Abstract
Forced expiratory volume in 1 s is currently the most widely used marker of chronic obstructive pulmonary disease (COPD) severity; however, it is a poor surrogate of the emphysematous component and the underlying pathophysiological mechanism, and therefore new markers are urgently needed. Neutrophil elastase (NE) is likely to play a key pathophysiological role in COPD and the current study explores a marker of NE activity as a potential indicator of COPD disease activity.
Aα-Val360 was measured in 81 subjects with a clinical diagnosis of COPD, both in the stable state and at presentation with an acute exacerbation, and comparisons were made using lung function tests and computed tomography imaging. The relationship of Aα-Val360 with disease progression was also assessed in 40 of the subjects over a 4-yr period.
Baseline Aα-Val360 related to physiological and radiological markers of disease severity, was higher at presentation with an acute exacerbation than in the stable state and (at least partly) related to disease progression over the subsequent 4 yrs.
We demonstrate that Aα-Val360 is a marker of cross-sectional COPD disease severity and possibly disease progression, and represents a new concept of specific biomarkers. This study therefore reports the first in vivo data to support the pathophysiological role of NE in COPD.
Chronic obstructive pulmonary disease (COPD) is a slowly progressive chronic disease characterised by airflow obstruction that is predominantly irreversible. Forced expiratory volume in 1 s (FEV1) is a recognised prognostic indicator commonly used as a clinical end-point in pharmaceutical trials. Furthermore, guidelines suggest the diagnosis of COPD is only made in people with symptoms and airflow obstruction as defined by a ratio of FEV1 to forced vital capacity (FVC) either <0.7 or the lower limit of normal (LLN) [1]. However, although short-term improvements in FEV1 are considered beneficial and often relate to improvements in patient symptoms, this physiological measure has significant flaws. First, the relationship to healthcare status is weak. Secondly, FEV1 is effort dependent and the day-to-day variability may be greater than the progressive decline observed in patients with COPD over many years [2], making it a poor surrogate for early phase II studies of potential disease-modifying agents. Finally, and importantly, it is recognised that COPD is a group of distinct pathological processes and the FEV1 relates poorly to the presence of emphysema and its severity as quantified by computed tomography (CT) densitometry [3]; therefore, a diagnosis made on spirometric grounds alone may be inappropriate [4]. There is therefore an urgent need to assess COPD more comprehensively and develop biomarkers relevant to the individual components that are validated as markers of disease prognosis and hence can be used as early read-outs for phase II clinical trials.
It is increasingly accepted that COPD is an inflammatory disease and, therefore, a number of potential biomarkers have been studied in subjects with COPD; however, few are central to the pathophysiological process and none have been effectively validated. For instance, although C-reactive protein (CRP), a nonspecific marker of inflammation, relates to mortality in people with mild-to-moderate (but not severe) COPD [5], elevated levels of this marker fail to predict a more rapid decline of FEV1 in longitudinal studies [6]. Also, densitometric analysis of CT scans are increasingly considered in clinical trials since lung densitometry is now accepted by the US Food and Drug Administration as a clinically meaningful end-point for the assessment of emphysema progression [7]. However, the methodology has yet to be standardised and a normal range of lung density defined; therefore, CT densitometry cannot yet be considered a validated biomarker and is unlikely to demonstrate the early response required for phase II proof-of-concept studies.
Further progress may be achieved by developing biomarkers of key pathological processes in COPD. Neutrophil elastase (NE) was implicated in the pathogenesis of COPD shortly after the observation of the presence of early-onset emphysema in people with α1-antitrypsin (α1-AT) deficiency [8]. At neutrophil degranulation, the NE concentration greatly exceeds that of its major inhibitor (α1-AT), even in healthy individuals; however, dilution occurs due to diffusion away from the point of release and NE is inhibited rapidly within the neutrophil microenvironment when an equimolar concentration with its inhibitors is reached. There is therefore an area of obligate proteolytic damage, even in healthy subjects, which is exponentially greater in patients with α1-AT deficiency, explaining their susceptibility to disease [9]. Although the role of NE in subjects with COPD unrelated to α1-AT deficiency is less well understood, the quantum proteolytic damage in this group of patients may be enhanced as abnormalities within circulating neutrophils relate to overall mortality [10]. Also, an inverse relationship exists between FEV1 and circulating neutrophil numbers [11], and neutrophils from subjects with COPD exhibit abnormal chemotactic responses [12]: adherence and migration under flow conditions [13], enhanced activity as indicated by the proteolytic destructive potential [14], and production of reactive oxygen species [15] compared to appropriate control subjects. Since NE has the potential to replicate many of the pathological features of COPD, a marker of NE activity may represent an ideal biomarker of COPD and disease activity. However, because of rapid inactivation in vivo, studies are yet to link NE activity conclusively with the pathogenesis of COPD and we have therefore developed an assay based on a pre-inhibition NE-specific fibrinogen cleavage product (Aα-Val360) [16]. The aim of the current study was therefore to explore the role of NE in the pathogenesis of COPD using Aα-Val360 as a marker of disease activity in a well-characterised cohort of subjects with symptoms and/or physiological evidence of COPD.
MATERIAL AND METHODS
Patients aged 40–80 yrs who were smokers or ex-smokers with a diagnosis of COPD (based on symptoms with and without supportive spirometry) who presented in primary care with an acute exacerbation were recruited to the study. All had a history of chronic bronchitis [17] and exertional breathlessness and had the normal PiM α1-AT phenotype (Heredilab, Salt Lake City, UT, USA). Exacerbations were defined by the presence of increased dyspnoea, cough and sputum production (although the volume of the latter had not always increased) and new or increased sputum purulence was a feature of a proportion of patients [18]. Spirometric confirmation of COPD was not used as an entry requirement, to allow the inclusion of participants with a broad range of phenotypes, and therefore subgroup analyses were performed for patients with chronic bronchitis and dyspnoea but with FEV1 and FEV1/FVC within the normal range, and also for subjects with COPD defined spirometrically. Normality was defined as within 1.64 standardised residuals (SRs), as recommended by American Thoracic Society/ European Respiratory Society guidelines, to overcome sex and age differences in lung function, a threshold which is often termed the LLN [19].
Subjects were assessed at the onset of an exacerbation and all provided a spontaneous sample of sputum over a 4-h period after waking. Sputum samples were analysed macroscopically using a standardised colour chart used to classify sputum colour (Bronkotest; Heredilab) and those with mucopurulent or purulent sputum (grade 3–8) were treated with antibiotics (oral cefuroxime), while those with mucoid sputum (grade 0–2) were not. The patients were assessed in detail 8 weeks after the episode (when in the stable clinical state) with full lung function tests and a high-resolution CT (HRCT) scan of the thorax.
Aα-Val360 was measured in plasma samples obtained both at presentation with the exacerbation and when clinically stable, using a highly specific assay as described previously [16]. In addition, comparisons were made with physiological parameters and visual assessment of the HRCT scan. All scans were assessed for the presence or absence of visible emphysema by an experienced thoracic radiologist using established criteria [20]. Plasma CRP was measured by ELISA using commercially available pre-prepared plates and standards (Binding Site; Birmingham, UK). Other inflammatory markers relevant to neutrophilic inflammation including myeloperoxidase (MPO), interleukin (IL)-8, and leukotriene B4 (LTB4) in sol-phase sputum samples and plasma α1-AT/NE complexes were also measured (as described previously [16, 21]) and related to the Aα-Val360 concentration. Aα-Val360 was also measured in plasma samples obtained from 39 healthy controls.
Finally, the patients were reviewed 4 yrs later (where possible) and repeat lung function testing and an HRCT performed at full inspiration using the same General Electric Prospeed Scanner (General Electric Medical Systems, Milwaukee, WI, USA). Densitometric analysis was performed to assess emphysema progression accurately using the voxel index (-950 HU) and the 15th percentile point in both the upper zone (level of aortic arch) and lower zone (level of inferior pulmonary vein).
Post-bronchodilator (salbutamol 400 μg and ipratropium 60 μg via a large-volume spacer) spirometry was performed using a wedge bellows spirometer (Vitalograph, Maids Moreton, UK) and gas transfer measurements by the single-breath carbon monoxide method. The European Community for Steel and Coal reference equations [22] were used to derive predicted values for spirometry, while the reference equation of Cotes [23] was used for transfer coefficient of the lung for carbon monoxide (KCO).
The study was approved by the South Birmingham Research Ethics Committee, Birmingham, UK.
Statistical analysis
Statistical analyses were performed using SPSS 17.0.1 (Chicago, IL, USA) for Windows. Data are presented as mean±se, normality was tested using the Kolmogorov–Smirnov test and statistical significance was taken as p<0.05.
Aα-Val360 and α1-AT/NE complex concentrations were not normally distributed and, therefore, correlations with other inflammatory markers and lung function were performed using Spearman’s rho. Multivariate analysis was performed using linear regression and stepwise entry of independent factors. Mann–Whitney U-tests were used to compare nonparametric values in subjects with and without visible emphysema on HRCT scan, while unpaired t-tests were used for parametric data. Comparisons were made between values obtained at the onset of an exacerbation and stable state data using paired t-tests (for normally distributed data) and paired Wilcoxon rank tests (for nonparametric data).
RESULTS
Initial stable state assessment
81 subjects (36 female and 45 male) with chronic bronchitis and exertional dyspnoea and a broad range of spirometric results (table 1) were included in the study. Of these patients, 58 achieved the spirometric criteria consistent with COPD (with a FEV1/FVC <LLN), while 61 subjects had a FEV1/FVC <0.7 and therefore met the alternative spirometric threshold for COPD.
The stable-state plasma Aα-Val360 concentration related to baseline FEV1 % predicted (r= -0.340, p=0.001) and KCO % pred (-0.246, p=0.013). Multivariate analysis accounting for age, sex, smoking history, height and sputum colour demonstrated that the stable-state Aα-Val360 was an independent predictor of KCO (standardised β coefficient -0.243, R2 change 0.048; p=0.037); however, the independent relationship with FEV1 fell short of conventional levels of significance (standardised β coefficient -0.231, R2 change 0.037; p=0.070). Importantly, in subjects with FEV1/FVC below the normal range, similar relationships were also observed between Aα-Val360 and FEV1 % pred (r= -0.297, p=0.013) and KCO % pred (r= -0.214, p=0.054).
In the full cohort, plasma Aα-Val360 showed a reasonable correlation with plasma α1-AT/NE complex in the stable state (r=0.459, p<0.001) (fig. 1); however, there was no relationship with high-sensitivity CRP (a nonspecific measure of inflammation). Also, the α1-AT/NE complex did not relate to either FEV1 % pred (r= -0.087, p=0.451) or KCO % pred (r= -0.172, p=0.126). Furthermore, the Aα-Val360 was significantly higher (p<0.001) in these subjects with chronic bronchitis and dyspnoea (n=80) with a median Aα-Val360 value of 20.76 nM (interquartile range (IQR) 13.99–25.44 nM) than healthy controls (n=39) with a median value of 3.50 nM (IQR 2.35–5.14 nM).
The relationship between a marker of neutrophil elastase (NE) activity (Aα-Val360) and elastase release (α1-antitrypsin (α1-AT)/NE complex). r=0.459, p<0.001.
Aα-Val360 in subjects with and without visible emphysema on HRCT
In the overall group, the plasma Aα-Val360 concentration was greater (p=0.013) in those with visible emphysema on HRCT (n=43) compared to those without (n=38). Also, subjects with visible emphysema had a significantly lower FEV1 % pred (p=0.014), FEV1/FVC (p<0.001) and KCO % pred (p<0.001) than those without (table 2). However, there was no difference in the plasma α1-AT/NE complex or sputum myeloperoxidase (MPO), leukotriene (LT)B4 and interleukin (IL)-8 (in the 55 subjects able to produce a spontaneous sputum sample for the stable-state assessment) between those with and without emphysema.
Subgroup analysis was performed for subjects with chronic bronchitis and dyspnoea but FEV1 and FEV1/FVC in the normal range as confirmed by SRs (table 2). This demonstrated that Aα-Val360 was also greater in those with visible emphysema (22.88 nM, IQR 14.09–42.17 nM; n=6) on HRCT compared to those without (13.98 nM, IQR 12.31–21.00 nM; n=17), which is similar to that seen in the larger cohort, but the difference just failed to achieve statistical significance in this subgroup (p=0.071). There was no significant difference in the FEV1 % pred in those with visible emphysema compared to those without. However, FEV1/FVC was lower in those with emphysema than those without (0.70±0.01 and 0.77±0.02, respectively; p=0.003). In addition, the KCO % pred was also lower (fig. 2) in those with visible emphysema (93.83±10.43% pred) than those without (114.88±5.09% pred; p=0.030).
Transfer coefficient for carbon monoxide (KCO) in subjects with “normal spirometry”, demonstrating the significantly lower values (p=0.016) in subjects with visible emphysema on high-resolution computed tomography compared to those without. % pred: % predicted.
Analysis of patients having FEV1/FVC <LLN demonstrated that the average Aα-Val360 was also greater in those with visible emphysema (n=37) compared to those without (n=21), but this difference was not significant (p=0.141). There was no difference in the FEV1 % pred or the sputum markers between the two groups of patients, although FEV1/FVC (p=0.06) and KCO % pred (p<0.001) were lower in subjects with visible emphysema on HRCT (table 2).
Aα-Val360 during an acute exacerbation of COPD
The Aα-Val360 was higher at the onset of the exacerbation than in the stable state even when stratified by sputum colour into visibly purulent or nonpurulent episodes (table 3). Furthermore, Aα-Val360 at the onset of the exacerbation was significantly greater (p=0.030) in subjects who presented with purulent sputum compared with those with mucoid sputum. Interestingly, although Aα-Val360 fell in both groups following resolution, the difference between these two groups persisted (p=0.024) (table 3). In addition, the stable-state sputum IL-8 (p<0.001) and plasma α1-AT/NE complex (p=0.036) were higher in those who experienced an exacerbation associated with purulent sputum. However, importantly, there was no longer a difference in the sputum colour or sputum MPO and LTB4 concentrations in the stable state between those who had presented with purulent and those with nonpurulent sputum. In the stable state, no correlation was seen between Aα-Val360 and sputum MPO (r=0.059, p=0.337); however, there remained positive correlations with the key neutrophil chemoattractants LTB4 (r=0.227, p=0.048) and especially IL-8 (r=0.486, p<0.001).
Longitudinal analysis
40 individuals were alive and consented to assessment with full lung function tests and densitometric analysis of HRCT scans at both baseline (stable state) and at follow-up 4 yrs later. Aα-Val360 obtained from subjects at baseline related cross-sectionally to both baseline and follow-up physiological and radiological measures (table 4). However, there was no relationship between α1-AT/NE complex concentration and any physiological or radiological parameter at either baseline or follow-up. Both physiological and radiological markers demonstrated disease progression in the 40 subjects (table 5); however, universal significance was not observed in all parameters. The absence of a significant change may at least partly be explained by the lack of sensitivity of these physiological and radiological tests for detecting disease progression in subjects with COPD and again supports the need for new markers of disease severity and activity.
There was a significant decrease in the FEV1 (p<0.001) and KCO (p<0.001) over the 4-yr period. There was also significant emphysema progression as measured by absolute change in the voxel index (-950 HU) and 15th percentile point in both the upper (p<0.001 and p=0.002, respectively) and lower (p<0.001 and p=0.021) zones. Baseline Aα-Val360 related to both the subsequent decline in KCO % pred (r= -0.406, p=0.008) and progression in lower zone emphysema expressed as change in the voxel index at -950 HU (r=0.306, p=0.027); however the relationship of the baseline Aα-Val360 with change in the lower zone 15th percentile point did not achieve statistical significance (r= -0.212, p=0.095). There was no significant association between Aα-Val360 and decline in FEV1 % pred (r= -0.196, p=0.113), or the upper zone voxel index (r=0.103, p=0.264) and 15th percentile point (r= -0.049, p=0.381). There was no relationship between the α1-AT/NE complex and any measure of decline. Receiver operating characteristic analysis of Aα-Val360 in subjects who demonstrated a decline in KCO % pred over the 4-yr period compared to nondecliners gave an area under the curve of 0.711 (p=0.037). An Aα-Val360 threshold of 11 nM would have a sensitivity and specificity of 91% and 43%, respectively, for the identification of subjects who will demonstrate a decline in KCO % pred over a 4-yr period, while a threshold of 22 nM would have a sensitivity of 33% and specificity of 93%.
In the subgroup analyses of patients with chronic bronchitis and dyspnoea but a baseline FEV1/FVC SR and FEV1 SR within the normal range (n=14), the baseline Aα-Val360 also correlated with subsequent decline in KCO % pred (r= -0.534, p=0.025), but not to spirometric or radiological progression. In those with a baseline FEV1/FVC SR <LLN (n=26), the Aα-Val360 related to radiological progression (change in lower zone voxel index r=0.348, p=0.041), but there were no significant relationships with other measurements.
DISCUSSION
We have developed a unique assay based on an NE-specific fibrinogen degradation product (Aα-Val360), which measures the damaging potential of NE at the point of release from the neutrophil prior to its inhibition by the surrounding protease inhibitors [16]. Previous authors have investigated the use of an alternative fibrinogen degradation product (formed by the cleavage of the fibrinogen α-chain at Aα-21); however, this was not pursued further because of its low specificity for NE and the very short half-life of this smaller fibrinogen fragment [24]. In contrast, the Aα-Val360 is highly specific and shows stability over time, both of which are important features of any biomarker [16].
In the absence of more suitable gold-standard markers, we opted to relate Aα-Val360 to physiological and radiological measures of COPD disease severity since (despite their flaws) they are widely used and reasonably well validated. We did not compare Aα-Val360 with other potential biomarkers since (to date) none have been effectively validated and therefore interpretation of relationships (if any) would be difficult. For example, while urinary desmosine may differ between healthy individuals, smokers with normal lung function and subjects with COPD, it does not correlate with FEV1 and is not influenced by augmentation therapy in α1-AT deficiency [25] (perhaps because it is neither organ- nor disease-specific).
However, we have previously demonstrated that Aα-Val360 relates to FEV1 and demonstrates a response to augmentation therapy in subjects with α1-AT deficiency [16], and in the current study we demonstrate that Aα-Val360 also relates to several specific features of COPD in subjects without α1-AT deficiency. First, it relates cross-sectionally to physiological and radiological markers of current COPD severity. However, this relationship is likely to be complex (and, therefore, consistent with the strength of the observed associations) since a marker of activity may also relate to the process leading to the current disease state (or severity) or to future disease progression. The strongest relationship was observed between Aα-Val360 and severity markers of the emphysematous process (gas transfer and the voxel index) and, in particular, Aα-Val360 was the best independent predictor of gas transfer. These data suggest NE activity is central to the pathogenesis of COPD but is likely to be of greatest relevance in the development of the emphysematous component.
Further confirmation of the relationship between NE activity and the emphysematous process was provided by our observation that Aα-Val360 is greater in subjects with visible emphysema on HRCT compared to those without. Although the difference was less marked in subjects who also had obstructive spirometry, this may be explained by the smaller number and discordance between the severity of the airway obstruction and alveolar destruction within individuals, since the average observed differences were similar to that seen for the whole cohort.
It is increasingly recognised that COPD is a heterogeneous disease, and the current study demonstrated that visible emphysema was present even in six of the 23 subjects with spirometry within the normal range, and those with visible emphysema also had a greater Aα-Val360 and significantly lower gas transfer and lower FEV1/FVC than those without visible emphysema, even though these physiological tests remained either largely or entirely within the normal range. This observation supports the use of a symptomatic diagnosis of COPD (chronic bronchitis and exertional dyspnoea) for inclusion of patients in the current study, since spirometric criteria would have excluded a number of subjects with either “early” disease or an emphysema-predominant phenotype. In those with normal spirometry, the difference in Aα-Val360 between these two subgroups (with and without visible emphysema on HRCT) fell just short of conventional levels of statistical significance; however the absolute difference in Aα-Val360 concentrations mirrored that observed in the entire cohort, suggesting that this is not due to chance alone but rather reflects the smaller number of subjects identified. There is therefore likely to be a subset of patients with an active NE-related disease process yet relatively mild physiological changes that do not meet current spirometric criteria for the diagnosis of COPD, who may be identified by Aα-Val360 and benefit from targeted therapeutic intervention to prevent deterioration to the more classical stages of COPD. Clearly, further studies are required to investigate this possibility in depth.
Secondly, Aα-Val360 is a specific marker of pre-inhibition NE activity, while the α1-AT/NE complex is a marker of total NE release as a result of neutrophil degranulation. We showed that, in general, the Aα-Val360 was related to the plasma α1-AT/NE complex, and that subjects with COPD had higher levels of both Aα-Val360 and α1-AT/NE complex than healthy controls, indicating both greater NE activity and neutrophil enzyme release. However, the differences between Aα-Val360 and other markers of neutrophil activation were emphasised by the absence of any correlation between either MPO (a marker of neutrophil degranulation) or α1-AT/NE complex and the physiological and radiological markers of COPD disease severity (either cross-sectionally or longitudinally), demonstrating that a measure of elastase release alone is a poor surrogate of the enzyme’s proteolytic activity and potential influence on disease progression.
Thirdly, Aα-Val360 also related to exacerbations of COPD, which are episodes known to relate to physiological progression [26]. Aα-Val360 was higher at the onset of an exacerbation than in the stable state 8 weeks later, reflecting greater NE activity which may at least partly impact on subsequent evidence of disease progression. Importantly, the Aα-Val360 was higher not only at the onset of an exacerbation in subjects who experienced an exacerbation associated with purulent (neutrophilic) sputum compared to those with mucoid sputum, but also remained higher in the stable state, supporting the concept that these episodes may mark subjects with a greater likelihood of progression. Furthermore, the higher Aα-Val360 in this group of patients was associated with a higher stable state sputum IL-8 and plasma α1-AT/NE complex concentration, demonstrating a greater ongoing inflammatory process leading to neutrophil recruitment, enzyme release and hence potential tissue damage. Although these data could represent a slower recovery, it is unlikely, since all subjects were seen 2 months after the exacerbation onset, when they were confirmed to be clinically stable. Additionally, the Aα-Val360 related both to the stable state and 4-yr follow-up physiological and radiological measures, and these relationships would be less likely if patients had not been in the stable state at the time of assessment (with further resolution after the study). In particular, there was no evidence of an ongoing bacterial trigger since there was no difference in subjective assessment of the stable-state sputum colour or objective measurement of MPO between the two groups of patients.
Although further studies are required, it is probable that patients who experience an exacerbation associated with purulent sputum have greater elastase activity in general (leading to tissue damage which enhances the subsequent risk of a bacterial infection) and, therefore, experience further exacerbations associated with purulent rather than mucoid sputum. It is also possible that subjects who experience a more severe exacerbation have a greater Aα-Val360 signal even following recovery and hence decline at a greater rate. However, there is currently no accepted inflammatory marker of exacerbation severity in subjects with COPD to confirm this concept, although it is possible that Aα-Val360 itself may fill this role in specific targeted studies.
Finally, there appears to be a relationship between NE activity (measured by Aα-Val360) and COPD disease progression. Current pathophysiological activity is likely to reflect not only preceding but also future disease progression and in the current study we demonstrated baseline Aα-Val360 related to deterioration and subsequent disease severity measured by gas transfer. The data also indicates a relationship with Aα-Val360 and disease progression measured by CT densitometry, consistent with cause and effect. Nevertheless, the exact contribution to the pathophysiology of COPD in general and emphysema in particular will require a much larger prospective trial in highly characterised patients. However, Aα-Val360 did not relate to spirometric decline, suggesting that tissue damage reflected by Aα-Val360 is more indicative of the emphysematous process, not only in the presence of established COPD but importantly even in those with emphysema but without airflow obstruction. Clearly larger studies, especially in this latter group, are now indicated, including data on longitudinal progression.
In summary, the current study reports the first in vivo data in human subjects which supports the role of NE in the pathophysiology of COPD without α1-AT deficiency. Furthermore, when considered in combination with previous circumstantial data, the current study suggests that NE may represent (at least part of) a final common pathway leading to tissue destruction in this disease process. Aα-Val360 is the first specific biomarker of pre-inhibition NE activity, which relates to cross-sectional measures of disease severity and exacerbations and appears to relate to disease progression in subjects with COPD. Although further work in a larger cohort of patients is required, particularly to explore the relationship with longitudinal physiological markers of disease progression in subjects at risk, Aα-Val360 thus represents a new concept of specific biomarkers that may be central to the pathophysiology of COPD.
Acknowledgments
A patent covering Aα-Val360 (US patent No. 6124107) was held by Merck, Rahway, NJ, USA, and we would like to thank them for providing us with the assay.
Footnotes
Support Statement
The study was partly funded by an unrestricted grant from GlaxoSmithKline.
Statement of Interest
A statement of interest for the study itself can be found at www.erj.ersjournals.com/site/misc/statements.xhtml
- Received November 13, 2011.
- Accepted March 21, 2012.
- ©ERS 2013
REFERENCES
