Quantification of increased MUC5AC expression in airway mucus of smoker using an automated image‐based approach

Microscopic analysis of mucus quantity and composition is crucial in research and diagnostics on muco‐obstructive diseases. Currently used image‐based methods are unable to extract concrete numeric values of mucosal proteins, especially on the expression of the key mucosal proteins MUC5AC and MUC5B. Since their levels increase under pathologic conditions such as extensive exposure to cigarette smoke, it is imperative to quantify them to improve treatment strategies of pulmonary diseases. This study presents a simple, image‐based, and high‐processing computational method that allows determining the ratio of MUC5AC and MUC5B within the overall airway mucus while providing information on their spatial distribution. The presented pipeline was optimized for automated downstream analysis using a combination of bright field and immunofluorescence imaging suitable for tracheal and bronchial tissue samples, and air–liquid interface (ALI) cell cultures. To validate our approach, we compared tracheal tissue and ALI cell cultures of isolated primary normal human bronchial epithelial cells derived from smokers and nonsmokers. Our data indicated 18‐fold higher levels of MUC5AC in submucosal glands of smokers covering about 8% of mucosal areas compared to <1% in nonsmoking individuals, confirming results of previous studies. We further identified a subpopulation of nonsmokers with slightly elevated glandular MUC5AC levels suggesting moderate exposure to second‐hand smoke or fine particulate air pollution. Overall, this study demonstrates a novel, user‐friendly and freely available tool for digital pathology and the analysis of therapeutic interventions tested in ALI cell cultures.


| INTRODUCTION
Computational-assisted microscopy advanced remarkably in recent years covering numerous biological and medical research fields (Eggerschwiler, Canepa, Pape, Casanova, & Cinelli, 2019;Tollemar et al., 2018;Vonbrunn et al., 2020). However, there is hardly any literature available on automated image processing to equally facilitate evaluation of mucosal specimen. Especially determining the expression levels of the major mucosal proteins MUC5AC and MUC5B is imperative to better assess pathologic mucosal conditions, which is of significant interest in both research and diagnostics.
While physiologic amounts of airway mucus are essential to lubricate the respiratory epithelium (REp), increased synthesis and density of airway mucus is a major cause for muco-obstructive pulmonary diseases and develops gradually upon exposure to airborne particles (Fahy & Dickey, 2010;Rubin, 2010;Tian & Wen, 2015). Cigarette smoke is a major source for harmful particles and severely elevates MUC5AC levels in both REp and submucosal glands (SMGs; Goldfarbmuren et al., 2020;Kesimer et al., 2017). However, a physiologic ratio of MUC5AC to MUC5B is a prerequisite for effective respiratory clearance (Hattrup & Gendler, 2008;Rose & Voynow, 2006;Roy et al., 2014). While MUC5B is produced abundantly in both REp and SMG, MUC5AC is primarily expressed in superficial goblet cells and found in SMG only under pathologic conditions (Okuda et al., 2019;Ostedgaard et al., 2017;Widdicombe & Wine, 2015).
Techniques to analyze mucus composition and properties are rare, require expertise, or special equipment, are often error-prone in their respective collection methods and fail to provide information on spatial expression patterns (Atanasova & Reznikov, 2019). Imagebased protocols to visualize and score the overall mucus level commonly employ periodic-acid Schiff reaction (PAS) or alcian blue (AlB) staining that requires manual manipulation and counting of target cells, is labor intensive and yields data in an "all-or-nothing" principle rather than intermediate values on staining intensity and ratio of mucosal proteins (Aterman & Norkin, 1963;Gregory et al., 2010;Patel et al., 2018;Sakamoto et al., 2000). Microscopy-based computational evaluation of mucosal specimen could facilitate quantitative analysis while enhancing reproducibility and minimizing inter-and intraobserver variability in interpreting histological images (Osborne et al., 2014;van Diest et al., 1997).
In this study, we present a simple and automated computational approach that is based on standard histological processing and the open-source software Cell Profiler (Carpenter et al., 2006;Kamentsky et al., 2011). In spite of the software's versatility, it is still often compromised in recognizing integral structures for example, blood vessels or glandular tubes in mucosal specimen. We addressed this issue by adapting established histologic and fluorescent staining techniques, which allowed our developed Cell Profiler pipeline to specifically target individual mucosal areas and assess number, ratio, and distribution of major airway mucins. The pipeline was validated by comparing mucus levels in tracheal tissue and air-liquid interface cell cultures of primary normal human bronchial epithelial cells (NHBE) derived from smoking and nonsmoking individuals.

| Specimen collection
Specimen from the distal trachea were collected in accordance with the Austrian law BGBl. 1 Nr. 108/2012 and approved by the ethical committee (Medical University of Graz [MU Graz], EK30-377ex17/18).
All samples were obtained from organ donors from the Division of Transplantation Surgery or during autopsies from the Institute of Pathology at the MU Graz. Samples were grouped in either smokers (no medical history in pulmonary diseases, current smoking) or "normal" subjects (neither medical history of pulmonary diseases nor records of smoking).
The mean age of donors was 72 ± 9.7 years for the normal (five male and one female) and 63.2 ± 12.5 years for smoking cohort (one male and five female).

| NHBE isolation and ALI culture
The specimen were washed with 0.9% NaCl/3% penicillinstreptomycin to remove blood and mucus plugs, mechanically separated from soft tissue or lung parenchyma and kept in incubation medium (IM; MEM, 1Â MycoZAP, 40 μg/ml tobramycin, 500 μg/ml DTT, 10 μg/ml DNase in PBS) for 4 hr at 4 C before incubation in digestion medium (DM) containing IM supplemented with collagenase solution (185 U/ml collagenase type II, 2 mg/ml BSA, 0.5 mM calcium chloride, 1 U amino acids 100Â, 5% FCS in DMEM) and 10 μg/ml DNase over night at 25 C. The epithelial cells were scrapped off the luminal side of the trachea four to five times into fresh MEM, spun at 200 g for 10 min and incubated in MEM supplemented with 5Â Antibiotic-Antimycotic (Gibco) for 2 hr at 37 C. The cells were then resuspended in PneumaCult™ ExPlus medium (Stemcell, Vancouver, Canada) supplemented with 1Â MycoZAP (Lonza, Basel, Switzerland) and seeded in gelatine-coated cell culture flasks. The cells were harvested at a confluency of 80-90% and seeded into fibronectincoated transwell inserts (VWR, Dublin, Ireland), at a density of 3 Â 10 5 cells per 12-well insert, and incubated with both apically and basally supplied medium for 3 days as suggested by the manufacturer.
The apical media was removed and PneumaCult™ ALI medium (Stemcell) was henceforth supplied in the basal compartment only.
The medium was changed every other day.

| Tissue processing
To adequately preserve airway mucus, both tracheal tissue and ALI cultures were fixed using Carnoy's reagent (ethanol, chloroform, glacial acetic acid; 6:3:1, v/v, [Puchtler, Waldrop, Conner, & Terry, 1968]) or 3.7% paraformaldehyde for 24 hr or 20 min, respectively. Dehydration and tissue processing for embedding in paraffin was conducted automatically using the Excelsior™ AS Tissue Processor (ThermoFisher Scientific). Paraffin-embedded blocks were cut to 5 μm sections using a rotary microtome (HM355 with STS & Cool-Cut, ThermoFisher Scientific, Waltham, MA, USA). The sections were mounted on SuperFrost Plus™ slides, heat-dried at 54 C for 2 hr and 48 C overnight, dewaxed using Histolab Clear ® (Histoslab ® , Askim, Sweden) for four times 5 min each and then rehydrated in a graded series of ethanol (100, 96, 70, and 50%) followed by washing in distilled water three times for 3 min each. Antigen retrieval was performed using 0.1 M sodium citrate buffer solution (pH 6) for 15 min at 93 C using a KOS Microwave Multifunctional Tissue Processor (Milestone, Sorisole, Italy).
Nuclei were counterstained using 4 0 ,6-diamidino-2-phenylindole (DAPI, 1:2000, ThermoFisher Scientific). Combined mouse and rabbit IgG (5 μg/ml, diluted in antibody diluent, Agilent Technologies) was used as negative control. The slides were dried using an ascending graded series of ethanol at 70, 90, and 100% for 3 min each and mounted with ProLong™ Gold Antifade Reagent (ThermoFisher Scientific). Image sets that did not cover any mucosal areas, showed inadequate preservation of morphology or contained both SMG and REp in tracheal specimen were excluded to avoid skewing data on spatial expression of mucosal proteins between SMG and REp. The final analysis included 28-98 image sets in SMG, 12-70 image sets in REp and 16-120 image sets in ALI cultures (all 200Â magnification). Representative images are shown in both bright field and IF (Figures 1 and 8).

| Computational analysis of mucosal parameter
Graphical data analysis was performed using the open-source software Cell Profiler (version 3.1.5.; (Carpenter et al., 2006;Kamentsky et al., 2011)). A "short" and an "extended" version of the pipeline were developed within this study and uploaded for public access (Figures 2 and 3). The pipelines are designed to first identify the AlB positive area (AlB + ) in bright field images to frame the general mucosal area within which the AlB staining intensity as well as the Cy3 (MUC5AC) and Cy5 (MUC5B) fluorescence intensities will be assessed as mean pixel intensities (henceforth referred to as mean intensity). This process calculates the mucin quantity within the total AlB + area also engrafting regions without mucin content therefore normalizing the mucin signal intensity to the overall mucosal area, which is why combined bright field and IF staining is imperative. The

| Short pipeline
The original AlB and IF images were loaded into Cell Profiler and matched as intended using "NamesAndTypes." To identify the AlB + areas as primary objects, the images were split into the individual RGB grayscale channels using the module "ColorToGray" (#1) and processed to distinguish the AlB + area by subtracting the RGB red (x 1) and RGB green (Â 0.1) image from the RGB blue image using the module "ImageMath" (#2). This step generates a negative of RGB red, in which AlB is mainly pronounced while excluding features of the surrounding tissue that are also shown in RBG green. Hence, only areas marked by AlB are emphasized (Figure 2a-e).
The module "IdentifyPrimaryObjects" (#3) subsequently identifies the highlighted area (AlB_calc) using an intensity threshold. Using "Over-layObjects" (#4) allows manually verifying the correct identification of the AlB + area, which represents the mucosal structures (Figure 2f-h). To accurately measure the fluorescence intensities of MUC5AC and MUC5B within the AlB + area, the background fluorescence derived from the FITC channel was corrected by using the module "ImageMath" (#5/ #7), which subtracts the FITC image from the Cy3 and Cy5 IF images

| Extended pipeline
In case the mucin intensities within the MUC5AC + and MUC5B + areas are of interest as a measure of local mucin expression, the "extended" version of the pipeline may be used. Here, the mucin signals of the background-corrected IF images (steps #5 and #7 of the "short" pipeline) are identified as primary objects ("IdentifyPrimaryObjects"; #9, #11) which may be manually verified by "OverlayObjects" (#10, #12, Figure 3d,h). Finally, the module "MeasureObjectIntensity" (#9/#13) allows calculating the area and staining intensities while "SaveImages" (#10/#14) and "ExportToSpreadsheet" (#11/#15) are added to export the resulting images and data. These three modules may be modified as required. The pipeline listed as modules in Cell Profiler is shown in Figure S3.
F I G U R E 1 Workflow and representative images for automated evaluation of airway mucus. (a) Schematic workflow for sample preparation and establishment of air-liquid interface cultures for automated analysis of airway mucus using the proposed microscopy-based approach. Representative bright field and IF images of Carnoy-fixed tracheal tissue derived directly from the same slide of (b) a normal individual and (c) a smoker. Image sets constitute 170-590 single 200Â images per bright field, DAPI (blue, nuclei), FITC (green, background only), Cy3 (red, MUC5AC), and Cy5 (yellow, MUC5B) channel. Enlarged images show representative regions of SMG and REp further illustrating spatial distribution of the mucosal proteins

| Validation of pipelines
The pipelines were validated manually for the accurate and complete identification of the individual mucosal areas by classifying the detected primary objects either as "correctly detected" (complete detection of AlB + area), "partly detected" (incomplete detection of AlB + area), "false positive" (excessive area detected outside AlB + area) or "false negative" (AlB + area not detected).
F I G U R E 2 Stepwise illustration of the presented "short" image analysis pipeline. The individual modules that constitute the pipeline are numbered #1-#11. The AlB stained bright field images (a) are split into RGB grayscale images (b-d) to allow off-set calculations (AlB_calc) (e) and threshold-based identification of the AlB + area (f,g). Subsequently, overlaying the identified objects onto the original bright field image allows manually verifying correct detection of the mucosal area (h). The original MUC5AC and MUC5B images (i) are background corrected by subtracting the FITC channel (j-l and n,o) and overlay the primary object identified from the bright field image onto the background-corrected IF images to verify the positioning of the mucin signals within mucosal areas (m,p). Modules #9-11 encompass steps to measure, save, and extract data The ratio is given in percent of mucosal structures present within each individual image set (N = 12) including areas counted as false positive (= 100%). The AlB + area in ALI cultures was correctly detected in 100% of images and was therefore not further evaluated.

| Data analysis and statistics
GraphPad Prism version 8.0.2 (GraphPad Software, http://www. graphpad.com) was used for statistical analysis and data representation. Data are shown as means ± standard deviation (SD). Data were tested for normality using Kolmogorov-Smirnov test followed by two-tailed Student's test for normally distributed data or one-way ANOVA with Tukey post-hoc test if >2 groups were compared.
Kruskal-Wallis with Dunn's multiple comparison testing was used for data not following Gaussian distribution and small sample sizes (N <6).

| RESULTS AND DISCUSSION
In order to visualize and evaluate the complete mucosal area while detecting the individual mucosal proteins MUC5AC and MUC5B directly on the same slide, we combined general staining methods on mucus for bright field imaging with targeted staining of mucosal proteins in IF. Downstream computational analysis of such color-coded individual targets allows direct comparability of staining intensities and fractions of airway mucins within the mucosal area. The entire workflow and timeline is schemed in Figure 1a.

| Sample preparation and staining conditions
Unmitigated fixation of airway mucus is a prerequisite for downstream mucosal analysis. While 3.7% paraformaldehyde did not properly retain airway mucus especially on ALI cell cultures ( Figure S1), we found the mucus integrity well preserved in Carnoy-fixed samples and proceeded using only the latter. We initially attempted to combine a general mucosal staining such as PAS or AlB with IF (PAS-AlB-IF) and F I G U R E 3 Representation of the "extended" pipeline to evaluate localized parameter of airway mucins. The background-corrected IF images from the "short" pipeline (a,e) are used to identify areas covered with MUC5B and MUC5AC as primary objects (b,c and f,g), which is verified manually by overlaying these objects onto the input images (d,h). Modules #13-15 represent modules #9-11 from the short pipeline encompassing steps to measure, save, and extract data therefore tested the specificity of the primary antibodies for MUC5AC and MUC5B in immunohistochemistry in comparison to PAS-AlB-IF.
However, PAS-AlB-IF did not return any staining of mucosal proteins (data not shown). This is most likely due to alterations of the epitopes during the oxidation reaction by periodic acid, which might hinder antibody binding. In contrast, linking AlB staining with IF yielded proper staining of MUC5AC and MUC5B at primary antibody dilutions of 1:100 (final concentrations 2 μg/ml for MUC5AC and 1 μg/ml for MUC5B, Figure S2b Perhaps the different antibody binding capacity following PAS or AlB staining is due to the lesser degree of epitope modification under applied conditions (concentration, incubation time, acidity) by acetic acid (AlB staining; 3%, 10 min, pK a = 4.75), than by periodic acid (PAS reaction; 2%, 60 min, pK a 1 = 1.64). The exact causality of the dissimilar antibody binding affinity, however, remains to be determined empirically.

| Detection of mucosal areas
Cell Profiler pipelines were developed to allow discerning the target areas based on their color and staining intensity using the same manually optimized global threshold across various samples (Figures 2 and   3). It was therefore necessary to validate the process of identifying mucosal structures in both effectiveness and accuracy. Our Cell Profiler pipelines detected the mucosal area based on AlB staining with an overall precision of 83% for correctly or partly detected SMG and in 81% of REp (Figure 4a,b). Since the areas that contributed to false results were generally small or showed a low level of staining intensity potentially caused by peripherally cut SMG, downstream analysis would only be affected to a minor degree. Identifying AlB + areas in ALI cultures was accurate across the image set and thus not further validated.

| AlB + area and mean intensity in SMG and REp
The verified pipelines were applied to the image data sets derived from nonsmoking and smoking individuals (Figure 2). Both the AlB + area and the AlB mean intensity within the SMG and REp were similar between nonsmokers and smokers (N = 6 per group; N = 5 for REp in the normal group, Table 1), allowing direct comparison of downstream calculations on mucin content within AlB + areas. As expected, the AlB + areas were generally larger in SMG than in the REp, while the REp showed stronger AlB mean intensities (Table 1, Figure S4).
These results nicely reflect the correlation between the optical densities of the basic AlB dye that binds to acidic glycosaminoglycans in the form of muco-polysaccharides in these different locations. Mucus is condensed before secretion leading to more intense stains in goblet cells primarily present in the REp (Fahy & Dickey, 2010).

| MUC5AC levels in SMG multiply 18-fold in smokers
MUC5AC expression levels of smokers were significantly increased within AlB + areas in the SMG compared to nonsmoking individuals ( Figure 5a), which was not the case in the REp (Figure 5b). Similarly, overall MUC5B levels surpassed MUC5AC in the SMG but not REp, impartial of smoking (Figure 5a In contrast to other studies, we found only a minor effect on MUC5B levels. It is reported in literature that both reduced  and elevated levels (Kesimer et al., 2017) were found to be a consequence of cigarette smoking. A reason for this variation may be the differing methodologies and location, as these studies measured MUC5B levels using transcriptomics in the tracheal epithelium and mass spectrometry of sputum, respectively, rendering direct comparison difficult.
Since the pipeline enabled the quantification of both mucins from within exactly the same area, we could use this data to directly assess their ratio and calculated a significant 16-fold increase of MUC5AC to MUC5B levels within the SMG of smokers (Figure 5e). Our study indicated MUC5B levels to remain rather stable upon exposure to cigarette smoke, which suggests that the increased ratio is almost entirely attributed to elevated MUC5AC expression. Increased levels of MUC5AC and MUC5AC/MUC5B ratio have also been reported in other mucosal pathologies such as asthma, opening the proposed methodology to its application in a wide range of respiratory diseases beyond smoking .  3.5 | MUC5AC elevation increases mucin area rather than density In addition, the "extended" pipeline allowed assessing whether the increased MUC5AC levels in the SMG of smokers attributes to either the expansion of MUC5AC + areas or the local increase in MUC5AC signal intensity by interrogating the spatial distribution and the staining intensity of the confined area that the mucosal proteins actually inhabits.
MUC5B was abundantly present and homogeneously distributed throughout the SMG and REp image sets as well as across all investigated patient cohorts (Table 2). MUC5AC, on the contrary, showed strong heterogeneous distribution across images, the area it occupies within the SMG, and between smokers and nonsmokers (Table 2).
Precisely, we found MUC5AC in the SMG in 55% of images from smokers compared to 14% in three out of six normal individuals that occupied 8% and <1% of the AlB + area, respectively. No difference was seen in the REp image sets (Table 2).
Further, we used the "extended" pipeline to test whether the staining intensity of the mucins within the respective mucin + areas increase locally upon elevated expression. We found that the localized signal intensity for both mucins was rather stable in smokers compared to nonsmokers in both the SMG and REp (Figure 5c,d).
Together, these results suggest that increased MUC5AC levels in smokers (Figures 1b and 5a) primarily expand the area occupied by the mucin rather than thickening its local density.

| Partial MUC5AC elevation in normal donors
As mentioned before, we identified MUC5AC in the SMG in three out of six normal donors. This was surprising since the general understanding of previous studies is that MUC5AC expression in the SMG occurs only in smokers and does not take place there for normal nonsmokers (Okuda et al., 2019;Widdicombe & Wine, 2015). However, the detected MUC5AC + areas were clearly distinguished from the background and targeted toward AlB + cells in the SMG (Figure 6a).
We therefore rated these signals as MUC5AC specific and subgrouped the donors as either normal 5ACÀ or normal 5AC+ for downstream analysis.
While MUC5AC levels per AlB + area were significantly distinct comparing normal 5ACÀ and normal 5AC+ donors to smokers individually, we found no statistic difference when comparing normal 5ACÀ to normal 5AC+ specimen, warranting to merge the subgroups for analysis (Figure 6b,c).
Nevertheless, our results prompted an intense literature search, which indeed revealed a certain inconsistency regarding MUC5AC in the SMG of normal individuals because some studies also reported sparse amounts of MUC5AC in the SMG of healthy donors to occupy about 2% of the total mucosal area (Caramori et al., 2009;Inoue et al., 2008). This percentage is even higher than the <1% observed here.
Although these differences might be attributed to the genetic makeup of donors or the used methodology during analysis, our findings reflect the inhomogeneous distribution of MUC5AC in SMG as it was found in only 14% of total images in only three out of six donors and might have been missed if only single images were analyzed. We therefore strongly recommend comprehensive analysis comprising wide areas of the specimen over single images, as done in this study. Additionally, these variations highlight the sensitivity of the pipeline as its threshold was optimized to also detect weak signals of MUC5AC.

| Positive correlation of MUC5B and MUC5AC expression in smokers and nonsmokers
Since the expression of both MUC5AC and MUC5B is reported to be induced by cigarette smoke Kesimer T A B L E 2 Percentage of mucin+ images per data set and mucin+ area within the total AlB+ area given in means ± SD in normal donors compared to smokers
Image analysis with the proposed pipeline confirmed these findings reporting both the percentage of images with MUC5AC + areas and the percentage of MUC5AC + area within the total AlB + area to be significantly increased in the REp ( F I G U R E 7 Scatter plot and correlation analysis computing a potential association between MUC5AC and MUC5B expression levels in SMG in both normal and smoking individuals using a Spearman correlation that is given as "r" while "p" reports the respective probability value 3.9 | Automated mucus analysis in 3D ALI cell cultures While the determination of mucosal compositions in tracheal tissue is of great value in digital pathology and retrospective analysis of clinical study cohorts, tissue samples themselves are severely limited when used to develop and test new therapeutics. To fill that gap, controllable 3D organoid models such as the ALI cell culture system are increasingly applied in studies concerning host-pathogen interactions (Caves et al., 2018;Marrazzo et al., 2016), drug efficiency (Min et al., 2013) or mucus production (Castellani, Di Gioia, Di Toma, & Conese, 2018) under various conditions (Miller & Spence, 2017), while minimizing the use of animal models in basic research (Derakhshanfar et al., 2018;Pfeiffer et al., 2014;Sosa-Hernández et al., 2018).
In order to demonstrate the broad applicability of our developed method, we applied the proposed pipelines to reconstituted ALI cell cultures of primary NHBE derived from both smokers and nonsmokers to evaluate potential mucus-modulating differences resulting from exposure to cigarette smoke prior to isolation and ALI establishment (Figure 8a,b).
We found no difference in the AlB + area when normalized to the membrane length but a trend toward increased AlB intensity in ALIs  (Table 3). However, both MUC5AC and MUC5B signal intensities within the AlB + area as well as the area of total mucus covered by the mucin were comparable in ALI derived from smokers and normal individuals (Table 3, Figure 8c-e). MUC5B levels surpassed MUC5AC both per AlB + area and mucin + area independent of cellular origin (Figure 8c).
We can therefore suggest that NHBE cells preconditioned to cigarette smoke normalize their mucin expression patterns when allowed to differentiate over the course of 3 weeks devoid of cigarette smoke.
While these are important implications for studies using ALI cell cultures in disease models, it was to be expected given that NHBE cells are isolated from the REp of the trachea and may thus only be compared to the REp and not SMG.

| General discussion
Overall, our developed methodology proved generally robust as both pipelines returned comparable values on target identification and staining intensities across different specimen of tracheal tissue and ALI cell cultures and allowed to quantify major mucosal parameters using only standard histological staining.
Although the data are derived from 2D images and cannot be related directly to the actual fraction of mucosal solids as derived by bronchoscopy or bronchoalveolar lavage and western blotting (Atanasova & Reznikov, 2019), the fold change on increased MUC5AC levels found in our study was very similar to those previously reported.
Still, the pipeline shows difficulties for example in distinguishing granular and secreted mucus when analyzing surface epithelial structures or ALI cell cultures. Secretion causes hydration and hence swelling of airway mucus, which in turn, enlarges the AlB + area while reducing the staining intensity (Fahy & Dickey, 2010). Consequently, the overall values on mucosal protein levels in the REp and ALI cultures is lower when the entire AlB + area is analyzed. This limitation may be addressed by adjusting the threshold of mucus detection or generating image masks that accurately define the target area.
Abnormalities in mucus quantity and protein content are not only a phenomenon of lung pathologies, but also common in the gastrointestinal and the reproductive tract (Davis, 2006;Rowe, Miller, & Sorscher, 2005;Viniol & Vogelmeier, 2018). Because mucopolysaccharides are also contained within mucus of these tissues, it might be possible to target them with the proposed pipeline. Furthermore, the primary antibodies used within the proposed method are interchangeable, widening the scope of experimental settings or therapeutic interventions when used on in vitro models such as ALI cell cultures (Castellani et al., 2018;Upadhyay & Palmberg, 2018) or animal models (Mercel et al., 2020). We commend the reported pipelines as a valuable and promising tool for quantitative image-based analysis to be used to facilitate research on muco-obstructive diseases.

| CONCLUSION
We developed Cell Profiler pipelines that enabled the automatic quan- Therefore, the proposed pipelines provide an innovative, robust, and high-processing computational tool to quantify so far only histologically visualizable objects. Since these pipelines are freely available online and user-friendly, they expand the computational toolbox on image-based analysis in research fields focused on muco-obstructive conditions.

DATA AVAILABILITY STATEMENT
The data that supports the findings of this study are available from the corresponding author on reasonable request. The pipelines and template data sets are freely available at https://cellprofiler.org/examples/published_pipelines.