Non-reversible tissue fixation retains extracellular vesicles for in situ imaging

Abstract

Extracellular vesicles (EVs) are secreted nanosized particles with many biological functions and a broad range of pathological associations. A major technical limitation to understanding the role of EVs in normal and diseased specimens has been the inability to visualize the spatial localization of EVs in tissue microenvironments. Here, we use bovine ocular tissue, the vitreous humor, as a model system to study EV imaging. We show that mammalian tissues crosslinked with conventional formaldehyde solutions result in significant EV loss, with subsequent reduced or negative EV signals; however, EV escape can be prevented by additional fixation with 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) that permanently holds these nano-sized particles and allows for visualization of EVs in normal and cancer tissues in situ.

Extracellular vesicles (EVs) are natural transport nanovesicles implicated in intercellular communication via transfer of biomolecules such as proteins, lipids, and RNA¹. Many cell types secrete exomeres (~35 nm)², exosomes (40–100 nm), larger microvesicles (100–1000 nm) or apoptotic bodies (1–5 μm)³ into fluids like breast milk, urine, and blood⁴. Pathophysiological functions for EVs are being implicated in diseases including cancers^{5, 6}.

Recent advances in EV imaging in living animals include using fusion proteins⁷, CRE recombinase with reporter proteins⁸, or multiphoton microscopy⁵. Yet, a major technical challenge in understanding EV biology is the inability to reproducibly image EVs in situ⁹. Identifying and solving for technical pitfalls that hinder EV imaging may help elucidate the structure and function of EVs in normal and diseased states.

To study EVs in tissues, we used the vitreous humor of the eye as a model system. The vitreous, located between the lens and the retina, is an optically clear, paucicellular tissue with abundant extracellular matrix¹⁰ (ECM). Vitreous EV-associated microRNAs have been described¹¹; however, normal vitreous EVs have not yet been imaged nor characterized in situ. We isolated bovine vitreous EVs and determined their concentration using nanoparticle-tracking analysis (NTA)¹², which revealed an EV concentration of at least 2.98 x 10⁷ ± 8.98 x 10⁶ particles per ml (mean ± s.e.m, Supplementary Fig. 1a). However, our repeated attempts to visualize vitreous EVs in situ using multiphoton (MPM), confocal or wide-field microscopy failed. Here, we show that standard formalin fixation results in loss of EVs from tissue due to temperature-dependent crosslink reversal, whereas fixation of proteins with 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) retains EVs and allows for EV imaging in situ.

Our studies focused on optimizing tissue fixation. Conventional fixation methods use 10% formalin to create protein-protein and protein-nucleic acid crosslinks. Tissue processing steps generally occur at or above room temperature; however, elevated temperatures are known to revert formalin crosslinks^{13, 14}. Therefore, we hypothesized that EVs escape from formalin-fixed tissues at elevated temperatures. To examine EV loss, we immersed formalin-fixed bovine vitreous in wash buffer at 37°C and collected the supernatant. We characterized EVs present in the wash buffer using transmission electron microscopy (TEM) or nanoparticle tracking analysis (NTA) (n = 3 tissues, Supplementary Fig. 1b–c). We examined the wash buffer with TEM, and EV loss was noted at incubation temperatures of 25°C or above. Fewer EVs escape at 4°C (Supplementary Fig. 1d–f). To permanently retain EVs within the tissue, we added a second fixation step using EDC to create a non-reversible crosslink between positively-charged amino group side chains and carboxyl groups of proteins. Under similar conditions, formalin-EDC fixation showed no evidence of EV loss to wash buffer (n = 3 tissues, Supplementary Fig. 1g–i). NTA also revealed substantial EV loss to wash buffers in formalin-fixed specimens, whereas for formalin-EDC-fixed specimens, the signal was below the detection threshold (n = 3 tissues, Supplementary Fig. 1j–l; Supplementary Table 1 and 2; Supplementary Video 1). These data suggest that formalin-fixed specimens experience temperature dependent EV loss, while additional fixation with EDC fixation retains EVs in tissues.

To visualize EVs in the ECM of vitreous tissue (Supplementary Fig. 2a), we compared conventional fixation (formalin alone) to formalin-EDC crosslinking, and then attempted to image EVs in situ. EVs are known to contain proteins; thus, we labeled protein in whole mounted specimens with carboxyfluorescein succinimidyl ester (CFSE) fluorescent dye in whole mounted specimens and then imaged with multiphoton microscopy. Formalin-fixed vitreous showed protein signal within cells but showed no extracellular signal (n = 3 tissues, Fig. 1a), suggesting that EVs were either absent or lost during processing. In contrast, formalin-EDC-fixed vitreous show a robust EV-shaped signal in the ECS (n = 3 tissues, Fig. 1b–c). Significantly more EVs were identified in formalin-EDC-fixed tissues (143.2 ± 23.8, EVs counted per image, mean ± s.d.), when compared to formalin-fixed tissues (1.2 ± 0.9 EVs counted per image, mean ± s.d., n = 3 tissues and 3 images recorded per tissue, P = 0.0007, Supplementary Fig. 2b). Vitreous EVs show a heterogeneous population of microvesicles and multivesicular bodies based on MPM imaging (Supplementary Fig. 2c). Similar findings were observed using standard wide-field fluorescent microscopy (n = 3 tissues, Supplementary Fig. 3). Furthermore, we noted similar intensity of CSFE signal in cells fixed with formalin or formalin-EDC, suggesting that the extracellular signal observed in figure 1b–c was independent of preferential binding of CSFE to EDC-fixed tissue (n = 3 cell cultures, Supplementary Fig. 4). To correlate the in situ optical microscopy findings with other methods used to visualize EVs, we studied vitreous EV ultrastructure using TEM¹⁵. Bovine vitreous tissue sections showed a substantial number of EVs in the ECM of the vitreous (Fig. 1d). Isolated bovine vitreous EVs stained with CFSE show intra-vesicular protein signal (n = 3 EV isolates, Fig. 1e). Moreover, we performed TEM on post-mortem human eyes and identified numerous EVs in the ECM near the vitreous base and ciliary body (Fig. 1f, g). These data suggest that EVs are present in the ECM of the vitreous. Here, we show that formalin-EDC fixation allows for imaging EVs in situ.

Figure 1: 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) fixation of bovine vitreous retains extracellular vesicles (EV), when compared to formalin fixation alone.

(a) Representative multiphoton microscopy (MPM) z-stack photomicrographs of whole mount bovine vitreous specimens fixed with formalin alone and stained with CFSE to mark protein (orange) and Hoechst to mark nuclei (purple). CFSE signal is observed surrounding the nuclei (left panel, open arrow), but not in the extracellular space (right panel, purple, open arrow). Nuclei staining show no extracellular signal (left panel, purple, open arrow). (b) Representative MPM z-stack photomicrographs of formalin-EDC-fixed vitreous stained with CFSE (orange) and Hoechst (purple). Overlay of image shows positive signal consistent with cell bodies (open arrowhead) and foci of extracellular protein signal (closed arrowheads, orange) consistent in size and shape with EVs. (c) Inset of b (white box), shows multiple round intracellular foci (left panel, open arrowhead, orange) surrounding the area of nuclear stains (right panel, open arrowhead, purple). Numerous focal extracellular protein signals are also observed (left panel, closed arrowheads, orange), consistent in size and shape with EVs, and no extracellular DNA is observed. (d) Representative transmission electron microscopy (TEM) photomicrographs of bovine vitreous tissue sections stained with uranyl acetate (UA) and lead citrate show a substantial number of EVs that are heterogeneous population size (arrowhead). The inset (upper right corner) is an enlargement of the area-enclosed box in the lower right corner and shows an EV (arrowhead). (e) Representative TEM photomicrograph of EVs isolated from bovine vitreous and stained with the electron dense protein stain, CSFE, depict EV morphology, with both smaller (arrowhead) and larger EVs (double arrowhead). (f, g) Representative TEM photomicrographs of postmortem human eye sections stained with UA and lead citrate show a substantial number of EVs in the extracellular matrix near the vitreous base (Vit), adjacent to the non-pigmented epithelium (NPE) of the ciliary body (smaller EVs marked with arrowhead, larger EVs with double arrowhead). Scale bars are (a) 40 μm, (b) 50 μm and (c) 10 μm, (d) 100 nm, (e) 200 nm, (f) 2 μm, (g) and 100 nm.

EVs are known to contain extracellular RNA¹⁶; therefore, we stained bovine vitreous nucleic acids with propidium iodide (PI), which marks DNA and RNA¹⁷. Confocal microscopy imaging of formalin-EDC-fixed vitreous stained with PI and CFSE show signals positive extracellular RNA and extracellular protein (n = 3 tissues, Fig. 2a–b). In contrast, fixation with formalin alone resulted in substantially less extracellular RNA and protein signal (Fig. 2c). The reduced extracellular RNA signal in formalin-fixed tissues is consistent with our prior studies demonstrating loss of microRNAs from formalin-fixed tissue specimens into wash buffers¹⁴. Moreover, PI staining is not adversely influenced by formalin-EDC fixation (n = 3 cell cultures, Supplementary Fig. 5). To support that the extracellular PI signal was RNA, we treated formalin-EDC-fixed vitreous with RNase, and noted a substantial reduction in extracellular signal (Supplementary Fig. 6). Interestingly, normal vitreous EVs express RNA, but show no DNA signal. These data support that formalin-EDC fixation enables evaluation of the differential expression of extracellular nucleic acids in situ.

Figure 2: Fixation of bovine vitreous with formalin-EDC retains vitreous extracellular vesicles (EV) and extracellular RNA in situ.

(a-b) Representative confocal fluorescent photomicrographs of formalin-EDC fixed whole mount bovine vitreous specimens stained with propidium iodide (PI, red) to mark DNA and RNA, Hoechst (blue) to visualize DNA and nuclei, and carboxyfluorescein succinimidyl ester (CFSE, green) to stain for protein. (a) Overlay of images from formalin-EDC-fixed bovine vitreous show positive signal consistent with cell bodies (a, open arrow) and foci of extracellular RNA (closed arrowhead) and extracellular protein (closed arrowhead) consistent in size and shape with EVs. (b) Representative confocal fluorescent photomicrographs of formalin-EDC-fixed vitreous show multiple round cellular foci (both panels, open arrowhead) and numerous focal signals of extracellular RNA (left panel, closed arrowhead, PI stain, red) and extracellular protein (right panel, closed arrowhead, CFSE stain, green) between the cells. (c) Representative photomicrographs of whole mount bovine vitreous fixed with formalin alone show signal for RNA (left panel, open arrowhead, PI, red) in the nucleus, similar to nuclei staining (middle panel, open arrowhead, Hoechst, blue). Formalin-only fixed vitreous show no foci of extracellular RNA signal (left panel). CFSE stain shows cellular protein signal (right panel, open arrow), but no EV-shaped extracellular protein signal (right panel, green, no punctate staining observed between open arrows). The cell size appears smaller in the formalin only fixation, presumably due to improved retention of cytoplasmic RNAs and protein with formalin-EDC fixation as compared to formalin fixation alone. Scale bars are (a) 25 μm, (b, c) 50 μm.

To broaden the usefulness of this technique, we imaged EVs in other tissues. First, we studied a murine metastatic breast cancer model by orthotopically transplanting 4T1 mouse mammary carcinoma tumor cells into the mammary fat pad of a mouse¹⁸, dissected the tumor, fixed the sample with formalin-EDC, labeled nucleic acids, and surveyed the tumor surface using MPM. The data showed extracellular RNA signal in the ECM and a heterogeneous population of EVs (n = 3 tissues, Fig. 3a). Moreover, extracellular DNA was detected within a subpopulation of larger EVs (Fig. 3b), consistent with other laboratories’ findings that extracellular DNA is present within tumor-derived EVs^{6, 19}. Remarkably, formalin-EDC fixation technique enabled the spatial localization of nucleic acid expression in a subpopulation of EVs within a tissue. Ultrastructural analysis of formalin-EDC-fixed mammary tumor tissues showed a heterogeneous population of EVs in the ECM near the tumor cell surface (Fig. 3 c, d). Tumors are known to secrete more EVs compared to normal tissues⁹, therefore, we compared the mammary tumor model to normal mouse mammary gland tissue. The results show a reduced number of EVs in normal tissues when compared to substantially more EVs present in mammary carcinoma tumor tissue sections (n = 3 tissues and 5 images captured per tissue, Supplementary Fig. 7). Finally, in a third tissue, formalin-EDC fixation preserved EVs near the surface of cultured human cells (n = 3 cell cultures, Supplementary Fig. 8 and Supplementary Table 3). These data support that formalin-EDC fixation retains EVs and allows for EV imaging in cancer specimens and other tissues.

Figure 3: Formalin-EDC fixation of metastatic breast cancer model allows for imaging of tumor extracellular RNA and extracellular DNA.

(a-b) Representative MPM z-stack photomicrograph of an formalin-EDC-fixed 4T1 mouse mammary carcinoma tumor cell line that was transplanted into the mammary fat pad of a mouse (syngeneic graft) shows EV-shaped extracellular RNA signal in the extracellular space (closed arrowheads). Tumors were dissected, fixed with formalin-EDC, and nucleic acids were labeled with PI (white signal only), DNA stained (Hoechst, blue), and images were captured near the tumor surface within the extracellular matrix. (a, right) An overlay image shows signal from a cell (open arrowhead, Hoechst, blue) and numerous foci of extracellular RNA (closed arrowhead, PI, white) between the cells consistent in size and shape with EVs. The photomicrographs show a heterogeneous population of EVs and highlighted are a small microvesicle (single arrowhead, ~270 nm), medium microvesicle (double arrowhead, ~850 nm) and an apoptotic body (arrowhead with asterisk, ~1.7 μm). PI (a, right, upper panel) and nuclei signal (a, right lower panel) are shown. (b) Representative MPM z-stack photomicrographs of an formalin-EDC-fixed 4T1 mouse mammary carcinoma tumor shows signal from cell (open arrowhead, Hoechst, blue) as well as co-localization of PI (RNA and DNA, white) with the DNA stain (Hoechst, blue) in the extracellular space (closed arrowhead), (c) Representative transmission electron microscopy (TEM) photograph of a formalin-EDC -fixed 4T1 mouse mammary carcinoma tumor shows a heterogeneous population of EVs (arrowhead) adjacent to a cell (labelled, Cell). Larger EVs are shown (double arrowhead, ~510 nm) and an exosome is marked (single arrowhead, ~146 nm). (d) Representative TEM photograph shows an EV (arrowhead, ~373 nm) connected to a cell membrane. Scale bars are (a, b) 5 μm, (c) 250 nm, and (d) 125 nm.

To determine if vitreous EVs expressed extracellular-vesicle-associated proteins, we conducted proteomic analysis using liquid chromatography mass spectrometry (LC-MS) comparing whole bovine vitreous with the EV isolated fraction. The data show that EV-associated proteins like TSG-101 were enriched in the EV fraction (Supplementary Table 4). Proteins implicated in ocular function and disease were also enriched in the EV fraction (Supplementary Table 5). To confirm that extracellular protein signals observed in formalin-EDC-fixed vitreous were indeed EVs, we conducted immunohistochemistry (IHC) for TSG-101. In our experience, formalin-EDC fixation is incompatible with IHC. Moreover, TSG-101 signal was not reliably detected in formalin-fixed tissues processed at room temperature. Since formalin crosslink reversal is temperature dependent¹³, we performed IHC at 4°C and immediately imaged the samples on a wide-field microscope at room temperature. Punctate TSG-101-positive signals were visualized in the ECM (n = 3 tissues, Supplementary Fig. 9a, d), consistent with the spatial distribution of CFSE-stained EVs in formalin-EDC-fixed tissues. Specificity controls showed no extracellular signal (Supplementary Fig. 9b–c). Unlike vitreous fixed with formalin-EDC (Fig. 2a, b), formalin-fixed samples showed no extracellular nucleic acid signal (Supplementary Fig. 9e), presumably from reversion of formalin nucleic acid cross-links¹⁴. These data show that vitreous EVs contain markers consistent with well-established EV studies¹⁵.

We sought to characterize vitreous EVs and determine if these EVs can transfer their RNA and protein cargo into target cells²⁰. We labeled bovine or human vitreous EV RNA with acridine orange (AO), purified the EV fraction to ensure the isolate was cell-free (Supplementary Fig. 10), and exposed retinal pigment epithelium cells (ARPE-19) to a bolus of the labeled EVs. We observed high transfection rates with bovine or human vitreous EVs, both of which were significantly higher than controls (n = 3 cell cultures, P<0.0005, Supplementary Fig. 11). EVs are also known to function as a vector to deliver recombinant proteins. Thus, we loaded bovine serum albumin (BSA, 66 kD) conjugated to fluorescein into bovine vitreous EVs via electropermeabilization, and treated ARPE-19 cells with the loaded bovine vitreous EVs. We observed robust transfection efficiency (n = 3 cell cultures, Supplementary Fig. 12a–c, P<0.0005). The controls resulted in no signal, demonstrating that uptake of BSA-fluorescein is EV-dependent. To evaluate whether vitreous EVs can transfect a functional protein, which must retain its conformational state to fluoresce, we loaded recombinant green fluorescent protein (GFP, 26.9 kD) into bovine vitreous EVs. ARPE-19 cells were transfected up to 88.3% of cells (n = 3 cell cultures, P<0.005 Supplementary Fig. 12d–e), significantly more than controls. These data show that vitreous EVs are capable of transferring RNA and recombinant protein in vitro.

Finally, we investigated vitreous EV transfection in vivo. We administered a dilute concentration of EVs loaded with BSA-fluorescein to rodent eyes through intravitreal injection. On day 3, EVs showed no evidence of retinal penetration (n = 3 injections, Supplementary Fig. 13a). At 3 weeks, we observed transfection of multiple retinal cell layers in vivo (Supplementary Fig. 13b–c). Specificity controls, PBS alone (Supplementary Fig. 13d) or EV samples mixed with BSA-fluorescein without electropermeabilization were negative. These data show that the vitreous EVs function as a vector for recombinant protein delivery in vivo.

In summary, conventional formalin fixation-based techniques are at risk for crosslink reversal as a function of temperature, and allow for EV escape from tissues, that result in negative signal. However, formalin-EDC fixation significantly improves retention of EVs in tissues and permits robust EV imaging in tissues in situ. This method illuminated a previously unidentified network of functional EVs in normal vitreous humor, a tissue long considered to have few biological functions. Finally, this fixation technique may be broadly applied for diagnostic purposes for diseases mediated by EVs such as cancer.

Data availability statement:

All data generated or analyzed during this study are included in this published article or its supplementary information files. The mass spectrometry data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD015234. The life sciences reporting summary, can be found in the online version of the manuscript.