopefully will lead to new therapies.
We propose the first genetic and biologic classifications of ISS. These classifications are designed to be flexible, and will allow the incorporation of new causes as they are discovered. They will help the clinician better direct diagnostic testing, and provide more accurate prognostic information and genetic counseling. Of equal importance, a biologic classification system for ISS is essential for the development of new disease-specific therapies, and helps to explain the phenotypes of various ISS syndromes.
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.