Article | Open

Bronchoalveolar Lavage Fluid Protein Expression in Acute Respiratory Distress Syndrome Provides Insights into Pathways Activated in Subjects with Different Outcomes

  • Scientific Reports 7, Article number: 7464 (2017)
  • doi:10.1038/s41598-017-07791-8
  • Download Citation
Published online:


Acute respiratory distress syndrome (ARDS) is associated with high mortality. We sought to identify biological pathways in ARDS that differentiate survivors from non-survivors. We studied bronchoalveolar lavage fluid (BALF) from 36 patients with ARDS (20 survivors, 16 non-survivors). Each sample, obtained within seven days of ARDS onset, was depleted of high abundance proteins and labeled for iTRAQ LC-MS/MS separately. Protein identification and relative quantification was performed employing a target-decoy strategy. A variance weighted t-test was used to identify differential expression. Ingenuity Pathway Analysis was used to determine the canonical pathways that differentiated survivors from non-survivors. We identified 1115 high confidence proteins in the BALF out of which 142 were differentially expressed between survivors and non-survivors. These proteins mapped to multiple pathways distinguishing survivors from non-survivors, including several implicated in lung injury and repair such as coagulation/thrombosis, acute phase response signaling and complement activation. We also identified proteins assigned to fibrosis and ones involved in detoxification of lipid peroxide-mediated oxidative stress to be different in survivors and non-survivors. These results support our previous findings demonstrating early differences in the BALF protein expression in ARDS survivors vs. non-survivors, including proteins that counter oxidative stress and canonical pathways associated with fibrosis.


Acute respiratory distress syndrome (ARDS) occurs as a response to infectious or inflammatory triggers and is characterized by acute tachypnea, refractory hypoxia, and loss of lung compliance1. Although a variety of conditions are triggers for ARDS, common risk factors include pneumonia (59%), extrapulmonary sepsis (16%), and aspiration (14.2%)2. Despite widespread adoption of lung protective ventilation3, early use of muscle relaxants4, extracorporeal membrane oxygenation5 and prone ventilation6, case fatality rates remain at 30–40%2, 7,8,9,10.

Regardless of the cause of ARDS, there is an unmet need to develop tools to reliably assess the mortality risk of patients early in ARDS evolution in order to design interventions to improve survival rates. The current definition of ARDS is based on physiological derangement11 and does not identify causal mechanisms. However, there is heterogeneity in disease susceptibility12,13,14,15,16 and manifestations. Only 50% of cases classified as ARDS have diffuse alveolar damage on biopsy17. Other lung-specific responses differ with respect to the degree of hypoxia2, ventilator driving pressure18, lung stiffness17, and collagen deposition19,20,21. A greater understanding of the differences in the disease biology that accounts for the variability in ARDS manifestations is critical to improve outcomes for ARDS patients22. Sophisticated modeling using latent class analysis of the ARDS Network cohort has identified subphenotypes in ARDS23. The subphenotypes with higher plasma concentrations of inflammatory markers are associated with an increased prevalence of sepsis requiring vasopressors and an inadequate response to therapeutic interventions such as fluid restrictive therapy24. Understanding differences in the underlying biologic processes that influence survival of ARDS patients early in the course of the illness could promote precision-medicine and exploration of new or existing therapeutic agents to improve survival22.

We have previously reported early differences in bronchoalveolar lavage fluid (BALF) between ARDS survivors and non-survivors25, although the study was conducted on pooled BALF and with only a single mass-spectrometry (MS) experiment. Here, we sought to identify biological processes and canonical pathways that differ in ARDS survivors and non-survivors by characterizing individual BALF samples using state-of-the-art label-based semi-quantitative protein expression profiling using iTRAQ (isobaric tagging for relative and absolute quantification) labeling in combination with fixed-gene set enrichment analysis. This study represents a critical first step in our effort to identify proteomic molecular endotypes in ARDS and identify potential signaling pathway targets to reduce ARDS-related mortality.


Characteristics of Study Subjects

In this study, we characterized ARDS BALF samples from 20 subjects who survived and 16 subjects that died during their hospital stay. Five patients (3 survivors and 2 non-survivors) had indirect lung injury from either sepsis (n = 4) or pancreatitis (n = 1), whereas 31 cases had direct lung injury such as pneumonia or aspiration pneumonia (17 survivors and 14 non-survivors). Table 1 outlines the risk factors for ARDS and previous pulmonary history for the study subjects. The median time to death after bronchoscopy in the non-survivors was seven days (IQR = 16.5 days). The median time from ARDS onset (defined as the day of intubation) to bronchoscopy was 2 days. The length of intensive care unit (ICU) stay was not significantly different between the two groups; although the length of hospital stay was lower in non-survivors, this did not reach statistical significance (Table 2).

Table 1: Pulmonary History and Clinical Risk Factors for ARDS in Study Subjects.
Table 2: Demographic and Clinical Characteristics of Study Subjects.

Survivors were younger than non-survivors (mean age 42 vs. 59 years, p = 0.01), but there were no statistically significant differences between the two groups in the time from ARDS onset to BALF collection (ARDS day) or in BALF leukocyte, neutrophil or lymphocyte counts (Table 2). The global internal standard consisted of BALF from 27 subjects who had respiratory failure from various etiologies, including pleural disease, mediastinal mass, myopathy/polyneuropathy, interstitial lung disease without exacerbation, radiation pneumonitis, infectious pneumonia, congestive heart failure and ARDS (one patient).

Proteins Identified by Database Search

The Protein-Pilot Proteomics System Performance Evaluation Pipeline Software (PSPEP) false discovery rate (FDR) summary showing the number of spectra, peptides and the proteins identified at 1% global FDR for the six iTRAQ experiments is shown in Table 3. Six iTRAQ liquid chromatography tandem MS (LC-MS/MS) runs identified 850, 606, 1055, 865, 976, and 879 proteins per run, respectively (Supplemental Table S1). The database search using the combined RAW files from all six LC-MS/MS experiments resulted in the identification of 1189 proteins at a local FDR ≤ 5% (Supplemental Table S2, master protein list tab). After manually removing misidentified proteins, reverse matches, contaminants, and proteins that were not completely removed by the high abundance protein depletion column, we retained 1115 unique proteins for further analysis. The fold change for these proteins can be found in Supplemental Table S2.

Table 3: PSPEP Protein Summary Report for Proteins Identified at ≤1% Global False Discovery Rate.

Differentially Expressed Proteins between ARDS Survivors and Non-survivors

For each iTRAQ eightplex, the fold changes for proteins identified were available with low (113–117) and high (118–121) reporter ion channels as the reference. Thus, we performed this analysis using the lower reporter ion channel and the higher reporter ion channel as a reference to calculate the relative abundance for all six iTRAQ LC-MS/MS experiments. Controlling for a q-value of ≤ 0.05 and using two internal standard reporter ion channels, we identified 181 and 166 proteins to be differentially expressed when we used the low or high reporter ion channel as the reference for calculation of fold change, respectively (Supplemental Table S3), with 142 proteins common between the two comparisons (Supplemental Table S3).

Proteins with the highest differential expression that were more abundant in survivors included galectin-10 (Q05315, log fold = −0.21 ± 0.12 vs 2.20 ± 0.17), cystatin-SA (P09228, log fold = 0.04 ± 0.02 vs 2.45 ± 0.26), glutamine gamma-glutamyltransferase 2 (P21980, log fold = 0.28 ± 0.09 vs 1.82 ± 0.24), prosaposin (C9JIZ6, log fold = 0.01 ± 0.07 vs 1.51 ± 0.21), and cathepsin D (P07339, log fold = 0.50 ± 0.11 vs 1.99 ± 0.19). Proteins with the highest differential expression with higher abundance in non-survivors included alpha-amylase 1 (P04745, log-fold 1.14 ± 0.20 vs −0.74± 0.04), club-cell secretory protein (P11684, log fold 2.03 ± 0.11 vs. 0.43 ±0.11), BPI fold-containing family B member 2 (Q8N4F0, log fold = 1.20 ± 0.19 vs 0.31 ± 0.09), metallothionein-2 (P02795, log fold = 2.31±vs 0.87± 0.15), and myoglobin (P02144, log fold 1.15 ±I 0.14 vs 0.13± 0.10).

Among the proteins that were more abundant in survivors were different isoforms of aldehyde dehydrogenase, alcohol dehydrogenase, and aldo-keto reductases (Table 4). Specifically, we observed significantly higher levels of isoform 2 of alcohol dehydrogenase 2, aldo-keto reductase family 1 C1, glutathione peroxidase 3, aldehyde dehydrogenase mitochondrial. There also was a statistical trend towards a higher abundance of glutathione-s-transferase A2 (q-value = 0.08).

Table 4: Proteins of Interest that Demonstrate a Difference in Expression between Survivors and Non-Survivors in Low Control.

Biological Relevance of the Differentially Expressed Proteins

We performed IPA core analysis to identify the canonical pathways which map to the differentially expressed proteins that met a threshold q-value of ≤0.05. The pathways that met the statistical threshold described in the methods (−log [B-H p-value] ≥ 1.3) and the proteins assigned to each canonical pathway are listed in Table 5. Figure 1 shows the degree of enrichment and the number of proteins with increased or decreased abundance for each pathway in survivors compared with non-survivors. The z-score was available for Complement System (1.667), Acute Phase Response Signaling (2.238), LXR/RXR Activation (3.162), Intrinsic Prothrombin Activation Pathway (1.00) and Coagulation System (0.447). Of the pathways listed in Table 5, only pathways that participate in fibrosis (Hepatic Fibrosis/Hepatic Stellate Cell Activation) were represented in an independent core analysis done on proteins that were more abundant in non-survivors than in survivors. Certain pathways such as Complement Activation, Fatty acid alpha oxidation, Histamine Degradation, Oxidative Ethanol degradation were mapped by separate IPA core analysis by the proteins that were high in survivors and also by some proteins that were high in non-survivors (although with limited coverage of the pathway). The remaining canonical pathways were mapped by proteins more abundant in survivors.

Table 5: IPA Canonical Pathways Represented by Proteins Differentially Expressed Between Survivors and Non-Survivors.
Figure 1
Figure 1

Canonical Pathways represented by proteins differentially expressed between ARDS survivors and non-survivors. Ingenuity Pathway Analysis (IPA) canonical pathways most significantly changed in ARDS survivors compared to non-survivors. The stacked bar chart displays the percentage of proteins in the canonical pathway that are more (light gray), less (dark gray) abundant in survivors or absent (white) our dataset. The secondary x-axis shows the −log of Benajmni-Hochberg corrected p-value indicating the statistical significance of each over-represented pathway (cutoff >1.3) and the line with solid diamonds is the −log of corrected p-value for that pathway. The numerical value on right side of each stacked bar is the number of genes in that canonical pathway in the IPA knowledgebase.

We also identified proteins with differential expression that are assigned to pathways that participate in fibrosis (Table 4) such as collagen alpha-1(V), insulin-like growth factor binding protein 4, vascular cell adhesion protein 1, fibronectin, and collagen alpha-1(XVIII). Additionally, there were several other proteins that participate in the pathophysiology of fibrosis, although they were not differentially expressed in survivors versus non-survivors. These included monocyte differentiation antigen CD14, angiotensinogen, insulin-like growth factor II, insulin-like growth factor binding protein 3, metalloproteinase inhibitor 1 and 2, 72 kDa type IV collagenase, matrix metalloproteinase-9, and collagens alpha-1(I), alpha-2(I), alpha-1(III), alpha-1(V), and alpha-3 (VI).

Proximity Extension Assay for Proteins Participating in Inflammation

As a validation of our iTRAQ findings, we selected five proteins identified in the iTRAQ experiments that were also present in the Olink PEA panel. These proteins were growth-regulated alpha protein (P09341-CXCLI), macrophage colony-stimulating factor 1 (P09603), monocyte chemotactic protein 1 (P13500), Protein S100-A12 (P80511), and Interleukin-18 (Q14116). The Olink results were consistent with those of the MS experiments in finding no difference in the protein level between survivors and non-survivors for growth-regulated alpha protein or S100 A-12. Additionally, the Olink analysis found no statistically significant difference in macrophage colony-stimulating factor 1 or monocyte chemotactic protein 1 by study group. This comparison was not possible with MS because high-quality quantitative information was not available; monocyte chemotactic protein 1 was only present in one of the six iTRAQ experiments, and error factor was not available for macrophage colony-stimulating factor 1 in two of the three iTRAQ MS/MS experiments. By MS, Interleukin-18 was identified in the BALF of only 12 ARDS subjects (6 survivors and 6 non-survivors), and although the levels were higher in non-survivors, the difference did not achieve statistical significance (q-value = 0.22). However, with Olink, Interleukin-18 was identified in all ARDS cases studied except one non-survivor and higher levels were seen in non-survivor compared to survivors (5.8 ± 0.88 vs. 4.9 ± 1.1, p = 0.04).


In this study, we successfully applied label based high-resolution protein expression profiling tools to comprehensively characterize BALF from individual cases of ARDS and identified differences between survivors and non-survivors. We also confirmed the finding of a coordinated response in survivors and a fibrotic signature in non-survivors. PEA results provide analyte level validation for a subset of inflammatory proteins that are present in ARDS BALF.

Oxidative stress occurs in ARDS due to the underlying pathophysiology, but is exacerbated by mechanical ventilation given the high oxygen fractions in inspired gas that is required to maintain adequate oxygenation. In many critically ill patients, limiting oxidative stress by either restricting oxygen exposure or treating with supplemental antioxidants is beneficial and can improve mortality26, 27. However, in ARDS the use of antioxidants has produced mixed results. Some studies show improved gas exchange28, shorter duration of ventilation28, 29, and improved APACHE II score30 while others show no benefit31, 32. In our prior work, Gene Ontology (GO) enrichment analysis, a form of over-representation analysis of fixed-gene sets, identified proteins more abundant in survivors that mapped to cellular cation homeostasis and iron ion homeostasis25 and some of these proteins are antioxidant and cytoprotective proteins regulated by nuclear factor, erythroid 2 like (NRF2)-antioxidant response elements (ARE)33. In the current study, the IPA core analysis revealed differential expression of several metabolic canonical pathways, such as fatty acid alpha-oxidation and ethanol degradation, represented by many of the same proteins (Table 4), including several that are thought to be enzymes involved in the detoxification of lipid peroxidation derived aldehydes34 and are regulated by NRF-ARE. These proteins include aldehyde dehydrogenase, alcohol dehydrogenase, aldo-keto reductases, and glutathione-S-transferase, which were found to be more abundant in survivors. Lipid aldehydes such as Malondialdehyde (MDA), hydroxynonenal (HNE), and acrolein are formed in conditions of oxidative stress35 by oxidation of polyunsaturated fatty acids. These byproducts are highly reactive, forming adducts with DNA or proteins35,36,37 to impact cellular homoeostasis by inactivating critical enzymes such as Na,K-ATPase38,39,40,41. The damage due to protein modification can be mitigated by mechanisms that convert the lipid aldehydes into less reactive alcohols by aldo-keto reductases or alcohol dehydrogenases, or to acids by aldehyde dehydrogenase. These reactions can occur spontaneously (phase 1 metabolism) or can be catalyzed by glutathione-S-transferase (phase 2 metabolism). In this study, enzymes that participate in detoxification of the lipid peroxides are higher in survivors suggesting a higher capacity to counteract lipid aldehydes. The fact that proteins that participate in phase 1 and 2 metabolisms of toxic byproducts of peroxidation are more abundant in survivors suggests that the host response to mitigate oxidative stress governs outcomes in ARDS. The precise lipid aldehyde and the specific mechanism will need further characterization for developing specific treatments in the ARDS cases lacking mechanisms to limit oxidant stress.

Similar to our previous findings in which differentially expressed proteins mapped to the GO term ‘collagen metabolic processes’25, several of the differentially expressed proteins identified in the current analysis mapped to pathways that participate in fibrosis. Although these proteins map to ‘hepatic fibrosis’, this is likely due to a better annotation of these proteins in liver fibrosis, and the fibrotic signature in our study represents a fibrotic response in the lung. In a recent autopsy study, pulmonary fibrosis was present in 51% of ARDS cases of pulmonary origin but only in 20% of cases of extrapulmonary origin17; collagen content in the lung is higher in pulmonary than in extrapulmonary ARDS17, 42. A majority of the cases in our study had direct pulmonary injury as a trigger which could explain why many of our subjects demonstrated a difference in matrix proteins. In conditions associated with pulmonary fibrosis, the extracellular matrix (ECM) and alveolar microenvironment regulate profibrotic genes43,44,45,46. Our previous study observed several ECM proteins, including collagen type 1, III and V, mucin 5a, and matrix metalloproteinase 9, to have a higher abundance in non-survivors25. In the current study, we identified proteins that participate in the pathophysiology of fibrosis, including ECM proteins such as collagens alpha-1 (XVIII), alpha-1 (I), alpha-2 (I), alpha-1 (III), alpha-1 (V), alpha-3 (VI), and fibronectin 1. Several of these proteins are involved in transforming growth factor beta and insulin-like growth factor-mediated pulmonary fibrosis. Of these, angiotensinogen, collagen alpha-1 (XVIII), collagen alpha-1 (V), fibronectin 1, insulin-like growth factor binding protein 4, and vascular cell adhesion protein 1 had differential expression between survivors and non-survivors. Though speculative, our findings regarding differential protein expression suggest a difference in host response mediated by transforming growth factor beta or insulin-like growth factor leading to early collagen deposition in non-survivors. The exact significance of this finding will need further evaluation. It is thus possible that mediators of fibrosis that increase after disease onset47 could be detected much earlier in BALF than with histology on lung tissue.

Congruent with the findings from our previous study, the variance weighted fold change showed higher levels of ceruloplasmin, plasminogen, antithrombin III and coagulation factor XII in survivors, whereas the level of club cell secretory protein (uteroglobin) was higher in non-survivors (Table 4). In contrast to our prior study with pooled BALF, we did not find higher moesin levels in non-survivors in this study but did observe that ezrin (a member of the ezrin, radixin and moesin family of proteins) was high in non-survivors. We acknowledge that the levels of some proteins that met the statistical threshold in our prior study – such as thioredoxin and S100-A9 – did not achieve statistical significance in this study; in fact, the direction of change was different between the two studies. It is possible that differences in the respective study populations or the inherent variability in protein levels over the course of illness could account for these differences. Utilizing a uniform time for sample collection from prospectively enrolled subjects would be beneficial in future studies. Despite the observed differences from our previous study mentioned above, the pathways and biologic processes represented by differentially expressed proteins between ARDS survivors and non-survivors found in the current study are consistent with those seen in the previous study with pooled patients25.

The results of this study should be viewed in light of several limitations. First, the outcome of interest is binary (survival versus non-survival) and can be influenced by some factors not controlled for in our study, such as comorbid conditions. In addition, although there was sufficient BALF to perform MS on individual samples, the overall sample size was relatively limited. Further, the MS method used data-dependent acquisition (DDA) which is biased against low-abundance protein quantification. Future work would benefit from utilizing data-independent acquisition (DIA) such as Sequential Windowed Acquisition of all Theoretical Fragment Ion Mass Spectrometry (SWATH-MS). Also, given the relative heterogeneity between survivors and non-survivors, there may be potential confounders which could have influenced the observed results. Despite these limitations, the current study provides new hypotheses for testing in a larger study cohort. A larger sample size would also permit the use of bootstrapping and other cross-validation tools for modelling phenotypic differences in heterogeneous and complex diseases such as ARDS.

Results from the current study validate several proteins previously reported to be upregulated in survivors in pooled BALF studies25. These proteins include plasminogen, antithrombin III, coagulation factor XII, and ceruloplasmin. Although all demonstrated an increase in expression, not all were statistically significant in the current analysis, likely due to the limited sample size in this study. One protein, club-cell secretory protein, was significantly higher among non-survivors, as was observed in our previous study.

This study furthers our knowledge about the differences in the biological processes activated in ARDS survivors and non-survivors. It also identifies potential future research areas, including determining the role of lipid peroxides and lung fibrosis. Given that a variety of conditions can lead to ARDS, optimal treatment of patients may differ by the mechanism active in individual subjects48, thus supporting phenotyping of ARDS for personalizing care23, 49.


Study Population

Eligible subjects consisted of individuals with ARDS as defined by American-European Consensus Conference (AECC) criteria50 treated at the University of Minnesota Medical Center between May 2009 and September 2012. This study included 36 subjects who underwent a clinically indicated bronchoscopy within seven days of their ARDS diagnosis and for whom excess BALF was available. The study was designed before the publication of the criteria established by the ARDS Definition Task Force (Berlin definition)11. Pooled BALF from 27 subjects who participated in our previous study51 was used as the control to determine the relative protein abundance and as global internal standards for comparison across the different iTRAQ LC-MS/MS experiments. A standard protocol for bronchoscopy was used for BALF collection25, 51.

ARDS subjects were divided into two groups: those who survived until hospital discharge (survivors) and those who died prior to discharge (non-survivors). Subjects with a history of HIV or viral hepatitis were excluded from the study.

Sample processing for protein profiling

After collection, all BALF samples were immediately placed on ice and centrifuged at 500 g at 4 °C for 10 minutes within 60 minutes of collection. Cell and debris free supernatant were stored at −80 °C and did not undergo any thaw-freeze cycles until sample processing.

We employed label-based semi-quantitative proteomics using eightplex iTRAQ reagent52 for our study. iTRAQ multiplexed sets of reagents for quantitative protein analysis place isobaric mass labels at the N-termini and lysine side chains of peptides. All resulting peptides are isobaric and chromatographically indistinguishable, but yield signature of reporter ions following collision induced dissociation that can be used to identify and quantify peptides in a digest mixture.

To compare protein abundance across different LC-MS/MS experiments and as a reference for determining relative protein abundance, we used the global internal standard. In each LC-MS/MS experiment, two iTRAQ reporter ion channels contained the global internal standard. The remaining six channels contained individual study samples (i.e., we performed 6 separate iTRAQ LC-MS/MS experiments to characterize the 36 ARDS cases in the study). The labeling strategy for the 36 BALF samples studied is outlined in Supplemental Table S4. To prevent reporter ion signal (channel) bias, survivor and non-survivor samples were randomly placed in different iTRAQ reporter ion channels in each experiment.

BALF containing at least 8 mg of protein (Bradford reagent, Bio-Rad cat#500-0006) was processed separately employing a protocol previously published with minor modifications25, 51. BALF was concentrated and desalted using Amicon 3-MWCO filters (Millipore Ireland Ltd, Cork, Ireland), depleted of high abundance proteins (Seppro IgY 14 spin column, Sigma-Aldrich, cat # SEP010) with appropriate buffer exchanges for trypsin digestion and labeling with iTRAQ followed by two-dimensional (2D) LC-MS/MS. The adequacy of the trypsin digestion was confirmed by analysis of 3 µg of the digested peptides with linear trap quadrupole MS (LTQ-MS). Equal amounts of the remaining peptide mixtures within each experiment were labeled with eightplex iTRAQ reagent per the manufacturer’s (AB Sciex, Framingham, MA) instructions51. 2D LC-MS/MS of the iTRAQ-labeled peptides was conducted as previously described51. Data-dependent MS acquisition was performed on a Thermo Scientific LTQ Orbitrap Velos system with higher energy collision induced dissociation (HCD) activation for peptide tandem MS. LC and MS experimental details were the same as previously reported51.

Database search for protein identification and quantification

RAW files generated directly from the mass spectrometer were imported into Galaxy-P platform53 for protein identification and quantification25, 51. Galaxy-P has also been used for proteogenomics analysis54,55,56 and metaproteomics studies57, 58. The MGF files were searched against the target-decoy version of Human UniProt database along with the contaminant sequences from the common Repository of Adventitious Proteins (cRAP, (88,304 sequences in total; Date Aug 1, 2014) using ProteinPilot version 4.5. PSPEP-FDR reports and protein and peptide-level summaries were generated within Galaxy-P as previously described54,55,56,57,58. The MS proteomics data have been deposited to the ProteomeXchange Consortium59 via the PRIDE partner repository with the dataset identifier PXD002672.

In each eightplex iTRAQ LC-MS/MS experiment, we used the global internal standard as a reference to determine the relative abundance and labeled two iTRAQ reporter ion channels in order to provide guidance regarding the FDR of quantitative differences (within each experiment). Specifically, the relative abundance of all proteins identified in each experiment when compared using the two reporter ion channels labeled with the global internal standard should be equal to 1; proteins that show a statistically significant difference thus indicate false positives. In our six iTRAQ LC-MS/MS experiments, the number of proteins that had a statistically significant difference and therefore false positive were 8, 15, 53, 11, 25 and 8, respectively, suggesting an adequate quantitative assessment.

The results of multiple iTRAQ LC-MS/MS experiments were aligned to compare protein-levels using Protein Alignment Template vs. 2.00p (AB Sciex)51, 60. For this alignment, we created a ‘reference master list’ by performing a database search using RAW files from all six iTRAQ LC-MS/MS experiments. To ensure that the proteins in this list were of high ID quality, a local FDR ≤5% was used as a threshold for protein identification in the reference master list as per the recommendation of the Protein Alignment Template. For the creation of feature tables with quantitative values, the threshold of ≤5% global FDR was used for individual sets for the six-iTRAQ LC-MS/MS experiments. Protein Alignment Template resulted in aligning the ratios (fold change), p-values and error factors for the proteins across replicate experiments by using accession numbers of isoforms within the protein summary and UniProt database. However, these included misidentified proteins (e.g., bovine albumin, pig trypsin, bovine casein), matches to reverse (decoy) protein sequences61, the contaminant protein sequences from the common Repository of Adventitious Proteins, and the proteins that were not completely removed by the IgY 14 depletion column or their protein fractions such as immunoglobulin chains. These proteins were manually removed from the final protein list (Supplemental Table S2, removed proteins tab).

To further investigate the difference in the BALF inflammatory proteins in ARDS survivors, we performed Olink proseek proximity extension assay (PEA)62. The inflammatory panel of the Olink PEA was performed on 29 un-depleted BALF samples (non-survivors = 14, survivors = 15). BALF was concentrated and desalted using a 3KD MW Amicon centrifugal filter (Millipore, Cork, IRL). Bradford assay (Bio-Rad, USA) determined the protein concentration of samples diluted with Dulbecco’s PBS to 0.5ug/uL and an equal amount (0.5 ug) of proteins per the manufacturer’s instructions.


Identification of differentially expressed proteins between ARDS survivors and non-survivors involved several steps that were similar to our prior studies with minor modifications51. We controlled for multiple comparisons by FDR corrected q-value ≤ 0.05. Log-transformed fold changes for all proteins that were identified in at least two-iTRAQ LC-MS/MS experiments for which error factors were available were compared between ARDS survivors and non-survivors using weighted two-sample t-tests with weights being the inverse variance of the log-transformed fold changes. The Storey method63 was used to control the FDR. This analysis was carried out using the lower reporter ion channel (113–117) and the higher reporter ion channel (117–121) in Statistical Analysis Software (version 9.3, SAS Institute Inc., Cary, NC).

To gain insight into the biological significance of differentially expressed proteins, we performed functional analysis using Ingenuity Pathway Analysis (IPA® QIAGEN, Redwood City Build 321510 M, Version 21249400). This analysis was performed on proteins with a q-value of ≤ 0.05 as the cutoff for differential expression. IPA core analysis was performed using the difference of the weighted log fold change between survivors and non-survivors. Thus, the proteins that were higher in survivors had a positive value and the proteins that were lower had a negative value. We focused on canonical pathways that met a Benjamini and Hochberg (B–H)–corrected p-value (obtained using the right-tailed Fisher exact test) of ≤0.05 (equivalent to −log [B–H p-value] ≥ 1.3). Additionally, IPA uses the z-score algorithm that informs the activation states of canonical pathways with ability to predict activation, inhibition, no change or inability to predict the activation state64. In our study, a positive score predicts activation in ARDS survivors. Several of the pathways in our dataset did not have a z-score assigned to them to predict their activation state.

Study Approval

The University of Minnesota Institutional Review Board (IRB) Human Subjects Committee approved this study (IRB # 0812M5623) and all methods were performed according to the relevant guidelines. Because we utilized excess BALF from clinically indicated bronchoscopies in this study, and as there was minimal risk to study participants, the IRB approved a waiver of consent for this study.

Additional Information

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


  1. 1.

    Ashbaugh, D. G., Bigelow, D. B., Petty, T. L. & Levine, B. E. Acute respiratory distress in adults. Lancet 2, 319–323 (1967).

  2. 2.

    Bellani, G. et al. Epidemiology, Patterns of Care, and Mortality for Patients With Acute Respiratory Distress Syndrome in Intensive Care Units in 50 Countries. JAMA 315, 788–800, doi:10.1001/jama.2016.0291 (2016).

  3. 3.

    Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. The Acute Respiratory Distress Syndrome Network. The New England journal of medicine 342, 1301–1308, doi:10.1056/NEJM200005043421801 (2000).

  4. 4.

    Papazian, L. et al. Neuromuscular blockers in early acute respiratory distress syndrome. The New England journal of medicine 363, 1107–1116, doi:10.1056/NEJMoa1005372 (2010).

  5. 5.

    Peek, G. J. et al. Efficacy and economic assessment of conventional ventilatory support versus extracorporeal membrane oxygenation for severe adult respiratory failure (CESAR): a multicentre randomised controlled trial. Lancet 374, 1351–1363, doi:10.1016/S0140-6736(09)61069-2 (2009).

  6. 6.

    Guerin, C., Reignier, J. & Richard, J. C. Prone positioning in the acute respiratory distress syndrome. The New England journal of medicine 369, 980–981, doi:10.1056/NEJMc1308895 (2013).

  7. 7.

    Rubenfeld, G. D. et al. Incidence and outcomes of acute lung injury. The New England journal of medicine 353, 1685–1693, doi:10.1056/NEJMoa050333 (2005).

  8. 8.

    Villar, J. et al. The ALIEN study: incidence and outcome of acute respiratory distress syndrome in the era of lung protective ventilation. Intensive care medicine 37, 1932–1941, doi:10.1007/s00134-011-2380-4 (2011).

  9. 9.

    Wang, C. Y. et al. One-year mortality and predictors of death among hospital survivors of acute respiratory distress syndrome. Intensive care medicine 40, 388–396, doi:10.1007/s00134-013-3186-3 (2014).

  10. 10.

    Esteban, A. et al. Evolution of mortality over time in patients receiving mechanical ventilation. Am J Respir Crit Care Med 188, 220–230, doi:10.1164/rccm.201212-2169OC (2013).

  11. 11.

    The ARDS Definition Task Force. Acute Respiratory Distress Syndrome: The Berlin Definition. JAMA 307, 2526–2533, doi:10.1001/jama.2012.5669 (2012).

  12. 12.

    Ahasic, A. M. et al. Adiponectin gene polymorphisms and acute respiratory distress syndrome susceptibility and mortality. Plos One 9, e89170, doi:10.1371/journal.pone.0089170 (2014).

  13. 13.

    Tejera, P. et al. Functional characterization of polymorphisms in the peptidase inhibitor 3 (elafin) gene and validation of their contribution to risk of acute respiratory distress syndrome. American journal of respiratory cell and molecular biology 51, 262–272, doi:10.1165/rcmb.2013-0238OC (2014).

  14. 14.

    O’Mahony, D. S. et al. Inflammation and immune-related candidate gene associations with acute lung injury susceptibility and severity: a validation study. Plos One 7, e51104, doi:10.1371/journal.pone.0051104 (2012).

  15. 15.

    Bajwa, E. K., Yu, C. L., Gong, M. N., Thompson, B. T. & Christiani, D. C. Pre-B-cell colony-enhancing factor gene polymorphisms and risk of acute respiratory distress syndrome. Crit Care Med 35, 1290–1295, doi:10.1097/01.CCM.0000260243.22758.4F (2007).

  16. 16.

    Gong, M. N. et al. Polymorphisms in the mannose binding lectin-2 gene and acute respiratory distress syndrome. Crit Care Med 35, 48–56, doi:10.1097/01.CCM.0000251132.10689.F3 (2007).

  17. 17.

    Thille, A. W. et al. Chronology of histological lesions in acute respiratory distress syndrome with diffuse alveolar damage: a prospective cohort study of clinical autopsies. Lancet Respir Med 1, 395–401, doi:10.1016/S2213-2600(13)70053-5 (2013).

  18. 18.

    Amato, M. B. et al. Driving pressure and survival in the acute respiratory distress syndrome. The New England journal of medicine 372, 747–755, doi:10.1056/NEJMsa1410639 (2015).

  19. 19.

    Auler Junior, J. O. et al. Adult respiratory distress syndrome: evidence of early fibrogenesis and absence of glucocorticoid receptors. Eur J Respir Dis 69, 261–269 (1986).

  20. 20.

    Zapol, W. M., Trelstad, R. L., Coffey, J. W., Tsai, I. & Salvador, R. A. Pulmonary fibrosis in severe acute respiratory failure. Am Rev Respir Dis 119, 547–554, doi:10.1164/arrd.1979.119.4.547 (1979).

  21. 21.

    Ingbar, D. H. Mechanisms of repair and remodeling following acute lung injury. Clinics in chest medicine 21, 589–616 (2000).

  22. 22.

    Beitler, J. R. et al. Personalized medicine for ARDS: the 2035 research agenda. Intensive care medicine 42, 756–767, doi:10.1007/s00134-016-4331-6 (2016).

  23. 23.

    Calfee, C. S. et al. Subphenotypes in acute respiratory distress syndrome: latent class analysis of data from two randomised controlled trials. Lancet Respir Med 2, 611–620, doi:10.1016/s2213-2600(14)70097-9 (2014).

  24. 24.

    Famous, K. R. et al. Acute Respiratory Distress Syndrome Subphenotypes Respond Differently to Randomized Fluid Management Strategy. Am J Respir Crit Care Med 195, 331–338, doi:10.1164/rccm.201603-0645OC (2017).

  25. 25.

    Bhargava, M. et al. Proteomic Profiles in Acute Respiratory Distress Syndrome Differentiates Survivors from Non-Survivors. Plos One 9, doi:10.1371/journal.pone.0109713 (2014).

  26. 26.

    Girardis, M. et al. Effect of Conservative vs Conventional Oxygen Therapy on Mortality Among Patients in an Intensive Care Unit: The Oxygen-ICU Randomized Clinical Trial. JAMA 316, 1583–1589, doi:10.1001/jama.2016.11993 (2016).

  27. 27.

    Manzanares, W., Dhaliwal, R., Jiang, X., Murch, L. & Heyland, D. K. Antioxidant micronutrients in the critically ill: a systematic review and meta-analysis. Critical care (London, England) 16, R66, doi:10.1186/cc11316 (2012).

  28. 28.

    Singer, P. et al. Benefit of an enteral diet enriched with eicosapentaenoic acid and gamma-linolenic acid in ventilated patients with acute lung injury. Crit Care Med 34, 1033–1038, doi:10.1097/01.ccm.0000206111.23629.0a (2006).

  29. 29.

    Bernard, G. R. et al. A trial of antioxidants N-acetylcysteine and procysteine in ARDS. The Antioxidant in ARDS Study Group. Chest 112, 164–172 (1997).

  30. 30.

    Soltan-Sharifi, M. S. et al. Improvement by N-acetylcysteine of acute respiratory distress syndrome through increasing intracellular glutathione, and extracellular thiol molecules and anti-oxidant power: evidence for underlying toxicological mechanisms. Human & experimental toxicology 26, 697–703, doi:10.1177/0960327107083452 (2007).

  31. 31.

    Kagan, I. et al. Preemptive enteral nutrition enriched with eicosapentaenoic acid, gamma-linolenic acid and antioxidants in severe multiple trauma: a prospective, randomized, double-blind study. Intensive care medicine 41, 460–469, doi:10.1007/s00134-015-3646-z (2015).

  32. 32.

    Rice, T. W. et al. Enteral omega-3 fatty acid, gamma-linolenic acid, and antioxidant supplementation in acute lung injury. JAMA 306, 1574–1581, doi:10.1001/jama.2011.1435 (2011).

  33. 33.

    Baird, L. & Dinkova-Kostova, A. T. The cytoprotective role of the Keap1-Nrf2 pathway. Arch Toxicol 85, 241–272, doi:10.1007/s00204-011-0674-5 (2011).

  34. 34.

    Curtis, J. M. et al. Protein carbonylation and metabolic control systems. Trends in endocrinology and metabolism: TEM 23, 399–406, doi:10.1016/j.tem.2012.05.008 (2012).

  35. 35.

    Ayala, A., Munoz, M. F. & Arguelles, S. Lipid peroxidation: production, metabolism, and signaling mechanisms of malondialdehyde and 4-hydroxy-2-nonenal. Oxidative medicine and cellular longevity 2014, 360438, doi:10.1155/2014/360438 (2014).

  36. 36.

    Hauck, A. K. & Bernlohr, D. A. Oxidative stress and lipotoxicity. Journal of lipid research 57, 1976–1986, doi:10.1194/jlr.R066597 (2016).

  37. 37.

    Kuzmenko, A. I., Wu, H., Bridges, J. P. & McCormack, F. X. Surfactant lipid peroxidation damages surfactant protein A and inhibits interactions with phospholipid vesicles. Journal of lipid research 45, 1061–1068, doi:10.1194/jlr.M300360-JLR200 (2004).

  38. 38.

    Miyake, H., Kadoya, A. & Ohyashiki, T. Increase in molecular rigidity of the protein conformation of brain Na + -K + -ATPase by modification with 4-hydroxy-2-nonenal. Biol Pharm Bull 26, 1652–1656 (2003).

  39. 39.

    Yang, J. H., Yang, E. S. & Park, J. W. Inactivation of NADP + -dependent isocitrate dehydrogenase by lipid peroxidation products. Free Radic Res 38, 241–249 (2004).

  40. 40.

    Fang, J. & Holmgren, A. Inhibition of thioredoxin and thioredoxin reductase by 4-hydroxy-2-nonenal in vitro and in vivo. J Am Chem Soc 128, 1879–1885, doi:10.1021/ja057358l (2006).

  41. 41.

    Cassidy, P. B. et al. Thioredoxin reductase is required for the inactivation of tumor suppressor p53 and for apoptosis induced by endogenous electrophiles. Carcinogenesis 27, 2538–2549, doi:10.1093/carcin/bgl111 (2006).

  42. 42.

    Negri, E. M. et al. Acute remodeling of parenchyma in pulmonary and extrapulmonary ARDS. An autopsy study of collagen-elastic system fibers. Pathology, research and practice 198, 355–361, doi:10.1078/0344-0338-00266 (2002).

  43. 43.

    Parker, M. W. et al. Fibrotic extracellular matrix activates a profibrotic positive feedback loop. The Journal of clinical investigation 124, 1622–1635, doi:10.1172/JCI71386 (2014).

  44. 44.

    Madtes, D. K. et al. Elevated transforming growth factor-alpha levels in bronchoalveolar lavage fluid of patients with acute respiratory distress syndrome. Am J Respir Crit Care Med 158, 424–430, doi:10.1164/ajrccm.158.2.9711112 (1998).

  45. 45.

    Pittet, J. F. et al. TGF-beta is a critical mediator of acute lung injury. The Journal of clinical investigation 107, 1537–1544, doi:10.1172/JCI11963 (2001).

  46. 46.

    Fernandez, I. E. & Eickelberg, O. The impact of TGF-beta on lung fibrosis: from targeting to biomarkers. Proceedings of the American Thoracic Society 9, 111–116, doi:10.1513/pats.201203-023AW (2012).

  47. 47.

    Marshall, R., Bellingan, G. & Laurent, G. The acute respiratory distress syndrome: fibrosis in the fast lane. Thorax 53, 815–817 (1998).

  48. 48.

    Papazian, L. et al. Diagnostic workup for ARDS patients. Intensive care medicine 42, 674–685, doi:10.1007/s00134-016-4324-5 (2016).

  49. 49.

    Prescott, H. C., Calfee, C. S., Thompson, B. T., Angus, D. C. & Liu, V. X. Toward Smarter Lumping and Smarter Splitting: Rethinking Strategies for Sepsis and Acute Respiratory Distress Syndrome Clinical Trial Design. American journal of respiratory and critical care medicine 194, 147–155, doi:10.1164/rccm.201512-2544CP (2016).

  50. 50.

    Bernard, G. R. et al. The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 149, 818–824, doi:10.1164/ajrccm.149.3.7509706 (1994).

  51. 51.

    Bhargava, M. et al. Proteome Profiling in Lung Injury after Hematopoietic Stem Cell Transplantation. Biol Blood Marrow Transplant 22, 1383–1390, doi:10.1016/j.bbmt.2016.04.021 (2016).

  52. 52.

    Ross, P. L. et al. Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Mol Cell Proteomics 3, 1154–1169, doi:10.1074/mcp.M400129-MCP200 (2004).

  53. 53.

    Boekel, J. et al. Multi-omic data analysis using Galaxy. Nat Biotechnol 33, 137–139, doi:10.1038/nbt.3134 (2015).

  54. 54.

    Anderson, K. J. et al. Proteogenomic Analysis of a Hibernating Mammal Indicates Contribution of Skeletal Muscle Physiology to the Hibernation Phenotype. Journal of proteome research 15, 1253–1261, doi:10.1021/acs.jproteome.5b01138 (2016).

  55. 55.

    Jagtap, P. D. et al. Flexible and accessible workflows for improved proteogenomic analysis using the Galaxy framework. Journal of proteome research 13, 5898–5908, doi:10.1021/pr500812t (2014).

  56. 56.

    Vermillion, K. L., Jagtap, P., Johnson, J. E., Griffin, T. J. & Andrews, M. T. Characterizing Cardiac Molecular Mechanisms of Mammalian Hibernation via Quantitative Proteogenomics. Journal of proteome research 14, 4792–4804, doi:10.1021/acs.jproteome.5b00575 (2015).

  57. 57.

    Rudney, J. D. et al. Protein relative abundance patterns associated with sucrose-induced dysbiosis are conserved across taxonomically diverse oral microcosm biofilm models of dental caries. Microbiome 3, 69, doi:10.1186/s40168-015-0136-z (2015).

  58. 58.

    Jagtap, P. D. et al. Metaproteomic analysis using the Galaxy framework. Proteomics 15, 3553–3565, doi:10.1002/pmic.201500074 (2015).

  59. 59.

    Vizcaino, J. A. et al. ProteomeXchange provides globally coordinated proteomics data submission and dissemination. Nat Biotechnol 32, 223–226, doi:10.1038/nbt.2839 (2014).

  60. 60.

    Champion, M. M., Williams, E. A., Kennedy, G. M. & Champion, P. A. Direct detection of bacterial protein secretion using whole colony proteomics. Mol Cell Proteomics 11, 596–604, doi:10.1074/mcp.M112.017533 (2012).

  61. 61.

    Tang, W. H., Shilov, I. V. & Seymour, S. L. Nonlinear fitting method for determining local false discovery rates from decoy database searches. Journal of proteome research 7, 3661–3667, doi:10.1021/pr070492f (2008).

  62. 62.

    Assarsson, E. et al. Homogenous 96-plex PEA immunoassay exhibiting high sensitivity, specificity, and excellent scalability. Plos One 9, e95192, doi:10.1371/journal.pone.0095192 (2014).

  63. 63.

    Storey, J. D. A direct approach to false discovery rates. J Roy Stat Soc B 64, 479–498 (2002).

  64. 64.

    Kramer, A., Green, J., Pollard, J. Jr. & Tugendreich, S. Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics 30, 523–530, doi:10.1093/bioinformatics/btt703 (2014).

Download references


The authors greatly appreciate Craig Solid, Solid Research Group, LLC for facilitation of the writing of this manuscript. We also appreciate the assistance by personnel at the Center for Mass Spectrometry and Proteomics, especially LeeAnn Higgins and Todd Markowski for MS analysis. GalaxyP is maintained by the Minnesota Supercomputing Institute at University of Minnesota and the infrastructure is supported by National Science Foundation grant 1147079. This project was funded by American Heart Association Scientist Development Grant (12SDG8160000, M.B PI) and the University of Minnesota CTSI KL2 Scholars Program (KL2 RR0333182, M.B.).

Author information


  1. Division of Pulmonary, Allergy, Critical Care and Sleep Medicine, University of Minnesota Medical School, Minneapolis, USA

    • Maneesh Bhargava
    • , Kevin Viken
    • , Peter Bitterman
    • , David Ingbar
    •  & Chris Wendt
  2. Biostatistical Design and Analysis Center, Clinical and Translational Science Institute, University of Minnesota, Minneapolis, USA

    • Qi Wang
  3. Biochemistry, Molecular Biology and Biophysics, University of Minnesota Medical School, Minneapolis, USA

    • Pratik Jagtap
  4. Minneapolis VAMC, Minneapolis, USA

    • Chris Wendt


  1. Search for Maneesh Bhargava in:

  2. Search for Kevin Viken in:

  3. Search for Qi Wang in:

  4. Search for Pratik Jagtap in:

  5. Search for Peter Bitterman in:

  6. Search for David Ingbar in:

  7. Search for Chris Wendt in:


M.B. was involved in the study design, data analysis and interpretation, and manuscript writing. K.V. processed BALF samples and galaxyP data analysis, P.D.J. performed galaxyP analysis, and Q.W. performed data and statistical analysis. P.B.B. and D.I. assisted in data interpretation and editing of the manuscript. C.H.W. was involved with the study design, data interpretation, and editing of the manuscript.

Competing Interests

The authors declare that they have no competing interests.

Corresponding author

Correspondence to Maneesh Bhargava.

Electronic supplementary material


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Creative Commons BY

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit