Animal care and handling were conducted in accordance with the National Institute of Health Guidelines for Animal Treatment, the Norma Oficial Mexicana NOM-036-SSA2-2002, and were approved by the Ethics Committee of the Medicine School at the National Autonomous University of Mexico.

Seventy-two male Wistar rats weighing 250–300 g were individually housed in acrylic boxes with free access to water and food (Purina, Minnetonka, MN, USA) and kept in a clear-air room. They were randomly divided into six experimental groups (n = 12 each): Group 1, exposed to an air stream free of ozone during 30 days; Group 2, exposed for 7 days to ozone; Group 3, exposed for 15 days to ozone; Group 4, exposed for 30 days to ozone; Group 5, exposed for 60 days to ozone; and Group 6, exposed for 90 days to ozone. The rats were exposed to ozone (0.25 ppm) daily for 4 h. Immediately after ozone exposure, the animals were returned to their home cages, and 2 h later they were processed for the following techniques: (1) Raman spectroscopy to follow the conformational structural changes of the Aβ peptides; and (2) immunohistochemistry Aβ 1–42.

Animals were put inside a chamber with a diffuser connected to a variable flux ozone generator (5 l/s) daily for 4 h. The procedure used has been described in other works (Pereyra-Muñoz et al., 2006; Rivas-Arancibia et al., 2010). The air feeding the ozone converter was filtered and purified. Ozone production levels were proportional to the current intensity and to the airflow. A PCI Ozone and Control System Monitor (West Caldwell, NJ, USA) was used to measure the ozone concentration inside the chamber throughout the experiment and to keep a constant ozone concentration.

The control group was treated in the same chamber described above using a flow of ozone-free purified air.

After the last exposure to ozone, the rats were deeply anesthetized using pentobarbital sodium (50 ml/kg). The brains were extracted, perfused with 4% paraformaldehyde, dehydrated, and embedded in paraffin. Afterwards, 20 μm slices were cut using a microtome (SHUR/Cut2500, General Data Healthcare). The sections were deparaffinized and hydrated before conducting the spectroscopic analysis.

The deparaffinized sections were first examined by light microscopy to define a study region where six different cells were randomly selected. Five single spectra were recorded for each selected cell at different cytoplasm points close to the membrane cell. In this way, we obtained 30 single Raman spectra for each section.

Raman spectra were obtained using a Witec Alpha300 R confocal microscopy system, with a laser excitation source of 532 nm at a power of 0.7 mW, 600 l/mm grating, and a 50× plane Fluor objective with a numerical aperture of 0.45 at a working distance of 4.5 mm. The spectral resolution used was 4 cm⁻¹ while a 576 × 400 pixel thermoelectric-cooled charge-coupled device (CCD) was used as detector.

Before each measurement, a system calibration was performed using a silicon chip in order to check the standard band position and the intensity. Data were measured over the 200–4000 cm⁻¹ spectral range. Each spectrum consisted of five accumulations with an integration time of 0.5 s to avoid tissue damage. As a reference, we used pure Aβ 1–42 (Sigma-Aldrich) to obtain its Raman spectra under similar experimental conditions.

The spectra were pre-processed for base line correction and cosmic ray removal using WITEC Project 2.0 software. After that, the 30 spectra obtained from each tissue section under study were vector-normalized with Wolfram-Mathematica software and, finally, each set of 30 spectra was averaged. This procedure was used for the control group as well as the groups exposed to ozone for 7, 15, 30, 60 and 90 days.

Additionally, a deconvolution of the amide I band (1600–1700 cm⁻¹) of the spectra was performed in order to visualize the subjacent sub-bands that compose the whole complex amide I band. The original spectra were baseline-corrected before the deconvolution analysis and smoothed with the Savitzky-Golay (SG) filter with five-point window. The deconvolution was performed using Origin 6.0 software, selecting the Gaussian peak type option. The number of sub-bands employed for this spectral analysis was defined by the second derivative of the whole amide I band. In addition, two bands at ~1604 cm⁻¹ and ~1612 cm⁻¹ were included with the amide I band during the deconvolution process since they were not baseline separated from the amide I feature. The values of the area under the curve of each sub-band were taken, in a first approximation, as proportional to the relative abundances of the peptide motifs associated with each sub-band.

Six animals from each group were anesthetized with pentobarbital sodium (50 mg/kg i.p.; Sedalpharma, Edo. de México, Mexico) and transcardially perfused with 4% paraformaldehyde (Sigma-Aldrich Chemie, Germany) in 0.1 M phosphate buffer (J.T Bakker E.U., New Jersey, Kentucky and Mexico; PB, Tecsiquim; pH 7.4). The brains were post fixed with 10% formaldehyde (J.T Bakker U.S.A., New Jersey, Kentucky and Mexico) for 24 h and embedded in paraffin (McCormick, St. Louis, MO, USA). Sagittal slices of the brain containing the hippocampi were cut at 5 μm on a microtome (ShurCut 2500. General Data Healthcare) and mounted on slides. Slices of each brain containing the hippocampi were deparaffinized and pretreated with heat retrieval solution (Biocare Medical, Concord CA, USA) and introduced into an electric pressure cooker (Decloacking Chamber, Biocare Medical) for 20 min. After being washed with distilled water and treated with 3% hydrogen peroxide (J.T Baker U.S.A., New Jersey, Kentucky and Mexico) for 5 min, the slices were rinsed again and treated with a blocking reagent (Background Sniper, Biocare Medical, Concord, CA, USA) for 10 min. They were washed with saline phosphate buffer (PBS, Merck, Germany) and incubated overnight at 4°C with Aβ 1–42 (purified monoclonal mouse antibody, diluted 1:50 (ANASPEC, Fremont, CA, USA). Sections were rinsed with PBS and treated with secondary antibody using Trekkie Universal Link (Starr Trek Universal HRP Detection; Biocare Medical) for 1 h. The sections were later washed with PBS, and then treated with Trekavidin–HRP Label (Starr Trek Universal HRP Detection, Biocare Medical) for 30 min. Bound antibody was visualized using 3,3′–Diaminobenzidine (DAB Substrate Kit, ScyTek, Logan, UT, USA) as the chromogen. The slices were washed with distilled water and contrasted with hematoxylin buffer solution (Concord, CA, USA). Representative brain slices from each group were processed in parallel. Following cover slipping with Entellan (F/550 ml Merck, Germany), the slices were examined with an Olympus BX41 Microscope (Japan) and photographed (Evolution VF-F-CLR-12 Media Cybernetics camera; Bethesda, MD, USA).

The Raman spectra of the “finger-print” region (600–1800 cm⁻¹) of rat hippocampus tissue were analyzed (Figure 1). The spectra presented correspond to the control groups (Figure 1a) and the group exposed to ozone for 90 days (Figure 1b), together with that of the pure Aβ 1–42 peptide used as a reference (Figure 1c). The complexity of the Raman spectra, both from a tissue (a and b) and from a pure peptide sample (c), is seen in a comparative way in Figure 1.

The regions conventionally used to make a structural study of proteins based on their Raman spectrum are amide I, amide III and skeletal backbone vibration C-C, which are shaded in Figure 1. It is important to note that the amide band I (1640–1680 cm⁻¹) of each spectrum presents a different profile in relation to bandwidth, wavenumber and intensity. From a direct observation of this band, it is not possible to estimate the dominant secondary conformation in each case. To achieve this purpose, the profile of the amide I band must be mathematically processed by a deconvolution analysis to obtain this information.

The regions of amide III (1200–1340 cm⁻¹) and skeletal backbone vibration C-C (890–980 cm⁻¹) help, in principle, to corroborate the secondary structure derived from the analysis of the amide I band. However, many other vibrational modes are also in this region, which come from other biomolecules that are present in the rat tissue and overlap with each other. This makes it difficult, in our case, to use these regions (amide III and skeletal backbone vibration C-C) to confirm the secondary structure of the Aβ 1–42 in our results.

The Raman spectrum of the commercial Aβ 1–42 peptide that is used as a reference for the experimental direct identification of the β-sheet conformation in the amide I band of this peptide is shown in Figure 2. The spectrum is dominated by a Gaussian type band centered at 1670 cm⁻¹ (Figure 2c) which is assigned mainly to the C = O stretching vibration of the β-sheet secondary structure in the Aβ 1–42 peptide (Barth and Zscherp, 2002; Rygula et al., 2013). Additionally, the deconvolution of the spectral region between 1600 cm⁻¹ and 1700 cm⁻¹ shows two minor bands: 1608 cm⁻¹ (Figure 2a) and 1634 cm⁻¹ (Figure 2b), which are assigned to the amino acid side chains and the aromatic amino acid ring mode vibration, respectively (Rahmelow et al., 1998; Rygula et al., 2013). The percentage of area under the curve for each of these bands is 89% for the β-sheet structure, 6% for the amino acid side chains and 5% for the aromatic amino acid ring mode vibration.

Confocal micrographs of hippocampal cells where the Raman spectra were obtained (Figure 3, columns A and B) along with the results from the mathematical deconvolutions of the amide I band (Figure 3, column C). The micrographs of the immunohistochemistry against Aβ 1–42 for each of the experimental groups (control and those exposed to ozone for 7, 15, 30, 60 and 90 days) are also shown (Figure 3, column D).

The deconvolution of the amide I band (Figures 3C1–C6) for the groups evidences the presence of vibrational modes assigned as follows: 1602–1610 cm⁻¹ amino acid side chains vibrations (Rahmelow et al., 1998), 1615–1635 cm⁻¹ aromatic amino acid ring mode vibrations (Rygula et al., 2013), 1644–1650 cm⁻¹ random coil structure (Sekhar et al., 2015), 1655–1668 cm⁻¹ α-helix secondary structure (Lippert et al., 1976; Rygula et al., 2013; Kurouski et al., 2015), 1670–1682 cm⁻¹ β-sheet secondary structure (Mahadevan-Jansen and Richards-Kortum, 1996; Wen, 2007; Kurouski et al., 2015), and 1683–1685 cm⁻¹ β-turn structural motifs (Barth and Zscherp, 2002). Table 1 summarizes the distribution of the sub-bands resulting from the deconvolution analysis for the control and all the ozone-exposed groups, indicating the percentage relative abundance and the assignment for each Raman vibrational mode.

In the control group the amide I band (Figure 3C1), is mostly composed by the corresponding α-helix peptide secondary structure (60%) (band 4), this band is centered at 1665 cm⁻¹. Additionally, there is a minor contribution of the vibrational motifs of the amino acids side chain (10%) (band 1), the aromatic amino acids ring mode (band 2) (5%), the unordered structure (10%) (band 3) and the β-turn type structure (15%) (band 5). On the other hand, the micrograph of the dentate gyrus shows absence of reactivity to Aβ 1–42 by immunohistochemistry (Figure 3D1).

The deconvolution of the amide I band for the groups exposed to ozone for 7 days (Figure 3C2), this band still shows a clear majority of the α-helix secondary structure conformation (47%) now centered at 1668 cm⁻¹ (band 4). Additionally, there is an increase in the abundance of the bands corresponding to the unordered structure (33%) (band 3), the vibrational modes of amino acids side chain (8%) (band 1), and the β-turn structure (12%) (band 5). For its part, the immunohistochemistry assay shows absence of immunoreactivity against Aβ 1–42 (Figure 3D2).

At 15 days of ozone exposure (Figure 3C3), there is an evident change in the amide I band profile, particularly in the region of the high wavenumbers where the formation of a shoulder at 1680 cm⁻¹ is observed (band 5); these changes are reflected in the deconvolution sub-bands. The predominant band is still the one corresponding to the α-helix secondary structure (45%), centered now at 1662 cm⁻¹ (band 4), together with the vibrations of the amino acids side chain (9%) (band 1) and the aromatic amino acid ring mode (13%) (band 2). The most significant change for this period of exposure is observed in the band associated to the β-turn structure (band 5), which increases its relative abundance (16%) and moves towards slightly lower wave numbers centering at 1681 cm⁻¹. For its part, in the microphotograph of the dentate gyrus in the hippocampus, the beginning of a diffuse extracellular immunoreactivity for Aβ 1–42 is observed (Figure 3D3).

In the group exposed to ozone for 30 days (Figure 3C4), the amide I band profile still shows the change in its profile for the high wavenumber side. The corresponding deconvolution now shows four sub-bands: one associated to the aromatic amino acids ring mode (6%) (band 2), the unordered secondary structure (30%) (band 3) and the α-helix secondary structure (41%) (band 4). The sub-band corresponding to the β-turn structures increases its abundance (23%) (band 5) and continues moving toward low wavenumbers. On the other hand, in the corresponding micrograph of the dentate gyrus in the hippocampus, Aβ 1–42 shows a clear increase of intracellular immunoreactivity, together with cell edema and extracellular diffuse immunoreactivity (Figure 3D4).

After 60 days of ozone exposure (Figure 3C5), the changes in the amide I band envelope are even more evident, a fact which is reflected in the relevant changes in the result of its deconvolution. Indeed, the sub-band corresponding to the α-helix secondary structure (band 4) considerably decreases its abundance (22%) while band 6 becomes predominant (52%) and is now centered at 1671 cm⁻¹, in clear accordance with the wavenumber of the previously assigned β-sheet structure of pure Aβ 1–42 presented in Figure 2. In a similarly evident way, the β-turn band (band 5) shows a significant decrease in its abundance (5%) and a deviation toward higher wavenumbers (1690 cm⁻¹). For their part, the bands of the amino acids side chains (band 1) and the mode of the aromatic amino acids ring (band 2) remain unchanged. In agreement with these spectral changes, the immunohistochemistry in the dentate gyrus of the hippocampus shows an intracellular increase of Aβ 1–42 immunoreactivity, which is also evident in the neuropil, together with cell swelling and pyknotic nuclei (Figure 3D5).

The deconvolution of the amide I band for the group exposed to ozone for 90 days (Figure 3C6) shows a similar distribution to that of the previous group. Once again, the most abundant sub-band (48%) corresponds to the β-sheet secondary structure (band 6) centered at 1672 cm⁻¹. The representative sub-band of the α-helix secondary structure (band 4) at 1655 cm⁻¹, decreases its relative abundance to 20%, while the one of the β-turn structure (band 5), centered at 1695 cm⁻¹, remains around 4%. Moreover, the sub-bands corresponding to the side chains of aromatic amino acids (band 1) and the breathing mode of the aromatic amino acids ring (band 2) have now a contribution of 15% and 13%, respectively. Once again, consistent with the spectral results, the micrograph of the dentate gyrus of the hippocampus shows increased Aβ 1–42 immunoreactivity, both intracellularly and at the neuropil, with morphological changes as swelling and pyknotic nuclei (Figure 3D6).

In relation with the optical micrographs of columns A and B in Figure 3, which shows the areas where the punctual Raman spectra were obtained, it is important to note the alteration of the morphology of neurons, in particular of the cell membrane, in correlation with the increase in the time of exposure to ozone. These observations are consistent with the results of the immunohistochemistry against Aβ 1–42 presented in column D of Figure 2.

Finally, in Figure 4 we show a general view of the trend of the major conformational secondary Aβ 1–42 motifs, resulting from the deconvolution analysis of the amide I band of the hippocampal tissue of rats, exposed to low doses of ozone for different periods. It is clear how the α-helix motif experiments a continuous decrease as a function of the increase ozone exposure time. The more pronounced decrease of the α-helix band between 30 and 60 days of ozone exposure, corresponds with the appearance and the strong increase of the band at 1671 cm⁻¹, associated with the β-sheet motif of the Aβ 1–42 sary structure. After 60 days of ozone exposure, the relative abundance of the α-helix and β-turn motifs seems to reach an almost steady state. On the other hand, there is not a clear trend in the changes of the β-turn and the non-ordered structures during the first 30 days of exposure to ozone. However, the near disappearance of these bands after 60 days of exposure seems to be correlated with the strong growth of the β-sheet secondary structure during the same period.

The chronic exposure to low doses of ozone leads to an increase in the number of cells expressing the Aβ 1–42 peptide from 30 to 90 days of exposure (Figure 5). Also, we can also observe that the average number of cells immunoreactive to beta-amyloid 1–42, increases in accordance with the increase of the abundance of the beta-sheet structure from 30 to 60 days, presented previously in Figure 4.