Monitoring in vivo changes in lung microstructure with ³He MRI in Sendai virus-infected mice

Abstract

Recently, a Sendai virus (SeV) model of chronic obstructive lung disease has demonstrated an innate immune response in mouse airways that exhibits similarities to the chronic airway inflammation in human chronic obstructive pulmonary disease (COPD) and asthma, but the effect on distal lung parenchyma has not been investigated. The aim of our study is to image the time course and regional distribution of mouse lung microstructural changes in vivo after SeV infection. ¹H and ³He diffusion magnetic resonance imaging (MRI) were successfully performed on five groups of C57BL/6J mice. ¹H MR images provided precise anatomical localization and lung volume measurements. ³He lung morphometry was implemented to image and quantify mouse lung geometric microstructural parameters at different time points after SeV infection. ¹H MR images detected the SeV-induced pulmonary inflammation in vivo; spatially resolved maps of acinar airway radius R, alveolar depth h, and mean linear intercept Lm were generated from ³He diffusion images. The morphometric parameters R and Lm in the infected group were indistinguishable from PBS-treated mice at day 21, increased slightly at day 49, and were increased with statistical significance at day 77 (p = 0.02). Increases in R and Lm of infected mice imply that there is a modest increase in alveolar duct radius distal to airway inflammation, particularly in the lung periphery, indicating airspace enlargement after virus infection. Our results indicate that ³He lung morphometry has good sensitivity in quantifying small microstructural changes in the mouse lung and that the Sendai mouse model has the potential to be a valid murine model of COPD.

chronic obstructive pulmonary disease (COPD) is a major cause of mortality and disability worldwide and is one of the only major diseases that is increasing (19). It is characterized by distal airspace enlargement (emphysema) and/or small-airway limitation, which is not fully reversible, and is associated with an abnormal inflammatory response of the lungs to noxious particles or gases (23). Murine models of COPD have been used effectively and are of critical importance to the study of the pathogenesis of this disease and potential therapeutic interventions (25). However, no existing model exhibits all the traits of COPD and each of them has important limitations. For example, although tracheal instillation of various types of proteases can produce early airspace enlargement with slow subsequent progression, they do not show significant inflammation and the airspace enlargement tends to be more panacinar rather than the centriacinar disease seen most often in humans without alpha-1 antitrypsin deficiency (4, 18). Moreover, it is difficult to extrapolate the findings on the acute response to instillation to the progressive chronic disease in humans (2). Although murine cigarette smoke-induced COPD has a significant inflammatory reaction and has linked the animal models to humans, the enlargement of airspace appears to be strain dependent (6), and the exposures to smoke and the morphometric analyses are not well standardized (2, 18).

Sendai virus (SeV) infection in mice is an experimental model of chronic lung disease with pathology that resembles asthma and COPD in humans (11). It has been widely investigated as a suitable model of respiratory viral infections (12). Recently, a new type of innate immune response has been identified in this model that drives the translation from respiratory viral infection into chronic inflammatory airway disease (14). The effect of SeV on distal lung airspaces, however, has not been investigated before; to our knowledge, there has been no study on lung microstructure alterations caused by viral infection.

With respect to quantification of lung structure, conventional stereological assessment of ex vivo fixed tissue specimens under the microscope is regarded as a gold standard (13), but its great efforts in sampling/observing and its invasive nature makes in vivo and longitudinal study difficult to impossible. The magnetic resonance imaging (MRI)-based ³He lung morphometry technique for performing in vivo lung morphometry has been implemented in humans (30) and was recently developed and validated in mice (21, 29) and dogs (8). In this approach, lung microstructure is modeled as a network of cylindrical acinar airways surrounded by alveoli (30). This model-based approach relies on prior detailed casts of pulmonary acini in humans, rabbits, and rodents (7, 24). ³He gas diffusion is highly restricted by alveolar walls, and diffusion measurements by MRI can provide information on lung microstructure by capitalizing on the anisotropy of the acinar airways. This noninvasive approach enables us to measure the same morphometric parameters as in quantitative morphologic techniques, but in vivo, serially, and with tomographic information inherent in imaging (29, 30).

Here we quantify mouse lung microstructure with ³He diffusion MRI at different time points after SeV infection in vivo. The main issue we attempted to address was whether SeV infection causes enlargement of distal lung airspaces. To our knowledge, this is the first investigation into the microstructure changes in this mouse model of chronic inflammatory lung disease and first in vivo study of this model via imaging.

MATERIALS AND METHODS

Animals.

All experiments were approved by the Washington University Animal Studies Committee. Five groups of five C57BL/6J 5- to 6-wk-old mice were used for SeV and PBS treatment: three groups of mice were infected by SeV and studied at 21 days, 49 days, and 77 days after infection; two groups of mice were treated with PBS and studied at 49 days and 84 days after treatment; one group of untreated healthy mice were studied at 12 wk.

Sendai mouse model induction.

SeV was obtained from ATCC (VR-105). The virus was twice plaque purified, propagated in 10-day-old embryonated chicken eggs, and stored at −70°C. Infectivity of SeV was assayed in VeroE6 cell line. Mice 5–6 wk old were infected intranasally with 2×10⁵ pfu/mouse of SeV in 30 μl and kept in a barrier facility until time of experiment.

Preparation of animals for imaging.

The mice were anesthetized with an intraperitoneal ketamine/xylazine mixture and orally intubated. Each animal was attached supine to a custom-built MR-compatible ventilator (5) (120 breaths/min, 0.25-ml tidal volume) and anesthetized with inhaled isoflurane for imaging experiments. Each breathing cycle consisted of pure O₂ delivered in the first half and ³He/N₂ gas mixture in the second half of inspiration, followed by a brief breath hold (∼35 ms) for image acquisition and then passive exhalation, resulting in a measured peak airway pressure of 15 cmH₂O. The respiration curve was measured during imaging and used to monitor pressure of the gas supplied during ventilation. The pressures used for all the animals involved in this study were strictly controlled and well maintained during imaging process.

Imaging.

Hyperpolarized ³He gas with a polarization level ∼40% was produced using spin exchange optical pumping with home-built and commercial polarizers (GE Healthcare) (27). MR imaging was performed on a Varian 4.7 T horizontal-bore magnet (40-cm magnet bore and 12-cm gradient bore) with a custom-built doubly resonant solenoid coil tuned to both ³He and ¹H frequencies (151.1 and 198.3 MHz, respectively). A trigger signal from the ventilator ensured that image data acquisition began at peak inspiration in each breathing cycle, the details of which have been previously described (29). Briefly, a two-dimensional (2D) multislice gradient-echo sequence was used for ¹H MR imaging for anatomical orientation and to detect inflammation (at peak inspiration of each breath one line of k space for all 5 slices were acquired during 17.95 ms, echo time TE = 0.99 ms, flip angle = 50°). Then the ³He images were acquired at precisely the same slice position as ¹H images and the same ventilation pressure. A 2D multislice gradient-echo sequence was used for ³He diffusion MRI with embedded bipolar diffusion-sensitizing gradients with six b values of 0, 1, 2, 4, 6, 9 s/cm² [matrix = 64 × 64; field of view (FOV) = 40 × 40 mm; slices = 5; thickness = 2.0 mm; at peak inspiration of each breath one line of k space at 1-b value for all 5 slices were acquired during 21.6 ms, TE = 2.44 ms; the maximum diffusion sensitizing gradient strength = 27.87 gauss/cm, length of each lobe δ = 440 μs, rise time τ = 150 μs, and no gap between lobes (Δ = δ), gradients along read out direction]. Both ¹H and ³He images were acquired in a k-space interleaved fashion. Five axial imaging slices nearly covered the entire lung. The SeV-infected mice were imaged as described above at days 21, 49, and 77 after infection; the PBS mice on day 49 and day 84 after treatment; and the untreated control group at age 12 wk.

Data analysis.

³He MR images of 6-b values are fitted to the cylindrical model of acinar airways on a pixel by pixel basis using Bayesian analysis software (21). This generates parametric maps including alveolar depth h, alveolar-duct radii R, the mean linear intercept (Lm), number of alveoli per unit lung volume(N_a), and alveolar surface-to-volume ratio (S/V) (29, 30).

The lung volumes were estimated from both the ³He ventilation images and ¹H images. For ³He images, the MRI signal voxels above a threshold that was set to ten times the value of noise were counted to segment the lung from background. For ¹H images, lung parenchyma was segmented by counting all connected voxels in the thoracic cavity minus the heart and large vessels. The trachea and major bronchi were manually removed in both methods.

Statistical analysis.

Standard one-way analysis of variance tests were used to compare results among groups. A P value of <0.05 was considered significant for the comparison.

Quantitative histology.

Mice were euthanized and the trachea was cannulated with an 18-gauge intravenous catheter and 4% paraformaldehyde was instilled into the lungs at 15 cm water pressure. This pressure was chosen to approximate the pressure administered by the custom ventilator during imaging. Lungs were inflation-fixed for a minimum of 15 min in the intact thorax and subsequently removed en bloc and fixed overnight. Following fixation, tissues were embedded in 4% agarose (BP1423, Fisher Scientific, Houston, TX) in distilled water. The agar block was placed into a custom cutting device with an integrated micrometer that allows precise movement of the stage; 2–3 mm agar sub-blocks were cut from the agar block with a fresh microtome blade. Sub-blocks were placed with cut surface next to a ruler, and photographs were taken pre- and postprocessing to determine the amount of shrinkage attributable to tissue processing for each slice to account for nonuniform shrinkage.

Tissue sections 5 μm thick were cut and stained with hematoxylin and eosin, and images of each entire slide were captured on a Nanozoomer virtual microscope (Hamamatsu, Bridgewater, NJ). Chord length was determined using the public domain NIH Image 1.63 program (1). Large airways, blood vessels, and areas of consolidated inflammation were excluded. The chord length subroutine draws separate horizontal and vertical reference grids and then calculates the length of line segments created by the intersection of these grids with the tissue walls. A less-automated approach using point counting and randomized test lines allows more information than Lm but was not employed here (15). For each set of lungs, at least 10 images per sub-block were analyzed; this represents >30–50 images acquired from sections that were at least 2 mm apart for each mouse and spanned 10–12 mm. Each sub-block was at approximately the same slice position and orientation as ³He MR images.

RESULTS

The signal-to-noise ratio (SNR) for ³He MR images ranged from 60 (for b = 9, standard deviation ∼ 12) to 170 (for b = 0, standard deviation ∼ 24). A representative set of ¹H MR images and HP ³He ventilation images is shown in Fig. 1.

Fig. 1.

Representative set of ¹H magnetic resonance (MR) images of PBS 49 day mouse (top) and corresponding set of HP ³He ventilation images (bottom).

In the groups of SeV-infected mice, the ¹H images clearly revealed inflammation, as expected, as areas of increased intensity (spin density), which corresponded to ventilation defects (dark areas) in ³He images at b = 0. This was further confirmed by histology (Fig. 2). Areas of inflammation increased with time after infection through 77 days. Because there was nearly no signal from those regions in ³He images, corresponding pixels were automatically eliminated from the parametric maps and subsequent analysis.

Fig. 2.

One representative ¹H MR image (top left) and corresponding ³He ventilation image (bottom right) of a Sendai (SeV) 77 day mouse. Microscope images correspond to the regions outlined in MR images. Lung parenchyma in the blue square is relatively normal, whereas the red square shows significant inflammation induced by SeV. This is clearly shown by the higher ¹H signal and lower ³He signal in the red square compared with the blue one.

The fitting by Bayesian analysis on a pixel-by-pixel basis was excellent for every mouse in this study, and the root mean square (rms) residues were all below 3% of the rms signals. Figure 3 shows a representative set of parameter maps for a SeV 21 day mouse. Parameter maps of R, h, Lm, N_a, S/V were generated for each mouse studied, and the individual data are summarized in Table 1.

Fig. 3.

Representative parametric (R, h, and Lm) maps obtained from a SeV 21 day mouse. Major airways are excluded. Color bar shows the range of the parameters (R, h, and Lm) from 35 to 135 μm.

Table 1.

Summary of morphometric parameters obtained via ³He MRI from PBS 49 day, PBS 84 day, SeV 21 day, SeV 49 day, SeV 77 day, and untreated control mice with histological comparison

Mouse	R, μm	h, μm	S/V, cm−1	Na, mm−3	Lm, μm	Lm by Histology, μm	Volume (3He MR), mm3	Volume (1H MR), mm3
PBS 49 day	102.8 ± 1.3†	48.8 ± 0.7	587 ± 15†	3204 ± 102†	70.1 ± 1.8†	53.7 ± 5.3	1,208 ± 135	1,055 ± 125
PBS 84 day	102.1 ± 0.8†	46.0 ± 0.5	562 ± 9†	3181 ± 69†	72.3 ± 1.4†	−	1,094 ± 218	940 ± 179
SeV 21 day	102.9 ± 1.1†	49.9 ± 0.2	587 ± 9†	3219 ± 71†	71.4 ± 2†	65.1 ± 5.6*†	1,155 ± 287	997 ± 253
SeV 49 day	103.5 ± 0.9†	48.4 ± 0.7	585 ± 15†	3216 ± 105†	71.2 ± 1.7†	59.7 ± 8.5†	1,250 ± 248	1,103 ± 174
SeV 77 day	105.7 ± 2.4*§†	48.1 ± 1	559 ± 27*§†	3020 ± 174*§†	75.4 ± 4.4*§†	66.9 ± 3.7*†	1,212 ± 220	1,063 ± 154
Control	97.2 ± 3.7*§	51.7 ± 3.2	698 ± 37*§	3841 ± 317*§	60.5 ± 3.5*§	52.0 ± 1.9	917 ± 64	808 ± 69

Values are means ± SD. Within each group all values were not significantly different between PBS mice and Sendai virus (SeV)-infected mice, except

^*P < 0.05 compared with the group of PBS 49 day mice,

^§P < 0.05 compared with the group of PBS 84 day mice, and

^†P < 0.05 compared with the untreated control group.

There is no statistically significant difference between parameters of PBS 49 day mice and PBS 84 day mice. R and Lm in the infected group were indistinguishable from PBS-treated mice at day 21 (R = 102.9 and Lm = 71.4 μm), increased slightly at day 49 (R = 103.5 and Lm = 71.2 μm, respectively), and were increased with statistical significance at day 77 (R = 105.7 and Lm = 75.4 μm, respectively; P = 0.02). Spatial variation of the morphometric parameters was apparent at each time point, with R and Lm more elevated at the lung periphery; these parameters increased as a function of time after infection (Fig. 4, Tables 1 and 2). The mean value of Lm measured by histology for each group, after shrinkage correction, is also shown in Table 1. We note that individual slices shrinkage differed substantially (as much as a factor of 2), but the effect of differential shrinkage is minimized upon averaging over all slices.

Fig. 4.

Representative R, Lm maps and transverse ³He MR ventilation images from groups of PBS 49 day, PBS 84 day, SeV 21 day, SeV 49 day, SeV 77 day at the same slice location. Elevation of R and Lm in the periphery increases with time after SeV infection.

Table 2.

Percentages of voxels with R > 105 μm and the percentages of voxels with Lm > 80 μm for different groups of mice

Mouse	R > 105 μm	Lm > 80 μm
PBS 49 day	17.8%†	10.3%†
PBS 84 day	13.9%	8.4%
Sendai 21 day	21.9%†	14.1%†
Sendai 49 day	24.1%*§†	15.8%*§†
Sendai 77 day	31.8%*§†	19.1%*§†
Control	6.9%*	3.4%*

^*P < 0.05 compared with the group of PBS 49 day mice;

^§P < 0.05 compared with the group of PBS 84 day mice;

^†P < 0.05 compared with the untreated control mice.

To quantify the expansion of acinar airspaces, we performed a threshold analysis on the parameter maps of R and Lm. The percentage of pixels in which R is greater than a range of different threshold values was calculated. The percentage is 100% when threshold = 0 and 0% when threshold is the maximum value of R within all pixels. Figure 5 shows the average calculated threshold curves from parameter maps of R of the six groups. At higher R thresholds, the percentage of pixels greater than the threshold grows with time after infection, indicating that there is a modest increase in acinar airspace size. (Similar results are seen with Lm, not shown in the figure.) Specifically, for R > 105 μm, the percentage of pixels increases from 17.8% for PBS 49 day (14% for PBS 84 day) to 31.8% for SeV 77 day (Table 2). Similar results are also shown for Lm in the table.

Fig. 5.

Percentages of voxels with R above a range of the threshold values for all groups of mice. Curves for control (brown dashed line), PBS 49 day (purple dashed line) and PBS 84 day (blue dashed line) have lower percentage values compared with SeV-infected groups (solid line) at high value range of R (100–105 μm shown here), indicating there are more pixel with higher R values in the SeV-infected groups than the control and PBS groups.

The total lung volume for each mouse from both ³He MR measurement and ¹H MR measurement is also shown in Table 1. Figure 6 confirmed the consistency of measured ³He MR and ¹H MR volume.

Fig. 6.

Linear regression of lung volumes measured by ³He-MR and ¹H-MR methods.

DISCUSSION

Noninvasive quantification of lung microstructure in mouse in vivo has been demonstrated using ³He diffusion MRI and validated by comparison to quantitative histology in our previous study (29). It can yield precise, quantitative information about acinar airways with tomographic information. In this study, we applied this technique to detect and quantify the in vivo lung microstructure changes well after initial SeV infection. Modest increases in alveolar duct radius distal to airway inflammation were measured by our imaging method, with some focal areas of greater increase in the lung periphery.

The SeV-infected mice show a similar increase in Lm via histology as via MRI compared with controls; however, there are several advantages to quantitative imaging. First and foremost, the invasive nature of histology prohibits its use in human longitudinal studies, whereas imaging can be performed serially and noninvasively, with results reflecting the in vivo lung. (Although serial imaging would have been possible in this study, animal regulations prevented reintroduction of infected mice into the animal facility. We have scanned many mice serially without incident.) Second, the precision of true stereological measurements requires significant work, careful specimen preparation, and an unbiased sampling scheme. In contrast, our imaging process takes less than 5 min per mouse from which whole lung morphometry with regional information is easily extracted. Furthermore, traditional quantitative histology is affected by tissue shrinkage. Whereas some methods suffer less than others (20), the estimation of shrinkage factor will significantly affect the final results. In our histological measurement, the shrinkage factor was estimated for each mouse by strictly following the tenets outlined in the ATS/ERS statement on the quantitative assessment of lung structure (13). Even a small estimation error of shrinkage factor will obscure the changes of Lm. This concern does not exist with the imaging method.

The estimated volume by ¹H and ³He MRI are tightly correlated, although the volumes measured by ¹H MRI are consistently smaller. This is because the ¹H MR image has four times the resolution of the ³He image. Overestimation of lung volume by ³He images may be understood by the fact that ³He pixels on the outer contour of the lung have larger size than ¹H pixels and are more likely counted as lung volume for partial-helium-containing pixels with high SNR. The estimated total lung volume can be used to examine whether there is subsequent airspace enlargement by lung expansion after SeV infection. The results from volume measurements showed that there is no connection between volume change and viral infection. Comparing Sendai 77 day mice with PBS 49 day mice, the average of R increased by 2.8%; if this were solely caused by volume expansion the average volume should have increased by 8.6% (V ∝ R³ for isotropic expansion). However, the actual average volume increase was only 2.5%. Although we cannot be precise about the relative amounts of destruction vs. volume expansion via these small increases, the volume changes serve as evidence that there is airspace enlargement not caused by volume expansion. A small limitation of this volume measurement was the fact that only five ¹H images were acquired; thus in a few cases, the most cranial or caudal parts of the lung were not included.

The increases in airspace size seen here are smaller than those reported for smoke exposed, C57BL/6J mice, although many of those increases are also quite modest (∼14%) (3, 6). An increase of Lm was also observed in other mouse strains with exposure to cigarette smoke, and the Lm change was shown to be strain dependent, with increases ranging from 0 to 38% (6, 10, 16, 17). Although the increase in Lm here in the SeV model (6%) is mild, the trend is clear and statistically significant. The SeV model holds some advantages as a model of chronic obstructive lung disease in that it shows some important physiological changes: chronic airways inflammation, airways hyperreactivity, and mucus cell hyperplasia, all of which are absent in the cigarette-smoking model (22, 26, 28).

The SeV mouse model is used as a model of chronic inflammatory lung disease that resembles asthma and COPD in humans, and previous studies have focused on airway inflammation. The chronic effects of the virus on expansion of distal lung parenchyma, however, have not been well studied. This study is the first report on the chronic effects of SeV infection on distal airspaces and the first study using imaging to quantify in vivo lung microstructure changes due to viral infection. It reveals a link between a small amount of acinar-airway dilatation and earlier virus infection. Moreover, our study reveals that SeV infection may indeed have the potential to become a valid murine model of COPD: there is significant pulmonary inflammation after SeV infection and subsequent progressive acinar-airway expansion.

Our technique is limited by a simplified lung model, that relies on a few reasonable assumptions. 1) We adopted the eight-alveolar model (eight alveoli around one alveolar duct) (9). Our method measures only two independent parameters (outer radius of acinar duct R and alveolar depth h). The measured changes in the alveolar length depend on R (30). This assumption holds well when alveoli expand proportionally in all directions. 2) We assume that during the diffusion time (2Δ = 880 μs, our shortest possible time given hardware restriction), most ³He atoms diffuse in a single acinar airway without escaping into adjoining airways. 3) We simplify the distributions of acinar-airway sizes; in reality, these parameters vary depending on the position and branching level of the acinar-airway tree. However, the ranges are relatively small: R = 70–100 μm, r = 40–75 μm, and h = 30–55 μm (21). We note that the precise mathematical model becomes invalid when tissue destruction become severe, although our previous study demonstrated that ³He morphometry still has good correlation with histology in patients with severe emphysema (30). The enlargement of acinar space in SeV models is modest, for which our technique is well suited.

In summary, we demonstrated successful in vivo monitoring of lung morphometry in mice with SeV, which has previously demonstrated chronic airway inflammation, particularly in the lung periphery. Increases in alveolar-duct radius R and mean linear intercept Lm, both measured by ³He diffusion MR images, imply that there is indeed a modest increase in alveolar-duct radius distal to airway inflammation, particularly in the lung periphery. Focal increases in R and Lm demonstrate mild tissue destruction 77 days after virus infection. The ability to noninvasively detect and quantify lung microstructure changes longitudinally brings an important new perspective to this potential mouse model of COPD.

GRANTS

This work was supported by the National Institutes of Health via grant numbers P50 HL084922, R01 HL090806, and R01 HL7003701.

DISCLOSURES

M. J. Holtzman is the principal investigator for research grants to Washington University from Hoffmann-La Roche and Forest Labs.

AUTHOR CONTRIBUTIONS

Author contributions: W.W., N.M.N., M.J.H., and J.C.W. conception and design of research; W.W., N.M.N., E.A., and J.C.W. performed experiments; W.W. and N.M.N. analyzed data; W.W. and N.M.N. interpreted results of experiments; W.W. prepared figures; W.W. drafted manuscript; W.W., N.M.N., and J.C.W. edited and revised manuscript; W.W., N.M.N., E.A., M.J.H., and J.C.W. approved final version of manuscript.