Advertisement
Research Article

Sputum Biomarkers and the Prediction of Clinical Outcomes in Patients with Cystic Fibrosis

  • Theodore G. Liou mail,

    ted.liou@utah.edu

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Frederick R. Adler,

    Affiliations: Department of Mathematics, University of Utah, Salt Lake City, Utah, United States of America, Department of Biology, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Ruth H. Keogh,

    Affiliation: Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, United Kingdom

    X
  • Yanping Li,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Judy L. Jensen,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • William Walsh,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Kristyn Packer,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Teresa Clark,

    Affiliation: Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Holly Carveth,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Jun Chen,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    Current address: Johnson & Johnson Pharmaceutical Research and Development, LLC, Spring House, Pennsylvania, United States of America

    X
  • Shaunessy L. Rogers,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    Current address: Department of Medicine, Oregon Health and Science University, Portland, Oregon, United States of America

    X
  • Christen Lane,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America, Intermountain Adult CF Center, University of Utah, Salt Lake City, Utah, United States of America

    X
  • James Moore,

    Affiliation: Department of Mathematics, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Anne Sturrock,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Robert Paine III,

    Affiliations: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America, Division of Respiratory, Critical Care and Occupational Pulmonary Medicine, University of Utah, Salt Lake City, Utah, United States of America

    X
  • David R. Cox,

    Affiliation: Department of Statistics and Nuffield College, Oxford, United Kingdom

    X
  • John R. Hoidal

    Affiliation: Department of Medicine, School of Medicine, University of Utah, Salt Lake City, Utah, United States of America

    X
  • Published: August 10, 2012
  • DOI: 10.1371/journal.pone.0042748

Abstract

Lung function, acute pulmonary exacerbations (APE), and weight are the best clinical predictors of survival in cystic fibrosis (CF); however, underlying mechanisms are incompletely understood. Biomarkers of current disease state predictive of future outcomes might identify mechanisms and provide treatment targets, trial endpoints and objective clinical monitoring tools. Such CF-specific biomarkers have previously been elusive. Using observational and validation cohorts comprising 97 non-transplanted consecutively-recruited adult CF patients at the Intermountain Adult CF Center, University of Utah, we identified biomarkers informative of current disease and predictive of future clinical outcomes. Patients represented the majority of sputum producers. They were recruited March 2004-April 2007 and followed through May 2011. Sputum biomarker concentrations were measured and clinical outcomes meticulously recorded for a median 5.9 (interquartile range 5.0 to 6.6) years to study associations between biomarkers and future APE and time-to-lung transplantation or death. After multivariate modeling, only high mobility group box-1 protein (HMGB-1, mean = 5.84 [log ng/ml], standard deviation [SD] = 1.75) predicted time-to-first APE (hazard ratio [HR] per log-unit HMGB-1 = 1.56, p-value = 0.005), number of future APE within 5 years (0.338 APE per log-unit HMGB-1, p<0.001 by quasi-Poisson regression) and time-to-lung transplantation or death (HR = 1.59, p = 0.02). At APE onset, sputum granulocyte macrophage colony stimulating factor (GM-CSF, mean 4.8 [log pg/ml], SD = 1.26) was significantly associated with APE-associated declines in lung function (−10.8 FEV1% points per log-unit GM-CSF, p<0.001 by linear regression). Evaluation of validation cohorts produced similar results that passed tests of mutual consistency. In CF sputum, high HMGB-1 predicts incidence and recurrence of APE and survival, plausibly because it mediates long-term airway inflammation. High APE-associated GM-CSF identifies patients with large acute declines in FEV1%, possibly providing a laboratory-based objective decision-support tool for determination of an APE diagnosis. These biomarkers are potential CF reporting tools and treatment targets for slowing long-term progression and reducing short-term severity.

Introduction

Cystic fibrosis (CF) is the most common lethal genetic disorder among Caucasians, but disease occurs world-wide. Approximately 10 million Americans carry mutations while 25,000 suffer actual disease [1]. Pancreatic enzyme therapy increased survival from infancy to 6 years of age and shifted the commonest cause of death from starvation to lung disease, which remains life-limiting for most patients [1], [2]. Median predicted survival now approaches 40 years [1] but remains well short of normal.

Therapeutic successes have greatly increased CF survival, however, this makes survivorship problematic for assessing treatment effects, prompting a search for alternative measurements [2][4]. Lung function, acute pulmonary exacerbations (APE) of CF requiring hospitalization and weight best predict survivorship clinically [2] but may themselves require large trial enrollments and years of observation to discern treatment effects.

In CF, airway inflammation underlies pulmonary deterioration, frequent APE and persistent malnutrition [5], [6]. Chronic airway infections incite intense, dysfunctional inflammation [5], [7], eliciting multiple deleterious signals [6]. However, these inflammatory signals may serve as immediate biomarkers of disease state. A biomarker that additionally predicts clinical outcomes might identify causal mechanisms for airway disease, pinpoint investigational targets, rapidly detect efficacy of new therapies, and distinguish patients most needing urgent interventions [6], [8]. Such biomarkers have previously been elusive in CF.

We hypothesized that high-mobility group box-1 protein (HMGB-1), a novel inflammatory cytokine, might be informative of future clinical events in CF. HMGB-1 is a highly conserved DNA-shepherding protein. However, when released by necrosis or secretion, HMGB-1 acts as a potent cytokine associated with delayed onset of prolonged inflammation [9], [10] and septic shock in animal models reversible with anti-HMGB-1 antibodies [11]. In this work, we investigated the potential of sputum HMGB-1 measurements to reflect concurrent clinical status and predict future outcomes following a physician-patient encounter [8].

Methods

Ethics Statement

This study was reviewed and approved by the University of Utah Investigational Review Board prior to enrollment of any patients. All participants were included only after obtaining written informed consent for sputum collection and analysis. Collection of information for the CF Foundation Patient Registry (CFFPR) was reviewed and approved at each participating institution and by the University of Utah Investigational Review Board. Our use of data from the CFFPR was also reviewed and approved by the CF Foundation. All adult patients included in the CFFPR gave written informed consent. All pediatric patients in the CFFPR gave assent if able to understand, and their parents or guardians gave written informed consent and permission to participate. The study was conducted under the supervision of the physicians involved (TGL, WW, HC, RP III, JRH), and patients were advised of risks, benefits and the right to withdraw from further involvement in the study at any point without repercussions. All data, particularly patient identifying data, were physically and electronically secured throughout the study.

Study Outline

From March, 2004 through May, 2011, we recruited sputum-producing outpatients from the Intermountain Adult CF Center at quarterly-regular and unplanned-sick visits. All visits were conducted in compliance with CF Foundation guidelines for outpatient care [12]. We collected sputum, recorded same-day lung function, weight, height, sputum culture results, number of prior-year APE, and recorded whether patients were in stable, mild exacerbation or APE states. We actively followed patients to lung transplantation, death, loss-to-follow-up, or May 31, 2011, the end of study, recording APE numbers and spirometry results [13]. Patients recruited from March, 2004 through September 2006 formed the study groups while remaining patients formed validation groups (Tables 1 and 2). Forced expiratory volume in 1 second (FEV1) was measured [13] and normalized to percent predicted FEV1 (FEV1%) [14]. We assessed generalizability of results by comparing our patients with sputum-producing adults in the 2006 CFFPR.

thumbnail

Table 1. Study Designa.

doi:10.1371/journal.pone.0042748.t001
thumbnail

Table 2. Analysis Descriptions.

doi:10.1371/journal.pone.0042748.t002

Acute Pulmonary Exacerbations

We diagnosed APE and hospitalized for symptoms and objective evidence of severe acute worsening of CF. Symptoms included increased cough, sputum production or dyspnea, chest pain or tightness, hemoptysis, fever, chills, arthralgias or decreased exercise tolerance. Patients had to meet one objective criterion: 10% decrease in FEV1% or percent predicted forced vital capacity, fever above 38.4°C, documented hemoptysis greater than 100 ml per episode, SaO2 below 90% or PaO2 below 60 mm Hg despite usual oxygen, increased supplemental oxygen requirements or three or more kg unplanned drop in weight within 3 months. Symptomatic patients without objective findings were classified as mild exacerbation and not hospitalized. These criteria were modified from previous APE definitions to allow prospective application at a patient encounter by excluding retrospective criteria such as prior antibiotic treatment [2], [4], [15].

Sputum Evaluations

Sputum samples underwent standardized processing (see Text S1). CF Foundation guidelines were followed to collect sputum cultures. Cultures were obtained within 6 months of sputum collections as bacterial infections are typically stable over such a period [16], a finding that has recently been supported by culture-independent methods [17]. We measured potential biomarkers previously identified as important in CF including granulocyte macrophage colony stimulating factor (GM-CSF) [6], [8], [18][25] using SearchLight multiplex assay services (Aushon Biosystems, Billerica, MA). We measured HMGB-1 by ELISA using commercially-available antibodies (R&D Systems, Inc, Minneapolis, MN) and previously published protocols [26] after confirming assay reproducibility on consecutive days (see Text S1). Biomarker measurements reported are log-scale.

Statistical Analysis

Through 5 analyses [27], we examined (1) associations between biomarkers and concurrent clinical disease measurements, (2) APE effects on biomarkers, (3) biomarker predictions of clinical outcomes using linear, logistic and quasi-Poisson regression [28], and proportional hazards modeling [29], [30] (See Tables 1 and 2 and Text S1 for details). Throughout this paper, p-values smaller than 0.05 are reported as statistically significant. In Analyses 1–3, we performed multiple univariate tests for associations between biomarker values and various clinical measurements, and we used stringent Bonferroni correction by dividing 0.05 by the number of tests performed to define the corrected p-value required for significance. We repeated Analyses 3–5 using sputum measurements from 4 sets of validation patients chosen so that no patient appeared simultaneously in a study group and comparison validation groups (Table 1). Mutual consistency of results was assessed by a weighted least squares analysis.

Results

Patients and Samples

During enrollment of the three study groups, 161 patients attended clinic at least once; 26 produced no sputum. Including validation groups, we recruited 97 patients and collected 149 samples: 56 samples were collected during stable, 31 during mild exacerbation and 62 during APE clinical states (Table 3 and Text S1). Every sputum-producing patient approached agreed to participate, but we were occasionally limited by availability of study personnel or by temporary lack of sputum production. All patients were lifetime non-smokers. Study patients were nearly indistinguishable from adult sputum producers in the 2006 CF Foundation Patient Registry (Table 4). Our patients had a weight-for-age z-score 0.24 units less than the national median and used oral steroids less frequently.

thumbnail

Table 3. Patient Characteristicsa.

doi:10.1371/journal.pone.0042748.t003

Study group patients were followed up to 7.1 years (median 5.9, interquartile range 5.0 to 6.6) through death, lung transplantation, loss to follow-up or end of study. All APE hospitalizations occurred within our center. Nine deaths occurred between 0.6 and 6.5 years, and six lung transplants between 1 and 6.3 years after enrollment. Three patients were lost to follow-up 2.7, 3.4 and 3.6 years after enrollment when they moved.

thumbnail

Table 4. Patient Comparisons with the 2006 CF Foundation Patient Registrya.

doi:10.1371/journal.pone.0042748.t004

Biomarkers were measured (Table S1) in three overlapping patient study groups after sample inventory demonstrated feasibility of planned analyses (Table 1). Group 1: the initial sputum sample from each of 80 randomly selected patients underwent SearchLight multiplex assays (Aushon); results for 24 were discarded due to unreliability (as samples thawed in transport to Aushon), leaving results from 56 patients for analysis. Group 2: we measured HMGB-1 by ELISA and obtained a second SearchLight multiplex assay (Aushon) on samples from the 26 patients that during repeated collections resulted in a pair of samples per patient: one from an APE and one from a stable point adjacent in time. Ten gave a clinically-stable-time-point sample first; 16 gave an APE-time-point sample first. The median number of days from stable to APE sample was 98 (interquartile range 91–154) and from APE to stable sample was 119 (interquartile range 84–212). Group 3: 76 patients from Group 1 had sufficient sample volumes to measure HMGB-1.

Analysis 1: Biomarker Measurements and Concurrent Clinical Status

We looked for associations with concurrent clinical findings to identify plausible biomarkers for CF airway inflammation (Table S2). Backwards model selection using biomarkers with statistically significant or borderline univariate associations (p<0.1) produced three final multivariate models. HMGB-1 and interleukin-(IL)-17A are associated with concurrent FEV1% (Figure 1A and S1A), IL-17A with weight-for-age z-score (Figure S1B), and HMGB-1 and interferon-(IFN)-α with number of prior-year APE (Figure 1B, Figure S1C, Table 5). (Discussion of these results can be found in Text S1.).

thumbnail

Figure 1. Key Univariate Relationships.

HMGB-1 had strong statistically significant associations with A) concurrent FEV1% and B) number of APE suffered in the year prior sputum sample collection. These results illustrate the immediate clinical relevance of HMGB-1. C) GM-CSF measured at APE time-points had an extremely strong univariate association with the size of the APE-associated decline in FEV1% for each of 26 patients in Group 2. D) Although the univariate relationship with APE-associated FEV1% decline was weak (Table S3), the addition of IL-5 as a covariate to GM-CSF significantly strengthened the multivariate linear regression model of APE-associated decline in FEV1% (Table 5).

doi:10.1371/journal.pone.0042748.g001
thumbnail

Table 5. Multivariate models for concurrent outcomes and APE-associated predictions.

doi:10.1371/journal.pone.0042748.t005

Analysis 2: Biomarker Behavior with an APE

We looked for biomarkers that changed in association with a clinical change from stable to APE state. For 19 of 22 biomarkers measured for Study Group 2 patients, the mean change between stable and APE states (Table S1) was positive indicating a non-random inflammatory state change (sign test, p = 0.006). C-reactive protein (CRP), IL-1β and IFN-α were statistically significantly higher (paired t-test, p<0.05) during an APE (Figure S1D-F). We investigated a potential effect of sample collection sequences (stable-then-APE and APE-then-stable) in sensitivity analyses. Mean IFN-α change was greater for the sequence stable-then-APE (t-test, p = 0.015). No sample-sequence-associated differences were found for CRP or IL-1β. (Discussion of these results can be found in Text S1.).

Analysis 3: Relationship of Biomarkers with APE-associated Clinical Changes

For the 26 Study Group 2 patients, mean FEV1% decline between stable and APE time points was 7.3 (SD 7.8) percentage points; 5 had no decline. Univariate linear regressions found statistically significant inverse associations between FEV1% decline and GM-CSF, IL-23, CRP and sputum total cell count (TCC) each measured at APE time points and the difference in IL-17 measurements at stable and APE time points (Table S3). Neither biomarker measurements at stable time points nor any other biomarker differences between stable and APE time points were associated with FEV1% decline. After backward selection of multivariate models, GM-CSF and to a lesser extent IL-5, both measured at APE time points, were significantly and independently associated with APE-associated decline in FEV1% (Figure 1C–D, Table 5).

Analysis 4: Predicting Future APE

Increased HMGB-1, adjusted for number of prior-year APE and follow-up time, was retained after backwards and forwards selection in a final multivariate quasi-Poisson regression model for the total number of APE occurring within 5 years of clinically-stable-time-point samples (Table 5). All other clinical measurements including P aeruginosa and S aureus infections, FEV1%, 5-year predicted survival [2] and other potential biomarkers measured at the stable-time point were not statistically significant predictors when forced into the final multivariate model (p = 0.23 for FEV1%; p>0.05 for all other covariates).

We have complete follow-up from clinically-stable time points to first APE, enabling proportional hazards modeling of time-to-first APE [29][31]. Univariate and multivariate models found HMGB-1 alone predicts time-to-first APE (Table 6, Figure 2A). Because of this result, we explored using median HMGB-1 level as a clinical test for predicting an APE within 6 months and found favorable preliminary results (see Text S1 and Figure S2).

thumbnail

Table 6. Proportional Hazards Models of Time-to-Eventa.

doi:10.1371/journal.pone.0042748.t006
thumbnail

Figure 2. Kaplan-Meier Curves for the Time from Stable Sputum Collection to First Event.

The curves illustrate the difference in time to A) first APE and B) death or censoring by listing for lung transplantation for patients with HMGB-1 measurements higher and lower than the value of 6.0 (log ng/ml). The value is the rounded median of the actual HMGB-1 data for both the 26 patients in A and the 76 patients in B. P-values shown are the results of log rank testing [52]. These graphs show the results of evaluation of HMGB-1 simplified to high or low values, which are consistent with the proportional hazards modeling [29] of the effects of HMGB-1 as a continuous variable. Models were tested for consistency with proportionality [31] (Table 6).

doi:10.1371/journal.pone.0042748.g002

Analysis 5: HMGB-1 as an Independent Predictor of Lung Transplantation or Death

Because of multiple other statistically significant associations, we assessed HMGB-1 ability to predict time-to-lung transplantation or death [29][31]. The composite outcome avoids 2 biases: 1) in a study of time-to-death, transplantation of end-stage patients creates an informative censoring bias and 2) in a study of time-to-lung transplantation, use of FEV1% to select transplantation candidates confounds the analysis. However, the composite outcome would still be confounded by modeling factors used to select transplantation candidates [2], [32], [33], thus potential covariates including FEV1% and prior-year APE were necessarily excluded from analysis. Other potential covariates including P aeruginosa and S aureus infection status do not alter lung transplantation candidacy and were included in the analysis. Follow-up time for the composite event was time-to-listing for transplantation, death, loss-to-follow-up or study end. Three patients lost-to-follow-up were treated as alive at the time of censoring. After multivariate proportional hazards modeling with backward and forward selection, we found that higher HMGB-1 values alone are strongly associated with shorter time-to-transplantation or death (Table 6, Figure 2B).

Validation

Analyses 3–5 were repeated using data from four validation patient groups (Tables 1 and 2) to validate predictive abilities of HMGB-1 and GM-CSF. All validation results were similar to primary results. We evaluated results for mutual consistency using a weighted least-squares analysis and found that differences in coefficients and hazard ratios were insufficiently discrepant to cause concern (Table 7).

thumbnail

Table 7. Testing Validation Results for Mutual Consistency.

doi:10.1371/journal.pone.0042748.t007

Discussion

Sputum HMGB-1 measurements predicted lung transplantation or death for CF patients during 7.1 years of follow-up (Table 6, Figure 2B). HMGB-1 measured during clinical stability surpassed FEV1% and all other biomarkers alone or in combination in ability to predict number of subsequent APE (Table 5) and time-to-first APE (Table 6, Figure 2A). Simultaneously, HMGB-1 measurements possess immediate clinical relevance by demonstrating strong independent associations with concurrent FEV1% and number of prior-year APE (Table 5) [2].

Our results confirm published findings from 32 CF patients that HMGB-1 inversely correlates with FEV1% and has an association with patients that suffer an APE. The earlier statistical findings were limited to concurrent clinical status [34]. Our study substantially extends these results finding evidence that HMGB-1 predicts future clinical events including survival (Tables 5 and 6).

HMGB-1 is required for normal cellular function but when released due to necrosis or secretion [9], [10] plays key roles in inflammatory arthritis [35], sepsis [11], and acute lung injury [36]. It mediates endotoxin-associated inflammation [10] through RAGE that is upregulated specifically in CF airways [7] and other receptors [37]. In rat models, HMGB-1-blocking antibodies reversed arthritis [35] and produced remarkably improved, durable survival despite delayed intervention following induction of septic shock by cecal ligation and puncture [10]. Increased HMGB-1 is associated with human acute lung injury [36] and sustained COPD-related inflammation [38]. In CF, HMGB-1 may increase pulmonary inflammation by attracting neutrophils [34], preventing their efferocytosis [39] and amplifying the effects of bacterial lipopolysaccharide and cytosine-phosphatidyl-guanosine-DNA constructs [40]. Soluble RAGE, an antagonist to HMGB-1-RAGE mediated inflammation, is absent in CF airways, potentiating HMGB-1 effects [7]. Our successful predictions (Tables 5 and 6, Figure 2) strengthen the argument that HMGB-1 may play a causal role in human inflammatory lung disease.

Low lung function shortens CF survival [2]. Sharp FEV1% declines define APE for most patients and may presage permanent lung function loss [5]. In this study, several biomarkers had significant univariate associations with acute APE-associated FEV1% decline, but GM-CSF and to a lesser extent IL-5 measured at APE time points together had the strongest association (Figure 1C–D, Table 5). These associations were independent of FEV1% itself and either cytokine measured during clinical stability. Our findings reflect previously known relationships in structure, expression and function of GM-CSF and IL-5, including the strong correlation between their measurements (Table S4) and the observation that one cytokine, in this case GM-CSF, is often more prominent than the other [41]. For GM-CSF at an APE time point, an approximate 1% increase was associated with a 10 point decline from the clinically-stable FEV1%. A patient’s GM-CSF measurement may quantify APE severity independently of prior clinical state and might provide laboratory-based objective support for hospitalization decisions.

GM-CSF maintains normal alveolar macrophage and innate immune responses to acute P aeruginosa pneumonia in mice [42]. In humans, elevated sputum GM-CSF is associated with increased COPD and asthma severity [43]. However, its role in chronically infected human airways is unclear. P aeruginosa or S aureus in CF airways [1] elicit increased epithelial GM-CSF secretion leading to prolonged survival and decreased apoptosis of airway neutrophils [44] and may prolong protease and reactive oxygen species releases [2].

Our definition of APE, based on experience and bedside utility, played a central role in defining events, sample collections and analysis of our major results with HMGB-1 and GM-CSF. It resembles prior definitions [2], [4], [15], but differs by excluding retrospective criteria such as a prior decision to give antibiotics [4] and by exchanging quantitative scoring [15] for bedside facility. Nevertheless our definition includes the key covariates of the existing quantitative model [15].

Our study focused on adult CF patients with moderately severe disease (Table 3), thus it provides no information easily extrapolated to children or mild or end-stage adult patients. However, we enrolled more than half of our large adult CF center and two-thirds of sputum producers. Our patients were nearly indistinguishable from the 70% of 2006 CFFPR adult patients nationally that produced sputum, a large, identifiable group that would benefit from new insights and therapies (Table 4) [1]. Volume of sputum collection could not be compared to the national population but had no effect in any of our analyses (not shown).

HMGB-1 was the key biomarker identified in multivariate models, proving to be superior to IL-8 and other biomarkers of previously established interest, indicating that HMGB-1 measurement alone is sufficient to report the information encoded in the other biomarkers evaluated (Table S1). However, we could not study all potentially significant biomarkers. To avoid compromising sample collection (see Text S1) [18], we excluded proteinases such as neutrophil elastase [45], [46], matrix metalloproteinase [47], [48] and proteinase-3 [49], [50] which may all cause lung tissue injury. Thus further studies are needed to explore the relative clinical predictive ability of these and still other biomarkers.

Our HMGB-1 and GM-CSF results successfully validated (Table 7) with additional but greatly dissimilar patients (Table 3). In every case, results were unaffected by commonly-prescribed potential modifiers of inflammation such as azithromycin and inhaled steroid therapies or by airway infections with Pseudomonas aeruginosa or Staphylococcus aureus. Nevertheless, this was still a single center study that was exploratory in nature, and collection of additional high-quality data from a larger cohort is needed to confirm our findings and explore directions of causality [51].

It is not feasible to include FEV1% or several other clinical covariates with HMGB-1 to predict time-to-lung transplantation or death due to biases from confounding and informative censoring. These biases arise because the same clinical covariates are used for selection of lung transplantation candidates. Prediction of either transplant or death alone is equally not feasible due to the same biases. Using a combined endpoint of lung transplant or death and excluding confounding variables avoided these biases but assumed that our clinical decisions [32] were accurate in identifying patients at highest risk of death [2]. These limitations are not specific to our study size, population or design. Rather, they apply to any population where lung transplantation is utilized. A positive byproduct results from avoiding these biases: this study focuses on causal pathways for airway inflammation leading to end-stage lung disease and avoids confusion from consideration of intermediate clinical findings like FEV1% that may not be causal [51].

This study identified airway-inflammatory biomarkers reporting on distinct pathophysiologic states that precede, accompany and follow an APE and provided clinically relevant information about permanent lung function losses on short and early mortality on long time scales. GM-CSF and HMGB-1 provide potentially useful new measurements for monitoring short and longer term treatment effects for CF. Both biomarkers provide focal points for additional clinical, epidemiologic and mechanistic investigations of CF airway inflammation. In particular, HMGB-1 provides pathophysiologic information predictive of APE and survival that might guide preemptive treatment and provide a novel target for therapy.

Supporting Information

Figure S1.

Biomarker Relationships with Concurrent Clinical Outcomes and Biomarker Changes between Stable and APE States. Relationships between A) IL-17A and concurrent FEV1%, B) IL-17A and weight-for-age z-score and C) IFN-α with prior year APE. Three biomarkers, D) CRP, E) IL-1β and F) IFN-α are increased in sputum collected at the time of diagnosis of an APE compared to their measurements during clinical stability using paired t-tests. Red symbols illustrate values for patients that underwent the sputum collection sequence stable-then-APE while blue symbols denote patients that underwent the sequence APE-then-stable. The number of days between stable and APE sputum collections had no effect on differences on measurements (multiple linear regression p>0.5). However, changes in these three biomarker measurements are unsuitable for clinical use to identify an APE due to substantial overlap of APE and stable time point values.

doi:10.1371/journal.pone.0042748.s001

(EPS)

Figure S2.

Receiver operator characteristic curves for potential markers of an APE within 6 months. Receiver operator characteristic curves were constructed for 9 biological or clinical markers of CF using APE within 6 months as the clinical discriminator. Log-scale biomarker values were derived from measurements of sputum samples from 26 patients during clinical stability (Study Group 2, Tables 12). Each plot has a color bar presenting the range of cut-off values of each biomarker used to create the points on each ROC curve. The area under the curve (AUC or accuracy) and 95% confidence interval for each AUC reported for each curve was the result of boot strapping of the AUC using 1,000 repetitions. HMGB-1 has the highest accuracy of any marker studied.

doi:10.1371/journal.pone.0042748.s002

(EPS)

Table S1.

Sputum Biomarker Measurements.

doi:10.1371/journal.pone.0042748.s003

(DOC)

Table S2.

Univariate Associations of Biomarkers with Clinically Relevant Concurrent Conditions.

doi:10.1371/journal.pone.0042748.s004

(DOC)

Table S3.

Univariate Associations of Biomarkers for APE-associated FEV1% drop.

doi:10.1371/journal.pone.0042748.s005

(DOC)

Table S4.

Correlations (p-values) between biomarkers, Study Group 1.

doi:10.1371/journal.pone.0042748.s006

(DOC)

Text S1.

Detailed Laboratory and Statistical Methods, Additional Results and Discussion.

doi:10.1371/journal.pone.0042748.s007

(DOC)

Acknowledgments

The authors thank Barbara Cahill, MD, Greg Elliott, MD and Wayne Samuelson, MD, all faculty at the University of Utah, who reviewed early versions of the manuscript and provided valuable critiques without compensation. We thank Hebe Quinton, MS, Dartmouth College, for technical assistance obtaining specific data about the Intermountain Adult CF Center from the US CFFPR for 2006 without compensation.

The national data on CF patients reported were supplied by the U.S. Cystic Fibrosis Foundation. The authors are responsible for interpretation and reporting of the CFFPR data and do not express opinions or recommendations for the Foundation.

Author Contributions

Conceived and designed the experiments: TGL FRA JM RP JRH. Performed the experiments: TGL YL JLJ WW KP TC HC JC SLR CL AS. Analyzed the data: TGL FRA RHK WW JM RP DRC JRH. Contributed reagents/materials/analysis tools: TGL FRA RHK WW DRC. Wrote the paper: TGL FRA RHK WW JRH DRC.

References

  1. 1. Cystic Fibrosis Foundation Patient Registry 2010 Annual Data Report (2011) Bethesda, Maryland: Cystic Fibrosis Foundation. Available: http://www.cff.org/LivingWithCF/CareCent​erNetwork/PatientRegistry/. Accessed 21 Jun 2012.
  2. 2. Liou TG, Adler FR, Fitzsimmons SC, Cahill BC, Hibbs JR, et al. (2001) Predictive 5-year survivorship model of cystic fibrosis. Am J Epidemiol 153: 345–52 doi:10.1093/aje/153.4.345.
  3. 3. Saiman L, Marshall BC, Mayer-Hamblett N, Burns JL, Quittner AL, et al. (2003) Azithromycin in patients with cystic fibrosis chronically infected with Pseudomonas aeruginosa: a randomized controlled trial. JAMA 290: 1749–1756 doi:10.1001/jama.290.13.1749.
  4. 4. Fuchs HJ, Borowitz DS, Christiansen DH, Morris EM, Nash ML, et al. (1994) Effect of Aerosolized Recombinant Human DNase on Exacerbations of Respiratory Symptoms and on Pulmonary Function in Patients with Cystic Fibrosis. N Engl J Med 331: 637–642 doi:10.1056/NEJM199409083311003.
  5. 5. Gibson RL, Burns JL, Ramsey BW (2003) Pathophysiology and management of pulmonary infections in cystic fibrosis. Am J Respir Crit Care Med 168: 918–51 doi:10.1164/rccm.200304-505SO.
  6. 6. Bonfield TL, Panuska JR, Konstan MW, Hilliard KA, Hilliard JB, et al. (1995) Inflammatory cytokines in cystic fibrosis lungs. Am J Respir Crit Care Med 152: 2111–8. doi: 10.1164/ajrccm.152.6.8520783
  7. 7. Makam M, Diaz D, Laval J, Gernez Y, Conrad CK, et al. (2009) Activation of critical, host-induced, metabolic and stress pathways marks neutrophil entry into cystic fibrosis lungs. Proc. Natl. Acad. Sci. U.S.A. 106: 5779–5783 doi:10.1073/pnas.0813410106.
  8. 8. Sagel SD, Chmiel JF, Konstan MW (2007) Sputum biomarkers of inflammation in cystic fibrosis lung disease. Proc Am Thorac Soc 4: 406–417 doi:10.1513/pats.200703-044BR.
  9. 9. Scaffidi P, Misteli T, Bianchi ME (2002) Release of chromatin protein HMGB1 by necrotic cells triggers inflammation. Nature 418: 191–195 doi:10.1038/nature00858.
  10. 10. Wang H, Bloom O, Zhang M, Vishnubhakat JM, Ombrellino M, et al. (1999) HMG-1 as a late mediator of endotoxin lethality in mice. Science 285: 248–251 doi:10.1126/science.285.5425.248.
  11. 11. Yang H, Ochani M, Li J, Qiang X, Tanovic M, et al. (2004) Reversing established sepsis with antagonists of endogenous high-mobility group box 1. Proc Natl Acad Sci U S A 101: 296–301 doi:10.1073/pnas.2434651100.
  12. 12. Cystic Fibrosis Foundation (1997) Clinical practice guidelines for cystic fibrosis. Bethesda, Maryland.
  13. 13. Lung function testing: selection of reference values and interpretative strategies. American Thoracic Society (1991) Am Rev Respir Dis. 144: 1202–18. doi: 10.1164/ajrccm/144.5.1202
  14. 14. Hankinson JL, John R Odencrantz, Kathleen B Fedan (1999) Spirometric reference values from a sample of the general U.S. population. Am. J. Respir. Crit. Care Med. 159: 179–187. doi: 10.1164/ajrccm.159.1.9712108
  15. 15. Rosenfeld M, Emerson J, Williams-Warren J, Pepe M, Smith A, et al. (2001) Defining a pulmonary exacerbation in cystic fibrosis. J Pediatr 139: 359–365 doi:10.1067/mpd.2001.117288.
  16. 16. Aaron SD, Ramotar K, Ferris W, Vandemheen K, Saginur R, et al. (2004) Adult cystic fibrosis exacerbations and new strains of Pseudomonas aeruginosa. Am. J. Respir. Crit. Care Med 169: 811–815 doi:10.1164/rccm.200309-1306OC.
  17. 17. Stressmann FA, Rogers GB, van der Gast CJ, Marsh P, Vermeer LS, et al. (2012) Long-term cultivation-independent microbial diversity analysis demonstrates that bacterial communities infecting the adult cystic fibrosis lung show stability and resilience. Thorax. Available: http://www.ncbi.nlm.nih.gov/pubmed/22707​521. Accessed 20 Jun 2012.
  18. 18. Kelly MM, Leigh R, Carruthers S, Horsewood P, Gleich GJ, et al. (2001) Increased detection of interleukin-5 in sputum by addition of protease inhibitors. Eur. Respir. J 18: 685–691. doi: 10.1183/09031936.01.00098501
  19. 19. Moser C, Jensen PØ, Pressler T, Frederiksen B, Lanng S, et al. (2005) Serum concentrations of GM-CSF and G-CSF correlate with the Th1/Th2 cytokine response in cystic fibrosis patients with chronic Pseudomonas aeruginosa lung infection. APMIS 113: 400–409 doi:10.1111/j.1600-0463.2005.apm_142.x.
  20. 20. Hartl D, Griese M, Kappler M, Zissel G, Reinhardt D, et al. (2006) Pulmonary T(H)2 response in Pseudomonas aeruginosa-infected patients with cystic fibrosis. J. Allergy Clin. Immunol 117: 204–211 doi:10.1016/j.jaci.2005.09.023.
  21. 21. Drumm ML, Konstan MW, Schluchter MD, Handler A, Pace R, et al. (2005) Genetic modifiers of lung disease in cystic fibrosis. N Engl J Med 353: 1443–1453 doi:10.1056/NEJMoa051469.
  22. 22. Bruscia EM, Zhang P-X, Ferreira E, Caputo C, Emerson JW, et al. (2009) Macrophages directly contribute to the exaggerated inflammatory response in cystic fibrosis transmembrane conductance regulator−/− mice. Am. J. Respir. Cell Mol. Biol 40: 295–304 doi:10.1165/rcmb.2008-0170OC.
  23. 23. Hubeau C, Le Naour R, Abély M, Hinnrasky J, Guenounou M, et al. (2004) Dysregulation of IL-2 and IL-8 production in circulating T lymphocytes from young cystic fibrosis patients. Clin. Exp. Immunol 135: 528–534. doi: 10.1111/j.1365-2249.2003.02385.x
  24. 24. Yang Y, Trinchieri G, Wilson JM (1995) Recombinant IL-12 prevents formation of blocking IgA antibodies to recombinant adenovirus and allows repeated gene therapy to mouse lung. Nat. Med 1: 890–893. doi: 10.1038/nm0995-890
  25. 25. McAllister F, Henry A, Kreindler JL, Dubin PJ, Ulrich L, et al. (2005) Role of IL-17A, IL-17F, and the IL-17 receptor in regulating growth-related oncogene-alpha and granulocyte colony-stimulating factor in bronchial epithelium: implications for airway inflammation in cystic fibrosis. J. Immunol 175: 404–412.
  26. 26. Davé SH, Tilstra JS, Matsuoka K, Li F, DeMarco RA, et al. (2009) Ethyl pyruvate decreases HMGB1 release and ameliorates murine colitis. Journal of Leukocyte Biology 86: 633–643 doi:10.1189/jlb.1008662.
  27. 27. R Development Core Team (2010) R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available: http://www.R-project.org.
  28. 28. Venables WN, Ripley BD (2010) Modern Applied Statistics with S. Springer New York.
  29. 29. Cox DR (1972) Regression models and life-tables. J R Stat Soc B Met 34: 187–220 doi:10.2307/2985181.
  30. 30. Cox DR, Oakes D (1984) Analysis of survival data. CRC Press.
  31. 31. Grambsch PM, Therneau TM (1994) Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 81: 515–526 doi:10.1093/biomet/81.3.515.
  32. 32. Liou TG, Adler FR, Cahill BC, Fitzsimmons SC, Huang D, et al. (2001) Survival effect of lung transplantation among patients with cystic fibrosis. JAMA 286: 2683–9. doi: 10.1001/jama.286.21.2683
  33. 33. Liou TG, Adler FR, Cox DR, Cahill BC (2007) Lung transplantation and survival in children with cystic fibrosis. N Engl J Med 357: 2143–52 doi:10.1056/NEJMoa066359.
  34. 34. Rowe SM, Jackson PL, Liu G, Hardison M, Livraghi A, et al. (2008) Potential role of high-mobility group box 1 in cystic fibrosis airway disease. Am. J. Respir. Crit. Care Med 178: 822–831 doi:10.1164/rccm.200712-1894OC.
  35. 35. Kokkola R, Li J, Sundberg E, Aveberger A-C, Palmblad K, et al. (2003) Successful treatment of collagen-induced arthritis in mice and rats by targeting extracellular high mobility group box chromosomal protein 1 activity. Arthritis Rheum 48: 2052–2058 doi:10.1002/art.11161.
  36. 36. Ueno H, Matsuda T, Hashimoto S, Amaya F, Kitamura Y, et al. (2004) Contributions of high mobility group box protein in experimental and clinical acute lung injury. Am. J. Respir. Crit. Care Med 170: 1310–1316 doi:10.1164/rccm.200402-188OC.
  37. 37. Park JS, Svetkauskaite D, He Q, Kim J-Y, Strassheim D, et al. (2004) Involvement of toll-like receptors 2 and 4 in cellular activation by high mobility group box 1 protein. J. Biol. Chem 279: 7370–7377 doi:10.1074/jbc.M306793200.
  38. 38. Ferhani N, Letuve S, Kozhich A, Thibaudeau O, Grandsaigne M, et al. (2010) Expression of high-mobility group box 1 and of receptor for advanced glycation end products in chronic obstructive pulmonary disease. Am. J. Respir. Crit. Care Med 181: 917–927 doi:10.1164/rccm.200903-0340OC.
  39. 39. Liu G, Wang J, Park Y-J, Tsuruta Y, Lorne EF, et al. (2008) High mobility group protein-1 inhibits phagocytosis of apoptotic neutrophils through binding to phosphatidylserine. J. Immunol. 181: 4240–4246. doi: 10.4049/jimmunol.181.6.4240
  40. 40. Hreggvidsdottir HS, Ostberg T, Wähämaa H, Schierbeck H, Aveberger A-C, et al. (2009) The alarmin HMGB1 acts in synergy with endogenous and exogenous danger signals to promote inflammation. J. Leukoc. Biol 86: 655–662 doi:10.1189/jlb.0908548.
  41. 41. Molfino NA, Gossage D, Kolbeck R, Parker JM, Geba GP (n.d.) Molecular and clinical rationale for therapeutic targeting of interleukin-5 and its receptor. Clinical & Experimental Allergy. Available: http://onlinelibrary.wiley.com/doi/10.11​11/j.1365-2222.2011.03854.x. Accessed 25 Jan 2012.
  42. 42. Ballinger MN, Paine R, Serezani CHC, Aronoff DM, Choi ES, et al. (2006) Role of granulocyte macrophage colony-stimulating factor during gram-negative lung infection with Pseudomonas aeruginosa. Am. J. Respir. Cell Mol. Biol 34: 766–774 doi:10.1165/rcmb.2005-0246OC.
  43. 43. Saha S, Doe C, Mistry V, Siddiqui S, Parker D, et al. (2009) Granulocyte-macrophage colony-stimulating factor expression in induced sputum and bronchial mucosa in asthma and COPD. Thorax 64: 671–676 doi:10.1136/thx.2008.108290.
  44. 44. Saba S, Soong G, Greenberg S, Prince A (2002) Bacterial Stimulation of Epithelial G-CSF and GM-CSF Expression Promotes PMN Survival in CF Airways. Am. J. Respir. Cell Mol. Biol. 27: 561–567 doi:10.1165/rcmb.2002-0019OC.
  45. 45. Carp H, Miller F, Hoidal JR, Janoff A (1982) Potential mechanism of emphysema: alpha 1-proteinase inhibitor recovered from lungs of cigarette smokers contains oxidized methionine and has decreased elastase inhibitory capacity. Proc Natl Acad Sci U S A 79: 2041–5. doi: 10.1073/pnas.79.6.2041
  46. 46. Mayer-Hamblett N, Aitken ML, Accurso FJ, Kronmal RA, Konstan MW, et al. (2007) Association between pulmonary function and sputum biomarkers in cystic fibrosis. Am. J. Respir. Crit. Care Med. 175: 822–828 doi:10.1164/rccm.200609-1354OC.
  47. 47. Hibbs MS, Hoidal JR, Kang AH (1987) Expression of a metalloproteinase that degrades native type V collagen and denatured collagens by cultured human alveolar macrophages. J Clin Invest 80: 1644–50 doi:10.1172/JCI113253.
  48. 48. Sagel SD, Kapsner RK, Osberg I (2005) Induced sputum matrix metalloproteinase-9 correlates with lung function and airway inflammation in children with cystic fibrosis. Pediatr Pulmonol 39: 224–32 doi:10.1002/ppul.20165.
  49. 49. Kao RC, Wehner NG, Skubitz KM, Gray BH, Hoidal JR (1988) Proteinase 3. A distinct human polymorphonuclear leukocyte proteinase that produces emphysema in hamsters. J Clin Invest 82: 1963–73 doi:10.1172/JCI113816.
  50. 50. Rao NV, Wehner NG, Marshall BC, Gray WR, Gray BH, et al. (1991) Characterization of proteinase-3 (PR-3), a neutrophil serine proteinase. Structural and functional properties. J Biol Chem 266: 9540–8.
  51. 51. Cox DR, Wermuth N (2004) Causality: A statistical view. International Statistical Review 72: 285–305 doi:10.1111/j.1751-5823.2004.tb00237.x.
  52. 52. Peto R, Peto J (1972) Asymptotically Efficient Rank Invariant Test Procedures. Journal of the Royal Statistical Society. Series A (General) 135: 185–207 doi:10.2307/2344317ips.