Mice. All animal procedures were performed in strict accordance with the National Institutes of Health (NIH) Guide for Care and Use of Laboratory Animals and approved by the Children’s Hospital Los Angeles Animal Care and Use Committee. A 6–8 weeks old male and female homozygous mice for the targeted mutation ROSA−mT/mG on a C57BL/6 background (Stock No. 007676) were purchased from Jackson Laboratory (Bar Harbor, ME, USA). ROSA−mT/mG is a ubiquitously expressed, cell membrane−targeted, two-color fluorescent Cre−reporter transgene16. In the absence of Cre recombinase, mTmG mice constitutively express membrane−targeted tdTomato red fluorescent protein. Following exposure to Cre recombinase, the tdTomato expression cassette is excised, and the rearranged mTmG transgene converts to an expression of a membrane−targeted GFP (enhanced green fluorescent protein). In this study, Cre mediated recombination was not used, and only the ubiquitously expressed membrane targeted tdTomato was imaged. The animals were housed in the animal facility allowed food and water ad libitum, and breeding was carried out to produce homozygote ROSA–mT/mG mice. The postnatal mice of age groups 7, 14, 21, and 28 days were used for further studies.
Inflated lung preparation. Inflated lungs were prepared from homozygous ROSA–mT/mG mice of all 4 different age groups. Anesthesia was induced via intraperitoneal injection of ketamine 100 mg/kg (Henry Schein Inc.) and xylazine 10 mg/kg (Akron Inc.) solution mixture. Once anesthetized, a cervical tracheotomy was performed and a 24G angiocatheter was positioned in the distal trachea and secured with a 3–0 silk tie. The mouse was then ventilated with the VentElite rodent ventilator (Harvard Apparatus) at 120 breaths/min, volume 400–500 µL, peak inspiratory pressure 21.5 cm H2O, and positive end−expiratory pressure 3 cm H2O.
Perfusion and fixation solutions were degassed under a vacuum for several hours before use. Access to the vasculature was achieved by surgical exposure of the abdominal viscera and insertion of a 24G safety intravenous catheter into the inferior vena cava. The descending aorta was severed, and blood was flushed from the vasculature with 100 ml saline solution (Baxter) containing 6U heparin (Sigma Aldrich) at 24 cm H2O. The breathing cycle was maintained during the flush. Once the perfusate ran clear, the breathing cycle was stopped, and tidal pressure was held at 20 cm H2O. Fixation was performed by perfusing 50–100 ml of a freshly prepared fixative solution containing 3% formaldehyde (Sigma Aldrich) prepared in 1X phosphate−buffered saline (PBS). After perfusion with fixative, the trachea was tied off with a silk tie to maintain lung inflation. The fixed, inflated lung with the attached heart was carefully removed from the chest cavity and submerged in the post–fix solution containing 2% MeOH–free formaldehyde (Polysciences Inc.) prepared in 1X PBS and allowed to post–fix for two days at 4°C. The fixative was removed by washing with twice daily buffer changes with 1 X PBS containing 0.02% NaN3 (Sigma Aldrich) for 4 days at 4°C. Then, the still−inflated lungs were transferred into a permeabilization solution mixture containing 1X PBS, 1.0% w/v Triton-X100 (Sigma Aldrich), and 0.02% w/v NaN3 (Sigma Aldrich) and stored submerged for several months at 4°C. The permeabilization solution was replaced with a fresh solution every month. As the lungs permeabilize, trapped air inside the lung tissue slowly escapes and is replaced by the buffer solution until the lung is no longer buoyant and sinks.
After samples were sufficiently permeabilized, the Triton–X100 was washed out by soaking the sample in a solution mixture of 1X PBS and 0.02% w/v NaN3 with daily buffer changes over a week at 4°C. The lungs were then dissected into their constituent lobes and stored at 4°C in a solution mixture of 1X PBS and 0.02% w/v NaN3. Mouse lungs prepared in this manner maintain their inflated morphology and exhibit robust fluorescent protein fluorescence for at least one–year storage.
Sample preparation for tomographic imaging. All infiltration and polymerization procedures were performed at room temperature and protected from light whenever possible to prevent photobleaching of the fluorescent proteins. Permeabilized lung samples were infiltrated by stepping through a step series of increasing concentrations of acrylamide and bis-acrylamide solution, 37.5:1 (BioRad) prepared in a 1X PBS solution containing TEMED [0.6 µL/ml] (BioRad). Acrylamide infiltration solutions were prepared fresh and degassed under vacuum for a minimum of 3 hours at room temperature before being added to the samples. Infiltrations were performed at room temperature in plastic 6 well dishes on a platform shaker with gentle shaking. Samples were equilibrated in each acrylamide concentration sequentially, 1X 4%, 1X 8%, and 2X 12% v/v, for 24 h before being transferred to a new well containing the next acrylamide solution. After infiltration, an embedding-polymerization solution consisting of 12% w/v acrylamide−PBS was prepared in 15 ml batches and degassed under vacuum for a minimum of 4 hours. Immediately before embedding, 70 µL of freshly prepared 10% w/v ammonium persulfate (BioRad) solution was added to the acrylamide solution and gently mixed to avoid excessive in−folding of air.
For casting the acrylamide, wells of glass spot plates (Pyrex) were filled with the embedding−polymerization solution and the lung samples were immediately transferred into the well and oriented. Note that gravity will draw the sample to the bottom of the well, and that will later be the top of the sample for imaging. The spot plate wells were then capped with 25 mm round cover glasses to seal out air which would impede polymerization. Absorbent paper towel tips were used to draw away excess solution and enhance the airtight seal of the cover glass onto the well opening. The remaining embedding−polymerization solution was placed aside in a sealed tube and served as an indicator of complete acrylamide polymerization. Once the acrylamide was fully polymerized, usually 2−3 h, the cover glass cap was lifted and the acrylamide hemisphere containing the embedded lung sample was liberated from the glass well and transferred to a solution of 1X PBS containing 0.02% w/v NaN3 (Sigma Aldrich) for storage.
Tomographic Imaging. Serial confocal tomographic imaging (referred to as Vibratome Assisted Sub-Surface Imaging Microscopy, Vibra-SSIM) was performed by coupling an upright confocal microscope to a modified vibrating microtome tissue sectioning device mounted adjacent to the microscope stage. The sectioning device assemblage consisted of the vibrating head, blade assembly, and knife holder from a salvaged Vibroslice Vibrating Microtome (Model #NVSLM, World Precision Instruments) that was mounted onto an XYZ stage (Newport Instruments) equipped with manual vernier micrometer actuators (Newport) on the X and Y axes for alignment. A CONEX-LTA-HS DC servo motorized actuator (Newport) controlled via CONEX-CC Controller software (Newport) was attached to the Z−axis and was used to control knife height. The sectioning device assemblage was mounted onto a 90° platform (Newport) and placed onto a damper rod with a mounting base (Newport). The entire sectioning device and post-assembly were bolted to the air table adjacent to the microscope stage. The knife vibration frequency was driven by voltage from a generic DC power supply. Single edge blades were purchased from Lafayette Instrument Co (Model 752/1/SS).
Before imaging, a sample embedded in an acrylamide hemisphere was glued with cyanoacrylate adhesive (Loctite 404) with its flat side down onto the bottom of a homemade acrylic tray. This try will serve as the sample bath. The sample bath was mounted to the motorized microscope stage. The sample bath was filled to above the top of the acrylamide embedded sample with room temperature 1X PBS that had been degassed under vacuum overnight. Imaging was performed using a Zeiss LSM710 confocal microscope on an Examiner upright stand and equipped with a fixed height motorized microscope stage (Prior Scientific). Zen Black software (Zeiss) was used to control the XY tiled, Z stack imaging. In addition to XY tiling, the motorized stage is also used to move the sample from the imaging location to the knife location for the removal of the sample thereby providing access to the next underlying region of the sample. The imaging session starts 10 microns below the upper surface of the acrylamide block. This ensured that the surface of the acrylamide block was never included in the resultant Z stack layer and any knife cutting induced distortions were avoided. The imaging-cutting sequence was standardized as follows: 60 microns Z stacks tiled in X and Y were acquired with the microscope. Then, the vibrating microtome was used to shave off the top 40 microns from the acrylamide block. Then, the microscope focus was lowered by 40 microns and another series of tiled 60 microns Z stacks was acquired. Then, the knife was lowered by 40 microns and another layer of the block was removed. Then the microscope focus was lowered by 40 microns and another series of tiled 60 microns Z stacks was acquired. The sequence was repeated until the desired extended Z regions were covered. This reiterative process generates consecutive, XY tiled, Z stack layers that overlap in Z with their next underlying XY tiled, Z stack layer.
All imaging was performed using a Zeiss 20x 1.0NA Plan-Apochromat water dipping objective. A 543 nm HeNe laser was used to excite tdTomato fluorescent protein and the emission was collected between 550−740 nm. Imaging was performed with Zen Black software (Zeiss). Images were collected in 16−bit depth. Tiled Z stacks were collected in 2×2 rectangular grids with 10% overlap between adjacent tiles. Auto Z brightness correction was used to mitigate depth−related decreases in signal intensity. A pinhole of 0.85 Airy units was used to collect 1.5 micron thick optical sections and the inter−section interval was 0.74 microns. Resultant Z stacks were saved in Zeiss CZI file format.
Post processing and extended Z stack compilation. The XY tiled Z stack layers were stitched together using either Zen Black or Zen Blue software (Zeiss). To compile extended image volumes, sequential composite image Z stacks were compared visually in the region of Z overlap between the two. Based on best correspondence in the Z overlap region, the bottom of the upper composite Z stack and top of the underlying stack was selected. Once the junctions in Z were determined for all of the sequential composite Z stacks, the optical sections were exported as 16−bit TIFF files and compiled into a larger stack of images that encompassed the entire extended volume. The sequential Z stacks often exhibited slight offsets in X and Y, most likely due to mild inaccuracies in the motorized stage movements. To correct this, the compiled larger Z stack was imported into FIJI software43 and processed with the Register Virtual Stack Slices44 plugin using the translation−only registration model with the shrinkage constraint option. After registration of the extended Z stacks, they were imported into Imaris 8.4.2 (Biltplane) or Amira 19.2 software for further analysis.
Image processing of confocal microscopy raw data. From the confocal microscopy raw data, the airspaces were extracted by first binarizing the tissue. The 16-bit images were binarized by using the Chan-Vese (CV) algorithm, a level set-based method17. A program was implemented in C + + using the semi-implicit Euler method, applying the Additive Operator Scheme to decrease calculation time45, 46. The parameters in the CV binarization were ε = 5/pi, µ = 0.00725125, λ = 200,000, and time step Δt = 2.0. These parameters were determined empirically by overlaying the output of the binarization procedure with the raw data in Imaris 8.4.2 software and validating the results, adjusting the parameters as required. The binarized volume was an output a stack of 16-bit TIFF images with each voxel of a value 65,535 or 0 representing tissue or airspace voxels, respectively.
The remaining steps were performed using MATLAB software. Since the binarization captured some of the noise presents in the volume, a 3D isotropic Gaussian Filter of standard deviation 3 was applied to the volume. After the Gaussian Filter was performed, the image was no longer binary, then a threshold was applied so that voxels below a certain threshold were set to 65,535 and the rest to 0.
Morphological closing was then performed on the images to remove connections between individual acini due to features of the lung such as Pores of Kohn and imaging artifacts such as light intensity variations. A 2-D disk of radius 11 was used as a structuring element on each image individually. Finally, the airspaces were extracted by running a connected component algorithm that grouped voxels so that voxels that shared a face (3-D 6-connectivity) belonged to the same set. Each connected component was then output as a separate stack of 16-bit TIFF images.
Spots creation in Imaris software to statistically analyze numbers and position of alveoli. The data from MATLAB pipeline was uploaded into Imaris 8.4.2 software. A connected component data volume was then created. A semi-transparent surface was created for the airspaces in the volume which show the alveoli. The volume rendering itself was then turned off to allow only the surfaces to be visible. After surfaces were created the spots feature was used to create easily visible spheres. The setting for spots depends on the size of the spaces in which we want to place the spots. The size that correctly positions one spot into each alveolus is diameter X, Y, and Z = 15 µm, Area = 225 µm2. This setting is specific to the consistently observed volume of the space in each alveolus throughout the connected components examined in this research. The spots (spheres) were then assigned colors that are visible in the alveoli. Then, alveoli sitting on cut surfaces of the connected components and therefore not completed were then easily visible and deleted to avoid measurement distortion. Imaris then counted and analyzed the number of alveoli as needed through statistical operations that are built−in Imaris. While rendering more than one connected component at a time, each component’s semi-transparent surface was given a contrasting color to its adjacent or if necessary, each set of spots in each component was also color adjusted for visual clarity (Fig. 3j – k).
Centerline Tree Determination in Amira. The data set was read into Amira 19.2 software and the centerline tree algorithm was applied unchanged. The algorithm was allowed to pick the root: setting − 1. The slope was varied between 0.5 and 5 as was the zero-value. The resulting centerline trees were visualized within the original semi-transparent volume. A specific value of these two parameters had the effect of clearly missing some small branches and end-nodes (alveoli) but exhibiting the correct scale to picking up most of the alveoli. Other values have an improper scale which results in fine surface details being labeled branch nodes with an excessive number of end-nodes.
This property was useful for visualization because one could prune smaller branches and alveoli as shown in Fig. 3b (slope = 3 and zero-value = 4). The value of the parameters to create the best balance between selecting every surface bump to be alveoli and missing smaller branches and end-nodes was determined to be a slope of 1.5 and a zero-value of 4 is shown in Figs. 3c – d. This set of parameters missed very few alveoli but over-determined their number and placed branching nodes close together. Hence the refinement is described below.
Generation Analysis of Lung Tree. The co-ordinates representing nodes and segments of the lung tree were obtained from Amira 19.2 software. These data served as input data to construct lung tree generation in MATLAB software. The lung tree originates from the root node is shown in red color in Fig. 3e. The MATLAB performed a merging operation based on the Euclidian distances and connections between nodes to eliminate the redundancy produced by Amira software. The root of the tree was then identified by the thickness analysis along all branches. From the root node, a growing process was adopted to reclusively check all branches to decide the generation number. All the nodes with no further connections are considered as the end nodes. These end nodes may partly indicate the position of alveoli.