From Biomarker Discovery to Clinical Evaluation for Early Diagnosis of Lung Surgery-Induced Injury

Lung cancer is one of the most common cancers in the world (Chiang et al., 2010; Landis et al., 1998). Surgical removal of the tumor mass offers the best chance for a cure in patients with non-small-cell lung cancer. A tumor in stages I (confined to the lung without nodal or distant metastasis), II (involvement of only lymph nodes within the lung), and IIIA (involvement of nodes on the same side as the tumor) is considered potentially resectable for cure (Martini et al., 1995). Based on tumor size and location, lung surgery is mainly divided into three types: wedge resection (removal of a small area in one lobe of either right or left lung), lobectomy (removal of one lobe from a right or left lung), and pneumonectomy (removal of an entire right or left lung). The mortality rate is much higher after pneumonectomy (61%) than lobectomy (35%) (Gunluoglu et al., 2011). Among the post-surgical factors, aberrant local inflammation and abnormal fluid drainage are the most common for inducing pulmonary edema. Excessive accumulation of fluid in the alveoli causes lung injury and hinders functional recovery. A severe form of acute lung injury results in acute respiratory distress syndrome. Sudden and life-threatening lung failure is the most detrimental factor in postsurgical mortality (Jordan et al., 2000). With the progression from lung injury to acute respiratory distress syndrome, proinflammatory cytokines are increased, such as interleukin-1┚ (Donnelly et al., 1996; Geiser et al., 2001) and tumor necrosis factor-┙ (Tremblay et al., 2002). However, the production of vascular endothelial growth factor is reduced in the early stage but not altered in the late stage (Medfor & Millar, 2006). Interleukin-1┚ and tumor necrosis factor┙ are further elevated in the sustained phase (Bhatia & Moochhala, 2004). The increases in proinflammatory cytokines are not correlated with injury-induced mortality (Donnelly et al., 1996). Corticosteroid which alters host inflammatory responses do not show beneficial effects in the early stage of acute respiratory distress syndrome (Kollef et al., 1995). Current reviews suggest that activation of inflammation-independent pathways in the early stage and inflammation-dependent pathways in the late stage contribute to the development of acute respiratory distress syndrome (Spragg et al., 2010; Bhatia & Moochhala, 2004).


Introduction
Lung cancer is one of the most common cancers in the world (Chiang et al., 2010;Landis et al., 1998). Surgical removal of the tumor mass offers the best chance for a cure in patients with non-small-cell lung cancer. A tumor in stages I (confined to the lung without nodal or distant metastasis), II (involvement of only lymph nodes within the lung), and IIIA (involvement of nodes on the same side as the tumor) is considered potentially resectable for cure (Martini et al., 1995). Based on tumor size and location, lung surgery is mainly divided into three types: wedge resection (removal of a small area in one lobe of either right or left lung), lobectomy (removal of one lobe from a right or left lung), and pneumonectomy (removal of an entire right or left lung). The mortality rate is much higher after pneumonectomy (61%) than lobectomy (35%) (Gunluoglu et al., 2011). Among the post-surgical factors, aberrant local inflammation and abnormal fluid drainage are the most common for inducing pulmonary edema. Excessive accumulation of fluid in the alveoli causes lung injury and hinders functional recovery. A severe form of acute lung injury results in acute respiratory distress syndrome. Sudden and life-threatening lung failure is the most detrimental factor in postsurgical mortality (Jordan et al., 2000). With the progression from lung injury to acute respiratory distress syndrome, proinflammatory cytokines are increased, such as interleukin-1 (Donnelly et al., 1996;Geiser et al., 2001) and tumor necrosis factor- (Tremblay et al., 2002). However, the production of vascular endothelial growth factor is reduced in the early stage but not altered in the late stage (Medfor & Millar, 2006). Interleukin-1 and tumor necrosis factorare further elevated in the sustained phase (Bhatia & Moochhala, 2004). The increases in proinflammatory cytokines are not correlated with injury-induced mortality (Donnelly et al., 1996). Corticosteroid which alters host inflammatory responses do not show beneficial effects in the early stage of acute respiratory distress syndrome (Kollef et al., 1995). Current reviews suggest that activation of inflammation-independent pathways in the early stage and inflammation-dependent pathways in the late stage contribute to the development of acute respiratory distress syndrome (Spragg et al., 2010;Bhatia & Moochhala, 2004).

Proteomics -Human Diseases and Protein Functions 34
To effectively reduce post-surgical mortality, early detection of acute respiratory distress syndrome may provide in-depth information for the design of management plans, including non-pharmacological therapies (Villar et al., 2011).

Proteomic analysis of bronchoalveolar lavage fluid in biomarker studies
To effectively identify the biomarkers of various lung diseases, bronchoalveolar lavage fluid from the lower airways and alveoli is collected for genomic or cytological analysis of cellular components (Meyer, 2007). This lung-specific fluid can be used for protein analysis. Identification of the non-cellular components in bronchoalveolar lavage fluid may provide valuable data for the early detection of acute lung injury.

Current advances in proteomic analysis of bronchoalveolar lavage fluid
In the past decades, over 100 human proteins or protein isoforms have been identified in bronchoalveolar lavage fluid from patients with various lung diseases (Wattiez et al., 1999;Lenz et al., 1993;Sadaghdar et al., 1992;Sabounchi-Schütt et al., 2001;Vesterberg et al., 2001). The major challenge today is to identify the lead proteins and validate the potential biomarkers (Turtoi et al., 2011a). To overcome this difficulty, integration of clinical studies with proteomic analysis of bronchoalveolar fluid is a potential solution (Turtoi et al., 2011b).

Sampling concern in proteomic analysis of bronchoalveolar lavage fluid
In clinical proteomics, the most difficult challenge before sample analysis is patient selection and sample collection (Apweiler et al., 2009). In the case of lung cancer patients, the major concern is to collect bronchoalveolar lavage fluid from those who may develop post-surgical lung edema. Although both bronchoalveolar lavage fluid and bronchial washings are collected using similar procedures, the former is collected from terminal alveoli after instilling more than 140 ml of sterile saline and the latter is collected from major airways after instilling less than 140 ml of saline. Because of the concern that excessive fluid accumulation may cause the complication of lung edema, bronchial washing is a better choice for conducting clinical proteomics.

Technical limitations in proteomic analysis of bronchoalveolar lavage fluid
In conventional proteomic analysis, two-dimensional gel electrophoresis provides good protein separation. However, it restricts the discovery of proteins with extreme biochemical properties such as size, isoelectric point, and solubility (Rabilloud, 2002). In comparison, one-dimensional gel electrophoresis provides easy comparison of banding patterns in protein profiling but is less efficient in protein separation. Moreover, the high salt concentration in the bronchoalveolar lavage fluid interferes with the resolution of protein separation to a lesser extent in one-dimensional gel electrophoresis (Plymoth, 2003).

Application of 1D gel with liquid chromatography and MS/MS in biomarker discovery
The rapid development of LC/MS/MS offers a better solution to one-dimensional gel electrophoresis (Schirle et al., 2003). A similar approach has been used to discover proteins www.intechopen.com From Biomarker Discovery to Clinical Evaluation for Early Diagnosis of Lung Surgery-Induced Injury 35 with molecular weight greater than 100 kDa (such as 2-macroglobulin). The discovery of hundreds of proteins in bronchoalveolar lavage fluid demonstrates its feasibility in biomarker identification Chang et al., 2007).

Sensitivity and specificity of the lead proteins after proteomic analysis
To accelerate the translation of biomarker discovery from bench to bedside, the development of techniques has been divided into 5 stages (Pepe et al., 2001). In Phase 1, potential biomarkers are discovered by various approaches, such as proteomic analysis. After the leads are identified by biochemical studies, measurable classifiers or outcomes are developed in Phase 2. Based on the analysis of their specificity and sensitivity, the cutoff point of the measurable outcome is determined and used in Phase 3. Based on patient history and clinical data, the number and nature of clinical cases is well defined in Phase 3. Suitable criteria for a clinical trial are determined in Phase 4. Phase 5 is a randomized trial to compare the specificity and sensitivity of the leads with those of current biomarkers in the market. Today, the importance of sensitivity and specificity in biomarker selection has shifted proteomic studies from large-scale analysis to clinically-relevant validation. In addition to large-scale analysis in protein or metabolite identification (Mou et al., 2011;Huang et al., 2011), the leads are selected based on their sensitivity and specificity.

Translational study from protein identification to clinical application
The purpose of this study was to discover potential biomarkers for the early detection of acute respiratory distress syndrome. To avoid sampling-induced complications, bronchial washings from lung cancer patients before and after surgical therapy (lobectomy) were collected. To reduce population heterogeneity, cancer stage, hormonal variation, and tumor location were well-defined. Only patients older than 60 years, had right lung cancer at stages IA and IB, and agreed to receive right lung lobectomy were recruited. Those patients who met the inclusion and exclusion criteria were selected as controls. The inclusion criteria were: defined cancer in any lobe of the right lung, non-smoker, age ≥60 years, elective operation, operation period <210 min, forced expiratory volume in 1 s (FEV1) >80%, and no prior major lung resection or thoracic irradiation. Exclusion criteria were: age <60 years, operation period >210 min, FEV1 <60%, emergency or urgent operation, and prior major lung resection or distant thoracic irradiation. Based on our criteria, 7 patients (5 females and 2 males with ages ranging between 61 and 77 years) were included as controls. A review was conducted from the medical records and prospective database. The study protocol was approved by the Human Medical Studies Committee at National Cheng Kung University Medical College Hospital. Informed consent was given by all participants or their legal guardians . Prior to a large-scale analysis of proteins in bronchial washings, the protein profiles of washings from different compartments of the lungs before and after lobectomy were compared. In the right lung where tumor tissues were identified, no clear bands at molecular weights >75 kDa were found in washings collected before or after lobectomy. In the left lung, no clear bands at molecular weights >75 kDa were found in the washings collected before lobectomy. After lobectomy, more bands at molecular weights >75 kDa were found. The intensity of each band was much greater (Fig. 1). Similar patterns were found in all samples studied. The banding pattern allowed us to hypothesize that the proteins at molecular weights >75 kDa are exuded into alveoli after surgery. One-dimensional gel electrophoresis coupled with LC/MS/MS allowed us to identify the proteins in 13 major bands. As listed in Table 1, 8 proteins had molecular weights >100 kDa, including 2-macroglobulin. To test our hypothesis that protein exudation is surgery-dependent, the relative abundance of 1-antitrypsin (47 kDa) and 2-macroglobulin (162 kDa) in bronchial washings was measured by Western blot analysis. 1-antitrypsin was found in washings collected before and after lobectomy (data not shown) but 2-macroglobulin was only found after lobectomy (Fig. 2). α2-macroglobulin→ α2-macroglobulin→ α2-macroglobulin→

Vascular endothelial growth factor and lobectomy-induced inflammation
Since vascular endothelial growth factor is a potent inducer of vascular permeability (Lee, 2005) and its expression is positively correlated with inflammation-induced protein exudation and leukocyte infiltration (Chang et al., 2005), it is plausible to suggest that an increase in vascular endothelial growth factor is associated with surgery-induced protein exudation and leukocyte infiltration.
As shown in Table 2, the vascular endothelial growth factor level was positively correlated with total protein concentration (y = 0.0025x + 1.0755, R 2 = 0.7359, P <0.05) and cell count (y = 0.0696x -0.6441, R 2 = 0.8463, P <0.05) but not with operation duration or PaO 2 /FiO 2 . The correlation analysis supported the hypothesis that the induction of vascular endothelial growth factor after surgery contributes to leukocyte infiltration and protein exudation.  Table 2. Clinical data of 7 patients who received right lung lobectomy without complications

α2-macroglobulin and α1antitrypsin in lobectomy-induced inflammation
The relative expression of 2-macroglobulin at bands 2, 4, and 5 from bronchial washings was correlated with protein concentration, leukocyte number, and the level of vascular endothelial growth factor (Fig. 3). Likewise, the relative expression of 1-antitrypsin at bands 5, 7, and 8 from bronchial washing was positively correlated with protein concentration, leukocyte number, and the level of vascular endothelial growth factor (data not shown). These data supported our hypothesis that the increase of vascular endothelial growth factor after surgery facilitates leukocyte infiltration and the exudation of acute-phase proteins (such as 1-antitrypsin and 2-macroglobulin) into alveoli.

Characterization of α2-macroglobulin and α1-antitrypsin in lobectomized patients with acute respiratory distress syndrome
Based on the report of the joint American-European Consensus Conference, the acute respiratory distress syndrome is well defined as follows: bilateral infiltrates on frontal chest radiography, the absence of left atrial hypertension (pulmonary capillary wedge pressure <18 mmHg or no clinical signs of left ventricular failure), and severe hypoxemia with a PaO 2 /FiO 2 ratio <200 mmHg (Bernard et al., 1994). Five patients who received lung surgery and met these criteria were studied.

Characterization of patients with acute respiratory distress syndrome
The group with lobectomy free of complications had levels of total protein and total leukocyte numbers in their bronchial washings similar to those who developed acute respiratory distress syndrome (P >0.05, Fig. 4). These data indicate that lung surgery induces inflammation (leukocyte infiltration and protein exudation) in the groups with and without the complication of acute respiratory distress syndrome. So, factors other than inflammation contribute to the development of this syndrome.  In lung cancer patients, an increase of vascular endothelial growth factor is positively associated with poor prognosis (P = 0.018; Han et al., 2001) but not with a worse postoperative year-survival rate (P = 0.0643; Liao et al., 2001). These reports are also consistent with our finding that the increase of vascular endothelial growth factor after lung surgery does not contribute to surgery-induced acute respiratory distress syndrome.

Protein profiling of bronchial washings from lobectomized patients with acute respiratory distress syndrome
Unlike patients with no complications, those with acute respiratory distress syndrome showed white or gray patches on the chest X-ray. In one-dimensional gel electrophoresis, the protein profiling of bronchial washings from patients without complications showed a much clearer banding pattern than those from patients with acute respiratory distress syndrome (Fig. 5). Eight bands from each gel were cut and subjected to LC/MS/MS for protein identification. No protein was identified in Lane 1. The most significant difference was that albumin appeared in almost every band of the samples from patients without complications but not in those with acute respiratory distress syndrome. In contrast, 1antitrypsin was identified only in bands 6 and 7 from the group without complications but was found in bands 2, 3, 4, 5, 6, and 7 in the group with the complication (Fig. 5).

α2-macroglobulin and α1-antitrypsin in bronchial washings from lobectomized patients with acute respiratory distress syndrome
As shown in Fig. 6, both 2-macroglobulin and 1-antitrypsin were detected in bronchial washings after surgery. After quantification, the total amounts of 2-macroglobulin at bands 2, 4, and 5 and 1antitrypsin at bands 5, 7, and 8 did not show any statistical difference between the groups with and without complications. The most important finding was lower levels of 1antytrypsin at bands 7 and 8 in the group without complications than the acute respiratory distress syndrome group (Fig. 6). It is likely that 1-antitrypsin variants at bands 5, 7, and 8 can be used as biomarkers for the early detection of acute respiratory distress syndrome. In bronchial washings collected from the patients with acute respiratory distress syndrome, leukocyte number was not correlated with the total amounts of 2-macroglobulin or 1antitrypsin. Our analyses again supported the notion that surgery-induced inflammation is not an important indicator in the early phase of acute respiratory distress syndrome. It has been reported that 1-antitrypsin can be produced by lung epithelial cells (Venember et al., 1994)  1-antitrypsin in A549, a lung epithelial cell line. The changes in 1-antitrypsin variants could be due to functional changes in lung epithelial cells.

Specificity and sensitivity of α1-antitrypsin variants as potential biomarkers for acute respiratory distress syndrome
It is of importance to turn the relative expression of 1-antitrypsin in bronchial washings into a measurable outcome because only the measurable outcome is used to determine the cutoff value. Based on the cutoff value, sensitivity (the proportion of subjects who test positive among those with the condition) and specificity (the proportion of subjects who test negative among those without the condition) can be calculated. As shown in Fig. 6, 1-antitrypsin variants at bands 7 (47 kDa) and 8 (40 kDa) had a lower abundance in the group without complications than the group with acute respiratory syndrome. To avoid variations in sample loading and the intensity in each calculation, the ratio of the expression of 1-antitrypsin at band 5 (70 kDa) to that at bands 7 and 8 was used as the measurable outcome. Based on this calculation, the cutoff value was 0.5. A ratio <0.5 was considered an indication of acute respiratory distress syndrome. Table 3 shows the ratio for each patient from the complication-free group. Four out of 7 patients had a ratio <0.5. The specificity of 1-antitrypsin for true negative patients was 0.43 (3/7). Table 4 shows the ratio for each patient from the complication group. Three out of 5 patients had a ratio <0.5. The sensitivity of 1-antitrypsin for true positive patients was 0.6 (3/5).

Patient No
Ratio of expression of 1-antitrypsin at band 5 to that at bands 7 and 8  Table 4. Ratio of the expression of 1-antitrypsin at band 5 to that at bands 7 and 8 in lobectomized patients with acute respiratory distress syndrome.

Further improvement of specificity and sensitivity for detecting acute respiratory distress syndrome using dual biomarkers
As shown in Tables 3 and 4, the sensitivity of 1-antitrypsin variants for detecting acute respiratory distress syndrome (0.6) was better than the specificity (0.43). The major concern is how to optimize the cutoff value and improve the specificity. In table 3, patients 1 and 6 with ratios <0.5 showed the lowest values in cell counts and protein concentration. Meanwhile, the expression of 2-macroglobulin was almost undetectable, which indicates minor inflammation in the patients. The lower ratio of relative expression of 1-antitrypsin at band 5 to that at bands 7 and 8 was false-positive. 1-antitrypsin was found in the lungs before and after surgery; 2-macroglobulin only occurred in the lungs after surgery. To avoid the lower levels of 1-antitrypsin variants which may create a false-positive result, 2-macroglobulin can be recruited as a second biomarker. The ratio of 1-antitrypsin variants was considered as a true result only when the sample expressed detectable 2-macroglobulin in bronchial washings. Accordingly, the specificity for true negative patients changed to 0.71 (5/7). The prediction for true negatives was improved.

From identification of leads to further validation using α2-macroglobulin and α1-antitrypsin variants as an example
After the discovery of potential biomarkers by proteomic analysis in this study, the first challenge was to identify the leads from the proteins discovered after developing a quick screening test. After Phase 1, the second challenge was to provide clear justification to optimize the cutoff values.

Contribution of this study to the discovery of biomarkers for detecting acute respiratory distress syndrome
Ideally, quantitative proteomic analysis should be used to reveal lobectomy-induced changes of all proteins in bronchial washings. However, the unique compartment of the lung allowed us to analyze exudate components which may not exist before surgery, such as 2-macroglobulin. Based on the important mechanism of surgery-induced inflammation in the early phase of lung injury, one-dimensional gel electrophoresis in this study was an easy and suitable tool to identify 2-macroglobulin as an indicator of vascular endothelial growth factor-mediated permeability. The second contribution of this study was to take advantage of one-dimensional gel electrophoresis with pattern analysis to reveal the pattern changes of 1-antitrypsin between the groups with and without post-surgical complications. The difference found allowed us to identify 1-antytripsin variants as biomarkers for the early detection of acute respiratory distress syndrome.

Limitations of this study
In this study, 1-antitrypsin variants were considered as biomarkers for acute respiratory distress. No mechanistic data are provided to explain why and how the formation of 1antitrypsin variants are related to the progression from surgery-induced inflammation to acute respiratory distress syndrome. The association between 1-antitrypsin variants and infection was first reported in 2010 (Zhang et al., 2010). The decrease of the 1-antitrypsin variant at 130 kDa and the increase of the variant at 40 kDa is associated with human immunodeficiency virus-induced infection. Glycoproteomic analysis shows that changes in 1-antitrypsin variants may be due to a shift of glycosylation. In future, glycoproteomic analysis of 1-antitrypsin variants should be further explored. Although the analysis of their specificity and sensitivity, the cutoff point of the measurable outcome, and criteria for patient selection are clearly and easily determined, the small number of clinical cases in this study limits the generalization of 2-macroglobulin and 1antitrypsin as markers for acute respiratory distress syndrome. To use them as measurable biomarkers in Phase 3, it is necessary to increase the number and the complexity of clinical cases for further validation on whether the cutoff points determined are suitable for early diagnosis of acute respiratory distress syndrome. One-dimensional gel electrophoresis does not offer a good way for protein separation. Comparative proteomic analysis only compares the intensity of each spot. These two 44 approaches may our discovery of new proteins. The technology of stable isotope dimethyl labeling coupled with LC/MS/MS permits further quantification of specific peptides of each protein and provides a better quantification tool after one-dimensional electrophoresis (Huang et al., 2006). This approach then compensates for the limitation of one-dimensional gel electrophoresis.

Conclusion
Both inflammation -dependent and -independent mechanisms contribute to the progression from lung injury to acute respiratory distress syndrome. Stage-dependent changes in biomarkers allow us to monitor the progression of the diseases and develop new treatments in a stage-dependent manner. In this study, 2-macroglobulin and 1-antitrypsin were positively correlated with vascular endothelial growth factor, clearly showing lobectomy-induced inflammation. The total amount of 1-macroglobulin can be used as a biomarker of increased vascular permeability in the lung. The severity of lobectomy-induced inflammation is similar to that of inflammation in acute respiratory distress syndrome but respiratory function becomes much worse in patients with the syndrome. Concomitantly, the patients with acute respiratory distress syndrome had lower levels of 1-antitrypsin at higher molecular weights and higher levels of 1-antitrypsin at lower molecular weights. Similarly, human immunodeficiency virus-induced infection is associated with the decreased abundance of 1-antitrypsin at higher molecular weights and the increased abundance of 1-antitrypsin at lower molecular weights (Zhang et al., 2010). Because 1-antitrypsin exists in lung epithelial cells (Venember et al., 1994), the changes of 1-antitrypsin variants in the patients with acute respiratory distress may reflect lung epithelial damage.