Injury Response of Resected Human Brain Tissue In Vitro

Abstract

Brain injury affects a significant number of people each year. Organotypic cultures from resected normal neocortical tissue provide unique opportunities to study the cellular and neuropathological consequences of severe injury of adult human brain tissue in vitro. The in vitro injuries caused by resection (interruption of the circulation) and aggravated by the preparation of slices (severed neuronal and glial processes and blood vessels) reflect the reaction of human brain tissue to severe injury. We investigated this process using immunocytochemical markers, reverse transcriptase quantitative polymerase chain reaction and Western blot analysis. Essential features were rapid shrinkage of neurons, loss of neuronal marker expression and proliferation of reactive cells that expressed Nestin and Vimentin. Also, microglia generally responded strongly, whereas the response of glial fibrillary acidic protein‐positive astrocytes appeared to be more variable. Importantly, some reactive cells also expressed both microglia and astrocytic markers, thus confounding their origin. Comparison with post‐mortem human brain tissue obtained at rapid autopsies suggested that the reactive process is not a consequence of epilepsy.

Introduction

Despite the huge impact brain injuries have on society 41, until now, the cellular processes and neuropathological alterations involved have remained elusive. Traditionally, the effects of damage in the human brain have been investigated using fixed tissue obtained at autopsy after different survival times following injury 10, 12, 37, or in vivo imaging techniques 17. However, tissue slice cultures of the adult human brain provide a novel alternative to study the properties of human central nervous system (CNS) cells in relation to ageing, disease or injury. In previous research, we have shown that cells in post‐mortem human brain tissue slices can stay alive for long periods in vitro and respond to experimental manipulation 38, 43. In epilepsy surgery, “normal” (ie, not directly adjacent to the epileptic focus and no recording of spontaneous epileptic activity by the electrode arrays placed on the cortical surface prior to resection) brain tissue frequently needs to be removed in order to reach the epileptic focus. This tissue may have been affected by epileptic events but is still relatively healthy at resection and many acute (ie, within 24 h) electrophysiological studies have successfully used such samples 2, 35. The presence of many healthy cells meant that damage inflicted by processing for culture naturally elicited an injury response. Here, we describe for the first time how reactive processes and their neuropathological implications develop in injured human brain tissue in vitro. Using immunocytochemistry, quantitative polymerase chain reaction (PCR) and Western blots we observed rapid degeneration of neurons. Concurrently, large proliferating cells expressing markers of glia activation, Vimentin and Nestin, emerged along blood vessels during the first week in vitro. These large reactive cells were not present in the tissue before it was cultured, although we did find a variable degree of activation among resident astrocytes and microglia cells. Cell division was found to play a key role in the emergence of reactive cells. Furthermore, we present evidence that the reactive cells expressed both astrocytic and microglia markers. The reactive process is not typical of epilepsy as reactive cells may also occur, albeit very infrequently, in post‐mortem tissue from control subjects, who died without a neurological disease. As far as we know, the collective cellular changes following severe damage of in vitro adult human brain tissue have not been described before.

Materials and Methods

Brain samples

The data presented in this study are based on brain specimens from 76 patients with temporal lobe epilepsy (Supporting Information Table S1). Only in very few cases (not mentioned in Supporting Information Table S1) was the quality or the amount of tissue deemed insufficient to start culture experiments. Resected brain tissue from the normal temporal neocortex was obtained during surgery performed to treat medication‐refractory epilepsy. This tissue was not needed for diagnostic purposes and would otherwise have been destroyed. During the operation, a thin, flexible flat electrode was gently placed on the pial surface to record spontaneous seizure activity or to distinguished areas involved in language processing. However, this procedure never caused any invasive damage such as observed with depth electrodes or subdural grids 13. The history of the tissue before culture is illustrated in Supporting Information Figure S1A. Please note that adjacent tissue samples from many of the epilepsy patients described in this study have been used for acute electrophysiological recordings 35. Benchmark for the resected tissue at the start of the culturing experiments was noncultured tissue specimens from 28 deceased control subjects without a history of epilepsy or other neurological disease (Supporting Information Table S2). This tissue was obtained within the framework of another experimental paradigm. It was acquired at rapid autopsies performed by the Netherlands Brain Bank and its history is illustrated in Supporting Information Figure S1B. The post‐mortem delay (PMD) of the majority of the control subjects was less than 7 h and exceeded 10 h in only one instance. Cultured slices from one of these subjects (C20) were used to illustrate the emergence of reactive cells in post‐mortem brain tissue. Permission for the use of human brain tissue for in vitro research was granted by the Ethical Medical Committee of the VU University Medical Center (VUmc) in Amsterdam, where the operations and autopsies took place. Brain tissue specimens were obtained with written informed consent and patient information was treated in accordance with the principles expressed in the Declaration of Helsinki.

Culture procedure

Brain tissue processing and culture procedures have been described in detail before 38, 39, 43. Briefly, neocortical tissue was cut into rod‐like pieces and chopped using a McIlwain tissue chopper, in such a way that 300‐μm‐thick slices comprising all cortical layers and a part of the white matter were obtained 39. The procedure did not involve mechanical or enzymatic dissociation of cells. Slices of these specimens were either harvested directly after tissue processing [ie, at day in vitro 0 (DIV 0)] or after different periods in vitro, ranging from DIV 1 to DIV 37 and occasionally longer. The shorter periods were mainly used to collect evidence for co‐localization of markers. Slices destined for culture were systematically distributed over 24‐well plates. Each well contained one slice with 1 mL (resected tissue) or 0.5 mL (autopsy tissue) R16 medium 28. The larger volume of medium used for resected tissue was necessary to prevent exhaustion of the nutrients. Three times a week, 200 μL (resected tissue) or 100 μL (autopsy tissue) of medium was replaced. Viability of the cells in the slices was assessed using a Live/Dead kit (L3224, Life Technologies, Bleiswijk, the Netherlands) according to the manufacturer's protocol.

Counts of dividing/reactive cells

To label dividing cells, 10 μM 5‐ethynyl‐2′‐deoxyuridine (EdU, Life Technologies) was added to the culture medium for varying periods of time. EdU staining did not require acid pretreatment 30 and was visualized with azide‐conjugated fluorochrome (Alexa‐594 or Alexa‐488) according to the protocol provided by the manufacturer (Life Technologies). In certain experimental conditions, mitomycin C (10 μg/mL final concentration, Sigma, Zwijndrecht, the Netherlands) was applied to inhibit cell division. In experiments with tissue from three patients, at different time points during culturing, slices (n = 4/time point/patient) were incubated with EdU for 24 h, briefly washed and immediately fixed. In each slice, images were taken at systematic random locations (ie, with a random start and at fixed distances) using a confocal laser scanning microscope (Zeiss 410, Oberkochen, Germany or Leica TSC SP5 II, Mannheim, Germany). At each location, the image was scanned at two planes, either 15 μm (Zeiss 410, 25× objective) or 8 μm (Leica SP5, 40× oil objective) apart, along the z‐axis, and stored in the red and green channel, respectively. In this way, we obtained sets of physical dissectors 33, which we used to estimate the density of DNA synthesizing cells.

Immunohistochemical procedures

Slices from epilepsy tissue were fixed overnight in 4% formaldehyde in PBS. Supporting Information Table S3 summarizes the primary antibody data. For diamino‐benzidine (DAB) staining, HRP‐conjugated antibodies (1:100, DAKO Cytomation, Heverlee, Belgium) or biotinylated antibodies (1:400, Vector Labs, Burlingame, CA, USA) were used. Biotinylated second antibodies were followed by incubation with ABC kit (1:800, Vector). In DAB protocols, the slices were pretreated with 3% hydrogen peroxide in 10% methanol to destroy endogenous peroxidase activity. Fluorescent second antibodies were conjugated with Cy‐2 (1:100) or Cy‐3 (1:400, Jackson ImmunoResearch, West Grove, PA, USA), or with Alexa‐488 or Alexa‐594 (1:800, Life Technologies). Sudan black (Brunswig Chemie, Amsterdam, the Netherlands) was used to reduce auto‐fluorescence. In some fluorescent procedures, ethidium‐bromide homodimer (2 μL/mL, Life Technologies) was added during the last antibody incubation to visualize the nuclei. For staining with PCNA, Olig2 or MIB‐1 slices were pretreated with 10 mM citrate buffer (pH 6.6, 90°C) containing 0.05% Tween. Fluorescence images were recorded using a confocal laser scanning microscope (Zeiss 410 or Leica TSC SP5 II), DAB stained images were obtained using a Zeiss Axioplan 2, equipped with an Evolution MP CCD camera (MediaCybernetics, Ottawa, Canada). Graded scores of GFAP, HLA, Nestin, NeuN, PCNA and Vimentin stained slices were used to semi‐quantitatively evaluate the extent of irregular neuron staining pattern, cell proliferation or glial cell reactivity in tissue slices as described before 40. NeuN scores ranged from 0 to −5, indicating progressive reduction of NeuN immunoreactivity: 0 apparently normal staining; −1, minor deviations; −2, means some layers show focal absence of NeuN⁺ neurons; −3, intermediate absence of staining; −4, substantial parts of layers lack NeuN⁺ neurons; −5, global absence of NeuN staining. Grades for GFAP were 0, no staining; 1, a few normal astrocytes or astrocytic processes; 2, many normal astrocytes; 3, many normal and some reactive astrocytes; 4, reactive astrocytes dominate with some aggregates; 5, many aggregates of reactive astrocytes and usually many GFAP⁺ fibers or processes. HLA was graded as follows: 0, no staining; 1, few resting/ramifying microglia; 2, many resting/ramifying microglia; 3, widespread microglia; 4, massive (reactive) microglia; 5, excessive reactive microglia. Grades for Vimentin in DIV 0 slices ranged from 0 (minimal staining of blood vessels) to 5 (strong staining of blood vessels and several astrocytes). Nestin staining in DIV 0 slices was usually limited to some indistinct background (grade 0) and background plus several blood vessels (grade 1). For both Vimentin and Nestin staining in cultured slices the grades from 0 to 5 reflect the abundance of reactive cells. Grades for PCNA (0 to 5) provide an indication of the number of PCNA⁺ nuclei. In case of doubt, intermediate scores were applied. For comparison of the scores after some time in vitro with those of DIV 0, slices from the same patient were used. The number of slices per time point ranged from 1 to 4, depending on the availability of tissue.

Rationale of the applied antibodies

Here, we summarize the antibodies used to identify different cell types in resected human brain tissue. Neuronal nucleic protein (NeuN) and microtubule‐associated protein (MAP2) were used to visualize neuronal cells and neurites 12, 19. Glial fibrillary acidic protein (GFAP) and astrocytic calcium‐binding protein (S100β) were meant to identify astrocytes 1, 18, 21, 26, 27, 31, 44, but it had been reported that S100β may also stain oligodendrocytes 8. Human leukocyte antigen (HLA), ionized calcium‐binding adaptor (Iba1) and lysosomal marker (CD68) were used to visualize microglia/macrophages 10, 22, 31, 34, 44. Antibodies against myelin basic protein (MBP), galactocerebroside (GalC) and oligodendrocyte marker O4 detected mature oligodendrocytes (12, 19). Nestin and Vimentin were used to stain astroglia precursors 27 and reactive cells 11, 18. In reactive astrocytes, Nestin correlates with Vimentin 11, 18. However, Nestin 34 and Vimentin 14 have also been associated with microglia. For proliferating cells, PCNA and Ki67 (MIB‐1) were used 1, 12, 19. Ki67 stained markedly fewer nuclei than PCNA in both cultured and noncultured slices, but the staining was very irregular and was therefore not employed for double staining or cell counts. Human transcription factor Olig2, a marker of oligodendrocyte and astrocyte precursors has also been associated by different investigators with subpopulations of reactive astrocytes and proliferating cells 1, 5, 6, 9, 12, 19. At DIV 0, it vaguely labeled nuclei in the cortex of certain patients. Unfortunately, NG‐2 chondroitin sulfate proteoglycan, another marker for oligodendrocyte precursor cells and (possibly) other cycling cells in the neocortex 5, 9, 12, 19, 25, 31, produced very irregular staining in our tissue. As far as we know, Tubulin βI has not been associated with any CNS cell type, but appeared to stain microglial cells, some neuronal somata and apical dendrites, and cells that were morphologically similar to pericytes in noncultured slices and to stain reactive cells in cultured slices. Blood vessels were stained with von Willebrand Factor (vWF; endothelial cells) or with laminin (basement membrane). As a reference for Western blots, we used β‐actin.

PAGE electrophoresis

At least 12 tissue slices per condition (DIV 0 or DIV 28) were used for protein extraction. Tissue homogenization was performed in 500 μL suspension buffer (100 mM NaCl, 10 mM Tris‐HCl, 1 mM EDTA, 0.5% Triton X‐100 (pH 8.0) to which the following protease inhibitors (Sigma) were added: chymostatin (f.c. 20 μM), leupeptin (f.c. 10 μg/mL), antipain (f.c. 20 μM), pepstatin (f.c. 1 μM), PMSF (f.c. 100 μg/mL). Protein concentration was determined using the BCA Protein Kit Assay (Thermo Scientific, Breda, the Netherlands). Per slot, a total of 10 μg of homogenate was loaded on a 5%, 8% or 10% polyacrylamide gel or precast NuPage Novex Bis‐Tris 4%–12% gradient gel (Invitrogen). Images of gels were obtained with an Odyssey infrared system (LI‐COR, Lincoln, NE, USA). BenchMark Pre‐Stained Protein Ladder (Life Technologies) or Precision Plus Protein Kaleidoscope standards (Bio‐Rad Laboratories, Veenendaal, the Netherlands) were used as molecular weight markers. We used β‐actin as a reference for protein load.

Reverse transcriptase quantitative PCR (qPCR)

RNA expression levels of several human genes were determined using reverse transcriptase qPCR (RT‐qPCR). Isolation of total RNA using Trizol (Invitrogen) was followed by DNAse I (Amplification Grade, Life Technologies) treatment and first‐strand cDNA synthesis using Superscript II (Life Technologies) reverse transcriptase. For each RNA isolation, 24 slices per time point (DIV 0, 10 or 28) were used. All steps were performed according to the manufacturer's recommendations. Supporting Information Table S4 summarizes information concerning the primers used for amplification. Amplification was detected with ABI 7300 real‐time PCR systems using SYBR green master mix (Life Technologies). Controls omitting reverse transcriptase or template were included in each PCR reaction. Crossing point (Cq) values, reaction efficiencies and the fluorescence threshold were estimated with version 12 of LinRegPCR shareware 29, which was then used to calculate the corresponding RNA quantities. For normalization of the qPCR data, we used eukaryotic elongation factor 1 alpha (EF1α) and glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH) as reference genes, which were selected based on expression stability among a group of putative reference genes 36, and on the proportion of explained variability regarding all target genes.

Statistical analysis

TIBCO Spotfire S+ software (version 8.2.0; TIBCO, Seattle, WA, USA) was used for statistical analyses and preparation of graphs. The chi‐square test was used to analyze gender vs. group. Correlative analyses involving immunocytochemical scores were assessed using the Spearman test. The paired t‐test was used to analyze differences in ¹⁰log‐transformed gene expression data of cultured slices with respect to noncultured slices. To compare immunocytochemical scores of GFAP, HLA, Nestin, NeuN, PCNA and Vimentin in slices before and after culturing, the Wilcoxon signed ranks test was used. Linear regression was used for the analysis of ¹⁰log‐transformed qPCR data from noncultured tissue from epilepsy patients and control subjects with respect to age and group effect. The comparison of immunocytochemical scores at DIV 0 of epilepsy patients and control subjects involved an ordinal response variable (scores) with many ties as well as two explanatory variables: one binary (group) and one continuous (age). As there is no standard test procedure for this, we used the Spearman rank method to test whether the scores correlated with age in either the control or the epilepsy group and we used the Mann–Whitney test to analyze the scores with respect to the group effect. The presented P‐values pertain to two‐sided alternative hypotheses. To adjust P‐values or construct confidence intervals corrected for multiple testing we used the Benjamini–Hochberg false discovery/coverage criterion (FDR/FCR) 3.

Results

In vitro injury caused degeneration of neurons

For this study, we used resected neocortex specimens from a total of 76 patients (Supporting Information Table S1 and Supporting Information Figure S1A) with temporal lobe epilepsy (TLE). Viability staining of unfixed brain slices is a convenient way to obtain a general idea about the presence of viable and dead nerve cells, and about several structural and morphological features of the tissue 38. When we assessed the viability of cells in resected specimens after a few days in vitro, the neurons in slices harvested from most of the epilepsy patients appeared less healthy and orderly as compared with neuronal nuclear protein (NeuN) staining of noncultured slices (Figure 1A–C). Moreover, for unclear reasons, the viability stain did not detect cells (dead or alive) in large areas of resected tissue (Figure 1A), although immunocytochemical stains showed the presence of many cells. For this reason, we did not perform viable cell counts. After about a week in vitro, the neurons started to lose NeuN expression, and this loss was accompanied by shrinkage of cell bodies and nuclei (Figure 1D). Although by 4 weeks NeuN and microtubule associated protein 2 (MAP2) staining had decreased considerably, the shrunken neurons persisted (Supporting Information Figure S2). These observations were corroborated by semi‐quantitative scores of NeuN immunostaining (median difference = −3, P = 0.0002, number of patients = 24) and Western blots of NeuN and MAP2 (Supporting Information Figure S3A,B). Moreover, gene expression of RNA binding protein FOX3 (RBFOX3, encodes NeuN), MAP2 and neuron specific enolase (NSE) was strongly down‐regulated (Figure 2). The data suggested that expression of RBFOX3 (−20.7‐fold at DIV 0 to −35.6‐fold at DIV 28) and of MAP2 (−7.9‐fold at DIV 10 to −11‐fold at DIV 28) continued to decline during culture, whereas expression of NSE did not change between DIV 10 and DIV 28 (Supporting Information Table S5).

Figure 1.

Resected brain tissue slices show an injury response in vitro. (A) Viability staining of resected brain tissue at 0 days in vitro (DIV 0) often detected very few cells. The panel shows viable neurons (green with a dark nucleus, arrows), dead cells (red nuclei, asterisk) and cells in an intermediate state (red nucleus in green cytoplasm, arrow heads). The sparseness of red nuclei suggests that there must have been many living cells. (B) An area with more neurons at DIV 5. Arrows, arrow heads and asterisks have the same meaning as in (A). NeuN immunostaining at DIV 0 (C) and DIV 9 (D) clearly demonstrate shrinkage of neuronal somata and nuclei (arrows) accompanied by a strong reduction of NeuN intensity during time in vitro. (E) IVR cells (arrow heads) and IVFL cells (arrows) visualized by viability staining (DIV 11). Before the appearance of IVR cells, Vimentin (F) stained some astrocytes (asterisk), peri‐vascular cells (arrows) and many blood vessels (DIV 2) whereas Nestin (G, DIV 0) stained blood vessels. IVR cells (arrow heads in H and I) were Vimentin⁺ (H, DIV 9) and/or Nestin⁺ (I, DIV 27); IVFL cells (arrows) were Vimentin⁺ (H, DIV 9). (J) Nestin) and Vimentin could be present to varying extent in the same IVR cells (arrows). Laminin staining at DIV 0 (K) and at DIV 28 (L) shows that the basal lamina and perivascular cells (arrows) may remain unperturbed during the reactive process. LD = Live/Dead viability staining with components: cakcein (green) and ethidium homodimer (red); g = green; r = red. Scale bars: A, B, E, G, H, I, J = 50 μm; C, D, F, K, L = 30 μm.

Figure 2.

Differences in gene expression levels at DIV 10 and DIV 28. Presented are the mean values and 95% confidence intervals of the changes from DIV 0 to DIV 10 and DIV 28, respectively. The expression of HLA ‐ DR continued to increase from DIV 10 to DIV 28. MAP2 and RBFOX3 expression showed a tendency to decrease over longer time in vitro, whereas the other genes remained more or less constant from DIV 10 onwards. GFAP expression was unchanged at DIV 10 and only marginally increased at DIV 28, while the difference in AIF 1 expression had returned to insignificance by DIV 28. The 95% confidence intervals were adjusted using the Benjamini–Hochberg procedure for multiple testing 3. The vertical gray lines mark the twofold change levels. The corresponding fold changes, adjusted P‐values and number of patients are described in Supporting Information Table S5.

Emergence of reactive cells

Concurrently with the degeneration of neurons, large reactive cells with long extending processes emerged along blood vessels (Figure 1E–J). Many of these in vitro reactive (IVR) cells expressed Vimentin and/or Nestin (Figure 1H–J). In several cases, also, variable numbers of in vitro foamy macrophage‐like (IVFL) cells emerged (Figure 1E,H). In tissue that was fixed at DIV 0 (ie, not cultured) no reactive cells were present and both Vimentin and Nestin predominantly labeled blood vessels (Figure 1F,G). To a variable extent, Vimentin also labeled resident astrocytes and pericytes (Figure 1F), whereas Nestin usually showed a grainy background (Figure 1G). With increasing time in vitro, these features were lost and replaced by IVR and IVFL cells (Figure 1H,I). Semi‐quantitative scores for Vimentin (median difference = +1, P = 0.004, number of patients = 24) and Nestin (median difference = +2.5, P = 0.0002, number of patients = 24) immunoreactivity, Western blot analysis (Supporting Information Figure S3C,D) and qPCR (Figure 2, Supporting Information Table S5) confirmed that these markers were strongly up‐regulated during the IVR process. The close association of many IVR cells with blood vessels suggested that blood vessel integrity might be concomitantly affected. Surprisingly, in tissue from some patients, blood vessels retained a normal appearance for several weeks in vitro (Figure 1K,L) although in slices from other patients, the vascular immunoreactivity for Laminin or von Willebrand factor (vWF) gradually declined. It is also noteworthy that morphologically normal pericytes (Figure 1H–K) might still be present at DIV 28 (Figure 1L), although reactive pericytes were observed as well (see below).

Role of glia

Microglia and astrocytes play important roles in the brain's reaction to injury. Slices from some epilepsy patients showed relatively high numbers of HLA⁺ microglia prior to their transfer to culture conditions. Some of these microglia cells displayed a ramifying phenotype, whereas others had an activated morphology (Figure 3A). However, further activation during culture was commonly observed (Figure 3B–D), which was also evident from the increased HLA scores (median difference = +2.5, P = 0.003, number of patients = 23), Western blot analysis (Supporting Information Figure S3E) and continued up‐regulation of HLA gene expression from +4.1‐fold at DIV 10 to +15‐fold at DIV 28 (Figure 2, Supporting Information Table S5). Many HLA⁺ IVR cells had large cell bodies with long hairy processes (Figure 3B), whereas a varying subpopulation consisted of IVFL cells (Figure 3C,D). Inside these IVFL cells, HLA seemed to be mainly confined to stubby appendages, some of which may actually be part of contacting processes from other cells (Figure 3C,D). Resident microglia also expressed ionized calcium‐binding adaptor 1 (Iba1) (Figure 3E,F). In cultured tissue, Iba1 predominantly labeled IVFL cells that did not necessarily co‐express HLA (Figure 3G). However, the brief and rather modest up‐regulated expression of Allograft inflammatory factor 1 (AIF1), the gene encoding for Iba1, indicated that the proportion of microglia‐derived foam cells may be limited. The lysosomal marker CD68, which is often used to detect microglia/macrophages 10, 22 presented an indistinct, faint grainy background pattern in noncultured slices (not shown), but was clearly detectable in IVFL cells of cultured slices, and, to a variable extent, co‐localized with Iba1 (Figure 3H).

Figure 3.

Microglia responded robustly to in vitro injury. (A–D) HLA ⁺ microglia may be reactive at DIV 0 but consistently showed further activation in vitro; ramifying (arrow) and reactive (arrow heads) microglia at DIV 0 (A). (B) At DIV 28, microglia displayed more pronounced reactive morphology with large cell bodies and long winding processes. (C and D) IVFL cells (asterisks) often had stubby HLA ⁺ appendages, which may be partly caused by contacting processes (arrows) from HLA ⁺ IVR cells (arrow head). (E) Nonreactive microglia cells that contain both HLA and Iba1 at DIV 0. (F) In some brain specimens, microglia was either predominantly HLA ⁺ (arrow head) or mainly Iba1⁺ (asterisk). (G) At DIV 23, IVR cells were only HLA ⁺, while IVFL cells could be either double labeled (arrow) or Iba1⁺ (asterisk). (H) At DIV 23, Iba1 and CD68 commonly co‐localized in IVFL cells, but with variable intensity. bv = blood vessel; g = green; r = red. Scale bars: E, G, H = 50 μm; A, B = 30 μm; F = 25 μm; C, D = 10 μm.

Resident GFAP⁺ astrocytes in the white matter of neocortex slices of many epilepsy patients showed some degree of activation, characterized by increased soma sizes and thickened processes (data not shown). However, unexpectedly, further activation of GFAP⁺ astrocytes in vitro was very variable (Figure 4) and often hypertrophied resident GFAP⁺ astrocytes could be found, apparently unchanged, at DIV 28. Inconclusive GFAP immunocytochemical scores (median difference = +1, P = 0.09, number of patients = 23), down‐regulated protein levels (Supporting Information Figure S3F), unchanged GFAP gene expression at DIV 10 and only marginal up‐regulation at DIV 28 (Figure 2, Supporting Information Table S5) corroborated this point. Double staining of GFAP and the astrocytic calcium‐binding protein S100β showed that in noncultured slices most astrocytes were S100β⁺/GFAP^‐ and that a minority was S100β⁺/GFAP⁺ (Figure 4A). Infrequently, S100β^– /GFAP⁺ astrocytic cell bodies were observed (Figure 4B). In cultured slices, apparently, many astrocytes transformed into S100β⁺ IVR cells, but did not consistently up‐regulate GFAP expression (Figure 4C,D). Surprisingly, the relative gene expression of S100β at DIV 0 (data not shown) and the up‐regulation during culture were hardly different from those of GFAP (Figure 2, Supporting Information Table S5).

Figure 4.

Astrocytes reacted ambiguously to in vitro injury. (A) At DIV 0, S100β⁺ astrocytes (arrow heads) outnumbered GFAP⁺/S100β⁺ astrocytes (arrows). (B) Very few resident astrocytes were GFAP ⁺/S100β^‐ (asterisk). (C) In some specimens, S100β⁺ IVR cells (arrow heads) contained few GFAP ⁺ filamentous structures (arrows) at DIV 28. (D) In other specimens, GFAP expression in IVR cells at DIV 23 ranged from relatively abundant (yellow somata, arrows) via moderate presence (arrow heads) to apparently absent (asterisks). bv = blood vessel; g = green; r = red. Scale bars: A, B, C, D = 50 μm.

Co‐localization of astrocytic and microglial markers in reactive cells

As is clear from the above, the reactive cells expressed several markers other than Vimentin and Nestin. In Supporting Information Figure S4, we show the presence or absence of co‐localization of Vimentin and Nestin with HLA, Iba1, CD68, GFAP and S100β. At DIV 0, Vimentin and Nestin were found in some endothelial cells (Supporting Information Figure S4A,B). HLA partly co‐localized with Nestin and Vimentin in IVR cells, but not in resident microglia of noncultured slices (Supporting Information Figure S4C,F). Iba1 did not co‐localize with Vimentin and Nestin at DIV 0, but was present in Vimentin⁺ IVFL cells in cultured slices (Supporting Information Figure S4G,I). Iba1 (mainly IVFL cells) and Nestin (usually IVR cells) remained mutually exclusive in cultured slices (Supporting Information Figure S4J). At DIV 28, CD68 and Vimentin were co‐localized in many IVFL cells (Supporting Information Figure S4K). IVFL cell scores in cultured slices based on DAB stainings for Vimentin, HLA and CD68 were not correlated, suggesting that there may be several different subpopulations of IVFL cells with variable expression of markers (Table 1). This is supported by the observed variable double labeling of IVFL cells by Iba1/HLA, Iba1/CD68 (Figure 3G,H) and Iba1/Vimentin (Supporting Information Figure S4K). Alternatively, the different marker expression may indicate a temporal variability in the emergence of IVFL cells. At DIV 0, GFAP was sometimes present in activated resident astrocytes (Supporting Information Figure S4L), but did not co‐localize with Nestin (Supporting Information Figure S4N), However, if GFAP was present in IVR cells, Vimentin and Nestin were often also present (Supporting Information Figure S4M,O). In noncultured slices, there was no overlap of S100β with Nestin and hardly with Vimentin, however, in cultured slices co‐localization of S100β with these reactivity markers was very robust (Supporting Information Figure S4P,T).

Table 1.

Correlation (Rho) of immunocytochemical markers for IVFL cells

Markers	N	Rho	P‐values
VIMENTIN/HLA	20	0.36	0.18
CD68/HLA	8	0.40	0.39
CD68/VIMENTIN	8	0.07	0.88

N = number of patients.

P‐values were adjusted for multiple testing 3.

Co‐localization of astrocytic and microglial cell markers with Vimentin and Nestin could be an indication of different subpopulations of reactive cells. Therefore, we investigated whether microglia and astrocytic markers could be simultaneously present in IVR or IVFL cells (Figure 5). While under normal conditions (ie, at DIV 0), no co‐localization of GFAP or S100β with HLA was observed (Figure 5A–C), in cultured slices, HLA⁺/GFAP⁺ and HLA⁺/S100β⁺ IVR cells were present (Figure 5B–D). Similarly, co‐localization of S100β with CD68 (Figure 5E), and of S100β with HLA (not shown) was commonly found in IVFL cells. As expected, GFAP and Iba1 did not show double labeling of cells in noncultured slices, and because Iba1 was mainly present in IVFL cells and GFAP immunoreactivity usually did not include IVFL cells, these markers remained separated in cultured slices (Figure 5F,G). Because of antibody incompatibility, co‐localization of Iba1 and S100β could not be determined.

Figure 5.

In vitro, reactive cells expressed both microglial and astrocytic markers. (A) As expected, GFAP (arrow) and HLA (asterisk) were strictly separated at DIV 0. (B) If GFAP ⁺ IVR cells were present at DIV 29 GFAP ⁺/HLA ⁺ double‐labeled IVR cells (arrows) were also observed. One double‐labeled cell had an IVFL morphology (asterisk). (C) Similarly, at DIV 0, S100β (arrow heads) did not co‐localize with HLA (arrows). In this specimen the microglia cells were already somewhat activated (arrows). (D) At DIV 23, S100β⁺/HLA ⁺ IVR cells (arrows) and IVFL cells (not shown) were found. Some IVFL cells (asterisks) were only S100β⁺ (D). Note that some nonreactive S100β⁺ astrocytes (arrow head) could still be found. (E) S100β did co‐localize with CD68 in foam cells at DIV 28. (F, G) GFAP and Iba1 (red) did not co‐localize at either DIV 0 or DIV 28. (F) Note the typical astrocytic (arrow) and microglial (asterisks) morphology at DIV 0. (G) At DIV 28, in this specimen, GFAP was present in IVR cells (arrows) and Iba1 in IVFL cells (asterisks). g = green; r = red. Scale bars: B, D, G = 50 μm; A, C, F = 25; E = 20 μm.

Participation of other cell types in the reactive process

Normal and reactive perivascular cells with long processes along blood vessels were observed using Vimentin and/or Tubulin βI staining [Supporting Information Figure S5A,F; cf. Figure 4 in 42 ]. Oligodendrocyte transcription factor 2 (Olig2), a marker of oligodendrocyte precursor cells and polydendron cells, labeled nuclei that were widely dispersed in noncultured neocortical slices (Supporting Information Figure S5G,P). In noncultured slices, there was no evidence of Olig2 co‐localization with Vimentin, Nestin or microglial markers (Supporting Information Figure S5I,K,O), but some GFAP⁺ astrocytes had Olig2⁺ nuclei (Supporting Information Figure S5M). In cultured slices, cytoplasmic co‐localization of Olig2 with Vimentin, Nestin or GFAP was observed (Supporting Information Figure S5H,J,L,N). A few HLA⁺ IVR cells had an Olig2⁺ nucleus (Supporting Information Figure S5P). Overall, immunocytochemical staining and gene expression (Figure 2, Supporting Information Table S5) suggested that Olig2 expression decreased in vitro. Markers of mature oligodendrocytes such as myelin basic protein (MBP), galactocerebroside (GalC) and oligodendrocyte marker 4 (O4) showed the same staining pattern at DIV 0 and at DIV 28, but occasionally, the staining in cultured slices was less intense (not shown). No IVR cells or IVFL cells expressing these markers were detected, indicating that mature oligodendrocytes did not participate in the reactive process. In a number of cases, a tumor was associated with the epilepsy (Supporting Information Table S1). Although the tissue we obtained was not adjacent to the tumor region, we cannot exclude that infiltrating cells have been present. However, the Mann–Whitney test showed that gene expression and the scores of immunoreactivity were not different between patients with or without tumor diagnosis (all adjusted P‐values > 0.20).

Time course of the reactive process

Next, we addressed the time course of the reactive process. At DIV 1, 24 h treatment of slices with mitomycin C prevented emergence of IVR and IVFL cells (Figure 6A,B), demonstrating that cell division is required for the reactive process. To count dividing cells at various time points, we labeled slices with a 24 h pulse of the thymidine analog 5‐ethynyl‐2′‐deoxyuridine (EdU). The observed time courses revealed complex waves of cell division that differed among patients (Figure 6C). Our data showed no significant DNA synthesis before the tissue was transferred to the in vitro conditions (Figure 6C). Although many small nuclei and some neuronal nuclei in noncultured slices expressed the cell cycle marker PCNA (Supporting Information Figure S6A–C,E,G,I), very few actually incorporated EdU during a 24 h pulse at DIV 0 or DIV 1 and even at later stages the ratio of (EdU⁺ and PCNA⁺)/PCNA⁺ cells remained modest (Figure 6D). At later stages, most IVR cells and still many nonreactive cells had PCNA⁺ nuclei (Supporting Information Figure S6D,F,H,J), while gene expression of PCNA was roughly threefold increased (Figure 2, Supporting Information Table S5). Changes in immunocytochemical scores of PCNA during culture could be positive or negative, depending on whether the initial scores were low or high, respectively (not shown). This suggests that there may be cycling cells that are independent of the reactive process. Interestingly, at DIV 0, the majority of PCNA⁺ nuclei also expressed Olig2 (Supporting Information Figure S6I). Unlike the gene expression data that were obtained at fixed time points (DIV 0, 10 and 28), material for immunoreactivity scores was harvested at a wider range of time points. This allowed us to explore whether the differences in the scores correlated with the time spent in vitro. It appeared that longer time in vitro correlated negatively with NeuN staining and positively with HLA and Vimentin staining (Supporting Information Figure S7).

Figure 6.

Cell proliferation is an essential part of the reactive process. (A) Nestin⁺ IVR cells in an untreated slice at DIV 15. (B) No Nestin⁺ IVR cells were observed at DIV 15 when slices were treated at DIV 0 for 24 h with 10 mg/mL mitomycin C (mito). (C) Counts of dividing cells showed that the reactive process involved roughly two waves of cell division; at different time points slices (four slices/time point/patient) were incubated with EdU for 24 h, washed briefly with medium and fixed. No labeled cells were observed in slices (n = 8) of a patient that were treated with EdU within 3 h after resection. (D) The proportion of (EdU ⁺ and PCNA ⁺)/PCNA ⁺ nuclei was initially very low, then strongly increased and declined again (four slices/time point). At all time points, the proportion was below 20%. E‐numbers correspond to patient numbers in Supporting Information Table S1. The flags indicate standard errors of the means. Scale bars: A, B, C, D = 100 μm.

Comparison of resected and post‐mortem brain tissue

We wondered whether the epileptic status of TLE patients would predispose resected tissue to the observed injury response. Therefore, we compared noncultured slices from epilepsy patients with noncultured slices from post‐mortem brain tissue of 28 control subjects (C) who died without any neurological cause (Supporting Information Table S2, Supporting Information Figure S1B). The percentages of males (TLE: 39%, C: 43%) and females (TLE: 61%, C: 57%) were comparable for the two groups (chi‐square test, P = 0.76). As reported before (11), the median age of control subjects (81 years) was roughly twice that of TLE patients (41 years), which was highly significant (P = 7 × 10⁻¹⁴). Another pertinent difference was that the brain of control subjects remained untouched and deprived of nutrients in the skull for several hours during the post‐mortem delay before it was dissected (Supporting Information Figure S1B). Additionally, post‐mortem tissue was obtained from the precentral gyrus as part of a different experimental paradigm.

Comparison of Figure 7A with Figure 1A,B illustrates that after a few days in vitro, neurons in post‐mortem tissue often displayed a more viable appearance than those in resected tissue. At DIV 0, neuronal morphology was often still comparable (Figures 1C and 7B), but immunoreactivity scores showed that slightly more neurons in tissue from epilepsy patients did not stain for NeuN (Table 2). Gene expression of NeuN (RBFOX3) was also lower in epilepsy patients than in control subjects (Table 3). However, these differences were minor in comparison with the in vitro events (Supporting Information Tables S5 and S7). There was no evidence for more activated resident GFAP⁺ astrocytes or HLA⁺ microglia in TLE tissue as compared with C tissue (Table 2). Together with the significantly lower AIF1 gene expression in epilepsy tissue, this indicated that the reactive process did not have an inflammatory origin. Furthermore, immunoreactivity scores for the markers of the IVR cells in epilepsy patients were either similar (Vimentin) to or lower (Nestin) than those in control subjects (Table 2). The higher abundance of PCNA⁺ nuclei suggested that epilepsy tissue contained many more proliferating resident cells (Table 2). There was no difference between epilepsy patients and control subjects as far as expression levels of the other genes that we investigated were concerned and, with the exception of HLA, none of the genes showed an age effect (Table 3). In another study (unpublished), we observed that cultured post‐mortem brain tissue from control subjects or subjects with a neurodegenerative disease may also show reactive cells (Figure 7C,D), albeit at a much lower frequency and less abundantly than epilepsy tissue. The infrequent occurrence excludes systematic investigation of this phenomenon in post‐mortem tissue. The stronger and more consistent response of resected tissue is probably caused by the higher metabolic state of the cells at the time when the damage was sustained. To prevent exhaustion of the medium, resected tissue was cultured at twice the volume used for post‐mortem tissue 37. Also noteworthy is that, apparently, viable neurons can be present near a reactive cell (Figure 7C), suggesting that the reactive process and neuronal degeneration may be separate processes.

Figure 7.

Post‐mortem neocortex from control subjects. (A) Viability staining of a slice from a deceased control subject at DIV 5; the neurons have a healthier appearance than those in resected tissue (cf. Fig. 1A,B). (B) NeuN stainings of post‐mortem tissue at DIV 0 did not differ from those of resected tissue (cf. Fig. 1C). (C, D) IVR cells (asterisks) at DIV 28 in the gray (C) and white (D) matter of post‐mortem neocortex from a control subject. (C) Note that viable neurons (arrows) may still be present in close proximity of an IVR cell (asterisk). LD = Live/Dead viability staining with components: calcein (green) and ethidium homodimer (red). Scale bars: A = 100 μm; C, D = 50 μm; B = 30 μm.

Table 2.

Scores of immunocytochemical markers in tissue of epilepsy patients (TLE) and control subjects (C) at DIV 0

A. Correlation of immunoreactivity scores with age (Spearman)
Marker	TLE			C
	NTLE	Rho	P‐values	NControl	Rho	P‐values
GFAP	74	0.31	0.04	28	−0.12	0.59
HLA	73	0.24	0.10	28	0.15	0.59
NESTIN	70	−0.34	0.03	27	−0.13	0.59
NeuN	74	−0.18	0.24	28	−0.13	0.59
PCNA	63	0.19	0.24	23	0.12	0.59
VIMENTIN	70	0.17	0.25	25	−0.23	0.38

B. Comparison of immunoreactivity scores between TLE and C (Mann–Whitney)
Marker	TLE		C
	NTLE	Median scores	NControl	Median scores	P‐values
GFAP	74	2.5	28	2	0.10
HLA	73	2	28	2.5	0.22
NESTIN	70	0.5	27	1	0.01
NeuN	74	−1.25	28	−0.5	0.04
PCNA	63	3.5	23	1.5	0.00001
VIMENTIN	70	2.5	25	2.5	0.96

P‐values were adjusted for multiple testing 3.

N_TLE = number of epilepsy patients; N_Control = number of control subjects.

Table 3.

Fold differences in gene expression in neocortical tissue of epilepsy patients (TLE) and control subjects (C) at DIV 0

Gene	NTLE	NControl	Fold difference	P Group	P Age
AIF1	17	6	−2.6	0.045	0.34
GFAPv12	22	12	−1.8	0.55	0.34
HLA.DR	22	12	−1.7	0.45	0.02
MAP2	24	12	1.4	0.46	0.61
NESTIN	24	12	1.2	0.30	0.46
NSE	24	12	−1.1	0.34	0.31
OLIG2	20	12	2.0	0.32	0.79
PCNA	24	12	1.1	0.20	0.09
RBFOX3	17	6	−2.4	0.03	0.17
S100B	17	6	1.1	0.46	0.26
VIMENTIN	24	12	1.6	0.43	0.34

P‐values were adjusted for multiple testing 3.

N_TLE = number of epilepsy patients. N_Control = number of control subjects. P _Group = P‐values for group. P _Age = P‐values for age.

Discussion

Healthy brain tissue that needs to be surgically resected to enable removal of diseased deeper brain structures could be an invaluable source to study many aspects of cellular functioning in the human brain, and is quite possibly preferable to post‐mortem tissue. However, resected tissue always responds to interruption of the circulation and processing of slices with a strong injury reaction. Thus, the relative integrity of the cytoarchitecture and microenvironment in resected tissue slices provides an ideal opportunity to study neuropathological consequences of extensive damage to the human brain.

The damage inflicted by our culture procedure caused neuronal degeneration. However, comparison with post‐mortem tissue from control subjects suggested that some mild neuronal degeneration may have been present before resection. For example, neurons in noncultured epilepsy tissue exhibited more irregular NeuN staining than in noncultured post‐mortem tissue 40 and RBFOX3 expression was slightly lower in epilepsy patients. In previous work, we estimated the number of viable neurons in cultured slices to be roughly 3000/mm³ 43, but in resected tissue, the Live/Dead staining could not provide a reliable estimate. Still, the changes that manifested shrunken neuronal somata and neurites, severe loss of NeuN and MAP2 staining, and of RBFOX3, MAP2 and NSE gene expression only occurred during the in vitro conditions. However, even then many neurons did not disappear, nor were there obvious signs of removal of neurons or neuronal debris by IVR and IVFL cells.

The neuronal degeneration was accompanied by the emergence of large reactive cells that expressed markers of reactivity, such as Vimentin and Nestin 11, 18, 27, 32. Similarly to the occurrence of reactive astrocytes in a rat model of acute stab wound injury 1, the reactive cells predominantly emerged along blood vessels. Vimentin stained both IVR and IVFL cells, whereas Nestin hardly stained foam cells. While both markers gave a good idea of the extent of the reactive process, they did not stain each and every reactive cell. The lack of IVR and IVFL cells directly after resection and tissue processing, and the time course of their emergence indicated that the reactive process is not a consequence of epilepsy. Moreover, the observation of IVR cells in cultured post‐mortem tissue confirms that this process is not unique for epilepsy tissue. However, the occurrence of IVR cells in cultured post‐mortem tissue is much less frequent than in resected tissue. This may predominantly be associated with the post‐mortem delay, when the brain is left untouched, and is deprived of nutrients and oxygen before the tissue is dissected.

Most animal studies of brain injury concentrate on reactive astrocytes 1, 5, 6, 9, 16, 18, 19, 21, 31. The appearance of Vimentin and Nestin in reactive astrocytes is invariably accompanied by up‐regulation of GFAP 27, 32, 44. The proliferation of reactive astrocytes and differentiation of NG2⁺ glial precursor cells is dependent on the presence of Olig2 6, 19, which is strongly up‐regulated in mice after cortical injury 5, 9. In the neocortex of many epilepsy patients, hypertrophied resident GFAP⁺ astrocytes were present before the in vitro reactive process started (at DIV 0), but they were not specifically located along blood vessels. Besides, the abundance of GFAP immunoreactivity was only moderately higher than in post‐mortem neocortical slices from control subjects. The presence of Olig2⁺ nuclei in some of these GFAP⁺ astrocytes might be a sign of acute activation and the presence of Olig2 in the cytoplasm of some IVR cells may indicate that Olig2⁺ precursor cells had differentiated into an astrocytic type. However, Olig2 expression in resected tissue tended to decrease following damage, which might partly explain the ambiguous GFAP response. It could also be that reactive human astrocytes have low proliferation rates 7. Alternatively, proliferating astrocytes might initially be GFAP negative and start to express GFAP much later 16. Thus, Vimentin‐ and Nestin‐stained cells in cultured slices could predominantly represent GFAP‐negative IVR astrocytes. This is also suggested by the fact that the mature astrocyte marker S100β 26 was abundantly present in IVR and IVFL cells. It might thus be that the use of GFAP as activation marker for human astrocytes leads to underestimation of their contribution to the reactive process. However, S100β is not exclusive for astrocytes 32 and it cannot be excluded that some S100β⁺‐reactive cells derived from oligodendrocyte precursors 8. Although proliferation of microglia and astrocytes may be detrimental to neuronal survival, there is also evidence that reactive astrocytes may actually be essential for the protection of neurons after traumatic injury 21, 44. The presence of viable neurons near a reactive cell in cultured post‐mortem tissue also indicated that the reactive cells need not be harmful. However, if the neuroprotective capacity of reactive astrocytes depended on up‐regulation of GFAP expression, the ambiguous astrocytic injury response might be unable to prevent neuronal degeneration in resected tissue.

Microglia react rapidly and profoundly to injury 20, 23, 31 and may remain activated for long periods afterwards 10, 22. Microglia played a dominant role in the injury response of resected tissue and the protracted increase in HLA gene expression suggested that they remained activated for at least 4 weeks. As previous studies have shown that reactive microglia can express Vimentin or Nestin 14, 34, some Vimentin⁺ and Nestin⁺ IVR cells might be microglia instead of GFAP^– astrocytes. As expected, neither Vimentin nor Nestin stained resident microglia in noncultured slices. However, both Vimentin and Nestin co‐localized with HLA in many IVR cells, and Vimentin also co‐localized with HLA, Iba1or CD68 in many IVFL cells. Moreover, most resident nonreactive HLA⁺ or Iba1⁺ microglia had disappeared or down‐regulated their marker expression in cultured slices. AIF1 gene expression returned to preculture levels by DIV 28, which may indicate that the Iba1⁺ IVFL cells only emerged for a short period after the injury.

As reactivity markers co‐existed with microglial or astrocytic markers, we decided to check for possible co‐expression of microglia and astrocyte markers in both cultured and noncultured slices and concluded that the distinction between microglia and astrocytes was lost during the in vitro injury reaction. As far as we know, this is the first time that reactive cells expressing both microglial and astrocytic markers have been reported in injured human brain tissue. Interestingly, it has recently been shown that expression of molecules involved in antigen presentation may be up‐regulated in reactive astrocytes 44. In the resected tissue, resident glial or precursor cells initially expressed their appropriate markers. Following damage, a substantial number of these cells apparently not only altered their morphology, but also their molecular expression patterns. As a result, the relative contribution of microglia and astrocytes to the injury reaction became confounded. As microglia and astrocytes derive from different developmental lineages it seems unlikely that co‐expression of their markers in IVR cells is a simple illustration of de‐differentiation of either cell type. Vimentin, Tubulin βI and Olig2 staining suggested that pericytes also reacted to injury and their contribution may even be underestimated because of their small size and vague staining. Presently, we have no evidence for a prominent role of oligodendrocytes in the reactive process.

It has been inferred that the neocortex of epilepsy patients contains roughly 3× more persistently cycling progenitor cells than the neocortex of control subjects 12. We found a somewhat smaller difference between the scores of immunoreactivity for PCNA of epilepsy patients and control subjects. While this suggests that there were many cycling cells in the resected tissue, the majority of the PCNA⁺ cells at DIV 0 had a small nucleus and incorporated hardly any EdU. It thus seems likely that the division rate of these resident cells was very low—as has been established in rodent studies 25, 31. The abundant co‐expression of PCNA with Olig2 in these nuclei at DIV 0 agrees with previous studies 12, 25, 31 and may concern cells that normally generate post‐mitotic NG2+glia cells in the gray matter and olygodendrocytes in the white matter 9. Unfortunately, the irregular staining obtained with the NG2 antibodies precluded further elucidation of the role of these cycling cells. The emergence of IVR and IVFL cells with their relatively large PCNA⁺ nuclei incorporated substantial amounts of EdU and their proliferation could be prevented by mitomycin C. However, even in cultured slices, the majority of the PCNA⁺ nuclei did not incorporate EdU during a 24 h pulse, suggesting that the presence of PCNA both at DIV 0 and in vitro may also be related to DNA repair or aberrant cell cycle entry 15. The variability of the cell proliferation curves may be inherent in patient‐based research. The timing of the first peak of IVR cell proliferation in human tissue is comparable with that reported for microglia after brain injury in mice 20, 31, while it is late as compared with proliferating astrocytes in rats 16 and mice 6.

In conclusion, the reactive phenomena observed in resected human brain tissue in vitro partly agree with results from post‐mortem human studies 10, 12, 13, 37, animal studies that mostly used in vivo experiments 1, 5, 6, 9, 11, 16, 18, 19, 20, 22, 31, 44 or purified mouse astrocyte cultures 44. The most conspicuous difference with the animal studies is the ambiguous reaction of GFAP⁺ astrocytes. Moreover, the co‐expression of microglial and astrocytic markers has not been reported before. Other interesting studies using organotypic cultures involved neonatal rat and mouse tissue 4, 24. Although such tissue is still subject to developmental processes, it proved to be amenable to experimental treatments which inhibited cell death 4, 24. If such manipulations also effectively prevent the described reactive and neurodegenerative response in adult resected tissue, they may guide the development of therapeutic strategies for human brain injury without burdening patients while simultaneously making resected tissue amenable to a wider range of research questions concerning human brain function.

Supporting information

Figure S1. History of investigated neocortex tissue. (A) During life, TLE patients contracted epilepsy that proved to be refractory to medication. During the operation, the tissue may be affected by anesthesia and possibly unknown factors until at resection the circulation was interrupted. (B) Autopsy tissue has experienced an agonal state before death and a post‐mortem delay during which the tissue has probably reached a lower metabolic state before it was dissected and manipulated. After resection or autopsy, the neocortical tissue was quickly placed into buffer (Leibovitz 15, Invitrogen) and transported at ambient temperature to the culture room where the tissue was chopped into slices (40). Subsequently, the slices were distributed (plating) over 24‐well plates containing culture medium and maintained for variable periods. Some slices were directly fixed (for routine immunostaining) or snap‐frozen (for qPCR or Western blots). *Health refers to absence of epilepsy or other neurological disease, but does not mean that the subjects never were ill or did not use medication. Brackets indicate the most common time span.

Figure S2. At DIV 0 MAP2 and NeuN displayed complementary aspects of neurons. (A–C) MAP2 mainly stained neurites and a minority of the neuronal somata (arrows); asterisks indicate neuronal cell bodies that were NeuN+ but were negative for MAP2. (D–F) At DIV 33, the size of the neurons and their nuclei was severely diminished (asterisks; cf. Figure 1C,D) and the immunoreactivity for both NeuN and MAP2 had decreased. Merged signals are in B, E. Scale bars: B, E = 20 μm.

Figure S3. Western blot analysis of resected tissue at DIV 0 (d0) and DIV 28 (d28). Protein levels of NeuN and MAP2 confirmed the general neuronal degeneration (A, B). Vimentin, Nestin and HLA protein levels reflected the emergence of IVR cells (D–E). Unexpectedly, GFAP protein levels were lower in four patients (three shown here) after 4 weeks in vitro (F). Although Nestin and, to a lesser extent, Vimentin and GFAP, give surprisingly complex banding patterns (possibly because of extra aggregation and breakdown in neuronal tissue slices), the marked changes between DIV 0 and DIV 28 , as compared with the standard, β‐actin, are unequivocal. Asterisks indicate areas of migration for presumed protein bands. Arrow heads indicate the β‐actin bands. Arrows in B, E indicate lanes that were removed. E‐numbers correspond to patient numbers in the Supporting Information Table S1. Kaleidoscope markers were used in A, B, D–F and benchmark markers were used in C and the β‐actin standard of D.

Figure S4. In IVR and IVFL cells, Vimentin and Nestin were co‐localized with different glial markers. However, at DIV 0, neocortical Vimentin and Nestin were mainly expressed by endothelial cells (A, B). (A) Two Vimentin⁺ endothelial cells (arrows) along a VWF⁺ blood vessel. (B) Several Nestin⁺ endothelial cells arrows) at a VWF⁺ vascular junction. (C) At DIV 0, no co‐localization of Vimentin and HLA was observed, although these markers could be present in adjacent (peri)‐vascular structures. Arrows point at HLA⁺ cells. At DIV 28, many IVR cells had a Vimentin⁺ (D, E) or Nestin⁺ (F) cytoskeleton with a HLA⁺ surface. (D) Arrow head points at a HLA^‐/Vimentin⁺ cell, while arrows indicate HLA⁺/Vimentin⁺ cells. (G, H) Similarly, at DIV 0 (G) Iba1⁺ cells (arrows) did not co‐express Vimentin, while at DIV 28 (H), IVFL cells expressed both Vimentin and Iba1 (arrows) and IVR cells expressed only Vimentin (arrow head). (I) At DIV 0, Iba1 and Nestin did not co‐localize. (J) At DIV 28, Iba1 labeled IVFL cells (arrows), whereas Nestin labeled IVR cells (arrow heads). (K) At DIV 28, lysosomal marker CD68 and Vimentin co‐localized to IVFL cells. (L) Co‐localization of GFAP and Vimentin in resident hypertrophic astrocytes (asterisks) in the white matter of a noncultured slice. (M) Resident hypertrophic GFAP⁺/Vimentin⁺ astrocytes (asterisks) intermingled with Vimentin⁺ IVR cells (arrow heads), some of which partly contained GFAP (arrows). (N) At DIV 0 the GFAP⁺ signal in astrocytic cells and processes was completely separated from the Nestin⁺ background. (O) When GFAP⁺ IVR cells were present at DIV 28, they could also contain Nestin. (P, Q) Astrocytes in noncultured slices (P) hardly showed co‐localization of Vimentin (arrow heads) and S100ß (arrows), but at DIV 28 (Q) most IVR cells were both Vimentin⁺ and S100ß⁺ (arrows). (R) Some IVFL cells expressed both Vimentin and S100ß, whereas others were only Vimentin⁺. (S) At DIV 0 S100ß⁺ astrocytes (arrows) showed no overlap with Nestin⁺ blood vessels . (T) At DIV 28 IVR cells expressed both S100ß and Nestin. bv = blood vessel; g = green; r = red. Scale bars: B, H, I, J, L, M, N, Q, S, T = 50 μm; G, P, R = 25 μm; C, D, F, K, O = 20 μm and A, E = 10 μm.

Figure S5. Some pericytes can be visualized with Vimentin or Tubulin isoform ßI. (A) A Vimentin⁺ pericyte with a nucleus stained with ethidium homodimer (arrow head) at DIV 0. (B) At DIV 28, some pericytes showed protrusions (arrow) indicating that they may have become reactive, whereas others have not (yet) developed such morphology (arrow head). (C) Tubulin ßI⁺ perivascular cells (arrows) along Laminin⁺ blood vessels at DIV 0. Arrow heads indicate Tubulin ßI^‐ pericytes. (D) Tubulin ßI⁺ pericyte (arrow) along a Laminin⁺ blood vessel. Note that many thin Tubulin ßI⁺ fibrils run inside the basement membrane. (E) Pericytes labeled with Vimentin (arrow head) or Tubulin ßI (arrows) at DIV 0 suggesting that at this stage, they may constitute different subtypes. (F) Tubulin ßI staining at DIV 28 illustrates that reactive perivascular cells may also adopt an IVFL morphology. Nuclei were stained with ethidium homodimer. (G) At DIV 0 variable numbers of small nuclei could be labeled with Olig2. However, sometimes, their immunoreactivity was very faint. The arrow points at two nuclei that recently may have finished cell division. (H) At DIV 28 Olig2 could be present in activated pericytes (arrows), where it was mainly located in the cytoplasm. (I) At DIV 0 Olig2⁺ nuclei (arrows) and Vimentin⁺ blood vessels showed no signs of co‐localization. (J) At DIV 28 reactive pericytes (arrows) and IVR cells (asterisk) could contain Olig2 and Vimentin in their cytoplasm. Single Vimentin labeled vasculature associated cells were also commonly observed (arrow head). (K) Olig2 nuclei were not associated with Nestin⁺ vascular structures at DIV 0. (L) Reactive pericytes at DIV 28 showed cytoplasmic co‐localization of Olig2 and Nestin. (M) At DIV 0 some GFAP⁺ astrocytes had Olig2⁺ nuclei (arrows), but most Olig2⁺ nuclei were negative for GFAP. (N) At DIV 28 a subpopulation of Olig2⁺ IVR cells also contained GFAP (arrows). Arrow heads mark GFAP^‐ IVR cells. (O) The Olig2⁺ nuclei (arrow heads) did not belong to microglia cells (green, arrows) at DIV 0. (P) At DIV 28, some HLA⁺ IVR cells contained a Olig2⁺ nucleus (arrow head), but most did not (arrows). bv = blood vessel; g = green; r = red. Scale bars: I, J, L, M, N, O, P = 50 μm; B, G, H = 30 μm; C, E, K = 20 μm; A, D, F = 10 μm.

Figure S6. Before and during culture, many cells contained PCNA⁺ nuclei. PCNA staining suggests that many cells were either replicating or repairing DNA. At DIV 0, PCNA was present in (peri)‐vascular and satellite cells (arrows in A–C, E, G). (A, B) Some neuronal nuclei (asterisks) expressed PCNA. Nuclei were counter‐stained with ethidium homodimer (E). (B) PCNA⁺ satelite cells (arrows) were commonly observed. (C) Strikingly, the PCNA⁺ peri‐vascular and satellite cells were negative for GFAP at DIV 0. Arrows point at PCNA⁺ nuclei, << indicates a resident nonreactive astrocyte with a PCNA^‐ nucleus (C). (D) At DIV 28, PCNA⁺ nuclei were both found in GFAP^‐ cells (arrows) and in reactive GFAP⁺ cells (arrow heads). (E) PCNA and Vimentin double staining at DIV 0. Arrows point at PCNA⁺ nuclei. (F) At DIV 28, all Vimentin⁺ reactive cells had a PCNA⁺ nucleus. Arrow heads indicate PCNA⁺ nuclei in IVR Vimentin⁺ cells; arrows indicate apparently nonreactive nuclei; asterisks indicate IVFL cells with a PCNA⁺ nucleus (F). (G) Illustration of a Nestin⁺ blood vessel with associated PCNA⁺ nuclei at DIV 0. Arrows point at PCNA⁺ nuclei. (H) Nestin⁺ IVR cells with PCNA⁺ nuclei (arrow heads) at DIV 28. (I) The majority of PCNA⁺ nuclei were also immunoreactive for Olig2 at DIV 0. However, some nuclei were only Olig2⁺ (arrow) or only PCNA⁺ (arrow heads). (J) At DIV 28, some Olig2⁺ IVR cells contained a PCNA nucleus (arrow). Single nuclei were either only PCNA⁺ or PCNA⁺/Olig2⁺ (arrow heads). For A–H, no sudan black was applied. bv = blood vessel; g = green; r = red. Scale bars: C, E, F, I, J = 50 μm; H = 30 μm; A, G = 25 μm; B = 10 μm.

Figure S7. The time spent in vitro had an effect on the extent of the changes of immunocytochemical scores of HLA, NeuN and Vimentin. The Spearman test showed that the immunocytochemical staining of NeuN decreased (Rho = −0.64, adjusted P = 0.006, n = 24), of HLA increased (Rho = 0.64, adjusted P = 0.006, n = 23) and of Vimentin increased (rho = 0.64, adjusted P = 0.006, n = 24) with longer time in vitro. For the other immunostainings, ie, GFAP (rho = −0.08, adjusted P = 0.71, n = 23), Nestin (rho = 0.38, adjusted P = 0.10, n = 24) and PCNA (Rho = 0.15, adjusted P = 0.64, n = 18), such a realtion did not exist. Material of some patients had been harvested at different time points. To obtain a set of independent data for testing, we discarded duplicated observations, while maintaining the widest possible range of time points. Thus, each dot corresponds to an individual patient and its value is score(DIV T) − score(DIV 0). For consistency, the same set of observations was used to test the median changes of the immunocytochemical scores caused by the in vitro conditions per se. It may be noted that, although the scores of Nestin were significantly higher in cultured slices (see text), they did not correlate with the time spent in vitro. The blue lines were drawn using smoothing splines with 3 degrees of freedom solely to indicate the direction of change.

Table S1. Clinicopathological data of resected brain tissue from patients with temporal lobe epilepsy.

Table S2. Clinicopathological data of autopsy brain tissue from control subjects.

Table S3. List of applied primary antibodies.

Table S4. Primer data for Homo sapiens genes used in this study.

Table S5. Changes in gene expression during the in vitro conditions.