Specific Detection of Pathological Three‐repeat Tau after Pretreatment with Potassium Permanganate and Oxalic Acid in PSP/CBD Brains

Abstract

Immunohistochemisty with RD3, a monoclonal antibody specific for three‐repeat (3R) tau, is sometimes hampered by diffuse neuronal staining on formalin‐fixed, paraffin‐embedded sections pretreated with formic acid and heating. Additional pretreatment with potassium permanganate followed by oxalic acid completely eliminated this diffuse RD3‐immunoreactivity (IR) in neurons. Furthermore, this additional pretreatment uniformly enhanced RD3‐IR, as well as RD4‐IR, a monoclonal antibody specific for four‐repeat (4R) tau, on pathological deposits with tau IR. This enhanced sensitivity and specificity may allow more reliable identification of 3R and 4R tau in pathological deposits, which may be variable dependent on disease and regions. Cerebral cortex and midbrain from 8 patients [5 progressive supranuclear palsy (PSP) and 3 corticobasal degeneration (CBD)] were screened for RD3‐ and RD4‐IR with this improved procedure. In addition to RD4‐positive structures found both in cerebral cortex and brainstem, RD3‐positive neurofibrillary tangles (NFTs) were also found in midbrain in 7 of these 8 cases but not in the cortex. Multi‐labeling study demonstrated that most of RD3‐negative neurons were positive for RD4. This reliable demonstration of pathological 3R tau deposits in the brainstem of PSP/CBD, so far presumably characterized by deposition of 4R tau, is useful to map tau‐positive lesions according to their biochemical composition.

INTRODUCTION

Isoform‐specific antibodies against tau are powerful tools to discriminate tau‐positive lesions on immunohistochemistry based on their biochemical features. de Silva and colleagues established a pair of monoclonal antibodies, RD3 and RD4, and their specific affinity to three‐repeat (3R) tau and four‐repeat (4R) tau, respectively, has been confirmed in several studies by different groups (2). Although epitope specificity of these antibodies has already been established by probing recombinant tau isoforms on immunoblot (2) or human brain homogenates 5, 6, harsh pretreatment with formic acid (>99% for 30 minutes) followed by autoclaving was required for immunohistochemical visualization of these tau epitopes on routinely processed (formalin‐fixed, paraffin‐embedded) human brain tissues 8, 9. In addition to expected pathological lesions such as neurofibrillary tangles (NFTs) (19), diffuse neuronal staining with RD3 has been noted since the initial (2) and subsequent report (4). Because this faint and diffuse RD3‐immunoreactivity (IR) in neurons with normal appearance is far more widespread than that obtained with AT8, a current standard to identify tau‐positive lesions 1, 12, and similarly observed even in normal brains without tau‐positive pathological deposits, it is possible that this faint RD3‐IR represents “normal 3R tau” retrieved in soluble fraction not only in normal brains but also in brains harboring tau‐positive deposits 5, 6. Because of this uncertainty, we have had a difficulty to interpret what is represented by this faint RD3‐IR, apparently distinct from AT8‐IR. Pretreatment of sections with potassium permanganate (KMnO₄) followed by oxalic acid (Ox), initially described for bleaching melanin granules (10), was found to be useful to eliminate non‐specific background of Gallyas silver impregnation method (GAL) (17). With the use of this pretreatment, we are now successful in eliminating this diffuse RD3‐IR found even in apparently normal neurons to visualize selectively “pathological 3R tau” in pathological deposits of tau. Furthermore, this pretreatment was found to be useful in enhancing the RD3‐ and RD4‐IR, which provides further advantage to localize 3R and 4R tau in formalin‐fixed, paraffin‐embedded archival sections.

MATERIALS AND METHODS

Influences of pretreatments on immunostaining with RD3 and RD4 were first evaluated on formalin‐fixed paraffin‐embedded sections of hippocampus from an 87‐year‐old female patient with Alzheimer disease (AD) and cerebral cortex from a normal control (59‐year‐old male). Frontal cortex and midbrain sections (formalin‐fixed, paraffin‐embedded sections) were obtained from 5 cases of progressive supranuclear palsy (PSP, 64‐year‐old female, 67‐year‐old male, 68‐year‐old male, 74‐year‐old female, 76‐year‐old male) and 3 cases of corticobasal degeneration (CBD, 49‐year‐old male, 61‐year‐old male, 67‐year‐old female). The diagnosis of PSP was based on the proposed criteria for histological diagnosis 7, 11. Histological diagnosis of CBD was based on the presence of extended cortical lesions with ballooned neurons and astrocytic plaques (16). Five µm‐thick sections, deparaffinized and rehydrated, were pretreated with following pretreatments. (i)no pretreatment; (ii)FA‐AC: formic acid (FA > 99% for 30 minutes) followed by autoclaving in 0.05 M citrate buffer at 121°C for 20 minutes (AC) 8, 9; (iii) periodic acid (1% for 15 minutes); (iv) pyridine (100% for 15 minutes); and (v) KMnO₄‐Ox: 0.25% KMnO₄ for 15 minutes followed by 2% Ox for 3 minutes at room temperature 10, 17. After inactivation of endogenous peroxidases with 0.1% hydrogen peroxide, pretreated sections were incubated first with the 5% horse serum diluted in 0.01 M phosphate‐buffered saline containing 0.03% Triton X–100 (PBST, Wako, Tokyo, Japan). They were then incubated with either 3R tau‐specific antibody (RD3 1:3000, Upstate, Lake Placid, NY, USA) or 4R tau‐specific antibody (RD4, 1:1000, Upstate) diluted in the same buffer for 2 days at 4°C (2). They were then incubated with biotinylated secondary antibody against mouse IgG (1:1000, ABC Elite, Vector, Burlingame, CA, USA) diluted in the same buffer for 2 h at room temperature. They were then incubated with avidin‐biotin‐peroxidase complex (1:1000, ABC Elite, Vector) for 1 h and visualized with diaminobenzidine and nickel ammnonium chloride. Specificity of RD3‐IR on immunohistochemistry was confirmed by co‐incubation of the synthetic antigen peptide (2) (RS3; KHQPGGGKVQIVYKPV, Operon, Tokyo, Japan) at final dilutions ranging 0.01–10 µg/mL with the primary antibody.

For double immunofluorolabeling, deparaffinized sections were pretreated with KMnO₄‐Ox followed by FA‐AC (optimal pretreatment for RD3, see results for detail). They were incubated with the mixture of RD3 (1:300) and an antihuman tau antibody (1:1000, rabbit polyclonal, generously provided by Professor Mori H., Osaka City University) diluted with PBST for 2 days at 4°C. The sections were then incubated with Alexa Fluor^® 488 conjugated with antimouse IgG antibody (1:200 Molecular Probe, Carlsbad, CA, USA) and Alexa Fluor^® 546 conjugated with antirabbit IgG antibody (1:200 Molecular Probe). After being mounted, the fluorolabeled sections were observed under the confocal laser scanning microscope (Leica SP5, Heidelberg, Germany). Alexa Fluor^® 488, which labeled the RD3 epitope, was excited by a 488‐nm beam and was detected through a light path ranging from 500–540 nm. Alexa 546, which labeled the human‐tau epitope, was excited by 543‐nm beam and was detected through a light path ranging from 600–640 nm. Silver impregnations, Campbell‐Switzer (CS) and Gallyas (GAL) methods were performed as summarized previously (15).

RESULTS

Hippocampal pyramidal neurons of AD brain, positive for AT8 (Figure 1A), were argyrophilic with silver impregnation with GAL (Figure 1B) and CS (Figure 1C). Without pretreatment, they were neither stained with isoform‐specific antibody RD3 (Figure 1D) nor RD4 (data not shown). After pretreatment with FA‐AC, RD4‐positive neurons and NFTs were visible (Figure 2A) but their intensity and the number were much smaller than those obtained with AT8 (Figure 1A). The number of RD4‐positive NFTs increased after additional pretreatment with KMnO₄‐Ox (Figure 2B). RD3‐positive neurons were visible in the cerebral cortex of normal control after FA‐AC treatment (Figure 2C), while this diffuse RD3‐IR was completely eliminated after additional treatment with KMnO₄‐Ox (Figure 2D). Although pretreatment with FA‐AC visualized RD3‐positive NFTs (Figure 2E, arrows), neurons exhibiting diffuse cytoplasmic staining for RD3 were always coexistent (Figure 2E empty arrowheads). The KMnO₄‐Ox pretreatment increased the number of RD3‐positive NFTs (Figure 2F) as was observed with RD4‐positive NFTs (Figure 2B). Moreover, this pretreatment eliminated the diffuse cytoplasmic RD3‐IR in neurons not containing NFTs to yield the results comparable with those with AT8 (Figure 1B) or with RD4 (Figure 2B). This RD3‐IR on NFTs (Figure 2E, arrows, and Figure 3) and that in neurons not containing NFTs (Figure 2E, empty arrowheads) were both eliminated when the RD3 antibody was co‐incubated with the synthetic peptide corresponding to the reported RD3 epitope up to the dilution of 0.1 µg/mL (Figure 2G and H), whereas this RD3‐IR was partially restored when the antigen peptide was diluted further to 0.01 µg/mL (Figure 2I and J).

Figure 1.

Tau‐immunoreactivity (IR) and argyrophilia in hippocampal pyramidal layer of Alzheimer disease (AD) brain . A. Abundant AT8‐IR in the pyramidal neurons and neuropil threads. B. Argyrophilic neurofibrillary tangles (NFTs) (Gallyas silver impregnation method). C. Argyrophilic NFTs (Campbell‐Switzer silver impregnation method). D. RD3‐IR (RD4‐IR as well, data not shown) was not detectable without pretreatment. Bar = 50 µm. A, D: counterstained with nuclear fast red.

Figure 2.

Enhanced specificity and sensitivity of RD3 and RD4‐immunoreactivity (IR) after pretreatment with potassium permanganate followed by oxalic acid in hippocampal pyramidal layer of Alzheimer disease (AD) brain. A. Some neurofibrillary tangles (NFTs) exhibited RD4‐IR after the standard pretreatment with formic acid followed by autoclaving (FA‐AC). B. Additional pretreatment with potassium permanganate followed by oxalic acid (KMnO₄‐Ox) prior to FA‐AC increased the number of RD4‐positive NFTs. C. Some NFTs exhibited RD3‐IR after the standard pretreatment with formic acid followed by autoclaving (FA‐AC). D. Additional pretreatment with KMnO₄‐Ox prior to FA‐AC completely eliminated this RD3‐IR. E. After the standard FA‐AC pretreatment, intense RD3‐IR in NFT (arrows) and diffuse RD3‐IR in neurons without NFT (empty arrowheads) were observed. F. Additional pretreatment with KMnO₄‐Ox prior to FA‐AC increased the number of RD3‐positive NFTs (arrows) but eliminated diffuse RD3‐IR in neurons without NFT. Co‐incubation with the synthetic peptide (KHQPGGGKVQIVYKPV, RD3 epitope) at the dilution of 0.1 µg/mL(or greater) eliminated RD3‐IR (G) not only in NFTs (E, arrows) but also in neuronal cytoplasm not harbouring NFTs (E, empty arrowheads). Co‐incubation with this peptide also abolished RD3‐IR in NFTs enhanced by additional pretreatment KMnO₄‐Ox prior to FA‐AC (H). Further dilution of the peptide to 0.01 µg/mL confirmed concentration‐dependent nature of the RD3‐IR with (I) or without (J) the additional pretreatment with KMnO₄‐Ox. (*G–J: the same blood vessel.) Bar = 50 µm: A, B, E, F. Bar = 100 µm: C, D, G–J.

Figure 3.

Enhanced specificity of RD3‐immunoreactivity (IR) and increased sensitivity of RD4‐IR after pretreatment with potassium permanganate followed by oxalic acid in cerebral cortex of progressive supranuclear palsy (PSP) brain. A. Numerous neurons containing diffuse cytoplasmic RD3‐IR (empty arrowheads) after formic acid followed by autoclaving (FA‐AC) pretreatment. B. Neurons (arrows) and glia (arrowheads) were positive for RD4 after FA‐AC. Additional pretreatment with potassium permanganate followed by oxalic acid (KMnO₄‐Ox) prior to FA‐AC completely eliminated diffuse RD3‐IR (C) but increased the number of RD4‐positive neurons and glia (D). These RD4‐positive structures, not detectable with Campbell‐Switzer (CS) (E), were fully demonstrated with Gallyas silver impregnation (GAL) (F). Bar = 50 µm.

This diffuse cytoplasmic staining with RD3 was similarly abundant in neurons of the cerebral cortex of PSP (Figure 3A, empty arrowheads) and CBD (data not shown) in all these eight cases examined. Similar diffuse staining was not detectable with RD4 (Figure 3B). The additional treatment with KMnO₄‐Ox completely eliminated this RD3‐IR (Figure 3C), whereas it enhanced RD4‐IR (Figure 3D) in contrast. CS demonstrated no argyrophilic lesions in the adjacent sections from cerebral cortex of PSP while GAL demonstrated abundant argyrophilic lesions in contrast.

In the substantia nigra of PSP brain pretreated with FA‐AC, RD3‐positive NFTs (Figure 4A, arrows) and RD3‐positive neurons without NFTs (Figure 4A, empty arrowheads) were observed. Additional pretreatment with KMnO₄‐Ox eliminated this RD3‐IR in neurons leaving NFTs more distinct (Figure 4C). RD4‐immunostain (Figure 4B) and GAL (Figure 4D) similarly demonstrated NFTs but not neurons without NFTs (Figure 4D). Similar RD3‐positive NFTs were identified in the midbrain (substantia nigra, oculomotor nuclei or central gray) in 7 out of 8 cases examined. RD3‐positive fraction in nigral neurons was highly variable but they are less than 10% of tau‐positive nigral neurons. Double immunofluorolabeling with RD3 (Figure 5A and B, green) and human tau (Figure 5E and F, red) after KMnO₄‐Ox and FA‐AC pretreatment demonstrated that RD3‐positive NFTs (green) were always positive for human tau (red) to yield yellow color in the merged images (Figure 5C and D, yellow) in hippocampal pyramidal layers of AD brains (Figs 5A, C, E and G) and nigral neurons of PSP brains (Figs 5B, D, F and H), which confirmed pathologic nature of these RD3‐IRs. Subsequent immunostain with RD4 and cell to cell comparison with fluorescence images demonstrated that each neuron positive for tau (E, F) was equally positive for RD4 (G, H).

Figure 4.

Presence of RD3‐positive neurofibrillary tangles (NFTs) in nigral neurons of progressive supranuclear palsy (PSP) brain . Neighboring sections of the substantia nigra from a patient with PSP. A. After formic acid followed by autoclaving (FA‐AC) pretreatment, a substantial number of nigral neurons (empty arrowheads) exhibited diffuse RD3‐immunoreactivity (IR). More intense RD3‐IR was found in NFTs (arrows). B. A larger number of NFTs were detectable with RD4 after FA‐AC pretreatment. C. Additional pretreatment with potassium permanganate followed by oxalic acid (KMnO₄‐Ox) prior to FA‐AC completely eliminated this diffuse cytoplasmic RD3‐IR, while RD3‐positive NFTs and neuropil threads were more intensely labeled. D. Gallyas silver impregnation (*: the same blood vessel, Bar = 100 µm).

Figure 5.

RD3‐immunoreactivity (IR) positive neurofibrillary tangles (NFTs) in Alzheimer disease (AD) and progressive supranuclear palsy (PSP) and their relation to RD4‐IR. Double immunofluorolabeling with RD3 (A, B; green) and antihuman tau (E, F; red) and their merged images (C, D), after pretreatment with KMnO₄‐Ox followed by FA‐AC. After recording fluorescent images, the same sections were heat‐treated and subsequently immunostained with RD4 (G, H). RD3‐IR (A, green) was detectable only in some of tau‐positive NFTs (C: red, antihuman tau) in the hippocampal pyramidal layer from AD brain (A, C, E, G) as well as in the substantia nigra of PSP brain (B, D, F, H). Each neuron, positive for tau (E, F), was equally positive for RD4 (G, H). Bar = 100 µm: A–D. Bar = 50 µm: E–H.

DISCUSSION

Immunohistochemistry with RD3 on formalin‐fixed, paraffin‐embedded human brains, expected to label pathological 3R tau deposits such as NFTs, has been reported to exhibit diffuse neuronal staining after conventional pretreatment with formic acid and heat retrieval 2, 4, 9. Although this diffuse RD3‐IR, is faint but extensive enough to include even normal neurons (Figure 2C), one may wonder whether this diffuse RD3‐IR represents deposition of pathological 3R tau or non‐specific staining. On the other hand, the specificity of RD3 antibody to 3R tau was convincingly established on Western blot by demonstrating its specific affinity to recombinant 3R tau (2). This specificity is confirmed even when normal tau from control brains is probed with RD3 (Figure 2C). This normal tau, which is retrieved in soluble fraction and is not extensively phosphorylated, is similarly retrieved in soluble fraction of brains with pathological tau deposits 5, 6. Because brains with pathological tau deposits also contain phosphorylated tau species retrieved in insoluble fraction in addition to this normal tau in soluble fraction, there are at least two different tau pools in the same brains with pathological tau deposits. This dual compartmentalization of tau into normal and pathological pools may be related to two different kinds of the 3R tau IRs detected with RD3 after conventional FA‐AC treatment in AD hippocampus; namely, NFTs (arrows Figure 2E) and diffuse RD3‐IR in neurons (Figure 2C and E arrowheads). Although this kind of extensive and diffuse IR is hardly detectable with other antibodies against tau, it is probable that this diffuse RD3‐IR (Figure 2C and E, empty arrowheads) and RD3‐positive NFTs (arrows, Figure 2E) represent different types of 3R tau deposition because both were absorbed in dose‐dependent fashion (Figure 2G and I) upon co‐incubation with the antigen peptide. Because this dual representation of 3R tau, however, does not distinguish normal or pathological nature of RD3‐IR by itself on histological sections, it is difficult to identify which 3R tau (normal and pathological) is involved in each RD3‐positive neuron (with or without NFT) without taking into account other morphological parameters

Our primary goal was directed to establish a protocol that may distinguish these 3R tau pools on histological sections. For this purpose, we tested various pretreatments and their combinations, expecting that some pretreatments may eliminate selectively this extensive and diffuse RD3‐IR in neurons with normal appearance but retain RD3‐IR in pathological tau deposits. Among pretreatments tested, KMnO₄‐Ox followed by FA‐AC completely eliminated this diffuse RD3‐IR in neurons with normal appearance. Because RD3‐IR in NFTs and neuropil threads was intensified, this KMnO₄‐Ox treatment conferred selective enhancement of RD3‐IR in NFTs and eliminated diffuse RD3‐IR in normal neurons in contrast. Comparison (Figure 2F) with AT8‐stained section (Figure 1A) or double staining with RD3/anti‐tau polyclonal (Figure 5A–F) demonstrated selective accumulation of RD3 epitope only on pathological deposits positive for AT8 or the anti‐tau antibody, both selective to pathological deposits. Moreover, subsequent RD4 immunostaining demonstrated that each of these tau‐positive neurons was also positive for RD4 in AD hippocampus and substantia nigra of PSP (Figure 5G and H). It is therefore now quite sure that RD3‐IR after pretreatment with KMnO₄‐Ox followed by FA‐AC represents exclusively “pathological 3R tau” without cross‐reacting to 3R tau in normal neurons or 4R tau.

Similarly, this diffuse RD3‐IR in neurons was extensive in the cerebral cortex of PSP/CBD brains (Figure 3A, empty arrowheads) when pretreated with FA‐AC, while adjacent sections stained with RD4 (Figure 3B) and GAL (Figure 3F) exhibited smaller number of tau‐positive neurons (Figure 3B, arrows) and glial cells (Figure 3B arrowheads). Interestingly, this diffuse RD3‐IR in neurons was nearly completely eliminated upon additional pretreatment with KMnO₄‐Ox (Figure 3C), suggesting again that this diffuse RD3‐IR does not necessarily represent pathological 3R tau. Because CS silver impregnation method visualizes pathological 3R‐tau lesions such as Pick bodies but not 4R‐tau lesions such as seen in PSP/CBD cortex, near complete absence of CS‐positive lesions in PSP/CBD (Figure 3E) also suggests the absence of pathological 3R‐tau lesions in the cortex and confirmed that this additional pretreatment was successful in eliminating non‐pathological (normal) 3R‐tau IR.

As this additional pretreatment with KMnO₄‐Ox eliminated RD3‐IR originated from non‐pathological tau, otherwise detectable as diffuse neuronal staining other than NFTs, it is now possible to probe these pathological 3R‐tau deposits selectively. Indeed, FA‐AC pretreatment of the midbrain sections visualized RD3‐positive NFTs (Figure 4A, arrows) and diffuse RD3‐IR in neurons (Figure 4A, empty arrowheads). As expected, additional KMnO₄‐Ox made RD3‐IR selective to pathological 3R tau deposits by eliminating diffuse RD3‐IR (Figure 4C). These pathological 3R‐tau deposits, although smaller in number than those observed with AT8 (Figure 4B), were identified in 7 out of 8 cases of PSP/CBD. We do not yet know, whether these brainstem lesions positive for 3R tau in PSP/CBD brains represent a concomitant senile change characterized by deposition of both 3R and 4R tau as seen in AD brains (14). It is now being recognized that biochemical distinction such as 4R tauopathy or 3R tauopathy is not absolute and there is some biochemical overlap between them. This relative predominance of 4R over 3R tau may explain why very small number (less than 10%) of brainstem neurons exhibited RD3‐IR. It remains to be clarified why pathological 3R‐tau deposits and CS‐positive lesions were usually absent in the cerebral cortex of PSP/CBD, where 3R tau is also detectable in insoluble fraction. It is possible that 3R tau are differently detected by immunohistochemistry and biochemistry. Although further improvements for better immunohistochemical detection is desirable, combined pretreatment with KMnO₄‐Ox followed by FA‐AC is the method of choice for selective immunolocalization of pathological 3R tau deposits in formalin‐fixed, paraffin‐embedded human brain samples 3, 13, which expand the applicability and reliability of RD3 immunohistochemistry to detect pathological 3R tau. This approach will expand and strengthen our understanding on isoform‐selective deposition of tau, which may vary according not only to lesions specific to disease categories but also to regions examined (18).