The electron microscopic and stereological analyses were performed on samples of archived lung tissue originally prepared for the studies by Zeltner et al. (1987) (infants) and Gehr et al. (1978) (adults). Fixation and use of the lungs were done according to the bioethical regulations of the University of Bern at the respective time. The use of the archived material was approved by the ethics committee of Hannover Medical School (Permission No. 2263–2014). The subjects from which the samples were taken had died from non-pulmonary causes. Three infant (C1, 26 days; C2, 30 days; C3, 6 months) and three adult (A2, 40 years; A4, 20 years; A8, 39 years) lungs of the originally reported lungs were picked for this study.

After death, lungs were fixed by instillation via the airways using a fixative containing 2.5% glutaraldehyde in phosphate buffer. The samples used in this study were subsequently incubated with osmium tetroxide and uranyl acetate before dehydration and embedding in epoxy resin. For further details, see Zeltner and Burri (1987), Zeltner et al. (1987) and Gehr et al. (1978). Semi- and ultrathin sections (60–80 nm) were obtained from 6–7 epoxy resin-embedded samples per individual. Semithin sections for conventional light microscopy were stained with toluidine blue and ultrathin sections were stained with lead citrate/uranyl acetate.

For CLSM, lung samples either from tumour far tissue of tumour resection specimens or downsizing material during lung transplantation of three patients were obtained from the Institute of Pathology at Hannover Medical School (ethics vote number: 1741-2013 and 3381-2016). Tissue was fixed in 4% buffered formaldehyde directly after surgery and afterwards embedded in paraffin.

For estimation of stereological parameters well-established guidelines and recommendations were used (Hsia et al. 2010; Ochs and Mühlfeld 2013). The volume fraction of interalveolar septa was determined using light microscopic images and a point grid consisting of a coarse and a fine grid. Images were obtained using a Zeiss AxioScan Z.1 slide scanner (Zeiss, Göttingen, Germany) and analysed using the Visiopharm stereology software (Visiopharm, Hørsholm, Denmark). The coarse grid, consisting of four points, was used to count points hitting air in gas exchanging parts of the lung. The fine grid, consisting of 24 points, was used to count points hitting interalveolar septa. All fields of view were obtained by systematic uniform random sampling (SURS) at an objective lens magnification of 20 × (Gundersen and Jensen 1987). The volume fraction of interalveolar septa regarding the parenchyma [VV (sept/par)] was calculated by dividing the number of points hitting interalveolar septa by the number of points hitting interalveolar septa and air multiplied by six. The volume fractions of AE1 cells and their mitochondria were determined by TEM (Morgagni, FEI, Eindhoven, The Netherlands). A point grid consisting of a coarse and fine grid was projected on fields of view obtained by SURS at a primary magnification of 8900 × using the STEPanizer (Tschanz et al. 2011). The coarse grid consisted of nine points and was used to count points hitting AE1 cells and the interalveolar septum. The fine grid (144 points) was used to count points hitting AE1 cell mitochondria. The volume fraction of AE1 cells per unit volume of interalveolar septum [VV (AE1/sept)] was determined by dividing the number of points hitting AE1 cells by the number of points hitting the interalveolar septum. The volume fraction of mitochondria per unit volume of AE1 cells was determined by dividing the number of points hitting mitochondria by the number of points hitting AE1 cells multiplied by 16. To calculate the total volumes of interalveolar septa, AE1 cells and AE1 cell mitochondria, the fractions were multiplied by the respective reference volume. Therefore, the total volume of parenchyma was needed, which was taken from the data published by Zeltner et al. (1987). Also, the fractions of cross-sectional mitochondrial profiles Q (ger.: “Querschnitt”) in the compartments defined below were determined. For determination of Q no test system was used, but every profile was counted (Sterio 1984).

For subsequent distribution analysis a modification of a method published by Mühlfeld et al. (2007) was used. Two compartments of AE1 cells were defined: (1) AE1 cell portions on top of capillaries, i.e. in the main parts of the gas-exchange interface and (2) AE1 cell portions above connective tissue pillars, i.e. playing a minor role in gas exchange. In short, an estimator for a compartment size was compared to the number of mitochondrial profiles in this compartment. Therefore, a point grid with 25 points was used to determine the number of points P hitting either the first or the second compartment defined above. Also, the number of mitochondrial profiles (N0) in each of the two compartments was counted. With these data, the number of expected mitochondrial profiles (NE) in each compartment was calculated based on the assumption of random distribution. Subsequently, an index of relative localization (IRL) was determined by dividing N0 by NE as an expression for random (IRL≈1), preferential (IRL > 1) or non-preferential (IRL < 1) localization. Chi-squared analysis was used to test the null hypothesis that “the observed and expected distributions of mitochondrial profiles are equal”.

For qualitative evaluation and quantitative colocalization studies of mitochondria in 10-µm-thick paraffin sections of AE1 cells of three adult human lungs were immunolabelled with (1) advanced glycosylation end product-specific receptor (AGER) antibody, (2) translocase of outer mitochondrial membrane-20 (TOMM-20) antibody and (3) DAPI to visualize (a) AE1 cell plasma membrane, (b) mitochondria and (c) nuclei, respectively. Fehrenbach et al. showed that AGER antibodies are specific for the basolateral membrane of AE1 plasma membranes in rat and human lung tissue by immunohistochemistry, double immunofluorescence and immunoelectron microscopy (Fehrenbach et al. 1998). For recombinant monoclonal TOMM-20 antibodies we relied on the manufacturer's (Abcam, Cambridge, UK) statement concerning the specificity of recombinant antibodies. In short, tissue was deparaffinized in xylene with decreasing isopropyl alcohol proportions. Sections were washed with distilled water, incubated twice in Dako retrieval buffer pH 6.0 (Dako, Glostrup, Denmark) at 700 W seven mins, cooled on ice for 30 min, washed with distilled water and incubated again with 5% donkey serum (Dianova, Hamburg, Germany), 1% bovine serum albumin (BSA; Serva, Heidelberg, Germany) and 0.3% Triton X-100 (TX-100; Sigma-Aldrich, Steinheim, Germany) in PBS for 60 min at room temperature. After washing, sections were incubated with polyclonal goat anti-AGER antibody (Bio-Techne, Wiesbaden, Germany; order number: AF1145) diluted 1:100 in PBS containing 1% BSA and 0.3% TX-100 over night. Then, sections were incubated with Alexa647-linked donkey-anti-goat antibody diluted 1:1000 in PBS (Thermo Fisher Scientific, Waltham, MA, USA), 1% BSA and 0.3% TX-100 for 60 min. Afterwards, the sections were incubated with monoclonal rabbit anti-TOMM-20 antibody (Abcam, Cambridge, UK; order number: ab186735) diluted 1:250 in PBS containing 1% BSA, 0.3% TX-100 and 5% goat serum (Biozol, Eching, Germany). Subsequently, the sections were transferred into PBS containing 1:1000 Alexa488-linked goat-anti-rabbit antibody (Thermo Fisher Scientific, Waltham, MA, USA), 1% BSA and 0.3% TX-100. Before and after each individual antibody containing step, the sections were washed with PBS-Tween first for 5 and then for 10 min. Finally, the sections were transferred into PBS containing 1:1000 DAPI (Thermo Fisher Scientific, Waltham, MA, USA) for 15 min and washed with PBS first for 5 and then 10 min. Sections were embedded in Mowiol 4–88 (Carl Roth, Karlsruhe, Germany) and sealed with a cover slip.

Immunohistochemically stained specimens of human lungs were investigated using a Zeiss LSM 980 microscope (Zeiss, Jena, Germany). Three sites in each specimen were investigated by taking an overview image and consecutive imaging along the z-axis (z-stacks). Z-stacks were generated on parts of the section that contained interalveolar septa and visible mitochondria. Z-stacks were recorded at 63 × objective lens magnification with a crop factor of 8 or 16 (final size: 26.52 µm × 26.52 µm × 10 µm or 13.26 µm × 13.26 µm × 10 µm) to avoid areas without interalveolar septa. Parameters like pin hole and z-distance were optimized to obtain z-stacks of 10 µm thickness and maximal resolution. The z-stacks and the software Imaris (Imaris × 64, version 8.2.1, Bitplane, Belfast, UK) were used to generate 3D reconstructions. The 3D-derived reconstructions were confirmed by a colocalization analysis of AE1 cell mitochondria and AE1 cell membrane based on the signals of Alexa488-linked mitochondria and Alexa647-linked AE1 cell membrane. The colocalization analysis was processed with the JACoP plugin (Bolte and Cordelières 2006) for Fiji software (Schindelin et al. 2012). We assumed that, due to the highest resolution of CLSM of approximately 0.2 µm (Elliott 2020) and the thin cytoplasmic extensions of AE1 cells with a thickness of 0.2–0.1 µm (Weibel 1971; Dobbs et al. 2010), mitochondria inside AE1 cells are colocalized with the basolateral AE1 cell plasma membrane. As reproducible colocalization parameters, Pearson’s coefficient (PC) with Costes’ automatic threshold (Costes et al. 2004) and the local maximum in Van Steensel’s cross-correlation coefficient (CCF) distribution (van Steensel et al. 1996) were used. Three cases of parameters were evident: (1) negative PC and local CCF maximum far from zero, (2) positive PC and local CCF maximum (almost) zero and (3) positive PC and local CCF maximum far from zero, therefore, showing (1) no colocalization, (2) colocalization and (3) partial colocalization. Furthermore, we classified the value of PC with a classification of Evans by (1) no, (2) weak and (3) very weak (Evans 1996).