|Home | About | Journals | Submit | Contact Us | Français|
Axonopathy is a pronounced attribute of many neurodegenerative diseases. In Alzheimer’s Disease (AD), axonal swellings and degeneration are prevalent and may contribute to the symptoms of AD senile dementia. Current limitations in identifying the contribution of axonal damage to AD include the inability to detect when this damage occurs in relation to other identifiers of AD because of the invasiveness of existing methods. To overcome this, we further developed the MRI methodology Manganese Enhanced MRI (MEMRI) to assess in vivo axonal transport rates. Prior to amyloid-beta (Aβ) deposition, the axonal transport rates in the Tg2576 mouse model of AD were normal. As Aβlevels increased and before plaque formation, we observed a significant decrease in axonal transport rates of the Tg2576 mice compared to controls. After plaque formation, the decline in the transport rate in the Tg2576 mice became even more pronounced. These data indicate that in vivo axonal transport rates decrease prior to plaque formation in the Tg2576 mouse model of AD.
Alzheimer’s disease (AD) is an age-related, neurodegenerative disease that is among the leading causes of dementia, afflicting 1% of people under the age of 60 to more than 40% of people over the age of 85 (Lindeboom and Weinstein 2004). Typical symptoms of AD are memory loss and a progressive decline of cognitive abilities (Lindeboom and Weinstein 2004). The pathological characteristics of AD are the presence of intracellular neurofibrillary tangles (NFTs) and the extracellular deposition of amyloid-beta (Aβ) aggregates, known as plaques (Gotz, Schild et al. 2004). NFTs are comprised of abnormally hyperphosphorylated tau protein, a microtubule-associated protein found primarily in axons that can block the axon when aggregated. Aβ plaques result from the sequential cleavage of the amyloid precursor protein (APP). In familial AD, mutations in APP can result in an increased amount of Aβ production leading to aggregation and formation of plaques.
Another attribute of neurodegenerative diseases includes a progressive neuronal deterioration resulting in abnormal neuronal structure and functioning, which ultimately leads to the death of the neuron. A common feature of many neurodegenerative diseases is a progressive perturbation in normal axonal transport rates (Jablonka, Wiese et al. 2004). In vitro data indicates that axonal transport deficits also occur in animal models of AD. For example, in different animal models of AD, the excessive accumulation of proteins such as the tau and Aβ appear to cause a slowing of fast axonal transport, and synaptic and neuronal loss as observed in cultured rodent neurons as well as in Drosophila (Buxbaum, Thinakaran et al. 1998; Morfini, Pigino et al. 2002).
Recently, it was shown that the addition of glutamate or Aβ in cultured rat hippocampal neurons, rapidly inhibited fast axonal transport (Hiruma, Katakura et al. 2003). It was proposed that the glutamate mechanism was through activation of NMDA and AMPA receptors and Ca2+ influx whereas the addition of Aβ was through actin polymerization and aggregation (Hiruma, Katakura et al. 2003). Alternatively, it is thought that abnormal phosphorylation of the microtubule associated protein, tau, could be responsible for alterations in axonal transport (Morfini, Pigino et al. 2002). Increased axonopathy also occurs in animals with normal tau and double transgenic mutations in APP and Presenillin1—a genetic indicator of increased risk of AD (Suh and Checler 2002; Wirths, Weis et al. 2006). Many of the transgenic AD mouse models also exhibit axonal swelling that likely interferes with normal axonal transport (Stokin, Lillo et al. 2005; Wirths, Weis et al. 2006). A recent study also indicates that axonal swellings occur early in the Tg2576 AD model (Stokin, Lillo et al. 2005; Wirths, Weis et al. 2006). However from these combined data, it is not clear if hyper-phosphorylation of tau or neurofibrillary tangle accumulations are causal or consequential of alterations in axonal transport rates (Morfini, Pigino et al. 2002). Currently, there are not any methods available to measure in vivo axonal transport.
The Tg2576 mouse model of AD overexpresses a mutated form of Amyloid Precursor Protein (APP) APPK670N,M671L under the control of the hamster prion protein promoter and exhibit accumulation of Aβ and eventual plaque formation as aging ensues (Hsiao, Chapman et al. 1996). At the age of 6–7 months, Tg2576 mice begin to accumulate insoluble forms of Aβ 42 and Aβ 40, which aggregate to form detectable plaques starting at the age of 10 months (Kawarabayashi, Younkin et al. 2001; King and Arendash 2002; Otth, Concha et al. 2002; Puig, Gomez-Isla et al. 2004). In this study, we focus on the role Aβ has on in vivo axonal transport rates using the Tg2576 mouse model of AD.
We utilized the magnetic resonance imaging (MRI) contrast agent manganese ion (Mn2+) in conjunction with a dynamic T1-weighted MRI sequences to assess the transport rates of Mn2+ ion. Mn2+ has been used as a contrast agent in MRI as Mn2+ is a calcium analogue and is also paramagnetic (Mendonca-Dias, Gaggelli et al. 1983; Burnett, Goldstein et al. 1984; Geraldes, Sherry et al. 1986; Cory, Schwartzentruber et al. 1987; Fornasiero, Bellen et al. 1987). In MRI, Mn2+ enhancement has been effective for trans-synaptic neuronal tract tracing enabling the in vivo mapping of neuronal tracts (Pautler, Silva et al. 1998; Pautler and Koretsky 2002; Saleem et al. 2002; Pautler, Mongeau et al. 2003; Chuang and Koretsky 2006).
In this study, we verified the use of Manganese Enhanced Magnetic Resonance Imaging (MEMRI) from simple anatomical tract tracings to dynamic tracings reflective of axonal transport. We utilize this novel technological development to assess in vivo axonal transport rates in Tg2576 mice before Aβ levels increase, during Aβ accumulation and after plaque formation. By focusing on the olfactory system of the mouse we are able to access a well-defined white matter projection with minimal invasiveness to the animal. Additionally, it has been documented that the olfactory system is targeted early in the time-course of AD making it an ideal target for monitoring disease progression (Attems, Lintner et al. 2005; Jellinger and Attems 2005).
Mn2+ is transported along microtubules via fast axonal transport making it possible to utilize MEMRI to dynamically measure the rates of Mn2+ transport, which is reflective of fast axonal transport rates, within the same animal before and during disease progression. In this study, we extend and verify the use of MEMRI from anatomic tracings to dynamic tracings reflective of axonal transport. We utilize this novel technological development to assess in vivo axonal transport rates in Tg2576 mice before Aβ accumulation, during Aβ accumulation and post plaque formation.
Baseline Mn2+ transport, temperature, and colchicine experiments were conducted utilizing eight-week-old C57/Bl6 inbred mice obtained from Baylor College of Medicine mouse facility (30 mice). Tg2576 mice overexpressing human SwAPP695(K670N/M671L), the Swedish mutation, were used to compare rates of axonal transport in normal and abnormal aging animals (Hsiao, Chapman et al. 1996). Male Tg2576 mice were crossed with C57Bl6/SJL F1 females to obtain Tg2576 overexpressing mice and littermate controls. A total of 28 Tg2576 animals were used for these studies.
Animals were anesthetized with ketamine/xylazine (0.75 g/ml)/(0.5 mg/ml) in phosphate buffered saline, 0.1 ml per 10 g body weight. 10 minutes after administration of the anesthesia, a nasal lavage of 4 μl of 0.75 g/ml MnCl2 dissolved in nanopure water was administered and animals were allowed to recover. One hour post lavage, animals were induced on 5% isoflurane and maintained with 2% isoflurane in 100% O2. Animals were placed in a prone position on a custom-built head holder with adjustable nose bar and secure ear pins. The respiratory rate was monitored with a pressure pad placed under the animal. Temperature was monitored by use of a rectal probe and maintained at 37ºC using both a water blanket and an air heating system (SA Instruments, Inc). Respiratory rates and temperature were monitored using the Model 1025 Small Animal Monitoring and Gating System software (SA Instruments, Inc).
Images were acquired utilizing a 9.4T, Bruker Avance Biospec Spectrometer, 21 cm bore horizontal scanner with 35 mm volume resonator (Bruker BioSpin, Billerica, MA). The imaging parameters to acquire olfactory multi-spin/multi-echo MEMRI images were as follows: TR = 500 ms; TE = 10.2 ms; FOV = 3.0 cm; slice thickness = 1mm; matrix = 128 x 128; NEX = 2; number of cycles = 15; each cycle took approximately 2 min 8 sec to acquire using Paravision software (Bruker BioSpin, Billerica, MA). Core temperature was maintained at 37ºC during scanning.
For the temperature experiments the animals were maintained at 37.0ºC SEM 0.1 for 40 minutes. (Experiment temperature was maintained for eight minutes prior to recording for 32 minutes for a total of 40 minutes) Next, the heating system (Small Animal Instruments) was adjusted to allow the animal to cool to 30.3ºC SEM 0.3 for 40 minutes and then the animal was returned to 37.0ºC SEM 0.1 for another 40 min (Figure 3).
Colchicine (Sigma C9754) (1 mg/kg) & vehicle controls (0.9% saline) were administered by lavage 24 hr prior to Mn2+ administration. Data was acquired using Paravision software (Bruker BioSpin).
The region of interest (ROI) was identified and placed on an axial slice 1 mm in front of the posterior of the olfactory bulb (OB). The ROI measured 0.23 X 0.23 mm and was vertically centered on the dorsal olfactory neuronal layer (ONL) (Paxinos and Franklin 2001). The ROI was determined by measuring the length of the olfactory bulb, locating the midpoint of this line, then extending this midpoint out to the ONL using a 90º angle. The pixel closest to the midpoint within 5% error was established as the ROI for all images (Paxinos and Franklin 2001). This method of defining the region of interest ensures that the widest point of the olfactory neuronal layer is considered in the measurement. Due to a chemical shift artifact on the left olfactory bulb the ROI was localized only to the right olfactory bulb (Sbarbati, Calderan et al. 2002). This ROI was copied for each cycle and each ROI value normalized to the unaffected muscle of the same slice. The small region of interest collected in this study is representative of a single fascicle of axons projecting into the olfactory neuronal layer (Akins and Greer 2006). Mouse olfactory bulb glomeruli, the main target of an olfactory neuron fascicle, range in size from 80–150 μm (Paxinos 1995; Monnier, Bahjaoui-Bouhaddi et al. 1999). The rationale for utilizing a smaller ROI was to focus upon the fascicle projecting onto a single glomerulus to minimize any variation introduced from fascicles projecting from other regions within the olfactory epithelium. Statistical analyses: linear regression, two-tailed t-tests, and one-way ANOVA were performed with Prism (GraphPad Software, Inc).
Brain sections were dissected from 2 and 8 month old mice and immediately frozen on dry ice. The samples were homogenized in lysis buffer (1% NP40, 9.975% glycerol, 0.15M NaCl, 0.5M Tris HCl pH7.5 and protease inhibitor (Sigma)). Protein concentrations were obtained using DC Protein Assay (BIO-RAD). 20 μg of protein were loaded per well and resolved by 10% SDS-PAGE. Protein was transferred onto a Nitrocellulose membrane (BIO-RAD). The blots were blocked with 5% milk in Tris-buffered saline with 0.1% Tween-20 and incubated with 6E10 (Signet, 1:1000) or APP-C (in house, 1:1000) primary antibody. Subsequently the blots were hybridized with goat-anti-mouse or goat-anti-rabbit HRS-conjugated secondary antibody, respectively (Vector Laboratories). Bands were visualized using ECL- Western blotting detection reagents (Amersham Biosciences).
Olfactory bulbs were obtained from a 7.5 month and 12 month old Tg2576 animals and a 12 month old littermate control. Mice were perfused with 4% PFA and plaques were detected using 6E10 monoclonal antibody (Signet) in conjunction with R.T.U. Vectorstain kit (Vector Laboratories).
Our previous work demonstrated that olfactory receptor neurons (ORNs) uptake Mn2+ and transport the ion to the olfactory neuronal layer (ONL) of the olfactory bulb (Pautler, Silva et al. 1998). To establish the rates of normal Mn2+ transport along this projection, control mice were lavaged intra-nasally with a solution of MnCl2 and then imaged one hour post administration utilizing our established paradigm (Pautler, Silva et al. 1998). A series of T1-weighted MRI images were acquired over the course of one hour in Mn2+ treated and control mice. MRI signal intensities in each data set were measured in the ONL and then normalized to unaltered muscle within the same slice. The location of this slice was always 1 mm anterior from the posterior edge of the olfactory bulb.
Figure 1A shows an example of the region of interest (ROI) located within the ONL as well as an example of the unaltered muscle utilized for baseline normalization. The inset of Figure 1B displays how the ROI on the olfactory bulb was selected using the length of the olfactory bulb and then finding the pixel on the ONL at the midpoint of the olfactory bulb. This pixel is representative of one fascicle within the ONL. Additionally, we tested motion correction software applied post image acquisition (Amira) and compared these data to data acquired without the motion correction. With this comparison so little motion was measured during the scans that motion correction software was determined to be unnecessary (data not shown).
The normalized signal intensity was then plotted versus time as shown in Figure 2. A linear regression curve best fit the data (Figure 2A). We verified that the gradual increase in signal intensity was due to the administered Mn2+ by comparing it to the slope of the control group not exposed to Mn2+ (Figure 2A). From Figure 2 it is evident that the slopes of Mn2+ treated and control animals are significantly different and that administration of Mn2+ results in signal intensity enhancement overtime. Figures 2B-I provide a qualitative verification of the enhancement measured in figure 2A. The slope of this curve is reflective of the rate of axonally transported Mn2+ (Figure 2).
After establishing the baseline rate of Mn2+ transport, two important regulators of axonal transport, body temperature and microtubule stability were tested to validate our hypothesis that the Mn2+ transport rates are reflective of axonal transport (Cosens, Thacker et al. 1976; Bamburg, Bray et al. 1986; Cancalon 1988).
To evaluate the dependence on temperature, three separate transport rates were recorded from one animal at three different temperatures in sequence while the animal was still in the magnet: 37.0ºC SEM 0.1, next 30.3ºC SEM 0.3 and then returned to 37.0ºC SEM 0.1. The MEMRI data for each temperature set was recorded and analyzed as for the previously described experiments. Each of the 15 scanning cycles lasted ~2 min for a total of approximately 32 minutes at each temperature. The difference in ΔSI/Time that occurred between normal physiologic temperature and the substantially reduced temperature was significant (Figure 3. One-way ANOVA 37.0ºC vs. 30.3ºC p-value < 0.001). Once the body temperature returned to 37.0ºC, the increase in Mn2+ signal intensity was restored. This result demonstrates that ΔSI/Time due to Mn2+ transport is temperature dependent
The dependence of Mn2+ transport upon microtubule integrity has previously been described (Hastie 1991; Han, Malak et al. 1998). Microtubules are the primary cytoskeletal element used in fast axonal transport and are composed of α- and β-tubulin dimers. Colchicine binds to tubulin prior to polymerization, thereby inhibiting microtubule assembly and reducing axonal transport (Hastie 1991; Han, Malak et al. 1998).
We confirmed the contribution of microtubule-based transport to our dynamic MEMRI measurement. Two groups were assessed: the first group was treated with a dose of 1 mg/kg of colchicine and the second group with saline 24 hours prior to the administration of Mn2+. Our results indicate that colchicine significantly reduced the ΔSI/Time when compared to control mice treated with saline (Figure 4). It is also important to note that these results demonstrate that colchicine did not stop Mn2+ from accumulating within the turbinates comprised primarily of axons. This data, in conjunction with the temperature studies, confirm that Mn2+ transport is dependent upon axonal transport activity and machinery as colchicine is known to bind tightly to tubulin and prevent microtubule assembly.
We chose to conduct our studies in the olfactory system due to the presence of early olfactory involvement observed in AD patients (Solomon 1994; Thompson, Knee et al. 1998; Attems, Lintner et al. 2005), and because of the relative ease of introducing Mn2+ to the olfactory receptor neurons by means of nasal lavage. Therefore, it was important to confirm that the mutant form of APP was present in the olfactory bulbs of the Tg2576 mice.
The presence of mutant APP in the olfactory bulbs of Tg2576 mice was confirmed using Western Blot analyses (Figure 5A) indicating that alterations in axonal transport due to the APP mutation should be detectable utilizing MEMRI in the olfactory bulb. The axonal transport rates in the Tg2576 and littermate wildtype mice were then assessed at ages 3–4, 7–8 and 11–14 months using the established MEMRI parameters (Figure 5B).
The youngest age group, 3–4 months, demonstrated no difference in the transport rate compared to controls (Figure 5B). However, Tg2576 mice in the 7–8 month old group demonstrated a significant difference in the transport rate compared to wildtype mice of the same age (p-value <0.05) (Figure 5B). Tg2576 animals at the third and oldest time point evaluated, 11–14 months, demonstrated an even more dramatic decrease in the transport rate compared to the wildtype group (**p-value < 0.015) (Figure 5B). It is important to note that the reported differences at 7–8 months took place during the expected increase in insoluble Aβ but prior to visible plaque formation that starts at about 10 months of age (Kawarabayashi, Younkin et al. 2001). This finding was confirmed with histology of the scanned animals that showed the olfactory bulbs of the Tg2576 animals were free of plaques at 7–8 months of age (Figure 6B). Plaques were present at 12 months of age in the Tg2576 and not present in the control animal (Figure 6).
The role of axonal transport in neurodegenerative diseases remains poorly understood. While there are indications that axonal transport deficits contribute to neurodegeneration, in vivo detection has been very difficult and invasive. The technique defined by this paper extends the use of MEMRI to quantify and compare in vivo axonal transport in normal and transgenic mouse models. This novel methodology opens up the possibility for in vivo, longitudinal studies that could contribute unique information in mouse models of neurodegeneration. Signal intensity data from thirty minutes of recording in the ONL allows for extrapolation of the rate (ΔSI/Time) at which Mn2+ is transported from the olfactory epithelium to the ONL. The difference in signal intensity produced by Mn2+ is also visible through qualitative assessment, as seen in Figures 2B-E. The data confirm that the nasal lavage and data acquisition are reliable and reproducible and that without Mn2+ the SI does not change.
It has been noted previously that physiological temperatures (36.9ºC in mice) are important for normal axonal transport (Cosens, Thacker et al. 1976; Van der Linden, Van Meir et al. 2004). To determine the effect of temperature on Mn2+ transport, baseline transport rates at 37.0ºC were compared to transport rates at 30.3ºC (Figure 3). The drop in temperature corresponded with a reduction in rate of Mn2+ transport indicating that Mn2+ is transported in a temperature dependent manner. Temperature dependence was further established by finding that Mn2+ transport rates returned to normal when the body temperature of the animal was reestablished at 37.0ºC.
Another component known to be necessary for normal axonal transport is the microtubule network. Although the role of microtubules in Mn2+ transport in anatomical MEMRI data has already been demonstrated (Sloot and Gramsbergen 1994; Pautler, Silva et al. 1998), we evaluated the dependence of dynamic Mn2+ transport upon microtubule integrity. The microtubule disruptor colchicine was utilized because it is known to bind tightly to tubulin and prevent the polymerization of microtubules (Hastie 1991; Han, Malak et al. 1998). By preventing assembly of microtubules, the ability of motor proteins to bind and transport cargo is also prevented. Colchicine has the added benefit of having been applied through nasal lavage and intracerebral injections in previous studies (Sloot and Gramsbergen 1994; Pautler, Silva et al. 1998). Here, the dose 1 mg/kg, a dose less than one-half that of previous studies, was used. Colchicine and vehicle controls (0.9% saline) were tested and results show that the colchicine significantly and dramatically blocked the rate of Mn2+ transport in the ONL. Data acquired using MEMRI confirmed Mn2+ enhancement in the olfactory turbinates at the base of the olfactory bulb, indicating continued Mn2+ influx into the cells and axonal areas following colchicine treatment. We paid particular attention to this because microtubule alterations are capable of changing calcium influx into the cell. The data in figure 4A-D indicate that Mn2+ influx was not hampered by the colchicine. The data shown here verify that Mn2+ transport is dependent upon microtubule-based axonal transport and that the MEMRI measurement is a quantifiable indicator of axonal transport.
Having established that Mn2+ transport measured by MEMRI is reflective of axonal transport, the axonal transport rates of Tg2576 animals and wildtype littermate animals were evaluated at ages 3–4, 7–8, and 11–14 months of age utilizing MEMRI. Significant differences in axonal transport between wildtype and Tg2576 mice were apparent in the 7–8 months aged group, at which point the axonal transport rate for APP overexpressing mice was 48% less than controls. When older animals, 11–14 months, were analyzed they were found to have an even more pronounced deficit, 82% less. We verified on scanned 7–8 month old animals that their olfactory bulbs were indeed negative for amyloid plaques compared to the olfactory of 12 month old Tg2576 animals with positive amyloid plaque accumulation in the olfactory bulb.
Studies analyzing Aβ accumulation, the cleavage product of APP, in these animals demonstrate that insoluble Aβ appears at 6–9 months (Kawarabayashi, Younkin et al. 2001). The definite biochemical changes take place by the 10th month leaving minimal Aβ deposition followed by diffuse plaques that are visible by 12 months (Kawarabayashi, Younkin et al. 2001). Combined with the present study, this evidence suggests that the axonal transport deficits are occurring in conjunction with the accumulation of insoluble Aβ and prior to Aβ plaque formation in this particular AD model.
The cause of the differences in axonal transport rates of the Tg2576 mouse model at different ages is unclear. One possibility could be that the calcium influx is decreased resulting in less Mn2+ available to be transported down the axon. However, in this mouse model the calcium influx increases with age instead of decreasing suggesting that Mn2+ entry into the cell is equally unimpeded as our images show (Xie 2004). It has also been reported that the cytoskeletal proteins α and β tubulin are genetically up-regulated in the Tg2576 animal (Reddy, McWeeney et al. 2004). This could result in abnormal microtubule function and further strengthens our data that axonal transport decreases abnormally compared to controls.
Magnetic resonance imaging (MRI) is one of the best imaging methodologies for studying soft tissues and monitoring biological processes as they occur in vivo. There are currently many attempts to identify and study AD animal models through MRI imaging. Numerous MRI studies have targeted the assessment of plaque formation with and without the use of endogenously applied contrast agents (Redwine, Kosofsky et al. 2003; Helpern, Jensen et al. 2004; Helpern, Lee et al. 2004; Jack, Garwood et al. 2004; Lee, Falangola et al. 2004; Song, Kim et al. 2004; Rohner, Staab et al. 2005; Sun, Song et al. 2005; Sykova, Vorisek et al. 2005; Vanhoutte, Dewachter et al. 2005). For example, MRI has been used to identify individual plaques in the Tg2576/Presinilin1 double mutant, a mouse model of AD (Jack, Garwood et al. 2004). Many studies have evaluated changes in anatomy, such as the high resolution, in vitro MRI study on fixed brains of transgenic mice containing the PDAPP mutation. The data from this in vitro study showed dentate gyrus shrinkage prior to the appearance of plaques (Redwine, Kosofsky et al. 2003). Other MRI studies include the assessment of alteration in the diffusion of water in the gray and white matter of mouse models of AD. This work demonstrated alterations in the diffusion of water in the Tg2576 mouse brain after plaque formation (Sun, Song et al. 2005). With the exception of the in vitro studies by Redwine et al, all MRI studies to date on AD mouse models have demonstrated alterations post plaque formation. Here we extend the use Manganese Enhanced MRI (MEMRI) to assess in vivo axonal transport, and therefore neuronal function, in the Tg2576 AD mouse model.
The data reported here establishes that functional deficits in axons begin prior to the appearance of Aβ plaques. This is supported by an in vitro study that found axonal swellings starting at four months in axons from mice also carrying the Swedish AD mutation (Stokin, Lillo et al. 2005). This finding implicates that axonal swellings do not form in response to amyloid deposition, but that the swellings may instigate the accumulation of Aβ similar to that seen in acute neuronal injury.
The evidence presented here indicates that MEMRI is a very promising tool for detecting early signs of abnormal physiological deficits. Using MEMRI we were able to identify and quantify the difference in axonal transport rates in vivo in the Tg2576 mouse model of AD during the progression of the AD phenotype. Data from the Tg2576 mouse indicates that deficits coincide with the timing of previously reported biochemical changes, but prior to histologically visible plaques. The ability to measure axonal transport rates in vivo opens up many new and exciting opportunities for the characterization of disease states as well as assessing the efficacy of diagnosis and therapeutic intervention.
The authors are grateful to Faridis Serrano, Mitch Deshazer, Lingyun Hu, Yingying Sun, X. Chen and N. Aithmitti for their assistance in the preparation of this manuscript. Funding for this work was provided by NIH/NINDS grant 1R21NS051274–01 [RGP], the Dana Foundation [RGP], RO1AG20670 [HZ] and RO1NS40039 [HZ].
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.