Immunohistochemical Markers for Quantitative Studies of Neurons and Glia in Human Neocortex

Human Tissue

The use of archival human brain tissue for this study was approved by the Danish Biomedical Research Ethical Committee for the Region of Southern Denmark (permission number S-20070065). All brains included in the study had been subjected to routine neuropathological examination, and no neuropathological changes had been observed.

For screening of antibodies, two tissue arrays were made from identical samples (tissue array A; Table 1): one as paraffin embedded material and one as frozen material. Both contained 16 specimens from the neocortex, white matter, or pons from normal adult, perinatal, and fetal brains. Samples from Donors 2, 3, 5, 8, 9, and 10 were kindly provided by Dr. Bente Pakkenberg, The Research Laboratory for Stereology and Neuroscience, Bispebjerg Hospital, Copenhagen, Denmark (Danish Biomedical Research Ethical Committee for the Region of Sealand, permission number KF 01-068/98 and KF 01-328/98). Samples from Donors 1, 4, 6, and 7 were from the biobank at the Department of Pathology, Odense University Hospital (OUH), Odense, Denmark.

To analyze the effect of fixation time on the staining result, tissue array B was composed of paraffin-embedded material from specimens sampled in the postcentral gyrus of the neocortex from 29 human brains fixed by immersion, ranging from 8 days to 10 years (Table 2). The samples from Donors 1 and 8–29 were from the biobank at the Department of Pathology, OUH. Samples from Donors 2–7 were from brains donated to the Department of Anatomy and Neurobiology, The Institute of Medical Biology, University of Southern Denmark. The material from Donors 2–6 was fixed by perfusion of the head with 4% Lillies phosphate-buffered formalin solution (PBFS), pH 7.0 (Lillies PBFS: 4% w/v formaldehyde, 75 mM phosphate buffer; Sygehusapotek Fyn, OUH) through the internal carotid artery. The brain was removed from the skull and immersed in 4% paraformaldehyde (PFA) in 0.15 M Sørensens phosphate buffer, pH 7.4, and fixed for 2 weeks at 4C before routine pathology. For long-term storage, this material was kept in 0.1% PFA in 0.15 M Sørensens phosphate buffer, pH 7.4, at 4C. The brain from Donor 7 was fixed by immersion only.

Paraffin sections were prepared as 5-μm-thick serial sections mounted on 75-μm capillary gap slides (S2024; Dako Cytomation, Dako Nordic a/s, Glostrup, Denmark). Sections were stored in sealed boxes at 4C. Sections from tissue array A were used within 6 months after sectioning, and sections from tissue array B were used within 2 weeks. Cryostate sections were prepared as 10-μm-thick serial sections, mounted on 75-μm capillary gap slides, and stored in sealed boxes containing silica gel at −20C and used within 1 year after sectioning.

Immunohistochemistry

Analysis of Staining Results and Statistics

The staining results from the screening of antibodies in tissue array A were scored by two investigators (LL and IDS) on a scale from 0 to 5: 0, no staining; 1, weak specific signal; 2, weak specific signal giving the impression of lack of stained cells or structures or high level of nonspecific staining; 3, acceptable level of specific signal; 4, high level of specific staining considered optimal for cell counting; 5, very high level of staining that, however, compromised the identification of the cellular nucleus. The results of the stainings for NeuN, GFAP, CNPase, and CD45 in tissue array B were scored by two investigators (BF and HDS) blinded to the fixation time and origin of the specimens. For each staining, five sections resulting from three separate experiments were rated on a scale from 0 to 4 (Table 4). The classification was inspired by the recommendations given by NordiQC (www.nordiqc.org) for evaluation of immunohistochemical staining results. For each staining, a mean score, SD, and SEM were calculated. Statistical analysis for correlation between scores and fixation time was done using the Spearman r test in GraphPad Prism 4 for Mac OS×. The Spearman r test is non-parametric and appropriate for statistics on rank-ordered data. Because the sex of the donor, the PMI, and the method of fixation by either immersion or perfusion represent potential confounding parameters to the expression of the antigen in the specimens, the data were stratified according to these parameters before the statistical analysis. The level of statistical significance was set at p<0.05. In figures, p values are indicated as follows: *p<0.05, **p<0.01, and ***p<0.001.

Photodocumentation

Sections were photographed using an Olympus DP70 digital camera mounted on an Olympus BX51 microscope and connected to a PC with the Olympus DP software (Olympus Danmark a/s, Ballerup, Denmark). Adobe Photoshop CS was used to adjust contrast of and to set up figures.

Identification of Immunohistochemical Markers for Visualization of Neurons and Glia

Effect of Epitope Retrieval and Choice of Detection System

In the initial screening, checkerboard titration was performed for each of the antibodies. The results were evaluated qualitatively selecting the combination of epitope retrieval and antibody concentration, resulting in a high level of specific signal and low level of background staining. In our hands, HIER yielded better results than PrER, even in antibodies in which the manufacturer suggested the use of proteolytic pretreatment such as MBP, GFAP, CD68(KP1), CD68(PGM1), and Ki67 (shown for GFAP and MBP in Figure 4). This observation was in agreement with results in recent test reports on www.ukneqasicc.ucl.ac.uk. No difference was observed using the microwave oven or the pressure cooker as a heating source (data not shown), whereas the choice of epitope retrieval buffer (TEG, TRS, Borg, or citrate buffer) greatly influenced the resultant staining depending on the specific antibody (Table 5). The antibodies against CD39 and CD169 gave rise to specific signal in cryostat sections only.

The performance of three detection systems for staining of short- and long-term fixed material was evaluated qualitatively. The Dako ChemMate LSAB-kit gave rise to a high signal level, with considerable background staining in tissue fixed for 24 hr (Figure 5A). The background staining did not compromise results in tissue stored in 4% Lillies PBFS, but the signal level was lower than that of both the Envision+ peroxidase-labeled polymer detection system and the Vectastain Universal Elite ABC kit (Figures 5D–5F). The Envision+ peroxidase-labeled polymer detection system resulted in very high signal levels, visualizing even fine astroglial processes both in tissue fixed in 4% Lillies PBFS for 24 hr (Figure 5B) and 10 years (Figure 5E). The Vectastain Universal Elite ABC kit produced slightly weaker specific signal, yet very low background staining both in short- and long-term fixed specimens (Figures 5C and 5F). Finally, to unveil nonspecific staining caused by the secondary antibodies or detection reagents, sections incubated without the primary antibody or with an inert isotype-specific antibody were included in each staining experiment. With few exceptions, these sections presented no signal, but when applying HIER and the Dako ChemMate LSAB system for staining of samples from adults fixed for 24 hr, a weak staining of white matter, neuropil, and pyramidal neurons was observed (Donors 6 and 7 in tissue array A; data not shown).

Effect of Fixation Method and Tissue Storage for Visualization of Specific Antigens

During the initial screening, we identified GFAP, NeuN, CNPase, and CD45/LCA as candidate markers for visualization of the cell bodies of the four major classes of brain cells, noting that the staining for GFAP and CD45/LCA appeared to be relatively robust, whereas NeuN and CNPase were sensitive to formalin fixation. We applied these antibodies in a protocol involving HIER in TEG buffer and detection with the Envision+ peroxidase-labeled polymer detection system to paraffin sections from tissue array B (Table 2). The staining quality of individual samples was scored on a scale from 0 to 4 (Table 4; Figure 6) from five different slides and by two independent observers blinded to the arrangement of the tissue array. For each sample, the mean score and SEM was calculated, whereafter the data were stratified for fixation method (Figures 7 and 8), sex (♂: n=16; ♀: n=9), and PMI (PMI ≤ 24 hr, n=7; PMI of 25–48 hr, n=6; PMI ≥ 49 hr, n=12; Figure 7). Results were analyzed by the non-parametric Spearman r test.

NeuN staining was observed in only two of the specimens fixed by simple immersion in 4% Lillies PBFS (Table 2, Donors 1 and 7a), having mean scores of 1.7 and 2.6, respectively. All specimens fixed by immersion for >2.5 months were scored 0, making stratification for sex and PMI redundant (Figures 7A and 7B). Specimens that had been fixed by perfusion before immersion in 4% PFA for 2 weeks obtained mean scores >2 (two of five specimens: 2–2.4; three of five specimens: 3.8–4; Figure 8A). However, although one specimen maintained a score of 3, most specimens had dropped to a mean score of <1 (four of five specimens: 0–0.6) after 36 months of storage in 0.1% PFA at 4C (Figure 8A).

The mean scores of the CNPase stainings correlated negatively to the storage time in 4% Lillies PBFS at room temperature (Figures 7C and 7D) in both sexes (♂: r = −0.80, p<0.001; ♀: r = −0.91, p<0.01; Figure 7C) and for specimens with a PMI ≤ 24 hr and a PMI ≥ 49 hr (r = −0.89, p<0.001; Figure 7D). The analysis of specimens from the perfusion- and immersion-fixed brains showed that the staining for CNPase deteriorated with storage time (Figure 8B), yet not as pronounced as observed for specimens from non-perfused brain stored in 4% Lillies PBFS at room temperature (Figures 7C and 7D).

In case of GFAP, the mean scores for the GFAP stainings correlated negatively with storage time in 4% Lillies PBFS at room temperature in specimens from male donors (r = −0.66, p<0.01) but not female donors (r = −0.36; Figure 7E) and in specimens with a PMI < 24 hr (r = −0.89, p<0.05; Figure 7F). Unlike CNPase, the analysis of the specimens from the perfusion- and immersion-fixed brains showed that the GFAP staining was well preserved, even with long-term storage of the specimens in 0.1% PFA at 4C (Figure 8C).

The mean score of the staining result for CD45/LCA was negatively correlated to the time stored in 4% Lillies PBFS at room temperature (Figures 7E and 7F). As shown for CNPase, this was independent of the sex of the donor (♂: r = −0.60, p<0.05; ♀: r = −0.60, p<0.05; Figure 7G). Regarding PMI, the mean scores were only negatively correlated to time of fixation for specimens with a PMI from 25 to 48 hr (r = −0.93, p<0.05). As observed for the GFAP staining, the mean scores for the CD45 staining of specimens from the perfusion- and immersion-fixed brains showed that CD45 was well preserved in this material, even with long-term storage of the specimens in 0.1% PFA at 4C (Figure 8D).

In the case of GFAP and CD45, the staining sensitivity seemed to be influenced by the cellular activation state. In some samples, astrocytes and microglia showed an activated phenotype with hypertrophic cell bodies and hypertrophic or blunted processes throughout sections, making these cells very easy to identify. In other specimens, microglia expressed very low levels of CD45 and showed a resting phenotype, with thin angulated processes (Figures 5G and 5H). In such samples, there were occasionally small areas corresponding to the territory of one or two microglial cells with no staining, suggesting that CD45 might not be expressed by all resting-like microglia in human brain.

Immunohistochemistry in Tissue Stored in Fixative Solution

Most of the literature describing staining with the antibodies tested in this study are based on staining of material fixed in formaldehyde fixatives for hours or a few days, including studies of the expression of NeuN (Sarnat et al. 1998b), vimentin (Sarnat 1998a), MBP (Zecevic et al. 1998), O4, and NG-2 (Back et al. 2001). Even studies testing the effect of HIER on immunostaining focused on specimens fixed up to 1 week (Pileri et al. 1997; Nadji et al. 2005). Only a few studies have addressed the use of immunohistochemistry on tissue kept in formalin for >1 year (Evers and Uylings 1994,1997). Evers and Uylings (1994,1997) reported good results for detection of neuronal antigens MAP-2, phosphorylated and non-phosphorylated neurofilaments, calbindin, parvalbumin, calretinin, and neuropeptide Y, but also reported that it was impossible to obtain satisfactory staining results in tissue kept in fixative for ≥8 years. In the initial screening, staining results for most antigens were excellent when applied to specimens from brains stored in fixative for 24 hr (Table 5), yielding results comparable to literature reports. Exceptions to this were the antibodies against CD11b, NG-2, and O4 yielding no staining, neither in paraffin nor cryostat sections. The absence of signal could be caused by degradation of the antigens because both NG-2 and O4 have been detected in unfixed, frozen sections of human brain material sampled with only very short postmortem delay (Chang et al. 2000; Back et al. 2001). In this study, most tissue was collected with a longer postmortem delay, raising the possibility that these antigens might be particularly sensitive to postmortem autolysis. With longer time of fixation, staining results varied, raising a demand for a more rigorous test of the effect of fixation time on the staining result than could be achieved with the suboptimally matched tissue samples in tissue array A (Table 1).

Based on the initial screenings, we tested the effect of fixation time on NeuN, CNPase, GFAP, and CD45, applying the HIER/TEG protocol for epitope retrieval and the Dako Envision+ peroxidase-labeled polymer detection system to tissue array B, which was composed of 29 different specimens sampled in the postcentral gyrus from adult donor brains fixed for 8 days to 10 years. Although perfect matching of the donors was not possible, all donors were selected on the basis of their medical history and diagnoses, avoiding inclusion of donors with conditions that could severely impact the brain. Most of the donor material was fixed by simple immersion of the brain in 4% Lillies PBFS at room temperature, as is the routine at the Department of Pathology, OUH, and at the Brain Bank at Bispebjerg Hospital (Pakkenberg and Gundersen 1997; Samuelsen et al. 2003). In addition, we included five sets of specimens sampled from donor brains that had been fixed by perfusion before immersion-fixation for 2 weeks in 4% PFA and storage in 0.1% PFA solution at 4C for 2–3 years. For both types of fixation, the staining results were excellent when brains had been fixed for 2–3 weeks, yet the results pointed at better preservation of sensitive antigens like NeuN and CNPase in perfusion-fixed specimens. NeuN was very sensitive to fixation, with complete disappearance of immunohistochemical signal after 2–3 months of storage in 4% Lillies PBFS, and the staining signal also deteriorated with storage time in 0.1% PFA at 4C. For GFAP and CD45, excellent staining was obtained in several specimens from immersion-fixed brains stored in 4% Lillies PBFS for up to 4 months, yet preservation of these antigens was improved in perfusion-fixed material stored in 0.1% PFA at 4C, which was also pronounced for CNPase. This difference could be caused by several things. First, the perfusion with fixative resulted in a more even and quick fixation of the brain tissue, preventing degradation of tissue antigens. This observation is in line with recent literature suggesting the use of perfusion fixation for human brains to improve the neuropathological examination caused by better preservation of morphology and tissue antigens (Adickes et al. 1997; Sharma and Grieve 2006; Schmitt et al. 2007). Second, it might be better to store the already fixed tissue at 4C in a solution containing less formaldehyde than the usual conduct of storage. The recent consensus report from The Consortium of Brainnet Europe II set a focus on the importance the preparation, fixation, and storage of brain tissue collected by brain banks to better preserve the tissue molecules (Schmitt et al. 2007), and our data add to the discussion on how to fix and store brain tissue.

Selection of Candidate Markers for Future Cell Counting Studies

The histological material used in this study did not conform to the demands set up for unbiased stereology. First, the 5-μm paraffin sections were not thin enough to qualify for counting of neurons or glial cells using the physical disector (Sterio 1984). For this purpose, sets of serial sections cut at 1–2 μm should be used. Second, the 10-μm-thick cryostat sections did not qualify as thick sections for cell counting by the optical disector (Gundersen 1986, West et al. 1991). To use this tool for counting of large cells like neurons, the final section thickness should be at least 25 μm (Andersen and Gundersen 1999; Dorph-Petersen et al. 2001), and penetration of the staining through the section thickness should be validated by analysis of the z-axis distribution of labeled cells (Dorph-Petersen et al. 2001; Gardella et al. 2003). Recent studies in mice have combined the use of immunohistochemistry and stereology for estimation of total numbers of NeuN⁺ neurons (Lyck et al. 2007), CD11b⁺ microglia (Wirenfeldt et al. 2003,2007), and GABAergic neurons (Muller et al. 2001; Lifshitz et al. 2007) in specific brain regions by using the optical fractionator.

Based on our results, we suggest NeuN for identification of neurons in the human neocortex, because β-tubulin III–, MAP-2–, NF(PAN)-, and NSE-labeled epitopes are also present in neuronal cell processes. GFAP and S100β can be used to label astroglia, although both markers might be suboptimal. GFAP labeling is low in the protoplasmic astrocytes in the neocortex (Korzhevskii et al. 2005) and high in activated astrocytes (Schmidt-Kastner and Szymas 1990, Jessen 2006). It was recently shown that the binding of the polyclonal rabbit anti-GFAP antibody from Dako Cytomation is sensitive to phosphorylation (Tramontina et al. 2007). S100β has been reported to also label subpopulations of oligodendroglial cells (Rickmann and Wolff 1995; Tiu et al. 2000a,b). Only CNPase and p25α identified the cell bodies of oligodendroglia in adult brain. Nkx2.2⁺ and PDGFα-R⁺ oligodendroglial precursor cells were observed in only fetal and newborn brain. Because the anti-p25α-antibody was characterized as a marker for oligodendrocytes just recently (Otzen et al. 2005; Skjoerringe et al. 2006), more knowledge on its specificity is needed before applying this marker in quantitative studies, rendering CNPase the candidate for identification of oligodendrocytes. In case of microglia, staining for CD45/LCA provided better visualization of the cell body than staining for CD68(KP1) and labeled much more microglia than staining for HLA-DR. Microglial expression of hematogenous cell markers is known to depend on the state of activation (Mittelbronn et al. 2001; Perry 2003; Raivich and Banati 2004; Ladeby et al. 2005). We observed some variation in intensity of staining and morphology of microglia, with appearance of coarse, intensely stained profiles in samples from some brains. This raises the question if perimortal activation of microglia may occur, influencing the detection of these cells. Finally, monocytes and perivascular cells also express these markers (Sasaki et al. 1996; Mittelbronn et al. 2001; Ladeby et al. 2005), and the investigator should be able to distinguish between these cells and microglia during quantitative studies. A recent paper suggested the use of an antibody against IBA-1 for labeling of microglia in human brain (Ahmed et al. 2007). This was not tested in this study but could be another potential marker for cell counting purposes.

Although the optical disector has already proven its value for counting of immunohistochemically labeled cells, it may be advantageous to use the physical disector because of transparency of sections with intense specific labeling of the neuropil. The presented staining methodology could easily be implemented in sets of parallel 1- to 2-μm paraffin sections for counting using the physical disector, because the staining method is already optimized for staining of thin sections in large batches. For example, it may be possible to use the protocols for labeling of neurons expressing β-tubulin III, MAP-2, NF(PAN), or NSE in a physical disector design, just as counting of GFAP⁺ astrocytes should be done using the physical disector because of limited penetration of staining into thick sections (Lyck et al. 2006). Furthermore, in a recent stereological study of microglial responses to axonal lesion, it was noted that reactive microglia were present in the tissue as “clusters,” making it difficult to distinguish individual cells when using the optical disector (Wirenfeldt et al. 2007). Thus, it is necessary to test the immunohistochemical staining in conjunction to actual application to clarify whether the staining is reproducible between subjects and whether it allows identification of the individual cells in the specimen in a way that conforms with the rules for stereological counting.

In conclusion, based on a screening of a range of frequently used antibodies along with epitope retrieval techniques and detection systems, we identified a panel of candidate markers for future stereological studies. We also observed considerable loss of immunohistochemical staining signal in tissue specimens stored for long periods of time in 4% Lillies PBFS at room temperature, which has been widely used in tissue banks. For some markers, this loss could be delayed and possibly reduced in the long term by perfusion fixation before storage in 0.1% PFA at 4C. We believe that the application of immunohistochemistry in future stereological studies of the human brain will be possible with careful selection and validation of staining methods and by the use of well-preserved material.