Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Jul 1.
Published in final edited form as: J Cyst Fibros. 2019 Dec 20;19(4):632–640. doi: 10.1016/j.jcf.2019.12.007

UTILIZING CENTRALIZED BIOREPOSITORY SAMPLES FOR BIOMARKERS OF CYSTIC FIBROSIS LUNG DISEASE SEVERITY

Scott D Sagel 1, Brandie D Wagner 2, Assem Ziady 3, Tom Kelley 4, John P Clancy 3, Monica Narvaez-Rivas 5, Joseph Pilewski 6, Elizabeth Joseloff 7, Wei Sha 8, Leila Zelnick 9, Kenneth D R Setchell 5, Sonya L Heltshe 10,11, Marianne S Muhlebach 12
PMCID: PMC7305052  NIHMSID: NIHMS1547172  PMID: 31870630

Abstract

Background

Circulating biomarkers reflective of lung disease activity and severity have the potential to improve patient care and accelerate drug development in CF. The objective of this study was to leverage banked specimens to test the hypothesis that blood-based biomarkers discriminate CF children segregated by lung disease severity.

Methods

Banked serum samples were selected from children who were categorized into two extremes of phenotype associated with lung function (‘mild’ or ‘severe’) based on CF-specific data and were matched on age, gender, CFTR genotype, and P. aeruginosa infection status. Targeted inflammatory proteins, lipids, and discovery metabolite profiles were measured in these serum samples.

Results

The severe cohort, characterized by a lower CF-specific FEV1 percentile, had significantly higher circulating concentrations of high sensitivity C-reactive protein, serum amyloid A, granulocyte colony stimulating factor, and calprotectin compared to the mild cohort. The mild cohort tended to have higher serum linoleic acid concentrations. The metabolite arabitol was lower in the severe cohort while other CF relevant metabolic pathways showed non-significant differences after adjusting for multiple comparisons. A sensitivity analysis to correct for biased estimates that may result from selecting subjects using an extremes of phenotype approach confirmed the protein biomarker findings.

Conclusions

Circulating inflammatory proteins differ in CF children segregated by lung function. These findings serve to demonstrate the value of maintaining centralized, high quality patient derived samples for future research, with linkage to clinical information to answer testable hypotheses in biomarker development.

Keywords: cystic fibrosis, biomarkers, proteins, lipids, metabolomics

1. INTRODUCTION

While there have been steady improvements in health outcomes in individuals with cystic fibrosis (CF), progressive lung disease remains the primary cause of morbidity and mortality. Current monitoring of lung disease relies primarily on spirometry to measure lung function which is relatively insensitive in individuals with mild lung disease, particularly children. Biomarkers reflective of lung disease activity and severity therefore have the potential to improve patient care and accelerate drug development in CF. Furthermore, clinical research in CF is hampered by a lack of sensitive biochemical measures of lung disease status. Lower airway specimens (sputum, bronchoalveolar lavage fluid) can clearly detect and quantify CF inflammation, but are not feasible for routine disease monitoring. Systemic (blood-based) markers would be preferred since blood measurements are easily standardized, repeatable, and can be obtained from subjects of any age and disease severity [1].

The U.S. CF Foundation (CFF) established a centralized biorepository to bank research quality biospecimens collected from CF patients during CFF-supported observational and interventional trials for future research [2]. These specimens, specifically collected for the purpose of banking have accompanying clinical information captured in the study of origin and are often linked to clinical outcome measures in the CFF Patient Registry (CFFPR) [3]. The link to detailed clinical information makes these specimens a powerful research resource, and the ability to access them can drive new CF biomarker explorations [2].

The primary aim of this study was to leverage banked specimens in the CFF Therapeutics (CFFT) Biorepository, collected from children with CF participating in clinical trials, to test the hypothesis that blood-based biomarkers discriminate CF children segregated by lung function. Targeted inflammatory proteins [4, 5] and lipids [6, 7], which were previously identified as abnormal in CF, were measured in serum from children with extreme lung function phenotypes who were otherwise matched based on age, gender, CFTR genotype, and Pseudomonas aeruginosa infection status. Further, discovery metabolite profiles were measured in these same serum samples based on a prior report of different serum metabolomes in CF and non-CF children [8]. An additional objective was to apply an expanded biostatistical analysis to correct for biased estimates that may result from selecting subjects using an extremes of phenotype approach [9].

2. METHODS

2.1. Sample selection

The CFFT Biorepository (https://www.cff.org/Research/Researcher-Resources/Tools-and-Resources/CFFT-Biorepository/), established in 2006 and currently managed by Fisher BioServices (Rockville, MD), has been collecting and storing biospecimens including serum from many observational and CF clinical trials. One of these studies is the Early Pseudomonas Infection Control (EPIC) observational longitudinal cohort study [10], in which serum samples were collected prospectively starting in 2005 and stored in freezers at −70°C. Banked serum samples were selected from children enrolled in the EPIC study who were either homozygous for the F508del mutation or compound heterozygous with one F508del mutation and were 11–18 years of age at the time of specimen collection (between the years of 2005–2010 when samples were collected). Patients with known CF-related liver disease or diabetes were excluded due to concerns of potential effects of these comorbidities on metabolite and lipid levels. Minimum sample volumes of 1.5 mL in separate aliquots had to be available to avoid freeze-thaw cycles. In our initial query to the CFFT Biorepository, 2,546 serum aliquots from 283 subjects were identified (Figure 1).

Figure 1.

Figure 1.

Open in a new tab

Flow diagram from initial sample query to the CFFT Biorepository to final selection of serum aliquots from 44 matched subjects in both cohorts. 1 Mild cohort was defined as having a CF-specific percentile >80% and the severe cohort with a percentile <45%. 2 To avoid thaw-freeze cycles, separate aliquots were requested.

The classification of lung function as ‘mild’ or ‘severe’ (i.e., the highest or lowest CF-specific percentile of lung function for age) was determined by the forced expiratory volume in one second (FEV1) for each patient nearest the time of specimen collection, and stability was ensured by examining multiple spirometric measurements during the two years before and after the time of the sample collection. Based on power calculations (see Supplementary Material), we targeted samples from at least 40 children in two cohorts. The mild cohort was defined as having a CF-specific percentile >80% and the severe cohort with a CF-specific percentile <45%. See the Supplementary Materials for additional details regarding the calculations of age and sex-specific CF percentiles of FEV1 % predicted. The two cohorts (mild and severe) were matched based on age, sex, CFTR genotype, and P. aeruginosa infection status. After classifying subjects into either the mild or severe cohorts, matching on the above criteria, and ensuring that there were at least two aliquots available without depleting the final samples from the Biorepository, aliquots from 44 matched subjects in both cohorts were selected (Figure 1).

One serum aliquot from each identified subject was sent directly to Metabolon® (Research Triangle Park, NC) for metabolomic analyses. Another aliquot from these same subjects was sent to the laboratory of Assem Ziady for lipid measurements. A third aliquot was sent to the University of Colorado for protein measurements. Institutional review boards at the Universities of North Carolina and Colorado and Emory University approved the analysis of de-identified blood samples from the CFFT Biorepository.

2.2. Protein measurements

Serum aliquots were analyzed in the CFFT Center for Biochemical Markers at the University of Colorado using validated assays for the following protein biomarkers: high sensitivity C reactive protein (hsCRP), serum amyloid A (SAA), granulocyte-colony stimulating factor (G-CSF), calprotectin, neutrophil elastase antiprotease complex (NEAPC), and transforming growth factor beta-1 (TGF-β1). Additional details about these assays can be found in the Supplementary Material. Multiple aliquots were available for 19 samples and analyzed to assess assay repeatability.

2.3. Lipid measurements

A total of 76 samples, 38 in both the mild and severe cohorts, were analyzed for fatty acid methyl esters. Linoleic acid (C18:2 cis; LA), eicosapentanoic acid (C20:5; EPA) and docosahexaenoic acid (C22:6; DHA) were extracted from serum samples by a specific methylation of non-esterified fatty acids in a one-step reaction using a previously described modified method[11]. Additional details can be found in the Supplementary Material.

2.4. Metabolomic profiling

Serum samples were analyzed at Metabolon® using gas- and liquid chromatography coupled with mass-spectrometry (GC/MS and LC/MS/MS platforms) as previously described[12, 13]. Quality control steps included preparation of separate aliquots that served as replicates and internal controls on separate runs and checking internal standard retention times and alignment. Compounds were identified based on chromatographic properties, mass spectra and retention indices and screened against Metabolon’s established metabolite library. Quantitative values were derived from integrated raw detector counts of the MS using normalized intensities scaled by the median values for each compound. This allowed compounds of widely different raw peak areas to be compared directly on a similar graphical scale.

2.5. Statistical Considerations

Details regarding power calculations based on previously published metabolomics data [8] are provided in the Supplementary Material.

2.5.1. Statistical analyses of proteins, lipids, and metabolites

For the targeted protein measurements, zero values were imputed for the samples with results below the lower limits of detection of the assays. There were some missing protein values due to insufficient sample to complete all six protein measurements. Values for each of the protein and lipid biomarkers were compared between disease severity groups using signed rank tests. In addition to paired analyses, two-sample comparisons were also made between disease severity groups since measurements may not have been available for each matched pair. Based on the skewness of the distributions, non-parametric tests were used and reported. A multivariate logistic regression model was created to identify the serum proteins and lipids which individually or in combination best discriminated between the mild and severe cohorts. Metabolomic results were log2 transformed prior to pairwise and group-wise analyses using parametric tests, since transformed metabolite data met normality assumptions. Correction for multiple comparisons was applied at a cutoff of 5% false-discovery-rate (FDR) at q<0.05[14]. Metabolites with > 25% missing (i.e. non-detectable) values in one of the cohorts were excluded. Partial Least Squares Discriminant Analysis (PLS-DA) was used in an attempt to identify a panel of metabolites that could discriminate the mild and severe cohorts.

2.5.2. Statistical analyses to correct for biased estimates resulting from an extremes of phenotype approach

A sensitivity analysis was conducted on a subset that more robustly fit the definition of mild and severe disease to examine serum protein differences between the groups. To that subset, an ascertainment-corrected maximum likelihood (ACML) regression approach[15, 16] was then applied to quantify and correct for the bias induced by the sampling design (i.e., having chosen those with extreme lung function instead of a random sample of all EPIC participants), allowing estimation of the usual regression parameters. Details of these methods can be found in the Supplementary Material.

3. RESULTS

3.1. Subject characteristics

The mild and severe cohorts were matched for age (mean ± SD, 14 ± 1.8 years), sex (18 female, 26 male), CFTR genotype (31 F508del/F508del, 11 F508del/minimal function mutation, 2 F508del/residual function mutation; see Supplementary Material for genotype distribution), and P. aeruginosa positivity (n=10) in the year of sample collection. The severe cohort had a lower FEV1 % predicted (using non-CF reference equations [17]) and lower BMI, higher prevalence of methicillin resistant Staphylococcus aureus (MRSA) infection, and more frequent use of macrolide and inhaled antibiotics compared with the mild cohort (Table 1). Note that the severe cohort had relatively reduced lung function compared to the mild cohort, based on CF-specific data, but not severely impaired lung function based on standard reference equations.

Table 1.

Clinical characteristics of the cohorts at the time of sample collection

Mild Cohort (n = 44) Mean (SD) Severe Cohort (n = 44) Mean (SD) P-value1
FEV1 % Predicted 111 (6.2) 90 (11.6) <0.001
BMI percentile 60 (22) 42 (27) <0.001
N (%) N (%)
MSSA infection 33 (75) 29 (66) 0.48
MRSA infection 6 (14) 15 (34) 0.04
Chronic macrolide use 15 (34) 26 (59) 0.02
Inhaled antibiotics 10 (23) 24 (55) 0.004
Dornase alfa use 37 (84) 41 (93) 0.31
Open in a new tab

SD: standard deviation, FEV1: forced expiratory volume in 1 second; BMI: body mass index, MSSA: methicillin sensitive Staphylococcus aureus, MRSA: methicillin resistant S. aureus

1

Group differences by t-test and Fisher’s exact as appropriate.

3.2. Serum protein measurements

The severe cohort had significantly higher circulating concentrations of hsCRP, SAA, G-CSF, and calprotectin compared to the mild cohort, while NEAPC and TGF-β1 were not significantly different between cohorts (Figure 2, Table S1). Comparing the samples with multiple aliquots, none of the serum protein concentrations were significantly different between the aliquots, suggesting the assays were repeatable (Table S2). Logistic regression revealed that serum G-CSF had the highest individual discriminatory ability for disease cohort (c=0.75), G-CSF and NEAPC had the highest combined discriminatory ability for any subset of two markers (c = 0.74), and G-CSF, NEAPC and LA had the highest combined discriminatory ability for any subset of three markers (c = 0.73) (Figure 3). Several circulating biomarkers are highly correlated with one another (Table S3).

Figure 2.

Figure 2.

Open in a new tab

Serum inflammatory protein and free fatty acid concentrations in mild and severe cohorts. The severe cohort had significantly higher concentrations of hsCRP, SAA, G-CSF, and calprotectin compared to the mild cohort. hsCRP: high sensitivity C reactive protein, SAA: serum amyloid A, G-CSF: granulocyte colony stimulating factor, NEAPC: neutrophil elastase antiprotease complex, TGF-β1: transforming growth factor beta 1, LA: linoleic acid, EPA: eicosapentanoic acid, DHA: docosahexaenoic acid.

Figure 3.

Figure 3.

Open in a new tab

Receiver operating characteristic (ROC) curves for the best individual serum biomarker (G-CSF) and best panel of two biomarkers (G-CSF, NEAPC) and three biomarkers (G-CSF, NEAPC and LA) for discriminating between the mild and severe cohorts. G-CSF: granulocyte colony stimulating factor, NEAPC: neutrophil elastase antiprotease complex, LA: linoleic acid.

3.3. Serum lipid measurements

Serum LA concentrations trended higher in the mild cohort compared with the severe cohort (p=0.06), while EPA and DHA concentrations were not significantly different between the two cohorts (p=0.14 and 0.36, respectively) (Figure 2; Table S1). Notably, none of the three lipids were significantly different based on age or sex.

3.4. Serum metabolomic profiling

Metabolomic profiling identified 435 distinct metabolites with known identity in the Metabolon® database. Of these, 42 and 45 compounds differed between mild and severe subjects as determined by group-wise and pairwise 2-sided t-tests at p<0.05, respectively. Thirty-five compounds were common to both analyses, with 26 being higher in severe compared to mild subjects. One metabolite, arabitol, a nucleotide sugar in the pentose and riboflavin pathway, remained nearly significant after FDR adjustment (q=0.051) (p=0.0002) and was 1.3-fold higher in mild versus severe subjects. Other metabolites in these pathways did not differ.

Metabolic pathways that showed multiple differentially expressed metabolites included bacterially derived compounds, lipids and anti-oxidants pathways (Table 2). Microbiome related pathways that differed included those derived from tryptophan and phenol metabolites (benzoate pathway). Several secondary bile acids were higher in subjects with mild disease, potentially reflecting differences in the gut microbiome. Anti-oxidants included methionine, the metabolic precursor of cysteine, and α-tocopherol, a form of vitamin E. Both were higher (unadjusted for multiple testing) in the mild cohort. Alterations in lipids included lower abundance of several ω-6 long chain fatty acids associated with inflammation in the mild cohort. To determine whether a panel of metabolites could distinguish the two cohorts, PLS-DA revealed an R2=0.641 and Q2=−0.0856, indicating poor model fit and predictive ability. No combination of metabolites was identified that discriminated between the two cohorts.

Table 2.

Pathways with multiple metabolites that differ in mild versus severe cohorts

BIOCHEMICAL SUB-PATHWAY Mild Severe Fold diff. p-value q-value
Arabitol Nucleotide sugars, pentose metabolism 1.118 0.867 1.29 0.0002 0.051
Indolelactate Tryptophan metabolism 1.133 0.894 1.27 0.0013 0.122
Para-cresol sulfate Phenylalanine & tyrosine metabolism 1.151 0.528 2.18 0.0016 0.122
4-vinylphenol sulfate (4-OH styrene) Tyrosine metabolism through benzoate pathway 1.398 0.629 2.22 0.0005 0.078
4-methyl-catechol sulfate Benzoate pathway - bacterial metabolite 1.163 0.601 1.93 0.0059 0.22
Catechol sulfate 0.14 0.074 1.87 0.0316 0.31
Glycodeoxy-cholate Secondary Bile acid and bile metabolism 0.61 0.234 2.59 0.0026 0.159
Deoxycholate 0.497 0.426 1.167 0.0139 0.225
Bilirubin (Z,Z) 1.04 0.664 1.57 0.0063 0.222
Methionine 1.084 0.934 1.1599 0.008 0.222
N-formyl- methionine Antioxidant related pathways 1.016 0.918 1.107 0.027 0.293
Alpha-tocopherol 1.04 0.772 1.34 0.011 0.224
eicosenoate 0.90 1.23 0.73 0.023 0.277
dihomo-linoleate ω-6 n fatty acids 0.90 1.21 0.74 0.029 0.300
3-hydroxybutyrate (BHBA) Fatty acid degradation / ketone body 0.99 1.45 0.68 0.024 0.334
Open in a new tab

Fold difference (diff.) indicates ratio mild/severe from paired t-test. P-value is uncorrected and q- value is corrected for multiple comparisons (false discovery rate-p).

3.5. Sensitivity analysis

Sensitivity analysis of a subset of participants (n=67) with robust ‘mild’ and ‘severe’ phenotypes as determined by averaging two years (one before and one after) of FEV1 data produced similar results (Supplementary Material Tables S4, S5). The ascertainment-corrected estimates, which are expected to increase power, also failed to identify the serum proteins NEAPC or TGF-β1 as independently associated with disease severity (online supplement Table S6).

4. DISCUSSION

This study evaluated whether banked serum specimens could be used to identify circulating biomarkers, either known candidate markers or novel ones, which discriminate children with CF segregated by lung disease severity. Subjects differed in CF-specific lung function percentiles but were matched for several important variables that influence lung disease status, including age, sex, CFTR genotype, and P. aeruginosa infection. We found that four out of six candidate serum inflammatory proteins (hs-CRP, SAA, G-CSF, calprotectin) were detected in higher concentrations in CF children with relatively impaired lung function (severe cohort) than in those with more preserved lung function. The lipid linoleic acid (LA) was higher in the mild group. While over 40 metabolites differed between the mild and severe cohorts, only one, arabitol, approached significance after adjusting for multiple comparisons. These findings serve to demonstrate the value of maintaining centralized, high quality patient derived samples for future research with linkage to clinical information to answer testable hypotheses in biomarker development. Further, the sensitivity analysis confirmed the findings from our extremes of phenotype approach to subject selection.

The CFF Biorepository was created to allow investigators access to specimens and associated data from various multicenter CF observational studies and clinical trials. It is a unique resource of biospecimens collected from a large broadly representative population of children and adults with CF. These specimens are available to all qualified researchers through a peer-review process. To date, over fifty applications to the CFF Biorepository have been submitted (Christopher Dowd, CFF, personal communication). In this study, we were interested in identifying and validating potential circulating biomarkers reflective of lung disease severity in CF. By utilizing a large biorepository of samples with linkage to a clinical sample database, we could select serum samples from a broad multicenter cohort of children with CF who met specific pre-defined diagnostic and clinical criteria. This provided us access to samples from two cohorts who differed relatively in FEV1, based on CF-specific data, but less so in absolute terms based on standard non-CF reference equations. These cohorts were well-matched across multiple demographic and clinical factors, which would be difficult to achieve in a single center study. Linkage to the CFFPR allowed for a review of longitudinal lung function data to ensure that these patients were rigorously classified into the mild and severe cohorts and that their lung functions were stable over time.

Our findings that four circulating proteins were present in different concentrations between the mild and severe cohorts suggest that these proteins may indicate underlying disease severity. The results add to a limited body of data regarding systemic inflammation in children with CF and corroborate several biomarkers identified in previous studies. Systemic inflammatory biomarkers correlate with key clinical events including pulmonary exacerbations and lung function decline [18–23]. They have also demonstrated a response to treatment of pulmonary exacerbations [5] and to treatment with azithromycin in a CF interventional trial [24]. When ROC analysis was performed, the discriminatory ability between the cohorts was moderate to good, with an AUC of 0.75 for the best individual biomarker (G-CSF). Combining biomarkers did not improve discriminatory ability of lung disease severity, suggesting that they likely represented common inflammatory processes rather than unique contributors to CF-related inflammation. Strong correlations among several of the biomarkers supports this notion. The results for hsCRP, SAA, G-CSF, and calprotectin could serve as a point of reference comparing novel biomarkers which might extend discrimination based on lung function and provide information regarding disease severity beyond these candidate protein analytes.

Several prior publications have reported that children with CF have essential fatty acid deficiency or abnormalities in comparison with control subjects. Our finding that LA, a biomarker of essential fatty acid status [25], is higher in the mild cohort is consistent with prior publications demonstrating that increased LA is associated with better weight gain and growth in infants and children with CF, and to a lesser extent with better lung function in CF [25–27]. However, we cannot conclude that serum LA is a biomarker of lung disease activity as the mild cohort was characterized by better BMI compared to the severe cohort and the higher LA concentrations among the milder subjects may primarily reflect better nutritional status in this cohort. Another possible explanation for lower LA in the severe cohort with lower lung function is increased utilization due to oxidative stress [28]. The metabolomic profiling indicated higher levels of pro-inflammatory fatty acids and lower levels of lipid peroxidation metabolites, though not significantly different after adjusting for multiple comparisons. Our findings add to genome-wide association studies and gene expression studies that have identified pathways involving inflammation and lipid metabolism as being associated with CF lung disease severity [29–31].

Metabolomic profiling showed differences in CF and non-CF children[8] and identified lipids and metabolites that changed in response to CFTR modulator therapy[32]. Some metabolomic differences have been detected in blood at initiation of antibiotics for exacerbations compared to disease baseline [33, 34]. Cross-sectional comparison of serum metabolomics here showed no distinction between cohorts when correcting for multiple comparisons. This apparent discrepancy between the current results and prior reports may indicate that metabolomic analysis is less powerful when convenience samples are examined, as samples from EPIC were obtained during routine CF blood draws independent of time or fed status. However, physiologically plausible pathways showed some differences for several compounds. These included compounds generated by fermentation of tryptophan and phenylalanine/tyrosine in the gut that may indicate microbial dysbiosis in the severe cohort. For example, p-cresol can be produced by Clostridium difficile and p-cresol sulfate can alter redox homeostasis by enhancing reactive oxygen species (ROS) production and decreasing glutathione availability [35, 36]. Differences in the secondary bile acids further support the notion of differences in gut microbiome with ineffective enterohepatic circulation in severe subjects. Further, several metabolites with anti-oxidant properties were higher in subjects in the mild cohort, e.g. polyphenols, histidine, methionine pathways which may indicate lower oxidative stress in those with milder disease [37, 38].

It is important to note that the purpose of this study was not to identify biomarkers that discriminate lung disease severity, but rather to identify serum biomarkers that discriminate CF children who were grouped by lung disease severity. The clinical utility of measuring biomarkers would be further enhanced and justified if they can discriminate disease activity among patients with the same disease severity or FEV1 % predicted, and even predict those subjects at risk of experiencing more rapid decline in lung function and/or more frequent pulmonary exacerbations. These questions will be valuable to consider in future investigations of serum biomarkers. There are additional limitations of the current study. The lack of significant differences of many of the lipids and metabolites between the mild and severe cohorts may be related to the markers themselves not having discriminatory value as biomarkers of lung disease activity. As noted previously, the results could also reflect a limitation in the use of stored samples in terms of timing of collection relative to meals and specimen stability. Many metabolites fluctuate depending on the timing of food intake [39], a fact that is unknown in these stored samples. The data on changes in lipids related to dietary habits in CF is complex, especially in regard to phospholipids [40, 41]. Prolonged sample storage is less likely a concern as it would have been expected to affect both groups to a similar extent and it has been reported that plasma fatty acid measurements were repeatable and unchanged after two years of sample storage [42]. Another limitation is that not all subject characteristics could be matched, including MRSA status, azithromycin and inhaled antibiotic use, and growth. The differences observed could reflect influences of these unmatched variables. While a higher prevalence of MRSA infection could contribute to increased circulating inflammatory protein concentrations in the severe cohort, azithromycin and inhaled antibiotics conversely reduce systemic inflammatory markers [24, 43]. Also, it is important to note that individuals with known CF-related liver disease or diabetes were excluded from these analyses. Thus, our findings do not reflect these comorbidities which we speculate may have limited selection of eligible subjects into the severe cohort. Finally, banking of these specimens pre-dates the availability of CFTR modulators. Thus, although these samples were collected from a broadly representative and highly treated cohort of CF children, they don’t reflect CFTR modulator treatment which is becoming available for more children with CF.

An “extremes of phenotype” approach to sample selection may lead to biased estimates of their association with outcomes of interest; therefore, additional sensitivity methods were applied to correct for this potential bias. While research has suggested that ascertainment-corrected maximum likelihood (ACML) estimation may have increased power to detect associations, both methods detected the same proteins as being different between the cohorts. The ACML method has an advantage of providing estimates of the differences in lung function that would be seen across ranges of the biomarkers were it to be measured in the population of interest. Although the findings of this study would need to be replicated in a larger sample, as illustrated here, the ACML method may be a useful tool in balancing the goals of biomarker discovery with the limitations of the resources at hand.

In summary, we demonstrate the successful selection and use of stored serum samples from the CFFT Biorepository to investigate potential biomarkers of lung disease severity in children with CF. We found that four candidate serum inflammatory proteins (hsCRP, SAA, G-CSF, calprotectin) were detected in higher concentrations and the lipid LA trended lower in CF children with relatively impaired lung function (severe cohort) compared to those with more preserved lung function. These data confirm the value of centralized specimen banking and linking to clinical datasets during clinical trials, validate previously identified inflammatory protein biomarkers, and could serve as a point of reference comparing future biomarkers of CF lung disease.

Supplementary Material

1

HIGHLIGHTS.

  • Serum samples from a biorepository with linkage to clinical information were selected

  • Children were categorized into two CF-specific lung function extremes of phenotype

  • Circulating inflammatory proteins differ in CF children segregated by lung function

  • Linoleic acid concentrations tended to be higher in those with preserved lung function

  • A sensitivity analysis confirmed the findings from our extremes of phenotype approach

Acknowledgements

The authors would like to thank Drs. Margaret Rosenfeld and Wayne Morgan, principal investigators of the CFF-sponsored EPIC observational study, for reviewing and approving our use of EPIC serum samples for this investigation, Alex Elbert at the CFF for generating CF-specific lung function percentiles using data from the CFFPR, and Christopher Dowd at the CFF for his assistance in identifying specimens in the CFFT Biorepository that met our specific pre-defined criteria.

Sources of support: This research was supported by Cystic Fibrosis Foundation Therapeutics (SAGEL11A0, MUHLEB11A1, and ZIADY11A0), NIH/NCATS Colorado CTSA #UL1 TR002535, and NIH/NIDDK P30 DK089507.

Footnotes

Conflict of Interest Statement

None of the study authors have any conflicts of interest to disclose beyond the grant funding listed.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

REFERENCES

  • [1].Muhlebach MS, Clancy JP, Heltshe SL, Ziady A, Kelley T, Accurso F, et al. Biomarkers for cystic fibrosis drug development. J Cyst Fibros. 2016;15:714–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [2].Clancy JP, Joseloff E. Take it to the Bank. J Cyst Fibros. 2017;16:532–3. [DOI] [PubMed] [Google Scholar]
  • [3].Knapp EA, Fink AK, Goss CH, Sewall A, Ostrenga J, Dowd C, et al. The Cystic Fibrosis Foundation Patient Registry. Design and Methods of a National Observational Disease Registry. Annals of the American Thoracic Society. 2016;13:1173–9. [DOI] [PubMed] [Google Scholar]
  • [4].Ratjen F, Saiman L, Mayer-Hamblett N, Lands L, Kloster M, Thompson V, et al. Effect of azithromycin on systemic markers of inflammation in cystic fibrosis patients uninfected with Pseudomonas aeruginosa. Chest. 2012:accepted. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [5].Sagel SD, Thompson V, Chmiel JF, Montgomery GS, Nasr SZ, Perkett E, et al. Effect of treatment of cystic fibrosis pulmonary exacerbations on systemic inflammation. Annals of the American Thoracic Society. 2015;12:708–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [6].Freedman SD, Blanco PG, Zaman MM, Shea JC, Ollero M, Hopper IK, et al. Association of cystic fibrosis with abnormalities in fatty acid metabolism. The New England journal of medicine. 2004;350:560–9. [DOI] [PubMed] [Google Scholar]
  • [7].Strandvik B Fatty acid metabolism in cystic fibrosis. Prostaglandins, leukotrienes, and essential fatty acids. 2010;83:121–9. [DOI] [PubMed] [Google Scholar]
  • [8].Joseloff E, Sha W, Bell SC, Wetmore DR, Lawton KA, Milburn MV, et al. Serum metabolomics indicate altered cellular energy metabolism in children with cystic fibrosis. Pediatric pulmonology. 2014;49:463–72. [DOI] [PubMed] [Google Scholar]
  • [9].Korn ELG, B.I. Analysis of Health Surveys: John Wiley & Sons, Inc.; 1999. [Google Scholar]
  • [10].Treggiari MM, Rosenfeld M, Mayer-Hamblett N, Retsch-Bogart G, Gibson RL, Williams J, et al. Early anti-pseudomonal acquisition in young patients with cystic fibrosis: rationale and design of the EPIC clinical trial and observational study’. Contemp Clin Trials. 2009;30:256–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [11].Lepage G, Roy CC. Specific methylation of plasma nonesterified fatty acids in a one-step reaction. J Lipid Res. 1988;29:227–35. [PubMed] [Google Scholar]
  • [12].Evans AM, DeHaven CD, Barrett T, Mitchell M, Milgram E. Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the small-molecule complement of biological systems. Analytical chemistry. 2009;81:6656–67. [DOI] [PubMed] [Google Scholar]
  • [13].Lawton KA, Berger A, Mitchell M, Milgram KE, Evans AM, Guo L, et al. Analysis of the adult human plasma metabolome. Pharmacogenomics. 2008;9:383–97. [DOI] [PubMed] [Google Scholar]
  • [14].Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A. 2003;100:9440–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [15].Zhou H, Chen J, Rissanen TH, Korrick SA, Hu H, Salonen JT, et al. Outcome-dependent sampling: an efficient sampling and inference procedure for studies with a continuous outcome. Epidemiology (Cambridge, Mass). 2007;18:461–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [16].Schildcrout JS, Garbett SP, Heagerty PJ. Outcome vector dependent sampling with longitudinal continuous response data: stratified sampling based on summary statistics. Biometrics. 2013;69:405–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [17].Quanjer PH, Stanojevic S, Cole TJ, Baur X, Hall GL, Culver BH, et al. Multi-ethnic reference values for spirometry for the 3–95-yr age range: the global lung function 2012 equations. The European respiratory journal. 2012;40:1324–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [18].Proesmans M, Els C, Vermeulen F, De Boeck K. Change in IgG and evolution of lung function in children with cystic fibrosis. J Cyst Fibros. 2011;10:128–31. [DOI] [PubMed] [Google Scholar]
  • [19].Cantin AM, Bilodeau G, Larivee P, Richter MV. Plasma biomarkers and cystic fibrosis lung disease. Clinical and investigative medicine Medecine clinique et experimentale. 2012;35:E173–81. [DOI] [PubMed] [Google Scholar]
  • [20].Reid PA, McAllister DA, Boyd AC, Innes JA, Porteous D, Greening AP, et al. Measurement of serum calprotectin in stable patients predicts exacerbation and lung function decline in cystic fibrosis. American journal of respiratory and critical care medicine. 2015;191:233–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [21].Quon BS, Dai DL, Hollander Z, Ng RT, Tebbutt SJ, Man SF, et al. Discovery of novel plasma protein biomarkers to predict imminent cystic fibrosis pulmonary exacerbations using multiple reaction monitoring mass spectrometry. Thorax. 2016;71:216–22. [DOI] [PubMed] [Google Scholar]
  • [22].Shoki AH, Mayer-Hamblett N, Wilcox PG, Sin DD, Quon BS. Systematic review of blood biomarkers in cystic fibrosis pulmonary exacerbations. Chest. 2013;144:1659–70. [DOI] [PubMed] [Google Scholar]
  • [23].Singh A, He M, Chen V, Hollander Z, Tebbutt SJ, Ng RT, et al. Blood biomarkers to predict short-term pulmonary exacerbation risk in children and adolescents with CF: A pilot study. Journal of cystic fibrosis : official journal of the European Cystic Fibrosis Society. 2019. [DOI] [PubMed] [Google Scholar]
  • [24].Ratjen F, Saiman L, Mayer-Hamblett N, Lands LC, Kloster M, Thompson V, et al. Effect of azithromycin on systemic markers of inflammation in cystic fibrosis patients uninfected with Pseudomonas aeruginosa. Chest. 2012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [25].Maqbool A, Schall JI, Garcia-Espana JF, Zemel BS, Strandvik B, Stallings VA. Serum linoleic acid status as a clinical indicator of essential fatty acid status in children with cystic fibrosis. Journal of pediatric gastroenterology and nutrition. 2008;47:635–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [26].Shoff SM, Ahn HY, Davis L, Lai H. Temporal associations among energy intake, plasma linoleic acid, and growth improvement in response to treatment initiation after diagnosis of cystic fibrosis. Pediatrics. 2006;117:391–400. [DOI] [PubMed] [Google Scholar]
  • [27].Stallings VA, Schall JI, Maqbool A, Mascarenhas MR, Alshaikh BN, Dougherty KA, et al. Effect of Oral Lipid Matrix Supplement on Fat Absorption in Cystic Fibrosis: A Randomized Placebo-Controlled Trial. Journal of pediatric gastroenterology and nutrition. 2016;63:676–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [28].Wood LG, Fitzgerald DA, Gibson PG, Cooper DM, Garg ML. Increased plasma fatty acid concentrations after respiratory exacerbations are associated with elevated oxidative stress in cystic fibrosis patients. The American journal of clinical nutrition. 2002;75:66875. [DOI] [PubMed] [Google Scholar]
  • [29].Corvol H, Blackman SM, Boëlle PY, Gallins PJ, Pace RG, Stonebraker JR, et al. Genome-wide association meta-analysis identifies five modifier loci of lung disease severity in cystic fibrosis. Nat Commun. 2015;6:8382. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [30].O’Neal WK, Gallins P, Pace RG, Dang H, Wolf WE, Jones LC, et al. Gene expression in transformed lymphocytes reveals variation in endomembrane and HLA pathways modifying cystic fibrosis pulmonary phenotypes. Am J Hum Genet. 2015;96:318–328. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [31].Polineni D, Dang H, Gallins PJ, Jones LC, Pace RG, Stonebraker JR, Commander LA, Krenicky JE, Zhou YH, Corvol H, Cutting GR, Drumm ML, Strug LJ, Boyle MP, Durie PR, Chmiel JF, Zou F, Wright FA, O’Neal WK, Knowles MR. Airway Mucosal Host Defense Is Key to Genomic Regulation of Cystic Fibrosis Lung Disease Severity. Am J Respir Crit Care Med. 2018;197:79–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [32].Kopp BT, McCulloch S, Shrestha CL, Zhang S, Sarzynski L, Woodley FW, et al. Metabolomic responses to lumacaftor/ivacaftor in cystic fibrosis. Pediatr Pulmonol. 2018;53:583–91. [DOI] [PubMed] [Google Scholar]
  • [33].Laguna TA, Reilly CS, Williams CB, Welchlin C, Wendt CH. Metabolomics analysis identifies novel plasma biomarkers of cystic fibrosis pulmonary exacerbation. Pediatr Pulmonol. 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [34].Muhlebach MS, Sha W, MacIntosh B, Kelley TJ, Muenzer J. Metabonomics reveals altered metabolites related to inflammation and energy utilization at recovery of cystic fibrosis lung exacerbation. Metabolism Open. 2019;3:100010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [35].Passmore IJ, Letertre MPM, Preston MD, Bianconi I, Harrison MA, Nasher F, et al. Para-cresol production by Clostridium difficile affects microbial diversity and membrane integrity of Gram-negative bacteria. PLoS pathogens. 2018;14:e1007191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [36].Sitkin SI, Tkachenko EI, Vakhitov TY. Metabolic dysbiosis of the gut microbiota and its biomarkers. Eksperimental’naia i klinicheskaia gastroenterologiia = Experimental & clinical gastroenterology. 2016;12:6–29. [PubMed] [Google Scholar]
  • [37].Rocha BS, Nunes C, Laranjinha J. Tuning constitutive and pathological inflammation in the gut via the interaction of dietary nitrate and polyphenols with host microbiome. The international journal of biochemistry & cell biology. 2016;81:393–402. [DOI] [PubMed] [Google Scholar]
  • [38].Watanabe M, Suliman ME, Qureshi AR, Garcia-Lopez E, Barany P, Heimburger O, et al. Consequences of low plasma histidine in chronic kidney disease patients: associations with inflammation, oxidative stress, and mortality. The American journal of clinical nutrition. 2008;87:1860–6. [DOI] [PubMed] [Google Scholar]
  • [39].Sedlmeier A, Kluttig A, Giegling I, Prehn C, Adamski J, Kastenmuller G, et al. The human metabolic profile reflects macro- and micronutrient intake distinctly according to fasting time. Sci Rep. 2018;8:12262. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [40].Maqbool A, Schall JI, Gallagher PR, Zemel BS, Strandvik B, Stallings VA. Relation between dietary fat intake type and serum fatty acid status in children with cystic fibrosis. Journal of pediatric gastroenterology and nutrition. 2012;55:605–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [41].Moukarzel S, Dyer RA, Innis SM. Complex Relation Between Diet and Phospholipid Fatty Acids in Children With Cystic Fibrosis. Journal of pediatric gastroenterology and nutrition. 2017;64:598–604. [DOI] [PubMed] [Google Scholar]
  • [42].Hodson L, Skeaff CM, Wallace AJ, Arribas GL. Stability of plasma and erythrocyte fatty acid composition during cold storage. Clinica chimica acta; international journal of clinical chemistry. 2002;321:63–7. [DOI] [PubMed] [Google Scholar]
  • [43].Toukan Y, Gur M, Hakim FH, Geffen Y, Bar-Yoseph R, Nir V, Bentur L. Intermittent inhaled tobramycin and systemic cytokines response in CF patients with Pseudomonas aeruginosa. Clin Invest Med. 2018;41:E136–E143. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1547172-supplement-1.docx (34.8KB, docx)

RESOURCES