X-34,A Fluorescent Derivative of Congo Red: A Novel Histochemical Stain for Alzheimer's Disease Pathology

Abstract

X-34, a lipophilic, highly fluorescent derivative of Congo red, was examined as a histochemical stain for pathological changes in Alzheimer's disease (AD). X-34 intensely stained neuritic and diffuse plaques, neurofibrillary tangles (NFTs), neuropil threads, and cerebrovascular amyloid. Comparison to standard methods of demonstrating AD pathology showed that X-34 correlated well with Bielschowsky and thioflavin-S staining. X-34 staining of NFTs correlated closely with anti-TAU antibody staining. A 1:1 correspondence of X-34 and anti-Aβ antibody staining of plaques and cerebrovascular amyloid was observed. Both X-34 and thioflavin-S staining were eliminated by formic acid pretreatment, suggesting that β-sheet secondary protein structure is a necessary determinant of staining. X-34 may be a general amyloid stain, like Congo red, because it also stains systemic amyloid deposits due to λ-light chain monoclonal gammopathy. In conclusion, X-34 is a highly fluorescent marker for β-sheet structures and intensely labels amyloid plaques, NFTs, neuropil threads, and vascular amyloid in AD brains. It can be used with both paraffin-embedded and frozen tissues as well as in combination with immunohistochemistry for double labeling. The intensity of staining and the simplicity and reproducibility of the technique suggest that it may be a useful addition to the standard techniques for evaluation of AD neuropathology.

Congo red has been used for decades to identify amyloid deposits in post-mortem tissue (Bennhold 1922; Puchtler et al. 1962). Structurally, Congo red is an azo dye that contains naphthalenesulfonic acid groups (Figure 1). Our laboratory has been working for several years to develop amyloid-binding Congo red analogues as in vivo neuroimaging agents (Klunk et al. 1994a, 1995). Both the azo groups (nitrogen-to-nitrogen double bonds) and the sulfonic acid groups of Congo red have disadvantages for in vivo applications. We have therefore developed derivatives of Congo red without these groups. We have previously reported that Chrysamine-G (Figure 1), an azo dye with less acidic, more lipophilic salicylic acid groups in place of the very acidic naphthalenesulfonic acid groups of Congo red, binds to the amyloid β-protein (Aβ) of Alzheimer's disease (AD) brain. Chrysamine-G is approximately 100 times more lipophilic than Congo red and enters the brain to a greater extent, making it a potential in vivo probe for amyloid (Klunk et al. 1994a, 1995). Because of its poor optical properties, Chrysamine-G only weakly stains neuritic plaques and cerebrovascular amyloid in postmortem tissue (Klunk et al. 1995). X-34 (1,4-bis(3-carboxy-4-hydroxyphe-nylethenyl)-benzene) (Figure 1) is another derivative of Congo red in which the naphthalenesulfonic acids are replaced by salicylic acids. Unlike Chrysamine-G, in X-34 the azo groups (N = N in Figure 1) present in Congo red are replaced by carbon-to-carbon double bonds (C = C in Figure 1). This latter change renders X-34 very highly fluorescent, and the salicylic groups confer lipophilic properties similar to Chrysamine-G. Because of the fluorescent properties and the likelihood that X-34 would stain amyloid similarly to Congo red, we explored the use of X-34 as a tissue stain for amyloid (Klunk et al. 1997).

Figure 1

Structures of Congo red, Chrysamine-G, and X-34. Selected protons are labeled on X-34 for correlation with NMR data (see text). Equivalent protons are omitted for clarity.

This study explores the properties of X-34 compared to standard staining methods used in the study of AD neuropathology, including antibodies to Aβ and TAU protein, Bielschowsky silver stain, and thioflavin-S. In addition, we define the optimal staining conditions, including optimal staining concentrations and staining times and optimal filter sets for fluorescence microscopy. Finally, we explore the specificity of this dye for β-sheet fibrils.

Materials and Methods

Synthesis of X-34

X-34 was synthesized using standard organic chemistry techniques through a Horner–Emmons reaction (March 1992) of 5-formylsalicylic acid (2.2 equivalents) (Aldrich; Milwaukee, WI) and p-xylylenediphosphonic acid tetraethyl ester (1 equivalent) (TCI America; Portland, OR) in dry dimethyformamide in the presence of potassium tert-butoxide (10 equivalents) at room temperature for 16 hr. After dilution of the reaction mixture 1:10 in water, the pH was adjusted to 5.0 and X-34 was extracted into ethyl acetate. The organic layer was concentrated on a rotary evaporator and X-34 was recrystallized by the addition of hexane. Structure was verified by NMR (500 MHz Bruker AM500) and mass spectrometry (see Results). Purity was verified by these techniques as well as reverse-phase HPLC (Vydac 214TP54 reverse phase C4 column; linear gradient of 20–50% acetonitrile with balance being 5 mM sodium phosphate buffer, pH 7.0, over 10 min at a flow rate of 1.4 ml/min; UV detection at 365 nm). Fluorescence spectra were obtained on a Perkin–Elmer MPF66 fluorescence spectrophotometer using 5-nm excitation and emission slit widths.

Tissue

Postmortem brain tissue from 12 autopsy-confirmed AD patients and six cognitively normal controls was obtained through the University of Pittsburgh Alzheimer's Disease Research Center Brain Bank. Age at death of AD patients was 71.3 ± 2.7 years (range 69–75) and that of controls was 77.8 ± 6.3 years (range 70–85). The postmortem interval was 4.0 ± 1.3 hr (range 2–5) for AD brains and 7.1 ± 3.7 hr (range 3–12) for controls. Frontal cortex was dissected from 1-cm-thick coronal slices and was placed in either 10% buffered formalin (7 days) or 4% paraformaldehyde (in 0.1 M potassium phosphate buffer, pH 7.4, 24–48 hr at 4C). Postmortem heart tissue from a 42-year-old woman with systemic amyloidosis due to λ-light chain monoclonal gammopathy was fixed in 10% formalin for 5 days and embedded in paraffin. Paraffin blocks were prepared by sequential dehydration in graded alcohol and vacuum infiltrated in paraffin before embedding and serial sectioning to a thickness of 8 μm. Serial sections were coded and numbered to enable sequential X-34/immunohistochemical analysis. In each series, all odd-numbered sections were stained with X-34. Even-numbered sections were stained by comparison methods. In this way, each comparison stain was both preceded and followed by an X-34-stained section. Before staining, paraffin sections were taken through two 10-min washes in xylene, two 5-min washes in 100% ethanol, two 5-min washes in 95% ethanol/H₂O, running tapwater for 10-min, and then 5 minutes in PBS (137 mM NaCl, 10 mM potassium phosphate, pH 7.6).

Paraformaldehyde-fixed tissue was cryoprotected by sequential incubation in 15% and 30% sucrose (in 0.1 M phosphate buffer), frozen, and cryosectioned on a Reichert Cryocut 1600 (Reichert–Jung; Leica, Buffalo, NY) to a thickness of 50 μm. Frozen sections were collected in Falcon 6-well plates (Fisher Scientific; Pittsburgh, PA) containing 0.1 M phosphate buffer, in a 1:6 series to allow serial analysis of X-34 and immunohistochemical and cytochemical markers. Sections were stored at 4C until assayed. Before staining, the sections were washed for 5 min in PBS.

When pretreated with formic acid, sections were soaked in 99% formic acid for 5 min and then rinsed for another 10 min in tapwater before going into the first PBS wash.

Autofluorescence Quenching Procedure and Thioflavin-S Staining

Thioflavin-S staining was performed by a modified thioflavin-S technique previously reported by Styren et al. (1991) and modified slightly by Guntern et al. (Guntern et al. 1992; Vallet et al. 1992). Thioflavin-S fluorescence was optimally observed using the UV filter set described below (U-filter). Unless otherwise noted, in this and all fluorescent staining procedures, inherent tissue autofluorescence from endogenous lipofuscin was minimized using a quenching step based on oxidative methods of lipofuscin modification (Barden 1983, 1984; Guntern et al. 1989).

Briefly, paraffin or frozen sections to be quenched are taken from the 5-min wash in PBS and placed in a potassium permanganate solution (0.25 g% in PBS) for 20 min, after which a brown color develops. The sections are washed twice in PBS for 2 min and then treated with a solution of 1.0 g% potassium metabisulfite and 1.0 g% oxalic acid in PBS until the brown color has been removed from the tissue (typically 1–6 min). Finally, the sections are washed three times in PBS for 2 min each.

Bielschowsky Staining

The Bielschowsky staining method was performed according to the Yamamoto and Hirano (1998) modification.

X-34 General Staining Procedure

Quenched or unquenched tissue sections are taken from PBS into a solution of X-34 in 40% ethanol/60% distilled H2O (adjusted to pH 10 by addition of 1 N NaOH) at the appropriate concentration (100 nM to 1 mM) for the appropriate period of time (0.5–10 min). The sections are then dipped briefly five times in tapwater before differentiation in 0.2 g% NaOH in 80% ethanol for 2 min. The sections are then placed in tapwater for 10 min before coverslipping with Fluoromount-G (Electron Microscopy Science's Fort Washington, PA).

Determination of Optimal X-34 Staining

Serial 8-μm paraffin sections of frontal cortex were selected from two autopsy-confirmed AD brains. A dilution vs time matrix was constructed in which sections were stained with 100 nM, 1 μM, 10 μM, 100 μM, or 1 mM solutions of X-34. At each concentration, sections were stained for either 0.5, 1, 5, or 10 min. An identical matrix was created for quenched and unquenched sections (see above). Blind analysis was performed by three investigators (SDS, WEK, and GCS) and the optimal concentration selected. The optimal condition was defined by clear identification of resident pathology, overall image quality, and qualitative analysis of signal-to-noise ratio. Although a matrix of this type will yield a number of equivalently stained sections, preference was given to the conditions that resulted in optimal staining with the lowest concentration of X-34.

X-34 Staining/Immunohistochemistry

Alternate paraffin and cryosections were quenched as described above and stained for 10 min in 100 μM X-34 and either coverslipped or further processed for immunohistochemistry or histochemistry (Styren et al. 1994). Briefly, cryosections were placed in blocking buffer (0.1 M phosphate buffer and 10% species-appropriate serum) and incubated overnight in the primary antibody to either Aβ (10D5 was a generous gift from Athena Neuroscience) or TAU (Table 1). Paraffin sections were encircled by a hydrophobic medium (HistoPaP; Electron Microscopy Sciences) before application of blocking buffer and primary antiserum. All sections were incubated in species-specific secondary antibodies conjugated with biotin. For fluorescence studies, biotinylated antibodies were visualized using rhodamine-linked avidin according to the manufacturer's directions (Vector Laboratories; Burlingame, CA). For horseradish peroxidase (HRP) studies, biotinylated antibodies were visualized by treatment with HRP-linked avidin followed by reaction in H₂O₂-catalyzed 3,3′-diaminobenzidine·4HCl (DAB) (0.05%). Cryosections were rinsed in 0.1 M phosphate buffer before mounting on gelatin-coated slides. Slides were allowed to dry at room temperature overnight. Fluorescent labeled sections were coverslipped with Flouromount-G (Electron Microscopy Sciences). All others were dehydrated in a graded ethanol series, cleared in Histoclear (Fisher Scientific), and coverslipped with Permount. Selected cryo-and paraffin sections were processed without primary antibodies to control for antibody specificity.

Table 1

Antibody characteristics

Antibody	Source	Dilution	Species
10D5 Anti-Aβ (1–28)	Athena Neuroscience	1:3000	Mouse monoclonal
Anti-TAU	DAKO	1:200	Rabbit polyclonal

Fluorescent sections were viewed using an Olympus Vanox AH-RFL-LB fluorescence microscope equipped with the following filter sets: U-filter (excites 360–370 nm, dichroic mirror DM400, 420 nm-longpass filter); V-filter (excites 400–410 nm, dichroic mirror DM455, 455-nm longpass filter); B-filter/FITC (excites 470–490 nm, dichroic mirror DM500, 515-nm longpass filter); G-filter/TRITC (excites 530–550 nm, dichroic mirror DM570, 590-nm longpass filter).

Results

Physical Properties of X-34

X-34 (Figure 1) is a bright yellow-green solid (decomposes at 300C before melting) which is soluble at low millimolar concentrations in aqueous solution at basic or neutral pH. It is very soluble in polar organic solvents, such as DMSO. When stored as a solid (routinely stored in the dark), we have seen no evidence of deterioration in samples over 12 months old. HPLC showed a single major peak at 6.9 min (see Materials and Methods). The 500-MHz ¹H NMR spectrum of X-34 in d₆-DMSO (TMS reference) showed (see Figure 1 for proton numbering) H₂ at δ 7.99 (d, J = 2.3 Hz, 2H); H₆ at δ 7.83 (dd, J = 2.3 and 8.7 Hz, 2H); H_X at δ 7.59 (s, 4H); H_A at δ 7.27 (d, J = 16.3 Hz, 2H); H_B at δ 7.14 (d, J = 16.3 Hz, 2H); H₅ at δ 7.00 (d, J = 8.7 Hz, 2H). Mass spectrometry showed an EI parent ion of 402 as expected. The fluorescence spectra showed a broad excitation maximum at 367 nm and a broad emission maximum at 497 nm (aqueous solution at pH 7.0).

Optimization of the staining method

X-34 staining was localized to structures with the morphological appearance of Aβ plaques, neurofibrillary tangles (NFTs), neuropil threads, and vascular amyloid deposits (Figure 2). X-34 staining was effective in both paraffin (Figure 2A) and frozen sections (Figure 2B). Note that, in the thick 50-μm section (Figure 2B), X-34 stains homogeneously through the entire thickness of the section, whereas antibodies often poorly penetrate sections of tissue this thick. This is best demonstrated by the homogeneous staining of short segments of two amyloid-laden blood vessels in the upper right and middle left of Figure 2B (red arrows).

Figure 2

AD frontal cortex. A and C–H are 8-μm paraffin sections and B is a 50-μm frozen section. (A) X-34 stain showing many plaques with and without dense cores, scattered NFTs, a focal area of many neuropil threads, and a single longitudinal profile of a small blood vessel at lower left (red arrow). The lower inset shows a ×4 magnification of the neuropil threads (NT). The upper inset shows a ×2.5 magnification of an area that illustrates the color differentiation of dystrophic neurites (DN), neuropil threads (NT), and NFTs (all cyan) from plaque amyloid (CORE, yellow-green). Bar = 125 μm. (B) X-34 stain showing massive numbers of neuropil threads, NFTs, plaques, and amyloidladen blood vessels (cylindrical profiles marked by red arrows). Bar =150 μm. X-34 (C) and 10D5 antibody to Aβ visualized with DAB (D) staining of adjacent sections, showing identical staining patterns of plaques with and without cores. Bars = 50 μm. X-34 (E) and Bielschowsky silver stain (F) of adjacent sections, showing good correspondence for the staining of plaques. Bars = 100 μm. Double labeling in the same section by X-34 (G) and a polyclonal anti-TAU antibody visualized with a secondary antibody conjugated to rhodamine (H), showing NFT and neuropil threads in H and these same structures as well as additional NFTs (red arrows) and plaques in G. Bars = 50 μm.

Optimal X-34 staining was observed using a violet filter cube (V-filter set). Under these conditions, X-34 fluorescence appeared as bright green/white fluorescence which was very resistant to fading with extended excitation. Different pathological entities fluoresced with slightly different colors, with NFTs and neuropil threads typically appearing cyan-white and plaques tending to appear more yellow-green (Figures 2A and 2G). In many cases, this allowed easy identification of dystrophic neurites within plaques with the single stain (Figure 2A, upper inset). Vascular amyloid typically stained with such high intensity as to appear white, and often saturated the photos before other structures could be properly exposed. Analysis of fluorescence with the UV filter cube (U-filter set) revealed diminished fluorescence compared to the violet filter cube, but X-34 staining was still clearly demonstrated with this filter set. Staining was very weak using the FITC filter cube (B-filter set). X-34 did not fluoresce at all with TRITC. This profile was very useful in distinguishing specific X-34 fluorescence from autofluo-rescence, which was apparent with all filter sets. Therefore, the choice of filter set appears critical for optimal demonstration of X-34 fluorescence.

Analysis of the staining concentration vs staining time matrix revealed an optimal X-34 concentration of 100 μM after incubation for 10 min in quenched sections. This concentration/time combination was determined by blind analysis by three of the authors (SDS, WEK, and GCS). Sections were examined at all times and concentrations from both the quenched and un-quenched matrices. In general, there were relatively small differences between the 0.5- and 10-min staining times except at the 100 nM concentration. Greater differences were observed across concentrations at a given time point. Our observations revealed that quenched sections were universally superior to unquenched sections owing to the interference of X-34 signal by lipofuscin autofluorescence. However, approximately 10-fold higher concentrations of X-34 were required to give the same intensity of staining in the quenched sections as in the unquenched sections. Although quenching produced optimal staining in our opinion, it was not essential. Using the violet filter set, unquenched lipofuscin autofluorescence was different in color (more yellowish) and morphology and could be easily distinguished from X-34 staining. In addition, X-34-specific fluorescence could be verified in unquenched sections by simply switching to the FITC or TRITC filter set in which X-34-specific fluorescence was either greatly diminished (FITC) or eliminated (TRITC), whereas the lipofuscin autofluorescence changed color but not intensity.

Although a number of concentration/time combinations revealed nearly equivalent images with high signal-to-noise ratios, the combination of 100 μM X-34 for 10 min was selected for the quenched sections. Unless otherwise noted, all reported analyses of X-34 staining of AD pathology were performed at this concentration/time combination. A combination of 10 μM X-34 for 10 min appeared optimal for unquenched sections. Because of the intensely fluorescent nature of X-34, one major factor in determining the optimal staining conditions was the amount of background fluorescence. Increased background fluorescence was observed with increasing X-34 concentration and was at least partly due to leeching of X-34 into the mounting medium. This could be best appreciated when air bubbles were present in the medium because there was virtually no background fluorescence where an air bubble overlaid the tissue.

X-34 staining by this method appears to be very stable. We have now had the opportunity to repeatedly observe the original stained sections over a period of more than 3 years. We see no apparent signs of degradation over this period of time in these X-34-stained sections (which are routinely stored in the dark).

Qualitative Comparison to Traditional Staining Methods

The apparent staining of the classical hallmarks of AD neuropathology, along with the intensity of X-34 staining and the simplicity of the method, suggested that X-34 could be useful as a sensitive stain for histochemical localization of AD pathology. Therefore, we qualitatively compared the staining of X-34 to several traditional methods of demonstrating AD neuropathology. Detailed quantitative comparisons were beyond the scope of this initial report.

Comparison of the well-established fluorescent marker of AD neuropathology, thioflavin-S, and X-34 showed that X-34 identified the same plaques, NFTs, and cerebrovascular amyloid demonstrated by thioflavin-S. In a qualitative sense, X-34 appeared to consistently stain these pathological structures more intensely than thioflavin-S. In addition, X-34 appeared to reveal plaques and NFTs not identified by thioflavin-S (Figures 3A and 3C). Qualitatively, X-34 appeared to more clearly demonstrate neuropil thread pathology (insets to Figures 3A and 3C). The staining by thioflavin-S was optimized by using the improved thioflavine S method for staining NFTs and senile plaques reported by Guntern et al. (1992). This method, which includes steps to quench tissue autofluorescence, has been reported to increase the signal-to-noise ratio between specifically stained structures and background compared to classical thioflavine-S staining procedures (Vallet et al. 1992). Figure 3C shows the results of optimal staining, illumination, and photography for thioflavin-S in our hands. The background in thioflavin-S-stained tissue is inherently higher than in tissue stained with X-34, constituting another advantage of X-34.

Figure 3

Adjacent 8-μm sections of AD frontal cortex stained with X-34 (A,B) or thioflavin-S (C,D). The sections shown in B and D were pretreated with formic acid before staining. Note that more plaques, NFTs, and neuropil threads are seen in the X-34-stained section (A) than the thioflavin-S-stained section (C) and that thioflavin-S demonstrates only the most intensely stained structures in the X-34 section. In sections pretreated with formic acid, the staining of both X-34 (B) and thioflavin-S (D) is eliminated [see vessel profiles (marked by arrows) for orientation]. Insets in A and C show a ×2 magnification of the area marked by red rectangles in the respective figures. Bar =100 μm.

Figure 4

Adjacent sections of heart from a case of systemic amyloidosis due to λ-light chain monoclonal gammopathy stained with X-34 (A) or Congo red (B). X-34 was visualized with fluorescence microscopy using the V-filter as described in Methods. Congo red was visualized with fluorescent microscopy using the TRITC (rhodamine) filter. Bar = 200 μm.

X-34 staining also was compared to other traditional, well-characterized immunohistochemical and silver impregnation methods of identifying pathology in AD brain. It has been reported that the Bielschowsky method and Aβ immunostaining are the most sensitive methods for detection of neuritic plaques (Wisniewski et al. 1989). Comparison of X-34 staining to anti-Aβ immunohistochemistry revealed an essentially one-to-one correlation between immunohistochemical localization of Aβ and cytochemical staining revealed with X-34 (Figures 2C and 2D). Comparison of Bielschowsky silver stain and X-34 in adjacent sections revealed that X-34 was able to identify the same markers as the Bielschowsky stain. The pathological profiles appeared slightly more numerous when observed with X-34 than with the silver staining (Figures 2E and 2F). Although many NFTs are apparent in Figures 2E and 2F, they show some variation in spatial distribution despite the fact that these are adjacent sections. To assess the staining of NFT and neuropil thread pathology by X-34, a double-labeling protocol for TAU and X-34 was used. Double staining of sections for TAU (Figure 2H) and X-34 (Figure 2G) showed that X-34 stains the same NFTs and neuropil threads as those demonstrated by the TAU antibody. In addition, X-34 appears to stain a few tangles in addition to those seen with this particular polyclonal antibody (arrows in Figure 2G) and more crisply delineates the neuropil thread pathology (as well as simultaneously identifying the amyloid plaques).

Requirement for β-pleated Sheets

It has been shown that staining with Congo red carries an absolute requirement for the existence of the β-pleated sheet secondary protein structure (Glenner et al. 1972; Glenner 1981). This property has been exploited to develop the Congo red method of assessing the aggregation of amyloid proteins into β-pleated sheet fibrils (Klunk et al. 1989, 1999). To investigate whether β-pleated sheet secondary protein structure also was necessary for X-34 staining, we compared X-34 to thioflavin-S staining in sections pretreated with formic acid with non-formic acid-treated adjacent sections also stained with either X-34 or thioflavin-S. Pretreatment with formic acid disrupts the β-pleated sheet structure of Aβ without cleaving or removing the protein and enhances anti-Aβ antibody staining (Kitamoto et al. 1987), although formic acid treatment can covalently modify the protein (Klunk and Pettegrew 1990; Orlando et al. 1992; Klunk et al. 1994b). Our analysis revealed that pretreatment with formic acid eliminates the binding of both X-34 and thioflavin-S to plaques, cerebrovascular amyloid, NFTs, and neuropil threads (Figures 3B and 3D). This provides direct evidence that X-34 and thioflavin-S bind only to β-pleated sheet protein configurations and not to the soluble or non-fibrillar forms of Aβ.

X-34 Staining of Systemic Amyloidosis

Although staining with Congo red carries an absolute requirement for the existence of the β-pleated sheet secondary protein structure, Congo red stains a variety of β-pleated sheet amyloid deposits composed of proteins with very different primary amino acid sequences (Glenner et al. 1972; Glenner 1981). That is, Congo red staining is more specific for β-pleated sheet secondary structure than for primary amino acid sequence (although the latter determines the secondary structure). Because X-34 is a derivative of Congo red, it was expected that it also would stain a wide variety of amyloid deposits in addition to the plaques and NFTs of AD. As a preliminary assessment of broad-range applicability, X-34 was used to stain cardiac tissue from a case of systemic amyloidosis due to λ-light chain monoclonal gammopathy (Figure 4). Compared to Congo red staining (Figure 4B), X-34 more intensely stained λ-light chain amyloid (Figure 4A) and showed an almost identical pattern of distribution.

Comparison of X-34 Staining in AD and Age-matched Normal Control Brain

Whereas X-34 staining of AD brain showed an abundance of plaques, NFTs, neuropil threads, and cerebrovascular amyloid (Figures 5D–5F), very little staining was observed in age-matched control brain (Figures 5A–5C). Rare cerebrovascular amyloid was noted in one control brain (Figure 5B). In all control brains studied, we consistently observed small stellate profiles that appeared to be intracellular (see arrows in Figures 5A and 5C). These profiles were concentrated in the molecular layer and in deep layers of neocortex, with a near absence in between. These profiles were not observed with Bielschowsky or Aβ immunostaining. The morphology of these profiles was clearly distinct from that of NFTs (see arrowheads in Figure 5F), neuropil threads, and plaques. The same type of stellate profile was present in similar areas of AD brain (arrows in Figure 5E), but was almost obscured by the much more obvious network of neuropil threads, NFTs, and plaques. The very extensive neuropil staining seen in AD sections (Figures 5D–5F) was absent in control brain (Figures 5A–5C).

Figure 5

X-34-stained 8-μm sections of frontal cortex from three different age-matched control brains (A–C) and three different AD brains (D–F). (A) Small, apparently intracellular stellate structures in the molecular layer (demonstrated by arrow). (B) A rare amyloid-laden vessel in middle cortical layers. (C) Stellate structures similar to those in A seen in deep cortical layers. (D) Extensive plaque, NFT, and neuropil thread pathology typical of X-34-stained AD brains. (E) In addition to the AD pathology, the stellate structures easily seen in control brain could also be found in AD brains (arrows). (F) The vast majority of NFTs (arrowheads) are readily distinguished from the stellate structures. Note the extensive neuropil pathology, which is not evident in control brains. Bar = 100 μm.

Discussion

X-34 is a highly fluorescent, lipophilic derivative of Congo red that intensely stains the neuropathological hallmarks of AD brain. The staining procedure is rapid, simple, and highly reproducible. It is applicable to both fresh/frozen and formalin-fixed, paraffin-embedded sections and can be used with immunohistochemical procedures for double labeling. The use of a violet filter cube was found to be important for optimal appreciation of the high fluorescence intensity. The specific profile of X-34 fluorescence suggests that it will also be useful in co-localization studies using laser confocal microscopy. The procedure requires little technical expertise to perform and is not subject to the inherent variability associated with immunohistochemistry, thioflavin-S, or Bielschowsky silver staining. This simplicity may be one reason for the fact that our staining results have been very reproducible among several technicians in two separate laboratories. Further evaluation of the reproducibility will await the application of this method in a much larger number of laboratories.

The very low background staining indicates that a large array of β-sheet structure is required for binding, because X-34 does not appear to stain native proteins that have only a few, relatively isolated β-strands. Previously proposed models hypothesize that a minimum of five adjacent β-sheet strands is necessary for Congo red (Klunk et al. 1989) and Chrysamine-G binding (Klunk et al. 1994a), and the same may also apply to X-34. The finding that formic acid pretreatment destroys X-34 (as well as thioflavin-S) staining is further proof of the importance of secondary structure (i.e., β-sheet) over primary amino acid sequence. Formic acid solubilizes precipitated aggregates of Aβ and destroys the β-sheet secondary structure (Kitamoto et al. 1987). Formic acid pretreatment of tissue sections enhances the staining with antibodies against Aβ (Kitamoto et al. 1987; Ikeda et al. 1989), proving that the peptide is still present. The combination of affinity for two separate fibrillar proteins (Aβ and TAU/NFT) along with the inability to stain either protein deposit after formic acid treatment (Figure 3B) strongly supports the notion that the determining characteristic of deposits stained by X-34 is extensive β-sheet fibril structure. A similar inhibition of Bielschowsky and periodic acid–methenamine silver (PAM) staining of compact and diffuse plaques by formic acid pretreatment has been previously reported (Yamaguchi et al. 1989).

The general affinity of X-34 for β-sheet fibrillar deposits can be viewed as disadvantageous because it decreases the specificity of X-34 for any one particular type of deposit. However, other neuropathological characteristics, such as morphology of the deposits, topology, and associated neuropathological features, will readily distinguish among most fibrillar deposits (e.g., plaques, NFTs, and cerebrovascular amyloid). The general affinity for β-sheet deposits also can be viewed as an advantage in the study of AD brain. In some studies, it may be technically advantageous to intensely stain both plaque and tangle pathology with a single simple method. This is particularly true if other double-labeling procedures are to be performed. In many circumstances, it would be advantageous to simultaneously identify and perhaps even quantitate a more comprehensive measure of the AD pathology than either plaques alone or NFT alone. X-34 allows the simultaneous demonstration of plaques (neuritic and non-neuritic), cerebrovascular amyloid, NFTs, and neuropil threads. In most AD cases, the extents of all of these neuropathological changes are highly correlated with one another other and the cumulative assessment may give a better measure of the effective progression of the disease than would any one of the pathological entities examined in isolation.

This intense staining of AD pathology by X-34 raises the question of whether X-34 staining might identify pathological structures in other neurodegenerative or neuropsychiatric diseases. For example, it would be expected that X-34 would stain plaques in prion disease, which have been shown to contain β-sheet fibrillar peptides (Jackson et al. 1999). Although it is beyond the scope of this initial study, follow-up studies are under way to look at X-34 staining in a variety of neuropathological diseases and in brains from patients of various ages who died from non-neurological causes. Our preliminary data showing that X-34 stains amyloid deposits in systemic amyloidosis due to λ-light chain monoclonal gammopathy supports the notion that X-34 is a general amyloid stain. Along with the inhibition of X-34 staining after disruption of the β-pleated sheet fibril secondary structure with formic acid, this indicates that, like Congo red, X-34 is more specific for β-pleated sheet secondary structure than for the primary amino acid sequence of the amyloid protein.

Our results from X-34-stained elderly control brains (Figures 5A–5C) have shown very little X-34 staining, supporting the specificity of X-34 for pathological fibrillization of endogenous proteins. This, coupled with the lipophilic nature of X-34, supports further pursuit of our original goal of developing X-34 or its derivatives as in vivo neuroimaging agents (using PET or SPECT for detection of radiolabeled derivatives of X-34) in living AD patients (Klunk et al. 1995). Thioflavin-S has already been used to stain superficial amyloid plaques in transgenic mice in vivo by administration through a thinned area of the skull and detection with two-photon microscopy (Christie et al. 1999). X-34, or related derivatives that are more lipophilic, may also be useful for this purpose and may hold the advantages of increased diffusion due to increased lipid solubility (thioflavin-S, like Congo red contains highly charged sulfonic acid groups). This lipophilicity, along with the greater fluorescence intensity, may allow visualization of more and deeper pathological structures.