Exhaled breath condensate identifies metabolic dysregulation in patients with radiation-induced lung injury

Abstract

Radiation-induced lung injury (RILI) is a consequence of therapeutic thoracic irradiation (TR) for many cancers, and there are no FDA-approved curative strategies. Studies report that 80% of patients who undergo TR will have CT-detectable interstitial lung abnormalities, and strategies to limit the risk of RILI may make radiotherapy less effective at treating cancer. Our lab and others have reported that lung tissue from patients with idiopathic pulmonary fibrosis (IPF) exhibits metabolic defects including increased glycolysis and lactate production. In this pilot study, we hypothesized that patients with radiation-induced lung damage will exhibit distinct changes in lung metabolism that may be associated with the incidence of fibrosis. Using liquid chromatography/tandem mass spectrometry to identify metabolic compounds, we analyzed exhaled breath condensate (EBC) in subjects with CT-confirmed lung lesions after TR for lung cancer, compared with healthy subjects, smokers, and cancer patients who had not yet received TR. The lung metabolomic profile of the irradiated group was significantly different from the three nonirradiated control groups, highlighted by increased levels of lactate. Pathway enrichment analysis revealed that EBC from the case patients exhibited concurrent alterations in lipid, amino acid, and carbohydrate energy metabolism associated with the energy-producing tricarboxylic acid (TCA) cycle. Radiation-induced glycolysis and diversion of lactate to the extracellular space suggests that pyruvate, a precursor metabolite, converts to lactate rather than acetyl-CoA, which contributes to the TCA cycle. This TCA cycle deficiency may be compensated by these alternate energy sources to meet the metabolic demands of chronic wound repair. Using an “omics” approach to probe lung disease in a noninvasive manner could inform future mechanistic investigations and the development of novel therapeutic targets.

NEW & NOTEWORTHY We report that exhaled breath condensate (EBC) identifies cellular metabolic dysregulation in patients with radiation-induced lung injury. In this pilot study, untargeted metabolomics revealed a striking metabolic signature in EBC from patients with radiation-induced lung fibrosis compared to patients with lung cancer, at-risk smokers, and healthy volunteers. Patients with radiation-induced fibrosis exhibit specific changes in tricarboxylic acid (TCA) cycle energy metabolism that may be required to support the increased energy demands of fibroproliferation.

INTRODUCTION

Thoracic radiation (TR) is a mainstay lung cancer treatment. However, ionizing radiation also damages healthy tissue, which can lead to short- and long-term injury (1). Radiation-induced lung injury (RILI) is a late effect of TR in which lung tissue develops interstitial scarring, or fibrosis, beyond 6 mo from TR (1). Technological advancements such as stereotactic body radiation therapy have reduced the risk of RILI, with only 5%–20% of patients developing clinical symptoms (such as shortness of breath, coughing, and dyspnea); however, up to 80% of patients receiving TR will develop fibrotic changes on CT imaging (1). Risk factors for developing radiation-induced fibrotic injury include tumor size and location, previous TR history, concurrent chemotherapy or immunotherapy, and radiation pneumonitis 3–6 mo after TR (2). The risk of RILI limits the dose of TR that can be applied to the tumor, potentially limiting the effectiveness of the TR at killing cancer cells (3). Currently, there are no FDA-approved treatments for RILI, although a planned clinical trial of pirfenidone has been announced (4). An improved understanding of the underlying mechanisms involved in this potentially devastating side effect is critical before novel therapeutic targets can be identified.

We were the first group to report that aerobic glycolysis is upregulated in idiopathic pulmonary fibrosis (IPF), including increased lactate and lactate dehydrogenase 5 (LDHA), in lung tissue and bronchoalveolar lavage fluid (BALF) (5–7). Subsequently, we found that ionizing radiation induces dose-dependent increases in LDHA expression, lactate production, and extracellular acidification in human lung fibroblast cultures, accompanied by increased myofibroblast differentiation and collagen expression (6). Genetic and pharmacological inhibition of LDHA protected against radiation-induced myofibroblast differentiation and blocked activation of the major profibrotic cytokine transforming growth factor (TGF)-β (6). Most significantly, pharmacological inhibition of LDHA activity significantly reduced radiation-induced lung fibrosis in a mouse model (8). Further discoveries from other groups have identified a strong relationship between glycolysis, lactate, and pulmonary fibrosis (9–13). However, disruptions in lactate and glycolysis are unlikely to be the only metabolic changes associated with fibrotic lung injury. Here, we used exhaled breath condensate (EBC) to assess the local metabolomic profile of patients with radiation-induced lung injury to determine if they exhibited disruptions in other metabolic pathways that might be associated with lung injury and fibrosis.

MATERIALS AND METHODS

Subject Recruitment and Collection of Exhaled Breath Condensate

All donors gave informed written consent. The study was conducted under the approval of the Virginia Commonwealth University (VCU) Institutional Review Board under protocol number HM20017903. Using chart review, we recruited patients who had received TR for lung cancer and who had developed fibrotic lung changes detectable by CT from the Virginia Commonwealth University (VCU) Health Massey Cancer Center located in Richmond, VA (14). Individuals with an extensive smoking history and healthy volunteers were recruited from the Early Detection Lung Cancer Screening Clinic at VCU Health. Patients with a new diagnosis of lung cancer were recruited before radiotherapy from the Department of Radiation Oncology at the Massey Cancer Center. Patient demographics and clinical parameters were retrieved from the subjects’ electronic medical records. Healthy control samples were collected between May 2021 and September 2021, and the at-risk smokers, pre-radiation cancer, and post-radiation samples were collected between November 2021 and March 2022. Exhaled breath condensate (EBC) was collected by having the subject breathe into a Rtube collection device (Respiratory Research, Austin, TX) for 15 min. The samples were kept chilled until processing, then centrifuged for 5 min at 330 g to remove cells and debris, aliquoted, and stored at −80°C until analysis.

Metabolomic Analysis in EBC

Samples were analyzed at the VCU Lipidomics and Metabolomics Shared Resource. Methanol was added to each EBC sample in a 1:1 ratio to precipitate proteins. Samples were cleared by centrifugation in a benchtop centrifuge at 5,000 rpm for 5′ and subjected to liquid chromatography/tandem mass spectrometry using the Q-Exactive Orbitrap instrument (Thermo Fisher Scientific, Bremen, Germany). Data were collected and features were identified using Compound Discoverer 3.1 (m/z cloud and ChemSpider). Signal intensity peaks were normalized to total detected signal from all features in the same sample as described (15). Features were filtered for a two-fold-change, statistically significant change compared with healthy controls in any group (adj. P < 0.05). Identified metabolites were imported into MetaboAnalyst 5.0 for pathway enrichment using SMPDB 2.0 and KEGG 103.0 databases. Alternative names for compounds were identified in PubChem.

Statistical Analysis

Data were analyzed using MetaboAnalyst 5.0 software (16), Prism 9 software (GraphPad), and R software (R Core Team 2013). Nonparametric data are shown as median and interquartile range. Differences between continuously distributed nonparametric data were assessed using Mann–Whitney tests. Two-tailed P values of <0.05 were considered significant. One-way ANOVA was conducted when comparisons included more than two groups. We performed univariate analyses between metabolite concentration and patient demographics. Continuous patient characteristics (age, pack years, forced vital capacity (FVC), diffusing capacity for carbon monoxide (DLCO), mean lung dose, and volume of lung irradiated) were tested for association using Pearson's correlation coefficient tests. For categorical variables (systemic and inhaled corticosteroids, comorbidities, cancer parameters, and pneumonitis), metabolite concentrations between groups were compared using two-sample Welch t tests. P values resulting from these analyses were adjusted for multiple comparisons using the false discovery rate (FDR) method.

RESULTS

We analyzed EBC from 10 patients with fibrotic changes after TR compared with three control groups: healthy, smokers at risk for lung cancer, and patients with lung cancer who were enrolled before their radiation treatment (Table 1). We identified striking changes between the radiation cases and the three control groups (Fig. 1A). Although there were minor differences between the healthy controls, smokers and cancer pre-radiation groups, the metabolic profile of seven of the 10 radiation cases was markedly different from the three control groups. Principal component analysis confirmed that three of the post-radiation cases segregate with the control groups in the first three principal components, while seven post-radiation cases are extremely divergent (Fig. 1B). We were unable to draw conclusions to explain why three of the post-radiation subjects exhibited a different pattern from the majority of cases as there were no marked differences in demographics or clinical pathology that could account for the differences.

Table 1.

Patient demographic and clinical characteristics

Demographic Characteristics	Healthy Control	Smoker Control	Pre-Radiation Control	Cases
n	9	10	10	10
Age, yr	69 (64, 74)	66 (63, 69)	72.5 (68, 78)	68.5 (62, 76)
Sex, male, n	3	5	7	6
Ancestry, Caucasian, n	9	9	6	6
Cigarette smoking, pack yr	0 (0, 8)	48 (42, 52)	39 (24, 64)	32 (5, 65)
Current systemic corticosteroids, n	n/a	1	5	1
Pulmonary function
FVC (% predicted)	n/a	n/a	86 (77.5, 94)3	71 (53, 96)1
DLCO (% predicted) uncorrected	n/a	n/a	42 (68, 82)2	56 (55, 81)2
Comorbidities
Other ILD, n	0	1	0	1
COPD, n	0	3	6	5
Diabetes, n	1	4	4	2
Cancer at diagnosis
Stage I/II, n	n/a	n/a	6	3
Stage III/IV, n	n/a	n/a	4	7
Current chemotherapy, n	n/a	n/a	1	0
Adenocarcinoma, n	n/a	n/a	5	5
Squamous cell carcinoma, n	n/a	n/a	2	4
Other cancer diagnosis, n4	n/a	n/a	3	1
Radiation therapy
Total treatment dose, Gy	n/a	n/a	n/a	48 (48, 60)
Dose per fraction, Gy	n/a	n/a	n/a	3 (2, 12)
Mean lung dose, Gy	n/a	n/a	n/a	5.29 (4.85, 15)1
Lung V20Gy, % volume	n/a	n/a	n/a	8 (6, 23)1
Clinical pneumonitis requiring steroid therapy, n	n/a	n/a	n/a	2

Values are median (IQR) unless otherwise indicated. n/a, not applicable. Current refers to the time of sampling. ¹One donor with missing data; ²two donors with missing data; ³three donors with missing data; ⁴other cancer diagnosis includes small cell lung cancer, mixed adenocarcinoma and squamous cell carcinoma, and metastatic carcinoma of the parotid gland.

Figure 1.

Identification of changing metabolites in exhaled breath condensate (EBC). A: heatmap of spectral features normalized to total detects with |2| fold change vs. healthy and p < 0.001 (n = 9–10 donors per group). Rows represent individual features, color = normalized to row average. Columns represent each individual donor. B: principal component (PC) analysis showing bi-plots for the first three principal components. C: Venn diagram of the named spectral features that are changing compared with healthy in each of the remaining participant groups. RILI, radiation-induced lung injury.

Of the 655 spectral features that were significantly upregulated or downregulated in the radiation case group compared with the healthy controls (P value threshold of 0.05 with false discovery correction), 235 could be matched to specific metabolites with high confidence. Although there were a small number of compounds dysregulated in each group (Fig. 1C), a majority of 111 identified compounds were uniquely dysregulated in the radiation case group (Fig. 1C). Seventy-four of the 111 compounds could be identified by MetaboAnalyst 5.0 software for pathways analysis, which revealed the enrichment of multiple pathways including glycolysis, fatty acid oxidation and biosynthesis, and the glutamate cycle. (Detailed Supplemental metabolomic data have been deposited online at https://doi.org/10.6084/m9.figshare.22377970.)

Consistent with previous observations of upregulated lactate production in patients with IPF and in a radiation fibrosis mouse model (8, 17), glycolytic energy metabolism was altered in the radiation case group. We detected significant increases in lactate in the radiation case EBC, as well as other markers of glycolysis including nicotinamide and DL-glyceric acid (Fig. 2A). Interestingly, we also observed significant alterations of other metabolites associated with energy metabolism via the TCA cycle. We observed decreased estriol accompanied by significantly increased taurine and carnitine (Fig. 2B) The glutamate pathway was altered, as indicated by increased sulfuric acid and L-pyroglutamic acid and decreased (+/−)-2-hydroxyglutaric acid (Fig. 2C). We also observed that palmitic acid, a lipid previously reported to be highly elevated in lungs of patients with IPF (9), was significantly upregulated in the EBC from the radiation injury cases, along with stearic acid and oleic acid (Fig. 2D).

Figure 2.

Metabolites associated with radiation-induced lung fibrosis are grouped by metabolic pathway. A: metabolites associated with lactate production. B: fatty acid oxidation pathway. C: glutamate pathway. D: fatty acid biosynthesis. E: selected targets related to the TCA cycle are shown for reference. P values are from one-way ANOVA; n = 9–10 donors per group. RILD, radiation-induced lung disease. [Image created with BioRender.com and published with permission.]

We looked very carefully for correlations between individual metabolites and demographic and clinical parameters including age, sex, smoking status, comorbidities, cancer stage, mean lung dose of radiation, volume of lung irradiated, occurrence of clinical pneumonitis and systemic steroid use. The only significant correlations were with mean lung dose, including (+/−)-2-hydroxyglutaric acid (R = −0.540, P = 0.017), acetic acid geranyl ester (R = −0.514, P = 0.024), and ostruthin (R = 0.506, P = 0.027). The data did not reveal significant correlations with other clinical parameters.

Interestingly, three of the radiation injury cases had a metabolic profile that was more similar to the lung cancer pre-radiation control group than to the other seven radiation cases (Fig. 1A, B). We were unable to identify any clinical parameters—including age, sex at birth, ancestry, smoking history, cancer stage, mean lung dose, or volume of lung irradiated—that correlated with the metabolic profile.

DISCUSSION

Radiation-induced lung injury (RILI) is a significant side effect of thoracic radiation (TR) We undertook this study to identify metabolic changes in patients with radiation-induced lung injury that could be associated with the incidence of pulmonary fibrosis. Although none of our radiation case subjects were diagnosed with clinical RILI, they all had CT-confirmed fibrotic lesions, which occur in up to 80% of cancer patients receiving thoracic radiation (10, 11). This ongoing risk of radiation-induced lung fibrosis even with improvements in beam technology (2), limits the dose of radiation that can be used to kill tumors, which highlights the ongoing need to understand the mechanisms of radiation-induced fibrotic change.

In this pilot study, we collected exhaled breath condensate (EBC) for metabolomic analysis. Collection of EBC is non-invasive and is amenable to repeated sampling and longitudinal studies (12, 13), and is increasingly recognized as a useful tool to identify diagnostic or prognostic biomarkers in lung disease (13, 18, 19). The EBC metabolome from the radiation case patient group was distinct from the healthy, smoker, and cancer pre-TR control groups, indicating that radiation may change lung metabolism in a specialized manner distinct from the effects of smoking history or cancer diagnosis (Fig. 1). Some of the changes suggest that lung cells adapt to alterations in energy demand concomitant with the development of fibrotic lesions. Significantly higher levels of nicotinamide and lactic acid in the EBC of case patients indicate upregulation of the nicotinamide and glycolysis pathways by the lung to generate pyruvate (Fig. 2A). We have previously reported that patients with IPF have upregulated glycolysis and production of lactate in BALF and lung tissue (5), and we were excited to see that finding replicated in our radiation cases. Fatty acid β-oxidation was increased in the radiation injury group, indicated by increases in taurine and carnitine (Fig. 2B), which is consistent with previous reports in radiation-induced injury (20). Fatty acid oxidation was also upregulated in macrophages in the mouse bleomycin model of lung fibrosis (21). Estriol is decreased, possibly because its conversion to taurine is accelerated. Interestingly, estrogen signaling has previously been shown to reduce following TR, and estriol treatment has been suggested as a radioprotective agent (22). Increased levels of sulfuric and pyroglutamic acid indicate increased activity of the glutamate pathway which can generate α-ketoglutarate as an alternative input to the TCA cycle (Fig. 2C). Finally, decreased phenylacetate and 2-hydroxyglutarate represent decreases in other sources of acetyl-CoA. Taken together, our results support the hypothesis that radiation upregulates glycolysis and conversion of pyruvate to lactate which promotes the incidence of fibrosis (23). Decreased pyruvate available to enter the TCA cycle is compensated for by upregulating the glutamine pathway and fatty acid β-oxidation (Fig. 2E).

RILI includes both an early pneumonitis phase and a late fibrotic phase. Only two of the 10 subjects in the fibrosis case group developed clinically significant pneumonitis after TR, and none had indications of pneumonitis on their CT scans used to identify fibrotic change. However, we cannot rule out that pneumonitis might contribute to the metabolic changes observed in the post-radiation case group. It is worth noting that our present study has identified several metabolic changes similar to those seen in IPF, including upregulated production of lactate (5), upregulated fatty acids including palmitic, stearic, and oleic acid (9), and disruptions in fatty acid β-oxidation (24) and glutaminolysis (25), suggesting, but not proving, that the changes we observe are largely due to post-radiation fibrosis rather than inflammation.

The primary limitation of this study is the small sample size. We arbitrarily chose 10 subjects per group because this was originally a pilot study for a planned, larger longitudinal study that will follow patients from diagnosis through the development of radiation fibrosis at one year or later. As a result, our ability to perform subgroup analysis in this pilot study is limited, and few correlations with treatment and demographic factors within the radiation case group could be uncovered. However, the data were statistically significant and meaningful even with a small sample, and we believe the results are striking and important. A power calculation using this pilot study found that a longitudinal study would need to enroll 24–30 subjects to have 80% power to detect significant (P ≤ 0.05 with FDR correction) changes in key metabolites that exhibited interesting but nonsignificant changes in the pilot study. A common limitation of untargeted metabolomics is that not all spectral features can be matched to known compounds, limiting subsequent analysis. Although targeted analysis can detect preselected compounds with greater specificity, it is subject to selection bias. In the future, targeted analysis could be used to refine our results by focusing on selected pathways. Finally, it is uncertain which cells in the lung contribute to the EBC. Contributions from macrophages, epithelial cells, and endothelial cells are likely, while, fibroblasts and myofibroblasts, the major scar-forming cells, have an unclear role in producing EBC. It is important to note that epithelial cells and macrophages can contribute to a pro- or anti-fibrotic environment in the lung that influences fibroblast phenotypes, and epithelial cells can contribute directly to fibrosis via epithelial-mesenchymal transition; this suggests that the EBC metabolome is highly relevant to fibrotic lung disease (26, 27).

Here, we report that patients with fibrotic lung changes post-TR exhibit a unique metabolomic signature in their EBC. This finding supports the concept that upregulation of glycolysis and lactate drives fibrosis. We also identified other significant metabolic alterations which support the hypothesis that fibrosis involves other compensatory changes to energy metabolism, although, the extent to which these changes are sufficient to promote fibrosis remains to be proven. We believe that a comprehensive understanding of metabolic alterations in pulmonary fibrosis will lead to novel treatment options.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental metabolomics: https://doi.org/10.6084/m9.figshare.22377970.

GRANTS

This work was funded in part by NIH Grant R01HL127001, the Pulmonary Fibrosis Foundation, and the Chandler-Pollock-Solimano-Thomas Pulmonary Fibrosis Research Fund. J.P.-L.O. was funded in part by a Ford Foundation Predoctoral Fellowship. M.A.T.F. is funded by NIH F32HL154525. Services in support of the research project were provided by the VCU Massey Cancer Center Lipidomics and Metabolomics Shared Resource, supported, in part, with funding from NIH-NCI Cancer Center Support Grant P30 CA016059. P.D.J. was supported by the Pulmonary Fibrosis Foundation Scholar program.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.P.-L.O., M.A.T.F., E.W., T.H.T., and P.J.S., conceived and designed research; J.P.-L.O., M.A.T.F., S.V.C., B.B.S., and J.C.A. performed experiments; J.P.-L.O., M.A.T.F., R.T., R.M.S.R., P.D.J., J.C.A., S.S., J.L., L.A.C., E.W., and T.H.T. analyzed data; J.P.-L.O., M.A.T.F., L.A.C., T.H.T., and P.J.S. interpreted results of experiments; J.P.-L.O. and M.A.T.F. prepared figures; J.P.-L.O. and M.A.T.F. drafted manuscript; J.P.-L.O., M.A.T.F., S.S., L.A.C., E.W., T.H.T., and P.J.S., edited and revised manuscript; J.P.-L.O., M.A.T.F., S.V.C., R.T., B.B.S., R.M.S.R., P.D.J., J.C.A., S.S., J.L., L.A.C., E.W., T.H.T., and P.J.S. approved final version of manuscript.