Cytochemical Demonstration of Oxidative Damage in Alzheimer Disease by Immunochemical Enhancement of the Carbonyl Reaction with 2,4-Dinitrophenylhydrazine

Abstract

Formation of carbonyls derived from lipids, proteins, carbohydrates, and nucleic acids is common during oxidative stress. For example, metal-catalyzed, “site-specific” oxidation of several amino acid side-chains produces aldehydes or ketones, and peroxidation of lipids generates reactive aldehydes such as malondialdehyde and hydroxynonenal. Here, using in situ 2,4-dinitrophenylhydrazine labeling linked to an antibody system, we describe a highly sensitive and specific cytochemical technique to specifically localize biomacromolecule-bound carbonyl reactivity. When this technique was applied to tissues from cases of Alzheimer disease, in which oxidative events including lipoperoxidative, glycoxidative, and other oxidative protein modifications have been reported, we detected free carbonyls not only in the disease-related intraneuronal lesions but also in other neurons. In marked contrast, free carbonyls were not found in neurons or glia in age-matched control cases. Importantly, this assay was highly specific for detecting disease-related oxidative damage because the site of oxidative damage can be assessed in the midst of concurrent age-related increases in free carbonyls in vascular basement membrane that would contaminate biochemical samples subjected to bulk analysis. These findings demonstrate that oxidative imbalance and stress are key elements in the pathogenesis of Alzheimer disease.

Although oxidative stress and consequent free radical damage are involved as a major cytopathological mechanism in a number of diseases and in normal aging, methods to document free radical involvement have not grown at the same rate as hypotheses. This gap has stemmed primarily from problems in quantifying free radicals in vivo, because they are rapidly removed by antioxidants or quenched by reaction with biopolymers. In addition, normal metabolism generates abundant free radicals, such that detecting them without knowing the balance of oxidants and antioxidants is not biologically meaningful. This aspect has been circumvented by methods to detect excess radical production. Although detection can be performed using exogenous agents that are specifically modified by free radicals, such as spin traps (Young et al. 1996), endogenous macromolecules can also be used as indicators because protein, DNA, and lipids are all chemically changed by oxidative stress (Stadtman 1993).

Modifications of proteins and polynucleotides take two forms: (a) adduction reactions by highly reactive intermediate products of lipid peroxidation or glycation, and (b) direct oxidative modification of the macromolecules. In the case of proteins, powerful oxidizing agents such as the hydroxyl radical directly modify amino acid side-chains, resulting in a diverse array of altered amino acids that are used to assess oxidative damage (Stadtman 1993). Among the most widespread of these modifications, and one that is considered specificfor oxidative damage, is the generation of free carbonyls that are not present on nonoxidized proteins (Levine et al. 1994). 2,4-Dinitrophenylhydrazine (DNP-H) reacts with free carbonyls and can therefore serve as a marker for the extent of oxidative damage to a given protein (Szweda et al. 1993). However, although this method has been validated as an indicator of oxidative damage for a number of tissues and isolated proteins (Smith et al. 1991), the procedure does not provide histological resolution. Such cellular localization is essential for analysis of tissue when one considers that the vascular basement membrane and extracellular matrix proteins, which can contaminate many biochemical samples, contain abundant oxidative modifications as a consequence of their long half-lives.

Here we report a highly specific and sensitive method to cytologically define oxidative damage that is based on the DNP-H reaction with free carbonyls but involves subsequent immunocytochemical enhancement. The technique is demonstrated using tissue sections from cases of Alzheimer disease and controls, because glycoxidative and lipoperoxidative damage are well documented in the pathological lesions of Alzheimer disease, neurofibrillary tangles and senile plaques (Yan et al. 1994; Smith et al. 1994,1995a,b,1996,1997; Montine et al. 1996; Sayre et al. 1997).

Materials and Methods

Tissue

The hippocampus extending from the entorhinal cortex to the full hippocampus from six cases of Alzheimer disease (ages 76–90), three age-matched controls (ages 70-78), and three younger controls (ages 32–64) was fixed in Methacarn (methanol:chloroform:acetic acid 60:30:10) at 4C for 24 hr. Tissue was dehydrated through ascending ethanol solutions and xylene and then embedded in paraffin (∼60C). Sections of 6 μm sections were prepared with a microtome (Leica), placed on Silane-coated glass slides (Sigma; St Louis, MO), and sequentially deparaffinized in xylene and rehydrated in descending ethanol. Endogenous peroxidase activity was inactivated by treating with 3% H₂O₂ in methanol for 20 min to block artifactual staining from endogenous peroxidase activity (Sternberger 1986). Although H₂O₂ might itself, or synergistically with tissue-bound iron, create carbonyl residues, deletion of this step had no effect on subsequent DNP-H labeling for the experiments shown here. However, a control that omits the H₂O₂ treatment step should be performed with each experimental paradigm. After blocking endogenous peroxidases, tissue was hydrated in ethanol (70, 50, and 30%) to 50 mM Tris-HCl, 0.15 M NaCl, pH 7.6 (TBS).

Dinitrophenyl Labeling

Sections were covered with 0.1-0.001% DNP-H in 2 N HCl. All incubations were done in a humidified plastic box. After a 1-hr incubation at room temperature (RT), sections were exhaustively rinsed in TBS, followed by a 30-min incubation in 10% normal goat serum (NGS) to block nonspecific binding sites. After rinsing with 1% NGS/TBS, a rat monoclonal antibody (LO-DNP-2; Zymed, San Francisco, CA) to dinitrophenyl (DNP) was diluted 1:100 in 1% NGS/ TBS and incubated with the sections at 4C for 16 hr. Sections were then rinsed with 1% NGS/TBS followed by goat antiserum to rat IgG (Boehringer-Mannheim; Indianapolis, IN) diluted 1:50 with 1% NGS/TBS. After rinsing in 1% NGS/TBS, rat peroxidase-anti-peroxidase complex (ICN) (1:250) in the same buffer was incubated with the section at RT for 1 hr, after which it was rinsed with 1% NGS/TBS. Peroxidase activity was localized by development for 5-10 min with 0.015% H₂O₂ in 50 mM Tris-HCl, pH 7.6, with 0.75 mg/ml 3,3′-diaminobenzidine (Sigma). Development of the sections was directly monitored for maximal contrast under the X10 objective of a Zeiss Axioskop 20 microscope.

Chemical and immunochemical controls were used to define carbonyl-specific binding. Chemical reduction of free carbonyls and Schiff bases was performed by incubating sections with 25 mM sodium borohydride (NaBH₄) in 80% methanol for 30 min at RT before incubation with DNP-H. Specific reduction of Schiff bases, while leaving carbonyls intact, was performed with 50 mM sodium cyanoborohydride in 0.1 M phosphate buffer, pH 6.0, for 1 hr at RT. Immunochemical specificity was demonstrated by omission of the antibody to DNP or the DNP-H treatment. Immunoabsorption of the antibody to DNP was performed by incubating the antibody (1:100) with 5 mM pyruvate 2,4-dinitrophenylhydrazone (stock 1 mg/ml in ethanol) at 4C for 16 hr and comparing the resulting immunoreactivity with unabsorbed antibody that had been similarly treated with 5% ethanol.

Results

DNP-H derivatizes free carbonyls on biomacromolecules, leading to DNP adduction to the carbonylcontaining biomacromolecule. In contrast, reaction of DNP-H with Schiff bases formed between exogenous carbonyl compounds and proteinaceous lysine sidechains leads to release of soluble dinitrophenylhydrazones with no fixation of the DNP moiety to the tissue. By examining the sites of DNP adduction from reaction with DNP-H, localization of oxidative damage at the cytological level can be achieved. Without subsequent antibody enhancement, DNP-H showed only faint yellow staining of the large blood vessels in the brains of aged individuals. No staining of other brain cells or of the pathological changes of Alzheimer disease was seen (data not shown). DNP-H reaction and consequent DNP adduction to the walls of large vessels may not be surprising because extracellular matrix proteins abundant in vessels accumulate oxidative modifications, probably owing to their long half-life (Salomon et al. 1997). Consistent with this interpretation, the large vessels of the aged individuals stained more intensely than those of the young (data not shown) (Salomon et al 1997). Although this assay was specific for free carbonyls, e.g., it is blocked with prior reduction of carbonyls with NaBH₄, but not NaCNBH₃, the value of the direct detection assay with DNP-H is limited because it does not detect the pathological sites of known oxidative damage in the lesions of Alzheimer disease. Therefore, sensitivity was increased by coupling the chemical reaction of DNP-H to an enzymelinked immunocytochemical technique to detect bound DNP. This method showed prominent DNP adduct formation not only in large vessels (not shown) but also in the cell bodies and apical dendrites of the pyramidal neurons of the hippocampus in cases of Alzheimer disease (Figures 1A and 1C). In contrast, DNP adducts were undetectable in the neuronal cytoplasm of young or age-matched controls and in microvessels (Figure 1B). In addition to cytoplasmic staining of populations of large pyramidal neurons, intraneuronal neurofibrillary tangles were also recognized (Figure 1A, arrows). Other populations of neurons, e.g., dentate granule cells, were also stained, but with lesser intensity. Senile plaques (including dystrophic neurites, β-amyloid deposits, microglia, astrocytes, and adjacent oligodendrocytes) showed no reaction with DNP-H.

The specificity of the method to detect free carbonyls was demonstrated by chemical and immunochemical controls. In the former case, reduction of free carbonyls with NaBH₄ blocked DNP-H binding (Figure 1D), whereas reduction of Schiff bases with NaCNBH₃ was again without effect (not shown). Immunochemical validation was demonstrated by performing all of the following: omitting DNP-H treatment, omitting the antibody to DNP, or absorbing the antibody to DNP with the DNP-H derivative of pyruvate (Figure 1E). These immunochemical controls were critical to ensure the use of an antibody detection technique that is highly sensitive and specific, e.g., we found several different rabbit antisera to DNP that showed such high background staining for neurons that it was often difficult to note specific staining, whereas the rat monoclonal antibody showed no background staining.

Tissue was fixed in the non-crosslinking fixative Methacarn (methanol:chloroform:acetic acid 60:30:10), but we had similar results using unfixed or samples fixed in 80% ethanol. In contrast, tissue fixed in formaldehyde-based fixatives, as well as sections from tissue fixed in Methacarn but later treated with formaldehyde, displayed nonspecific DNP-H adduction throughout the section, making them unsuitable for study.

Figure 1

Free carbonyls were detected in cases of Alzheimer disease (A) not only in the pathological lesions of the disease, i.e., neurofibrillary tangles (arrows) but also in cytoplasm of neurons (large arrowheads). By contrast, in age-matched (B) and young controls none are labeled. DNP-H adduction in a case of Alzheimer disease (C) is completely blocked by prior reduction by NaBH₄ (D), omission of DNP-H, or adsorption of anti-DNP with DNP-pyruvate (E). ∗ indicates landmark present in each adjacent serial section (C-E). Bars = 50 μm.

Figure 2

Schematic representation of the relationship between aging- (——–) and disease (——)-related oxidative damage, showing that the accumulation of damage in long-lived extracellular matrix proteins (A) represents the vast majority of total modifications in tissue (C). Therefore, using whole-tissue homogenates, disease-related increases (Δ) in neuronal carbonyls (B) might not be appreciated in relation to the total (C). Using in situ techniques as described herein, the increased neuronal carbonyls are readily apparent. Note: Scale bars for carbonyls in A and C are a magnitude higher than in B.

Discussion

2,4-Dinitrophenylhydrazine (DNP-H), when coupled to immunochemical enhancement, can detect oxidative damage produced by a pathological condition, Alzheimer disease. It is of interest that the adduction of DNP-H was limited to cell bodies of neurons. This is essentially the same localization noted for adducts of the highly reactive lipid peroxidation product hydroxynonenal (Sayre et al. 1997) and for nitrotyrosine (Smith et al. 1997), the result of peroxynitrite-mediated damage. In all these cases, neuronal cytoplasm showed evidence of damage whereas glia and senile plaques showed none.

Absence of oxidative damage in the senile plaque is surprising in light of reports that senile plaques are the source of oxygen radicals in Alzheimer disease. First, β-amyloid, the primary component of senile plaques, is suggested to produce free radicals through a novel mechanism (Butterfield et al. 1994). Second, β-amyloid interaction with the RAGE receptor of microglia leads to free radical production (Yan et al. 1996). Third, senile plaques accumulate iron, which may be available to generate hydroxyl radicals through Fenton chemistry (Grundke-Iqbal et al. 1990; Connor et al. 1992; Levine et al. 1994; Jefferies et al. 1996; Smith et al. 1997). Yet, when we examine free carbonyls, as here, lipid peroxide adducts (Sayre et al. 1997), or peroxynitrite-mediated damage (Smith et al. 1997), we find no increase in damage in senile plaques. Absence of oxidative damage in senile plaques argues either that β-amyloid and the surrounding cells are not susceptible to oxidative damage, that radical production by senile plaques is balanced by antioxidant defenses, or that some neuron-specific macromolecules are very susceptible to oxidation. Although it can be argued that β-amyloid, because of its tight β-structure and low lysine content, may be less susceptible than lysinerich cytoskeletal proteins, cytoskeletal proteins also accumulate as paired helical filaments in the abnormal neurites surrounding senile plaques. Why cytoskeletal proteins exhibit no oxidative damage in senile plaque neurites and do exhibit damage while in the neuronal cell body as neurofibrillary tangles is a major unresolved issue of our findings. Nevertheless, our results do not support the senile plaque as the primary site of free radical imbalance. Because reactive oxygen species (ROS) have short diffusion distances through tissue, our findings instead implicate an ROS source within the cell body. Although the paired helical filaments might play this role (Yan et al. 1994; Smith et al. 1995a), their absence in most neurons showing increased free carbonyls, as well as their presence in senile plaque neurites, does not strongly argue for their involvement in neuronal oxidative damage. Another possibility is mitochondria, because mitochondrial respiration is associated with ROS production and because significant mitochondrial abnormalities (Davis et al. 1997) and altered metabolism (Smith et al. 1997) are found in AD. In this regard, the specificity of damage to neuronal vs glial cytoplasm is expected if ROS depends on high metabolic activity, because neurons are among the most oxidative cells of the body. This, together with possible differences in either the susceptibility or the clearance of damaged proteins, may be responsible for the observed difference. It is tempting to speculate that neuronal damage involves not only the cytoskeletal proteins that bear glycoxidative (Ledesma et al. 1994; Yan et al. 1995) and lipoxidative (Smith et al., unpublished observation) adducts but also nuclear components. Furthermore, the latter finding is consistent with the increased DNA fragmentation noted in Alzheimer disease (Su et al. 1994), because oxidative strand cleavage and repair of oxidative base modification are major causes of DNA fragmentation that may be incorrectly ascribed to apoptosis (Tsang et al. 1996).

Our finding of increased neuronal DNP-H adduction is consistent with biochemical studies demonstrating increased DNP-H-derived adducts by assay of tissue homogenates from normal aging and Alzheimer disease (Smith et al. 1991). However, in light of the large number of free carbonyls in the vasculature of all samples, particularly of the aged, the specificity and sensitivity of whole-tissue analysis to detect disease related increases, particularly for an age-related disease, are limited. Therefore, disease-related neuronal damage could be masked by damage to vessels when whole-tissue homogenates are used (Figure 2).

Another approach is to use the enhanced DNP-H immunochemical technique to identify specific proteins modified on immunoblots or by isolation (Smith et al. 1991; Shacter et al. 1994; Nakamura and Goto 1995). Coupling of cytological and biochemical approaches not only may give greater specificity and sensitivity to oxidative damage measurement but may also lead to a greater appreciation of its dynamic nature.