Isolation of Living Neurons from Human Elderly Brains Using the Immunomagnetic Sorting DNA-Linker System

Abstract

Isolation and culture of mature neurons from affected brain regions during diseased states provide a well-suited in vitro model system to study age-related neurodegeneration under dynamic conditions at cellular levels. We have developed a novel technique to isolate living neurons from rapidly autopsied human elderly brains, and have succeeded in keeping them alive in vitro. Specifically, the parietal cortex blocks were fractionated by density gradients and further enriched for neurons by an immunomagnetic sorting DNA-linker technique. The postmortem interval averaged 2.6 hours. After isolation and purification of neurons using this technology, the cells were maintained in vitro for 2 weeks. Our evaluation revealed that 80% of the isolated cells were neurons and they exhibited neurotransmitter phenotypes (glutamate and γ-aminobutyric acid) as well as glutamate receptors. Studies on cell viability and calcium influx suggest that these isolated living cortical neurons still retain their typical neuronal functions. Our present study demonstrates that neurons isolated from human elderly brain autopsies can survive in vitro and maintain their functional properties. Our study has opened an opportunity to apply such neurons to dynamic pharmacological studies of neurological disorders at the single-cell level.

Isolation of aged neurons from elderly brain autopsy samples and their survival in vitro are clearly necessary, because we believe that the cells from the human species, an appropriate developmental or disease state, and the afflicted brain regions would provide a more suitable in vitro model system for studies of age-related neurodegeneration. However, little success has been made in human elderly brain tissues by using conventional approaches. Historically, isolation of neurons from human brain autopsies or aged rodent brains was pioneered by McKhann and colleagues, 1 Iqbal and Tellez-Nagel, 2 Farooq and colleagues, 3 and Brewer 4 although those in vitro cells were isolated either from frozen tissue or elderly rat brains. Thus, keeping human neurons from elderly brains alive in vitro was not attempted. We have developed a novel way to isolate and enrich neurons from rapidly autopsied human elderly brain samples, by using an immunomagnetic sorting technique following a conventional gradient isolation, and have succeeded in keeping these neurons alive in vitro and characterizing them in detail. Thus, this new approach has several advantages. First, this approach allows us to harvest neurons with an excellent purity that was previously limited by the single use of gradient isolation by which heterogeneous cell types were obtained. Second, it can be adapted to study molecular and cellular mechanisms of particular cell types for which phenotype-specific antigens are available. Third, it allows us to identify, study, and follow living neurons from human elderly healthy brains and diseased brains, such as Alzheimer’s disease, at the single-cell level.

Materials and Methods

Neuronal Isolation and Cultures

Brain tissue for culture was obtained from rapidly autopsied brains of geriatric patients (n = 30) enrolled in the Sun Health Research Institute brain bank. The average age of the patients was 78 years and the gender was well matched. The postmortem interval was 2.6 hours on the average. The parietal cortex blocks were cut into slices, digested with papain (Worthington, Lakewood, NJ) and overlaid onto a discontinuous density gradient of Optiprep (Nycomed, Olso, Norway) ranging from ∼1.006, 1.006 to 1.030, 1.030 to 1.037, 1.037 to 1.045, 1.045 to 1.100, and 1.100∼ g/ml (15 to 60%) (top-to-bottom), respectively. The lower one-third of phase 5 (phase 5c) (1.085 to 1.100 g/ml) contained the richest cells bearing the morphology consistent with neurons, 5 thus brain cells fractionated into phase 5c were processed for further purification of neurons, which was accomplished by incubating the cells with tetanus toxin C fragment (Boehringer, Indianapolis, IN) that binds to cell-surface gangliosides present in neurons. 6,7 The cells then reacted with an anti-TTC mouse IgG (Boehringer) were subsequently separated by addition of Dynabeads (Dynal, Oslo, Norway), which had been coated with recombinant streptavidin via DNA linker (Dynal) followed by conjugation with a biotinylated goat anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA). After selecting the bound cells in a Dynal magnetic particle concentrator and removal of the beads with DNase I, the enriched neuronal preparations were cultured in Neurobasal A with B27 supplement with anti-oxidants (Life Technologies, Inc., Grand Island, NY), containing 5 ng/ml of basic fibroblast growth factor, 50 ng/ml of nerve growth factor, 10 ng/ml of brain-derived neurotrophic factor, 10 ng/ml of epidermal growth factor (Life Technologies, Inc.) and 100 nmol/L of retinoic acid (Sigma Chemical Co., St. Louis, MO). Moreover, no antibiotics, glutamate, or aspartate were included in our neuron culture because they are toxic to the neurons.

Classical Cytochemical Analysis

The phase 5c cells 2 weeks after plating (2w-phase 5c cells) were processed for the classical cytochemical stains, including Nissl (1% cresyl violet), Oil Red O, and pyronin Y/methyl green. Methods for these commonly used stains are given in detail elsewhere. 8-10 When pyronin Y is used in combination with methyl green, pyronin Y stains RNA red, whereas methyl green stains DNA green in bright field. Pyronin Y preferentially binds to double-stranded RNA by intercalation, resulting in enhancement of red fluorescence. DNA binding of pyronin Y is blocked by the nonfluorescent dye, methyl green. 10,11

Immunocytochemistry

The following antibodies were used in this study: anti-neurofilament protein (NFP) SMI33 (1:2000, Sternberger, Lutherville, MD), anti-microtubule-associated protein-2(1:400, Boehringer), anti-neuron-specific enolase (1:3000, Polysciences, Warrington, PA), anti-neuron-specific class III β-tubulin (βIII) (1:2000, Berkley, Richmond, CA), anti-calbindin (1:1000, Swant, Bellinzona, Switzerland), anti-neuron cell surface antigen (NeuSu) (1:500; Chemicon, Temecula, CA), anti-tetanus toxin C fragment (1:500, Boehringer), anti-neuronal nuclei (NeuN) (1:50, Chemicon), and anti-RNA-binding neuronal protein called Hu (1:1000; gift of Dr. H. Furneaux, Sloan Kettering Institute, New York, NY). 12-16 Moreover, antibodies to glial fibrillary acidic protein (GFAP) (1:100; DAKO, Carpinteria, CA), NG2 chondroitin sulfate proteoglycan 17 (1:1000, Chemicon), LN-3 (1:200, ICN, Costa Mesa, CA), von Willebrand factor (vWF) (1:50, DAKO), fibronectin (1:1000, Sigma), and galactocerebroside (GalC) (1:100, Boehringer) were used for nonneuronal identification. An antibody that detects central nervous system stem cells, anti-Musashi (1:1000; gift of Dr. H. Okano, Keio University, Japan) 18-25 was also used. Antibodies against neurotransmitter-synthesizing enzymes and neurotransmitter receptors used were anti-phosphate-activated glutaminase (PAG) (1:1000; gift of Dr. T. Kaneko, Kyoto University, Japan), 26-28 glutamate decarboxylase (GAD) (1:2000, Chemicon), choline acetyltransferase (ChAT) (1:1000, Chemicon), N-methy-d-aspartate receptor type 1 (NMDAR1) (1:1000; PharMingen, La Jolla, CA), and α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA)-type glutamate receptor types 2 and 4 (GluR2/4) (1:1000, PharMingen).

The 2w-phase 5c cells were fixed with 4% paraformaldehyde followed by fixation with methanol and incubated with a primary antibody followed by a biotin-conjugated secondary antibody (Vector Laboratories, Inc., Burlingame, CA; Jackson ImmunoResearch). Vectastain Elite kits (Vector Laboratories, Inc.), 3,3′-diaminobenzidine (DAB), or Nova RED (Vector Laboratories, Inc.), and Cy3- or Oregon Green 514-strepatavidin (Molecular Probes, Eugene, OR) were used for immunoenzymatic and fluorescence detection. Sudan Black B (1%) in 70% ethanol was used to quench autofluorescence. 29 No nonspecific binding occurred when the antibodies that had been obtained from private individuals were preabsorbed to the respective recombinant antigens, which is consist with previous publications. 13-16,19-25,28 Deletion and replacement of primary antibodies with nonimmune sera were techniques used as controls throughout to certify the specific binding of commercially available antibodies. The optimal dilution ratio for each primary antibody was determined using adequate positive and negative control cells. Amplification of the immunolabeling using the Tyramide Signal Amplification kit (NEN, Boston, MA) was applied in parallel for detecting neurotransmitters and their receptors.

Western Blotting

Cells were lysed with 1% Triton X-100 and 0.1% sodium dodecyl sulfate. Then 20 μg of cell lysate protein was electrophoresed on a 10% polyacrylamide-sodium dodecyl sulfate gel, and transferred onto a polyvinylidene difluoride membrane (Millipore, Bedford, MA). The membrane was blocked in 5% nonfat milk, reacted with the anti-βIII or GFAP antibody, and then with a horseradish peroxidase-conjugated secondary antibody (Boehringer).

Cell Viability Tests

Four different assays of cell viability were performed for the 2w-phase 5c cells. For the trypan blue exclusion test, cells were incubated for 5 minutes with 0.04% trypan blue (Sigma). 30,31 Methods for the other assays are given in detail elsewhere. 32-35 LIVE/DEAD Viability/Cytotoxicity kits (Molecular Probes) used a 30-minute incubation of the cells with 4 μmol/L of acetoxymethyl (AM) ester of calcein (calcein AM) and 6 μmol/L of ethidium homodimer (EthD-1). LIVE/DEAD Reduced Biohazard Viability/Cytotoxicity kits (Molecular Probes) used a 15-minute incubation of the cells with the mixture of SYTO 10 and DEAD Red recommended by the manufacture. The cells were cultured with addition of 0.5 mg/ml of tetrazolium salt such as 3-(4,5-dimethyl thiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) (Molecular Probes) at 37°C, with micrographs taken 2 and 4 hours after the addition. 35

Calcium Imaging

The 2w-phase 5c cells were loaded with 10 μmol/L of fluo-3 AM (Molecular Probes) in low potassium buffer 36-38 (in mmol/L: NaCl, 137; NaHCO₃, 4.2; Na₂HPO₄, 0.34; KCl, 5.4; KH₂PO₄, 0.44; MgSO₄, 0.81; CaCl₂, 1.26; d-glucose, 5.55; HEPES, 20; pH 7.4) for 1 hour at 37°C in a CO₂ incubator, then processed for calcium optical recordings. Calcium influx was induced by exposing the cells to a 50 or 100 mmol/L of potassium buffer identical to the low potassium buffer, except that 92.4 or 42.4 mmol/L of NaCl and 50 or 100 mmol/L of KCl, respectively, were used. Calcium influx was also examined by using glutamate at concentrations ranging from 1 to 500 μmol/L. The fluo-3 fluorescence in the loaded cells was measured with Fluoview software (Olympus) and a confocal laser-scanning microscope system mounted to an inverted microscope (Olympus). Excitation was provided by the 488-nm line of an argon laser (NEC), and emission was collected at a wavelength >505 nm. Images were obtained every 2 seconds, and the significant changes in intracellular intensity were confirmed by recording background fluorescence intensity in parallel. Other buffers used were: calcium-free low potassium buffer 36-38 (in mmol/L: NaCl, 137; KCl, 5.4; MgCl₂, 2.07; EGTA, 5.0; d-glucose, 5.55; HEPES, 20; pH 7.4); manganese (or cobalt)-containing low potassium buffer 37,38 (in mmol/L: NaCl, 137; KCl, 5.4; MgCl₂, 2.07; MnCl₂, or CoCl₂, 4.0; d-glucose, 5.55; HEPES, 20; pH 7.4).

Results

Neuronal Characterizations

Our gradient isolation technique yielded six phases from the brain tissues. With phase microscopy, the lower one-third of phase 5 (phase 5c) most abundantly contained cells bearing the morphology consistent with neurons, 5 along with much debris (Figure 1a) ▶ . The neuron-like cells accounted for 62.1 ± 3.6% (n = 6) of the phase 5c cellular components, when observed after 2 weeks in vitro. After further purification with the immunomagnetic sorting technique, in which magnetic beads coated with streptavidin via DNA linker were used (Figure 1, b and c) ▶ , observation with phase microscopy after 2 weeks in vitro revealed that 86.9 ± 3.3% (n = 6) of the phase 5c cells were neuron-like cells (Figure 1d) ▶ . The immunomagnetic purification from the phase 5c yielded approximately 1 × 10 6 cells/g tissue weight. These cells were multi-polar to mono-polar, fusiform, or triangular, and bore cell processes (Figure 1; e, f, and h) ▶ . A relatively large nucleus and nucleolus were conspicuous (Figure 1g) ▶ .

Figure 1.

Immunomagnetic sorting and phase microscopic morphology. a: Cells from phase 5c after the gradient isolation were maintained in vitro for 2 weeks. This population contained much cell debris. Arrows indicate the cells showing a bright cytoplasm. b and c: The cells were further purified by the immunomagnetic sorting technique. Magnetic beads coated with streptavidin via DNA linker were used. The phase 5c cells were incubated with tetanus toxin C fragment (TTC) followed by exposure to a mouse anti-TTC IgG, to which then the magnetic beads complexed with an anti-mouse IgG were reacted. The beads shown readily attached to the cells. d: The purified cells using the magnetic beads were maintained in vitro for 2 weeks. A relatively pure neuronal-cell population is observed. e, f, and h: Neuron-like cells obtained through the above-mentioned procedures were maintained in vitro for 2 weeks. The vast majority of such phase 5c cells exhibit the morphological characteristics of neurons. Brown lipofuscin granules, a typical hallmark for aged neurons, are seen. g: In such a neuron-like cell, the nucleus and nucleolus are conspicuous (arrow). Scale bars: 100 μm (a); 50 μm (b, c, e, f, g, and h); 75 μm (d).

To characterize these neuron-like cells, classical cytochemical techniques were used: Nissl, Oil Red O, and pyronin Y stains. With Nissl staining, the nuclei were relatively unstained but the nucleoli were stained blue or purple (Figure 2, a and b) ▶ . The granular materials that accumulated in the cytoplasm were usually stained blue or greenish-blue (Figure 2b) ▶ . With Oil Red O or pyronin Y staining, these granules were stained red under bright-field viewing (Figure 2, c1 and d) ▶ . With pyronin Y combined with methyl green, these granules produced a relatively low red fluorescence (Figure 2,c2) ▶ , suggesting that these granules are lipofuscin, which has been reported to include lipids, 39 but not Nissl bodies. Nissl staining evaluations suggested that 81.5 ± 6.1% (n = 6) of the 2w-phase 5c cells were neurons.

Figure 2.

Classical cytochemistry. The phase 5c cells isolated through the immunomagnetic sorting were maintained for 2 weeks in vitro. a: Nissl-1. The nuclei of the phase 5c neuron-like cells are weakly stained at best (arrows). Lipofuscin granules are intensely stained blue or purple. An arrowhead indicates nonneuronal cells. b: Nissl-2. The lipofuscin granules are stained greenish-blue, a somewhat different color, compared to true Nissl bodies. The light blue staining is observed around the nucleus and in the nucleolus (arrow). c: Pyronin Y/methyl green. Under bright-field optics (c1), the lipofuscin granules are stained red. Under fluorescence optics (c2), however, these granules show relatively low red fluorescence, and by contrast the cytoplasm exhibit intense red fluorescence (white arrow). d: Oil Red O. The lipofuscin granules are labeled red. Scale bars: 50 μm (a, c1, c2); 25 μm (b); 100 μm (d).

To further ensure that these cells were neurons, we characterized them with immunocytochemistry. For neuron-specific cytoplasmic markers, antibodies against NFP (SMI33), microtubule-associated protein-2, neuron-specific enolase, βIII, and calbindin were used. The representative staining for NFP(SWI33) or βIII is shown in Figure 3A, a and b ▶ series. For a neuron-specific cell surface marker, antibody for NeuSu was used (Figure 3A, c) ▶ . Lastly, we used antibodies against neuron-specific nuclear protein, NeuN and Hu 12 on these neurons. Figure 3A, d2 ▶ , demonstrates a representative immunofluorescent staining of NeuN. The fluorescent staining shown was not quenched by using Sudan Black B 29 for exclusion of overlapping by autofluorescence. Thus, the immunofluorescence observed with the neuron staining is specific. Quantitative evaluations on the 2w-phase 5c cells revealed that 60 to 80% (n = 10) of the total cells were immunoreactive to these neuronal markers (Table 1) ▶ ▶ . Meanwhile, quantitative analysis of percentages of cells showing immunoreactivity for nonneuronal antigens such as GFAP, LN-3, vWF, NG2 17 in the total cell population (Figure 3B, a to d) ▶ from the 2w-phase 5c cells in Table 1 ▶ ▶ is consistent with the above findings whereas positive control cells, including astrocytes, microglial cells, endothelial cells, and progenitor cells show strong immunoreactive response, respectively (Figure 3B; e to h) ▶ . We found that our phase 5c neurons lacked immunoreactivity for the neural RNA-binding protein Musashi 18 (Figure 3B, d) ▶ , excluding the possibility that the neurons isolated were differentiated from in vitro replication of stem cells. Western blot analysis further demonstrated that a single immunoreactive band for the neuron marker, β III at 50 kd was identified from our phase 5c neuron samples from autopsies of human elderly patients whereas this band was not observed for microglia or astrocyte samples, which were prepared in our facility 40 (Figure 3C) ▶ . By contrast, an immunoreactive band for the astrocyte marker GFAP at the appropriate molecular weight of 50 kd was observed in the astrocyte samples but not in the microglia or phase 5c cells (Figure 3C) ▶ .

Figure 3.

Immunocytochemistry of neuronal and nonneuronal markers. A: The phase 5c cells after 2 weeks in vitro were characterized for neuron-specific markers. Most of the phase 5c cells were consistently labeled with antibodies directed against the neuronal antigens. Neuron-specific cytoplasmic markers used were: neurofilament protein (NFP, SMI33) and neuron-specific class III β-tubulin (βIII). NFP was immunoreactive to the neurons, shown with DAB staining (a1) and also with Cy3 immunofluorescence (a3), a2 is phase contrast of a3. Arrows in a1 indicate unstained cells. Similarly, the neurons also show βIII-immunoreactive staining using DAB (b1) and Cy3 immunofluorescent methods (b3). b2 is phase contrast of b3. Neuron-specific cell surface marker is NeuSu. c1 is the same field visualized with phase contrast and c2 is NeuSu immunofluorescence using Oregon Green marker. Arrows in c1 indicate unstained cells and a white arrowhead in c2 indicates lipofuscin-derived cytoplasmic autofluorescence. Neuron-specific nuclear marker, neuronal nuclei (NeuN) shows labeled neuronal nuclei. d1 is the same field visualized with phase contrast and d2 is NeuN immunofluorescence using Oregon Red marker. Arrows in d1 also indicate unstained cells. B: The phase 5c cells after 2 weeks in vitro were characterized for nonneuronal markers. Less than 10% of the total phase 5c cell population was immunoreactive to nonneuronal markers tested. The neurons show negative DAB immunoreactivity to antibodies of GFAP (a), LN-3 (b), von Willebrand factor (vWF) (c), and Musashi (d) with a DAB marker. Arrows in a to c indicate occasionally stained cells. In parallel, each positive control cell type is demonstrated: primary astrocyte (e), 40 primary microglia (f), 40 isolated capillary (g), and NG108 neuroblastoma cell (h). C: Western blot analysis for the neuronal marker βIII (a) and the astrocytic marker GFAP (b) was performed using protein extracts obtained from primary microglia (Mic) and astrocyte (Ast) cultures, and isolated phase 5c cells (Neu) 2 weeks after plating from rapid autopsies of human elderly brains. Arrows indicate specific immunoreactive bands for βIII (a) and GFAP (b). Arrowheads illustrate bands for β-actin. Scale bars: [50 μm (A:a1, a2, a3, b1, b2, b3; B d, f, g, and h); 100 μm (A:c1, c2, d1, d2; B: a, b, c, and e).

Table 1A.

Percentages^* of the Phase 5c Cells Labeled with Neuron-Specific Markers

Markers (cytoplasmic)	Positive cells (%)	Markers (cell surface)	Positive cells (%)	Markers (nucleus)	Positive cells (%)
NFP	78.8 ± 5.9†	NeuSu	63.2 ± 8.1	NeuN	69.3 ± 12.3
MAP-2	68.9 ± 1.0	TTC	86.8 ± 9.0	Hu	80.6 ± 12.1
NSE	94.0 ± 7.8
βIII	69.4 ± 1.1
Calbindin	51.8 ± 3.1

NFP, neurofilament protein (SMI33); MAP-2, microtubule-associated protein-2; NSE, neuron specific enolase; βIII, neuron-specific class III β-tubulin; NeuSu, neuron cell surface antigen; TTC, tetanus toxin C fragment; NeuN, neuronal nuclei.

*Each percentage value was calculated by using 10 of × 400 photomicrograms obtained from each of 10 autopsy cases.

^†Mean ± SD.

Table 1B.

Percentages^* of the Phase 5c Cells Labeled with Nonneuron-Specific Markers

Markers	Positive cells (%)	Cells detected
GFAP	9.4 ± 7.3†	Astrocytes
LN-3	8.9 ± 2.5	Microglia/macrophages
vWF	9.5 ± 7.8	Endothelial cells
NG2	3.7 ± 3.6	O-2A progenitors, type 2 astrocytes
Fibronectin	6.8 ± 1.0	Fibroblasts

GFAP, glial fibrillary acidic protein; vWF, von Willebrand factor; O-2A, oligodendrocyte-type 2 astrocyte.

*Each percentage value was calculated by using 10 of ×400 photomicrograms obtained from each of 10 autopsy cases.

^†Mean ± SD.

Neurotransmitter Phenotypes

To examine what types of neurotransmitters these neurons contain, the neurons were evaluated for synthesizing enzymes for two typical cortical neurotransmitters, glutamate and γ-aminobutyric acid (GABA). We found that the antibody against the glutamate-synthesizing enzyme PAG 27 was immunoreactive to presumed glutamatergic neurons after the autofluorescence was quenched by Sudan Black B (Figure 4A, a2) ▶ . Meanwhile, PAG-transfected cells were used as a positive control (Figure 4A, a3) ▶ . Thus, we believe that the immunofluorescent staining was specific. By using the similar approach, we found that the antibody against GAD (67 kd) immunoreactively labeled presumed GABAergic neurons 41 (Figure 4A, b2) ▶ . Lastly, we used an antibody to the acetylcholine-synthesizing enzyme ChAT to label presumed cholinergic neurons 42 and found little immunoreactive staining on these neurons (Figure 4A, c2) ▶ . To quantitatively examine proportions of these neurons we isolated, we counted each type of neurotransmitter immunoreactive neurons. We found that PAG-positive neurons comprised 55.3 ± 18.5% (n = 6) of the 2w-phase 5c neuronal population, a low percentage compared to the estimated 70 to 80% prevalence of glutamatergic cells in the intact neocortex 43 (Table 2) ▶ . GAD-immunoreactive cells comprised 52.3 ± 12.3% (n = 6) of the 2w-phase 5c neuronal population, a percentage somewhat higher than that observed in the human cortex in situ 44,45 (Table 2) ▶ . These results are consistent with a relative resistance of GABAergic neurons to ischemia and a heightened sensitivity of glutamatergic neurons to ischemia, as previously reported. 46-48 Immunoreactivity for ChAT was evident in only 9.5 ± 1.9% (n = 6) of the 2w-phase 5c neuronal population. A similarly low percentage of ChAT-positive neurons have been observed in sections from the human adult neocortex 49 (Table 2) ▶ . Many neurons in the cerebral cortex receive glutamatergic input. 50 We examined whether our neurons expressed glutamate-related receptors by using antibodies directed against NMDAR1 and GluR2/4. Immunoreactivity for the NMDAR1 and GluR2/4 was present in 74.2 ± 10.2% and 77.1 ± 6.4% (n = 6) of the 2w-phase 5c neurons, respectively. This immunoreactivity was typically localized to the cell membrane, a result more clearly visualized after Sudan Black B treatment to reduce lipofuscin autofluorescence 29 (Figure 4B, a2 and b2) ▶ .

Figure 4.

Neurotransmitter and receptor phenotypes. A: The phase 5c cells after 2 weeks in vitro were evaluated for phenotypes of neurotransmitters by immunocytochemistry enhanced with the tyramide signal amplification technique. a: PAG for glutamatergic neurons; b: GAD for GABAergic neurons; and c: ChAT for cholinergic neurons. The first (a1, b1, and c1) and second (a2, b2, and c2) columns in each series are the same field visualized with phase contrast and fluorescence using Oregon Green marker, respectively. Specific fluorescence in the cytoplasm was observed after lipofuscin autofluorescence was quenched by a Sudan Black B treatment. Arrows and arrowheads in the first column indicate significantly stained and unstained cells, respectively. The third column demonstrates specific immunofluorescent staining in each corresponding transfected cells as positive controls: a3, transiently transfected PAG HEK293 cells; b3, transiently transfected GAD HEK293 cells; c3, transiently transfected ChAT HEK293 cells. B: a, NMDA receptor type 1 (NMDAR-1); b, AMPA-type glutamate receptor types 2/4 (GluR²4). The first (a1 and b1) and second (a2 and b2) columns in each series are the same field visualized with phase contrast and fluorescence using Oregon Green marker, respectively. Specific fluorescence in the cytoplasmic membrane was obtained after lipofuscin autofluorescence was quenched by a Sudan Black B treatment. Arrows and arrowheads in a1 and b1 indicate significantly stained and unstained cells, respectively. The third column pictures are transiently transfected NMDR1 HEK293 cells (a3) and transiently transfected GluR²4 HEK293 cells (b3) as positive controls. Scale bars: 50 μm (A:a1 to a3; B: a1 to d3, b1 to b3); 100 μm (A:b1 to b3, c1 to c3).

Table 2.

Neurotransmitter Phenotypes of Neurons in the Phase 5c

Markers	Neurotransmitters	Positive cells (%)*	Percentage in the intact cortex‡ (%)
PAG	Glutamate	55.3 ± 18.5†	70–80
GAD	GABA	53.2 ± 12.3	30–40
ChAT	ACh	9.5 ± 1.9	∼10

PAG, phosphate-activated glutaminase; GAD, glutamate decarboxylase; ChAT, choline acetyltransferase; ACh, acetylcholine.

*Each percentage value was calculated by using 10 of ×400 photomicrograms obtained from each of 6 autopsy cases.

^†Mean ± SD.

^‡Reported by immunohistochemistry in humans or primates.

Biochemical Evaluation of Neuronal Viability

To evaluate whether these neurons were alive, four different assays of cell viability were performed after 2 weeks in vitro. With the trypan blue exclusion test, the 2w-phase 5c cells with neuronal morphology ranged from having a clear cytoplasm to having a pale blue cytoplasm, with only occasionally an intense blue staining (Figure 5a) ▶ . Trypan blue can passively be taken by injured cells through their permeable membrane or even enter viable cells with some injury by trypsinization. 30,31 Thus, these isolated neurons were further examined by using more rigorous cell viability techniques. With calcein AM plus EthD-1 (Figure 5b) ▶ , 80.8 ± 10.5% (n = 10) of the 2w-phase 5c neurons significantly showed green cytoplasmic fluorescence, indicative of viability, 32,33 and 6.7 ± 1.6% (n = 10) exhibited red nuclear fluorescence, indicating cell death 32,33 (Table 3) ▶ . To further confirm these findings, we conducted this assay for isolated dead neurons from frozen tissue of the same brain region. We found that all cells demonstrated red nuclear fluorescence (Figure 5e) ▶ .

Figure 5.

Biochemical evaluation of cell viability. The phase 5c cells after 2 weeks in vitro were assayed for their viability. a: Trypan blue (0.04%). Trypan blue incorporation into our neurons ranges from a clear cytoplasm (black arrows) to a pale blue cytoplasm (white arrows), with occasionally intense blue staining (arrowhead). Cell viability was further confirmed by the following rigorous cell viability: b: Calcein AM/ethidium homodimer (EthD-1) staining. Green cytoplasmic fluorescence and red nuclear fluorescence are observed. Arrows indicate yellowish autofluorescence that are easily identified because of distinct wavelengths. c: SYTO 10/DEAD Red staining. White arrows and white arrowheads (c1) indicate unlabeled cells with both dyes and red-labeled nuclei with indication of being dead, respectively. Orange arrowheads indicate viable cells (stained with SYTO10 in the presence of DEAD Red). Differences are observed between the nuclear labeling when the cells were incubated for 15 minutes (c2) and 1 hour (c3). A yellow arrowhead in c2 indicates a labeled nucleus, and an arrow and an arrowhead in c3 indicates an unlabeled cell and a labeled nucleus, respectively. Meanwhile, dead neurons isolated from frozen tissues stained with Calcein AM/EthD-1 and SYTO 10/DEAD Red are demonstrated in e and f, serving as control groups. To further rule out a possibility of autofluorescence, the tetrazolium salt MTT was used on these neurons. d: The exposure of MTT to living neurons produces purple formazans. Dark blue formazans are shown within living cells as a granular duster in approximately 80% of the neurons (arrows), whereas no formazans are formed in nonstained cells (arrowheads), indicating cell death. Scale bars: 100 μm (a, b, c1, e, f); 75 μm (d); 50 μm (c2, c3).

Table 3.

Percentages^* of the Living and Dead Neurons Estimated by Different Cell Viability Tests

Live markers	Living cells (%)	Dead markers	Dead cells (%)
Calcein AM	80.8 ± 10.5†	Ethidium homodimer	6.7 ± 1.6
SYTO 10	82.1 ± 5.9	DEAD Red	10.1 ± 8.3
MTT	77.0 ± 8.7

MTT, 3-(4,5-dimethyl thiazol-2-yl)-2,5-diphenyl tetrazolium bromide.

*Each percentage value was calculated by using 10 of ×400 photomicrograms obtained from each of 6 or 10 autopsy cases.

^†Mean ± SD.

Next, we used the mixture of SYTO 10 and DEAD Red, by which living cells appear fluorescent green and dead cells appear fluorescent red. 32,33 After a 15-minute incubation, 69.9 ± 9.2% (n = 6) of the neurons exhibited fluorescent green nuclei, and 8.3 ± 7.3% (n = 6) showed red nuclear fluorescence (Figure 5c1) ▶ . When the incubation time was 1 hour, 82.1 ± 5.9% (n = 6) fluoresced green and 10.1 ± 8.3% (n = 6) exhibited red or yellowish orange fluorescent nuclei (Figure 5, c2 and c3) ▶ (Table 3) ▶ . Such a yellowish signal may mean that both SYTO 10 and DEAD Red have bound nucleic acids. 34 To further verify these results, the dead neurons isolated from the frozen tissue of the same brain region were used as control cells, showing that all of the dead neurons exhibited red nuclei (Figure 5f) ▶ .

Lastly, besides using Sudan Black B to avoid any possibility of lipofusin autofluorescence, we also used tetrazolium salts such as MTT, which are evaluated in bright field but not by fluorescence. Specifically, we examined mitochondrial function with six different formazans. 51 Tetrazolium salts, including MTT, form purple or orange formazans after exposure to the tetrazolium salts in living but not dead cells. After incubation of the 2w-phase 5c cells with MTT (Figure 5d) ▶ , fine granular purple formazans were recognized in 77.0 ± 8.7% (n = 6) of the 2w-phase 5c neuron-like cells (Table 3) ▶ . The formazans exhibited a granular appearance, as in a previous report. 35

Calcium Imaging Studies

To examine whether these neurons retain a typical neuronal function, calcium influx in the neurons was monitored using the calcium-imaging technique. When neurons from individuals (<60 years old) were depolarized by applying 50 to 100 mmol/L of potassium chloride (high KCl), we found that 78.3 ± 14.4% (n = 6) of the neurons exhibited the expected rapid increase in intracellular fluo-3-derived fluorescence intensity 2 to 5 seconds after the high KCl application (Figure 6a) ▶ . The KCl-induced fluorescence increments were quenched if calcium in the buffer was subsequently replaced with manganese (Figure 6b) ▶ or cobalt (not shown), and no increments occurred if calcium-free buffer was used initially (Figure 6c) ▶ . On the other hand, the percentage of neurons exhibiting KCl-induced fluorescence increases was lower (42.5 ± 25.6%, n = 6) if the neurons were isolated from relatively older (>70 years old) individuals. Furthermore, as expected in that more than 70% of our 2w-phase 5c neurons express glutamate (NMDA and AMPA) receptors, an increased calcium influx was observed when they were loaded with fluo-3 AM followed by treatment with 100 to 500 μmol/L of glutamate (Figure 6d) ▶ . They exhibited the expected rapid increase in intracellular fluo-3-derived fluorescent intensity 5 to 10 seconds after the glutamate treatment. The result suggests that glutamate can induce a typical neuronal increase of calcium influx. From the above experiments on calcium influx, no calcium influx was observed in the dead neurons isolated from frozen brain tissue (Figure 6, b4 to b6, c3 to c4, d3 to d4) ▶ .

Figure 6.

Calcium influx in human aged cortical neurons. The phase 5c cells after 2 weeks in vitro were loaded with fluo-3 AM and depolarized by 50- to 100-mmol/L KCl (high KCl) or 100- to 500-μmol/L glutamate. a: Pseudocolor-enhanced photomicrograms. Pictures are the same field before (a1) and 2 seconds after (a2) high KCl application. In the color scale bars, upper colors show higher fluorescence intensity. High KCl-induced fluorescence increments are quenched when calcium in the buffer is subsequently replaced with manganese (b1, before; b2, 5 seconds after; b3, 10 seconds after replacement). The same treatments of high KCl and replacement of manganese were applied on dead neurons isolated from frozen tissue of the same brain region and no calcium influx was observed (b4 to b6). Furthermore, neurons exhibit no high KCl-induced increase in intracellular fluorescence intensity when calcium is initially removed (c1, before; c2, 10 seconds after high KCl application) whereas no calcium influx was observed with the same treatment in the dead neurons isolated from frozen brain tissue (c3 and c4). b1 to b6 and c1 to c4 were obtained at the different voltages for the 488-nm line of an argon laser. Furthermore, 500 μmol/L of glutamate induces calcium influx (d1, before; d2, 10 seconds after the treatments) whereas no calcium influx was observed with the same treatment in the dead neurons isolated from frozen brain tissue (d3 and d4). Scale bars, 75 μm.

Discussion

To our knowledge, this is the first report of a new strategy for the identification, isolation, and enrichment of primary neurons obtained from rapidly autopsied human elderly brains using the immunomagnetic sorting technique bonded with the neuronal TTC. By providing an approach to identifying neurons while they are alive, even if they are aged, the use of immunomagnetic sorting technology with cell surface-specific markers such as TTC will allow the specification of phenotype to be studied on a single-cell level in real time. Furthermore, when used in conjunction with other antigens, this strategy might permit the enrichment of any neurotransmitter phenotypes or cell types with the phenotype-specific markers.

Besides our novel technique providing a unique opportunity, several other factors also contributed to this successful study. First, we believe that a short postmoterm period (<3 hours) is extremely important for neuronal survival for the reason that a low yield or failure of neuronal survival has been obtained from autopsies with postmortem periods >6 hours. We also believe that our optimal combination of neurotrophic factors and anti-oxidants has played a critical role in the survival of these aged neurons.

Neurons with external applications of 50 to 100 mmol/L of KCl or 100 to 500 μmol/L of glutamate increased Ca²⁺ influx. Because ion concentrations in the bathing buffer for our neurons were adjusted by decreasing sodium ion concentration as the KCl concentration was increased, thus, these neurons still existed under the same osmolarity of the bathing solution, and behaviors of these neurons in this culture are not expected to be different. 36-38 Moreover, it is possible that some of these isolated neurons may be degenerating neurons that have weak calcium signaling because they are from aged or diseased brains. 52,53 Even if unusual Ca²⁺ events may be involved, our report is the first to demonstrate KCl or glutamate-induced calcium influx in human aged neurons, indicating the possibility of further pharmacological studies.

Lastly, it should be emphasized that we understand that the neurons are not perfectly healthy and that they are isolated from autopsied brains of aged or sick patients. It is quite labor intensive to isolate and maintain aged or sick neurons in vitro. Nevertheless, RNA and neuronal proteins from our isolated living neurons are preserved, and enzymes and membranes are still functional, indicating that our study has provided the future possibility to apply such human aged neurons to dynamic pharmacological studies for age-related neurological disorders at the single-cell level.