We analyzed samples from patients diagnosed with AD and control individuals without neurological disease. The age and gender of the subjects are listed in Supplementary Table S1. All samples were supplied by a brain bank (Banco de Tejidos CIEN, Madrid) and were analyzed anonymously. The transfer of samples was carried out according to national regulations concerning research on human biological samples. The Ethics Committee of the Universidad Autónoma de Madrid approved the study. In all cases, written informed consent was obtained.

Anti-chitin antibodies were generated as previously described (Walker et al., 1990, 1991). Briefly, rabbits (female New Zealand) were immunized with reacetylated chitosan suspended in phospate-buffered saline (PBS; 20 mM sodium phosphate buffer, pH 7.4, 0.15 M NaCl). After weekly subcutaneous injection (0.8 ml of PBS containing 0.1–0.6 mg reacetylated chitosan) for 4 weeks, animals were ear bled using a vacuum cuff. The antiserum was obtained and purified by affinity chromatography on protein A-agarose. Bound antiserum was eluted with 0.2 M glycine buffer, pH 2.9, neutralized, and dialyzed. To obtain the fraction that bound di-N-acetylchitobiose, the antiserum was further purified by affinity chromatography on ovalbumin-agarose. The unbound material, which had no binding activity toward ovalbumin (undetectable by western blotting), was stored at -20°C until needed. These antibodies were obtained at Mercer University (USA) before IACUC was instituted. Nevertheless, we followed all USDA-approved animal care regulations regarding immunization and subsequent ear-bleeding at that time.

Antibodies against C. famata enolase were obtained by injection of a purified fusion protein (maltose-binding protein plus enolase). The enolase gene was obtained by PCR amplification using the primers CGCCGCGGATCCATGGCCGTCACTAAGTTATT and CGCCGCGTCGACTTATAATTGAGAAGCAGCGT. The amplifed fragment was cloned into the pMAL vector (New England Biolabs, Ipswich, MA, USA). After expression, the fusion protein was purified on maltose columns and eluted with 0.5 ml PBS. Rabbits were injected every 3 weeks with 1 mg of fusion protein suspended in Freund’s adjuvant. A peptide corresponding to a region of fungal β-tubulin (DVVRREAEGCDS) was purchased from PolyPeptide Group (Strasbourg, France) and conjugated to keyhole limpet hemocyanin (KLH). This peptide sequence is conserved in several fungal species and is absent in human β-tubulin. Rabbits and rats were inoculated with 0.60 mg of β-tubulin peptide-KLH previously mixed with an equal volume of Freund’s adjuvant every 2 weeks, up to three times in total. The antibody titer and specificity of the antiserum were tested by immunohistochemistry. The protocols employed were approved by the ethics committee of Centro de Biologia Molecular “Severo Ochoa” (identification number: ES280790000180). Animal welfare and methods for sacrifice dictated by the European Union have been followed. Animals were injected with sodium pentothal as the anesthetic before bleeding.

CNS tissue was embedded in paraffin following standard techniques and cut into 5 μm sections using a microtome (Microm HM355s; Microm, Walldorf, Germany). For immunohistochemical analysis, paraffin was removed and sections were rehydrated and boiled for 2 min in 10 mM citrate buffer and then incubated for 10 min with 50 mM ammonium chloride. Tissue sections were then incubated for 10 min with PBS/Triton X-100 (0.1%) followed by 20 min with 2% bovine serum albumin in PBS. Sections were incubated overnight at 4°C with mouse monoclonal antibodies against human α-tubulin (Sigma), human phospho-PHF-tau, clone AT100 (Thermo Scientific), human neurofilament protein, clone 2F11 (Dako), or rabbit polyclonal antibodies raised against enolase, β-tubulin, or chitin, all at 1:50 dilution. Thereafter, sections were washed with PBS and further incubated for 1 h at 37°C with donkey anti-mouse IgG secondary antibody conjugated to Alexa 555 (Invitrogen) at 1:500 dilution for α-tubulin, tau and neurofilament, and donkey anti-rabbit IgG secondary antibody conjugated to Alexa 488 (Invitrogen) at 1:500 dilution for anti-fungal antibodies. To visualize nuclei, sections were stained with 4,6-diamino-2-phenylindole (DAPI; Merck) and treated with autofluorescence eliminator reagent (Merck). The use of this reagent is important to avoid autofluorescence since lipofuscin is present in the aging brain. All images were collected and analyzed on a Zeiss LSM710 confocal laser scanning microscope equipped with the upright microscope stand AxioImager.M2 (Zeiss), running ZEN 2010 software. The spectral system employed was Quasar + 2 PMTs. Images were deconvoluted using Huygens software (4.2.2 p0; Scientific Volume Imaging) and visualized and processed with Fiji/ImageJ software (NIH, Bethesda, MD, USA). Stacks of 3D images were collected with the high-speed, high-resolution A1R+ confocal microscope (Nikon) combined with an inverted microscope, running NIS Elements 4.40 software.

Fungal proteins were precipitated with 10% trichloroacetic acid. HeLa cells were grown in Dulbecco’s modified Eagle medium supplemented with 10% fetal calf serum. These cells were collected in sample buffer (0.37 M Tris–HCl, pH 6.8, 0.1 M DTT, 2% SDS, 17% glycerol, and 0.024% bromophenol blue) and boiled for 5 min. Fungal and HeLa cell proteins were fractionated by SDS-PAGE (15% polyacrylamide), transferred to nitrocellulose membranes by wet immunotransfer and processed for western blotting after blocking the membranes with 1% bovine serum albumin. Incubation with specific rabbit antibodies for peptide β-tubulin-KLH and enolase at a 1:200 dilution and specific mouse antibody for human α-tubulin at a 1:500 dilution was performed for 1 h. Donkey anti-rabbit and sheep anti-mouse IgG horseradish peroxidase-conjugated antibodies (Amersham Biosciences) and the ECL kit (Amersham) were used to detect bound antibodies.

Our previous findings strongly indicated the presence of fungal structures in brain tissue from AD patients as revealed by positive staining with antibodies raised against whole fungal cells (Pisa et al., 2015a,b). Although the antibodies were specific and did not recognize human antigens, the number of proteins and other cellular components immunoreacting with the polyclonal antibodies were unknown. We therefore generated three antibodies against individual fungal components to further evaluate the existence of these specific proteins/polysaccharides in brain tissue. Anti-chitin antibodies have been successfully tested against a number of fungal pathogens (Walker et al., 1990, 1991) and selectively recognize chitin, which is not present in human tissues. Additionally, we raised an antibody against the glycolytic enzyme enolase cloned from C. famata, and against a peptide specific for fungal β-tubulin. The three rabbit polyclonal antibodies were first tested using immunofluorescence microscopy on different Candida species. Anti-chitin, immunoreacted strongly with C. parapsilosis and C. tropicalis, while its reactivity with C. albicans. C. krusei, and C. glabrata was weaker, suggesting that the chitin present in the latter yeast species is differentially recognized by the antibody perhaps due to differences in the structure or in the amount of chitin present in the cell wall (Supplementary Figure S1A). Additionally, the anti-enolase antibody immunoreacted strongly with C. krusei and C. glabrata, whereas the anti-β-tubulin antibody immunoreacted strongly with C. albicans. C. tropicalis, and C. krusei. Rat polyclonal antibodies raised against the β-tubulin peptide also immunoreacted with the different Candida spp. (Supplementary Figure S1A). Thus, all three rabbit antibodies and a rat antibody immunoreacted with Candida spp. to different degrees. Western blotting analyses showed that anti-enolase and anti-β-tubulin antibodies recognized the corresponding protein bands in C. famata but not in HeLa lysates (Supplementary Figure S1B). Conversely, a commercial anti-human tubulin antibody detected tubulin in HeLa lysates but not in C. famata. Further assessment of the selectivity of the three fungal antibodies was performed by testing their reactivity against human HeLa cells by immunocytochemistry. As expected, none of the three antibodies immunoreacted with components of HeLa cells (Supplementary Figure S1C).

Chitin is a polysaccharide of N-acetylglucosamine that is an integral part of the fungal cell wall. Chitin-like bodies have been observed in AD brains stained with the flurochrome dye calcofluor (Castellani et al., 2005). Moreover, levels of chitinase (chitotriosidase), a human enzyme synthesized by macrophages, are very high in serum and CSF from AD patients (Choi et al., 2011; Watabe-Rudolph et al., 2012; Rosen et al., 2014; Melah et al., 2015). We therefore tested for the first time specific anti-chitin antibodies to survey potential fungal structures in AD brains. Tissue sections from the entorhinal cortex of 10 AD patients were double stained with an anti-chitin antibody (1:50 dilution, shown in green) and a mouse monoclonal anti-human α-tubulin antibody, which was used to identify microtuble structures (shown in red). Cell nuclei were visualized by DAPI staining (shown in blue; Figure 1). A number of structures were stained with the anti-chitin antibody (Figure 1); some of which were in the form of punctate bodies or yeast-like structures. On several occasions, the stained structures were located inside the nucleus, while in other sections they were detected around the nucleus or were extracellular. Hyphal structures could also be readily detected with the anti-chitin antibody (Figure 1, AD10). In summary, different structures that immunoreacted with the anti-chitin antibody were detected in all 10 AD patients examined. Since the morphology of these structures is similar to fungal cells and hyphae, and they most probably contain chitin, it seems plausible to suspect the existence of mycoses in the entorhinal cortex of the patients.

To further assess the presence of specific fungal components in AD brain tissue, we employed rabbit polyclonal anti-enolase and fungal anti-β-tubulin antibodies. First, tissue sections of the entorhinal cortex were double stained with an anti-enolase (green) and a human anti-α-tubulin antibody (red; Figure 2, Upper panels). Entorhinal sections were also double stained with anti-β-tubulin (green) and a mouse monoclonal anti-neurofilament antibody (red; Figure 2, Lower panels). In all cases, nuclei were visualized by DAPI staining (blue). Fungal enolase was clearly detected in brain tissue from AD patients. The immunostained structures were similar to those described with anti-chitin antibodies but the staining was more robust, perhaps due to the better reactivity of the antibody against a protein rather than a polysaccharide. Of particular interest was the finding of intranuclear fungal structures (Figure 2, panel AD6). Immunostaining with the anti-β-tubulin antibody highlighted the presence of a number of punctate bodies in the cytoplasm of a neuron (Figure 2, Lower panel AD2). The immunostained punctate bodies are reminiscent of those observed after yeast infection of cultured cells or mice (Pacheco et al., 2007; Pisa et al., 2015a). In addition, several rounded fungal cells were clearly visible in neuronal nuclei (Figure 2, Lower panels). As controls for the different antibodies employed in this work, immunohistochemistry analyses of AD sections were carried out without primary antibodies. Certainly, in presence of anti-enolase antibodies fungal structures are detected, whereas when this primary antibody is omitted the green structures are not observed (Supplementary Figure S2). Thus, when the secondary antibody against rabbit polyclonal antibodies is employed, no fungal structures are revealed. On the other hand, omission of the secondary antibody that recognizes mouse monoclonal antibodies does not visualize human α-tubulin.

To validate these results, we carried out double immunostaining using rabbit polyclonal anti-chitin antibodies and rat polyclonal fungal anti-β-tubulin antibodies. Curiously, whereas some putative fungal structures were recognized by both antibodies, other structures were immunostained with either one or the other of these antibodies (Supplementary Figure S3). Moreover, when the same structure was recognized by both antibodies, the pattern of staining observed with each antibody could differ (see Lower panels in Supplementary Figure S3).

To further determine whether some of the putative fungal structures were intranuclear, we carried out orthogonal projection analysis. Figure 3 shows that the anti-chitin antibody immunoreacted with intranuclear material from patients AD7 and AD10 and an intranuclear fungal hypha is clearly visible (Figure 3A). Remarkably, 3D analysis of one neuronal nucleus demonstrated the presence of one yeast cell inside the nucleus. Since intracellular infection by fungal cells requires that they be alive, we conclude that infection of the neurons occured when the patient was alive, and was not the consequence of post-mortem contamination.

Having shown that fungal bodies could be detected in the entorhinal cortex of several patients, we next sought to evaluate different CNS regions. To this end, sections from the lateral frontal cortex (LFC), cerebellar cortex (CEC), entorhinal cortex/hippocampus (ERH) and choroid plexus (CP) of one AD patient were immunostained with the three anti-fungal antibodies described above (Figure 4, shown in green). For double immunostaining studies, we used monoclonal anti-human phosphorylated tau with anti-chitin, mouse monoclonal anti-human α-tubulin with anti-enolase, and monoclonal neurofilament antibody with anti-β-tubulin (shown in red). Results showed that a number of fungal structures were detected in the four different regions, pointing to the idea that fungal infection is disseminated in different brain regions (Figure 4). It is possible that the fungal species and/or the structures recognized by each antibody could be different. Once again, the morphology of the structures immunostained with the anti-fungal antibodies was similar to those described in the entorhinal cortex. Accordingly, intranuclear yeast-shaped cells were found in the LFC with anti-chitin antibodies. Also, hyphal structures were detected in the LFC, CEC and ERH regions with the anti-β-tubulin antibody. In CP, several yeast cells of about 1–2 μm are evidenced inside a blood vessel using the anti-enolase antibody. Collectively, these results lead us to conclude that specific fungal macromolecules such as chitin and proteins are detected in fungal structures widespread in different brain regions from the AD patient analyzed.

Corpora amylacea (CA) are small rounded bodies of 10–50 μm that are very abundant in the CNS of patients with neurodegenerative diseases, including AD (Rohn, 2015; Pisa et al., 2016). The composition of CA has been analyzed in some detail. They mainly contain polyglucans and only a small percentage (4%) corresponds to proteins (Robitaille et al., 1980; Nishimura et al., 2000). The precise origin of CA remains enigmatic, but it is thought that they accumulate in elderly people and their formation occurs over long periods (Sfanos et al., 2009). We recently found that fungal proteins can be detected in CA with polyclonal antibodies raised against whole fungal cells (Pisa et al., 2016). Thus, it was of interest to examine whether specific fungal components could be detected in CA using the new antibodies described here. Thus, CA were localized from entorhinal cortex brain sections and were immunostained as described (Figures 1 and 2). Of note, fungal enolase could be clearly detected in the vast majority of CA, which showed a high immunoreactivity with the antibody (Figure 5, Upper panels). By contrast, CA showed low reactivity with the anti-β-tubulin antibody, and not all CA were immunolabeled (Figure 5, Middle panels). Curiously, anti-chitin reactivity was not detected in any of the CA, suggesting that chitin is not incorporated into these bodies (Figure 5, Lower panels). Therefore, we conclude that some specific fungal proteins can be detected in CA, and enolase is a particularly good marker to use for detection. This may be due to the fact that enolase is both a secreted and cytoplasmic protein, whereas β-tubulin is cytoplasmic and forms part of the fungal cytoskeleton. The finding that fungal chitin is not found in CA may be due to the fact that CA are mainly composed of polyglucans.

Previous work from our group has found that control subjects do not contain appreciable amounts of fungal proteins in brain sections (Pisa et al., 2015a,b). Nevertheless, to compare the burden of chitin, enolase and β-tubulin with AD brain tissue, we performed immunohistochemistry on brain sections from control individuals. We evaluated six control subjects (described in Supplementary Table S1) and the more representative results are shown in Figure 6. We failed to detect chitin immunoreactivity in brain tissue from control subjects (Figure 6I–L). Similar results were obtained using anti-enolase and anti-β-tubulin (Figure 6A–H). Rare positivity could be detected in control samples (Figure 6M–P). Nevertheless, this low burden of fungal infection in control subjects may be of interest to understand some symptoms such as the stimulation of the immune system in elderly people (Cribbs et al., 2012; Heneka et al., 2015). In conclusion, immunohistochemistry analyses with three new antibodies comprehensively demonstrate the presence of fungal components in brain tissue from AD patients, but these fungal structures are very scanty in control individuals.