Spatial persistence of reduced specific ventilation following methacholine challenge in the healthy human lung

Abstract

Specific ventilation imaging was used to identify regions of the healthy lung (6 supine subjects, ages 21–41 yr, 3 men) that experienced a fall in specific ventilation following inhalation of methacholine. This test was repeated 1 wk later and 3 mo later to test for spatial recurrence. Our data showed that 53% confidence interval (CI; 46%, 59%) of volume elements that constricted during one methacholine challenge did so again in another and that this quantity did not vary with time; 46% CI (28%, 64%) recurred 1 wk later, and 56% CI (51%, 61%) recurred 3 mo later. Previous constriction was a strong predictor for future constriction. Volume elements that constricted during one challenge were 7.7 CI (5.2, 10.2) times more likely than nonconstricted elements to constrict in a second challenge, regardless of whether the second episode was 1 wk [7.7 CI (2.9, 12.4)] or 3 mo [7.7 CI (4.6, 10.8)] later. Furthermore, posterior lung elements were more likely to constrict following methacholine than anterior lung elements (volume fraction 0.43 ± 0.22 posterior vs. 0.10 ± 0.03 anterior; P = 0.005), and basal elements that constricted were more likely than their apical counterparts to do so persistently through all three trials (volume fraction 0.14 ± 0.04 basal vs. 0.04 ± 0.04 apical; P = 0.003). Taken together, this evidence suggests a physiological predisposition toward constriction in some lung elements, especially those located in the posterior and basal lung when the subject is supine.

NEW & NOTEWORTHY The spatial pattern of bronchoconstriction following methacholine is persistent over time in healthy individuals, in whom chronic inflammation and airway remodeling are assumed to be absent. This suggests that regional lung inflation and airway structure may play dominant roles in determining the spatial pattern of methacholine bronchoconstriction.

INTRODUCTION

Asthma is a disease characterized by airflow obstruction that varies both temporally and spatially in the lung. The temporal variability is a hallmark of the disease: many patients experience long, symptom-free intervals punctuated by acute exacerbations that vary in severity between and within individuals. The spatial variability of the disease is displayed in the uneven, patchy distribution of affected lung tissue during an asthma exacerbation. This was first demonstrated in postmortem histology of subjects who died from an acute asthma attack (9) and has since been corroborated by imaging studies spanning a range of modalities: γ-scintigraphy (10, 11), single-photon emission computed tomography (19), positron emission tomography (15, 37), high-resolution computed tomography (2, 12, 25), and hyperpolarized gas MRI (1, 23, 32). This spatially heterogeneous obstruction results in regions of markedly low ventilation. Perfusion is actively regulated within these ventilation defects, by both hypoxia-mediated and non-hypoxia-mediated mechanisms (12, 18), to compensate for this low ventilation, but ventilation-to-perfusion ratio mismatch still occurs. This mismatch remains the most significant cause of hypoxemia in these patients (28).

Recently, imaging studies have begun to explore the interaction of the spatial and temporal modes of variation. Specifically, they seek to determine whether the spatial distribution of airway obstruction is a pattern that repeats throughout the natural history of the disease or whether it varies between exacerbations. Magnetic resonance imaging techniques have been most useful so far in addressing this question, as the lack of ionizing radiation makes the techniques amenable to repeated experiments. In hyperpolarized helium studies of young, moderate asthmatics given methacholine, de Lange et al. (6) showed 69% of postmethacholine defects recurred at the same location in studies performed 2 days apart. In a subsequent study in asthmatics (this time with exercise instead of methacholine to elicit bronchoconstriction), it was shown that the number of defects that recur in the same place decreases with time: from 67 to 38% for time intervals spanning 31–85 days, respectively (5). Another hyperpolarized helium study estimated the rate of spatial recurrence to be lower; in an experiment performed on patients with exercise-induced bronchoconstriction, Niles et al. (24) found that 20% of baseline defects and 29% of postexercise defects recur on separate days.

This study adds to the ongoing discourse regarding the spatial-temporal variability of repeat bronchoconstriction by studying the phenomena in healthy, nonasthmatic adults. We chose to study healthy nonasthmatics to examine the repeatability of bronchoconstriction in isolation from the possible underlying effects of chronic asthmatic inflammation and hypersensitivity. This insight into how the normal lung functions under abnormal conditions will both further our understanding of the healthy lung and act as a point of comparison for studies of pathological lung function.

Based on modeling (21) and empirical (13, 14) work, we hypothesized that the spatial pattern of bronchoconstriction would be mostly reproduced throughout multiple methacholine-induced constriction events. Modeling studies (21, 37, 39, 40) have predicted that lung regions are more likely to constrict due to local airway branching structure and/or regional lung inflation and that relatively small inhomogeneities in these properties can lead to largely deterministic patterns of bronchoconstriction (21). An imaging study by Harris et al. (13) confirmed that ventilation defects were indeed more likely to form in regions with a lower degree of lung inflation. Since airway structure and regional lung inflation are unlikely to change significantly with time in subjects without ongoing pulmonary disease, we theorized that the spatial pattern of bronchoconstriction in the healthy lung would be stable.

METHODS

Subjects and Study Sessions

This study was approved by the University of California San Diego Human Subjects Research Protection Program. Eight volunteers were recruited via advertisement and participated after giving written, informed consent. All subjects were MRI-compatible and had no history of significant pulmonary or cardiovascular disease. One subject was excluded from the study due to a methacholine challenge result that fell within the range characteristic of asthma (see below). One subject chose to withdraw from the study following the first study session. The remaining six subjects completed the entire study (for subject information, see Table 1).

Table 1.

Subject demographic data and methacholine challenge results

Subject	Sex	Age, yr	Height, m	Weight, kg	Seated FEV1, l [%Predicted]	Methacholine PC20*, mg/ml	Methacholine dose, mg/ml
S1	M	21	1.60	59	3.82 [100]	2.78	4
S2	F	27	1.57	45	3.05 [101]	6.32	8
S3	M	27	1.88	96	5.59 [110]	10.00	16
S4	F	28	1.70	77	3.54 [101]	5.15	8
S5	M	41	1.78	83	3.87 [92]	10.90	16
S6	F	27	1.64	66	3.25 [99]	3.27	4
Mean (SD)		29 (7)	1.70 (0.12)	71 (18)	3.85 (0.91) [101 (6)]	6.40 (3.40)	9 (5)

Three male (M) and three female (F) subjects were recruited by word of mouth. Average age was 29 yr, average height was 1.70 m, and average weight was 71 kg. Forced expiratory volume in 1 s (FEV₁) was measured seated as part of the subject-screening process to ensure normal lung function. Methacholine challenge testing to determine PC₂₀ was performed supine so as to mimic the subjects’ posture during future scanning visits. Methacholine dose for single-dose bronchoprovocations before scanning (far right column) was the final dose administered during the ascending-concentration challenge test. PC₂₀, concentration of methacholine that elicits 20% drop in FEV₁, determined by interpolation from an empirically determined dose-response curve.

^*Supine measurement.

Once enrolled into the study, subjects participated in four study sessions (Fig. 1). The first was a methacholine dose determination. The second session involved two MR scans, one at baseline and one postbronchoprovocation. The third session, 1 wk later, and fourth session, 3 mo later, involved one postbronchoprovocation MR scan.

Fig. 1.

Components of the 4 study sessions, with approximate time in minutes for the scanning sessions shown along the x-axis. During the 1st visit, methacholine dose (see Table 1) was determined by methacholine challenge testing performed in the supine posture per American Thoracic Society guidelines. The 2nd visit contained 2 forced expiratory volume in 1 s (FEV₁) measurements and specific ventilation imaging sessions: 1 at baseline without any drugs administered and a 2nd following single-dose administration of methacholine. All data were acquired within the ~75-min steady-state methacholine plateau (3). After the postmethacholine scan, return to baseline lung function was achieved with albuterol. The 3rd and 4th study sessions, which occurred 1 wk and ~3 mo later, contained only the postmethacholine specific ventilation imaging scan. During those visits, FEV₁ was measured before and after methacholine.

Methacholine Bronchoconstriction

Methacholine dose determination was performed for each subject according to the procedures established by the American Thoracic Society (26) for methacholine challenge testing, with the modification that the subject was supine during all parts of the process to duplicate his or her position in the MR scanner. Provocholine (methacholine chloride United States Pharmacopeia; Methapharm, Brantford, Ontario, Canada) was administered in increasing doses via a KoKo dosimeter (nSpire Health, Longmont, CO) air nebulizer system. The drug was administered in stages of five breaths each, starting at 0.03125 mg/ml and doubling concentration every stage to a maximum dosage of 16 mg/ml. Before the first stage, a dose of nebulized saline diluent was given. Following each stage of saline and drug administration, spirometry was performed to determine forced expiratory volume in 1 s (FEV₁). The drug titration was terminated when the fall in FEV₁ exceeded 20% of the subject’s postdiluent value. The concentration of methacholine that elicited a 20% drop in FEV₁ (PC₂₀) for each subject was determined from his or her dose-response curve. The dose chosen for the rest of the study (Table 1) was the final dose administered in the titration, the lowest dose that elicited a >20% reduction in FEV₁.

During each subsequent imaging session, bronchoconstriction was achieved via supine, single-dose administration of the final dose from the subject’s PC₂₀ determination. Provocholine was delivered using the same KoKo dosimeter nebulizer equipment from the dose determination, using the same five-breath procedure. The drug was administered while the subject lay supine and fully instrumented on the scan table, in the control room, to minimize the amount of time necessary between drug administration and MR imaging. FEV₁ determination was performed postbronchoconstriction and before MR scanning. Methacholine action peaks between 1 and 4 min after inhalation, plateaus for an average of 75 min, and then diminishes for an average of 57 min (3). The 20-min specific ventilation imaging (SVI) protocol took place 15–35 min after inhalation of the drug, during the plateau period in most subjects when smooth muscle activation would be in steady state. The SVI processing algorithm (described below) will report any non-steady-state behavior as a fitted temporal average of specific ventilation during the imaging period.

Imaging Sessions

Timeline.

MR images were acquired at baseline and following three separate pharmaceutically induced bronchoconstriction episodes. Both proton density and specific ventilation images were acquired during each condition. An experimental timeline is shown in Fig. 1.

A detailed description of SVI is presented in Sá et al. (30). SVI as implemented in this study employs new hardware to streamline the experimental protocol (4) and a slightly altered stimulus block design to boost sensitivity to low specific ventilation units (see below). An overview of the protocol is presented here.

Experimental setup.

Each subject lay supine on the scan table in the operator room and was fitted with a face mask (dead space 73–113 ml, depending on mask size; Hans Rudolph) attached to a three-dimensional-printed flow-bypass device (dead space 41 ml; Ref. 4). After bronchoconstriction and/or FEV₁ measurement, subjects were wheeled into the scan room and placed in the scanner. The input side of the flow-bypass system was attached to a tank of medical oxygen via 1/4-in. tubing and a three-way manual valve. When actuated, oxygen flowed through the flow-bypass face mask adapter at ~120 l/min. This rate was chosen to exceed maximal inspiratory flow rate and thus ensure that the subjects were breathing 100% oxygen while the valve was turned on. Actuation of the valve resulted in an abrupt switch from room air to oxygen, or vice versa, and occurred during expiration so that the concentration of inspired gas changed in a stepwise fashion between subsequent inspirations.

SVI air-oxygen stimulus.

The physical basis of the SVI measurement (see Ref. 30 for full description) is that T₁-weighted signal intensity in the lung is dependent on the partial pressure of oxygen in the tissue and blood, which is in equilibrium with the alveolar gas (which itself provides no appreciable signal). Therefore, by changing the inspired concentration of oxygen, we change the blood and tissue oxygenation and the associated T₁-weighted signal intensity. Importantly, the rate of signal equilibration following an instantaneous change in inspired oxygen concentration is related to the rate at which fresh gas is replacing resident gas in that voxel. This rate of replenishment, when it is expressed as the change in volume of a voxel during inspiration divided by its previous end-expiratory volume, is termed specific ventilation. Thus rapid switches between inhaled air and 100% oxygen (and vice versa) act as stimuli for which we measure a response, MR signal intensity over time, which allows us to sample specific ventilation spatially. Before this study, an SVI experiment was implemented with 20 breaths between switches to allow for maximum sensitivity to the most physiologically relevant range of SV values, 0.08–0.5, and 1 40-breath interval to give sensitivity to lower SV voxels (specific ventilation values between 0.01 and 0.1).

For the purposes of this experiment, the 40-breath block of oxygen was delivered earlier in the experiment than has been described previously (30), this time during breaths 21–60 instead of breaths 181–220. This choice was made to enhance sensitivity to lung regions with specific ventilation between 0.01 and 0.1 by equilibrating these slow-ventilated regions to a higher signal (and thus higher signal-to-noise-ratio) range early in the experiment. This alteration does not change the sensitivity to higher specific ventilation units, which are unaffected by the switch in block position.

SVI acquisition.

Successive multiplanar image sets were acquired every 5 s. Subjects were trained to breathe between image acquisitions using the scanner noises as audible cues, each breath returning to a “relaxed” lung volume (functional residual capacity) and pausing postexpiration for the following image acquisition. Data were collected in the right lung to avoid cardiac motion artifacts. Four adjacent 15-mm sagittal slices were selected so that the imaging range included as little of the hilar vessels medially or the chest wall laterally as practical. The location of the vertebral spinous process at the level of the right hilar point was recorded and used as a landmark for future scans to ensure the same area was being imaged each time. With the use of a 1.5-T Excite HDx TwinSpeed MRI system (General Electric Medical Systems, Milwaukee, WI), four two-dimensional images were consecutively acquired with a single-shot fast spin-echo after a global inversion pulse during each breath hold. Consequently, each imaging slice had a unique inversion recovery time of 1,100, 1,335, 1,570, and 1,805 ms, respectively (medial to lateral). An eight-element torso coil was used to provide greater sensitivity than the built-in body coil. Each image in the multiplanar set had a 40 × 40 cm field of view acquired at 128 × 128 voxel resolution. Images were reconstructed by the scanner onto a 256 × 256 matrix.

Image Processing

Custom-written MATLAB (MathWorks, Natick, MA) code was used for image processing. Data from each scanning session were processed separately and compared during the analysis phase. The processing steps for each day were as follows.

For each of the 4 imaging slices, a time series of 220 images, each image corresponding to a separate postexpiratory breath hold, was assembled. A generalized dual-bootstrap iterative closest point algorithm (42) was used to register all images in the time series to a reference image chosen manually to be as close as possible to functional residual capacity. This publicly available algorithm identifies features based on edges and corners in the images to be registered and iteratively performs affine transformations based on features within a region that grows with each iteration. Multiple seed points are used to create multiple transformations via this region-growing approach, and only those that pass a set of decision criteria are accepted to compute a final transformation (42). Images in which the subject’s postexpiratory lung volume required >10% correction (4 ± 2 images out of 220) to match the reference lung volume were removed from the time series and treated as missing data. This limit was chosen because we (1a) have shown that lung deformations within 10% can be accurately registered using an affine transformation.

A 220-image (breath) time course of signal intensity was constructed for every voxel within a manually drawn lung region of interest. The time course for each voxel was then correlated with 50 modeled theoretical time courses for specific ventilation values equally spaced in log₁₀ from 0.01 to 10: these bin values were chosen to match the 50-compartment model employed by multiple-breath washout studies (29). The modeled time course with the maximum correlation coefficient was determined to be the specific ventilation of the voxel. The SV thus computed was only retained if the correlation was significant (P < 0.05) and the null hypothesis of no correlation rejected. Voxels failing to pass this test (<1%) were treated as missing data.

For each of the 4 slices, a spatial map of specific ventilation and histogram of SV distribution was created. For further spatial analysis (intercondition/day comparison and identification of bronchoconstricted regions, described below), SV maps were smoothed using a geometric algorithm with a kernel size of 7 voxels (~1 cm²).

Data Analysis

Specific ventilation heterogeneity.

Maps of specific ventilation were used to construct distributions of SV for each subject at baseline and during the three episodes of bronchoprovocation. A log-Gaussian function was fit to each of these histograms, and the full-width half-maximum of the fitted distribution was reported as a measure of SV heterogeneity (29).

Identification of bronchoconstricted regions.

To measure the change in specific ventilation maps between baseline and during methacholine challenge, we first normalized the maps so that the total imaged specific ventilation, the sum of all voxel specific ventilation values, of each multiplanar set was equal between conditions. Total imaged specific ventilation is a function of tidal volume, which can potentially change between conditions. An increase or decrease in tidal volume induces a global increase or decrease (respectively) in specific ventilation across the lung (27). Therefore, a decrease in tidal volume following methacholine-induced bronchoconstriction could cause “false-positive” indications of bronchoconstriction that are due to global but not regional changes in specific ventilation.

For the purposes of this study, we defined a bronchoconstriction event as a postnormalization 50% decrease in specific ventilation from baseline to a value below the center of the baseline distribution (e.g., 50% decrease to SV < 0.18 for subject 2; see Fig. 2). This imposed two conditions: 1) that specific ventilation be significantly decreased and 2) that the postmethacholine specific ventilation value of a constricted lung region implies that it is ventilating less than average. Importantly, these criteria effectively exclude two types of lung regions: 1) regions with perpetually low specific ventilation that are not necessarily responding to methacholine and 2) regions with high specific ventilation at baseline that are affected by the drug but not to the point that they could potentially generate regions with a low ventilation-to-perfusion ratio.

Fig. 2.

Left: multiplanar specific ventilation (SV) maps for 1 subject (S2) from all 4 imaging sessions. Each row of images represents 1 day of imaging, and each column represents a sagittal slice of the subject’s right lung. Rows are organized from medial on the left to lateral on the right. Right: SV histograms for subject S2 from baseline (blue) and 3 constricted (red, yellow, and purple) imaging sessions. SV values from all 4 planar maps were combined into 1 histogram for each session.

After normalization, the multiplanar image sets were spatially correlated. Because imaged thoracic cage volume varied between baseline and bronchoconstriction, on average by 18 ± 10%, beyond the limit within which we can confidently register images accurately, 10% (1a), we did not perform image registration in the sense that we did within a 220-image temporal set (see above). Instead, we divided each multiplanar set into 12 (anterior-posterior) × 12 (apical-basal) × 4 (medial-lateral, defined by imaging plane) volume elements, the extent of which were redefined for each condition. Arithmetic mean specific ventilation of the voxels within each of these volume elements was measured, and comparisons were made between volume elements from different conditions with corresponding 12 × 12 × 4 coordinates. Elements in which the specific ventilation fell to less than half of the baseline value and to less than the center of the baseline distribution were identified as constricted and used to create binary constriction masks (Fig. 3). The volume of each element (number of voxels inside the element) was taken into account, and all further processing was weighted by the volume of each element. For example, a constricted element in the middle of the lung containing 80 voxels would contribute more to the whole lung constriction fraction than would an element on the edge of the lung that contains only 30 voxels.

Fig. 3.

Processing flow for subject S2’s midclavicular (plane 2) specific ventilation map, shown for baseline and each of the 3 constrictions. After normalization, maps were downsampled to a 12 × 12 grid of volume elements and then projected onto the baseline image frame for comparison. All further data processing steps were performed on these resampled maps. Volume elements with fractional specific ventilation (column 4) <0.5 were identified as constricted (column 5) if the end value of the element is less than the center of the baseline distribution (center of the blue histogram in Fig. 2, right). This process was repeated for all 4 planar maps.

Spatial persistence of constriction between episodes of bronchoconstriction.

Binary constriction masks were compared between conditions to determine 1) the percentage of voxels that constrict in a subsequent methacholine challenge, given a prior constriction, and 2) the predictive power of a prior constriction for a future constriction. These metrics were stratified for time between constrictions, 1 wk vs. ~3 mo, and for number of constrictions: one previous constriction predicting a second vs. two previous constrictions predicting a third. Odds ratios (exact usage defined in Fig. 4) were used as measures of predictive power. Volume elements that constricted following all three methacholine administrations were classified as “persistent.”

Fig. 4.

A: pairs for interday comparison. Comparisons were always made going forward in time, i.e., the percentage of day 1 constrictors (yellow regions in the 1st column) that constricted again 1 wk later (became orange in 2nd column) or the odds ratio for a 1-wk-later constrictor to constrict ~3 mo later. B: odds ratios for pairs were defined as the odds of a previous constrictor constricting again divided by the odds of a previous nonconstrictor constricting for the 1st time.

Regional trends in bronchoconstriction across subjects.

On each sagittal lung image, regions were defined by thirds in the apical-basal direction and thirds in the anterior-posterior direction. The 4 sagittal slices were grouped 2 × 2 to form a medial group and a lateral group. Thus 18 lung regions were defined, each with a unique spatial designation, e.g., apical-anterior-medial. In practice, this was done by resampling the 12 × 12 × 4 grid of volume elements (from above) to a 3 × 3 × 2 grid of regions. Regional tendency to bronchoconstrict following methacholine was characterized in our study by the average volume fraction of constriction in a region over the 3 study sessions. Likewise, tendency toward persistent constriction was characterized by the volume fraction of persistent (constricted in response to all 3 methacholine administrations) voxels. Likelihood of a constricted region to persist over time, or persistence of constriction, was quantified as the ratio of the 2 fractions above, persistent to average constricted.

RESULTS

Demographic Data, Baseline Pulmonary Function Tests, and Methacholine Dose Determination

Subject demographic data, seated FEV₁ measurements, and methacholine challenge results are shown in Table 1. Average seated FEV₁ between subjects was 101 ± 6% of predicted (Table 1). Subsequent FEV₁ measurements were all made supine. Average supine FEV₁ at baseline (premethacholine) was 92 ± 5% of predicted (Table 2), which suggested that, as expected, supine posture significantly reduced FEV₁ (P = 0.002, paired t-test, 1-tailed).

Table 2.

Lung-wide metrics of constriction for each study day

Subject	Baseline, Supine FEV1, l [%Predicted]	Day 1 Constricted FEV1, l [%Change]	One-Week-Later Constricted FEV1, l [%Change]	Approximately-3 mo-Later Constricted FEV1, l [%Change]	Average Constricted FEV1, l [%Change]
S1	3.55 [93]	2.86 [−19]	2.91 [−18]	2.28 [−36]	2.68 [−25]
S2	2.91 [96]	2.53 [−13]	2.68 [−8]	2.51 [−14]	2.57 [−12]
S3	4.99 [99]	3.92 [−21]	3.42 [−32]	3.79 [−24]	3.71 [−26]
S4	3.05 [86]	2.34 [−22]	2.28 [−24]	2.62 [−13]	2.41 [−21]
S5	3.70 [88]	3.14 [−15]	2.90 [−22]	2.65 [−28]	2.90 [−22]
S6	3.11 [92]	2.38 [−24]	2.49 [−20]	2.03 [−35]	2.30 [−26]
Mean (SD)	3.56 (0.79) [92 (5)]	2.86 (0.60) [−20 (4)]	2.78 (0.40) [−21 (8)]	2.65 (0.61) [−26 (10)]	2.76 (0.51) [−22 (6)]

	Baseline Imaged Thoracic Cage Volume, cm3	Day 1 Constricted Imaged Thoracic Cage Volume, cm3 [%Change]	One-Week-Later Constricted Imaged Thoracic Cage Volume, cm3 [%Change]	Approximately-3 mo-Later Constricted Imaged Thoracic Cage Volume, cm3 [%Change]	Average Constricted Imaged Thoracic Cage Volume, cm3 [%Change]
S1	851	927 [9]	821 [−4]	905 [6]	884 [4]
S2	560	732 [31]	632 [13]	627 [12]	664 [19]
S3	930	1,206 [30]	1,221 [31]	1,198 [29]	1,208 [30]
S4	559	785 [40]	723 [29]	683 [22]	730 [31]
S5	509	518 [2]	544 [7]	649 [28]	570 [12]
S6	497	561 [13]	521 [5]	599 [21]	560 [13]
Mean (SD)	650 (190)	790 (250) [21 (15)]	740 (260) [26 (11)]	780 (230) [20 (9)]	770 (250) [18 (10)]

	Day 1 Volume Fraction of Constricted Elements	One-Week-Later Volume Fraction of Constricted Elements	Approximately-3 mo-Later Volume Fraction of Constricted Elements	Average Volume Fraction of Constricted Elements
S1	0.31	0.27	0.30	0.29
S2	0.25	0.14	0.26	0.22
S3	0.37	0.55	0.44	0.45
S4	0.33	0.27	0.18	0.26
S5	0.18	0.10	0.38	0.22
S6	0.26	0.13	0.25	0.21
Mean (SD)	0.28 (0.07)	0.24 (0.17)	0.30 (0.09)	0.28 (0.09)

	Baseline SV Distribution Width	Day 1 SV Constricted FWHM [Change]	One-Week-Later Constricted SV FWHM [Change]	Approximately-3 mo-Later Constricted SV FWHM [Change]	Average Constricted SV FWHM [Change]
S1	0.43	0.92 [0.49]	0.95 [0.52]	0.75 [0.32]	0.87 [0.44]
S2	0.32	1.03 [0.71]	0.89 [0.58]	1.20 [0.32]	1.04 [0.72]
S3	0.43	0.91 [0.48]	1.81 [1.37]	1.02 [0.59]	1.25 [0.81]
S4	0.57	0.46 [−0.11]	1.19 [0.61]	0.4 [−0.13]	0.70 [0.12]
S5	0.61	0.55 [−0.06]	0.39 [−0.22]	1.15 [0.54]	0.69 [0.09]
S6	0.32	1.00 [0.68]	0.50 [0.18]	0.87 [0.54]	0.79 [0.47]
Mean (SD)	0.45 (0.12)	0.81 (0.25) [0.37 (0.36)]	0.95 (0.51) [0.51 (0.53)]	0.90 (0.28) [0.46 (0.34)]	0.89 (0.22) [0.44 (0.30)]

Data are presented as means (SD). Response to methacholine, quantified change in FEV₁, thoracic cage volume, or FWHM is expressed as percentage or absolute change. All measurements in this table were made in the supine posture. Supine forced expiratory volume in 1 s (FEV₁) was measured in all subjects before baseline scanning on day 1. FEV₁ was measured supine after methacholine administration on all 3 days to quantify response to the drug. Imaged thoracic cage volume is the size of the lung in MR images acquired at functional residual capacity. Volume fraction of constricted elements is the fraction of lung voxels that pass our 2 criteria for constriction [>50% decrease in specific ventilation (SV) from baseline to a value below the center of the baseline distribution]. Full-width half-maximum (FWHM) is the width of a log-Gaussian function fit to the specific ventilation distribution (see Fig. 2) acquired during each condition.

Response to Methacholine Bronchoconstriction: Pulmonary Function Tests and Whole Lung Imaging Measures

Metrics of assessing bronchoconstriction are shown in Table 2. FEV₁, imaged thoracic cage value, and the width of the specific ventilation distribution all changed significantly between baseline and bronchoconstriction, both on average and for each individual episode (P < 0.05, paired t-tests). There were no statistically significant differences (P > 0.05, paired t-tests) between bronchoconstriction episodes in any of the four metrics.

Spatial Persistence of Constriction between Bronchoconstriction Episodes

On average, 53% CI (46%, 59%) of volume elements that constricted on a given day constricted again on a following day (Fig. 5). This finding was independent of the time between constrictions: 46% CI (28%, 64%) repeated 1 wk later and 56% CI (51%, 61%) repeated 3 mo later. Likewise, the odds ratio (Fig. 5) of a given constricted volume element constricting again during a later challenge was 7.7 CI (5.2, 10.2) and did not vary based on the time between challenges; odds ratios were 7.7 CI (2.9, 12.4) for 1 wk later and 7.7, CI (4.6, 10.8) for 3 mo later. The amount of constriction varied considerably between subjects (Fig. 6, Table 2).

Fig. 5.

Left: volume fraction of constricted elements that repeatedly constrict during a following methacholine challenge. Right: odds ratio for the likelihood of a constricted voxel to constrict again during a following challenge. In both cases, subject values are shown as dots and connected by gray lines, and group means are shown as solid black lines. Subject values may be the average of multiple measurements, i.e., “Any time” represents the average of all 3 pairs (Fig. 4) for a subject and ~3 mo later represents the average of the 2 3-mo pairs (Fig. 4). *P = 0.038, paired t-test, 1-tailed.

Fig. 6.

Constriction frequency maps for all subjects. Imaging planes are arranged from medial on left to lateral on right, and subjects (S) are arranged vertically. The color of a lung region indicates the number of times the region constricted following methacholine challenge. Red volume elements constricted 3/3 times, orange constricted 2/3, yellow constricted 1/3, and black constricted 0/3. Red units, those that constricted during all 3 challenges, are classified in this study as “persistent constrictors.”

Of volume elements that constricted during the first two methacholine administrations, 68% CI (57%, 79%) constricted for a third time 90 days later (Fig. 5). This was significantly higher (P = 0.01, paired t-test, 1-tailed) than the 56% of volume elements in the 90-day pairs that repeated. The odds ratio of twice constrictors repeating a third time was 9.8 CI (4.1, 15.6). This odds ratio was higher than for the 90-day pairs, but this trend did not reach statistical significance (P = 0.22, paired t-test, 1-tailed).

Regional Trends in Bronchoconstriction across Subjects

Volume elements in the posterior lung were more likely to constrict than those in the anterior lung (P = 0.005, paired t-test, 1-tailed), and persistently constricting volume elements were more likely to be found in the posterior lung than the anterior lung (P = 0.03, paired t-test, 1-tailed). Although there was no statistically significant apical-basal gradient in overall constriction fraction (P = 0.35, paired t-test, 1-tailed), volume elements in the basal lung that do constrict were more likely than apical units to persist throughout all 3 challenges (P = 0.01, paired t-test, 1-tailed). There was no statistically significant medial-lateral bias toward constricted volume, persistent volume, or persistence of constriction in the healthy lungs we tested (P = 0.63, P = 0.79, and P = 0.52, respectively, paired t-tests, 2-tailed).

DISCUSSION

In healthy subjects, previous bronchoconstriction of a lung element following methacholine challenge is a strong predictor for future bronchoconstriction of that same element following methacholine challenge weeks or months later. The predictive value of a previous constriction does not diminish over 3 mo. There is a higher tendency toward bronchoconstriction in the posterior lung when supine and a tendency for basal lung units that constrict once to constrict persistently during each of three challenges given over 3 mo. These results are in agreement with previously published data from hyperpolarized He imaging (13, 14, 17, 41).

Airway smooth muscle activation following methacholine inhalation is counteracted, at the level of the airway, by tethering forces from surrounding lung tissue and by transmural pressure differences at the airway wall (41). Venegas et al. (37), with an integrative model of these forces acting on 12 generations of airways, showed that interdependence between serial and parallel airways gives rise to a self-organized patchy distribution of bronchoconstriction, even in a symmetric lung model (although in a symmetric lung model, the locations of the patches are random). Later, Leary et al. (21) showed that branching asymmetry in the model led to “local triggers” for this patchiness and found that when the degree of branching asymmetry of the simulated lung matched that of an anatomically realistic model (34), the spatial pattern of bronchoconstriction was highly persistent, with an average local variation between successive bronchoconstriction episodes of <3% (21).

This persistence due to branching asymmetry may help explain the spatial persistence of bronchoconstriction found in our study but likely is not the sole determining factor. Differences in regional lung inflation, or, rather, the differences in transmural pressure that are associated with differences in lung inflation, also cause lung regions to constrict persistently. Harris et al. (13) provided evidence for this effect in a PET study, which showed that ventilation defects were more likely to form in regions with a lower degree of lung inflation premethacholine. This principle is also supported by our data, which show that gravitationally dependent lung units are more likely to constrict than their nondependent counterparts. Since the dependent lung is compressed by the weight of the lung above it (16, 38), it is relatively less inflated. Thus the parenchymal forces holding open the airways and preventing airway collapse are less potent.

Since airway tree structure (absent remodeling from disease) and regional lung inflation (absent changes in posture or breathing) are not expected to change in healthy individuals between scanning sessions, we expected the forces driving regional tendency to bronchoconstrict following methacholine to be roughly the same and thus the pattern of constriction to be mostly persistent. This was borne out in the data.

However, there are potential factors other than innate airway asymmetry or regional inflation that could cause certain areas of the lung to be predisposed to constriction in our study, among which is that a reproducibly uneven pattern of methacholine deposition in the lung could lead to more drug response in regions receiving a higher dose. Indeed, the fact that posterior lung regions were more likely to constrict could be due to preferential deposition of methacholine in the posterior lung. It is well-known that the gravitationally dependent (in this case, posterior) lung ventilates more than the nondependent (in this case, anterior) lung (22, 38). If methacholine simply followed the pattern of ventilation, one would expect more deposition in the posterior lung. Since the action of methacholine is mostly topical or local, this could have led to higher rates of constriction and persistence of constriction in the posterior lung.

Limitations

Our experiment is subject to certain factors that limit the scope of our conclusions. The most significant of these is that the supine posture in which all elements of the experiment were performed limits the interpretation of our findings. There are (at least) three factors that could lead to a persistent pattern of postmethacholine bronchoconstriction in healthy individuals: 1) airway asymmetry acting as local triggers for constriction, 2) lower inflation in the gravitationally dependent lung (and in other regions) predisposing these regions, and 3) preferential deposition of methacholine and increased local action of the drug due to gravity and/or airflow at the time of administration. The repeatable pattern of constriction that we see in our subjects is likely influenced by all of these factors, but the degree to which each contributes is difficult to quantify. Since airway asymmetry is insensitive to posture and the other two factors are not, experiments in other postures may help give weight to the three factors.

A second limitation is that the four-slice imaging volume encompassed only 80% of one (the right) lung. It has been shown that specific ventilation imaging in a single, sagittal lung slice (~8% of total lung volume) accurately represents the heterogeneity of specific ventilation in a healthy lung (29). This is because 1) in the healthy lung (without methacholine), no spatially localized heterogeneity is expected and 2) gravity is a major source of specific ventilation heterogeneity in a healthy subject and a sagittal slice captures a cross-section of this gravitational gradient. However, in this study, in which we imposed an additional layer of SV heterogeneity, it is possible that the extent of that added effect could not have been captured by four imaging slices in one lung. However, the lack of a medial-lateral gradient in either constricted or persistent lung fraction suggests that exclusion of the most medial and lateral lung is unlikely to have biased our results. Any systematic difference between the behavior of the right lung and left lung (which was not imaged) postmethacholine was not captured by our experimental design.

Comparison with Other Studies

Although this was the first imaging study to probe the spatial reproducibility of methacholine bronchoconstriction in healthy individuals, other studies have done so in mild, moderate, and severe asthmatics. Direct comparison between those studies and ours is challenging, as the constriction metric used in previous studies, manual or semiautomated identification hypointense regions on hyperpolarized He scans, is not directly analogous to the constriction metric we have defined here, a 50% decrease in SV from baseline to a below average SV. However, some comparison is possible. Our finding, that 53% CI (46%, 59%) of voxels that constricted in one methacholine challenge did so in a second (Fig. 5), fits within a range previously reported by de Lange et al. (5). Those data showed that the rate of spatial recurrence of ventilation defects in asthmatics is between 38 and 67%, depending on the time between examinations. Unlike this previous result, the rate of recurrence in our study did not depend on the time between challenges. It is probable that there is a temporal disease process at work in the de Lange et al. (5) asthmatic study population that causes the spatial persistence of ventilation defects to decrease over time and that this process did not occur in our healthy individuals given methacholine.

Other studies on the recurrence of impaired ventilation during bronchoconstriction, using a variety of approaches and metrics, have reported a wide range of results: a separate study by de Lange et al. (6) showed that 69% of postmethacholine ventilation defects recur in the same place during a 2nd challenge; a study by Niles et al. (24) showed that 29% of postexercise defects recur in the same place during a 2nd exercise-induced exacerbation; and a study by Svenningsen et al. (33) showed that the ratio of intermittent to persistent ventilation defects in asthmatics postexercise was 20:1 (persistent defined as constricting in 3 out of 3 exercise challenges 5 ± 2 days apart). The discrepancy between our study and these is partially explainable by the different methods for eliciting, measuring, and quantifying bronchoconstriction and even more fundamentally by the wide spread in the results of the existing studies. The latter, which suggests a fundamental uncertainty in our understanding of the process and in the evidence to elucidate it, highlights the need for more studies of this type.

The spatial gradients of constriction and persistence (Fig. 7) that we report here directly corroborate previous reports by Svenningsen et al. (33) and Harris et al. (13, 14), both of which involved asthmatic (as opposed to healthy) subjects. In the study by Svenningsen et al. (33), in which asthmatics were imaged with hyperpolarized helium postexercise on three separate occasions, it was shown that both intermittent ventilation defect percentage (somewhat equivalent to our overall constricted fraction) and persistent ventilation defect percentage (somewhat equivalent to our persistently constricted fraction) increase from the anterior to the posterior lung and from the apical to the basal lung. This finding is reflected in our data. Importantly, both the methods of eliciting bronchoconstriction, exercise vs. supine methacholine administration, and the subject populations, asthmatics vs. healthy young volunteers, are different in the two studies, despite the similar results. This suggests that the pattern of methacholine distribution, and perhaps even underlying presence or lack of asthma, both play relatively minor roles in the spatial gradients of bronchoconstriction found in the studies. More likely, it is regional differences in airway structure or a predisposition due to differences in regional lung inflation that serve to drive the effect.

Fig. 7.

Top row (red): the average fraction of voxels that constrict during each of the 3 challenges (F̄) is shown as a function of anterior-posterior, apical-basal, and medial-lateral position. Middle row (green): the fraction of voxels that persistently constricted during all 3 methacholine challenges (F_p) is shown as a function of position. Bottom row (purple): the fraction of persistent voxels divided by the average fraction of voxels that constrict (F_p/F̄) is shown as a function of position. Lines on box plots represent the median and 1st/3rd quartiles of the data, and dots represent subject values. Note that although the basal volume elements were no more or less likely to constrict than their apical counterparts, P = 0.35 by paired, 1-tailed t-test, those that do were more likely to persist through all 3 trials, P = 0.01 by paired, 1-tailed t-test. *Difference in regions that was significant with P < 0.05 by paired, 1-tailed t-test.

The 2009 (14) and 2012 (13) studies by Harris et al. (also corroborated by our data) lend support to the second of these possibilities, that the spatial pattern of bronchoconstriction is influenced by the systematically decreased inflation of the dependent lung compared with the nondependent lung. Harris et al. (14) showed that, whether the subject was prone or supine during imaging, bronchoconstriction following prone methacholine administration preferentially occurred in the gravitationally dependent lung. The group (13) went on to show that this gravitational dependence of constriction following methacholine existed both in mild asthmatics and healthy nonasthmatics, although the gravitational dependence was stronger in healthy nonasthmatics. This supports the conjecture above, that healthy individuals and mild asthmatics have spatial tendencies toward bronchoconstriction that are largely based on the same physiological phenomena. The caveat to this claim is that, in the mild asthmatics who were studied, the amount of airway remodeling from chronic inflammation is likely low and thus a minor factor governing constriction (13).

Potential Applications

With the emergence of new surgical and pharmaceutical treatments, quantification and localization of pathological lung regions is becoming important for informing treatment and evaluating efficacy. Surgical techniques that target specific regions of the lung, such as bronchial thermoplasty (8, 35), could potentially benefit from precise localization of airflow-obstructed regions before treatment, especially if these regions are shown to repeat between exacerbations. Furthermore, quantification of airflow deficits, both temporally repeatable and nonrepeatable, holds promise as a means of assessing disease progression and treatment response. Indeed, hyperpolarized imaging-derived ventilation defect volume has been shown to be more sensitive in some cases than standard pulmonary function tests for depicting short-term changes in chronic obstructive pulmonary disease (20). This work fits into the ongoing effort to transform functional MR imaging into a clinical tool for guiding therapy and for disease/treatment monitoring. The style of oxygen-enhanced imaging used in this study, specific ventilation imaging (SVI), has been validated (29), used to measure the gravitational gradient of ventilation in the lung (30), and recently integrated with blood flow measurements to map ventilation-to-perfusion ratio in the healthy lung (31). This study represented an extension of the technique in two regards: 1) it was the first application of the technique to the study of pathological lung function, and 2) it was the first study to implement a multislice form of SVI that extended the imaging volume from ~20 to ~80% of the right lung.

Conclusion

Specific ventilation imaging has shown that, in the healthy lung, an area of lung that undergoes a dramatic reduction in SV postmethacholine is much more likely to do so again in a successive challenge than is an area that resists the first challenge. This remains true whether considering challenges 1 wk apart or ~3 mo apart. Furthermore, it has shown that tissue in the posterior, in this case gravitationally dependent, lung is more likely to be susceptible to constriction and that tissue in the basal lung, although not necessarily more likely to constrict on average, is more likely to constrict persistently if it constricts at least once. Further studies involving changes in posture during drug administration and imaging are needed to determine whether regional vulnerability to constriction is driven by anatomy, drug delivery, or local transmural pressure due to regional lung inflation.

The overall repeatability and the spatial gradients of constriction in our study match previous hyperpolarized helium MR imaging work done in patients with asthma (5, 6, 13, 14, 33) despite the fact that our subjects have no history of asthma or any other pulmonary disease. This is evidence that methacholine constriction of healthy subjects may be a representative model for asthmatic bronchoconstriction, at least for the purposes of studying the spatial patterns of bronchoconstriction.

GRANTS

This work was supported by the National Heart, Lung, and Blood Institute (NHLBI) Grant R01-HL-119263. E. T. Geier’s work was supported by NHLBI Grant F30-HL-127980-02.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

E.T.G., G.K.P., and R.C.S. conceived and designed research; E.T.G., I.N., and R.C.S. performed experiments; E.T.G., I.N., and R.C.S. analyzed data; E.T.G., R.J.T., G.K.P., and R.C.S. interpreted results of experiments; E.T.G. and I.N. prepared figures; E.T.G. drafted manuscript; E.T.G., R.J.T., G.K.P., and R.C.S. edited and revised manuscript; E.T.G., I.N., R.J.T., G.K.P., and R.C.S. approved final version of manuscript.