Post-mortem human brain tissues were obtained from the NSW Brain Tissue Resource Center (BTRC) in accordance with the University of Sydney's Human Research Ethics Committee (HREC# 2019/HE000531). BTRC protocols have been previously described (3) and updated further in this issue (Stevens et al., 2025). AD cases were pathologically confirmed using an abridged version of the current consensus criteria (13). Here, 7 μm sections were cut from 2.5 × 2.5 cm × 3 mm formalin-fixed, paraffin-embedded (FFPE) standard block (3 weeks fixation in 15% neutral buffered formalin) or prolonged fixation (5 years in 10% neutral buffered formalin). For three cases (AD1, 2, and 3), a second adjacent block to the original standard block, incorporating the entorhinal cortex, was re-cut from long-term fixed tissue. One additional AD case also utilized here had the standard block only (AD4). Sections mounted on SuperFrostPlus+ slides and dried overnight.

After rehydration, slides were subjected to a 15-min incubation with 90% formic acid to enhance amyloid-beta antigenicity, then washed for 3 × 5-min in RO H₂O. Heat-induced epitope retrieval (HIER) was carried out using a BioCare DeCloaker for 30 min at 110 °C and 6 psi. Here, citrate buffer (pH 6) and Opal AR6 were used for traditional mIF and Opal protocols, respectively. After cooling, slides destined for traditional mIF were briefly washed in RO H₂O, then incubated for 10 min in 3% hydrogen peroxide (H₂O₂) in methanol to quench endogenous peroxidase activity. Opal AR6 buffer contains a proprietary endogenous peroxidase and therefore did not require this step. Prior to staining, slides for both protocols were demarcated with a PAP pen and permeabilised for 5 min with Tris-Buffered Saline with 0.1% Tween 20 (TBS-T).

Traditional mIF was achieved using iterative double labeling cycles separated by beta-mercaptoethanol (BME) elution buffer (EB). Each cycle consists of a 1-h blocking step, followed by overnight incubation of the primary antibody mixture in the dark. After 3 × 5-min washes in TBS-T, slides were incubated for 20 min with the secondary fluorescent mixture consisting of anti-rabbit AF488, anti-mouse AF568 (both at 1:200), and DAPI. Primary antibody combination and dilutions are detailed in Table 1.

Opal immunolabelling of TMAs followed Akoya Bioscience protocol, wherein proteins of interests are iteratively labeled over several cycles of single detection. In brief, each cycle consisted of a 10-mine blocking step with Antibody Diluent/Block, followed by primary antibody incubation using the same diluent (Table 2). Slides were washed 3 × 5-min in TBS-T, then incubated with Opal Anti-Ms + Rb HRP-secondary antibody for 10 min. Slides were washed again prior to the 10 min incubation with a specific Opal fluorescence reporter (Table 2). Antibody elution was performed in a similar manner to the traditional mIF protocol. Finally, slides were incubated for 5 min using DAPI. After sequential washing with RO H₂O and TBS-T, slides were mounted with ProLong™ Diamond Antifade (Invitrogen, Thermo Fisher Scientific, Carlsbad, CA).

The preparation of the BME-based elution buffer (EB) has been previously reported (10, 14, 15). In brief, the EB was made fresh for each run and preheated to 56 °C in a shaking water bath for a minimum of 30 min before incubation of stained sections for a further 30 min with moderate agitation. This was further modified to immersion only, without agitation to minimize potential tissue damage. To ensure proper removal of BME from tissue, slides were first rinsed with ddiH₂O, followed by 4 × 15-min immersion washes in RO H₂O. Slides were returned to TBS-T for subsequent staining cycles.

Traditional mIF sections were imaged on a Zeiss Axio Scan.Z1, Carl Zeiss Microscopy GmbH, Jena, Germany. Slides were imaged at 20 × using the apochromat 20 × /0.8NA, M27 objective, with a resolution of 0.32 micron per pixel. Exposure times were set per channel, per case, to account for tissue variability. Opal tissues were imaged using the PhenoCycler™-Fusion system (Akoya Biosciences, Marlborough, MA). Here, exposures were automatically selected for each fluorophore. Images were acquired using the 20 × objective, providing a maximal pixel resolution of 0.5 micron per pixel. Image tiles were selected in Phenochart and processed in inForm^®, where spectral unmixing was performed to separate each fluorophore emission spectrum and correct for autofluorescence signals from lipofuscin, elastic membranes, and residual red blood cells.

Aligning traditional IF images was done using QuPath (16) and Fiji (17). In brief, the same region of interest (ROI) was selected across the three iterative IF images and exported as TIF files. These composite files were initially separated into their constituent markers using Image > Color > Split Channels, generating independent grayscale images for each fluorophore. All relevant channels were then reassembled via Image > Color > Merge Channels using DAPI as reference for registration.

The four neuropathologically confirmed AD cases had a mean age was 92.8 years (Table 3). 7 μm sections from the two FFPE blocks of three cases (standard and long-term), along with standard blocks from AD4, were cut and held at 4 °C prior to TSA-based (OPAL) or non-TSA (traditional) mIF.

Typically, we apply HIER between the rounds in mIF. To explore the utility of a non-HIER approach, we substituted the BME protocol for the second and third rounds of a three-round, double plex traditional mIF. The order of antibody pairs was dictated by the formic acid pre-treatment for beta-amyloid (Aβ) (1st round) and separation of the two tau antibodies [phospho-tau (AT8-Tau) and total tau]. These two antibodies have very similar staining profiles in AD tissue, but the former is the industry standard (18), while the latter subjectively shows more non-NFT tau pathology (12, 19). Pairing of antibodies was then species-specific: 1: Aβ (mouse, Monoclonal) and NeuN (rabbit, polyclonal), 2: AT8-Tau (mouse, monoclonal) and Iba1 (rabbit, polyclonal) and 3: GFAP (mouse, monoclonal) and Total tau (rabbit, polyclonal; Table 1). The BME protocol showed no apparent changes in antigen retrievability or tissue damage over the latter two rounds of mIF (Figure 1). Notably, there was no obvious non-specific binding to residual primary antibodies in the preceding rounds. However, tissue damage did arise from our coverslip removal protocol between the 2nd and 3rd round (Figure 1I) where agitation with a rotary shaker was used to accelerate dislodgement. We have since omitted the agitation with the caveat that this takes longer. However, the fact that coverslip removal is required between rounds remains an inherent risk of mIf. mIF has been a key development in studying disorders like AD, as hallmark pathologies and their inter-relationships with different brain cells can be both characterized and quantified. For example, Figure 1J shows the typical reactive astrocytes, with their prominent soma and processes, adjacent to and surrounding the Aβ plaque, but not necessarily NFTs (Figures 1G, H).

We next set out to determine whether the BME protocol, above, and traditional mIF would show the same efficacy in long-term fixed tissue as seen in the standard blocks. Long-term fixed tissue has known problems with epitope retrieval (4), but it may also be more susceptible to tissue damage from repetitive HIER. To test this idea, we cut a second block adjacent to the standard entorhinal cortex block from three AD cases (AD1–3) that had each been stored for ~5 years in 10% formalin. We repeated the above traditional mIF protocol in parallel with the standard (short fixed) and long fixed tissue sections. There were no significant differences in antigen retrievability or tissue damage for short and long fixed tissue for AD1 (Figure 2) or the other two AD cases (data not shown). All five antibodies successfully labeled their respective antigens across the three cycles. However, an increase in background was consistently observed for NeuN staining in long-term fixed tissue. This negative correlation between NeuN staining and fixation time with has been previously reported (20) and is likely to reflect NeuN's epitope availability being relatively sensitive to cross-linking. Figure 2J shows the complete overlap of AT8-Tau and total Tau antibody signals in NFTs while Figure 2R shows how microglia are often found within Aβ plaques, presumably phagocytosing Aβ over a period to convert early diffuse plaques into mature, cored plaques. The plaque-microglial interactions seen here can be compared to the circumferential pattern of astrocytes as they combine to “wall-off” Aβ plaques from normal brain parenchyma (shown in Figure 1J).

The TSA-based miF techniques like OPAL have theoretically greater flexibility than traditional approaches with the accompanying Akoya hardware having a nine-channel capacity (8 protein markers + DAPI). The tyramide covalent bonding also promises greater signal strength from FFPE tissue, particularly with markers like NeuN, which are adversely affected by fixation time. Here we compared traditional mIF described above with a six-round, single-plex OPAL run. Both techniques worked well with the BME protocol (Figure 3). Staining of Aβ plaques (Figures 3B, I) and NFTs (Figures 3C, J) were similar for both techniques. However, the signal intensity of all cell-specific markers for microglia (IBA1; Figures 3D, K), astrocytes (Figures 3E, L) and neurons (Figures 3A, H) were better with OPAL. Neuron detection was far greater with OPAL, consistent with the idea that epitope retrieval is fastidious for this pan-neuronal marker (4). The merged OPAL image also showed the common occurrence of tau-positive dystrophic neurites at the periphery of an (neuritic) Aβ plaque (Figure 3N) (21). Dystrophic neurites (dendritic origin), along with neuropil threads (axonal) and the NFTs in neuronal cell bodies are the three different types of neurofibrillary or Tau pathology seen in the AD brain (22).

A major advantage of TSA-based miF over traditional mIF is that only one imaging session is required at the end of the procedure, compared to imaging between rounds. This not only saves time, but traditional mIF requires intensive image alignment and registration. Furthermore, tissue movement can occur between rounds, including from coverslip removal, preventing accurate image superimposition. This issue was demonstrated in the current study through our use of two different antibodies that identify the same Tau Pathology (12). The Tau antibodies were included in rounds 2 and 3, respectively. When the images from the three rounds were superimposed on each other, the NeuN-positive granule cells in the dentate gyrus (stained in round 1) are clearly seen, but NFTs in the adjacent subiculum (white box and inset) are duplicated due to the slight movement of the tissue between rounds two and three (Figure 4).

The discipline of neuropathology has seen an evolution of staining techniques over the last 30 years that has allowed an increasingly greater number of targets to be demonstrated in the same FFPE section. This evolution is well-illustrated by Braak's staging of AD, which began with modified silver staining (22) but was then superseded by AT8-Tau immunostaining (18). Immunostaining with chromogens was then gradually replaced by fluorophores and immunofluorescence because of the greater scope for multiplexing and a better signal-to-noise ratio. However, this was a slow process in human post-mortem tissue, due to autofluorescence largely stemming from immersion fixation protocols. The previous use of quenching agents like Sudan black and TrueBlack (19, 23), and more recently sophisticated unmixing software with algorithms to subtract the autofluorescence (12) has largely circumvented the issue. mIF is a very powerful technique for observing cell-to-cell interactions in the normal brain and how these change in and around pathological entities in disease. However, the advent of single-cell biology and the rise of cell subtype markers have led to an ever-increasing demand for greater breadth of multiplexing. Furthermore, the latest spatial transcriptomic technologies are now being combined with multiplex immunofluorescence in the same section. This is a powerful paradigm switch because previously molecular studies were done on frozen tissue using the other hemisphere to that used for miF, and one needed to assume that pathology was equivalent on both sides of the brain, if matching the extent of pathology to molecular signatures.

TSA-based mIF is relatively expensive requiring commercial kits and associated hardware. Traditional mIF remains a good option for labs on a budget but both are potentially limited by the multiple cycles of HIER. As a human post-mortem brain bank, we were interested in testing a gentler stripping protocol across our short and long-term fixed tissue. Our results show similar performance between the commercial OPAL approach and traditional mIF and that a BME-based stripping protocol can perform as well as HIER, with less risk of tissue damage.

Ultimately, the selection of either TSA-based or traditional mIF should be driven by the biological questions of each study, as neither technique is inherently superior but rather offers distinct contextual advantages. Traditional mIF provides a straightforward and cost-effective approach for protein detection using either primary- or secondary-conjugated fluorescent antibodies. Given its simplistic workflow, the detection of 3–5 markers of interest is achievable within a single day for standard sections (5–10 μm), or two or more days for thicker sections (50–100 μm) that are more suitable for confocal microscopy (24). However, traditional mIF capabilities are fundamentally constrained by the number of compatible primary and secondary antibody pairings. This limits the technique's ability to deeply phenotype complex samples without resorting to staining serial sections or performing multiple sequential staining rounds. Indeed, researchers have used traditional mIF to stain 100 targets in post-mortem brain tissue across 10 staining rounds (11). However, such endeavor requires careful consideration of both species and immunoglobulin subtypes. This can constrain certain target to specific fluorescent channel. For Opals, there is an inherent flexibility wherein any rabbit or mouse antibody can be paired with any of the eight fluorescent reporters. This enables dynamic multiplexing, wherein higher priority or more lowly expressed targets can be paired with higher intensity fluorescent reporters, and vice versa.

In contrast, TSA-based platforms rely on the irreversible and sequential deposition of fluorescence reporters onto tyramide residues near the markers of interest (8). As such, multiple antibodies from the same host can be used without risks of cross-reactivity. The enzymatic amplification afforded by TSA's secondary antibody significantly increases signal intensity, enabling the detection of proteins with low-abundance or those with fixation-compromised epitopes. Although current TSA platforms can support up to 9-plex (inclusive of DAPI), fluorophore intensity and performance vary across tissue types. For example, Opal 480 performs poorly in brain tissue due to strong autofluorescence from lipofuscin, red blood cells, and elastin-rich structures, and this is all further magnified in long-term fixed tissue. Whilst TSA-based platforms offer a clear path toward higher plex immunolabelling, the costs associated with hardware (i.e., imaging machines such as PhenoImager HT and Phenocycler) and spectral deconvolution software (e.g., InForm) can be prohibitive, and its workflow, like traditional mIF does inherently increase with “plexity.”

Here we chose to explore the AD brain with its distinctive hallmark pathologies. It is here that mIF really comes into its own in terms of observing how major brain cell types interact with each other and the pathologies. There is now a demand for cell subtype markers with the latest single-cell studies in the AD brain suggesting 16 different types of microglia, for example (25). Alternative multiplexing techniques like Phenocycler Fusion currently offer 40-plex, but these kits are also expensive and limited to smaller, focused studies. Lastly, spatial platforms such as Visium (26) offer the opportunity to study the molecular make-up of cells relative to their proximity to plaques and tangles (27). This is an extremely powerful paradigm when combined with mIF. Spatial biology is likely to usher in a renaissance in neuropathology and a period of unheralded discovery toward closing the knowledge gap between the molecular machinery in the human brain and complex behavioral traits and anomalies.

In summary, traditional mIF, when used in combination with BME protocols, remains a powerful method for visualizing the biology and pathology in the human brain. It performs well in short and long-term fixed tissue. TSA-based systems are superior in terms of signal-to-noise and scalability but are more expensive. They do have the major advantage of requiring a single imaging session, compared to multiple cycles of strip-stain-image, and a greater risk of tissue movement and distortion.