Preclinical Four-Dimensional Functional Lung Imaging and Quantification of Regional Airflow: A New Standard in Lung Function Evaluation in Murine Models

Abstract

The current standard for lung function evaluation in murine models is based on forced oscillation technology, which provides a measure of the total airway function but cannot provide information on regional heterogeneity in function. Limited detection of regional airflow may contribute to a discontinuity between airway inflammation and airflow obstruction in models of asthma. Here, we describe quantification of regional airway function using novel dynamic quantitative imaging and analysis to quantify and visualize lung motion and regional pulmonary airflow in four dimensions (4D). Furthermore, temporo-spatial specific ventilation (ml/ml) is used to determine ventilation heterogeneity indices for lobar and sublobar regions, which are directly compared to ex vivo biological analyses in the same sublobar regions. In contrast, oscillation-based technology in murine genetic models of asthma have failed to demonstrate lung function change despite altered inflammation, whereas 4D functional lung imaging demonstrated diminished regional lung function in genetic models relative to wild-type mice. Quantitative functional lung imaging assists in localizing the regional effects of airflow. Our approach reveals repeatable and consistent differences in regional airflow between lung lobes in all models of asthma, suggesting that asthma is characterized by regional airway dysfunctions that are often not detectable in composite measures of lung function. 4D functional lung imaging technology has the potential to transform discovery and development in murine models by mapping out regional areas heterogeneously affected by the disease, thus deciphering pathobiology with greater precision.

Traditionally, lung function in preclinical models has been measured using the forced oscillation technique. This technology is based on involuntary changes in pressure, volume, and airflow in response to oscillatory airflow waves (1, 2). In clinical medicine, quantification of expiration lung volume is the clinical standard for assessing lung function (3). Both pulmonary function tests are accurate and provide overall measurements of the whole lungs, but they are unable to provide any information on the regional heterogeneity in lung function, may miss significant localized disease, and thus are a cause for discontinuity in translational medicine.

Advanced techniques using magnetic resonance imaging (MRI) of patients with asthma and ovalbumin (OVA) mouse models with hyperpolarized helium contrast showed ventilation defects of lower and central lung regions that are likely associated with airway narrowing and collapse caused by remodeling of the airway wall (4–7). However, in all these studies, static MRI images did not provide any information about alterations in airflow.

Clinical correlates such as ventilation heterogeneity (VH), originally measured by multiple-breath nitrogen washout (8), a well-recognized feature of asthma (9) that has deleterious effects on gas exchange (10), is a strong determinant of airway hyperresponsiveness, independent of airway inflammation (11). Moreover, ventilation defect percentage (VDP) is a measure of regional airway obstruction and air trapping in asthma (12) and has been demonstrated to be an important biomarker of severe asthma exacerbation and hospitalization (13, 14). To date, lung imaging is either static or quasi-dynamic (accumulated over time), or the technology is prohibitive in availability of cost; therefore, we aimed to describe a novel dynamic technological advanced modality that is relatively simple and affordable.

Four-dimensional X-ray velocimetry (4DXV) technology (three-dimensional space and time) is a unique functional lung imaging approach to quantify and visualize the dynamic motion of lung and regional pulmonary airflow (15, 16). The foundational design and development for four-dimensional (4D) lung imaging was conducted by Dr. Fouras and the Laboratory for Dynamic Imaging at the Monash University (17–24). Cross-disciplinary innovation led to a second-generation small animal scanner, used in this work, and has enhanced capabilities for greater resolution and mapping of a range of manifestations and severities of dysfunction associated with heterogeneous diseases. The ability to study dynamic alterations in airflow at a regional level now enables us to overcome the limitations of status quo technology, which have previously been dampened by functional imaging modalities unavailable for translational discovery and development.

Methods

Animals

Male or female mice at the age of 12 weeks were used in three murine models of airway inflammation: 1) wild-type (WT) mice (Jackson Laboratory); 2) Eotaxin1,2^−/− (Eotaxin-1 and 2 double-knockout) mice on the BALB/c background (25); and 3) ARG2^−/− (arginase-2 knockout) and iNOS^−/− (inducible nitric oxide synthase knockout)/ARG2^−/− mice on a C57BL/6 background (26, 27) (Lerner Research Institute).

Models of Airway Inflammation

OVA model mice were allergen sensitized by an intraperitoneal injection of 10 µg OVA and 20% aluminum hydroxide (Sigma-Aldrich) in 100 µl of saline. Two weeks after sensitization, animals were daily challenged with aerosolized OVA (1% w/v in saline) for 7 days (25, 28, 29). Imaging and airway hyperreactivity measurements were performed 1 day after the final allergen exposure.

House dust mite extract/complete Freund’s adjuvant (HDME/CFA) model (30) mice were subcutaneously sensitized with 100 µg HDME (Dermatophagoides farinae), mixed with 100 μl saline and 100 μL CFA. Two weeks after the sensitization, mice received a single intranasal challenge (100 μg HDME/50 μl saline). Animals were allowed to develop airway inflammation for the next 2 days before measurements were performed.

Measurements of Airway Hyperreactivity

Animals were anesthetized with an intraperitoneal injection of 50 mg/kg pentobarbital, followed by tracheotomy. Airway hyperreactivity to increasing doses of nebulized methacholine was measured using a FlexiVent Fx2 system (SCIREQ) (28, 31).

Dynamic Functional Lung Imaging

4DXV combines two technologies: four-dimensional computed tomography (4DCT) and X-ray velocimetry (XV). The imaging system is based around a rotating anode X-ray source with an elliptical spot size of 58 × 80 µm. The relatively small spot combined with the relatively large propagation distance (source-to-sample distance of approximately 350 mm; sample-to-detector distance of 1,700 mm) provides for propagation-based phase contrast enhancement of the images, providing for significant edgeenhancement within images. 4DCT is performed on a specialized small animal X-ray scanner (Permetium, 4DMedical) to achieve the necessary tissue and airway resolution and contrast of the lungs. A key feature of this technology is a microfocus X-ray source to achieve phase contrast enhancement of the lung tissue (32, 33). The X-ray velocimetry technology is based on and modified from particle image velocimetry, commonly used to monitor air velocity in aeronautic application (31). In this case, X-ray velocimetry is applied to the 4DCT data to measure the velocity of lung tissue motion and airflow by extension. Pulmonary airflow at the tissue level is mapped to the airway tree, obtained from the computed tomography (CT), to provide a quantitative measurement of flow in the airways (16, 19). The various stages of the preclinical functional lung imaging workflow are illustrated in Figure 1. Velocimetry is performed on the volumetric CT data by computing three-dimensional cross-correlations to measure the displacement between small regions of the CT at consecutive time points, each corresponding to a different phase of the respiratory cycle. The result is a vector indicating the direction and magnitude of motion or velocity of that region. Calculation of the divergence of the total vector field yields the tissue expansion of each region, which is a surrogate for regional ventilation accounting for the magnitude of tissue compressibility (21). The CT at start-inspiration is used to identify the airway tree. Regional ventilation is mapped to the airway tree to calculate regional airflow through each airway. The supplying airway for each lobe is identified and used to create a mask that is segmented by lobe. The regional ventilation data are then segmented into lobes to calculate lobe-specific ventilation statistics.

Figure 1.

Preclinical four-dimensional (4D) functional lung imaging data acquisition and analysis. A typical workflow is illustrated. (A) An X-ray image of a mouse positioned in the scanner. A series of volumetric data is obtained by 4D computed tomography (4DCT) imaging gated to the respiratory cycle. The 4DCT scans also provide images to reconstitute the airway tree and perform lobe segmentation. (B) A single 4DCT scan slice from one time point is shown. (C) Next, velocimetry analysis is applied on 4DCT data from each time point to obtain a 4D map of local lung tissue displacement. (D) The displacement gradient is then used to calculate time-resolved ventilation. The regional ventilation is mapped to the airway tree to obtain time-resolved regional airflow. (E) Lobar ventilation metric is also obtained by mapping the data to lobe segmentation. A typical functional lung report summarizes the regional ventilation in color maps, distribution histograms, and ventilation metrics. (F) An anterior-posterior (AP) slice and three axial slices of the specific ventilation are shown. Arrows indicate areas with high ventilation (blue) and low ventilation (red). (G) The ventilation distribution illustrates the regional ventilation heterogeneity (VH) and ventilation defect percentage (VDP). A narrow or wide ventilation distribution indicates low or high VH, respectively. (H) A metrics table provides values of tidal volume, VDP, and VH. VH small scale quantifies VH in local areas (alveolar to lobar size), whereas VH large scale measures the heterogeneity in larger regions (lobar and larger) areas. (I) Schematic representation of sublobar segmentation. 3D = three dimensional; A = accessory; I = intermediate; IF = inferior; L = left; S = superior.

Ventilation Settings

Mice were weighed and anesthetized using pentobarbital. Upon no reaction to a toe pinch, a tracheotomy was performed, and the trachea was cannulated (20 ga 1/2″ Luer Stub, Instech Laboratories). The mouse was placed in a holder, the cannula was attached to the ventilator, and the front limbs were taped above the head to secure the animal in position. The endotracheal tube was attached to a custom-built small animal volume-controlled ventilator (AccuVent, 4DMedical) calibrated to a volume of 0.2 ml. Peak inspiratory pressure of approximately 14 cm H₂O was needed to achieve a volume of 0.2 ml, although this varied slightly between animals. Positive end-expiratory pressure was 2 cm H₂O, with respiratory rate of 129 breaths per minute.

Scanner Settings

Mice were mounted in a vertical position in front of the X-ray source on a custom high- precision rotation stage. The source-to-detector distance was fixed, and translation of the stage enabled the mice to be moved toward and away from the source to alter the zoom factor as required. Mice were rotated through 360 degrees while a flat-panel detector captured images. Image acquisition was triggered by the ventilator and was gated to collect images at 14 time points throughout the breathing cycle, with an effective pixel size of ∼19.0 μm. Six hundred projections were collected at each time point, for a total of 8,400 projections.

Image Processing and Resolution

The X-ray images were analyzed using a cross-correlation–based motion measurement technique to obtain complete quantification of the 4D motion field of the lungs. Each measurement voxel represents a region of ∼1.8 mm³ of lung tissue isotropically spaced every 0.3 mm in space at each of the 14 gated acquisition phase-points per breath.

Visualization and Summary Statistics in Preclinical Functional Lung Imaging

The spread of specific ventilation across the lungs over a full inhalation is depicted and summarized in a standard visualization and summary statistics report (Figure 1). These outputs are standard for every animal. Specific ventilation is defined as the ratio of the volume of gas entering a region of the lung (ΔV) after an inspiration, divided by the end-expiratory volume (V0) of that same lung region: specific ventilation = ΔV/V0. For reporting purposes, the specific ventilation is normalized to the overall mean for that measurement. A color map depicting high, normal, and low areas of regional normalized specific ventilation is shown on a coronal slice and three axial slices (Figure 1F). A ventilation distribution histogram (Figure 1G) shows the distribution of the normalized specific ventilation across the lungs. The distribution also illustrates the VH, which is defined as the ratio of the interquartile range to the mean specific ventilation. A low VH indicates uniform ventilation throughout the lung and is typically observed in healthy mice. High VH signifies substantial variability in the lungs. VDP is calculated as the percentage of specific ventilation below 0.6 of the mean. The threshold selected for VDP is more appropriately considered to be a function of lung physiology than of imaging modality or parameters. Because the hyperpolarized MRI approach and 4D imaging method presented here both seek to quantify ventilation within the lung, it seems appropriate to use the same thresholds. Values of all the measurements are also summarized in a table format (Figure 1H) for statistical analyses.

Lobar and Sublobar Comparisons between Imaging and Ex Vivo Biological Analyses

Functional lung imaging allows lobar and sublobar segmentation of the specific ventilation. The technology is able to image the individual airways in mice, but microdissection of each airway for ex vivo biological studies is challenging, because mouse airways are very small. The natural anatomical curvature of the lobes does facilitate reproducible sublobar dissection of the tissue (Figure 1I). The CT at start-inspiration was used to identify the airway tree. Regional ventilation was mapped to the airway tree to calculate regional airflow through each airway. The supplying airway for each lobe was identified and used to create a mask that was segmented by lobe. The regional ventilation data were then segmented into lobes to calculate lobe-specific ventilation statistics. Using this approach, we are able to segment all functional lung imaging data on a lobar basis or a sublobar basis that precisely matches the dissections and therefore the ex vivo biological analyses.

Statistical Analysis

JMP 14 software was used for all statistical analysis. Specific tests for each experiment are mentioned in the figure legends. A log transformation was performed on skewed data for statistical analysis. ANOVA was performed for multiple-group comparison. P values < 0.05 were considered significant. Data are shown as mean ± SEM unless described differently.

Results

Strengths of Functional Lung Imaging over Traditional Methods of Lung Function Assessments

We used two genetic deletion mouse models that have been intensively studied by our group to illustrate the stringency of functional lung imaging and various ways in which preclinical functional lung imaging data are presented. It has long been known that there is a disconnect between eosinophilic airway inflammation and airway hyperreactivity; both occur independently from each other (34–36). The Eotaxin1,2^−/− mice do not develop eosinophilia in the OVA allergen–induced airway inflammation model, and the deficiency of these two major eosinophilic chemoattractants did not affect global lung function (37). We have previously used this model to study the contribution of angiogenic pathways to eosinophilic airway inflammation (25, 31), and we confirmed using FlexiVent that eotaxin-1/2 deficiency does not affect airway resistance in the OVA model (see Figure E1 in the online supplement). In contrast, visualization of the regional ventilation using functional lung imaging of wild-type and Eotaxin1,2^−/− in the OVA model showed at first sight that specific ventilation is increased in the knockout mice (Figures 2A and 2B). Quantification of VDP and VH confirmed that absence of eotaxin-1/2 does benefit lung function (Figures 2C and 2D). These findings show that, unlike previous reports using traditional technology, functional lung scanning demonstrates that eotaxin-dependent recruitment of eosinophils mediates lung function.

Figure 2.

Functional lung imaging of lung function in two knockout models of airway inflammation that have no effect on global airway hyperreactivity. Two knockout mouse models were used to illustrate the strength of this novel technology. In the first model, wild-type (WT) BALB/C and Eotaxin1,2^−/− mice in ovalbumin (OVA)-induced airway inflammation were used. Coronal slices of the regional ventilation of (A) WT and (B) Eotaxin1,2^−/− are shown. Quantitative metrics (C) VDP and (D) VH show that eotaxin-1/2 deficiency improved regional lung function. (E–G) The second model involved mice ARG2^−/− (arginase-2 knockout [KO]) and ARG2^−/−/iNOS^−/− (inducible nitric oxide synthase KO) on a C57BL/6 background. Scans of one coronal and three axial slices, and specific ventilation distribution histograms of WT, ARG2 KO, and iNOS^−/−/ARG2^−/− mice in a standard house dust mite extract (HDME) model are shown. (H and I) VDP% and VH for each genotype show that the combined deficiency of iNOS and ARG2 decreased VDP% and VH. Each circle or diamond represents one data point from a female or male mouse, respectively. Two-tailed t test was performed for group comparisons. DKO = double knockout.

The other knockout model was previously used to reveal arginine metabolic control of airway inflammation (26, 27). Arginine is the common substrate for iNOS and ARG2. ARG2^−/− mice develop a severe airway inflammation phenotype (27) that was diminished in INOS^−/−/ARG2^−/− mice in a standard HDME model (26). Using standard FlexiVent measurements, we reported that these genetic deletions do not affect airway hyperreactivity (26) (Figure E2). In contrast, functional lung image scans and distribution diagrams of the specific ventilation in Figures 2E–2G showed better regional ventilation in iNOS^−/−/ARG2^−/− relative to wild-type and ARG2^−/− mice in the HDME model.

Quantification of the VDP and VH (Figures 2H and 2I) demonstrated that in comparison with wild-type animals, lung function was also better in iNOS^−/−/ARG2^−/− animals. The results show that, in contrast to previous reports, arginine metabolism regulates regional lung ventilation.

Lobar and Sublobar Functional Lung Imaging in CFA/HDM Mouse Model of Airway Inflammation

Using our segmentation and dissection methods, we analyzed lobar heterogeneity in a commonly used HDME model. Axial slices of the specific ventilation, VDP, and VH are shown in Figures 3A–3F. Control mice receiving CFA alone (Figures 3B and 3C) showed higher VDP in the accessory lobe, whereas VH was homogenous across lobes. HMDE-exposed mice (Figures 3D–3F) exhibited lobar heterogeneity in both VDP and VH. Quantification of the 4D imaging metrics in the subregions in an HDME-exposed mouse demonstrated marked intralobar heterogeneity in specific ventilation, VH, and VDP (Figure E3). MRI imaged regional ventilation defects are found in patients with asthma (38) and involve specific lobes, dependent on the severity of the disease (39).The functional lung imaging results here show that quantification of lobar and sublobar ventilation is now also possible in murine models of airway inflammation.

Figure 3.

Quantification of lobar heterogeneity in HDME model of airway inflammation. CT images were used to identify the airway tree in each mouse. The airflow in each airway was quantified by mapping regional ventilation to the airway tree. For each lobe, the delivering main airway was identified and a lobar mask was generated. The regional ventilation measurements were then segmented into lobes to obtain lobe-specific ventilation metrics. Axial slices, VDP, and VH for (A–C) control-treated and (D–F) HDME allergen–exposed mice are shown. Each circle or diamond represents one data point from a female or male mouse, respectively. ANOVA was performed for comparisons across groups. CFA = complete Freund’s adjuvant.

Discussion

Here we describe a novel lobar and sublobar quantitative assessment of lung function in mice. Researchers and clinicians have long known that asthma presents heterogeneously in varying airways, yet the disease is still diagnosed using a breath test that measures the average cumulative airflow in all airways. This standard-of-care diagnostic approach dates back as far as the 1800s and, although reliable, cannot identify subregions of the disease process (40). Here we described and demonstrated the capabilities of a novel functional lung imaging technology in preclinical mouse models of asthma. A standard data report provides easy-to-interpret visualization and metrics of regional ventilation.

As an illustration of the strength of this new technology, we used two models that have been well-studied by our group. In an Eotaxin1,2^−/− model we showed that eotaxin deficiency improved lung ventilation in OVA allergen–exposed animals. In contrast, current standard airway hyperreactivity measurements showed no effects, as reported previously (26, 27, 37). Arginine metabolism regulates airway inflammation, as demonstrated in clinical studies and mice knockout for arginine-converting enzymes, ARG2 and iNOS (26, 27). Although global airway hyperreactivity is not affected in these knockouts, functional lung imaging showed improved lung function in house dust mite allergen–exposed iNOS^−/−/ARG2^−/− mice. Our data suggest that the previously reported disconnect between airway inflammation and lung function in these two models was inaccurate, because traditional technology is not sensitive enough. We do not know if this holds true for other models, where airway inflammation and global airway hyperreactivity are uncoupled (34–36). However, there will be a paradigm shift in lung biology if further studies reveal that regional, but not global, inflammation and lung function are closely interconnected. Another possibility is that when global pulmonary function tests, but not 4DMedical Imaging, are normal in a disease model, areas of increased ventilation coexist with areas of reduced lung function. Regional ventilation is most likely governed by physiological and mechanical stress relationships that are beyond the scope of this Major Technical Advances paper but would be investigated in future studies using a hypothesis-driven paradigm. 4D preclinical medical imaging enables longitudinal imaging without tracheotomy, but the effects of low-dose irradiation on lung function are unknown.

Lobar analyses of the specific ventilation showed consistent differences across the lobes in allergen-naive mice, predominantly in T-helper cell type 2 (Th2) or Th17 models of asthma. The underlying biology behind the regional ventilation heterogeneity is currently unknown and is a fundamental question. The presence of heterogeneity in healthy lungs suggests that it is part of physiology, perhaps attributable to the physical properties of the airways. For example, local airflow is determined by the driving pressure and mechanical time constant in the region, and an airway with increased local stiffness or a narrow airway will have lower airflow. Innate immune cells resident in the lungs may also be having local effects on the airway tone. In allergen-exposed lungs, the airway inflammation is not equally distributed (41). Several mediators originating from inflammatory cells are known to contribute to airway constriction (42, 43). Also, airway remodeling, including goblet cell metaplasia, airway fibrosis, and the proliferation of airway smooth muscle cells, is most likely also heterogeneous across the lungs.

Asthma is a cluster of airway diseases manifesting with variable symptoms and showing an inconsistent response to therapy. In current clinical practice, lung function is measured by gathering data at the openings of large airways yielding measurements of whole-lung function rather than providing detailed information about local regions of the tracheobronchial tree. Functional lung imaging allows us to measure regional ventilation, which has proven invaluable in the mapping of regional areas that are heterogeneously affected by asthma, allowing us to decipher the pathobiology of this disease with more precision. In addition, this new technology allows gathering of important clinical information to help tailor drug regimens to focal regions of the airways and provide new biological insights for more effective treatment strategies.