Imaging the extracellular matrix structure and remodeling in healthy, fibrotic, and emphysematous human precision-cut lung slices Authors: Yuqing Deng, Joseph K Hall, Yu Wang, Jae Hun Kim, Elizabeth Bartolák-Suki, Hadi T Nia, Holger Behrsing, Kenneth R Lutchen, Ramaswamy Krishnan, Béla Suki Published: 2025 ABSTRACT AbstractHuman precision-cut lung slices (hPCLS) have been used for studying extracellular matrix (ECM) remodeling in lung diseases; yet, quantitative assessment of disease-induced structural changes remains limited. We utilized multiphoton microscopy combining second harmonic generation (SHG), two-photon excited fluorescence (TPEF) and optical coherence tomography (OCT) to characterize ECM organization in hPCLS from healthy, pulmonary fibrosis (PF), and chronic obstructive pulmonary disease (COPD) donors. We applied quantitative metrics including fractal dimension, texture correlation analysis, fiber diameter and waviness measurements to assess ECM structural alterations. PF samples exhibited significantly increased fractal dimension and elevated collagen content, while COPD samples showed reduced fractal dimension and collagen content. Collagen fiber waviness, diameter and structure from texture analysis were significantly higher in diseased samples. Tissue stiffness, as a physiological readout measured from tensile tests, was significantly higher in PF than in healthy and COPD samples. The fractal dimension and collagen content estimated from standard histology showed a high correlation with collagen content (R2 = 0.92). Alveolar airspaces from OCT and histological images had slightly higher mean diameters from healthy samples than from PF and COPD samples, with diseased tissues showing increased heterogeneity characterized by collapsed alveoli clustered around enlarged airspaces. Elastase treatment of hPCLS significantly reduced texture correlation while waviness showed a similar trend, suggesting a possible role of elastic fibers in maintaining collagen network organization. Our comprehensive image and functional stiffness analyses enable quantitative assessment of ECM remodeling and offer valuable tools for monitoring structure-function alterations in lung disease.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12931-025-03464-7.Keywords: Second harmonic generation, Optical coherence tomography, Histology, Pulmonary fibrosis, Chronic obstructive pulmonary disease, Human precision-cut lung slices, Extracellular matrix remodeling Introduction Acute and chronic lung diseases represent a significant global health burden, causing millions of deaths annually [1, 2]. The pathology of lung disease is complex and heterogeneous, requiring physiologically relevant experimental models that reflect the complex cellular interactions and tissue architecture of the human lung. While the use of animal models is common, they are limited in their relevance to the human disease. Human precision-cut lung slices (hPCLS) have emerged as a powerful ex vivo model system that preserves the native cellular diversity, extracellular matrix (ECM) composition and structure, as well as mechanical properties of the human lung in vivo, increasing the efficiency of basic science and drug development [3–5]. In human lung diseases such as pulmonary fibrosis (PF) and chronic obstructive pulmonary disease (COPD), hPCLS allow for direct visualization and quantification of disease-specific structural remodeling, such as changes in ECM organization and composition [6, 7]. The ability to maintain viable human tissue in culture while also preserving the native tissue architecture makes hPCLS an invaluable tool for studying biological, structural and functional changes that lung diseases bring about. Second harmonic generation (SHG) microscopy is an advanced imaging tool capable of probing the structural organization of the lung ECM without requiring any exogenous labels [8, 9]. SHG takes advantage of the intrinsic optical properties of type I collagen to provide high-contrast, three-dimensional visualization of fibrillar collagen structures with submicron resolution. Combining SHG with two-photon excited fluorescence (TPEF) imaging of elastin enables simultaneous characterization of the two major load-bearing components of the lung ECM: collagen and elastic fibers [10]. Early applications demonstrated the utility of this approach by visualizing ECM remodeling and differentiating normal from fibrotic and COPD human lung tissues using quantitative SHG metrics [11–13]. Despite its high resolution, SHG microscopy has a limited field of view (FOV), constraining observations at the alveolar scale. This limitation can be overcome by utilizing other imaging techniques such as optical coherence tomography (OCT) or histology, which allow a substantially larger FOV for scanning the whole hPCLS sample. While these techniques have been applied to lung tissue imaging individually [8, 11, 13, 14], the combined advantages of these complementary imaging modalities and their quantitative analysis have not yet been applied to hPCLS. The aims of the present study were to develop and apply a comprehensive multimodal approach to characterize the structural differences between healthy, COPD, and PF hPCLS. By combining multiscale quantitative analysis of images from SHG, OCT, and histology with functional tissue stiffness measurements, we demonstrate how structural features at both the tissue and fiber scales vary in disease-specific patterns. We also applied this methodology to hPCLS before and after elastase treatment to show the feasibility of performing longitudinal studies of ECM remodeling without destructive processing. Materials and methods Human precision cut lung slices Deidentified and cryopreserved hPCLS were purchased from the Institute for In Vitro Sciences (Gaithersburg, MD). Following thawing, these slices retain biomass, viability, tissue integrity, and inflammatory markers [15]. Donor information is summarized in Table 1. Table 1Donor information and PCLS numbersConditionsagesex#hPCLSHealthy66male769male360male3Fibrotic67male165female762female3COPD67female467female6 Donor information and PCLS numbers Multiphoton microscopy (MPM) Multiphoton imaging was performed using a Bruker two-photon microscope equipped with a Nikon LWD 16× water immersion objective. The excitation wavelength was set to 880 nm, and images were acquired at a PMT gain of 950. The SHG signal from collagen structures was detected through a 460/50 nm bandpass filter, while the TPEF signal from elastic fibers was simultaneously collected through a 525/50 nm bandpass filter on a separate channel. Z-stack images were obtained through a depth of 48 μm with laser power controlled via a Pockels cell. For each hPCLS, 4–5 stacks were collected, resulting in at least 25 stack images per condition. Optical coherence tomography (OCT) imaging and alveolar airspace diameter measurements Decellularized hPCLS were imaged using a ThorLabs Telesto Series spectral-domain OCT system with a super luminescent diode light source (center wavelength: 1300 nm) and a 10× objective. A 5 × 5 mm field of view (FOV) image was collected for each hPCLS. Alveolar airspace areas were hand-traced and measured in FIJI, and equivalent diameters were computed from the areas assuming circular geometry. These equivalent diameters may also contain contributions from alveolar ducts and terminal bronchioles. Since the number weighted mean alveolar diameter in the healthy lung free of fixation artifacts was measured to be 325 μm [16], we performed a follow-up analysis where airspaces with equivalent diameters larger than 350 μm were excluded from the healthy group to remove potential contributions from terminal bronchioles and artifacts. This threshold was applied only to the healthy group to establish a reference value, while the diseased groups were analyzed without this threshold to avoid biasing comparisons due to their inherently heterogeneous airspace distributions. The effects of different threshold values were separately analyzed (see Supplementary Table S2). To compare airspace sizes between different samples, we also computed the area weighted mean diameters for each hPCLS as follows [17]: where is the equivalent airspace diameter and <…> indicates the arithmetic mean. The parameter D2 is sensitive to both the heterogeneity and the mean airspace size and hence it is suitable to characterize heterogenous structures in diseased lungs [17]. For the healthy group, values larger than 1.5 times the interquartile range of the dataset were excluded. Collagen fiber visualization Hematoxylin–Van Gieson (HVG) staining was used to visualize collagen in which collagen appears bright pink whereas other tissue components such as smooth muscle are yellowish to brownish and nuclei are blue to faint black. All staining components were from Electron Microscopy Sciences (Hatfield, PA) and the staining was performed using the manufacturer’s protocol. Briefly, 7 μm paraffin embedded sections were deparaffinized by washing 3 times for 5 min in xylenes and hydrating through 100-95-70% ethanol to distilled water, 3 times for 5 min each step. Weigert’s working hematoxylin step was applied for 10 min (Weigert’s Iron Hematoxylin A, #26044-05 and Weigert’s Iron Hematoxylin B, #26044-15). After 5 min wash in distilled water, the sections were stained for 3 min in Van Gieson’s Solution (#26046-05), dehydrated in 95% ethanol, 100% ethanol (2 times each), cleared twice with xylene and mounted with Permount (#17986-01). Microphotographs of the whole hPCLS area was captured using a slide scanner of a total of 15 hPCLS slides. Structural analysis The structure of collagen fibers in hPCLS was quantified using fractal analysis and texture analysis on the SHG channel. To reduce variation across Z-depth, the measurements were averaged across 8 scans in a Z-stack image. The fractal dimension was calculated using the box-counting method in MATLAB (Moisy, 2008; available at https://www.mathworks.com/matlabcentral/fileexchange/13063-boxcount). The algorithm overlays grids of varying box sizes (r) onto the binary image and counts the number of boxes (n) required to cover the structure of interest. The fractal dimension (D) was then calculated as the negative slope of a linear regression fit to the r-n data on a log-log graph. We determined the linear regression region on the log-log plot by selecting the range from 0 to at least 12.8 μm that yielded the highest R² value (Supplementary Fig. S1). The D can be considered as a measure of the self-similar fractal properties with higher values indicating a more complex and more space-filling collagen fiber structure. The correlation metric from MATLAB’s texture analysis was used to quantify local collagen fiber structure in the SHG images. The images were first analyzed using the Gray Level Co-occurrence Matrix (GLCM), which evaluates pixel pair relationships to extract key texture features. The correlation coefficient measures the linear pattern dependencies between neighboring pixels in an image. To visualize how structural features in an image create correlations, we created synthetic images with different correlation values. Higher correlation values increased the size of similar-valued regions (Supplementary Fig. S2). In the context of SHG imaging of lung tissue, this metric characterizes the “neighboring structure” around a fiber, i.e. how organized or structured the collagen fibers are within the tissue. Collagen content The main ECM components were quantified from signals from the SHG and TPEF channel signals. The pixel values of images from both channels were normalized to between 0 and 1 and binarized using a threshold value of 0.15. The total tissue content was first calculated as the sum of binarized signals from both channels (CH1 and CH2). Then, the relative SHG amount, a measure of collagen content, was determined by dividing the sum of pixels in the SHG channel by the total tissue content and expressed as a percentage. The collagen content data were further normalized by dividing with the mean of the healthy group and presented as amounts relative to the healthy group. Collagen fiber network and waviness The SHG channel was selected for collagen fiber imaging and analysis. A binarized image was obtained after applying the Frangi vesselness filter in MATLAB (Mathworks, Natick, MA) to enhance the fiber structures. The image was then skeletonized to obtain a fiber network (Fig. 3aii), which was then separated into discrete segments containing individual fiber-like structures (Fig. 3aiii). The contour length of each fiber segment was determined using the geodesic distance transform in MATLAB and the waviness of a fiber was calculated as the contour length divided by the end-to-end distance. Analysis of collagen fiber properties. a Representative grayscale SHG image (i) and its binarized version (ii) with measurement of fiber waviness (iii), calculated as the ratio of contour length (white) to end-to-end distance (blue line). (iv) Broken fibers can be found in COPD slices. b Fiber waviness increased in PF (1.08 ± 0.02) and COPD (1.08 ± 0.03) samples compared to healthy controls (1.07 ± 0.02). c Compared to healthy fiber diameters (n = 625, 5.3 ± 1.9 μm), collagen fibers had larger diameter values in both PF samples (n = 493, 5.7 ± 2.1 μm) and COPD samples (n = 383, 6.0 ± 1.9 μm). Fibers in COPD were also significantly thicker than in PF. Box plots are superimposed on estimated distributions. The separate distributions correspond to individual donors Collagen fiber diameter measurement Collagen fiber diameters were determined by manual analysis of images by an observer blinded to experimental groups. Line segments were drawn perpendicular to fiber orientation across different regions of each image (n ~ 20 lines per image). For each line segment, an intensity profile was generated showing signal intensity distribution along the line. Fiber width was determined by measuring the full width of each intensity peak represented by horizontal green lines on the profile plot (Supplementary Fig. S3). This measurement allowed the quantification of fiber diameter distributions while accounting for potential variations in signal intensity. Decellularization and elastin digestion Decellularization and elastin digestion were performed according to the protocol described in [18]. Briefly, hPCLS were decellularized by washing the samples in 2% Triton-X-100 buffer (10 mL/g tissue) at 4 °C for 24 h, with buffer replacement at 4 and 8 h. The samples were then transferred to a 0.1 M sodium chloride (NaCl) solution containing 0.1% sodium dodecyl sulfate (SDS) and agitated at 4 °C for 24 h to remove the remaining cellular structures. The SDS/NaCl buffer was replaced by deionized water and the tissue was shaken at 4 °C for another 24 h. Following SDS treatment, the tissue was washed in deionized water with agitation at 4 °C for 24 h. The decellularized tissue was then subjected to elastin degradation using porcine pancreatic elastase (Elastin Products, Owensville, MO, USA) at 60–80 units/g in phosphate-buffered saline (PBS) at 37 °C for 2 h. Stiffness determined from uniaxial stretching The stiffnesses of the hPCLS were measured using a uniaxial stretching device following the protocol previously established [6]. Briefly, hPCLS strips of ~ 1 mm in width and 3–8 mm in length were prepared using a blade. The actual thickness was measured from the OCT images. During stretching, one side of the sample was attached to a force transducer (model 400 A, Aurora Scientific, Ontario, Canada) using krazy glue. The other side was connected to a dual-mode lever arm system (model 300B, Aurora Scientific, Ontario, Canada), which generated prescribed displacement waveforms. A quasi-static tensile stretch was then applied to the strips using a 0.005-Hz triangular wave generated by the computer. The output force and displacement signals were converted to stress and strain by normalizing force with the cross area and displacement with the stress-free initial length. The slope of a regression line between stress and strain between 0 and 10% strain provided an estimate of the stiffness, which is equivalent to the Young’s modulus of the sample. Collagen percent and fractal dimension in histology staining Collagen content in histological images was quantified using custom-built color-based segmentation in MATLAB R2024a (Matheworks, Natick, MA). Airway regions were manually excluded from the analysis (Fig. 4c). Collagen fibers were segmented using hue-based thresholds combined with saturation and brightness criteria to identify pink-stained regions within the tissue boundaries. The collagen-to-tissue ratio (%Collagen) was calculated as the ratio of collagen-positive pixels to total tissue pixels within the analyzed region. Fractal dimension analysis was then performed on the resulting collagen masks using the box-counting method previously described for SHG signal analysis. Alveolar sizes from OCT images. a A representative OCT image from a healthy hPCLS showing alveolar structure with blue-shaded areas highlighting individual alveolar airspaces (inset). Gray color in the background shows agarose inside the hPCLS. b Alveolar diameters measured from OCT images (n = 321, 261, and 429 for healthy, PF, and COPD, respectively). The overlaid distributions correspond to individual donors. Alveolar structure from histological slide Images were first binarized followed by segmentation to individual airspaces using the watershed transform. This allowed labeling each airspace and measuring its area from which the equivalent diameter was determined. Larger airways, approximately 0.1 to 0.5% of the total airspaces, were manually excluded from the analysis. Since the mean alveolar diameter in healthy fixed lungs from standard histology was measured to be 225 μm [19], airspaces with equivalent diameters larger than 250 μm in the healthy group were excluded in a follow-up analysis. Statistics For normally distributed data, comparisons between fractal dimensions, collagen contents, and stiffness values of different groups were conducted using ANOVA with post-hoc Tukey HSD test while for not normally distributed data, a linear mixed-effects model (supplementary table. S1) was used in MATLAB (Mathworks, Natick, MA). The linear mixed-effects model analyzed measurements from all hPCLS from 7 donors across 3 conditions (COPD, healthy, and PF), with either hPCLS or donor as the random effect to account for individual differences. Outliers were removed using MATLAB’s isoutlier() function before running the tests. Comparisons between fiber waviness values, fiber diameters, and alveolar sizes were obtained using the Kolmogorov-Smirnov test after combining all the individual measurements from all images and donors in each group. To estimate the standard deviation (SD) of D2 from histological images, we used a bootstrap method. A P value < 0.05 was used for statistical significance. Results Tissue structure on MPM images In healthy tissue, collagen fibers were well-organized, that is, mostly single distinctive fibers were seen running aligned with the septal walls, and the alveoli were relatively uniform in size (Fig. 1a). In contrast, PF samples exhibited thickened and disorganized collagen fibers (Fig. 1b), while COPD samples showed enlarged airspaces often surrounded by many smaller alveoli with perhaps less collagen (Fig. 1c). Quantitative analysis revealed distinct differences among the three conditions. In the fractal analysis, the plot of the number of boxes against box size often deviated from a straight line for different box sizes (see Supplementary Fig. S1) indicating the possibility that the fractal dimension depended on the scale. PF samples had significantly higher fractal dimensions compared to both healthy (P < 10− 7) and COPD tissues (P < 10− 14), suggesting increased structural complexity and space-filling properties of collagen fibers (Fig. 2a). The correlations from texture analysis also showed significant differences between healthy and diseased hPCLS (Fig. 2b), with lower values in healthy than COPD (P = 0.007) and PF (P = 3 × 10⁻4) tissues. Representative MPM and OCT images of human precision-cut lung slices (hPCLS). Top row shows MPM images of (a) healthy hPCLS with well-organized collagen (cyan) and elastic (red) fibers within alveolar septal walls of relatively uniform sizes, (b) hPCLS from a lung with pulmonary fibrosis (PF) exhibiting thickened and disorganized collagen fibers, and (c) hPCLS from a lung with chronic obstructive pulmonary disease (COPD) displaying enlarged airspaces. Bottom row shows corresponding OCT images of the same tissue conditions. MPM images were acquired using simultaneous second harmonic generation (SHG, cyan) for collagen type I visualization and two-photon excited fluorescence (TPEF, red) for tissue autofluorescence. The SHG and TPEF signals often overlap, appearing white. Yellow arrows: collapsed alveoli. Scale bars = 200 μm (MPM) and 1 mm (OCT) Analysis of collagen structure and content in hPCLS. a Comparison of fractal dimensions showing significantly higher values in PF samples (n = 47, mean ± SD = 1.68 ± 0.04) compared to both healthy (n = 51, 1.62 ± 0.06) and COPD samples (n = 40, 1.60 ± 0.05), indicating increased structural complexity of collagen networks in fibrotic tissue. b Correlation coefficients from texture analysis with progressively higher values from healthy (0.946 ± 0.009) to IPF (0.953 ± 0.006), and to COPD samples (0.951 ± 0.01). c Normalized percentage of collagen content (SHG signal relative to signal from SHG and TPEF) revealing significantly elevated collagen in PF samples (1.05 ± 0.03) and reduced collagen signal in COPD samples (0.96 ± 0.10) compared to healthy controls (1.00 ± 0.04). Different symbols in (a), (b) and (c) represent individual datapoints from different donors. d No correlation (R2 = 0.1) was found between fractal dimension and normalized percent collagen across all tissue types. The correlation increased (R2 = 0.26) when the COPD data were excluded. Data on panels (a), (b) and (c) are presented as box plots showing the median (center line), the interquartile range (box), superimposed on individual values Quantitative analysis of collagen content from MPM images revealed significant differences among healthy, PF, and COPD samples (Fig. 2c). PF tissue demonstrated significantly increased normalized collagen content compared to healthy (P = 6 × 10− 5) and COPD (P = 3 × 10− 12) tissue. COPD samples demonstrated significantly reduced collagen signal intensity compared to healthy (P = 0.005) tissues. Samples from 2 donors in the COPD group generated a large variance in collagen content. Collagen content did not correlate with fractal dimension when all groups were analyzed together (R2overall = 0.1, Fig. 2d). However, limiting the analysis to PF and healthy groups revealed a moderate correlation (R² = 0.26, Fig. 2d). Collagen fiber structure We quantified fiber waviness and diameter in hPCLS (Fig. 3a). The analysis revealed that despite the large variabilities, both fiber waviness (Fig. 3b) and diameter (Fig. 3c) significantly increased in the diseased samples (Fig. 3). The COPD sample showed higher fiber diameters than healthy (P = 2 × 10− 6) and PF (P = 0.005) samples. Alveolar structure from OCT images Measured from OCT images (Fig. 4a), the alveolar airspace size in the PF tissue had a mean ± SD value of 267 ± 114 μm while the COPD tissue had a value of 269 ± 128 μm, both significantly lower (P = 7 × 10− 10 and P = 3 × 10− 9, respectively) than that in healthy hPCLS which had a value of 314 ± 132 μm (Fig. 4b). Using the interquartile range (IQR) method, 8.7% and 8.1% of the total data were detected as outliers in the IPF and COPD datasets, respectively, whereas in the healthy data, only 4.4% were identified as outliers. Shapiro-Wilk testing revealed that alveolar diameter measurements were not normally distributed (all P < 0.001). Log-transformation normalized the healthy group data (P = 0.07) but not the disease groups (IPF: P = 5 × 10− 5; COPD: P = 2 × 10− 4). Limiting airspace diameters to < 350 μm in the healthy group lowered the mean to 243.1 ± 58 μm. The corresponding D2 values in the PF and COPD groups were 368 ± 87 μm and 410 ± 88 μm, respective, both significantly different from those in the healthy group 272 ± 13 μm (P = 0.0003 and P = 0.005; Supplementary Fig. S4). Tissue structure on histological samples and stiffness Healthy and diseased tissue structures on histological images are compared in Fig. 5a. Alveolar airspace diameters show considerable variability (Fig. 5b) even in the healthy tissue. The diameter values in the healthy, PF and COPD groups were 164 ± 106 μm, 138 ± 89 μm, and 150 ± 77 μm, respectively. All three diameter distributions were statistically significantly different from each other. Limiting airspace diameters to < 250 μm in the healthy group lowered the mean to 125 ± 61 μm. The corresponding D2 values in the healthy, PF and COPD groups were 174 ± 7 μm, 256 ± 14 μm and 263 ± 15 μm, respectively. The D2 of the healthy group was significantly smaller than that of the COPD group (P = 0.03) whereas the difference between healthy and PF values almost reached statistical significance (P = 0.08) (Supplementary Fig. S5). Histological and biomechanical characterization hPCLS. a Stained sections from healthy (i), PF (ii), and COPD (iii) hPCLS. The inset in panel is (i) magnified (top) and shown using color-based segmentation (bottom) of pixels with collagen (cyan). b Airspace diameters (boxplots) and their distributions per individual donor show statistically significant differences among all conditions (healthy: n = 839; PF: n = 3404; COPD: n = 1845). c Correlation between percent collagen and fractal dimension (R² = 0.92). Dashed line: linear regression fit. Each point represents an individual histology image (healthy: blue circles, PF: yellow triangles, COPD: red squares). d Tissue stiffness measurements from the same donors. PF tissue (n = 34, 2.5 ± 1.4 kPa) demonstrated significantly increased stiffness compared to healthy (n = 34, 1.4 ± 0.7 kPa) and COPD (n = 9, 1.1 ± 0.6 kPa) The fractal dimension had a high correlation with collagen content (Fig. 5c). The scatter plot in Fig. 5c also shows that PF datapoints predominantly appeared in the upper right region, indicating elevated fractal dimension and collagen content in the PF group. Although COPD samples demonstrated lower mean values for both collagen percentage and fractal dimension compared to healthy samples, the histological images did not exhibit typical airspace enlargement characteristic of COPD for samples from this donor (Fig. 5aiii). This observation was also reflected in the stiffness measurements (Fig. 5d), because there was no significant difference between healthy and COPD samples, despite the mean being 21% lower in the latter. However, PF samples exhibited significantly higher stiffness than both healthy (P = 0.008) and COPD (P = 0.023) samples. This finding is consistent with the dense tissue architecture observed in the corresponding histological image (Fig. 5aii). Structural changes in the collagen networks due to elastin digestion We investigated the effects of elastase treatment on hPCLS. After elastase digestion, there was a reduction in the TPEF signal (Fig. 6a), which primarily originated from elastic fibers [11]. The absence of elastic fibers led to a significant decrease (P = 0.02) in the texture-based correlation value evaluated for the collagen fiber structure (Fig. 6b). The fractal dimension did not change whereas fiber waviness tended to decrease after the elastic fibers were removed, but neither reached statistical significance due probably to limited sample size. Effects of elastase treatment on collagen network structure in healthy hPCLS. a Representative MPM images showing collagen (cyan) and elastic (red) fiber structure before (left) and after (right) elastase treatment. b Texture correlation analysis showed a significant reduction in correlation values after elastase treatment (P = 0.02). c Fractal dimension of the collagen network did not change following elastase digestion. d Fiber waviness showed a strong tendency to decrease after digestion but the difference between the groups were not statistically significant. Data in (b), (c) and (d) are presented as box plots showing the median (center line), the interquartile range (box), and the range of outliers (whiskers). Individual outliers are shown as circles. Waviness values in panel (d) were obtained from the same hPCLS but not from the same fibers Discussion In this study, we have demonstrated the utility of MPM, OCT, and histology combined with quantitative image analysis for characterizing structural alterations that were observed in hPCLS from healthy, PF and COPD donors. Quantitative metrics (e.g., fractal dimension, collagen content, fiber structure, waviness and alveolar size) revealed significant differences between healthy and diseased hPCLS. Specifically, we observed higher fractal dimension, local collagen texture correlation and content in PF samples. In COPD samples, the fractal dimension, a measure of structural complexity and space-filling capacity, and image-based collagen content decreased but the local texture correlation remained higher than in healthy hPCLS. Furthermore, the fiber waviness and diameter were higher in diseased hPCLS. Histological analysis demonstrated only mild airspace enlargement in the COPD donor, which was consistent with the results of mechanical testing demonstrating a 21% drop that did not reach significant difference between COPD and healthy samples which may explain the contrasting results with our recent study [6]. The dense tissue observed in histological images of PF samples correlated with significantly higher stiffness values. Using color segmentation, a strong correlation was also observed between collagen fractal dimension and collagen content in histological images. To probe the interaction between elastin and collagen fibers, we quantified the structure of elastase-treated hPCLS and found that enzymatic digestion significantly altered the local collagen structure as evidenced by a reduced correlation from texture analysis. Analysis of the box-count plots (Supplementary Fig. S1) revealed non-linear behavior, indicating scale-dependent fractal dimensions. We focused our analysis on the small-scale regime, which corresponds to sizes smaller than the length of a single alveolar wall segment and maximum observed fiber diameter (Fig. 3c), thus capturing fractal and correlation properties at the fiber level. In this regime, PF samples showed significantly increased fractal dimensions compared to healthy tissue, while COPD samples showed a lower value. This pattern paralleled our observations of collagen content, with changes showing opposite directions in PF and COPD samples. Interestingly, no correlation was seen between fractal dimension and % collagen; however, a stronger positive relationship was observed when excluding the COPD results (Fig. 2d), indicating that the destructive COPD affects the organization of the collagen network structure. In contrast to MPM images, there was a high correlation between fractal dimension and collagen content in the histological images with R² = 0.92 (Fig. 4c). This may indicate that collagen fiber structure was altered during fixation and drying during the process of histological slide preparation. This is supported by the fact that alveolar airspace sizes derived from histological images were significantly smaller than those from OCT images (Fig. 5b vs. Fig. 4b). While previous studies have reported increased collagen content in COPD tissue [20], we found that despite significant regional variation, it decreased significantly in COPD samples (Fig. 2c) albeit with increased fiber diameters (Fig. 3c), indicative of massive collagen remodeling. This is further supported by the fact that both PF and COPD samples exhibited increased texture correlation compared to healthy tissue (Fig. 2b). Texture correlation measures the linear dependency of gray levels between neighboring pixels along fibers with lower correlation values indicating greater variation between adjacent pixels, resulting in smaller regions of similar values (see Supplementary Fig. S2). Therefore, the lower correlation values observed in healthy samples suggest a more pronounced collagen micro-structure which tends to disappear with tissue remodeling in both diseases. To further investigate how remodeling affects network complexity at a larger scale, we also quantified the number of nodes at fiber intersections in skeletonized 2D MPM images (Supplementary Fig. S7). However, no significant correlations were observed between the number of nodes and the fractal dimension or the correlation values. The differences in collagen network structure likely manifest in different mechanical properties in healthy and diseased lung tissue [21]. Indeed, tissue stiffness was higher in PF samples than in healthy or COPD samples (Fig. 5d). The effects of this network level complexity on stiffness were likely outweighed by mechanical rupture (Fig. 3a iv) that resulted in reduced stiffness in COPD compared to PF. Multiphoton microscopy enabled simultaneous visualization of elastic fibers (TPEF) and collagen fibers (SHG), allowing quantitative assessment of ECM composition. Our analysis revealed significantly higher collagen content in PF hPCLS compared to both COPD and healthy samples (Fig. 2c). This result is consistent with findings that reported increased collagen signal in SHG imaging [22] and excess deposition of elastic fibers [23]. Tilbury et al. also reported an increase in the collagen/elastin ratio in fixed PF tissues compared to healthy tissue using the same MPM imaging technique [12]. COPD samples also showed differences in ECM composition, exhibiting a slightly reduced collagen signal intensity (Fig. 1a vs. 1c). While previous studies have reported increased elastin and collagen deposition in COPD [24], our findings of reduced SHG signal intensity likely reflect changes in collagen fibrillar organization and composition of collagen types rather than changes in total content. While types I and III are the main collagens in lung tissue, type I collagen forms thicker fibrils that generates stronger SHG signals [25]. During COPD progression, studies have reported enhanced collagen degradation alongside deposition of immature and disorganized type III collagen and increased type III to type I ratio [24, 26–29]. These factors could contribute to reduced SHG intensity as well as single fiber mechanical properties [18]. Our analysis revealed a significant difference in fiber waviness between healthy and PF as well as healthy and COPD samples (Fig. 3b). Histological slides of fixed human lung tissue demonstrated an age-related decrease in collagen fiber waviness together with an increase in fiber thickness [30]. We hypothesized that similar fiber-level alterations might occur during fibrotic disease progression. Using hPCLS, which preserve native ECM structure, we found the opposite in PF: fiber diameters, texture correlation as well as waviness increased in PF. Although fiber thickening occurs in both aging and PF, the PF tissue showed higher waviness compared to healthy controls, whereas aged lung tissue typically exhibits straighter fibers than young tissue [30]. This suggests excessive deposition of disorganized collagen molecules and micro-fibrils in PF. Hence, age-related increases in collagen crosslinking [31] appear to affect fiber waviness differently than the excessive deposition of immature collagen observed in disease states, possibly due to different cross-link types in aging and PF. Indeed, while upregulated lysyl oxidases govern collagen crosslinking in human idiopathic PF [32], enzymatic crosslinking is gradually replaced by non-enzymatic ageing-related increase in crosslinking by advanced glycation endproducts [33]. Additionally, the stiffer tissue in PF compared to healthy likely resulted in a smaller lung volume at the generation of the hPCLS which is expected to increase waviness [22]. Fiber rupture in diseased lungs can also influence waviness, as demonstrated in a COPD sample (Fig. 3a iv), where broken and retracted fibers in enlarged airspaces display very high waviness at their untethered ends. Although correlations between fiber waviness and fractal dimension have been reported in arterial and tendon tissues [34, 35], the lack of correlations in the current study (Supplementary Fig. S6) may be due to the fact that fractal dimension is a global measure whereas waviness of a single fiber is a local measure. Nevertheless, it is interesting to note that a local unexpectedly high waviness as in Fig. 3a iv can be used to detect septal wall rupture in COPD which should correlate with the progressive nature of the disease. Chow et al. reported a decrease of collagen fiber waviness when digesting arterial ECM using elastase [34]. With limited sample size, we found a significantly decreased texture correlation reflecting local collagen structure and a tendency of fiber waviness to decrease, albeit not significantly, after elastase digestion (Fig. 6b and d, respectively). The decreased correlation values indicate that elastase alters the small-scale collagen fiber structure. Interestingly, this change occurred in the opposite direction to that observed in COPD lung tissue, despite both conditions involving elevated elastase levels [36]. This suggests that elastase alone cannot fully recreate the ECM alterations characteristic of COPD, which likely requires additional factors such as mechanical forces driving wall rupture and collagenase activity [18, 37, 38]. Nevertheless, the correlation results in Fig. 6b highlight the important role of elastin in maintaining collagen fiber organization in lung tissue. The OCT images provide a large FOV with potential of capturing enlarged airspace in COPD and honeycombing structures in PF (Fig. 1) [39, 40]. The average alveolar airspace diameter from OCT images in healthy hPCLS was 314 ± 132 μm, which is higher than previous findings [41], indicating perhaps a high inflation volume during hPCLS preparation. However, this value is in good agreement with a more recent study that reported a mean diameter of 325 μm [16]. The alveolar diameter distribution in COPD and PF revealed many small alveoli. The coefficient of variation (SD/mean × 100%) for COPD (47%) and PF (43%) samples were higher than that for healthy samples (41%), indicating increased alveolar airspace size heterogeneity in disease. This distribution aligns with observations in PF human lung explants, for which collapsed airspaces caused by wall thickening as well as overinflation of adjacent healthy regions were reported [42, 43]. The wide distribution of alveolar diameters in COPD samples demonstrates a greater variation in alveolar size, due to the characteristic airspace enlargement in COPD [44] that results from mechanical force-induced fiber and septal wall ruptures [45] as also demonstrated in Fig. 3a iv. Airspace enlargement also leads to collapse of alveoli surrounding the enlarged area. Indeed, while only 6% of healthy airspaces had a diameter < 150 μm, this increased to 8% in IPF and 12% in in COPD samples. Our method has some limitations. One arises from the inherent constraints of tissue slicing: unlike microCT, our approach cannot reconstruct the full 3D architecture, and the analysis is confined to the cutting plane orientation. In addition, there may be a bias in slice selection for studies to avoid hPCLS with small tissue content which may have contributed to our findings that the COPD samples represented a mild disease case. There may also be some shrinkage in the direction perpendicular to the plane of the hPCLS. Furthermore, given that lung pathology varies across different regions [46], and the precise location of our tissue samples within the lung was not known, correlating our findings with clinical severity may present challenges. The preparation of hPCLS also introduces variables that could affect tissue mechanics, as different agarose concentrations and inflation pressures during sample preparation may influence alveolar structure including waviness as well as tissue mechanics [6]. Lastly, the relatively small number of donors also limits a robust statistical analysis across disease stages. In conclusion, this study characterized ECM remodeling in hPCLS from healthy, COPD and PF donors using a combination of method including MPM, OCT, mechanical testing and histology. The results obtained from live, decellularized and fixed imaging across multiple spatial scales, from individual fiber properties to network-level features in human lung samples, provide new insight into ECM and fiber remodeling in the two main diseases of lung ECM. Therefore, the tools presented here have the potential to be utilized in the future in various drug discovery and mechanistic studies using living human lung slices. Supplementary Information Supplementary Material 1. Acknowledgements Imaging was performed in the Boston University Micro and Nano Imaging Facility and Precision Measurement Laboratory. Authors’ contributions YD designed studies, carried out experiments and wrote paper, JKH analyzed images, YW analyzed images, JHK analyzed data, EB carried out histologic staining, HTN wrote paper, HB provided tissue and wrote paper, KRL helped with statistics and wrote paper, RK wrote paper, BS designed studies and wrote paper. Funding Part of this study was supported by NIH 1R43 HL176301 and NIH-DP2HL168562. Data availability The datasets supporting the conclusions of this article are) included within the article and the Supplement. Declarations Competing interests R.K. and B.S. are co-founders and J.H.K., R.K and B.S. are equity holders of Mechanobiologix LLC. H.B is a founder of Advanced In Vitro, LLC. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The datasets supporting the conclusions of this article are) included within the article and the Supplement.