ABSTRACT
Many light-microscopy protocols have in recent years been published for visualization of ultrastructure in the kidney. These protocols present researchers with new tools to evaluate both foot process anatomy and effacement, as well as protein distributions in foot processes, the slit diaphragm and in the GBM. However, these protocols either involve the application of different complicated super-resolution microscopes or lengthy sample preparation protocols. We here present a fast and simple, 5-hour long procedure for the full three-dimensional visualization of podocyte foot processes using conventional confocal microscopy. The protocol combines and optimizes different optical clearing and tissue expansion concepts to produce a mild swelling, sufficient for resolving foot processes using a diffraction-limited confocal microscope. We further show that the protocol can be used for common visualization of large-scale renal histology, pathology, and kidney protein distributions. Thus, we believe that our fast and simple protocol can be beneficial for routine conventional microscopic evaluation of kidney tissue integrity both in research and in the clinic.
INTRODUCTION
In recent years, a range of novel optical imaging and sample preparation protocols have been presented both by our group and others, handing kidney researchers and pathologists the opportunity to use light microscopy for the analysis of renal ultrastructure1–8, reviewed in9,10. The power of these new techniques as tools for quantifying renal ultrastructure is exemplified in recent publications from our group and others3,11–13. Importantly, super-resolution STED microscopy together with tissue clearing was a crucial tool for establishing the newly proposed model for albumin filtration referred to as the GBM compression model11,14. Although powerful, all of these optical protocols require the use of advanced diffraction-unlimited microscopy techniques, such as stimulated emission depletion (STED) microscopy5, single-molecule localization microscopy (SMLM)4 or structured illumination microscopy (SIM)2,3, or involve the introduction of complex polymer chemistry into the sample to clear and/or physically expand it1,6,7. Many of the protocols are also very tedious in duration and require a substantial amount of hands-on labor. This extra layer of complexity might be what is keeping these new methods from being routinely used by both researchers and clinical pathologists. Further, many of the protocols lack access to deep 3D imaging of foot processes (FPs) beyond 5-15 µm in depth1–3,7,8. Therefore, we set out to develop a drastically simplified and fast all-optical protocol, which clears and spatially swells kidney tissue samples sufficiently to allow for 3D high resolution analysis of renal anatomy and pathology using a conventional diffraction-limited microscope available in almost all life-science and pathology institutes.
Previous protocol simplification have shown that the introduction of acrylamide polymers into PFA-fixed samples can be totally omitted, while still preserving tissue integrity upon SDS clearing15. Further, another publication shows that tissue samples can be expanded more than a factor of two using a simple polymer-free protocol based on immersion in different aqueous solutions16. On the basis of this, we removed acrylamide infusion from our more tedious and complex published protocols5,6, and changed the sample embedding medium to allow for a mild (1.3-fold linear) swelling of kidney tissue samples. By also optimizing imaging parameters, we achieved high enough effective resolution for three-dimensional visualization of podocyte FPs and the GBM using a conventional confocal microscope.
RESULTS
Optical clearing and mild swelling of kidney tissue
Kidney samples were treated according to our simple clearing/swelling protocol, schematically depicted in Fig. 1a. Vibratome sectioning of kidney samples and biopsies is not absolutely necessary, but is helpful for later staining (shorter incubation times), mounting (easier to mount flat samples) and imaging (more sample volume accessible close to objective). It should however be stressed that the agarose embedding and Vibratome sectioning takes maximum 15 minutes for a biopsy-sized sample (∼3-4 slices per full biopsy) and is thus neither a complex nor time-consuming addition to the protocol. If no Vibratome is available, samples can also be cut manually with a razor blade, although it is difficult to cut samples thinner than around 0.5 mm using this method. Of note is that the SDS de-lipidation step was not primarily added for optical transparency, but to ensure sufficient swelling of samples16. Further, if omitting the SDS de-lipidation, we show that staining quality becomes poor (Fig. S1a), as has also been shown previously by our group5. Moreover, to minimize the time needed for sufficient de-lipidation, the temperature was raised compared to our original optical clearing protocol. De-lipidation of the kidney tissue at 70°C for 1 hour produced good results without sample degradation occurring at even higher temperatures (Fig. S1b). Furthermore, as shown by Murakami et al.16 urea and urea-like molecules can effectively swell tissue samples resulting in an increase in effective spatial resolution. 8 M urea embedding indeed resulted in a linear swelling of 1.3 times (data not shown) but was not compatible with the lectin staining we wanted to use for the visualization of both glomerular filtration barrier structures and large-scale histology. We thus lowered the urea concentration to 4 M in a mixture with 80% wt/wt fructose (dubbed FRUIT)17, which preserved a linear swelling factor of about 1.3 (Fig. 1b) while still allowing lectin labelling (later see Fig. 2-5). In early versions of the simpler and faster protocol, staining of specific proteins (nephrin/podocin/collagenIV) was carried out using primary and secondary antibodies which had to be incubated 2 hours each to allow for sufficient penetration. To shorten this time, amine-reactive fluorophores were directly conjugated to primary antibodies using NHS chemistry. Initially, using the primary conjugated antibodies in a standard staining buffer (PBST) the signal and labelling efficiency was not high enough for outlining individua foot processes. However, by diluting antibodies in an alternative staining buffer called Scale-CUBIC-1 (10 mM HEPES + 10% Triton-X), which has been reported to increase the signal around 2-fold in immunostaining applications18, staining quality was increased to a satisfactory degree (Fig. S1c). Thus, the protocol was further simplified and the duration could be decreased by 2 hours, resulting in a protocol taking only 5 hours from collection of the tissue to acquired image. To compare this protocol to known protocols for visualizing podocyte foot processes, several essential parameters were evaluated (Fig. 1c-f, boldface black font = conventional microscopy protocol, gray font = super-resolution or electron microscopy protocol, our protocol highlighted with a square). Evaluated parameters include number of protocol steps, number of reagents used, hands-on time and total protocol duration. Compared to previously published protocols using conventional microscopy, our fast and simple protocol outperformed them on all evaluated parameters. The protocol presented by Pullman et al.2 is slightly faster in total, but with the drawbacks that the protocol requires super-resolution SIM microscopy, more hands-on labor, more sample interactions and number of reagents. Further, the later protocol is performed on thin cryo-sections, thus lacking access to large in situ imaging depths. Of note is that our protocol is less complex than standard paraffin-embedding or cryo-sectioning when all evaluated parameters are considered.
Optimization and validation of our protocol for resolving foot processes
Because our fast and simple protocol gives only a moderate swelling factor of 1.3, confocal imaging parameters had to be optimized to achieve sufficient resolving power for imaging foot processes. Mouse kidney samples were treated according to the protocol and stained for nephrin using different fluorophores. As is common knowledge, the resolution of a confocal microscope depends both on the pinhole size and the wavelength of the excitation/emission light19. We thus used a smaller pinhole diameter of 0.3 airy units (AU) together with a green-emitting dye Alexa 488, and in this way achieved high enough resolution to resolve foot processes in a wild-type mouse kidney (Fig. 2a). Thus, whenever mouse foot processes are to be visualized, fluorophores with an excitation wavelength of maximum 488 nm (e.g. Alexa-488, Alexa-405, CF405M) are used. With these optimized imaging conditions, we were able to visualize whole mouse glomeruli in 3D, at an effective resolution of around 115 nm, well enough to resolve individual foot processes (Fig. 2b-e). Moreover, we show that with the two-hour long primary labelled antibody incubation step, a penetration depth of at least 70 µm is achieved (Fig. S2a-c). Using a glycerol objective with a correction collar, this gives access to larger z-depths than any of the previously developed protocols, with the exception being our clearing/STED protocol5. Since the resolution of a confocal microscope is highest in the xy-dimension, capillaries must preferentially be oriented in this imaging plane for optimally resolving foot processes. This imaging condition holds true for all the other optical protocols mentioned in this article as well. Keeping the glomerular capillaries in their 3D in situ setting allows us to screen through the whole glomerulus. We show that in every glomerulus, there are around 0.2 (human) to 0.3 (mouse) xy-oriented areas per µm in z (Fig. S2d). This implies that we can with our protocol find plenty (around 10 large field-of-view areas), allowing us to quantify and easily follow also subtle variations in renal nanoscale structures11,14. This shows the importance of having access to the depth-dimension when performing high-throughput quantitative analyses of glomerular health. Importantly, we also validated the protocol for also resolving foot processes in human patient material (Fig. 2e), which is less challenging as compared to mouse, due to the larger dimensions of human FP’s11,13. Using wheat germ agglutinin (WGA) lectin with affinity towards the glycocalyx of FP’s, we were moreover able to resolve foot processes also in a cross-sectional “transmission-EM-like” manner, in both mouse and human samples (Fig. 2f-g). Additionally, we co-stained for nephrin, with the expected localization in the slit membrane, showing that we can get data on protein localization in addition to morphological information. Taken together, we have demonstrated that our simple clearing and swelling protocol can in just a few hours provide three-dimensional information regarding renal ultrastructure and protein expression at a resolution previously only accessible with more complex protocols.
Ultrastructure Pathology in Mouse Samples
As previously shown2,3,5,6,11,12, “en face” foot process imaging of the SD can readily be utilized to visualize and quantify foot process effacement. We validate that this rationale can be applied also with our new protocol by using a recently published mouse model for FSGS, with two compound heterozygous point mutations in the podocin gene11 (Fig. 3). By staining for nephrin, the SD pattern could be visualized in three dimensions in whole glomeruli of WT and mutated mice (Fig. 3a-b). As previously shown3,11, the SD length per area can be semi-automatically quantified using ImageJ macros, and applying this morphometric analysis gave a result consistent with what has previously been published for this mouse line11 (Fig. 3c). Further, glomerular capillaries were imaged in cross-section showing both effacement of foot processes as well as alterations to the GBM in the mouse model for FSGS (Fig. 3d-e). The GBM thickness was semi-automatically measured, showing a significant increase in mutated mice modelling FSGS disease progression (Fig. 3f).
Ultrastructure Pathology in Human Samples
To further validate our protocol for possible clinical use, we applied it to human tissue from patients diagnosed with different types of kidney disease. Two patients with congenital nephrotic syndrome (CNS), either with a homozygous mutation in the NPHS1 gene (encoding for nephrin), or trans-associated mutations in the TRPC6 and NPHS2 genes (encoding for the TRPC6 channel and podocin). We also applied the protocol to two patients diagnosed with focal segmental glomerulosclerosis (FSGS) and minimal change disease (MCD). Using our protocol, we could indeed visualize and quantify typical hallmarks of kidney disease such as foot process effacement as well as pathologies in the GBM (Fig. 4). Images of renal ultrastructure could be acquired both “en face” (Fig 4a-e) and from the side (Fig. 4g-k). We could moreover observe the expected absence of nephrin in slit diaphragms of the patient with a homozygous mutation in the NPHS1 gene20 (Fig. 4b,h). As for mouse samples (Fig. 3), filtration barrier morphology could be quantitatively evaluated by measuring the slit diaphragm length per area and the GBM thickness using semi-automatic ImageJ macros (Fig. 4(f, l), S3(a-b)). Of note, expected GBM thickening is not observed in any of the diagnosed patients as compared to control (as observed for mice, Fig. 3f). The explanation likely lies in the fact that GBM thickness increases with age21. The nephrectomized control kidney was taken from an older patient while all the other samples are from infants or young children, so GBM thickening would probably be observed if compared to age-match control samples.
Large-Scale Pathology in Mouse and Human Samples
Since not only ultrastructural, but also histological changes to renal morphology is of high interest both in research and in the clinic, we investigated the possibility of our protocol to assess large-scale morphology using fluorescence-based histological analysis. We show that using a basic 10X/0.45NA air objective, we could detect large scale pathologies such as focal partial or total loss of nephrin (Fig. S4a), global loss of nephrin in glomeruli from the patient with an NPHS1 mutation (Fig. S4a), different degrees of interstitial fibrosis (Fig. S4b), glomerulosclerosis (Fig. S4c), focal thickening of Bowman’s capsule (Fig. S4c), atherosclerosis (Fig. S4d), loss of brush border in tubules (Fig. S4e) and dilatation of tubular lumen (Fig. S4d). Also in the mouse model for FSGS, we were able to detect both segmental sclerosis of glomeruli as well as tubular alterations (Fig S5).
Visualizing Large- and Ultra-Scale Alterations in Patients Diagnosed with Focal Segmental Glomerulosclerosis (FSGS) and IgA nephropathy
Moreover, to further exemplify the possible use of the protocol in patient samples, we investigated if we could detect pathology features important for diagnosing a case of focal segmental glomerulosclerosis (FSGS) of the cellular type (Fig. 5). We show the visualization of most pathology features mentioned in the clinical scoring statement from the pathologist, such as focal total/segmental sclerosis, tubular atrophy and endothelial hypercellularity in glomerular capillaries when comparing to present gold standard methods such as optical histology and electron microscopy (Fig. 5a-j). Both qualitatively and quantitatively (Fig. 5k-l, S3) we see that our fast and simple protocol provides comparable diagnostic information as standard pathology methods presently used for renal pathology. To further investigate the diagnostic possibilities of our protocol, we also applied it to a biopsy from a patient diagnosed with IgA nephropathy (IgAN) (Fig. 6) and were able to confirm the visualization of many diagnostic features. The biopsy showed no severe pathologies in the filtration barrier as shown both by EM and our protocol (Fig. 6a-b) and a comparative analysis resulted in no obvious differences between EM and our protocol (Fig. 6c-d). Next, the histology of the IgAN kidney was evaluated both with standard histology (Fig. 6e) and “fluorescence histology” (Fig. 6f). We could visualize the alterations (or lack of alterations) mentioned in the statement from the pathologist such as interstitial infiltration of cells (Fig. 6g-h), normal glomerular capillaries (Fig. 6i) and mesangial expansion and hypercellularity (Fig. 6j).Visualization of immune deposits is critical for diagnosing IgAN and other immune complex diseases. Therefore, we stained the IgAN biopsy using an IgA antibody and were able to clearly localize the deposits both on the large scale (Fig. 6k-l) and in detail (Fig. 6m). The staining pattern was equivalent to that of standard immunofluorescence (IF) (Fig. 6n-o). By imaging foot processes “en face”, partial mild effacement could also be visualized (Fig. 6p) and quantified (Fig. 6q). Even though we only got access these two patients where comparisons to standard pathology was available, the results indicate that parallel evaluations using our protocol and standard pathology yield similar pathological data regarding immune deposits (IF), histology and ultrastructure (EM).
DISCUSSION
We here present a fast, optimized and simplified way of performing high-throughput visualization of both large-scale (e.g. tubules, glomeruli) and nanoscale (e.g. foot processes, filtration barrier) renal structures. This renal analysis could be achieved in just five hours (from dissection to analysis) which is substantially shorter than most previous clearing and expansion protocols. As a comparison, the turn-around time for full renal diagnostics (LM, IF and EM analysis) is a day or two22. In terms of total duration, complexity, hands-on work and number of reagents, our workflow outperforms previously published microscopy protocols for foot process analysis and glomerular filtration barrier imaging in the kidney. Moreover, we show that the resolution we achieve is high enough to draw conclusions about pathologies in the filtration barrier using simple confocal microscopy, a standard tool already available in most research facilities and pathology labs. Since the swelling of the samples is mild, the effective field-of-view is larger than for other more tedious expansion protocols, allowing also for high-throughput imaging of large-scale morphology and pathology of human samples. It should be stressed that SDS clearing (de-lipidation) is in the protocol not added primarily for optical transparency, but rather to allow for sufficient labelling and swelling of samples. We show, however, that the optical transparency is beneficial when performing imaging of foot processes as a large imaging volume is accessible, allowing for finding and screening more suitable regions of the sample for quantitative analysis.
Furthermore, an important finding of our study is that our protocol performs complementary and beneficially vs. IF/Cryosections, histology and EM, i.e. protocols presently used for standard analysis of sectioned kidney tissue. Still, we believe that the biggest advantage over standard protocols is that this protocol can potentially merge all three workflows of standard microscopy into just one with apparent savings in terms of working hours and material/reagents/equipment needed.
Recent studies have applied Airyscan microscopy, which utilize variable pinhole settings on a small camera array detector in combination with deconvolution and signal reassignment image processing23,24. In this study we show that with mild swelling and subsequent superb staining and imaging conditions, just conventional confocal imaging with a closed-down pinhole provides sufficient resolution without mathematical post-processing of images. Thus, our simple and fast 3D-compatible protocol can be performed with any standard confocal microscope equipped with a suitable high NA objective (oil/NA 1.4 and glycerol/NA 1.3 tested in this work).
For possible use in clinical pathology, the protocol must be further validated for evaluation of the plethora of pathological parameters present in other types of kidney disease as well. As one example, it is in some cases crucial to stain for amyloid deposits (presently visualized using Congo red histology stain) to diagnose amyloid nephropathy. We believe such detection issues are mostly a matter of developing new non-histological staining strategies, and in fact fluorescence-based methodologies for amyloid visualization have been demonstrated by others25. Moreover, the resolution of our protocol is not optically sufficient to visualize nanoscale fibrillary deposits, so for some special cases one might still have to apply EM, or a combination of our protocol with super-resolution fluorescence microscopy for accurate diagnosis. In this work we present comparative data between our protocol and EM for the evaluation of glomerular filtration barrier morphology which indicates that the two approaches yield very similar results. However, it should be stressed that the analysis is carried out on a small number of patients, so a more thorough analysis is in the future necessary to further establish this point. To achieve this, a focused collaborative effort involving both pathologists, sample preparation experts and imaging experts has to be initiated. After such further validations and optimizations, a fast and simple all-scale optical analysis might be at hands for use in renal diagnostics. This thorough validation has not been within the scope of this proof-of-principle study, where the aim was to develop a fast optical protocol for kidney ultrastructure imaging.
Presently, development of improved optical kidney diagnostics is explored by several researchers. There have recently been demonstrations that pathological parameters can be extracted from cleared kidney tissue using both non-labelled (autofluorescence) and fluorescently labelled samples26,27. Furthermore, a recent study shows that “virtual histology” can be performed using deep learning algorithms to auto-fluorescence images28 and neural network-based segmentation and classification of kidney pathologies has also been shown29,30. Merging novel sample-preparation methods (as the ones shown in this study) with new image analysis tools poses many interesting digital possibilities in the future, and we could likely see changes to how we image and analyze kidney samples in the clinic soon.
MATERIALS AND METHODS
Mouse, rat and human kidney tissue
Rats used in the experiments were purchased from Charles River, Germany. Animals were anesthetized (intraperitoneal injection of pentobarbital), the aorta cut, and kidneys dissected. All experiments were approved by Stockholm North Ethical Evaluation Board for Animal Research and were performed in accordance with animal welfare guidelines set forth by Karolinska Institutet.
All mouse experiments were approved by the State Office of North Rhine-Westphalia, Department of Nature, Environment and Consumer Protection (LANUV NRW, Germany) and were performed in accordance with European, national and institutional guidelines. Mice of 100% C57BL/6N background were used. After anesthesia with Ketamine and Xylazine, mice were euthanized by cardiac perfusion with Hank’s Balanced Salt Solution (HBSS) and fixated as described below.
Patient material was obtained from two children suffering from Steroid-Resistant Nephrotic Syndromes due to either point mutations in NPHS1 or compound-heterozygous point mutations in TRPC6 and NPHS2, as well as three patients diagnosed with focal segmental glomerulosclerosis, minimal change disease and IgA nephropathy, respectively. Control human tissue was collected from patients that were nephrectomized due to renal tumors. Tissue sample was dissected from the non-tumorous pole of the kidney and showed normal histological picture in routine histological examination. All procedures were approved by the Ethics Commission of Cologne and the regional Ethical Committee of Stockholm and conducted in accordance with the declaration of Helsinki. When applicable, patients or the patient’s parents gave informed consent.
Mouse Model for FSGS
Mice with two compound-heterozygous point mutations, PodR231Q/A286V, were generated as previously described11. Mice were sacrificed at 20 weeks of age as stated above.
Fixation
Human, rat and mouse kidneys were fixed in 4% PFA in 1X PBS at room temperature for 1-3 hours. The human kidney tissue from a renal carcinoma patient was fixed in Duboscq-Brasil solution for 3 hours, stored in 70% EtOH and hydrated in 1X PBS for 5 minutes prior to the protocol.
Sectioning
If needed, samples were embedded in 3% agarose and sectioned using a Vibratome to 300 µm thickness.
De-lipidation
Lipids were removed from kidney samples by incubation in clearing solution (200 mM boric acid, 4% SDS, pH 8.5) at 70°C for 1 hour with shaking at 500 rpm.
Labelling
Samples were washed for 5 min in 1X PBS with 0.1% TritonX-100 (PBST) prior to immunolabelling to remove SDS. Samples were then incubated in primary antibody (+ lectins, DRAQ5 for latest version of protocol (primary conjugated antibodies)) diluted in PBST (older version) or 10nm HEPES with 10% TritonX-100 (Scale-CUBIC-1, newer version) at 37°C for 2 hours with shaking at 500 rpm, followed by washing in PBST for 5 minutes at 37°C. For the latest version of the protocol with primary conjugated affinity molecules, the staining protocol was terminated here. For the earlier version (primary + secondary) samples were then incubated in secondary antibody (if primary AB not conjugated to fluorescent dye), lectins and DRAQ-5 diluted in PBST at 37°C for 2 hours with shaking at 500 rpm, followed by 5 minutes washing in PBST at 37°C.
Mounting
Samples were incubated in 75% wt/wt fructose (1 mL of de-ionized water added to 4g of fructose) containing 4M Urea at 37°C with shaking at 500 rpm for 15 minutes and then placed in a MatTek dish with a cover slip on top (to prevent evaporation) prior to imaging.
Antibodies and probes
DRAQ-5 (Thermo Fisher, 62251) was used at a dilution of 1:500. Wheat Germ Agglutinin (Thermo Fisher, W11261) and PHA-E (Vector Laboratories, B-1125) was conjugated to Alexa-488 (Thermo Fisher, A20000) and Atto-594 (AttoTech, AD 594-31), respectively, using NHS chemistry and were used at a 50 µg/mL concentration. Also, WGA conjugated to CF405M (Biotium, 29028-1) was used at the same concentration. Primary antibodies used in this study were rabbit polyclonal antibody to podocin (Sigma-Aldrich, P0372) used at a dilution of 1:100, sheep polyclonal antibody to nephrin (R&D systems, AF4269) used at a dilution of 1:50, rabbit polyclonal antibody to collagen IV (Abcam, ab6586) used at a dilution of 1:50 and a FITC-conjugated rabbit polyclonal antibody to IgA (Dako, F0204). The polyclonal sheep anti-nephrin antibody was also directly conjugated to Alexa-488 using NHS chemistry. A rabbit monoclonal antibody to collagen IV (Abcam, ab256353) was conjugated to Alexa-555 using NHS chemistry. Secondary antibodies used in this study were donkey anti-goat conjugated to Alexa-405 (Abcam, ab175664), DyLight-405 (Jackson Immuno Research, 705-475-003), Alexa-488 (Thermo Fisher, A-11055), Alexa-594 (Thermo Fisher, A-11058) or Abberior STAR635P (Abberior, 2-0142-007-2, dilution 1:50) and donkey anti-rabbit conjugated to Alexa-555 (Thermo Fisher, A-31572) or Cy3 (Jackson Immuno Research, 711-165-152), all used at a dilution of 1:100 unless other stated.
Imaging
Images were acquired using a Leica SP8 confocal system (20X NA 0.75 multi-immersion objective (oil used) or 100X NA 1.4 oil objective as stated in figure legends). Imaging with the 100X objective always implies a pinhole setting of 0.3 AU (as calculated for 594 nm light) if not stated otherwise. Standard histology imaging was performed using a Leica DM3000 system (Fig. 5) or an Olympus BX45 system (Fig. 6), standard immunofluorescence was performed using a Nikon Microphot-FXA and EM performed with a Zeiss 912 (Fig. 5) or a Hittachi 7700 (Fig. 6) transmission electron microscope.
Image processing and analysis
Images were smoothened (unless other stated) by replacing each pixel value by the mean of its 3×3 neighborhood using Fiji/ImageJ. The morphometric analyses described below are visually presented in Supplementary Figure S3.
Slit diaphragm length measurement
Slit diaphragm (SD) length per area was determined using ImageJ/Fiji. In brief, the fluorescent signal was enhanced using background subtraction, mean filtering and contrast enhancement. The SD signal was detected with the help of the ridge detection plugin as previously published by us and Siegerist3,5,11. In the IgAN-patient (Fig 5 p,q), the measurement of slit diaphragm length was done manually. This macro can be downloaded from GitHub at https://gist.github.com/github-martin/e699d18ae6cfa5b1a6aace15d3c3544c.
GBM thickness measurement
The GBM thickness was evaluated using an ImageJ macro where the inner and outer boundaries of the GBM were manually assigned based on the collagen IV and nephrin staining or for EM images based on the electron density grayscale (see Fig. S3).
Slit diaphragm width
For EM images slit diaphragms were visually identified and the distance between them were manually measured using line profiles in ImageJ. The slit diaphragms were identified as nephrin-positive “spots” along the filtration barrier and the distance between them were measured in the same way as for EM images.
Schemes
The scheme in Fig. 1a was produced by D.U-J apart from the vector graphics for the Eppendorf tube and the razor blade which were downloaded from https://pixabay.com/vectors/vial-tube-fluid-laboratory-41375/ and https://pixabay.com/illustrations/razor-blade-blade-sharp-cutting-220323/ respectively (free to use).
DISCLOSURE
All the authors declared no competing interests.
Supplementary Information
ACKNOWLEDGEMENTS
This study was supported by grants from the Swedish Research Council (VR 2013-6041) and infrastructure development support from the Swedish Foundation for Strategic Research (RIF14-0091). We thank the CECAD Imaging Facility for their support in the acquisition of microscopy data.