|Home | About | Journals | Submit | Contact Us | Français|
Rationale: Because acute lung injury is a sporadic disease produced by heterogeneous precipitating factors, previous genetic analyses are mainly limited to candidate gene case-control studies.
Objectives: To develop a genome-wide strategy in which single nucleotide polymorphism associations are assessed for functional consequences to survival during acute lung injury in mice.
Methods: To identify genes associated with acute lung injury, 40 inbred strains were exposed to acrolein and haplotype association mapping, microarray, and DNA-protein binding were assessed.
Measurements and Main Results: The mean survival time varied among mouse strains with polar strains differing approximately 2.5-fold. Associations were identified on chromosomes 1, 2, 4, 11, and 12. Seven genes (Acvr1, Cacnb4, Ccdc148, Galnt13, Rfwd2, Rpap2, and Tgfbr3) had single nucleotide polymorphism (SNP) associations within the gene. Because SNP associations may encompass “blocks” of associated variants, functional assessment was performed in 91 genes within ± 1 Mbp of each SNP association. Using 10% or greater allelic frequency and 10% or greater phenotype explained as threshold criteria, 16 genes were assessed by microarray and reverse real-time polymerase chain reaction. Microarray revealed several enriched pathways including transforming growth factor-β signaling. Transcripts for Acvr1, Arhgap15, Cacybp, Rfwd2, and Tgfbr3 differed between the strains with exposure and contained SNPs that could eliminate putative transcriptional factor recognition sites. Ccdc148, Fancl, and Tnn had sequence differences that could produce an amino acid substitution. Mycn and Mgat4a had a promoter SNP or 3′untranslated region SNPs, respectively. Several genes were related and encoded receptors (ACVR1, TGFBR3), transcription factors (MYCN, possibly CCDC148), and ubiquitin-proteasome (RFWD2, FANCL, CACYBP) proteins that can modulate cell signaling. An Acvr1 SNP eliminated a putative ELK1 binding site and diminished DNA–protein binding.
Conclusions: Assessment of genetic associations can be strengthened using a genetic/genomic approach. This approach identified several candidate genes, including Acvr1, associated with increased susceptibility to acute lung injury in mice.
A major challenge to critical care is to reliably predict survival of patients with acute lung injury. Individual susceptibility varies greatly (i.e., patients presenting with the same severity score can have markedly different clinical outcomes). For this reason, studies have begun to investigate the role of genetics in determining survival during acute lung injury. However, because acute lung injury is a sporadic disease produced by heterogeneous precipitating factors, previous genetic analyses are mainly limited to case-control studies that evaluate candidate genes associations.
This study presents a strategy in which single nucleotide polymorphism associations are assessed for functional consequences to survival during acute lung injury in mice. This approach identified several candidate genes, and Acvr1 may be particularly worthy of future investigations in acute lung injury.
Acute lung injury is marked by pulmonary edema resulting from increased epithelial and endothelial permeability, decreased clearance of edema fluid, and disruption of surfactant-associated protein function (1–3). This condition can be induced directly (e.g., through smoke inhalation) or indirectly (e.g., through sepsis) (1). In smoke inhalation, the severity of acute lung injury (2) and cardiovascular dysfunction (4) are critical determinants of morbidity and mortality (5–7). Acrolein, a potent irritant (8–10), is the major chemical in smoke responsible for pulmonary edema (11, 12).
A major challenge to critical care is to reliably predict and enhance survival in acute lung injury (7). Individual susceptibility varies greatly (i.e., patients presenting with the same severity score can have markedly different clinical outcomes) (13). For this reason, studies have begun to investigate the role of genetics in determining survival during acute lung injury (14–17). However, because acute lung injury is a sporadic disease produced by heterogeneous precipitating factors, previous genetic analyses are mainly limited to case-control studies that evaluate candidate genes associations.
Constant acquisition of genome-wide information on numerous species, including more than 40 mouse strains (18–21), makes the mouse a useful model to facilitate rapid evaluation of the genetic basis of human physiology and pathophysiology (22, 23). Complementing conventional single gene mapping approaches, genome-wide mapping has mainly used quantitative trait locus (QTL) analysis in mice, which has been a valued tool to identify candidate genes responsible for complex traits underlying human disease (24–26). However, a major obstacle of identifying genes is the difficulty of resolving large QTL regions (10–20 cM) into sufficiently small intervals to make positional cloning practical (27). In this study, haplotype association mapping is used to obtain a dense single-nucleotide polymorphism (SNP) map, which offers finer genetic resolution in the identification of genetic determinants of acute lung injury.
This study was performed in accordance with the Institutional Animal Care and Use Committee of the University of Pittsburgh (Pittsburgh, PA) and mice were housed under specific pathogen-free conditions. Forty inbred strains (n = 507 mice, 6–8 wk old, female; The Jackson Laboratory, Bar Harbor, ME) were used in the study. Mice were exposed to filtered air (control) or acrolein (10 ppm, 24 h) as generated and monitored as described previously (8), and survival time was recorded. Haplotype association analysis was performed using threshold –log(P) value of 6.0 for significant and 4.0 for suggestive linkage as described previously (23). To examine acrolein-induced changes in lung histology, bronchoalveolar lavage, and lung transcripts, 129X1/SvJ and SM/J mice (two strains that represented the opposite ends of the phenotypic spectrum of the analyzed strains) were exposed to filtered air (0 h, control) or acrolein (10 ppm for 6, 12, or 17 h). Microarray analysis and quantitative real time-polymerase chain reaction (qRT-PCR) were used to determine transcript levels of genes previously associated with acute lung injury and to contrast genes identified. Three 5′ untranslated region (UTR) SNPs in Acvr1 were genotyped in 28 strains and the consequences of the rs6406107 variant on DNA–protein binding was assessed by electrophoretic mobility shift assay. Additional details of the methods are presented in the online supplemental.
Mice from 40 mouse strains were exposed to acrolein and survival time recorded. The mean survival time varied among mouse strains (Figure 1A). The most polar strains varied approximately 2.5-fold from 16.5 ± 0.9 hours (MOLF/EiJ) to 41.1 ± 4.5 hours (129X1/SvJ). Using the data from all the mice, a haplotype association map was obtained for acrolein-induced acute lung injury (Figure 1B). Significant [−log(P) ≥ 6.0] associations were identified on chromosome 2 with suggestive [−log(P) ≥ 4.2] associations identified on chromosomes 1, 2, 4, 11, and 12. Five genes (Rfwd2, Cacnb4, Galnt13, Acvr1, and Rpap2) had three to four significant or suggestive SNP associations within the gene boundary, whereas two genes (Ccdc148 and Tgfbr3) had one significant or suggestive SNP association within the gene boundary (see Table E1 in the online supplement). Evaluation of SNP associations determined the allelic frequency ranged from 0.20 to 0.54 (or 73 and 197 mice, respectively) among the 40 strains tested for survival (n = 365 mice). Because a sufficient number of mice carried the “minor” allele (≥ 10%), the percentage of phenotype explained by each SNP association was determined using the difference between the mean survival time of mice carrying either allele and dividing it by the total difference between the two polar strains (24.6 h) (Figure E1). The amount of phenotype explained by the SNP associations was substantial (range: 17–34%) (Table E1).
A number of SNPs with association probabilities −log(P) of 5.0 to 7.0 were identified in a small region on chromosome 2 (54–59 Mbp). Genes in this region with one or more significant SNP associations included Galnt13, Acvr1, and Ccdc148 (Figure 1C). Because any SNP association may encompass blocks of associated variants that reside near the identified SNP (28), functional assessment was performed on SNPs within 91 genes that mapped within 1 Mbp of a significant/suggestive SNP association. Using 10% or greater allelic frequency and 10% or greater phenotype explained as threshold criteria, 16 candidate genes qualified for further analysis.
To further assess susceptibility, a sensitive and a resistant mouse strain were contrasted by detailed characterization. For the sensitive strain, we selected SM/J and avoided wild-derived (e.g., MOLF/EiJ, SPRET/EiJ, or PERA/EiJ) or repository (e.g., RIIIS/J) strains. The mean survival time for the SM/J strain (17.2 ± 0.46 h) was not significantly different from that of MOLF/EiJ. The 129X1/SvJ was selected as representative of a resistant strain. Histological analyses of these strains were consistent with acute lung injury. Perivascular enlargement and leukocyte infiltration were more evident in the sensitive SM/J strain than in the resistant 129X1/SvJ strain (Figure 2). Similarly, bronchoalveolar lavage protein concentration (Figure 3A), polymorphonuclear leukocyte percentage (Figure 3B), and nitrite concentration (Figure 3C) increased sooner in the SM/J than in 129X1/SvJ mice.
Lung transcript levels were also compared between SM/J and 129X1/SvJ strains using microarray analysis (Table E2). Previously identified transcripts that are altered in mouse lung during acute lung injury include IL-6 (29) and metallothionein 2 (MT2) (30). Microarray results indicated that IL6 transcripts at 6 or 12 hours increased more in SM/J than in 129X1/SvJ lung (Figure E2, top). These differences were confirmed by qRT-PCR (although microarray increases were less than qRT-PCR increases). In contrast, MT2 transcripts increased in both strains when compared with strain-matched controls but were not different between the strains (Figure E2, bottom).
Further evaluation of microarray transcripts that increased with exposure (fold log 2 ≥ 1.0; P < 0.01) revealed enriched pathways including transforming growth factor-β (TGFB) signaling, cell death, and nuclear factor, erythroid derived 2, like 2 (NFE2L2)-mediated oxidative stress response for transcripts that increased with exposure (Figure 4; Table 1). Transcripts in TGFB signaling (e.g., SM/J −log(P) = 5.8 vs. 129X1/SvJ −log(P) = 1.2 at 12 h) or cell death pathway (e.g., SM/J −log(P) = 24.6 vs. 129X1/SvJ –log(P) = 15.4 at 12 h) were increased more in SM/J as compared with 129X1/SvJ mice. In contrast, the response of SM/J and 129X1/SvJ were similar in the NFE2L2-mediated oxidative stress response (e.g., SM/J –log(P) = 5.8 vs. 129X1/SvJ –log(P) = 6.0 at 12 h). The latter is noteworthy in that NFE2L2 activation leads to accumulation of transcripts encoding enzymes that may be viewed as protective (e.g., heme oxygenase [decycling] 1 [HMOX1]) during acute lung injury (31). One transcript in the TGFB signaling pathway, secreted phosphoprotein 1 (SPP1), was also evaluated using qRT-PCR. At 6 or 12 hours, SPP1 transcript increased in SM/J mice, but was not significantly different from control in 129X1/SvJ mice (Figure E3). The increase at 12 hours was significantly greater in the SM/J lung as compared with the 129X1/SvJ lung.
For microarray transcripts that decreased with exposure, enriched pathways included glucocorticoid receptor signaling, lipid metabolism, and retinoic acid receptor activation (Figure 4). In general, pathways with decreased transcripts were similar between these strains as determined by the enrichment probability −log(P). Individual transcripts among these pathways were decreased more in the SM/J than in 129X1/SvJ mice, however.
Additional pathways that were enriched in SM/J but were not in 129X1/SvJ mice included (1) increased transcripts: IL-10 signaling, mTOR signaling, and immune cell trafficking; and (2) decreased transcripts: lysine biosynthesis (Table 1). Pathways that were enriched in 129X1/SvJ but not in SM/J mice included (1) increased transcripts: phospholipid degradation, sphingosine-1-phosphate signaling, protein ubiquitination pathway, cell movement, and antiapoptosis; and (2) decreased transcripts: ephrin receptor signaling.
Having found that SM/J and 129X1/J mice differ in the time-course of acrolein-induced lung injury, the identified 16 candidate genes were then evaluated by qRT-PCR. Initially, basal levels were compared and transcript levels of 11 of the 16 genes were significantly different between strains (Figure 5).
Next, the time course of transcripts of the 16 candidate genes was measured during exposure and compared with the strain-matched control. This analysis identified six transcripts (ACVR1, ARHGAP15, CACNB4, CACYBP, RFWD2, and TGFBR3) that were different between the strains with exposure (Figure 6). For each of these transcripts, the corresponding gene was interrogated to determine whether it contained SNPs that could alter putative transcriptional factor recognition sites (i.e., expression SNP or eSNP). Examining the sequence variability within −2,000 bp in the 5′UTR through intron 1, five genes, Acvr1, Arhgap15, Cacybp, Rfwd2, and Tgfbr3, were identified to have eSNPs that could eliminate ELK1, GAL4, HES-1, YY1, and NF-1 binding motifs as determined by Transcription Element Search Software (32), respectively (Table 2). The survival time associated with either allele was then determined for the mouse population. Mice possessing these eSNPs exhibited mean survival time of 17–30% of the difference between the mean survival times of the most polar strains. The sixth gene, Cacnb4, lacked an eSNP associated with a difference in phenotype in the 5′UTR but contained an SNP that would produce a variant splicing transcript.
In addition, four transcripts (CCDC148, FANCL, MYCN, and TNN) were altered by exposure but were not different between the strains (Figure E4). The corresponding genes were then interrogated to determine whether they contained a coding nonsynonymous SNP (i.e., cnSNP) that could produce amino acid changes. Two genes, Ccdc148 and Fancl, had Gln81Arg and Asp1Ala amino acid substitution, respectively (Table 2). Although lacking a cnSNP, Mycn had an eSNP that could eliminate a MTF-1 binding motif. In Tnn, three SNPs (rs4877414, rs46709434, and rs49224805) could produce Ala310Thr, Lys333Gln, and Arg562Cys amino acid substitutions, respectively. Mice having either allele in the identified SNPs exhibited phenotypic differences in survival of 11 to 28%.
The predicted amino acid substitutions (missense mutations) were examined using Prosite (33, 34) to determine whether they are situated in functional domains of the encoded protein. In CCDC148, the rs27956803 Gln81Arg substitution resulted in a gain of a predicted protein kinase C phosphorylation site at amino acid 79 to 81. The Asp1Ala substitution in FANCL was not in a predicted functional domain. In TNN, three amino acid substitutions are located in fibronectin type III domains and the Ala310Thr substitution resulted in a loss of a predicted protein kinase C phosphorylation site.
Last, MGAT4A transcripts did not change significantly with exposure in either strain of mouse tested. As noted above, basal transcript levels were different between the strains. The corresponding sequence was therefore interrogated for eSNP and cnSNP, and none were observed. The 3′UTR was then examined and an SNP was observed in exon 15/3′-UTR, which is in a region that may alter message stability.
Together, the difference in transcript levels combined with the in silico sequence assessment yielded 11 candidates (Mgat4a, Rfwd2, Tnn, Cacybp, Arhgap15, Cacnb4, Acvr1, Ccdc148, Tgfbr3, Fancl, and Mycn) with probable functional associations to diminished survival resulting from acrolein-induced acute lung injury (Table 2). Candidate genes with functional SNP associations were located in eight chromosomal regions (Mbp): 1 (37), 1 (161), 2 (43), 2 (52), 2 (58), 5 (107), 11 (26), and 12 (12) and independently could explain 11 to 30% of the phenotype.
Three SNPs (rs33408603, rs6406068, and rs6406107) were identified in the 5′UTR region of Acvr1. However, few strains had previously been genotyped (129X1/SvJ, C57BL/6J, and A/J for rs33408603, and CZECHII/EiJ and C57BL/6J for rs6406068 or rs6406107). To obtain additional genotype information on these SNPs, 28 strains were genotyped by PCR amplification–DNA sequencing (Table E3). One SNP, rs33408603 (A/G), was observed only in the 129X1/SvJ strain (A-allele). The remaining SNPs exhibited an identical haplotype in all strains typed (i.e., strains with the rs6406068 C-allele had the rs6406107 A-allele). The allele frequency of this SNP haplotype was 12% and could explain 19% of the phenotype. Analysis of the possible consequence of these SNPs indicated that the rs6406107 A-allele would result in the loss of a putative ELK1, member of ETS oncogene family (ELK1) transcription factor binding site. In vitro, an Acvr1 promoter region–derived oligonucleotide containing the rs6406107 G-variant more effectively competed the nuclear protein-binding capacity of a labeled probe than that containing the A-variant (Figure 7).
Past studies using candidate gene approaches have been promising and have provided valuable insights into the genetic architecture of the complex trait of acute lung injury; however, these studies also have had recognized limitations (28, 35–37). For several reasons it has been difficult to assemble a large enough cohort of clinically identical individuals to perform a genome-wide association study in humans. Adding to this barrier are the difficulties in establishing functional significance of genetic associations (38). Thus, using genome-wide association approaches to identification of candidate genes and related pathways (39) in mice could be a valued additional approach (21).
Previously, considerable interstrain differences have been reported in acute lung injury susceptibility (24, 25). In this study with acrolein, we observed an approximately 2.5-fold difference in survival time between the most sensitive and resistant strains, and thus this trait was amenable to mapping. In past decades, QTL mapping has identified chromosomal regions containing genes affecting cancer, diabetes, hypertension, and other disease-related phenotypes. However, a major obstacle of identifying QTL genes is the difficulty of resolving large chromosomal regions (10–20 cM) into sufficiently small intervals to make positional cloning practical. With advances in genome sequencing, dense SNP maps have proven successful in the refinement of previous QTL regions and the identification of new genetic determinants of complex traits (21).
One concern about any SNP association study is that variants identified in genome-wide association study may confer a relatively small increment of risk (explaining 1–3% to the population variance) (28, 35–38). Here, we present a strategy to overcome this limitation in which SNPs within 1 Mbp of an association are assessed for functional consequences to survival in the mice tested. A threshold of 10% or greater survival time between the most sensitive and resistant strains was set for further assessment of candidates. Type I error was diminished by comparing the phenotype (i.e., individual survival time) of every member of the population with the corresponding strain genotype. This approach also limits the associated SNPs to those with feasible functional consequence to expression (eSNP) or amino acid sequence (cnSNP).
Of the candidate genes identified, only Mycn has been associated with respiratory failure. Mice with a Mycn mutation (that reduces expression through alternative splicing) die at birth due to respiratory failure (40). However, this effect is probably due to defective lung development (41) and abnormal branching morphogenesis (42), because complete ablation of the gene is embryonic lethal. The other candidate genes identified in this study have not previously been associated with acute lung injury. Nonetheless, these genes are present in the lung and the corresponding proteins have functions related to the cell stress signaling in lung injury.
As with any study, this investigation has several limitations. Although acrolein-induced acute lung injury has relevance to smoke inhalation, numerous other agents (either chemical or infectious) can elicit acute lung injury (43). Until these other forms or phenotypes (e.g., lavage protein) of acute lung injury are evaluated, generalization to other forms of this condition is not warranted. Supportive evidence by another genetic approach (e.g., traditional back-cross or F2 cohort) that likewise identified the candidate genes identified with this haplotype association mapping analysis would strengthen the associations found in this study. In addition, although this approach may improve the assessment of the results obtained from a haplotype mapping analysis, functional assessment of each candidate gene will require further studies (e.g, gene-targeted or transgenic mice). Last, animal models are limited in that species differences in lung structure and function can diminish applicability to humans. Despite these limitations, the identified genes and related pathways may help to direct future human genetic studies that evaluate such pathways using selected tagSNPs.
One of the leading candidate genes was Acvr1. In mice, Acvr1 gene targeting produces a morphological gastrulation defect, which is embryonic lethal (44–46). In humans, mutations in the glycine-serine–rich (GS) activation domain or sites that interact with the GS domain of ACVR1 have been associated with fibrodysplasia ossificans progressiva, which is characterized by progressive heterotopic ossification that can lead to respiratory difficulties (47, 48). The ACVR1 GS domain is involved in phosphorylation and human mutations lead to a gain of function (i.e., augmented signaling) (49). In mice, postnatal expression of mutant ACVR1 can lead to ectopic ossification, but only when combined with infection. Corticosteroid inhibited ossification in mice, suggesting that mutant ACVR1 and inflammation are both required for ectopic ossification (50).
Although named activin A receptor, type 1 (a.k.a. activin receptor-like kinase 2, Alk2), ACVR1 does not bind activins (a.k.a. inhibins), but binds BMP2, 4, 6, and 7, through a heteromeric complex with BMPR2 (51–54) (Figure E5, Table E4). Critical to respiratory organogenesis and development (55, 56), BMPs elicit various effects in adult tissues through type I and II receptors, which in turn phosphorylate receptor-regulated R-SMAD protein (primarily SMAD1, 5, and 8) (57, 58). On activation, R-SMAD proteins associate with the common mediator SMAD4 and translocate to the nucleus, where they act as transcription factors to regulate expression of target genes, including the inhibitory SMADs (I-SMAD), SMAD6, and SMAD7 (59). SMAD6 and SMAD7 inhibit/modulate TGFB/BMP signaling by interfering with the activation of other SMADS (59–61), thereby providing negative feedback control (62).
In this study we found that lung ACVR1 transcripts decreased more in the sensitive SM/J than the resistant 129X1/SvJ mice. In vitro, an Acvr1 promoter region SNP rs6406107 diminished nuclear protein-binding capacity of a labeled probe and was associated with decreased survival in mice. This SNP could lead to diminished ACVR1 transcript levels because the A-allele would eliminate a putative ELK1 binding site. ELK1 can be phosphorylated by a TNF-initiated, JUN-mediated mechanism in pulmonary epithelial cells (63). A possible consequence of decreased ACVR1 transcripts would be augmented TGFB/BMP signaling. This is consistent with increases in transcripts encoded by TGFB target genes noted in the microarray. Previously, TGFB/BMP signaling has been associated with adverse effects during acute lung injury (64–66). Active TGFB1 is increased in edema fluid obtained from patients with acute lung injury (67–70), and TGFB1 decreases pulmonary endothelial (71) or epithelial integrity (72) and diminishes epithelial fluid transport (73, 74). The determination whether ACVR1 may serve to limit TGFB/BMP signaling during lung injury is worthy of future investigation.
Compared with 129X1/SvJ mice, several stress-response transcripts (e.g., CACYBP, DNAJA1, FKBP4, HSPA1A, NFAT5, and STIP1) decreased more in the SM/J lung. Inasmuch as these transcripts encoded chaperon proteins, a decrease could destabilize existing proteins (i.e., aggregation and unfolding of newly translated proteins) and diminish the effectiveness of ubiquitin-proteasome pathway (75, 76). Ubiquitin-proteasome proteins may interact with other candidates identified (77). For example, ACVR1 interacts with FANCL (78) and modulates SMURF2 ubiquitination of various proteins, including SMADs. Two candidate genes, Rfwd2 and Cacybp, encode proteins with ubiquitin-protein ligase activity. Besides limiting TP53 (a.k.a. p53) (79), RFWD2 suppresses JUN (80) and FOXO1 (81) and other transcription factor accumulation and thereby limits cell stress (82). For example, JUN diminishes surfactant-associated protein B synthesis and thereby increases mortality during acute lung injury (3, 83), and FOXO1 is critical to maintaining claudin 5 in acrolein-induced acute lung injury (84).
In summary, haplotype association mapping, microarray/qRT-PCR analyses, and in silico SNP analysis identified 11 candidate genes (Acvr1, Arhgap15, Cacnb4, Cacybp, Ccdc148, Fancl, Mycn, Mgat4a, Rfwd2, Tgfbr3, and Tnn) associated with acrolein-induced acute lung injury in mice. Several genes were related and encoded receptors (ACVR1, TGFBR3), transcription factors (MYCN, possibly CCDC148), and ubiquitin-proteasome (RFWD2, FANCL, CACYBP) proteins that may interact to modulate cell signaling. The Acvr1 SNP rs6406107 eliminates a putative transcription factor binding site and diminished DNA–protein binding making this gene worthy of future investigations in acute lung injury.
Originally Published in Press as DOI: 10.1164/rccm.201006-0912OC on February 25, 2011
Supported by National Institutes of Health grants ES015675, HL077763, and HL085655 (G.D.L); HL091938 (Y.P.P.D.), HG003749 and LM009662 (M.M.); HL084932 and HL095397 (N.K.); AT003203, AT005522, CA113793, and CA134433 (M.Y.); and HL075562 (D.R.P.).
This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org
Author Disclosure: G.D.L. is employed by the University of Pittsburgh. V.J.C. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. P.L. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. K.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. A.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. K.G. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. A.S.J. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. K.A.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. M.D. is employed by the University of Cincinnati. H.P.-V. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. R.A.D. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Y.P.P.D. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Q.L. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. L.J.V. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. M.M. is employed by the University of Cincinnati. N.K. was a consultant for Sanofi-Aventis and Stromedix. He received lecture fees form Medimmune and owns a patent on use of microRNAs to diagnose and treat IPF Us or peripheral blood biomarkers to diagnose and predict outcome in IPF. M.Y. is employed by the Medical College of Wisconsin. D.R.P. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript.