Accelerated Lung Function Decline and Mucus–Microbe Evolution in Chronic Obstructive Pulmonary Disease

Abstract

Rationale

Progressive lung function loss is recognized in chronic obstructive pulmonary disease (COPD); however, no study concurrently evaluates how accelerated lung function decline relates to mucus properties and the microbiome in COPD.

Objectives

Longitudinal assessment of mucus and microbiome changes accompanying accelerated lung function decline in patients COPD.

Methods

This was a prospective, longitudinal assessment of the London COPD cohort exhibiting the greatest FEV₁ decline (n = 30; accelerated decline; 156 ml/yr FEV₁ loss) and with no FEV₁ decline (n = 28; nondecline; 49 ml/yr FEV₁ gain) over time. Lung microbiomes from paired sputum (total 116 specimens) were assessed by shotgun metagenomics and corresponding mucus profiles evaluated for biochemical and biophysical properties.

Measurements and Main Results

Biochemical and biophysical mucus properties are significantly altered in the accelerated decline group. Unsupervised principal component analysis showed clear separation, with mucus biochemistry associated with accelerated decline, whereas biophysical mucus characteristics contributed to interindividual variability. When mucus and microbes are considered together, an accelerated decline mucus–microbiome association emerges, characterized by increased mucin (MUC5AC [mucin 5AC] and MUC5B [mucin 5B]) concentration and the presence of Achromobacter and Klebsiella. As COPD progresses, mucus–microbiome shifts occur, initially characterized by low mucin concentration and transition from viscous to elastic dominance accompanied by the commensals Veillonella, Gemella, Rothia, and Prevotella (Global Initiative for Chronic Obstructive Lung Disease [GOLD] A and B) before transition to increased mucus viscosity, mucins, and DNA concentration together with the emergence of pathogenic microorganisms including Haemophilus, Moraxella, and Pseudomonas (GOLD E).

Conclusions

Mucus–microbiome associations evolve over time with accelerated lung function decline, symptom progression, and exacerbations affording fresh therapeutic opportunities for early intervention.

At a Glance Commentary

Scientific Knowledge on the Subject

Although mucus hypersecretion and altered inflammation predispose to microbial colonization, infection, and exacerbations in chronic obstructive pulmonary disease (COPD), no work to date assesses how accelerated lung function decline relates to mucus alteration and microbiome change in COPD.

What This Study Adds to the Field

Prospective longitudinal analysis of individuals with COPD experiencing accelerated lung function decline exhibit concurrent changes in their mucus–microbiome associations not present in those without lung function decline. Mucus–microbiome associations evolve over time with lung function decline, symptom progression, and exacerbations. This affords fresh therapeutic opportunities for early intervention targeting the mucus–microbiome interface in COPD.

Chronic obstructive pulmonary disease (COPD) is characterized by respiratory symptoms resulting from abnormalities in the airway and/or alveoli that underlie persistent and often progressive airflow obstruction (1, 2). COPD progression is commonly measured using the rate of decline in lung function, but not all patients exhibit accelerated lung function decline (3–7). The Evaluation of COPD Longitudinally to Identify Predictive Surrogate End-points (ECLIPSE) study revealed high variability over a 3-year period, with 38% of patients experiencing >40-ml decline in lung function per year, whereas others experienced an increase in FEV₁ (8). Current COPD treatments focus on symptom management and exacerbation reduction, but no pharmacological intervention has reliably slowed lung function decline rate and disease progression. A critical understanding of the inflammatory, microbiological, and immune changes accompanying accelerated lung function decline may lead to interventions that prevent or delay COPD progression.

Current evidence suggests that changes in mucus composition and microbiomes play independent but significant roles in the pathogenesis, progression, and clinical outcomes of COPD (5, 9–11). Of the airway mucins, MUC5AC (mucin 5AC) and MUC5B (mucin 5B) are elevated in COPD, with MUC5AC disproportionately increased and associated with airway obstruction, small airway dysfunction, and increased exacerbation frequency (5, 9). Separately, alterations in the airway microbiome are linked to underlying inflammatory endotypes, disease progression, and exacerbation frequency in COPD. However, the relationship between the airway microbiome and mucus composition has not been extensively studied, and this relationship has never been longitudinally assessed (12–14).

Although mucus hypersecretion and altered inflammation predispose to microbial colonization, infection, and exacerbations in COPD, no work to date evaluates the relationship between mucus alteration, microbiome changes, and accelerated lung function decline in COPD. This study of the London COPD cohort incorporates prospective longitudinal analysis of the biochemical and biophysical properties of mucus, with concurrent microbiome assessment, to evaluate individuals with COPD experiencing accelerated lung function decline relative to individuals with unchanged or increased lung function. Some of the results of these studies have been previously reported in the form of an abstract (15, 16).

Methods

The London COPD Cohort Study Participants

The London COPD cohort is an ongoing longitudinal observational cohort with rolling recruitment of >200 active patients with mild to severe COPD, as described previously (17). Ethics approval was granted by the London-Hampstead Ethics Committee (REC reference: 09/H0720/8), and all participants provided written informed consent. Maintenance treatment is optimized according to Global Initiative for Chronic Obstructive Lung Disease (GOLD) recommendations, and patients are followed every 3–6 months when stable (not during an exacerbation event) (18, 19). Participants included in sputum sample analysis had not used antibiotics for at least 4 weeks and had experienced no exacerbations in the previous 8 weeks. Exacerbations were identified as 2 consecutive days with at least two major symptoms (increased dyspnea, sputum purulence, or sputum volume) or a major symptom and a minor symptom (cold, wheeze, sore throat, or cough) (20). Spontaneously expectorated sputum from a deep cough was collected in sterile containers at these visits and stored at −80°C. Participant exacerbation frequency was calculated based on the number of daily diary card events recorded during their study participation, and individuals were asked to contact the clinic when they developed exacerbation symptoms before starting antibiotics or steroids. Health status was measured using the COPD assessment test (CAT) at recruitment and each subsequent visit. The St. George’s Respiratory Questionnaire was completed at recruitment and annually thereafter.

From individuals who provided sputum samples at two different visits (the earliest referred to as the initial visit and the most recent follow-up visit), intervening FEV₁ decline rate was assessed by evaluating the FEV₁ difference, adjusted for the time elapsed between these visits. We identified individuals representing the two extremes of lung function decline: an accelerated decline group with >50 ml/yr FEV₁ loss (n = 30) and a nondecline group exhibiting no FEV₁ decline or even a slight FEV₁ increase (n = 28), respectively (total n = 58; all White individuals). To ensure avoidance of any unintentional bias based on use of FEV₁ differences to define the clinical groups between the two selected time points (chosen because of availability of clinical specimens), we further evaluated and confirmed the FEV₁ trends and relationships over time in each participant, using linear regression and all available FEV₁ measurements since cohort inception. This demonstrated the representativeness of the two clinical groups as intended and the stability of the FEV₁ relationship over time.

Only spontaneously expectorated sputum from a deep cough was used, and no induction was performed. Sputum quality was clarified using previously published and established methodology (21). To ensure consistency and standardization of assessments, all 116 sputum specimens were shipped under strict temperature control and integrity checked before use in experiments. Mucus measurements and microbiome laboratory analysis were performed at Nanyang Technological University, Singapore.

Details of biochemical and biophysical mucus assessments and shotgun metagenomic sequencing of the airway microbiome are provided in the online supplement. All sequence data from this study have been uploaded to the National Center for Biotechnology Information Sequence Read Archive under project accession: PRJNA941079.

Statistical Analysis

All continuous data were tested for normality using the Kolmogorov-Smirnoff test. The distribution differences of continuous variables (clinical, mucus, and all microbes) between lung function decline and GOLD classification were assessed using either a t test or Mann-Whitney U test, as appropriate. In paired group comparisons, either a paired t test or Wilcoxon signed-rank test was used, as appropriate. For the comparison of three or more groups, the Kruskal-Wallis test was used, followed by Dunn’s post hoc testing with Benjamin-Hochberg correction for multiple comparisons. Categorical clinical data were evaluated for differences between groups using the chi-square test. Patients were categorized based on their symptom burden and exacerbation risk according to the GOLD classification (18, 19). Correlation analysis between clinical and mucus profiles, while controlling for participants’ study duration, was performed using pairwise Spearman partial correlation in R (22).

To measure changes in mucus concentration over time, time-corrected delta-change mucus was calculated for each participant by taking the pairwise difference (i.e., Mucus_Follow-up − Mucus_Initial) divided by their study participation duration. Principal component analysis (PCA) was performed on the delta-change mucus computed across the two-sample time points for each respective individual. Data were scaled and centered before PCA analysis and implemented in R using the prcomp function. To compare differences between groups, delta-change mucus between groups was assessed using permutational multivariate ANOVA (PERMANOVA) on pairwise patient distance matrices.

Microbial α-diversity was assessed using the Shannon diversity index, and β-diversity was assessed using the Bray-Curtis dissimilarity index. Principal coordinate analysis was performed to visualize differences in β-diversity using the vegan package in R (23). Differences in microbial composition between the initial and follow-up time points were assessed using PERMANOVA with the Bray-Curtis dissimilarity index. Partial least square discriminant analysis (PLS-DA) was performed on the delta-change microbial composition and the concatenated delta-change microbiome (restricted to the top 10 distinguishing taxa) and delta-change mucus computed across the two-sample time points for each respective individual as appropriate. Time-corrected delta-change in microbial composition was calculated as the pairwise difference of the centered log ratio (CLR)-transformed microbiome composition between the two time points [i.e., CLR(microbiome_Follow-up) — CLR(microbiome_Initial)] divided by their study participation duration. A small positive value (0.001) was added to all microbiome datasets before computing the CLR transform, to address the problem of zeroes. PLS-DA was performed using the mixOmics package in R, to allow for distinguishing between groups (i.e., accelerated decline and nondecline) with parameter number of components = 2 (24). Top distinguishing taxa were calculated as the taxa with the highest magnitude (length of the vector) in the two-dimensional loading plot. Differences in the delta-change microbial composition and the concatenated delta-change microbiome (restricted to the top 10 distinguishing taxa) and delta-change mucus composition between groups were assessed using PERMANOVA on pairwise patient distance matrices derived using Euclidean distance. Differential abundance of top 10 distinguishing taxa was performed using Mann-Whitney U tests with Benjamin-Hochberg correction for multiple comparisons, on the inverse CLR-transformed time-corrected delta-change microbiome. Microbiome datasets are compositional and require transformations (such as CLR) before conventional arithmetic operations (such as subtraction used for computing delta-change), with limitations to interpreting the transformed data (25, 26). An inverse transformation was applied after analysis of transformed data to facilitate interpretation. Prevalence of the top 10 taxa was determined based on the inverse CLR transformed of time-corrected delta-change microbiome, by counting the number of patients with the microbe >1% abundance divided by the total patients separately for both accelerated and nondecline groups. The quality and performance of the PLS-DA models were evaluated using the area under the receiver operating characteristics curve (AUC) with leave-one-out cross-validation. All data were scaled and centered before PLS-DA analysis. Microbial and mucus characteristics were compared using the Mann-Whitney U test with Benjamin-Hochberg correction for multiple comparisons. All statistical analyses were performed using a custom script written in R (v4.1.2) and graphs designed using GraphPad Prism (v9.5). P values <0.05 were considered statistically significant.

Results

Demographics and Clinical Characteristics

At enrollment, the 58 participants (69% male) had a median age of 71.3 years (interquartile range [IQR], 12.2 yr) and had accrued a median 54.5 (IQR, 40) pack-year smoking history, with 25 (43%) continuing to smoke (Table 1). Participants belonged to GOLD grades 1, 2, 3, and 4 (distributed 10%, 41%, 34%, and 17%, respectively), with an overall mean FEV₁% predicted of 53.0% (SD, 20.6%), with most (78%) using triple inhaled therapy. Mean time between initial and follow-up visits was 2.4 years (SD, ±2.6 yr) (range, 8 months to 12 years, across the years 2007–2019), with similar follow-up time between groups (Table 2). In the accelerated decline group (n = 30), FEV₁ declined by a median 156 ml (IQR, 79 ml/yr; P < 0.0001) or a mean 4.76% predicted decline per year (SD, 3.44%; P < 0.0001). In the nondecline group (n = 28), FEV₁ increased by median 45 ml (IQR, 135 ml/yr; P < 0.0001) or a mean 2.13% predicted increase per year (SD, 5.3%; P < 0.0001). Relative to the nondecline group, members of the accelerated decline group were more commonly ongoing smokers (63% vs. 21%, respectively; P < 0.01), recorded higher initial FEV₁% predicted values (mean, 63% vs. 43%; P < 0.0001) and had lower-risk GOLD classification (A/B/C/D: 6/13/1/10 vs. 0/10/3/15; P < 0.05) (Table 1). After adjusting for the duration of each participant’s enrollment in the study, we detected no statistically significant differences in either baseline symptom scores (CAT and St. George’s Respiratory Questionnaire scores) or exacerbation frequency during follow-up between the groups (Table 2 and Figure E2 in the online supplement).

Table 1.

Demographics of the Study Cohort

Characteristics	All COPD	Accelerated Decline	Nondecline	P Value
Number of patients, n	58	30	28	NA
Age, yr, median (IQR)	71.3 (12.2)	68.1 (11.9)	74.1 (13.3)	ns
Sex, male/female, n*	40/18	24/6	16/12	ns
BMI, kg ⋅ m−2, median (IQR)	27.1 (7.8)	26.3 (8.7)	27.3 (9.0)	ns
Current/ex-smoker, n*	25/33	19/11	6/22	<0.01
Smoking pack-years, median (IQR)	54.5 (40)	60.0 (31.8)	45.4 (44.0)	ns
Sputum producer, yes/no, n*	38/20	22/8	16/12	ns
Participation in study, yr, mean ± SD	2.4 ± 2.6	2.4 ± 2.8	2.4 ± 2.4	ns
FEV1, L, median (IQR)	1.44 (0.85)	1.76 (0.62)	1.01 (0.73)	<0.0001
Change in FEV1 over the study period, ml/yr, median (IQR)	−77.49 (203.81)	−156.01 (79.09)	45.48 (135.78)	<0.0001
FEV1% predicted, mean ± SD	53.02 ± 20.62	62.77 ± 18.47	42.58 ± 18.12	<0.0001
Change in FEV1% predicted/yr, over the study period, median (IQR)	−1.57 (7.02)	−4.67 (3.44)	2.13 (5.30)	<0.0001
FVC, L, median (IQR)	2.69 (1.09)	2.90 (1.19)	2.16 (0.72)	<0.0001
FEV1/FVC ratio, mean ± SD	0.52 ± 0.14	0.55 ± 0.12	0.47 ± 0.16	ns
CAT score, mean ± SD	17.5 ± 7.1	15.9 ± 7.2	19.1 ± 6.9	ns
SGRQ score, mean ± SD	46.8 ± 14.9	42.1 ± 13.9	51.0 ± 15.1	ns
Annual exacerbation frequency over the study period, median (IQR)	1.7 (2.3)	1.7 (2.4)	1.7 (2.2)	ns
Hematocrit ratio, mean ± (SD)	0.43 ± 0.05	0.43 ± 0.04	0.42 ± 0.05	ns
GOLD grade (I/II/III/IV), n†‡	6/24/20/8	5/17/7/1	1/7/13/7	<0.01
GOLD classification (A/B/C/D), n†‡	6/23/4/25	6/13/1/10	0/10/3/15	<0.05
GOLD classification (A/B/E), n†§	6/23/29	6/13/11	0/10/18	ns
Medication, n (%)
LABA/LAMA/ICS*	45 (77.6)	22 (73.3)	23 (82.1)	ns
LABA/ICS*	2 (3.4)	0 (0.0)	2 (7.1)	ns
LAMA/ICS*	1 (1.7)	1 (3.3)	0 (0.0)	ns
LABA/LAMA*	0 (0.0)	0 (0.0)	0 (0.0)	ns
LABA monotherapy*	4 (6.9)	4 (13.3)	0 (0.0)	ns
LAMA monotherapy*	6 (10.3)	3 (10.0)	3 (10.7)	ns

Definition of abbreviations: BMI = body mass index; CAT = COPD Assessment Test; COPD = chronic obstructive pulmonary disease; GOLD = Global Initiative for Chronic Obstructive Lung Disease; ICS = inhaled corticosteroids; IQR = interquartile range; LABA = long-acting β2 agonist; LAMA = long-acting muscarinic antagonist; NA = not applicable; ns = nonsignificant; SGRQ = St. George’s Respiratory Questionnaire.

^*Chi-square test for two groups.

^† Kruskal-Wallis test followed by Dunn’s post hoc testing for three or more groups.

^‡ Defined according to GOLD criteria 2017 (19).

^§ Defined according to GOLD criteria 2023 (18).

Table 2.

Demographics of the Study Cohort at the Initial and Follow-Up Time Points

Characteristics	Accelerated Decline			Nondecline
	Initial	Follow-Up	P Value	Initial	Follow-Up	P Value
Number of patients, n	30	30	NA	28	28	NA
Current/ex-smoker, n*	19/11	14/16	ns	6/22	5/23	ns
Smoking pack-years, median (IQR)	60.0 (31.8)	53.6 (42.4)	ns	45.4 (44.0)	46.0 (55.0)	ns
FEV1, L, median (IQR)	1.76 (0.62)	1.26 (0.68)	<0.0001	1.01 (0.73)	1.28 (0.93)	<0.0001
FEV1% predicted, mean ± SD	62.77 ± 18.47	47.53 ± 16.71	<0.0001	42.58 ± 18.12	51.86 ± 20.21	<0.0001
FVC, L, median (IQR)	2.90 (1.19)	2.63 (1.42)	<0.001	2.16 (0.72)	2.52 (0.97)	<0.001
FEV1/FVC ratio, mean ± SD	0.55 ± 0.12	0.46 ± 0.12	<0.0001	0.47 ± 0.16	0.50 ± 0.16	ns
Hematocrit ratio, mean ± (SD)	0.43 ± 0.4	0.41 ± 0.04	ns	0.42 ± 0.05	0.41 ± 0.05	ns
CAT score, mean ± SD	15.9 ± 7.2	17.0 ± 7.3	ns	19.1 ± 7.0	19.6 ± 6.6	ns
SGRQ score, mean ± SD	42.1 ± 13.9	45.7 ± 23.1	ns	53.3 ± 23.6	49.0 ± 16.8	ns
GOLD Grade (I/II/III/IV), n†‡	5/17/7/1	1/10/15/4	ns	1/7/13/7	1/14/7/6	ns
GOLD Classification (A/B/C/D), n†‡	6/13/1/10	3/16/1/10	ns	0/10/3/15	0/16/0/12	ns
GOLD Classification (A/B/E), n†§	6/13/11	3/16/11	ns	0/10/18	0/16/12	ns
Medication, n (%)
LABA/LAMA/ICS*	22 (73.3)	24 (80)	ns	23 (82.1)	20 (71.4)	ns
LABA/ICS*	0 (0)	0 (0)	ns	2 (7.1)	2 (7.1)	ns
LAMA/ICS*	1 (3.3)	1 (3.3)	ns	0 (0)	0 (0)	ns
LABA/LAMA*	0 (0)	1 (3.3)	ns	0 (0)	0 (0)	ns
LABA monotherapy*	4 (13.3)	3 (10)	ns	0 (0)	2 (7.1)	ns
LAMA monotherapy*	3 (10)	1 (3.3)	ns	3 (10.7)	4 (14.3)	ns

Definition of abbreviations: CAT = Chronic Obstructive Pulmonary Disease Assessment Test; GOLD = Global Initiative for Chronic Obstructive Lung Disease; ICS = inhaled corticosteroids; IQR = interquartile range; LABA = long-acting β2 agonist; LAMA = long-acting muscarinic antagonist; NA = not applicable; ns = nonsignificant; SGRQ = St. George’s Respiratory Questionnaire.

^*Chi-square test for two groups.

^† Kruskal-Wallis test followed by Dunn’s post hoc testing for three or more groups.

^‡ Defined according to GOLD criteria 2017 (19).

^§ Defined according to GOLD criteria 2023 (18).

Longitudinal Change in Biochemical and Biophysical Mucus Properties

Longitudinal mucus assessment using paired samples (total 116 specimens) revealed that biochemical and biophysical mucus parameters changed differently between the two study groups. Within the accelerated decline group, we observed increases in percentage mucus solids (+52.4%; P < 0.0001; biochemical), total mucin concentration (+23.7%; P < 0.0001; biochemical), MUC5AC concentration (+52%; P < 0.001; biochemical), MUC5B concentration (+46.1%; P < 0.001; biochemical), DNA concentration (20.1%; P < 0.001; biochemical), and MUC5AC/MUC5B ratio (+63.8%; P < 0.05; biochemical) (Figure 1). By comparison, within the nondecline group these values trended downward (−4.5%, −30.4%, −4.5%, −1.5%, −10.5%, and −15.4%, respectively), although these changes were not statistically significant.

Figure 1.

Mucus biochemical profiles in chronic obstructive pulmonary disease with accelerated decline in lung function demonstrate significantly elevated (A) mucus solids, (B) total mucins, (C) MUC5AC, (D) MUC5B, (E) MUC5AC/MUC5B ratio, and (F) DNA concentration compared with individuals exhibiting no lung function decline between the initial and follow-up time points. P values calculated using Wilcoxon signed-rank test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. DNA = sputum DNA concentration; MUC5AC = mucin 5AC; MUC5B = mucin 5B; ns = nonsignificant.

Similarly, the accelerated decline group showed a significant longitudinal increase in complex viscosity of mucus (+72.3%; P < 0.0001; biophysical), reflecting increases in both its elastic (+68.5%; P < 0.0001; biophysical) and viscous (+63.2%; P < 0.01; biophysical) components, and reduced tanδ (−69.8%; P < 0.05; biophysical), indicative of an increased elastic versus viscous mucus dominance (Figure 2). In the nondecline group, mucus complex viscosity, constituent elastic and viscous components, and tanδ trended toward an increase, yet these changes were nonsignificant and smaller than those in the accelerated decline group (+26.9%, +7.1%, +54.8%, −51.7%). Despite a substantial change to mucus properties in the accelerated decline group over time, it is important to recognize that the nondecline group had much higher baseline values, levels of which are comparable to the follow-up time point in the accelerated decline group. It is important to highlight that Figures 1 and 2 present data without accounting for each individual’s duration of study participation.

Figure 2.

Mucus biophysical profiles in chronic obstructive pulmonary disease with accelerated decline in lung function demonstrate significantly elevated (A) complex viscosity, constituent (B) elastic and (C) viscous components, and (D) ratio of viscous to elastic components (tanδ) compared with individuals exhibiting no lung function decline between the initial and follow-up time points. P values calculated using Wilcoxon signed-rank test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Rheological measurements performed at 10 rad/s. δ = ratio of the viscous to elastic effects; ns = nonsignificant; Pa = pascals; Pa.s = pascal-seconds; tanδ = the tangent of the phase angle.

After accounting for study duration between individuals, we observed that relative to the nondecline group, members of the accelerated decline group showed significant longitudinal increases in total (P < 0.01), individual mucins (P < 0.0001 for MUC5AC and MUC5B), complex viscosity (P < 0.01), and elastic component (P < 0.05) (Figures E3 and E4). No statistically significant changes over time were detected in the other mucus parameters. Although active smoking can influence biochemical and biophysical mucus properties (27, 28), the observed differences in smoking status between the accelerated decline and nondecline groups did not have any appreciable effect on our presented analyses (Figures E5 and E6).

Having demonstrated significant longitudinal biochemical and biophysical mucus change in the accelerated decline group, we next performed unsupervised PCA of the longitudinal changes observed in mucus parameters (i.e., delta-change mucus), corrected for each individual’s duration of study participation. This revealed clear separation (P < 0.0001) between the accelerated and nondecline groups (Figure 3A). To best appreciate the relative contribution of the various mucus variables driving these differences, we constructed a loadings plot (Figure 3B). Briefly, each mucus parameter is indicated as a vector with direction (arrows) and magnitude (length) indicating the direction and degree of contributed variance (i.e., longer vectors indicate stronger relationships). Although accelerated decline group membership was associated with mucus biochemical changes over time (i.e., total and individual mucins: MUC5AC and MUC5B and DNA concentration), biophysical mucus characteristics (i.e., complex viscosity and its constituent elastic and viscous components) were not. Complex viscosity is a measure of the mucus’s resistance to deformation, combining aspects of its thickness (viscosity) and its ability to return to its original shape after being stretched or compressed (elasticity). These biophysical properties contributed to the interindividual variability observed among those belonging to the nondecline group (Figures 3A and 3B).

Figure 3.

(A) Unsupervised principal component analysis (PCA) plot demonstrating significant differences in mucus profile between accelerated decline (black circles) and nondecline (red squares) groups, with each point representing the delta-change between the two time points (initial and follow-up) after correction for each individual’s duration of study participation. Each study participant is represented by a symbol and colored according to group status, with ellipses representing 95% confidence intervals. Permutational multivariate ANOVA (PERMANOVA): P < 0.0001. (B) Individual loadings plot outlining the direction and degree of contributed variance of the biochemical (orange circles and direction arrows) and biophysical (teal squares and direction arrows) mucus components driving the observed variability seen in the presented PCA (A). P values calculated using PERMANOVA using Euclidean distance. Data centered and scaled before PCA analysis. 5AC/5B ratio = ratio of MUC5AC to MUC5B; DNA = sputum DNA concentration; G′= elastic component; G″ = viscous component; MUC5AC = mucin 5AC; MUC5B = mucin 5B; η* = complex viscosity; tanδ = the tangent of the phase angle.

As the longitudinal changes within the accelerated decline group reflected changes driven by biochemical rather than biophysical properties (Figure 3), we conducted exploratory analyses examining how these mucus parameters related to known clinical correlates of accelerated lung function decline, specifically the change in COPD symptoms (assessed by CAT score and its respective components) and exacerbation frequency in the accelerated decline group (Figure E7) (29). In line with published studies, longitudinal change in total and individual (MUC5AC and MUC5B) mucins correlated positively with exacerbation frequency; however, here we report mucus viscosity (as complex viscosity, η^*) and its malleability (viscous component [G″]), both key biophysical mucus elements, relate to increased symptoms but not exacerbation frequency (9). Taken together, these data support the existence of interrelationships between respiratory symptoms, exacerbation frequency, and longitudinal change in mucus properties among those patients with COPD experiencing accelerated decline. No significant difference was observed when examining the change between the initial and follow-up time points in mucus biochemical (Figure E8) and biophysical (Figure E9) composition, corrected for each individual’s study duration when assessed between the respective GOLD classification A, B, and E, at either time point.

Mucus–Microbiome Relationships and Lung Function Decline

Recognizing the close link between mucus and microbes (13, 30), we proceeded to evaluate the airway microbiome using shotgun metagenomics. No change in microbial composition or α- or β-diversity was observed between time points (initial and follow-up) or between the accelerated decline and nondecline groups (Figure E10). Assessment of distinguishing taxa performing PLS-DA, with correction for each participant’s duration in the study, revealed microbial change associated with accelerated decline status, driven predominantly by the increased prevalence of Achromobacter, Klebsiella, and Bordetella genera, and to a lesser extent Parvimonas (Figures 4A and 4B). This is further supported by Mann-Whitney–based differential abundance testing of the top 10 distinguishing taxa, which revealed significant increases in Achromobacter, Klebsiella, and Bordetella in the accelerated group, whereas Gemella and Veillonella are elevated in the nondecline group (Figure E11). To provide an integrated view, we performed a combined PLS-DA analysis on the concatenated mucus and microbial change (restricted to top 10 distinguishing taxa) relevant to accelerated decline (Figure 4C).

Figure 4.

(A) Supervised partial least square discriminant analysis (PLS-DA) classifying patients by decliner status and airway microbiomes. Accelerated decline (black circles) and nondecline (red squares) groups represent the pairwise difference of the centered log ratio (CLR)-transformed microbiome composition between the two time points (initial and follow-up) after correction for each individual’s duration of study participation. Each participant is represented by a symbol and colored according to group status, with ellipses representing 95% confidence intervals. Permutational multivariate ANOVA (PERMANOVA): P < 0.0001. (B) Individual loadings plot outlining genus-level airway microbiome change (purple direction arrows) driving the variability observed in the presented PLS-DA (A). (C) PLS-DA classifying patients by decliner status and incorporating biochemical and biophysical mucus assessment with airway microbiomes (i.e., the top 10 distinguishing microbial taxa identified in B). Accelerated decline (black circles) and nondecline (red squares) groups represent the pairwise difference in delta-change mucus and the CLR-transformed microbiome composition between the two time points (initial and follow-up) after correction for each individual’s duration of study participation. Each participant is represented by a symbol and colored according to group status, with ellipses representing 95% confidence intervals. PERMANOVA: P < 0.0001. (D) Individual loadings plot outlining biochemical (orange direction arrows) and biophysical (teal direction arrows) mucus assessments with airway microbiomes (purple direction arrows) (i.e., top 10 genus-level distinguishing microbial taxa) driving the observed variability seen in the presented PLS-DA (C). P values calculated using PERMANOVA using Euclidean distance. Data centered and scaled before principal component analysis analysis. MUC5AC = mucin 5AC; MUC5B = mucin 5B; 5AC/5B ratio = ratio of MUC5AC to MUC5B; DNA = sputum DNA concentration; Tan(δ) = the tangent of the phase angle.

When considering mucus and microbes together, an accelerated decline mucus–microbiome association emerges, characterized by increased total mucins, MUC5AC, and MUC5B together with prominent airway Achromobacter and Klebsiella (Figure 4D). In contrast, within the nondecline group we found no obvious mucus–microbiome relationships. The increased distinguishing potential for the accelerated decline group is gained by the incorporation of mucus assessments compared with microbiomes alone (i.e., 29% increase; microbiome alone, mean AUC with leave-one-out cross-validation of 0.6; combined microbiome + mucus model, AUC of 0.89). Comparable results were obtained when incorporating exacerbation frequency, CAT score, and smoking pack-years, with no appreciable effect on the observed mucus–microbiome associations (Figure E12).

Mucus–Microbiome Relationships to COPD Activity

Although significant lung function decline in COPD is associated with increased mucins MUC5AC and MUC5B and the microbial genera Achromobacter and Klebsiella, current GOLD criteria do not classify patients based on the rate of decline in lung function. Accordingly, we used the GOLD (2023) ABE assessment tool to categorize COPD at the initial time point to better expound mucus–microbial phenotypes (unrelated to lung function decline) that may be applicable to practice (Figure 5). Group A, exhibiting milder symptoms (CAT score: <10) and infrequent exacerbations (fewer than two exacerbations per year), showed a low concentration of mucins. Their mucus was less thick (reduced viscosity) and less springy (reduced elasticity), making it more easily deformable or moldable (increased malleability), with an overrepresentation of Veillonella and Gemella. Group B, experiencing more severe symptoms (CAT score: >10) but still with infrequent exacerbations, had higher concentrations of MUC5AC and MUC5B mucins. Their mucus showed low complex viscosity, meaning it was thinner, but exhibited a relative increase in elasticity, indicating a more rubber-like quality that resists deformation and an airway microbiome populated by Rothia and Prevotella. Individuals in Group E, characterized by frequent exacerbations (more than two exacerbations per year) regardless of symptom severity, demonstrated elevated concentrations of MUC5AC and MUC5B, an increased MUC5AC/MUC5B ratio, and high DNA concentration. This group’s mucus was notably thicker and stickier (high viscosity), potentially impeding airway clearance and contributing to exacerbations, with an airway microbiome typified by potentially pathogenic microorganisms (PPMs), including Haemophilus, Moraxella, and Pseudomonas. Taken together, elevated MUC5AC and MUC5B, increased mucus viscosity, and a shift in airway microbes from commensal to PPMs are associated with disease progression when assessed by worsening symptoms, lung function decline, and increased exacerbation frequency. Finally, when the change between the initial and follow-up time points in microbiome composition, corrected for each individual’s duration of study participation, was assessed between the respective GOLD classifications A, B, and E, no significant differences were detected (Figure E13), suggesting that lung function decline as opposed to GOLD ABE may have greater importance in defining mucus–microbiome evolution in COPD.

Figure 5.

Summary of the (A) GOLD (2023) ABE assessment tool and related (B) mucus biochemical profile, (C) mucus biophysical profile, and (D) airway microbiome profile (18). (E) A combined assessment summarizing mucus properties and airway microbiome relationships within the respective categories as follows: A: low risk, fewer symptoms; B: low risk, more symptoms; E: high risk, frequent exacerbations irrespective of symptoms. P value calculated using Mann-Whitney U test with Benjamin-Hochberg correction for multiple comparisons. 5AC/5B ratio = ratio of MUC5AC to MUC5B; ABE = A (low symptoms, low risk), B (high symptoms, low risk) and E (high symptoms, high risk); DNA = sputum DNA concentration; CAT = Chronic Obstructive Pulmonary Disease Assessment Test; GOLD = Global Initiative for Chronic Obstructive Lung Disease; G* = Complex modulus; G′ = elastic component; G″ = viscous component; mMRC = modified Medical Research Council test; MUC5AC = mucin 5AC; MUC5B = mucin 5B.

Discussion

Among patients with COPD, we show that accelerated FEV₁ decline may reflect changes in mucus physiology and microbial presence. Many COPD studies consider mucus physiology, microbial change, or lung function decline, but we explore how these disease components may be longitudinally linked within the well-characterized London COPD cohort. Links between mucus biochemistry, including increased mucus solids, total and individual (MUC5AC and MUC5B) mucins, and DNA concentration, have been previously described in the literature. However, our concurrent and longitudinal assessment of accompanying biophysical mucus and microbiome change is a novel addition to the field, potentially offering fresh insight into COPD progression defined by lung function decline (5, 9).

Using detailed longitudinal data, we demonstrate that, relative to nondecline, accelerated lung function decline was more commonly accompanied by changes in mucus, with increases in total and individual mucin concentration (MUC5AC and MUC5B). After adjustment for the pairwise difference in the delta-change between the two time points (initial and follow-up), it was revealed that out of the mucus parameters assessed, only five measures—total mucus, MUC5AC, MUC5B, complex viscosity, and its elastic component—remained significantly different after correcting for participant study duration, although no change to overall microbiome composition and/or diversity indices were observed over time or between COPD decline groups. Our combined assessment of mucus and microbial change demonstrated that accelerated decline is distinguished by increased total and individual mucin (MUC5AC and MUC5B) concentration together with prominent airway Achromobacter and Klebsiella. No comparable mucus–microbiome relationships were observed in the nondecline group. The GOLD (2023) ABE assessment tool focuses on symptom burden and exacerbations as opposed to lung function decline to guide treatment in COPD. In our analysis using the GOLD ABE assessment tool to evaluate mucus–microbiome evolution, we observed notable shifts corresponding to the progression of the disease, as characterized by symptoms and exacerbation history. Initially, these changes were marked by a low complex viscosity in mucus. To clarify, complex viscosity is a measure of the mucus’s resistance to flow, incorporating aspects of both its thickness and stickiness. During the early stages, the mucus exhibited a shift from being malleable—indicating higher viscosity and a tendency to flow easily—to becoming more elastic, akin to a more rubber-like consistency. This transition was accompanied by the presence of microbial commensals, primarily observed in Group A and B, with a shift toward increased mucus viscosity, suggesting a transition to a thicker mucus state. This stage was also characterized by elevated concentrations of mucin and DNA, together with the emergence of PPMs, predominantly in Group E. These findings illustrate a dynamic interaction between mucus properties and the microbiome, which may evolve with the advancing stages of the disease. Assessments of mucus–microbiome change in relation to lung function decline and GOLD classification was initially performed independently, and then collectively, revealing the superiority of spirometric decline over time, as opposed to GOLD classification, in defining the observed mucus–microbiome evolution in COPD.

Cigarette smoke–induced impairment in mucociliary transport, increased airway dehydration, and increased mucus viscosity are key mechanisms by which smoking may drive COPD pathogenesis and progression. Individual mucin (MUC5AC and MUC5B) concentrations are important biomarkers of airway inflammation and have key roles in airway defenses, COPD progression, and exacerbations (5, 9, 31, 32). Consequently, there is burgeoning interest in developing mucus-targeting therapeutics for COPD; however, to be effective, mucin structure, configuration, function, and granule location all need careful consideration (33–35). Our findings illustrate the close links between mucus and microbes in the COPD airway, suggesting that a holistic, dual-focused therapeutic approach targeting mucus and microbes concurrently should be considered; however, the inherent complexity and interconnected relationship between mucus biology, rheology, and microbial composition (and function) requires dedicated mechanistic studies to better inform potential future interventional strategies targeting mucus and infection in COPD before translation of our findings can be realized.

The airway microbiome remains stable over time in paired samples, despite individual variations in lung function decline. This stability may be attributed to the presence of a core microbial community that remains consistent over time, including host–microbe interactions, making it difficult to identify specific microbiome changes associated with accelerated decline. Performing supervised analysis, those with accelerated decline exhibited increased presence of Achromobacter and Klebsiella. Achromobacter is a gram-negative bacterium already associated with severe disease, frequent exacerbations, hospitalizations, and mortality in COPD (36–38). Clinical isolates of Achromobacter demonstrate antimicrobial resistance and in cystic fibrosis have been linked to rapid lung function decline (39–42). Similarly, Klebsiella spp. are often associated with COPD exacerbations including antimicrobial resistance, and use of the inhaled corticosteroid fluticasone is linked to its delayed pulmonary clearance (43–46). Klebsiella pneumoniae is implicated in fluticasone-associated pneumonia in COPD, and recent work demonstrates a growth advantage and metabolic dysfunction in COPD suggesting that the organism possesses a functional repertoire to respond to and/or potentially use corticosteroids, a common COPD treatment, in the airway microenvironment (47).

Microbes share intimate interrelationships with airway mucus that provide nourishment for growth, providing mucin glycans as substrates for degradation to maintain a highly diverse microbe community (48, 49). Changes in mucin glycosylation have been observed in several inflammatory diseases, which are promoted by T helper 2 (T_H2) immune responses, particularly IL-13, that regulate the expression of glycosyltransferases and sulfotransferases both involved in glycan biosynthesis, thereby modulating mucin glycosylation (50, 51). Notably, bronchial mucins from cystic fibrosis sputum exhibit higher concentrations of sulfation, sialyation, and fucosylation than healthy individuals, making them preferential ligands for pathogenic microorganisms such as Pseudomonas aeruginosa in the airways (52). When mucus and microbes are considered together, an accelerated decline mucus–microbiome association emerges, reflecting increased mucin (MUC5AC and MUC5B) concentration and the presence of Achromobacter and Klebsiella. How this association may be used to identify accelerated lung function decline among patients with COPD and how it relates to underlying mechanisms warrants further investigation. Changes in the composition of the microbial community may partly reflect variation in mucus composition, or vice versa. Therefore, future microbiome studies should consider incorporating measures of mucus consistency to identify disease markers, as mucus consistency could potentially represent a confounding variable (53, 54). Studying comparable mucus–microbe associations in early or pre-COPD states (including after inhaled treatments) may provide insights into how our findings relate to COPD pathogenesis and inform the development of mucus/microbiome targeting therapeutics (10, 55).

Our work emphasizes the heterogeneous and complex nature of COPD progression. Although we identified patients with a steep negative slope in FEV₁ decline over time, we also identified a group exhibiting no decline or even a slight increase in FEV₁ over time. This latter group has been previously identified in the ECLIPSE study, in which 8% of individuals exhibited a >20 ml/yr increase in FEV₁ over 3 years (8). Such diverse patterns in FEV₁ change over time raises essential questions about factors influencing such distinct lung function trajectories and responsiveness to treatment.

A major strength of our study is that it prospectively studies well-characterized patients with COPD from the established London COPD cohort using longitudinal data and clinical samples collected over a prolonged time period. This enables analysis of mucus (biochemical and biophysical) and microbiome composition longitudinally, allowing, for the first time, insight into interrelationships in association with lung function decline. These longitudinal data allow us to show differences between those at both ends of the lung function decline spectrum. However, lung function decline can manifest in a wide variety of individual trajectories, and the cohort grouping we selected for this study represents only the two extreme ends of this spectrum, leaving intervening decline rates (or trajectories) unstudied (7). Uniform changes over time in lung function, the microbiome, and mucus cannot be assumed, as there is potential to miss changes in the decline rate, or sputum, between the initial and follow-up time points in either group. In addition, our nondecline group mainly consisted of individuals with already moderate-to-severe COPD at the initial time point. Some could view these individuals as having already progressed to a severe disease state, with further FEV₁ decline unlikely because of the already significantly impaired baseline lung function compared with the milder, but rapidly progressing, accelerated decline group. FEV₁ values of the accelerated group decreased to that of the nondecline group in follow-up, reinforcing the importance of considering individual patient trajectories in understanding COPD progression and interpreting our study results. In our analytical approach, using PLS-DA to identify differentially abundant taxa has limitations. Although it identifies the most distinguishing features, this does not inherently adjust for multiple comparisons, and it should be noted that the microbial and mucus signatures presented in Figure 5 illustrate the microbial and mucus characteristics that significantly differed between GOLD classifications only at the initial time point. Although this study establishes significant associations between mucus–microbiome relationships and lung function decline in COPD, we cannot make assertions on causality and whether these mucus–microbiome changes are a cause or consequence of accelerated lung function decline in COPD. Further exploration of these parameters as potential biomarkers for COPD progression remains an important area for future research. Shotgun metagenomics was used for the evaluation of microbiomes in this study. Although providing significant benefits compared with targeted amplicon approaches (e.g., 16S rRNA), including less susceptibility to PCR amplification bias, little to no influence of copy number variation, and the ability to interrogate beyond composition to study functional aspects of the microbiome, such as the antimicrobial resistance or functional pathway analysis, we did not pursue the latter largely because of our modest sample size and the high likelihood of being unable to draw clear conclusions. Future work, therefore, should incorporate larger cohorts with dedicated focus on functional, interaction, and mechanistic aspects of the microbiome–mucus relationship in COPD, including the study of other microbial kingdoms, such as the virome and mycobiome, respectively. We did not account for antibiotic exposures beyond the 4-week period preceding study recruitment, and, therefore, our results should be interpreted accordingly. Finally, the careful selection of patients led to a relatively modest sample size, and the practically necessary mucus sample freeze–thaw cycle could influence rheology measurements.

In conclusion, we link accelerated FEV₁ decline among patients with COPD to a mucus–microbiome association comprising increased total and individual mucins and the presence of Achromobacter and Klebsiella. This evolution of mucus–microbe relationships in COPD provides novel insight informing therapeutic developments to target the mucus–microbiome interface.