Improved immunostaining of nanostructures and cells in human brain specimens through expansion-mediated protein decrowding

Proteins are densely packed in cells and tissues, where they form complex nanostructures. Expansion microscopy (ExM) variants have been used to separate proteins from each other in preserved biospecimens, improving antibody access to epitopes. Here we present an ExM variant, decrowding expansion pathology (dExPath), which can expand proteins away from each other in human brain pathology specimens, including formalin-fixed paraffin-embedded (FFPE) clinical specimens. Immunostaining of dExPath-expanded specimens reveals, with nanoscale precision, previously unobserved cellular structures, as well as more continuous patterns of staining. This enhanced molecular staining results in observation of previously invisible disease marker-positive cell populations in human glioma specimens, with potential implications for tumor aggressiveness. dExPath results in improved fluorescence signals even as it eliminates lipofuscin-associated autofluorescence. Thus, this form of expansion-mediated protein decrowding may, through improved epitope access for antibodies, render immunohistochemistry more powerful in clinical science and, perhaps, diagnosis.

Expansion microscopy (ExM) enables physical expansion of biological specimens, thereby permitting nanoscale resolution imaging on diffraction-limited microscopes (15,16).Briefly, ExM starts by covalently anchoring biomolecules, or labels against biomolecules, to a swellable hydrogel densely and evenly synthesized throughout a preserved biological specimen.Then, an enzymatic or protein-denaturing treatment softens the mechanical properties of the specimen.Water then causes the polymer network to expand, and thus the anchored molecules to be pulled uniformly away from one another.Given the difficulty of labeling many epitopes in their natural, densely packed state, we asked whether, in human tissues, conventional antibodies introduced in the post-expansion, decrowded, state could access previously undetectable epitopes.Some expansion protocols preserve protein antigens throughout the expansion process (Table S1) (17)(18)(19)(20)(21)(22)(23)(24)(25), and are thus compatible with post-expansion immunostaining.However, most of these existing post-expansion staining protocols either require specialized fixative compositions (17,18,21,22,25), and thus are incompatible with archival clinical samples, or cause tissue cracks and anisotropy due to incomplete tissue softening (19), or had uncharacterized nanoscale isotropy (20).In addition, none of these studies underwent quantitative comparison of structures or cells in the same specimen of human tissue with pre-versus post-expansion staining, key to understanding whether the decrowding of proteins contributed to visualization of previously invisible structures.
We previously developed expansion pathology (ExPath), a form of ExM that prepares human specimens for expansion microscopy, using pre-expansion antibody staining (6).
Here we present decrowding ExPath (dExPath), an expansion pathology variant that preserves protein epitopes for post-expansion staining.dExPath can be applied to formalinfixed paraffin-embedded (FFPE) human tissues, as well as other standard formats of interest in basic and applied biology, such as4%-paraformaldehyde (PFA)-fixed mouse brain tissue.We validated dExPath systematically, comparing, within the same specimen of human brain tissue, immunostaining intensity and continuity between pre-and post-expansion staining, showing improvements in both intensity and continuity, and revealing new features, including disease marker-bearing cell populations (in human glioma specimens) that were previously invisible.Furthermore, dExPath eliminates autofluorescence associated with lipofuscin, an aggregated product commonly found in brain tissue, in addition to autofluorescence reduction resulting from the loss of autofluorescent molecules shown in prior expansion protocols (6).dExPath supports multi-round immunostaining, enabling highly multiplexed imaging of protein targets within the same human brain specimen.

Rationale for dExPath technology
We first prepared tissue to enter the expansion pipeline (Fig. 1A; involving tissue deparaffinization and re-hydration, for FFPE samples) (6), followed by protein anchoring and gel formation (Fig. 1B).In contrast to the original ExPath protocol, which uses protease digestion to soften the specimen (feasible because fluorescent antibodies, which are partly protease-resistant, are applied pre-expansion and anchored to the polymer network for later imaging), we used a buffer to maximally enable protein separation for post-expansion staining.We used higher concentrations of sodium dodecyl sulfate (SDS) (20% weight/ volume (w/v)) than in earlier protein-preserving protocols (Table S1) (17)(18)(19)(20)(21)(22)(23)26), reasoning that this could better denature proteins, and minimize non-covalent intra-and inter-protein interactions (27,28).We included a new ingredient, the reducing agent β-mercaptoethanol (100 mM), which we reasoned could cleave inter-molecular disulfide bridges between proteins (24,(27)(28)(29)(30)(31).We used the same high concentration of ethylenediaminetetraacetic acid (EDTA) (25 mM) as in original ExPath, which proved to be useful for isotropic tissue expansion (6), possibly through de-stabilization of metal-mediated protein interactions (28)(29)(30).We used a higher temperature than in original ExPath, adapted from a form of proExM that uses autoclaving to expose samples to 121°C (Fig. 1C) to strongly denature and loosen bonds between proteins in the sample, allowing them to separate during washes (which drives partial tissue expansion, ~2.3x;Fig. 1D).Antibodies were applied at this post-decrowding state (Fig. 1E; see Table S2 for antibodies used in this work), instead of the fully expanded [~4x] state, because full expansion requires sample immersion in water, which can hinder antibody binding (17)(18)(19)(20)(21)(22).Multiplexing is possible because these antibodies can be stripped using the same buffer, and then new antibodies applied (fig.S1), a strategy previously demonstrated by other post-expansion staining protocols but not on human tissues (18,20,32).
High grade glioma tissues are known to undergo abnormal endothelial proliferation, leading to some areas of tissue with abnormally large amounts of vascularity and extracellular matrix (ECM).These areas can be identified under conventional clinical microscopy (33) but present a challenge to isotropic expansion of tissue (6).To circumvent this problem, we devised a modified form of dExPath using collagenase treatment prior to softening (fig.S2).Thus, dExPath was designed to provide a methodology for isotropic tissue expansion, enabling preservation, and post-expansion as well as multiplexed staining, of decrowded proteins in both normal and pathologic human and rodent brain tissues.

Validation of dExPath expansion isotropy in brain tissue
We validated the isotropy of dExPath on normal and diseased FFPE 5-μm-thick human brain tissues (standard for clinical samples), using the pre-vs-post distortion analysis used to validate earlier expansion protocols (6,15,18,19,34,35).We performed antigen retrieval followed by pre-expansion immunostaining against microtubule-associated protein 2 (MAP2, a neuronal dendritic marker) (36), and the intermediate filament protein vimentin (37)(38)(39), on normal human hippocampus (Fig. 2A) and on high-grade glioma tissues (located in the human cortex or white matter) (Fig. 2B), respectively.We performed standard immunostaining (6,(40)(41)(42) and obtained pre-expansion images using a super-resolution structured illumination microscope (SR-SIM) (Fig. 2A-B).Next, we performed dExPath (modified to use pre-expansion staining prior to anchoring and gelation, to facilitate distortion comparison between pre-and post-expansion images of the same sample, outlined in fig.S3), obtaining post-expansion images of the same fields of view (Fig. 2C-D) using a confocal microscope.We observed low distortion between pre-and post-expansion images, similar to previous versions of ExM applied to mouse brain tissue (6,15,19) (Fig. 2E-F).Our modified form of dExPath using collagenase treatment prior to softening was used to compare pre-and post-expansion images of the same specimen, outlined in fig.S4; low distortion was obtained on high-grade glioma tissues with a high degree of ECM (fig.S5).Thus, dExPath isotropically expands archival clinical samples of FFPE normal brain and brain tumor tissues by ~4x without the need for enzymatic epitope destruction (6,19), or specialized fixatives (17,18,21,22).

dExPath enables visualization of decrowded proteins revealing previously invisible cells and structures
To investigate whether post-expansion immunostaining could enable detection of previously inaccessible protein epitopes, we compared pre-vs.post-expansion staining of normal human hippocampus (Fig. 4A-F), supratentorial high-grade glioma tumor specimens (Fig. 4G-R), and low-grade glioma tumor specimens (Fig. 4S-X).Tissue samples were imaged pre-expansion, after antigen retrieval and antibody staining (Fig. 4A, G, M, S), and after expansion without restaining (Fig. 4B, H, N, T); and after expansion and re-staining with the same antibodies (Fig. 4C, I, O, U; experimental pipeline in Fig. S3).All tissue states were imaged using identical confocal imaging settings.
To quantify the improvement in labeling post-expansion vs. pre-expansion, we constructed a binary image "signal" mask, for each stain, that corresponded to pixels that were positive (above a manually selected threshold) for a given stain in both pre-expansion and postexpansion staining images.We also created a second "background" mask, for each stain, that corresponded to pixels that were negative (below the threshold mentioned before) in both pre-and post-expansion staining images; a "double negative" background mask corresponded to the pixels that were negative in both of these background masks.Next, we constructed regions of interest (ROIs) that were small enough (0.2 microns) to fit entirely within the signal mask for a given stain, but that were at least an ROI-width away from the signal mask for the other stain; we also constructed ROIs that were fully contained within the double negative mask, and similarly far from pixels that were positive in either signal mask.Finally, we calculated intensities averaged across the ROIs for the same locations in the expanded (Fig. 4B) vs. expanded-and-restained (Fig. 4C) images.In regions positive in the MAP2 and GFAP signal masks (Fig. 4D, left and Fig. 4E, right), we saw increases of both signals in their respective ROIs (MAP2, p = 0.0003, 2-tailed paired t-test; n = 3 tissue samples from different patients; GFAP, p = 0.0007, 2-tailed paired t-test; n = 3 tissue samples from different patients).Of course, we would not expect MAP2 to occur in GFAP-positive regions, nor GFAP in MAP2 regions.Thus, these two proteins give us the opportunity to assess whether post-expansion antibody application suffers from nonspecific staining.Indeed, GFAP was consistently low in both pre-expansion and post-expansion images (p = 0.0004, 2-tailed paired t-test; n = 3 tissue samples from different patients) in locations within the MAP2 signal mask.Similarly, MAP2, imaged in the GFAP signal mask, was consistently low in pre-and post-expansion images (p = 0.003, 2-tailed paired t-test).
In the double negative regions, MAP2 intensities were consistently low in pre-decrowding and post-decrowding states (Fig. 4F, left), as were GFAP intensities (Fig. 4F, right).Thus, staining in the double negative regions was similar to that in the single negative regions, supporting the idea that the nonspecific staining is extremely low.
Next, we examined vimentin and α-SMA in high-grade glioma tissue.Vimentin is expressed in some tumor cells (70), some activated microglia (71), as well as all endothelial cells (38) and some pericytes (72).Thus, vimentin would be expected to sometimes be near, or even overlapping, with α-SMA (in pericytes) and sometimes to be well-isolated from α-SMA (in other cell types) (68,69,73).We observed vimentin and α-SMA signals in the blood vessel wall and surrounding the vessel lumen (Fig. 4M).Vimentin signals were also observed in cells (putative tumor cells or activated microglia) outside of blood vessels (Fig. 4M); with similar observations after 4x expansion (Fig. 4N).However, after post-expansion re-staining (Fig. 4O), new vimentin-positivity appeared in cells, far from blood vessels, that were previously vimentin-negative (Fig. 4M-O).We analyzed vimentin ROIs far away from α-SMA, and found the vimentin staining to go up in these ROIs (Fig. 4P, left; p = 0.0008, 2-tailed paired t-test; n = 3 tissue samples from different patients); in α-SMA ROIs, vimentin also went up significantly (Fig. 4Q, left; p = 0.0001, 2-tailed paired t-test; n = 3 tissue samples from different patients), as expected.In contrast, α-SMA was very little located in the vimentin ROIs Fig. 4P, right; p = 0.0001, 2-tailed paired t-test; n = 3 tissue samples from different patients), and went up in α-SMA ROIs (Fig. 4Q, right; p = 0.04, 2-tailed paired t-test; n = 3 tissue samples from different patients) to some extent -clearly, not all proteins are equally crowded in all cells; perhaps α-SMA is relatively uncrowded to begin with.As before, double negative staining was consistently low (Fig. 4R).
Having validated the decrowding aspect of dExPath, we next examined whether the improved immunostaining facilitated by dExPath improved images in comparison to those obtained by a similar-resolution super-resolution method that does not decrowd epitopes, SR-SIM.We first performed antigen retrieval and stained high-grade glioma and normal hippocampus with anti-vimentin or anti-MAP/anti-GFAP.Samples were imaged by SR-SIM (Fig. 5A,B), followed by the first part of the dExPath protocol (chemical softening and expansion) (fig.S3A-D) to acquire confocal images post-expansionwith pre-decrowdingstaining (Fig. 5C-D).Next, we performed the last part of the dExPath protocol to acquire confocal images post-expansion with post-decrowding-staining (Fig. 5E-F; fig.S3E-F).Both SR-SIM and post-expansion confocal images of pre-decrowding-stained tissue revealed punctate patterns for vimentin (Fig. 5A, 5C), MAP2, and GFAP (Fig. 5B,5D).In contrast, these stains revealed continuous structures after post-decrowding staining (Fig. 5E, 5F), as well as new structures that had not been observed (compare Fig. 5A,C,E vs. 5B,D,F).Thus, dExPath may provide a general solution to the problem of punctate staining appearance in brain tissues, for continuous signals, in super-resolution microscopy (2,9,11,13,80).

dExPath retains proteins with improved visualization of targets in normal, glioma, Alzheimer's disease (AD) and Parkinson's disease (PD) human brains
We next investigated whether post-decrowding immunostaining improved visualization of protein targets using validated, commercial antibodies useful for pathological analysis.We observed improved visualization compared to standard histopathological chromogenic analysis in immediately adjacent tissue sections of normal human cortex, Alzheimer's disease (AD) human cortex, and Parkinson's disease (PD) human cortex (fig.S8).For chromogenic analysis, tissues were imaged pre-expansion, after antigen retrieval, primary antibody staining followed by biotinylated secondary antibody staining and use of 3,3'-diaminobenzidine (DAB).For dExPath, immediately adjacent tissue sections underwent post-decrowding staining with the same primary antibodies, and fluorescent secondaries, under the same conditions (experimental pipeline in fig.S1).dExPath images yielded expected biological targets such as putative neurons (via anti-MAP2 antibody), neurofilaments (via anti-NF-L), AD plaques (via anti-β-amyloid), and PD aggregates (via anti-α-synuclein), but demonstrated better resolution, especially for densely-packed filamentous structures such as MAP2 and NF-L, and more structural detail with amyloid plaques and α-synuclein aggregates, vs. standard DAB analysis.
We surveyed a panel of antibodies commonly used by clinical pathology labs, and found that dExPath could produce high quality images, enabling detection of protein targets across glioma, normal, AD, and PD human tissues (fig.S9).We also observed that dExPath does not lead to protein loss (fig.S10).We demonstrated that dExPath enables multiple rounds of staining and imaging on the same tissue sample, allowing multiplexed imaging with nanoscale resolution of human brain tissues (fig.S11 and fig.S12).dExPath worked well on mouse brain tissue, fixed with standard PFA (fig.S13).dExPath could be applied to tissue sections thicker than used in standard pathology preparations (50-100 μm) (fig.S14; Movie S1 and S2).

dExPath reveals cell populations exhibiting combinations of disease-state markers in human glioma tissue
Our prior experiments using glioma tissues (Figs. 4 and 5) demonstrated that post-expansion staining increases the intensity, continuity, and number of structures stained for vimentin, Iba1 and GFAP vs. pre-expansion staining.We next asked whether this could lead to detecting more cells carrying specific antigen combinations, which might alter interpretation of clinical biopsies as well as basic understanding of brain tumor biology.For example, cells with both GFAP and vimentin have been reported to be more aggressive than vimentinnegative/GFAP-positive tumor cells (91,(123)(124)(125).
For identification of cells, one may want the enhanced staining afforded by post-expansion staining, without incurring the time cost of imaging an expanded specimen.Thus, we compared the initial pre-decrowded immunostained state to the post-decrowded immunostained state after tissues were shrunk back down to almost their native size (~1.3x) by adding salt.We imaged low-grade glioma tissue sections serially i) after antigen retrieval and pre-decrowding immunostaining (Fig. 7A); ii) after dExPath softening, washing with PBS (which results in an expansion factor of ~2.3x), and tissue shrinkage (via adding salt to attain expansion factor of 1.3x, Fig. 7B); iii) after ~4x expansion (~4x, Fig. 7C); iv) after post-decrowding immunostaining, washing (~2.3x) and shrinkage (~1.3x) (Fig. 7D); and v) after a final expansion step back to ~4x (Fig. 7E).In this way we could decouple the effects of improved resolution from those of improved staining, in the same specimen.
By comparing samples at pre-decrowding vs. post-decrowding staining stages, both in the shrunken ~1.3x state (Fig. 7A,B,D), we observed that post-decrowding immunostaining (Fig. 7D) was able to reveal additional vimentin-, GFAP-, and Iba1-positive staining not detected in the pre-expansion (Fig. 7A) or pre-decrowding staining (Fig. 7B) states, despite the lack of physical magnification in all three cases.Some regions showed increased signal, after post-decrowding immunostaining.For example, some regions showed new structures that were GFAP-and vimentin-positive (compare Fig. 7D-E vs. 7A-C), or Iba1, GFAP and vimentin positive (compare Fig. 7D-E vs. 7A-C).Indeed, when we examined the fraction of pixels that were positive for one or more stains in single z-slices of pre-expansion (Fig. 7A) and post-decrowding (Fig. 7D) images, they increased significantly (Fig. 7F; p <0.05).
These increases in stain-positive pixels translated into increases in the number of cells identified with a label (Fig. 7G; vimentin, p = 0.032, 2-tailed paired t-test, n = 3 tissue samples from different patients; GFAP, p = 0.0071, 2-tailed paired t-test, n = 3 tissue samples from different patients; Iba1, p = 0.0011, 2-tailed paired t-test, n = 3 tissue samples from different patients).Thus, the number of cells corresponding to some tumor cells, some activated microglia, as well as all endothelial cells and some pericytes of mesenchymal origin (vimentin), or astrocytes and glioma cells (GFAP), or macrophages and microglia (Iba1), increased dramatically, suggesting that many cell types important for glioma pathology and response may be quantitatively underestimated by conventional immunostaining.
We next examined the counts of cell types defined by multiple labels.As mentioned earlier, a cell with both GFAP and vimentin is an aggressive tumor cell (91,(123)(124)(125), a cell with Iba1 and vimentin is an activated macrophage or microglial cell (71,74,127), and a cell with Iba1 and GFAP is either a macrophage or microglial cell that phagocytosed a GFAP expressing cell (astrocyte or tumor cell) or a cell type with a dual astrocytic and macrophage/microglia molecular phenotype (75-77, 79, 122).In each case, the dually labeled cell is qualitatively different from a singly labeled one.Cells positive both for GFAP and vimentin, identified as aggressive/invasive tumor cells, increased in number by about 6-fold with post-expansion vs. pre-expansion staining, suggesting that more aggressive/ invasive tumor cells are present than previously thought (Fig. 7G, p = 0.0035, 2-tailed paired t-test, n = 3 tissue samples from different patients).Amongst GFAP-expressing cells, we observed a ~30% increase in the fraction that were vimentin-positive (Fig. 7H, p = 0.036, 2-tailed paired t-test, n = 3 tissue samples from different patients), suggesting that even in low-grade gliomas, a vast majority of tumor cells may be aggressive.Cells double-labeled with Iba1 and vimentin increased by about 4-fold using post-decrowding vs. pre-decrowding staining (Fig. 7G, p = 0.0030, 2-tailed paired t-test, n = 3 tissue samples from different patients), suggesting that a majority of activated macrophages and microglia might currently be overlooked.
Similarly, cells double-labeled with Iba1 and GFAP increased by about 10-fold with postdecrowding vs pre-decrowding staining (Fig. 7G, p = 0.00043, 2-tailed paired t-test, n = 3 tissue samples from different patients).These dual labeled cells are indicative of two cell populations.One population is that of macrophages or microglia, which have phagocytosed GFAP-expressing cells or debris in the tumor tissue sample from astrocytes or tumor cells) (75-77, 79, 122).Macrophage or microglial phagocytosis of GFAP-expressing cells and their debris may support tumor growth via removing of debris such as apoptotic corpses from the tumor microenvironment (75).The second cell population might be a cell population found in diseased states such as stroke and neurodegenerative states (76,77,122) and recently found to be present in glioblastoma (79), in which cells share the molecular signatures of both Iba1 expressing cells (macrophages or microglia) and GFAP expressing cells (astrocytes or tumor cells) (75,79).We show a substantial increase of these Iba1-GFAP dually labeled cells, which can have a protumorigenic role in low-grade gliomas (75,79).Approximately 80% of Iba1-expressing cells also exhibited GFAP post-expansion, versus only 20% pre-expansion (p = 0.000094, 2-tailed paired t-test, n = 3 tissue samples from different patients, Fig. 7H).In summary, we observed an increase in the percentage of immune cells with phenotypes of importance for the growth of low-grade gliomas.

Discussion
We describe here a form of expansion microscopy, dExPath, that enables immunostaining of decrowded proteins, for nanoscale visualization of previously unseen biological structures and cell populations in human clinical tissue specimens.By isotropic magnification of human tissues, together with antigen preservation, we achieve protein decrowding, improving the accessibility of target epitopes by antibodies (3-5, 9-12, 128).dExPath works across both normal and diseased brain tissue (low-and high-grade gliomas) types, improving immunostaining for many molecular targets.dExPath enables immunostaining of previously inaccessible cells or subcellular features in normal brain and tumor tissues.Post-expansion staining not only improved the continuity of staining for existing structures, but revealed new, previously invisible, structures of appropriate morphology -as has been noted before in mouse brain tissue (26), but now shown for human brain tissue.It also increased cell counts, including cell types involved with tumor aggression and immune response.These results suggest that post-decrowding staining could, potentially, uncover cell populations that may contribute to early tumor progression but remain undetected with common histological methods.Further studies correlating the presence of these cell populations with clinical outcomes will be necessary to quantify and apply increased clinico-pathological accuracy.
While the clinical applications of these findings need to be explored further, in the context of patient outcomes, treatment regimens, and other relevant factors beyond the scope of this technology paper, to be relevant for any potential future diagnostic or prognostic use, our results show potential for dExPath as a research tool for clinicians and researchers to uncover immunostaining patterns previously unseen in diseased tissues, with further potential, when validated in clinical contexts, for improved diagnostics.
Post-decrowding staining may increase the number of spatially accessible epitopes on a target protein, increasing labeling density of the antibodies and their associated fluorescent signals.Previous studies demonstrate improvement in immunostaining by using small-sized probes (~3nm) (3,4,9,13,14,80).dExPath supports the use of conventional off-the-shelf antibodies and can therefore be applied immediately in research settings.
dExPath removed the autofluorescence from lipofuscin found in senescent brain tissues (43)(44)(45)(46)(47)(48)(49)(50), improving IHC-mediated detection of intracellular structures.While other methods exist for the masking or quenching of lipofuscin autofluorescence, such as with Sudan Black B (51), they have been associated with limitations including interruption of antibody binding, and reduction of on-target fluorescence (44,46,49,50).dExPath provides for highly multiplexed immunostaining of decrowded proteins, by retaining protein antigenicity across sequential rounds of antibody stripping and re-staining.These capabilities could be useful for mapping cellular and molecular types and states in normal and diseased tissue microenvironments.
This study examined several antigen targets that have been commonly used as molecular markers to identify specific cell types or states important in normal or diseased brains.For example, dExPath revealed abundant, clearly resolved GFAP-positive filaments in non-diseased human brain tissue, via its decrowding capability.GFAP is involved in physiological and injury-induced functions in which the precise role of this protein remains unknown but its spatial localization appears critical for function (for example, formation of glial scars (129) (130), maintenance of myelinated sites (131), lining of the blood-brain barrier (132), etc.).Visualizing GFAP-positive filaments will facilitate studies of cellular responses to brain injury in clinically relevant human contexts.
Our triple staining experiment (vimentin, Iba1, GFAP) of low-grade glioma tissues showed that dExPath can reveal substantially increased colocalization between these cell type markers, with implications for the analysis of cell populations in glioma biology.For example, our detection of an increased number of previously undetected double-labeled GFAP-and vimentin-positive cells in low-grade glioma tissue may represent a malignant cell subpopulation in these tumors (133)(134)(135)(136)(137), usually not detected histologically.Similarly, cells double-labeled with Iba1 and vimentin (interpreted as activated macrophages or microglial cells (71,74,127)) may represent a smoldering status of immune activation that could have clinical relevance in these tumors, and cells double-labeled with Iba1 and GFAP may represent an increase in the number of phagocytic macrophages/microglia, or possibly an increase in tumor cells with phagocytic properties with increased invasive ability (74, 78, 79) (138)(139)(140).
While we have primarily focused on glioma tissues for this study, dExPath could be applied to other malignancies and/or neuropathologies (AD and PD).
Our study has limitations.Here we optimized dExPath for normal and pathologic brain tissues.However, its application in other tissue types may require additional optimization; for example, a preliminary application of standard dExPath to lymph node tissue did not yield the same low distortion seen for brain (fig.S15), suggesting that additional softening would be useful for tissues more fibrous than the soft brain; lymph nodes studies are beyond the scope of our clinical expertise.dExPath uses low cost, commercially available reagents and instruments (confocal microscopes) found in a conventional basic science laboratory.However, confocal microscopes and some of the reagents, all of which are commercially available, are not standard in clinical laboratories, which might limit the immediate use of dExPath in a standard clinical setting.
dExPath requires multiple manual steps that could benefit from automation to become more efficient for routine use.dExPath may be amenable for automation, given the low amount of tissue deformation present (Fig. 2), the excellent protein preservation evidenced (figs.4-7; figs, S9, S13), even following multiple rounds of immunostaining and stripping (fig.S11,  S12), and the previously published compatibility of gel-embedded tissues for automated multiple rounds of staining and imaging in an RNA context (141).Image analysis would benefit from software that supports accurate quantification and automated registration, currently not part of routine clinical laboratory workflows.Lightsheet microscopy could accelerate imaging of expanded samples (35,105), but again is not routinely used in the clinic.Studies demonstrating the clinical benefit of dExPath in prognosis or personalized medicine would be required for ultimate clinical adoption.All of these limitations represent urgent opportunities for our field.
In conclusion, dExPath achieves protein decrowding and highly multiplexed immunostaining of archival clinical samples while enabling nanoscale resolution imaging on conventional microscopes, all accomplished using low cost, commercially available reagents and instruments found in conventional basic science or pathology laboratories.We anticipate broad utility of dExPath in many scientific and clinical contexts to advance our understanding of molecular relationships in pathological states and improve diagnostic capabilities.

Study design
The aim of this study was to optimize, and characterize, decrowding expansion pathology (dExPath), for use in normal and pathologic human brain tissues.The goal of this technology is to preserve protein epitopes for post-decrowding staining using commercially available antibodies while expanding tissues isotropically so that nanoscale resolution can be achieved on conventional microscopes.We characterized the ability of dExPath to eliminate tissue autofluorescence associated with lipofuscin and other molecular phenomena (β-amyloid plaques).We characterized the ability of dExPath to support multi-round post-expansion immunostaining.For this we used human tissue microarrays with specific numbers of samples and experimental replicates as indicated in the figures and main text.No blinding, randomization, or prior power calculations were performed.

Statistical analysis
For sample sizes n < 20, individual level data in tabular format can be found in supplemental Table S3.2-tailed paired student's t test was used when comparing two groups.Statistically significant results were marked with * in figures, with specific P-values noted in the text.
Statistical analyses were performed using GraphPad prism software.

Human and animal samples
The normal brain, low-and high-grade glioma human samples used in this study were all 5-μm-thick formalin-fixed paraffin-embedded (FFPE) tissue microarrays obtained from US Biomax or GeneTex.The use of excess deidentified archival specimens does not require informed consent from the subjects.
All procedures involving animals were conducted in accordance with the US National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Massachusetts Institute of Technology Committee on Animal Care.Male and female 12-16 week old, wild type (Swiss Webster) mice were used in this study.Mice were deeply anesthetized with isoflurane and perfused transcardially with ice cold phosphate buffered saline (PBS) followed by ice cold 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS).Brains were harvested and postfixed in the same fixative solution at 4°C overnight.Fixed brains were incubated in 100 mM glycine for 1-2hrs at 4°C and sectioned to 10 or 100 μm-thick slices.
For tissue samples that were designated for pre-expansion immunostaining, following format conversion, we applied antigen retrieval to enhance immunostaining (6,41,42).Antigen retrieval was performed by incubating tissues in either the softening buffer (20% (weight/volume) sodium dodecyl sulfate (SDS), 100 mM β-mercaptoethanol, 25 mM ethylenediaminetetraacetic acid (EDTA) and 0.5% Triton-X in 50 mM Tris at pH 8 at RT for 1 hr (Fig. 1A) or by microwave heating for 1 min in 5mM citric acid buffer, 0.5% Triton-X, pH 6, because it provided improved collagen staining (fig.S4A).Antigen retrieval was then followed by 3 washes in 1x PBS for 5 min each and blocking at 37°C for 30 min with MAXblock blocking buffer (Active Motif, #15252) (6).Immunostaining was performed by diluting primary antibody in MAXbind Staining buffer (Active Motif, #15253), and incubating tissue samples in the antibody solution at 37°C for 1 hr, at RT for 2.5 hr or at 4°C overnight.The same procedure conditions were applied for secondary antibodies.Primary and secondary antibodies used in this work are listed in Table S2.All pre-expansion stained tissues were immersed in VectaShield mounting media (Vector Laboratories, #H-1000-10) and covered with a No. 1 coverslip prior to imaging Anchoring and gelation-Anchoring and gelation were performed according to previously published protocols (6,19), and briefly summarized below.Acryloyl-X (a.k.a.6-((acryloyl)amino)hexanoic acid, succinimidyl ester, here abbreviated AcX, Thermo Fisher Scientific, #A20770) powder was dissolved in anhydrous dimethyl sulfoxide at a concentration of 10 mg/ml and stored in aliquots in a desiccated environment at −20°C.Tissues underwent anchoring by incubation with AcX at a concentration of 0.1 mg/ml in 1x PBS with 0.5% Triton-X, at 4°C for 30 min, followed by 1.5 hrs at 37°C, and then x3 washes with 1x PBS at RT for 5 min each.Next, a monomer solution composed of 1× PBS, 2 M sodium chloride (NaCl), 8.625% (w/v) sodium acrylate, 2.5% (w/v) acrylamide and 0.10% (w/v) N,N′-methylenebisacrylamide (Sigma-Aldrich) was prepared, aliquoted and stored at −20°C.Gelling solution was prepared by mixing the monomer solution with the following chemicals, in the order shown: (1) 4-hydroxy-2,2,6,6-tetramethylpiperidin-1oxyl (abbreviated as 4-HT; final concentration, 0.01% (w/v)) as an inhibitor of gelation, (2) tetramethylethylenediamine (abbreviated as TEMED; final concentration, 0.2% (w/v)) as an accelerator of gelation, and (3) ammonium persulfate (abbreviated as APS; final concentration, 0.2% (w/v)) as an initiator of gelation.Tissue sections on glass slides were covered with gelling solution, and then a gel chamber was constructed by first placing two No. 1.5 square coverslips (22 mm x 22 mm) as spacers, one at each end of the glass slide and flanking the tissue section in the middle; then, a rectangular coverslip is placed on top of spacers, to enclose the gel chamber, in which the tissue sample is fully immersed in the gelling solution and sandwiched by the glass slide and the top coverslip.Samples were first incubated in a humidified atmosphere at 4°C for 30 min, which slows down gelation rate and enables diffusion of solution into tissues and subsequently incubated in a humidified atmosphere at 37°C for 2.5 hrs to complete gelation (Fig. 1B Softening-After gelation, all coverslips were gently removed from the glass slide that carries the gelled tissue.Excessive gel around the tissue sample was trimmed away using a razor blade.Then, tissues were incubated in the softening buffer, which consists of 20% (w/v) SDS, 100 mM β-mercaptoethanol, 25 mM EDTA, 0.5% Triton-X, and Tris 50 mM at pH 8, at 37°C for 30 min followed by 1 hr in an autoclave at 121°C, followed by cooling to RT for 30 min (Fig. 1C

Expansion with post-decrowding methoxy-xO4 or NHS ester staining-After
softening, tissue underwent either decrowding (Fig. 1D; fig.S1D), expansion with postexpansion methoxy-O4 staining, expansion without post-expansion immunostaining (fig.S3D), or decrowding with NHS ester post-expansion staining (fig.S9).After softening, the gelled tissue has detached from the slide and is floating freely in the softening buffer.The tissue is transferred into a clear polystyrene petri dish plate by slowly decanting the buffer solution which contains the gelled tissue into the plate.Then, using a pipette, the excess buffer is removed and discarded.Then 1x PBS is added to the well plate to fully cover the tissue and the petri dish plate is gently shaken at RT to remove excess softening buffer.Then while the gelled tissue was free floating in 1x PBS we used a flat, wide mini paintbrush which we then placed underneath the gelled tissue, ensuring that the paint brush was covering most of the gelled tissue undersurface area, and transferred it into a clear 6-well plate (Clearstar) that contained 1x PBS to completely submerge the tissue.Then the well plate was gently shaken at RT for 3 min.Then the excess 1x PBS was removed using a pipette and new 1x PBS was added to cover the tissue and the well plate gently shaken at RT for 3 min.This process was repeated a total of 5 times, which results in tissues reaching an expansion factor of ~2.3x (Fig. 1D; fig.S1D; fig.S2E; fig.S3E).For decrowding, tissues were washed 5 times with 1x PBS at RT for 3 min each.At this stage, tissues were at a partially expanded state, with ~2.3x linear expansion factor.For expansion with methoxy-x04 (Biotechne, #4920), tissues were stained with methoxy-x04 at a concentration of 0.01 mg/ml in 1x PBS for 2 hours at RT, and then additionally washed in deionized water 3-5 times at RT for 3 min each, expanding the hydrogel-embedded tissue to an expansion factor of ~4x (6,15,19).For decrowding with NHS ester post-decrowding staining, we used Alexa Fluor 647 succinimidyl ester at the same concentration and conditions as with pre-expansion staining.For expansion without post-expansion immunostaining, tissues were then additionally washed in deionized water for 3-5 times at RT for 3 min each, to expand the hydrogel-embedded tissue to an expansion factor of ~4x (6,15,19) (fig.S3D).A subset of tissue samples was imaged by confocal microscopy at this state, with methods described in the section Image Acquisition, to obtain the post-expansion, pre-decrowding staining images.
Post-decrowding immunostaining-Tissues underwent decrowding by washing 5 times with 1x PBS at RT for 3 min which results in an expansion factor of ~2.3x (Fig. 1D; fig.S1D; fig.S2E; fig.S3E).Then, while the gelled tissue was free floating in 1x PBS, we used a flat, wide mini paintbrush placed underneath the gelled tissue, ensuring that the paint brush was covering most of the gelled tissue undersurface area, and transferred it into a new 6-well plate (CellVis) that was subsequently used for imaging containing 1x PBS.We prepared antibody solutions by diluting primary antibody in MAXbind Staining buffer (Active Motif, #15253) and performed post-decrowding immunostaining (Fig. 1E S3F) by incubating tissue samples in the antibody solution at 37°C for 1 hr, at RT for 2.5 hr or at 4°C overnight.Next, excess antibody solution was removed with a pipette and the tissues were washed 3 times with MAXwash buffer (Active Motif, #15254) at RT for 3 min, each time removing excess MAXwash buffer with a pipette.The same procedure conditions were applied for secondary antibodies.For tissues that underwent both pre-expansion and post-decrowding staining, antibody concentrations and incubation conditions were identical to ensure quantitative comparisons pre-and post-expansion.Immunostained tissues were imaged at ~2.3x linear expansion or further expanded by washing with deionized water at RT for 3-5 times for 3 min each to ~4x linear expansion (Fig. 1E; dig.S3F).Then, excess deionized water was removed from the well plate, and the gelled tissue was ready for imaging with a confocal microscope by imaging through the clear glass of the well plate.We performed confocal microscopy at the ~4x state, adhering to the protocol described in the Image Acquisition section, to obtain post-decrowding staining images.Once imaged, tissue can be stored by adding 1x PBS to completely cover tissue, also ensuring the container is covered so tissue does not dry out.Fully ~4x expanded gelled tissues are friable and can easily break if the user tries to lift them up and out of the well plate with a paint brush.To transfer gels between containers do so in the ~2.3xstate by washing ~4x expanded gelled tissues in 1x PBS 5 times at RT.A step by step protocol is found in Supplementary Materials (S1).
Tissue shrinking-We shrunk tissues to a ~1.3x linear expansion factor by treating with in a high-ionic-strength buffer (1M NaCl + 60 mM MgCl2) (19) following the softening (fig.S3C) and washing with PBS (which results in an expansion factor of ~2.3x) or following post-decrowding staining and washing with PBS (which results in an expansion factor of ~2.3x).Specifically, we washed the tissues 3-5 times with this buffer at RT for 3 min each, until no more tissue shrinkage was observed.We then performed confocal microscopy at this stage, with methods described in the following section Image Acquisition, to obtain pre-decrowding or post-decrowding staining at shrunken state images.

Imaging processing methods
Image acquisition-For confocal imaging, we used a spinning disk confocal system (CSU-W1, Yokogawa) on a Nikon Ti-E microscope.The objective lenses that we used include a 40× 1.15 NA water immersion objective or 10× 0.20 NA air objective.The excitation lasers and emission filters that we used to image each fluorescent dye are the following: 405 nm excitation, 450/50 nm emission filter; 488 nm excitation, 525/40 nm emission filter; 561 nm excitation, 607/36 emission filter; 640 nm excitation, 685/40 emission filter.The following acquisition and display settings apply to all images shown in this study, unless otherwise specified: (1) within the same experiment (as grouped by figures and described in the Results and Figure Legends), all images were obtained with the same laser power, camera settings, and objective lens.(2) For all image display in all figures except Fig. 4, brightness and contrast settings were first individually set by the automated adjustment function in ImageJ, and then manually adjusted (raising the minimum-intensity threshold and lowering the maximum-intensity threshold) to improve contrast for features of interest.For image display of Fig. 4 brightness and contrast settings of images were adjusted so that 1% of the pixels were saturated.None of these changes in the brightness and contrast settings, throughout the entire study, affects the downstream quantitative analysis of fluorescent intensities, which were always applied on raw images, as specified in Results and captions.underwent format conversion, antigen retrieval, and immunostaining for GFAP and α-SMA (α-smooth muscle actin), and DAPI staining (fig.S3A).White box in (i) marks a region with sparse and discontinuous signals that is shown magnified and in separate channels at the right (GFAP in (ii) and α-SMA in (iii)).α-SMA-staining of pericytes that are enveloping blood vessels (dashed arrow in (i)).Discontinuous GFAP regions (dotted arrow in (ii)).Solid arrows in (i) and (ii) show regions that were GFAP-negative pre-expansion (G), for comparisons to post-expansion staining panels later in this figure .(H) As in B, but for panel G. (I) As in C, but for panel G. (J) As in D, but for the GFAP (cyan) and α-SMA (magenta) channels, in GFAP-positive ROIs.(K) As in D, but for the GFAP (cyan) and α-SMA (magenta) channels in α-SMA-positive ROIs.(L) As in D, but for the GFAP (cyan) and α-SMA (magenta) channels in double negative ROIs.(M) Representative (pre-expansion confocal image (single z slice) of 5-μm-thick human high grade glioma tissue (cortex or white matter).Sample underwent format conversion, antigen retrieval, and immunostaining for vimentin and α-SMA, and DAPI staining (fig.S3A).White box in (i) marks a region including part of a blood vessel that is shown magnified and in separate channels to the right (vimentin in (ii) and α-SMA in (iii)).Vimentin and α-SMA-staining of the blood vessel wall (dashed arrow in (i)) which surrounds the vessel lumen (asterisk in (i)).A vimentin-positive cell outside the blood vessel (dotted arrow in (i)).Solid arrows in (i) and (ii) show regions that were vimentin-negative pre-expansion (M), for comparison to post-expansion staining panels later in this figure.

Fig. 1 .
Fig. 1.Decrowding expansion pathology (dExPath) for post-expansion immunostaining of human tissue and other formaldehyde-fixed specimens.(A-E) Workflow for expanding formalin-fixed paraffin-embedded (FFPE) or formaldehydefixed human or mouse brain specimens.Key modifications of proExM/ExPath protocols are highlighted in green.PFA, paraformaldehyde; PBS, phosphate buffered saline; RT, room temperature; AcX, Acryloyl-X; SDS, sodium dodecyl sulfate.(A) Tissue samples undergo conversion into a state compatible with expansion.(B) Tissue samples are treated so that gel-anchorable groups are attached to proteins, then the sample is permeated with an expandable polyacrylate hydrogel.(C) Samples are incubated in a softening buffer to denature, and loosen disulfide bonds and fixation crosslinks between, proteins.(D) Softened samples are washed in a buffer to partially expand them.Linear expansion factor is shown in parentheses.(E) Samples are stained and then expanded fully by immersion in water.

Fig. 2 .
Fig. 2. Isotropy of dExPath.(A-B) Representative pre-expansion super resolution structured illumination microscopy (SR-SIM) images of healthy human hippocampus (A) and human cerebrum high-grade glioma brain tumor tissue (B) which underwent processing as in fig.S3A with staining for MAP2 and DAPI (A), or for vimentin and DAPI (B).(C-D) Post-expansion images of the same fields of view as in (A-B), respectively.Samples underwent anchoring, gelation and softening (as in fig.S3B-C), another round of DAPI staining, ~4x linear expansion (as in fig.S3D), and imaging with confocal microscopy.(E-F) Root-mean-square (RMS) length measurement errors obtained by comparing pre-and post-expansion images such as in A-D (n = 4 samples, each from a different patient, E; n = 3 samples, each from a different patient, F).Line, mean; shaded area, standard deviation.Images are sum intensity z-projections, either of SR-SIM (A-B), or confocal (C-D) image stacks, both covering an equivalent tissue depth in biological units.Brightness and contrast settings: first set by the ImageJ auto-scaling function, and then manually adjusted to improve contrast for the stained structures of interest; quantitative analysis in (E-F) was conducted on raw image data.Scale bars (in biological units: physical sizes of expanded samples divided by their expansion factors, used throughout this manuscript, unless otherwise noted): (A-D) 5 μm.Linear expansion factors: (C-D) 4.0x.

Fig. 3 .
Fig. 3. dExPath removal of lipofuscin autofluorescence.(A-C) Pre-expansion confocal images (single z slices) of a neuron in a a 5-μm-thick normal human cortex sample(format conversion as in Fig. 1A).Images were acquired for 3 fluorescent channel settings: (A) 488 nm ex/ 525 nm em; (B )561ex/607em; and (C) 640ex/685em.(D) Mean fluorescence intensities from pre-expansion images, averaged across regions of interest (ROIs) that exhibited prominent lipofuscin (left bar graph), as well as across background ROIs (right bar graph); (n = 4 tissue samples, each from a different patient).Brightness and contrast settings: first set by the ImageJ auto-scaling function, and then manually adjusted to improve contrast for lipofuscin; quantitative analysis was conducted on raw image data.Box plot: individual values (open circles; 3 measurements were acquired from each patient), median (middle line), mean (dotted line), first and third quartiles (lower and upper box boundaries), lower and upper raw values (whiskers).Statistical testing: 2-tailed paired t-test (non-Bonferroni corrected) was applied to lipofuscin vs. background, for pre-expansion mean fluorescence intensities for each spectral channel.*, p < 0.05; ns, not significant.(E-G) Shown are post-expansion confocal images after the sample from A-C was treated with anchoring, gelation,softening and decrowding (as in Fig. 1B-D), DAPI staining, and ~4x linear expansion, without post-decrowding immunostaining.Sum intensity z-projections of image stacks corresponding to the biological thickness of the original slice, taken under identical settings and of the same field of view as A-C and displayed under the same settings.(H) Mean fluorescence intensities, from post-expansion images, averaged across the same lipofuscin (left) and background (right) ROIs used in panel D. Plots and statistics as in D. (I-K) Confocal images as in (E-G), after post-decrowding immunostaining for MAP2 (microtubule-associated protein 2), giantin, and synaptophysin (labeled with antibodies in the same spectral ranges as indicated above A-C), as well as stained for DAPI (not shown; used for alignment), and then re-expanded to ~4x linear expansion.(L) Representative pre-expansion confocal image of a tissue sample

Fig. 4 .
Fig. 4. dExPath-mediated protein decrowding reveals cells and structure not detected in preexpansion staining forms of expansion microscopy.(A) Representative pre-expansion confocal image (single z slice) of 5-μm-thick FFPE normal human hippocampal tissue.(Sampleunderwent processing as in fig.S3A and immunostaining for MAP2 and GFAP.White box in (i) marks a region with sparse and discontinuous signals that is shown magnified and in separate channels at the right (MAP2 in (ii) and GFAP in (iii)).MAP2 staining of a putative cell body (asterisk in (i)) and dendrite (upper dashed arrow in (i)).GFAP staining of a putative astrocytic process (lower dashed arrow in (i)) and discontinuous GFAP regions (dotted arrows in (iii)).Solid arrows show regions that were MAP2-negative (ii) or GFAP-negative (iii) in pre-expansion images (A), for comparison to post-expansion staining panels later in this figure.(B) Shown is a post-expansion confocal image after processing as in fig.S3B-D and imaging at ~4x linear expansion.Sum intensity z-projection of an image stack covering the biological thickness of the original slice (used for all expanded images throughout this figure); images were of the same fields of view as in (A), using identical hardware settings.Asterisks and arrows as in (A).(C) Post-decrowding stained confocal images of the same fields of view as in (A-B) after decrowding and additional immunostaining for MAP2 and GFAP and re-expansion to ~4x (fig.S3E-F), using identical hardware settings.Asterisks and arrows as in (A).(D) Quantification of fluorescence intensities for raw data of images post-expansion such as those of B (NR, "not restained") and C (R, "restained"), averaged across MAP2-positive ROIs, for the MAP2 channel (cyan) and the GFAP channel (magenta).Box plot: individual values (open circles; 3 measurements were acquired from each patient), median (middle line), mean (dotted line), first and third quartiles (lower and upper box boundaries), lower and upper raw values (whiskers).Statistical testing: 2-tailed paired t-test (non-Bonferroni corrected) *, p < 0.05.(E) As in D, but for GFAP-positive ROIs, for the MAP2 channel (cyan) and the GFAP channel (magenta).(F) As in D, but for double negative ROIs, for the MAP2 channel (cyan) and the GFAP channel (magenta).(G) Representative pre-expansion confocal image (single z slice) of 5-μm-thick FFPE human high-grade glioma.Sample (N) As in B, but for panel M. (O) As in C, but for panel M. (P) As in D, but for the vimentin channel (cyan) and the α-SMA channel (magenta), in vimentinpositive ROIs.(Q) As in D, but for the vimentin channel (cyan) and the α-SMA channel (magenta), in α-SMA-positive ROIs.(R) As in D, but for the vimentin channel (cyan) and the α-SMA channel (magenta), in double negative ROIs.(S) Representative pre-expansion confocal image (single z slice) of 5-μm-thick human low grade glioma tissue (cortex or white matter).Sample underwent format conversion, antigen retrieval, and immunostaining for ionized calcium binding adapter molecule 1 (Iba1) and GFAP, and DAPI staining (fig.S3A).White box in (i) marks a region with sparse and discontinuous signals that is shown magnified and in separate channels to the right (Iba1 in (ii) and GFAP in (iii)).Iba1 staining of discontinuous regions (dotted arrow in (ii)).Solid arrows in (i) and (ii) show regions that were Iba1-negative pre-expansion (S), for comparison to post-expansion staining panels later in this figure.(T) As in B, but for panel S. (U) As in C, but for panel S. (V) As in D, but for the Iba1 channel (cyan) and the GFAP channel (magenta), in the Iba1-positive ROIs.(W) As in D, but for the Iba1 channel (cyan) and the GFAP channel (magenta), in GFAP-positive ROIs.(X) As in D, but for the Iba1 channel (cyan) and the GFAP channel (magenta), in the double negative ROIs.Scale bars: (A-C) panel i, 9 μm; ii, 1.7 μm; (G-I) i, 7 μm; ii, 0.7 μm; (M-O) i, 8 μm; ii, 0.8 μm; (S-U) i, 8 μm; ii, 0.8 μm.Linear expansion factors: (B,C) 4.1x; (H,I) 4.0x; (N,O) 4.3x; (T,U) 4.2x.

Fig. 5 .
Fig. 5. dExPath-mediated protein decrowding reveals cells and structures not detected by SR-SIM imaging of unexpanded tissues.(A-B) Representative pre-expansion SR-SIM images of FFPE 5-μm-thick human tissue (processed as in Fig. S3A).(A) High-grade glioma tissue stained for vimentin and DAPI.Solid and dashed white boxes in (i) mark two separate regions shown magnified in (ii) (solid box) and (iii) (dashed box), respectively.Dotted arrows mark regions that appear as punctate and discontinuous in pre-expansion SR-SIM images for vimentin in (ii) and (iii), and solid arrows mark regions that were negative for vimentin in (iii), for comparison to post-expansion staining panels later in this figure.(B) Normal human hippocampus tissue stained for MAP2, GFAP and DAPI.(A) Solid white box in (i) shown magnified in (ii) for MAP2 and in (iii) for GFAP.Arrows as in (A) but for MAP2 and GFAP, in their respective images.(C-D) Shown are representative samples used for (A-B) after processing for post-expansion imaging Fig. S3B-D) and not restained.Sum intensity z-projection of an image stack covering the biological thickness of the original slice (used for all expanded images throughout this figure); images were of the same fields of view as in (A-B).Arrows as in (A-B).(E-F) Images of the same fields of view as in (A-B) after decrowding and additional restained for vimentin (E), or MAP2 and GFAP (F), followed by DAPI staining and re-expansion to ~4x (Fig. S3E-F), imaged using identical hardware settings as in (C-D).Arrows as in (A-B).Brightness and contrast settings in images (A-F): first set by the ImageJ auto-scaling function, and then manually adjusted to improve contrast for stained structures.Scale bars (in biological units): (A, C, E) left column, 8.3 μm; middle and right columns 840 nm; (B, D, F) left column, 6.0 μm; middle and right columns 500 nm.Linear expansion factors: (C) 4.1x; (D) 4.3x; (E) 4.1x; (F) 4.2x

Fig. 6 .
Fig. 6. dExPath removes autofluorescence from amyloid plaques and preserves detection of disease markers in Alzheimer's disease (AD).(A) Representative pre-expansion confocal image (single z slice) of an amyloid β plaque within a FFPE 5-μm-thick sample of AD human cortex.The samples underwent processing as in Fig. 1A.(i) Images were acquired for the fluorescent channel setting of 488ex/525em.Solid arrow points to an amyloid β plaque.(ii) Mean fluorescence intensities from preexpansion images, averaged across regions of interest (ROIs) taken at amyloid β plaque ([plaque], left bar) and background ROIs ([Bck], right bar); Brightness and contrast settings: first set by the ImageJ auto-scaling function, then manually adjusted to improve contrast for

Fig. 7 .
Fig. 7. dExPath reveals previously undetected cells defined by single and multiple markers of importance to glioma biology.(A)Representative pre-expansion confocal image (single z slice) of a 5-μm-thick FFPE human low-grade glioma specimen.Sample was immunostained for vimentin, Iba1 and GFAP, and DAPI staining (Fig.S3A).Left panel, overlay of all 4 channels; right three panels, individual channels (not including DAPI).Dotted arrows show regions that were vimentin and GFAP negative in pre-expansion images, and solid arrows show regions that were Iba1, GFAP and vimentin negative in pre-expansion images, for comparison to post-decrowding staining panels later in this figure.(B) Sample used for (A) after anchoring, gelation), softening (fig.S3B-C), washing with PBS (which results in an expansion factor of ~2.3x), tissue shrinkage (via adding salt) to ~1.3x of the original size, and imaging.Single z slice image centered at the same midpoint of the original slice; images were of the same field of view as in (A), using identical hardware and software settings.Arrows as in (A).(C) Sample used for (B) after expansion (fig.S3D) for imaging at ~4x linear expansion.Sum intensity z-projection of an image stack covering the biological thickness of the original slice; images were of the same field of view as in (A), using identical hardware and software settings.Arrows as in (A).(D) Post-decrowding stained confocal images of the same field of view as in (A) after decrowding and additional immunostaining for vimentin, Iba1, and GFAP, tissue shrinkage (fig.S3E-F) and imaging at shrunken state.Arrows as in (A).(E) Sample used for (D) after expansion (fig.S3D) for imaging at ~4x linear expansion.Arrows as in (A).(F) Pixel level analysis of the percent of single or double positive stained pixels, from pre-expansion (gray boxes) and post-decrowding at shrunken state (white boxes) images, for vimentin (V), Iba1 (I), GFAP (G), Iba1 and vimentin (I&V), vimentin and GFAP (V&G), and Iba1 and GFAP (I&G).Values represent the percentage of positive pixels among all pixels in the field of view with 3 different values per sample each corresponding to a different field of view Box plot: individual values (open circles; 3