Application of direct stochastic optical reconstruction microscopy (dSTORM) to the histological analysis of human glomerular disease

Abstract Electron microscopy (EM) following immunofluorescence (IF) imaging is a vital tool for the diagnosis of human glomerular diseases, but the implementation of EM is limited to specialised institutions and it is not available in many countries. Recent progress in fluorescence microscopy now enables conventional widefield fluorescence microscopes to be adapted at modest cost to provide resolution below 50 nm in biological specimens. We show that stochastically switched single‐molecule localisation microscopy can be applied to clinical histological sections stained with standard IF techniques and that such super‐resolved IF may provide an alternative means to resolve ultrastructure to aid the diagnosis of kidney disease where EM is not available. We have implemented the direct stochastic optical reconstruction microscopy technique with human kidney biopsy frozen sections stained with clinically approved immunofluorescent probes for the basal laminae and immunoglobulin G deposits. Using cases of membranous glomerulonephritis, thin basement membrane lesion, and lupus nephritis, we compare this approach to clinical EM images and demonstrate enhanced imaging compared to conventional IF microscopy. With minor modifications in established IF protocols of clinical frozen renal biopsies, we believe the cost‐effective adaptation of conventional widefield microscopes can be widely implemented to provide super‐resolved image information to aid diagnosis of human glomerular disease.


Introduction
Kidney disease is highly complex and challenging to diagnose, typically requiring light microscopy (LM), immunohistology, and electron microscopy (EM). EM is useful in the histopathology of~50% of native kidney biopsies and essential for the diagnosis of~20% [1][2][3], making the use of EM a standard technique in many countries for native kidney biopsy examination following LM and immunofluorescence (IF) imaging. EM is required for the diagnosis of kidney diseases associated with structural abnormalities of the basement membrane (e.g. inherited abnormalities of collagen type IV alpha chains), diseases with fibrils (e.g. fibrillary and immunotactoid glomerulonephritides), and rare genetic diseases such as Fabry's disease or lecithin cholesterol acyl transferase deficiency. It is also routinely used to document morphological changes in podocytes, and to document shape, substructure, and position relative to the glomerular basement membrane (GBM) of immune complexes and/or complement fragment deposits.
However, EM instrumentation is not available to much of the world's population and the number of expert clinical EM staff and facilities is decreasing where it is available. For both developing and developed countries, it would be useful to find a cheaper alternative to EM to enhance diagnosis beyond what is possible with LM and IF, and it would be useful to simplify and accelerate the diagnostic workflow by reducing the number of instruments required. Recently, optical microscopy has been extended below the diffraction limit with superresolved microscopy (SRM) techniques such as structured illumination microscopy (SIM) approaches [4,5], stochastically switched single-molecule localisation microscopy (SMLM) techniques such as photo-activated localisation microscopy [6,7] and stochastic optical reconstruction microscopy (STORM) [8], and RESOLFT [9] techniques such as stimulated emission depletion microscopy [10,11]. Of these SRM techniques, SMLM approaches have the simplest requirements for instrumentationmaking them cost-effective and potentially accessible to a broad spectrum of laboratoriesand can provide spatial resolution below 50 nm, which approaches that of EM. SMLM utilises sequential emission and localisation of stochastically 'blinking' fluorophores that are sufficiently sparse at any given time to permit the position of each emitter to be determined with high precision by determining the centre of the recorded intensity distribution. This emitter blinking can be realised in many ways, e.g. by photoswitching fluorophores to emit in the detection band or otherwise, by photobleaching to terminate emission or by utilising appropriate chemical buffers to facilitate reversible photoswitching of fluorophores in and out of dark states, as demonstrated in the technique described as direct STORM (dSTORM) [12]. In previous work developing an approach we described as 'easySTORM', we have shown that dSTORM can be robustly implemented at a relatively low cost (<£20,000) using multimode diode lasers and multimode optical fibres to provide superresolved images over large (>120 × 120 μm) fields of view [13]. We here show that this approach can be applied to clinical histological sections to provide superresolved IF imaging using clinically approved antibodies an approach we describe as 'histoSTORM'.
We specifically explore the potential to replace EM with histoSTORM in the diagnosis of kidney disease and to potentially provide a widely accessible clinical tool based on much lower cost instrumentation. This follows earlier work using SIM [14] and STORM [15] to study renal podocyte substructure and protein organisation in the GBM. Although this prior work demonstrated the potential of super-resolved IF, it was realised with expensive commercial SRM instrumentation and the study utilising STORM was undertaken with mouse tissue and non-clinically approved antibodies. We aim to develop a low-cost approach that could be accessible by clinicians in low-and middle-income countries by utilising 'easySTORM' to image clinically relevant proteins, such as immunoglobulin G (IgG) in GBM, with existing clinically validated antibodies and to develop practical protocols to work with existing biopsy samples such as frozen sections or formalin-fixed paraffinembedded (FFPE) sections. We note that STORM has previously been applied to research pathology, e.g. to study epigenetic modulation [16] and the progression of cancer [17], but not to clinical histological sections using clinically approved antibodies.

Sample preparation
The protocol to prepare kidney biopsies for histoSTORM is provided in Table 1. Frozen kidney biopsy sections of 3 μm thickness on slides were circumscribed with the addition of silicon (Polycraft ZA22 Mould RTV Addition Cure Mould Making Silicone Rubber; MB Fibreglass, Newtownabbey, Northern Ireland) and polymerised at room temperature to achieve rubber consistency and form a well of 0.1 ml volume. For FFPE sections, paraffin was removed with xylene, using two treatments for 5 min and then ethanol washes of 2 min with decreasing ethanol dilutions in water at 100, 75, 50, 25, and 0%. Samples were allowed to dry and additional silicon was applied to reinforce the sample well. Acetone fixation for 10 min followed by three quick washes with phosphate-buffered saline (PBS) was followed by antigen retrieval at 37 C for 27 min with 4 ml of Protease Type 24 (P8038; Sigma-Aldrich, Dorset, UK) at 0.125 mg/ml in prewarmed PBS covering the whole slide. Slides were then washed three times for 5 min at room temperature in a Coplin jar with 50 ml of diluted PBS (pH 7.4) in water at a ratio of 1:10. Samples were then incubated for 10 min in 0.1 ml of PBS with 1 mg/ml sodium borohydride to reduce tissue autofluorescence, followed by three further washes as previously described. Unspecific antibody-binding sites were reduced on samples by blocking with 0.1 ml of 3% (w/v) bovine serum albumin (BSA) in PBS at room temperature for 10 min.
The tissue samples were then labelled using 0.1 ml of a cocktail of primary antibodies -Laminin 439 dSTORM and human glomerular disease MAB1920 (Millipore, Watford, Hertfordshire, UK) and IgG (A0423; Agilent Dako, Stockport, Cheshire, UK) or Rabbit Isotype X0936 (Agilent Dako)diluted at 1:10,000 ratio in 3% (w/v) BSA at room temperature for 20 min. The samples were then washed three times as previously described and treated with 0.1 ml of a cocktail of secondary antibodiesgoat anti-mouse IgG H+L diluted at 1:2,000 ratio (A32727; Invitrogen, Ther-moFisher Scientific, Loughborough, Leicestershire, UK) and 0.25 mg/ml goat anti-rabbit IgG H+L (16837; AAT Bioquest, Stratech, Ely, Cambridgeshire, UK) diluted at 1:2,000 ratio in 3% BSA at room temperature for 20 min. The secondary antibodies were conjugated to either Alexa Fluor 555 (ThermoFisher Scientific) or iFluor 647 (AAT Bioquest). The samples were then washed a further three times and then fixed with acetone for 5 min.
To improve image quality, the tissues were then chemically cleared by immersing in 0.1 ml of 70% 2'2 thiodiethanol (TDE) (166782; Sigma-Aldrich) in PBS for 10 min at room temperature [16]. Samples were then treated with 0.1 ml of STORM buffer (mercaptoethylamine 50 mM, D-lactate 10 mM, and 60% TDE in PBS and 0.75 U/ml of Oxyrase-EC [SAE0010; Sigma-Aldrich]) at room temperature for 30 min. Immediately prior to imaging, the sample was mounted in a fresh preparation of STORM buffer.

Imaging
We employed our easySTORM [13] implementation of dSTORM that utilises a set of fibre-coupled multimode laser diode sources (Laserbank; Cairn Research Ltd, Faversham, Kent, UK) with a standard inverted microscope frame (Axiovert 200; Carl Zeiss GmbH, Jena, Germany) and an excitation beam coupling unit (OptoTIRF; Cairn Research Ltd) that can be configured for epifluorescence or total internal reflection. We note that these commercial components can be replaced with much lower cost home-built equivalents [18] and a complete super-resolved IF microscope can be assembled for a component cost less than £20,000 [18]. For the results presented here, dSTORM of the histological sections was undertaken using a ×100, 1.46 numerical aperture oil lens in the epifluorescence microscope with an sCMOS camera (Photometrics Prime 95B, Tucson, AZ, USA).
To image structures labelled with Alexa Fluor 555, excitation at 520 nm was initially set to 1,045 μW/cm 2 for 5-10 s to activate fluorophore blinking and then the power was decreased by~50% during dSTORM image acquisition. To image structures labelled with iFluor 647, excitation at 635 nm was initially set to 2,500 μW/cm 2 at the sample plane for 5-10 s and then reduced by~50% during dSTORM image acquisition The camera integration time was set to 30 ms, with images acquired at a frame rate of 33 Hz.
Super-resolved images were reconstructed by Thun-derSTORM with drift correction enabled, as reported previously [13]. Two-channel images were generated with a cross-correlation function that can be found at https://github.com/yalexand/Imperial-ClusDoC.git on the Alexa Fluor 555 and iFluor 647 channels.

Results and discussion
In this initial small study, we applied histoSTORM with standard clinically approved antibodies to both FFPE and frozen histological sections and, while we expect the sample processing protocols can be further optimised, we believe that the exemplar results presented below show ultrastructure of clinical interest that could potentially aid diagnosis.

E Garcia et al
Membranous glomerulonephritis (Figure 1) is characterised by subepithelial immune complex deposits containing IgG in the GBM with thickening of this structure [19]. The two-colour widefield IF image in Figure 1D shows a capillary loop with the GBM in green stained with an anti-human laminin probe (green, Alexa Fluor 555) and IgG deposits on the epithelial side of the filtration barrier in red (IgG, iFluor 647). dSTORM images rendered at 25 nm per pixel reveal well-defined subepithelial deposits that are consistent with those observed by EM ( Figure 1F,I).
Whilst there is no clear definition of structures in the widefield IF images ( Figure 1D,G), details of immune deposits at subepithelial regions and a gradient of content of immune deposits are readily observed in dSTORM ( Figure 1E,H). Green regions of the dSTORM images indicate areas of GBM free of immune deposits, yellow areas indicate overlap of laminin and immune deposits, and red areas indicate clusters of immune deposits.

dSTORM and human glomerular disease
Lupus nephritis is characterised by glomerular deposition of polyclonal IgG in various areas of the glomerulus [20,21]. Figure 2 shows deposits of IgG (red, iFluor 647) and basement membrane staining (laminin, green, Alexa Fluor 555) in a glomerular capillary of stage IV lupus nephritis. In this stage, mesangial ( Figure 2G), subendothelial ( Figure 2H), and subepithelial ( Figure 2I) IgG deposits are readily observed with dSTORM and recapitulate the distribution of high electron density IgG deposits documented with EM ( Figure 2C).
EM is also routinely used to measure the thickness of the GBM [22]. The large field of view (of the order of 120 μm × 120 μm) with resolution below diffraction

442
E Garcia et al limit provided by the easySTORM platform allows evaluation of GBM thickness in glomerular capillaries and documentation of other aspects of the glomerulus. Figure 3A,C shows widefield epifluorescence images of an FFPE section from a biopsy of minimal change disease, where the laminin in the GBM is labelled with iFluor 647. Figure 3B,D shows the corresponding STORM images and the ability of the STORM images to enable GBM thickness measurements below the diffraction limit is confirmed. Figure 3F,G shows line sections through the widefield and STORM images of the GBM. Figure 3E shows an electron micrograph with GBM width of 281 nm.
In conclusion, we have demonstrated that histoSTORM of frozen or FFPE kidney biopsy sections can provide additional information compared to conventional widefield IF. histoSTORM can precisely locate subepithelial, subendothelial, and mesangial immune complex deposits, which can aid the diagnosis of glomerulonephritis, and enables the thickness

dSTORM and human glomerular disease
of the GBM to be measured with sufficient resolution to aid diagnostic assessments where EM is not available. We note that recalibration of the thickness of the GBM measured using histoSTORM relative to that using EM may be required to establish diagnostic criteria.
As well as being much cheaper to implement and sustain, compared to EM, the sample preparation for histoSTORM is similar to IF and the larger fields of view enable much faster imaging than EM. This could allow, for example, multiple capillaries to be routinely documented within a glomerulus, including at multiple planes along specimen depth. We note that emerging image processing tools, including those based on machine learning, could further enhance the ability of histoSTORM to probe ultrastructure and to diagnose disease.
While this initial study does not establish that histoSTORM can fully replace EM in renal diagnosis, it does provide evidence of added value relative to LM and IF. However, further prospective studies of large case series would be required to establish its clinical utility. histoSTORM could be useful as an auxiliary techniqueas could other advanced optical microscopy techniques such as SIM [14] and could refine current classification stages of glomerular lesions and other renal pathologies, noting that these are periodically revisited and modified based on new findings and progress in understanding of mechanisms of disease and tissue injury [20,21]. There may also be a clinical role for correlative STORM/EM, as previously presented in mouse kidney frozen sections [15].
Ultimately, histoSTORM may not be able to replace EM for all renal diagnoses but we believe that it has the potential for wide clinical impact, especially in less well-resourced settings where EM is not available.