Sputum Cellularity and MRI Ventilation Defects in Severe Asthma

Abstract

Background

Airway inflammation is a hallmark feature of asthma characterized by elevated eosinophils and/or neutrophils. Eosinophils in sputum can contribute to ventilation defects. The functional consequence of other types of cellularity on ventilation is unknown.

Research Question

What is the effect of different sputum cellular phenotypes on abnormal ventilation evaluated by ¹²⁹Xe MRI in patients with severe asthma?

Study Design and Methods

Eighty-five patients with severe asthma and 15 healthy control participants underwent ¹²⁹Xe ventilation MRI. Sputum cytometry was performed to evaluate airway inflammation and to stratify patients with asthma into 4 cellular phenotypes (paucigranulocytic, eosinophilic, neutrophilic, and mixed-granulocytic). Abnormal ventilation, quantified as the postbronchodilator MRI ventilation defect percent (VDP), was compared between asthma sputum cellular phenotypes and to healthy control participants. Demographics, clinical characteristics, and sputum cytokine levels of patients with paucigranulocytic asthma with MRI VDP above and below the upper limit of normal were also compared.

Results

In patients with asthma, ventilation was abnormal (MRI VDP above the upper limit of normal) for 44% (14 of 32) of those with paucigranulocytic sputum, 64% (14 of 22) with eosinophilic bronchitis, 75% (9 of 12) with neutrophilic bronchitis, and 89% (17 of 19) with mixed-granulocytic bronchitis. Compared to healthy control participants, MRI VDP was higher in participants with asthma with eosinophilic (adjusted P = .0002), neutrophilic (adjusted P = .0001), and mixed-granulocytic phenotypes (adjusted P < .0001) but not a paucigranulocytic phenotype (adjusted P = .051). Among participants with paucigranulocytic asthma, those with an MRI VDP above the upper limit of normal were older (P = .006), had higher fractional exhaled nitric oxide (P = .02), and higher CT mucus scores (P < .0001).

Interpretation

Our results indicate that in severe asthma, ventilation is abnormal in the presence of intraluminal cellular inflammation, irrespective of phenotype. Abnormal ventilation in paucigranulocytic asthma may be due to airway mucus and was shown to often be associated with elevated fractional exhaled nitric oxide.

Graphical Abstract

Take-Home Points.

Study Question: What is the effect of different sputum cellular phenotypes on abnormal ventilation evaluated by ¹²⁹Xe MRI in patients with severe asthma?

Results: We reaffirm the contribution of airway eosinophils to ventilation defects and newly demonstrate that elevated airway neutrophils, with or without concomitantly elevated airway eosinophils, are associated with abnormal ventilation. In addition, we observed that a subset of patients with normal cellularity (ie, paucigranulocytic phenotype) also have abnormal ventilation. These patients are characterized by older age, raised fractional exhaled nitric oxide, and greater mucus burden.

Interpretation: Our results show that in severe asthma, ventilation is abnormal in the presence of intraluminal cellular inflammation, irrespective of phenotype. Abnormal ventilation in paucigranulocytic asthma may be due to airway mucus and was shown to often be associated with elevated fractional exhaled nitric oxide.

Analysis of sputum to determine inflammatory cellular patterns has led to the recognition of 4 distinct airway inflammatory phenotypes (eosinophilic, neutrophilic, mixed-granulocytic, and paucigranulocytic)¹ of asthma that provide a pathobiological understanding of disease, and has been a focal point of precision therapy.^2,3 Eosinophil-predominant^4,5 and neutrophil-predominant^{6, 7, 8} asthma have been associated with more severe disease, frequent exacerbations, and loss of lung function among other adverse outcomes. Some studies suggest patients with asthma who have elevated airway eosinophils and neutrophils, known as mixed-granulocytic inflammation, have the most severe disease.^{8, 9, 10} Although there are important clinical repercussions for the 4 well-recognized airway inflammatory phenotypes of asthma,^{5,8, 9, 10} differences in their functional consequence on airflow obstruction remains unclear.

Ventilation defects are the functional consequence of airflow obstruction and can be directly visualized in vivo at high resolution by hyperpolarized gas MRI. In patients with asthma, MRI ventilation defects are spatially related to narrowed and obstructed airways quantified by CT imaging,^11,12 and are sensitive to both inflammatory and noninflammatory components of asthma pathology.^{11, 12, 13, 14, 15} We previously demonstrated that ventilation defects, evaluated by hyperpolarized ³He MRI, that persist after bronchodilators are greater in patients with asthma with uncontrolled eosinophilic bronchitis as compared with those with controlled eosinophilic bronchitis,¹³ and that ventilation defects improve when airway eosinophilia is controlled with anti-type 2 (T2) biologic therapy.¹⁶ The association of MRI ventilation defects with neutrophilic¹⁵ and eosinophilic¹⁷ bronchoalveolar lavage (BAL), and sputum eosinophil count,¹³ have also been reported. These findings suggest the pathophysiologic abnormalities responsible for ventilation defects that persist after bronchodilator use are largely the functional consequence of uncontrolled airway inflammation. Differences in abnormal ventilation between airway inflammatory phenotypes of severe asthma have not been examined. Furthermore, the cause of residual ventilation defects in patients identified as paucigranulocytic, who have no intraluminal cellular inflammation, remains unclear.

In patients with severe asthma, we hypothesized that elevated levels of inflammatory cells within the airway lumen, irrespective of the specific cellular phenotype, would be associated with abnormal ventilation quantified by ¹²⁹Xe MRI. Therefore, the objective of this study was to characterize the ventilation defect burden, evaluated by ¹²⁹Xe MRI, in specific sputum cellular phenotypes of severe asthma.

Study Design and Methods

Study Participants and Design

This was a single-center retrospective analysis of 85 adults with severe asthma and 15 age-matched healthy control participants who completed research assessments at St. Joseph’s Healthcare Hamilton (Hamilton, ON, Canada) under ethics board-approved (Hamilton Integrated Research Ethics Board 3932, 4758, 13308) and registered protocols.^{18, 19, 20} All participants with severe asthma were under the care of a respirologist, and their diagnosis of severe asthma was in accordance with the Global Initiative for Asthma treatment step criteria.²¹ During a single study visit, all healthy control participants provided demographic information and underwent hyperpolarized ¹²⁹Xe ventilation MRI, and all participants with severe asthma provided demographic information, reported their asthma control and quality of life using the 5-point Asthma Control Questionnaire²² and Asthma Quality of Life Questionnaire,²³ and performed spirometry and hyperpolarized ¹²⁹Xe ventilation MRI after bronchodilator administration (four 100-μg doses of salbutamol). Sputum was collected from participants with asthma on the same day as ventilation MRI, and they were retrospectively grouped on the basis of their cellular inflammatory phenotype derived from sputum cytology.

Sputum Cellular Phenotypes

Intraluminal cellular inflammation was evaluated by quantitative cytology of spontaneous or induced sputum. Sputum was induced, processed, and quantified according to standardized methods, as previously described.²⁴ Irrespective of the method of sputum collection, both spontaneous and induced sputum have demonstrated good agreement to measure airway inflammation.²⁵ Participants were retrospectively classified into the following sputum cellular phenotypes determined by sputum cell counts^24,26,27: (1) eosinophilic (≥ 2.3% eosinophils and/or moderate or many eosinophil-free granules), (2) neutrophilic (≥ 64.4% neutrophils and total cell count ≥ 9.7 × 10⁶ cells/g, and/or an absolute neutrophil count ≥ 4.9 × 10⁶ cells/g), (3) mixed-granulocytic (eosinophilic and neutrophilic), and (4) paucigranulocytic (neither eosinophilic nor neutrophilic). Participants who were unable to provide a sufficient sputum sample (< 0.08 g of sputum separated from saliva)²⁸ following 3 rounds of saline inhalation (21 minutes total in duration) were considered paucigranulocytic.

Other Biomarkers of Airway Inflammation

Inflammation was also assessed using fractional exhaled nitric oxide (Feno) and blood eosinophil count. Feno was measured with a portable exhaled NO analyzer (Niox Vero; Niox). Peripheral blood was collected to determine the peripheral blood eosinophil count. Cytokine levels in sputa were measured with an automated enzyme-linked immunosorbent assay system (Ella ELISA; Bio-Techne). The cytokines assayed were IL-5, IL-4, IL-13, and IL-33 (T2 inflammation) and IL-1β, B cell-activating factor, IL-6, IL-12p70, IL-15, IL-17A, IL-18, interferon-γ, and tumor necrosis factor-α (T1/T17 inflammation).²⁹

Lung Imaging Acquisition and Analysis

Anatomical proton (¹H) and hyperpolarized ¹²⁹Xe ventilation MR images were acquired after bronchodilator administration using a 3-T MRI system with broadband imaging capabilities (General Electric Health Care) as previously described.³⁰ Participants were instructed to inhale 1 L of gas (100% N₂ for ¹H MRI and a ¹²⁹Xe-N₂ mixture for ¹²⁹Xe MRI) from functional residual capacity, and 16 to 18 coronal slices were acquired under breath-hold conditions. Quantitative MRI evaluation was performed by a single trained observer (S. S.) using semiautomated segmentation software as previously described.³¹ Ventilation defect burden was quantified as the ¹²⁹Xe MRI ventilation defect percent (VDP), defined as the ventilation defect volume normalized to the thoracic cavity volume.³¹ In a subset of participants, postbronchodilator inspiratory chest CT scans were acquired at total lung capacity using a 64-slice LightSpeed VCT scanner (General Electric Health Care). CT data sets were reviewed by a pulmonologist (H. S. or C. V. G.) for the presence of mucus plugs, and the CT mucus score was determined as previously described.³²

Statistical Methods

This analysis was conducted retrospectively, without a priori sample size calculation. However, a post-hoc calculation for 1-way analysis of variance (ANOVA) revealed an effect size of 0.45 and power (1 – β) of 0.94, indicating that there was likely sufficient statistical power to detect if the MRI VDPs differed between the 5 study groups (ie, 4 sputum cellular phenotypes and healthy control participants). Data were tested for normality by Shapiro-Wilk test, and when data were not normal, nonparametric tests were performed. Differences in participant demographics, clinical characteristics, and MRI VDP between sputum cellular phenotypes were determined using a 1-way ANOVA or Kruskal-Wallis test, and between-group differences were evaluated using a Dunnett test or the Dunn multiple comparisons test, respectively. Unadjusted (not corrected for multiple comparisons) and adjusted (corrected for multiple comparisons) P values are reported. The upper limit of normal (ULN) of VDP was determined as the mean + 2 SD of the healthy control group. To understand potential contributors to abnormal ventilation in paucigranulocytic asthma, participants were dichotomized on the basis of postbronchodilator VDP above and below the ULN. Between-group differences in demographics, clinical characteristics, and cytokine levels in sputa were evaluated by unpaired t tests or Mann-Whitney tests. Receiver operating characteristic (ROC) curves were generated to independently assess the diagnostic performance of inflammatory markers in identifying abnormal ventilation. The direction of the ROC analysis was set such that larger values for all test variables indicated a positive outcome (ie, abnormal ventilation). Univariate relationships were evaluated with Spearman (ρ) correlation coefficients. Statistical analyses were performed with GraphPad Prism version 9.3.1 (GraphPad Software), SPSS (version 25.0; IBM SPSS Statistics), and R Statistical Software version 4.3.3 (R Foundation for Statistical Computing). All results were considered statistically significant when the probability of making a type I error was less than 5% (P < .05).

Results

Participant Characteristics

Table 1 summarizes demographic data and clinical characteristics for 85 participants with severe asthma (all and separated by cellular phenotype) and 15 healthy control participants. Sputum cellular phenotype classification of participants with asthma revealed 32 (38%) participants with the paucigranulocytic phenotype, 22 (26%) participants with the eosinophilic phenotype, 12 (14%) participants with the neutrophilic phenotype, and 19 (22%) participants with the mixed-granulocytic phenotype. Asthma cellular phenotypes were similar with respect to participant sex, BMI, asthma control, quality of life, spirometry, and inhaled/oral corticosteroid dose (P > .05 for all).

Table 1.

Participant Demographics and Clinical Characteristics

Demographic/Clinical Characteristic	Healthy (n = 15)	Asthma (n = 85)	Asthma by Sputum Cellular Phenotype				P Valuea
			Pauci (n = 32)	Eosinophilic (n = 22)	Neutrophilic (n = 12)	Mixed (n = 19)
Age, y	46 (33)	55 (22)	48 (19)	60 (27)	59 (22)	55 (13)	.03b
Female, No. (%)	8 (53)	48 (56)	21 (66)	11 (50)	7 (58)	9 (47)	—
BMI, kg/m2	25 (6)	29 (8)	29 (8)	26 (10)	28 (8)	28 (9)	.06c
Smoking history (pack-years)	—	0 (3)	0 (2)	0 (7)	0 (3)	0 (1)	.36b
ACQ-5 score	—	2.5 (2.0)	2.8 (2.0)	2.4 (2.2)	2.6 (2.8)	2.5 (2.2)	.65c
AQLQ score	—	4.3 (2.2)	4.0 (1.9)	5.0 (1.9)	3.9 (2.0)	4.8 (2.2)	.29c
Post-BD spirometry
FEV1, % pred	99 (23)	72 (30)	72 (27)	75 (34)	77 (39)	66 (36)	.78c
FVC, % pred	102 (17)	86 (18)	85 (19)	89 (17)	88 (26)	82 (19)	.89c
FEV1/FVC, %	82 (9)	65 (18)	66 (18)	64 (22)	63 (16)	63 (14)	.65c
Inflammatory biomarkers
Sputum eosinophils, %	—	3.4 (14)d	0 (0)e	11.5 (19)f	0.5 (1)	11.8 (17)	< .001b
Sputum neutrophils, %	—	58 (44)d	45 (34)e	33 (34)f	89 (17)	70 (31)	< .001b
Sputum total cell count, × 106 cells/g	—	6.5 (13)g	3.8 (3)e	3.2 (4)	16.3 (42)	20.8 (22)	< .001b
Feno, ppb	—	37 (50)	28 (36)	59 (61)	30 (30)	44 (63)	.01b
Blood eosinophils, × 109 cells/L	—	0.3 (0.4)	0.1 (0.3)	0.5 (0.7)	0.2 (0.2)	0.5 (0.6)	< .001b
IgE, kIU/L	—	160 (254)h	168 (283)i	214 (244)j	56 (113)k	365 (519)l	.09b
Asthma medications	—
ICS dose, μg/d	—	1,000 (500)	1,000 (500)	750 (500)	1,000 (688)	1,000 (500)	.44b
OCS dose, mg/d	—	0 (1.3)	0 (5.4)	0 (1.3)	0 (0)	0 (0)	.40b
OCS dependent, No. (%)	—	21 (25)	11 (34)	5 (23)	2 (17)	3 (16)	—
Anti-T2 biologic, No. (%)	—	18 (21)m	10 (31)m	4 (18)	2 (17)	2 (11)	—
Omalizumab, No. (%)	—	1 (6)	1 (10)	0	0	0	—
Mepolizumab, No. (%)	—	3 (17)	1 (10)	2 (50)	0	0	—
Benralizumab, No. (%)	—	4 (22)	3 (30)	0	1 (50)	0	—
Reslizumab, No. (%)	—	4 (22)	1 (10)	1 (25)	1 (50)	1 (50)	—
Dupilumab, No. (%)	—	5 (28)	3 (30)	1 (25)	0	1 (50)	—
Tezepelumab, No. (%)	—	2 (11)	2 (20)	0	0	0	—
Post-BD 129Xe ventilation MRI
VDP, %	1.3 (0.8)	5.2 (8.5)	2.6 (4.8)	5.1 (12.9)	6.7 (14.5)	10.8 (14.4)	.0002b

Data are presented as median (interquartile range) except where indicated otherwise. Em dash means not determined. % pred = % predicted; ACQ-5 = 5-item Asthma Control Questionnaire; AQLQ = Asthma Quality of Life Questionnaire; BD = bronchodilator; Feno = fractional exhaled nitric oxide; ICS = inhaled corticosteroid by fluticasone propionate equivalent; OCS = oral corticosteroid by prednisone equivalent; Pauci = paucigranulocytic; Mixed = mixed-granulocytic T2 = type 2; VDP = ventilation defect percent.

^aSignificance of difference (P < .05) between groups determined by 1-way ANOVA or Kruskal-Wallis test.

^bKruskal-Wallis test.

^cOne-way ANOVA test.

^dn = 68.

^en = 16.

^fn = 21.

^gn = 69.

^hn = 49.

ⁱn = 24.

^jn = 10.

^kn = 7.

^ln = 8.

^mn = 1 dual biologic (dupilumab and benralizumab).

Contribution of Inflammation to Abnormal Ventilation

Figure 1 shows representative ¹²⁹Xe ventilation MRI coronal slices (in cyan) coregistered to the ¹H anatomical MRI (in grayscale) for a representative healthy control participant, and participants with asthma who have paucigranulocytic, eosinophilic, neutrophilic, and mixed-granulocytic sputum cellular phenotypes. Qualitatively, ¹²⁹Xe MRI revealed that ventilation was more heterogeneous and there were more ventilation abnormalities observed for each inflammatory phenotype, as compared with the healthy control participants and paucigranulocytic phenotype. Figure 2 shows the ventilation defect burden quantified by MRI VDP for healthy control participants and all participants with asthma subdivided by sputum cellular phenotype. The upper limit of normal for MRI VDP in the healthy control group was 3.7%, and the ventilation defect burden was abnormal (MRI VDP ≥ ULN) for 44% of patients with asthma with paucigranulocytic sputum, 64% of those with eosinophilic bronchitis, 75% with neutrophilic bronchitis, and 89% with mixed-granulocytic bronchitis. MRI VDP was different between healthy control participants and sputum phenotypes (Kruskal-Wallis test, P < .0001). Compared with healthy control participants (VDP = 1.3% [0.8]), MRI VDP was greater for participants with asthma and eosinophilic bronchitis (5.1% [12.9], unadjusted P < .0001/adjusted P = .0002), neutrophilic bronchitis (6.7% [14.5], unadjusted P < .0001/adjusted P = .0001), and mixed-granulocytic bronchitis (10.8% [14.4], unadjusted P < .0001/adjusted P < .0001) but not paucigranulocytic sputum (2.6% [4.8], unadjusted P = .01/adjusted P = .051) after adjustment for multiple comparisons. MRI VDP was also different between sputum cellular phenotypes of asthma (Kruskal-Wallis test, P = .0002). Compared with participants with paucigranulocytic sputum, MRI VDP was greater for those with neutrophilic bronchitis (unadjusted P = .009/adjusted P = .03) and mixed-granulocytic bronchitis (unadjusted P < .0001/adjusted P < .0001), but not eosinophilic bronchitis (unadjusted P = .02/adjusted P = .07) after adjustment for multiple comparisons.

Figure 1.

¹²⁹Xe ventilation MRI of a representative healthy control participant, and participants with asthma with a paucigranulocytic, eosinophilic, neutrophilic, or mixed-granulocytic sputum cellular phenotype. Sputum cytospin is shown alongside 5 center coronal ¹²⁹Xe ventilation MRI slices (in cyan) coregistered to the anatomical ¹H MRI (in grayscale). A, A 47-year-old healthy man with normal ventilation (MRI VDP = 0.8%). B, A 59-year-old woman with paucigranulocytic sputum (1.0 × 10⁶ cells/g, 0.0% eosinophils, 47.0% neutrophils, with no free granules) and normal ventilation (MRI VDP = 2.3%). C, A 68-year-old woman with eosinophilic sputum (4.3 × 10⁶ cells/g, 48.0% eosinophils, 21.8% neutrophils, with many free granules) and abnormal ventilation (MRI VDP = 15.4%). D, A 58-year-old woman with neutrophilic sputum (72.7 × 10⁶ cells/g, 0.5% eosinophils, 97.0% neutrophils, with no free granules) and abnormal ventilation (MRI VDP = 20.1%). E, A 52-year-old man with mixed-granulocytic sputum (15.4 × 10⁶ cells/g, 15.3% eosinophils, 72.3% neutrophils, with many free granules) and abnormal ventilation (MRI VDP = 19.4%). Eos = eosinophils; Neut = neutrophils; TCC = total cell count; Mixed = mixed-granulocytic; VDP = ventilation defect percent.

Figure 2.

Abnormal ventilation quantified by ¹²⁹Xe MRI VDP for all participants with asthma subdivided by sputum cellular phenotype (paucigranulocytic, eosinophilic, neutrophilic, and mixed-granulocytic), and healthy control participants. Box-and-whisker plots show minimum to maximum with individual VDP values for all participants superimposed on the plot. The plus sign within each box represents the mean. ∗Adjusted significance of difference (P < .05) from the healthy control group determined by Kruskal-Wallis with the Dunn multiple comparisons test. ^†Adjusted significance of difference (P < .05) from the paucigranulocytic group determined by Kruskal-Wallis with the Dunn multiple comparisons test. Upper limit of normal for VDP was determined as the mean + 2 SD of the healthy control group and is displayed as the horizonal line at 3.7%. The proportion of participants with asthma with a VDP greater than the upper limit of normal is reported. Open circles indicate those participants from whom insufficient sputum was obtained after 3 rounds of saline. BD = bronchodilator; Pauci = paucigranulocytic; Mixed = mixed-granulocytic; ULN = upper limit of normal; VDP = ventilation defect percent.

All participants, regardless of sputum cell count, were stratified based on established thresholds of elevated Feno (≥ 25 ppb) and blood eosinophil count (≥ 300 cells/μL) (e-Fig 1). MRI VDP was not different in participants with and without elevated Feno (e-Fig 1A; P = .70) or blood eosinophil count (e-Fig 1B; P = .80). Considering both biomarkers, MRI VDP was also not different between participants classified by either a high Feno and/or high blood eosinophil count (e-Fig 1C; Kruskal-Wallis test, P = .67).

ROC curves were used to assess the performance of markers of airway inflammation as predictors of abnormal ventilation defect burden, using the upper limit of normal (MRI VDP = 3.7%) as the diagnostic threshold. Measures of inflammation included sputum total cell count (n = 69, area under the curve [AUC] = 0.72), sputum eosinophils (n = 68, % of total cell count AUC = 0.53; n = 68, absolute cell count AUC = 0.65), sputum neutrophils (n = 68, % of total cell count AUC = 0.63; n = 68, absolute cell count AUC = 0.70), Feno (n = 85, AUC = 0.59), and blood eosinophils (n = 85, AUC = 0.49). The optimal threshold for maximum sensitivity and specificity to indicate abnormal ventilation for each marker is summarized in Table 2.

Table 2.

Measures of Inflammation as Indicators of Abnormal Ventilation Quantified by ¹²⁹Xe MRI Ventilation Defect Percent

Measure	No.	ROC AUC	Optimal Threshold	Youden Index	ACC	SENS	SPEC	PPV	NPV
Blood eosinophils	85	0.49	900 cells/μL	0.12	46%	19%	94%	83%	40%
Feno	85	0.59	78 ppb	0.19	52%	31%	87%	81%	42%
Sputum TCC	69	0.72	5.1 × 106 cells/g	0.46	74%	76%	70%	86%	54%
Sputum eosinophils
% of TCC	68	0.53	1.0%	0.23	66%	73%	50%	78%	44%
Absolute count	68	0.65	0.3 × 106 cells/g	0.31	62%	56%	75%	84%	42%
Sputum neutrophils
% of TCC	68	0.63	60.7%	0.28	62%	58%	70%	82%	41%
Absolute count	68	0.70	2.5 × 106 cells/g	0.41	71%	71%	70%	85%	50%

ACC = accuracy; AUC = area under the curve; Feno = fractional exhaled nitric oxide; NPV = negative predictive value; PPV = positive predictive value; ROC = receiver operating characteristic; SENS = sensitivity; SPEC = specificity; TCC = total cell count.

Contributors to Abnormal Ventilation in Paucigranulocytic Asthma

Among paucigranulocytic participants with asthma, 14 of 32 (44%) had an MRI VDP greater than the ULN, indicating the presence of noncellular contributors to airflow obstruction. To better understand these residual defects, we compared demographics, clinical characteristics (Table 3), and levels of T2 and T1/T17 cytokines in sputum (e-Fig 2) among patients with paucigranulocytic asthma with MRI VDP below or above the ULN. Patients with paucigranulocytic asthma with an abnormal ventilation defect burden were older (median difference of 13 years; P = .006), with a lower prevalence of obesity (67% vs. 21%; P = .02), lower blood eosinophils (median difference of 0.1 × 10⁹ cells/L; P = .04), higher Feno (median difference of 16 ppb; P = .02), higher CT mucus score (median difference of 6; P < .0001), and greater airflow limitation assessed by spirometry (FEV₁ and FEV₁/FVC). As shown in e-Figure 2, sputum T2 and T1/T17 cytokine levels were not significantly different between MRI VDP subgroups. Considering VDP as a continuous variable, older age (ρ = 0.56, P = .0009), higher CT mucus score (ρ = 0.76, P < .0001), lower sputum IL-18 (ρ = –0.54, P = .02), and lower blood eosinophil counts (ρ = –0.38, P = .03) were correlated with higher VDP, whereas higher Feno was not (Fig 3). A multivariable linear regression model was generated, using the forward approach to evaluate the relative influence of age, Feno, blood eosinophils, airway mucus (CT mucus score), and sputum IL-18 on VDP (R² = 0.43, P = .004). Airway mucus (unstandardized β = 0.60, P = .004) significantly added to the prediction of VDP, but the other variables in the model did not (P > .05).

Table 3.

Demographics and Clinical Characteristics for Patients With Paucigranulocytic Asthma With Normal (MRI VDP < ULN) and Abnormal (MRI VDP ≥ ULN) Ventilation Defect Burden

Characteristic	MRI VDP < ULNa (n = 18)	MRI VDP ≥ ULNa (n = 14)	P Valueb
Age, y	41 (19)	54 (13)	.006c
Patient sex (female), No. (%)	13 (72)	8 (57)	.47
BMI, kg/m2	33 (11)	29 (6)	.04c
Patients with obesity, No. (%)	12 (67)	3 (21)	.02
Smoking history, pack-years	0 (0.6)	0 (9.4)	.25d
ACQ-5 score	2.7 (2.1)	2.8 (2.0)	.71c
AQLQ score	3.6 (1.9)	4.5 (1.7)	.23c
Asthma onset, y	7 (22)	23 (29)	.34d
Asthma duration, y	34 (33)	38 (28)	.45c
Post-BD spirometry
FEV1, % pred	76 (32)	63 (30)	.02c
FVC, % pred	87 (18)	82 (17)	.21c
FEV1/FVC, %	74 (14)	59 (14)	.002c
Inflammatory biomarkers
Sufficient sputum obtained, No. (%)	7 (39)	9 (64)	.29
Sputum total cell count, × 106 cells/g	3.8 (3.0)	3.6 (3.0)	.49c
Sputum eosinophils, %	0.3 (0.7)	0.0 (0.3)	.35d
Sputum neutrophils, %	47.0 (21.9)	37.9 (47.3)	.99c
Feno, ppb	24 (25)	40 (77)	.02d
Blood eosinophils, × 109 cells/L	0.1 (0.4)	0.0 (0.1)	.04d
Asthma medications
ICS dose, μg/d	1,000 (600)	1,000 (500)	.72d
OCS dose, mg/d	0 (6.6)	0 (5.6)	.93d
OCS dependent, No. (%)	6 (33)	5 (36)	1.0
Anti-T2 biologic, No. (%)	6 (33)	4 (29)e	1.0
Omalizumab, No. (%)	1 (17)	0	—
Mepolizumab, No. (%)	1 (17)	0	—
Benralizumab, No. (%)	0	3 (75)	—
Reslizumab, No. (%)	0	1 (25)	—
Dupilumab, No. (%)	2 (33)	1 (25)	—
Tezepelumab, No. (%)	2 (33)	0	—
CT mucus
Mucus score	2 (3)f	8 (5)g	< .0001d
≥ 4 plugged segments	5 (33)	11 (92)	.005

Data are presented as median (interquartile range) except when indicated otherwise. Em dash means not determined. % pred = % predicted; ACQ-5 = 5-item Asthma Control Questionnaire; AQLQ = Asthma Quality of Life Questionnaire; Feno = fractional exhaled nitric oxide; ICS = inhaled corticosteroid by fluticasone propionate equivalent; OCS = oral corticosteroid by prednisone equivalent; T2 = type 2; ULN = upper limit of normal; VDP = ventilation defect percent.

^aUpper limit of normal for VDP is 3.7%.

^bSignificance of difference (P < .05) between groups determined by unpaired t test or Mann-Whitney test. Proportions were compared by Fisher exact test.

^cUnpaired t test.

^dMann-Whitney test.

^en = 1 dual biologic (dupilumab and benralizumab).

^fn = 15.

^gn = 12.

Figure 3.

Univariate relationships with ¹²⁹Xe MRI VDP in patients with paucigranulocytic asthma. Relationship of age (A, ρ = 0.56, P = .0009), Feno (B, ρ = 0.16, P = .38), CT mucus score (C, ρ = 0.76, P < .0001), sputum IL-18 (D, ρ = –0.54, P = .02), and blood eosinophil count (E, ρ = –0.38, P = .03) with post-bronchodilator ¹²⁹Xe MRI VDP. BD = bronchodilator; Feno = fractional exhaled nitric oxide; ppb = parts per billion; VDP = ventilation defect percent.

Discussion

This study built on previous work to further characterize the functional consequence of airway inflammatory phenotypes on abnormal ventilation in patients with severe asthma. Our findings demonstrated that the ventilation defect burden, measured by ¹²⁹Xe MRI VDP, was higher in the presence of cellular inflammation, irrespective of phenotype. Specifically, we reaffirm the contribution of airway eosinophils to ventilation defects, while newly demonstrating that elevated airway neutrophils, with or without concomitantly elevated airway eosinophils, are associated with abnormal MRI VDP. Interestingly, we observed that a subset of patients with normal cellularity (ie, paucigranulocytic phenotype) had abnormal ventilation. Thus, motivating our exploratory analysis of the paucigranulocytic phenotype, which demonstrated that these patients were characterized by older age, raised Feno, and a greater mucus burden compared with patients with normal ventilation.

It is well established that airway smooth muscle dysfunction contributes to abnormal ventilation,^14,33 and that ventilation is improved following bronchodilator therapy.^34,35 Therefore, to mitigate the effect of bronchoconstriction due to smooth muscle dysfunction on ventilation defect burden, the current study evaluated ventilation defects following bronchodilator inhalation. Under postbronchodilator conditions, participants with eosinophilic, neutrophilic, and mixed-granulocytic phenotypes were observed to have abnormal ventilation with an MRI VDP greater than that of healthy control participants. It has previously been demonstrated that sputum eosinophils are associated with ventilation defects observed by ³He MRI in asthma.¹³ In a relatively small study, Fain and colleagues¹⁵ acquired BAL from patients with asthma and observed that a greater ³He MRI ventilation defect burden in the sampled lobe (but not the whole lung) was associated with a higher percentage of neutrophils but not eosinophils. We observed that 75% of patients with a neutrophilic phenotype had abnormal ventilation, and their median VDP was 6.7%. To the best of our knowledge, this study is the first to demonstrate that elevated sputum neutrophils may also contribute to ventilation defects observed by ¹²⁹Xe MRI in asthma.

There was a notable signal indicating that patients with asthma with a mixed-granulocytic phenotype have the highest ventilation defect burden. In this group, 89% of patients had an abnormal VDP, compared with 64% and 75% in those with the eosinophilic and neutrophilic phenotypes, respectively. Their median VDP was 10.8%, which was nearly double that of patients with eosinophilic or neutrophilic inflammation in isolation. This observation aligns with prior research, which revealed that patients in whom both sputum eosinophils and neutrophils were elevated exhibit the lowest lung function^10,36 and the greatest loss of lung function over 3 years,⁹ as assessed by spirometry. In our smaller cohort, spirometry was not significantly different between sputum phenotypes. This suggests that MRI VDP may be more sensitive than spirometry in capturing the impact of cellular airway inflammation on lung function. Taken together, our observations provide evidence that elevated levels of inflammatory cells (eosinophils and/or neutrophils) in the airway contribute, in part, to abnormal ventilation in asthma.

In our cohort of relatively modest size, we also evaluated various clinical measures of inflammation as classifiers of abnormal ventilation defect burden. The best performer was sputum total cell count, with an ROC AUC of 0.72, indicating good diagnostic accuracy for abnormal ventilation. Absolute sputum eosinophils and neutrophils were more effective than corresponding sputum cell counts (represented as percentages of total cell count) in distinguishing abnormal ventilation. Blood eosinophils, with an ROC AUC of 0.49, performed near random guessing and had the poorest performance, and Feno was a modest predictor of abnormal ventilation with an ROC AUC of 0.59. The optimal threshold for blood eosinophils (900 cells/μL) was three-fold greater than the clinical cutoff, and Feno (78 ppb) was relatively near the clinical cutoff of high T2 inflammation; both had poor sensitivity (blood eosinophils, 19%; Feno, 31%) but excellent specificity (blood eosinophils, 94%; Feno, 87%) at the identified thresholds. This suggests that peripheral eosinophils and Feno at higher than currently accepted thresholds are specific tests, effective for confirming abnormal ventilation but not for ruling it out. Collectively, these observations suggest that a greater luminal burden of inflammatory cells is more accurate than indirect markers of airway inflammation to identify abnormal ventilation. However, these findings may be confounded by the varying effects of corticosteroids and anti-T2 biologics on surrogate markers of T2 inflammation.

Furthermore, based on prior research of the asthmatic airway milieu, the overwhelming luminal burden of inflammatory cells may likely be accompanied by mucus, extracellular traps, and/or Charcot-Leyden crystals, resulting in an aggregated contribution to ventilation defects.^37,38 Concurrent with the inflammatory component, a subset of patients with abnormal ventilation may have impaired smooth muscle relaxation after bronchodilator use, likely due to remodeled airways, resulting in persistent airflow obstruction. Because of the lack of a complete CT data set, we speculate that in a subset of patients residual ventilation abnormalities after bronchodilator administration may be due to a bidirectional relationship between pathologic remodeling and uncontrolled airway inflammation.³⁹ In addition, it is important to note that all inflammation measures had negative predictive values between 40% and 55%. This indicates that a paucity of inflammation does not guarantee normal ventilation and that other noninflammatory factors, including mucus plugging and airway remodeling, may contribute to abnormal ventilation.

Related to this, nearly one-half of the patients with asthma in the paucigranulocytic group (14 of 32; 44%) had abnormal ventilation. This subset with normal sputum cellularity and abnormal ventilation was characterized by older age, elevated Feno, and higher CT mucus scores compared with patients with paucigranulocytic asthma with normal ventilation. Multivariate analysis revealed that mucus plugging was the strongest contributor to abnormal VDP in patients with the paucigranulocytic phenotype. It is important to consider that all patients were receiving high-dose inhaled corticosteroids, 25% were receiving oral corticosteroids, and 21% were receiving an anti-T2 biologic for their asthma. Although the intensity of these therapies did not differ between phenotypes, it is likely they were effectively controlling eosinophilic inflammation in a subset of our cohort. In the abnormal ventilation group, the raised Feno suggests residual inflammation, not resolved by corticosteroids and anti-T2 biologics that may be a marker of mucus plugging.³⁷ Interestingly, the sputum cytokine profile in patients with the paucigranulocytic phenotype with abnormal ventilation demonstrated trends of lower sputum IL-4 and elevated sputum tumor necrosis factor-α, B cell-activating factor, and IL-6. These observations suggest a non-T2 pathway potentially contributing to residual plugging. It is important to note that sputum cytometry does not consider mast cells, which are submucosal and hence evade detection during sputum differential cell count assessment. Aberrant mast cell activity was previously demonstrated in patients with paucigranulocytic sputum⁴⁰ and associated with mucus plugging in asthma.⁴¹ Therefore, we speculate that a plausible contributor to the high mucus burden in patients with paucigranulocytic asthma may be mast cells and associated mediators. More work is required to better characterize the airway plugs in patients with absent or controlled cellular inflammation assessed by sputum.

There are limitations to this study that require consideration. First, although our study represents what is, to our knowledge, the largest report of ¹²⁹Xe MRI in patients with severe asthma, the sample sizes for individual sputum cellular phenotypes were relatively small and unequal. Disparities in sample size and low statistical power for multiple comparisons should be considered when interpreting the results. It should also be acknowledged that we considered patients to be paucigranulocytic if they were unable to expectorate sputum after three rounds of sputum induction. Related to this, various thresholds for sputum cell counts have been used to define inflammatory phenotypes of asthma, which could impact the generalizability of our results. For example, the derived ULN for neutrophil percent has ranged from 49%^27,42,43 to 93%,⁴³ with and without consideration of total cell counts with most sources using thresholds in between, while abnormal eosinophil counts have been cited between 2% and 3%.^27,42,43 In addition, cell counts and sputum phenotypes can vary over time,⁴⁴ and in response to therapies including corticosteroids, anti-T2 biologics, and antibiotics. Therefore, it is important to consider that our analysis reflects the phenotype determined on the basis of a single sputum sample collected at the time of ¹²⁹Xe MRI, and that the underlying phenotype may be masked by current asthma therapies. Furthermore, we were unable to interrogate regional association of inflammatory cell counts with focal ventilation abnormalities. Although sputum cell counts localize inflammation to the airways, the relationship of ventilation defects with luminal cellularity may be better established with cell counts from BAL samples of targeted segments with and without ventilation defects, as previously proposed by Fain et al.¹⁵ We suspect that if the relationship were interrogated regionally by image-guided bronchoscopy, we would establish the presence of inflammatory cells and/or airway remodeling localized to lung segments with observable ventilation defects. Finally, it is important to note that our primary analysis only considered inflammatory granulocytes in the airways, and CT assessment of airway mucus was limited to the paucigranulocytic group. Therefore, we did not attempt to ascertain the relative contribution of various obstructing mechanisms (eg, cells, mucus, airway remodeling) across the sputum phenotypes.

Interpretation

This study reveals the ventilation defect burden measured by ¹²⁹Xe MRI for distinct airway cellular phenotypes of asthma, including those not previously examined (eosinophilic, neutrophilic, mixed-granulocytic, and paucigranulocytic). We observed that each type of cellular inflammation is associated with an increased ventilation defect burden, and noncellular contributors to airway luminal obstruction may be identified by the elevation of Feno or better yet, mucus score, as a predictor of ventilation defects in paucigranulocytic asthma. Characterizing the ventilation defect burden associated with inflammatory phenotypes by MRI adds new insight to asthma pathophysiology.

Funding/Support

H. S. was supported by a Research Career Development fellowship award from the Western University Schulich School of Medicine and the Department of Medicine. A. T. was supported by the Canadian Institutes of Health Research Canadian Graduate Scholarship-Master’s Program. C. V. G. was supported by a Frederick Hargreave Clinical Research Fellowship award. M. M. holds the AstraZeneca Chair in Respiratory Diseases. P. N. was supported by the Frederick E. Hargreave Teva Innovation Chair in Airway Diseases. S. S. was supported by the Canada Research Chairs program and holds a Tier 2 Canada Research Chair.

Financial/Nonfinancial Disclosures

The authors have reported to CHEST the following: H. S. reports personal fees from Valeo, Sanofi, AstraZeneca, and GlaxoSmithKline, outside the submitted work. C. V. G. reports personal fees from AstraZeneca, Sanofi, and GlaxoSmithKline, outside the submitted work. M. M. reports grants from the Canadian Institutes of Health Research; grants from Methapharm Specialty Pharmaceuticals; personal fees from AstraZeneca and GlaxoSmithKline; and consultant fees from AstraZeneca, Sanofi, and Respiplus, outside the submitted work. P. N. reports grants and personal fees from AstraZeneca, GlaxoSmithKline, and Teva; grants from Sanofi, Foresee, and Cyclomedica; and personal fees from Equillium and Arrowhead Pharma, outside the submitted work. S. S. reports grants and personal fees from Cyclomedica, personal fees from GlaxoSmithKline, grants from Genentech, and personal fees from Polarean Imaging, outside the submitted work. None declared (A. T., M. K., C. H., N. R., E. S., Y. F., N. K., K. Z., N. S. T., K. R.).