Noninvasive Monitoring of the Response of Human Lungs to Low‐Dose Lipopolysaccharide Inhalation Challenge Using MRI: A Feasibility Study

Development of antiinflammatory drugs for lung diseases demands novel methods for noninvasive assessment of inflammatory processes in the lung.

A IRWAY INFLAMMATION is believed to play a vital role in the pathogenesis of many lung diseases. Specifically, progression of chronic obstructive pulmonary disease is associated with neutrophilic airway inflammation. 1 Segmental challenge of the lung using lipopolysaccharide (LPS) causes a local neutrophilic inflammatory response, thereby providing a disease model for development of targeted antiinflammatory drugs. 2 Cell counts from bronchoalveolar lavage then enable a quantification of the inflammatory response for treatment monitoring. 3 However, these procedures entail repeated bronchoscopies and thus cause additional burden to the subjects.
An alternative approach to segmental challenge is low-dose LPS inhalation, causing a mild inflammatory response in the whole lung. Collection of induced sputum is generally well-tolerated and inflammatory cell counts then similarly enable quantitative assessment of inflammation for testing the efficacy of investigational new drugs. 4 However, the origin of sputum from the airways is less controlled and the potential of sputum analysis for accurate disease monitoring at the alveolar level is thus limited. 5 Further, repeated sputum induction has been shown to alter cell counts at later times, which limits its use for repeated assessments. 6 Magnetic resonance imaging (MRI) provides spatially resolved diagnostic information and can be repeated without detrimental effects. It has been shown that 1 H turbo inversion recovery magnitude (TIRM) MRI and T 1 mapping MRI are sensitive to airway inflammation after segmental allergen challenge in asthmatics. 7,8 However, TIRM MRI only provides semiquantitative information on local fluid content, and 1 H T 1 is not very specific for a certain pathology. 9 MRI and spectroscopy of inhaled hyperpolarized 129 Xe enable noninvasive assessment of gas uptake to lung parenchyma and blood and yield specific, quantitative information on lung function at the alveolar level. 10 Models of 129 Xe uptake dynamics have been developed, providing physiologic parameters like septal wall thickness, surface-to-volume ratio, and capillary transit time of blood. 11 The specificity of parameters derived from 129 Xe dissolved-phase MRI and spectroscopy to lung structure and function at the alveolar level suggests that changes in these parameters could be quantitatively associated with the degree of inflammation in the alveolar region and could potentially be used as surrogate markers, for example, for cell counts.
The purpose of this study was to test the feasibility of hyperpolarized 129 Xe, 1 H T 1 mapping and dynamic contrastenhanced (DCE) MRI for monitoring the response of human lungs to inhaled LPS challenge. The potential of MRI-derived biomarkers for noninvasive assessment of inflammation was tested by comparison to inflammatory cell counts from sputum analysis.

Study Design
This prospective study (ClinicalTrials.gov Identifier NCT03044327) was approved by the Institutional Review Board and all subjects gave written informed consent. The study was conducted at a single research center between February and November 2017 and included inhaled LPS challenge of the lung. MRI was performed before and 6 hours after inhaled LPS challenge. The timepoint for MRI after LPS challenge was chosen to coincide with the time of maximal inflammatory response in terms of cell counts, although there is still some controversy about this point. 12 Subjects were included if they had normal lung function (forced expiratory volume in 1 sec [FEV 1 ] >80% of predicted normal, FEV 1 /forced vital capacity [FVC] >70%) and were nonsmokers with a smoking history of less than one pack-year. Exclusion criteria were clinically relevant history of allergies, elevated level of immunoglobulin E, lower respiratory tract infections within 4 weeks of the first MRI, and general MRI contraindications.

Study Population
For this study, 13 healthy volunteers were recruited and recruitment was stopped after 10 subjects successfully completed the study protocol. One volunteer had to be excluded due to an overreaction to the sedative used during bronchoscopy for segmental LPS challenge in a different part of the study protocol not presented in this work. Another volunteer was excluded due to respiratory tract infection unrelated to the study protocol. A third volunteer was excluded after screening due to an elevated level of immunoglobulin E.

LPS CHALLENGE AND CELL COUNTS.
For inhaled LPS challenge, 2 μg LPS was dissolved in sterile water and administered using a breath-controlled nebulizer. 12 Inspiration was adjusted to a low flow rate of 150 mL/s, which is expected to improve deposition of LPS to the alveoli. 13 Sputum was induced by inhalation of nebulized saline and immediately processed.
Transmitter calibration was performed using a dedicated pulse sequence 14 in a separate breath-hold. Subjects inhaled 1 L of a gas mixture containing 0.3 L hyperpolarized xenon with natural isotope ratio and 0.7 L nitrogen for this purpose.
A 3D-radial multiecho sequence and a 3-point Dixon algorithm were used for interleaved imaging of the hyperpolarized 129 Xe dissolved phase, ie, 129 Xe in red blood cells (RBC) and tissue/plasma (TP), and gas phase (GP). 15 Sequence parameters were repetition time 19 msec, echo times 0.74/2.36/3.98 msec for dissolved phase, 0.74/2.36 msec for gas phase, flip angle 23 for dissolved phase / 0.4 for gas phase, reconstructed resolution 7.6 × 7.6 × 17.6 mm, whole-lung coverage, scan time~10 sec. Spectroscopy of the dissolved phase (~18 Hz resolution), ventilation imaging (stack-of-stars gradient-echo sequence, 2.4 × 2.4 × 17.6 mm resolution, 14 coronal slices, scan time~6 sec) and a fast 1 H scan (gradient echo sequence, reconstructed resolution 2.0 × 2.0 × 5.0 mm, scan time~3 sec) were also performed during this second breath-hold. For these acquisitions, subjects inhaled a 129 Xe gas mixture of volume 1/3 FVC, starting from residual volume, containing 1 L hyperpolarized xenon with isotopically enriched 129 Xe fraction (Linde, München, Germany) and air similarly as in previous studies. 15 129 Xe and air were kept in two different Tedlar bags of 1 L capacity (Jensen Inert Products, Coral Springs, FL) for this purpose and mixed only during inhalation by the subject.
Dynamic gas-uptake measurements of 129 Xe were made using a chemical shift saturation recovery (CSSR) spectroscopy sequence. 16 Spectra were acquired in a third breath-hold after inhalation of a gas mixture containing 0.5 L hyperpolarized 129 Xe and 0.5 L nitrogen from a 1 L Tedlar bag and subsequently room air to achieve full lung inflation. The spectroscopic bandwidth was AE16.7 kHz at 32.6 Hz resolution; sampled delay times ranged from 3-700 msec, resulting in a measurement time of~9 sec. Saturation of the dissolved phase was performed using a 2.4 msec long rectangular 90 pulse at 198 ppm. The residual 129 Xe magnetization was used for ventilation imaging in the same breath-hold as before. 1 H T 1 maps were obtained in typically 8-9 slices covering the lung during breath-hold at end-tidal inspiration with the central slice at the tracheal bifurcation. An inversion recovery fast low angle shot sequence 9 was used for imaging, repetition/echo time 3/0.66 msec, flip angle 8 , resolution 3.9 × 3.9 mm, slice thickness 15 mm, no gaps, 32 inversion times (103-6055 msec). DCE MRI was performed after intravenous injection of 0.033 mmol/kg bodyweight gadopentetate using a time-resolved angiography with stochastic trajectories sequence 17 with repetition/ echo time 2.37/0.8 msec, voxel size 2.0 × 2.0 × 5.0 mm, temporal resolution~1.3 sec, 30 frames.

Data Analysis
The MRI reader was blinded to the results of induced sputum analysis. For computation of whole-lung dissolved-phase ratios in 129 Xe dissolved-phase MRI, an automatically determined mask based on signal-to-noise ratio (SNR) thresholding was applied to the data. This mask excluded voxels with SNR less than 5 in the first echo of either gas or dissolved phase. Quantification of ventilation defect percentage was performed using a mask of the thoracic cavity obtained by application of a region-growing algorithm in the 1 H gradient echo images and signal binning as described previously. 18 The masks for 1 H T 1 mapping and DCE MRI were similarly determined by application of a region-growing algorithm within the thoracic cavity and manual refinement.
DCE data were analyzed using model-free deconvolution 19,20 implemented in a self-developed MatLab script (MathWorks, Natick, MA), yielding maps for the parameters: blood volume, blood flow, and mean transit time. The arterial input function was determined by drawing a region of interest in the pulmonary artery. Large vessels were removed using cross-correlation analysis. 21 For 129 Xe CSSR measurements, Patz et al 11 developed a model function of gas uptake to the dissolved phase as a function of CSSR delay time. This function with the free parameters septal wall thickness, surface-to-volume ratio, and capillary transit time was fitted to the uptake curves of the TP peak area to obtain whole-lung estimates of these parameters. Lineshape analysis of the highresolution 129 Xe dissolved-phase spectra was performed by fitting two complex Lorentzians in the frequency domain, yielding chemical shifts and T 2 * relaxation times.

Statistical Analysis
Statistical tests were performed using MatLab R2014b (MathWorks, Ismaning, Germany). For comparison of data after challenge with baseline, the Wilcoxon signed-rank test was used (α = 0.05 twosided). Correlations of results from MRI with results from sputum analysis were assessed using Pearson's correlation coefficient and significance of the correlation was assessed by a permutation test. Since this was an exploratory pilot study, no sample size justification is given.

Results
Subject demographics and baseline clinical characteristics are shown in Table 1. Due to erroneous calibration of the MRI transmit system, the dissolved-phase imaging and highresolution spectroscopy data after LPS inhalation in one subject contained only noise. The same subject did not produce sufficient amounts of sputum for assessment of cell counts.
Exemplary MR images before and after LPS challenge are shown in Fig. 1 and results from all MR measurements are summarized in Supporting Table 1. No clear changes of gas distribution were observed in 129 Xe ventilation imaging and the ventilation defect percentage was not changed significantly, P = 0.16. Both FEV 1 and FVC as percent of predicted value were significantly reduced after inhaled LPS challenge ( Table 2). Cell counts from induced sputum are summarized in Supporting Table 2. Percentages of macrophages, neutrophils, and monocytes in sputum were significantly different compared to baseline, P = 0.002, while this was not the case for the percentage of lymphocytes.
There was a marked increase in neutrophil percentage in sputum after LPS inhalation (Fig. 2a). A significant reduction of the group median of whole-lung RBC-TP was observed in 129 Xe dissolved-phase imaging after inhaled LPS challenge, 0.31, compared to baseline, 0.40, P = 0.004 (Fig. 2b). There also was a significant reduction in whole-lung RBC-GP in dissolved-phase imaging from 0.46 to 0.42, P = 0.020, and a strong trend for elevated TP-GP, P = 0.074. Figure 3 summarizes the results of CSSR spectroscopy before and after inhaled LPS challenge. The capillary transit time was significantly increased after inhalation of LPS, 2.48 sec, compared to baseline, 1.96 sec, P = 0.020. Exemplary 129 Xe dissolved-phase spectra for the whole lung are depicted in Fig. 4a. Lineshape analysis exhibited a significant increase in T 2 * relaxation time of the TP phase after inhalation of LPS compared to baseline (Fig. 4b). The chemical shift of the RBC resonance was significantly reduced compared to baseline (Fig. 4c). 1 H T 1 was significantly elevated after LPS challenge, 1187.8 msec, compared to baseline, 1157.6 msec, P = 0.027 (Fig. 2c). No significant changes of pulmonary blood volume, P = 0.64, pulmonary blood flow, P = 0.17, and mean transit time, P = 0.11, were observed in DCE MRI. A significant correlation between the change of RBC-TP and change of neutrophil fraction from sputum analysis was observed (Fig. 5).

Discussion
The feasibility of 129 Xe MRI, 1 H T 1 mapping and DCE MRI was investigated for noninvasive monitoring of the response of human lungs to inhaled LPS challenge as well as their relation to cell counts as established markers of inflammation. The results of our study show the feasibility and added value of MRI through noninvasive assessment of the inflammatory response at the alveolar level when compared to induced sputum collection. While the RBC-TP ratio is sensitive for gas diffusion limitation, 22 it is clear that there is also a contribution from perfusion to this quantity. 23 The results of DCE imaging : (a) CSSR gas uptake measurements at baseline and after inhaled LPS challenge. Surface-to-volume ratio is proportional to initial slope for low delay times, septal wall thickness is given by time to saturation and the slope for high delay times increases with blood flow velocity, ie, inverse of the capillary transit time. The level of saturation is proportional to the volume ratio of parenchyma to air spaces. (b) No significant change was observed for the CSSR-derived septal wall thickness parameter after LPS inhalation, P = 0.193. (c) There was a trend to an increased surface-to-volume ratio after challenge, P = 0.084. (d) A significantly increased (P = 0.020) capillary transit time corresponding to reduced blood flow velocity is observed after inhaled LPS challenge. There is one outlying data point with a very high capillary transit time of 65.6 sec after LPS. Without this subject, there still was a significant increase, P = 0.040. CSSR, chemical shift saturation recovery; LPS, lipopolysaccharide; S/V, surface-to-volume ratio; TP, tissue/ plasma.
showed no significant changes in pulmonary parenchymal perfusion and pulmonary blood volume after LPS challenge. This suggests that the reduction of RBC-TP after inhaled LPS challenge is primarily attributable to pulmonary edema and not due to parenchymal hypoperfusion. The increase in proton T 1 is consistent with this interpretation, since water content increases T 1 . 8,24,25 The increased capillary transit time parameter from 129 Xe CSSR measurements after inhaled LPS challenge is likely due to vasodilation, leading to reduced blood flow velocity at the capillary level. The fact that no such change is observed for mean transit time from DCE MRI suggests that hyperpolarized 129 Xe MR is more specifically probing blood flow at the level of capillaries.
The increase of TP T 2 * may suggest a reduced local heterogeneity of magnetic flux density due to reduced air/alveolar surface interfaces due to septal edema and increased fluid content of the alveoli after LPS inhalation 26 or a slower chemical exchange between TP and RBC phases. A similar increase of TP T 2 * has previously been observed in idiopathic pulmonary fibrosis. 22 It is also known that 129 Xe chemical shift in RBCs is a marker for blood oxygenation 27 and the reduction in chemical shift hence may suggest reduced oxygenation due to an oxygen diffusion limitation. Further research is necessary in order to validate this finding.
Septal thickening and a decreased septal surface density have been reported using histology and septal thickening has also been observed previously using CSSR in mouse models of inflammatory lung injury using LPS. 28,29 One would have  expected the CSSR septal wall thickness and potentially also surface-to-volume ratio parameter to significantly increase after inhalation of LPS, which was not the case. This might have to do with the relatively low dose of inhaled LPS and small sample size in our study. In a previous reproducibility study, 16 the mean coefficient of variation was found to be 4.7% for the wall thickness parameter and 12.2% for the surface-to-volume ratio. Another reason might be the dependence of these two quantities on lung inflation, 30,31 which makes the method prone to errors despite the standardization of breathing maneuver.
It would have been possible in principle to perform this feasibility study in large animals. We decided, however, to perform it in healthy human volunteers since the translation of some results would not have been clear, for example because the spectral properties and even number of resonances of the hyperpolarized 129 Xe dissolved phase are different among different species. 32 Recently, Svenningsen et al 33 found the volume of ventilation defects in hyperpolarized gas MRI after application of a bronchodilator to be quantitatively associated with the amount of eosinophils in sputum in severe asthmatics and thus suggested ventilation imaging for assessment of the inflammatory component of airway disease. Ventilation defects have also previously been observed using 3 He MRI in mice after inhalation of a higher dose of 4-5 μg LPS. 34 In comparison, our results indicate that in the presented human low-dose LPS inhalation model there are no clear changes in ventilation distribution, although the nonsignificant increase of group median ventilation defect percentage by 1.5% after LPS may point to a small change below our detection limit.
The correlation of change in RBC-TP and change in neutrophil fraction in induced sputum constitutes initial evidence that MRI is able to quantify the degree of inflammation in terms of cell counts. This relation still needs to be confirmed in future studies with larger cohorts.
One of the limitations of this study is the relatively small sample size. This makes the assessment of relationships between quantities derived from MRI and physiologic processes derived from cell counts difficult. Previous studies with positron emission tomography were able to show correlations of [ 18 F]fluorodeoxyglucose uptake and neutrophil counts from bronchoalveolar lavage in various diseases, 35,36 although also only a weak correlation with r 2 = 0.21 was observed in a segmental LPS model of lung inflammation. 37 Pulmonary perfusion measurements using DCE MRI also have various limitations with regard to physiological variations like the dependence on heart rate and inspiratory level 38 as well as technical limitations. 39 A limitation of our spectroscopic lineshape analysis is the negligence of non-Lorentzian effects in spectral broadening. 26,40 In conclusion, hyperpolarized 129 Xe dissolved-phase MRI provides promising biomarkers, especially the RBC-TP ratio, for assessment of lung function at the alveolar level, which enable longitudinal monitoring of the response of human lungs to LPS challenge. 129 Xe MRI and 1 H T 1 mapping yield important complementary regional information in addition to the cell count derived from induced sputum and show promise as tools for monitoring inflammation in drug development.