Stereologic estimation of hippocampal GluR2/3- and calretinin-immunoreactive hilar neurons (presumptive mossy cells) in two mouse models of temporal lobe epilepsy


Address correspondence to Thomas M. Freiman, Department of Neurosurgery, University Medical Center, Breisacher Strasse 64, D-79106 Freiburg, Germany. E-mail:


Purpose:  Hippocampal mossy cells receive dense innervation from dentate granule cells and, in turn, mossy cells innervate both granule cells and interneurons. Mossy cell loss is thought to trigger granule cell mossy fiber sprouting, which may affect granule cell excitability. The aim of this study was to quantify mossy cell loss in two animal models of temporal lobe epilepsy, and determine whether there exists a relationship between mossy cell loss, mossy fiber sprouting, and granule cell dispersion.

Methods:  Representative hippocampal sections from p35 knockout mice and mice with unilateral intrahippocampal kainate injection were immunolabeled for GluR2/3, two subunits of the amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptor and calretinin to identify mossy cells. Mossy fibers were immunostained against synaptoporin.

Key Findings:  p35 Knockout mice showed no hilar cell death, but moderate mossy fiber sprouting and granule cell dispersion. In the kainate-injected hippocampus, there was an 80% and 85% reduction of GluR2/3- and GluR2/3/calretinin-positive hilar neurons, respectively, and dense mossy fiber sprouting and significant granule cell dispersion. In the contralateral hippocampus there was a 52% loss of GluR2/3-, but only a 20% loss of GluR2/3-calretinin-immunoreactive presumptive mossy cells, and granule cell dispersion; no mossy fiber sprouting was observed.

Significance:  These results indicate a probable lack of causality between mossy cell death and mossy fiber sprouting.

In temporal lobe epilepsy (TLE) most patients show hippocampal sclerosis, known as Ammon’s horn sclerosis (AHS), characterized by variable cell loss in the hilus, CA1, and CA3 (Houser, 1990; Hermann et al., 1992; Armstrong, 1993). Several TLE animal models show similar patterns of neuronal loss (Ben-Ari, 1985; Liu et al., 1994; Leite et al., 1996; Bouilleret et al., 1999; Zhang et al., 2009). Among hilar cells, glutamatergic mossy cells (MCs) constitute about half of the neurons and are considered as one of the most vulnerable neuronal groups (Sloviter, 1987; Scharfman & Schwartzkroin, 1989). The other population of hilar neurons consists of inhibitory γ-aminobutyric acid (GABA)ergic interneurons (Seress & Ribak, 1983; Buckmaster & Jongen-Relo, 1999). It is suspected that MC death leads to loss of target neurons for mossy fibers (MFs), causing autoinnervation by sprouting back toward granule cells (Sutula et al., 1989; Blümcke et al., 2000). Furthermore, loss of MC innervation on inhibitory GABAergic interneurons is thought to cause reduced inhibition of granule cells (GCs; Sloviter, 1991; Sloviter et al., 2003). Finally, partial MC survival might be proconvulsive through direct excitation of GCs (Ratzliff et al., 2002, 2004).

Despite the importance of MCs, publications concerning a precise estimation of MC numbers in TLE are sparse, most likely because an unambiguous identification of MCs is difficult with common staining methods, despite their distinct microanatomy with “thorny excrescences” (Frotscher et al., 1991). Blümcke et al. (1999, 2000) and Seress et al. (2004, 2009), reported a loss of MC in human TLE; however, stereology was not applied. In rodent TLE models, MCs have been described as highly vulnerable (Sloviter, 1991; Magloczky & Freund, 1993; Scharfman et al., 2001). Buckmaster and Jongen-Relo (1999) and Jiao and Nadler (2007) reported a loss of MCs. However, they used intraperitoneal kainate or pilocarpine injection as tools for epilepsy induction. Both substances, however, lead to generalized brain damage with bilateral AHS and generalized seizures (Mello et al., 1993; Liu et al., 1994). In contrast, the human pathology of TLE shows unilateral AHS and nongeneralized, hippocampal seizures (Zentner et al., 1995).

To address this controversy we stereologically examined the survival of MCs in two TLE mouse models, namely p35 knockout mice (Wenzel et al., 2001; Patel et al., 2004) and mice with unilateral intrahippocampal kainate injection (Bouilleret et al., 1999; Riban et al., 2002) to show differences between a genetic model with generalized hippocampal maldevelopment and an interventional unilateral hippocampal lesion model. Hilar MCs were identified by immunolabeling for GluR2/3 (Fujise et al., 1998; Fujise & Kosaka, 1999; Sloviter et al., 2001, 2003), subunits of the amino-3-hydroxy-5-methyl-4-isoxazolepropionat (AMPA) receptor, predominately expressed in glutamatergic neurons (Leranth et al., 1996; Petralia et al., 1997) and by immunostaining for calretinin (CR), a marker of a subgroup of mouse MCs (Fujise et al., 1998). We estimated the number of GluR2/3+, GluR2/3+CR+, and CR+ hilar neurons by double immunohistochemistry, and correlated these results with the occurrence of granule cell dispersion (GCD) and MF sprouting, the latter identified by synaptoporin (SPO) immunolabeling.



All animals (naive control, p35−/−, and kainate-injected mice) were treated according to the guidelines of the European Community Council Directive (86/609/EEC), approved by the government representative (Regierungspraesidium Baden-Wuerttemberg, Freiburg, registration no.: 35/9185.81/G-08/72). Twenty adult male C57BL/6 mice (BioMed Center, Albert-Luwigs-University, Freiburg, Germany) (10–20 weeks of age) were included in this study. Ten of these mice were subjected to unilateral intrahippocampal KA injection. Ten p35−/− mice (10–20 weeks of age) were generated on a mixed Sv129/C57BL6 background (Chae et al., 1997) and inbred for at least five generations in our colony. Their genotypes were confirmed by polymerase chain reaction (PCR) analysis.

Unilateral intrahippocampal kainate injection

This animal model has been described in detail by Bouilleret et al. (1999) and Heinrich et al. (2006). In brief, C57BL/6 mice were anesthetized intraperitoneally (ketamine, 100 mg/kg, xylazine 5 mg/kg, atropine 0.1 mg/kg in 0.9% NaCl) and placed in a stereotaxic frame. Fifty nanoliter of a 20 mm kainate solution (Sigma-Aldrich, Munich, Germany) in 0.9% NaCl were injected through a steel cannula (outer diameter, 0.28 mm), connected to a 0.5 μl microsyringe (Hamilton, Bonaduz, Switzerland) via PE-20 tubing into the right dorsal hippocampus (coordinates: anteroposterior: −1.9 mm; mediolateral: −1.5 mm; dorsoventral: −1.9 mm). After recovery from anesthesia, animals experienced a status epilepticus, characterized by mild cloni of forelimbs and rotations. After 2 weeks chronic seizures became apparent with behavioral arrest, sometimes associated with facial movements, seldom with motor seizures (vocalization, mastication, forepaws cloni). Histology of the injected hippocampus showed cellular loss in CA1, CA3c, and hilus and massive GCD.

Immunohistochemistry for GluR2/3, calretinin, and synaptoporin

Animals were anesthetized with an overdose of ketamine (150 mg/kg, i.p.) and were transcardially perfused with 4% paraformaldehyde/0.1 m phosphate buffer (PB), pH 7.4 for 10 min. Mouse brains were postfixed in the same solution for 24 h at 4°C, followed by sectioning (50 μm, coronal plane) on a vibratome (Leica vibratome VT 1000 S, Bensheim, Germany). For immunostaining tissue sections were washed in 0.1 m PB, immersed for 30 min in 0.25% Triton in 0.1 m PB, preincubated in 10% normal horse serum/0.1 m PB, and finally incubated in antiserum for 24 h at 4°C in 0.1 m PB containing 1% normal horse serum, 0.1% NaN3, 0.05% Triton (Linaris, Wertheim, Germany). Antisera: GluR2/3-antibody, 1:300, rabbit polyclonal, former Chemicon no. AB1506; now Millipore, Billerica, MA, U.S.A. This antibody recognizes the nearly identical carboxy terminal sequences of GluR2 and GluR3 subunits (immunogen EGYNVYGIESVKI); no cross-reaction with GluR1 or GluR4 is described. Anti-calretinin (CR), 1:1,000, goat polyclonal, former Chemicon no. AB1550. Anti-synaptoporin (SPO), 1:400, rabbit polyclonal, Synaptic Systems no. 102002, Goettingen, Germany. The sections were incubated with appropriate secondary antibodies (Cy3-α-rabbit or Cy2-α-goat, 1:250; Jackson ImmunoResearch Laboratories, West Grove, PA, U.S.A.) for 3 h in the dark at room temperature followed by washing for 1.5 h in PB and counterstaining with 4′,6-diamidino-2-phenylindole (DAPI, Merck, Darmstadt, Germany) for 10 min. Finally sections were mounted on gelatin-coated slides and coverslipped (Immu-Mount; ThermoShandon, Dreieich, Germany). Penetration of immunohistochemistry amounted to 25 μm on both section sides (Leranth et al., 1996; Fujise & Kosaka, 1999) and was considered to be sufficient in initially 50-μm thick sections.

Quantitative analysis of granule cell layer diameter

The granule cell layer (GCL) width measurement was developed by Houser (1990); in short: five consecutive DAPI-stained 50-μm thick coronal hippocampal sections in the midseptal level were chosen, because the GCL forms its characteristic C-shaped structure (Fig. 1A). Ten to 20 consecutive measurements at 25-μm intervals were taken along the central 250–500 μm straight inferior GCL limb (Fig. 2C). The perpendicular distance from the hilar edge to the outer border of the most distal GC somata was measured (Zeiss AxioVision software, Axioplan2 microscope, Oberkochen, Germany). Group differences were analyzed using mean, 95% confidence interval (CI95) and Mann-Whitney U-test with significance level at p < 0.05.

Figure 1.

Stereologic methods. (A) The hippocampus was sliced along the septotemporal axis (arrow). (B) Cutting the hippocampus into 50-μm coronal sections resulted in about 50–60 slices. Every fifth slice was taken (section sampling fraction, ssf; 1, 6, 11, 16, 21, …) resulting in 10–12 slices/sections per animal. (C) The region of interest, the dentate gyrus, is indicated in gray in the background of the Δx*Δy intervals (100*100 μm grid) and defined in low power magnification with the StereoInvestigator software using the following three margins: On two sides with the border of granule cell layer (GCL)/inner molecular layer (IML), which is clearly distinguishable by the combined yellow GluR2/3+CR+ background (Fig. 2C–F). On the third side we have chosen the beginning of the GluR2/3-only immunoreactive background of the CA3 region. The neurons were counted within the optical disectors (red/green), which were located inside the Δx*Δy intervals. (D) Within the disector, defined by the area (A(f) = 40*40 μm) and the height (h = 10 μm), only neurons with distinct neuron morphology were counted. (E) The MCs were large, multipolar neurons, at least positive for GluR2/3; they were always easily distinguishable from other cells such as pyramidal, granule cells (GCs), or interneurons (Fig. 2A,B). Only those cells were counted, whose nucleus did not touch the red exclusion borders or was not within 2 μm from the top of the slice surface. The asterisk indicates the counted neurons and the cross the excluded neurons.

Figure 2.

Hilar neurons in naive control, p35−/−, and kainate-injected mice analyzed with fluorescent DAPI labeling (blue), fluorescent immunohistochemistry against GluR2/3 (red), and calretinin (green). (A) High power magnification of a hilar GluR2/3-only immunoreactive multipolar large hilar neuron (GluR2/3+, asterisk), presumably an MC and a calretinin-only positive neuron (CR+, arrowhead) presumably a large hilar interneuron. (B) Four GluR2/3- and calretinin-immunoreactive hilar neurons (GluR2/3+CR+, arrow), and two GluR2/3+ hilar neurons (asterisks). Both cell types were presumably MCs. (C) In control animals (CTRL), GluR2/3+ and GluR2/3+CR+ hilar neurons were visible in the hilus (H). The cell bodies in the GCL showed red immunolabeling against GluR2/3 and blue DAPI nuclear staining. The IML appeared as a yellow band because of intense double staining of GluR2/3 (red) and CR (green). The outer molecular layer was only weakly immunopositive for GluR2/3 and did not show any CR immunoreactivity. The CA3 region (asterisk) displayed an increased GluR2/3-immunoreactivity compared to the hilus. (D) In p35−/− mice, the GCL showed less dense layering visualized by DAPI staining. The IML showed a mild dissolution toward the red GluR2/3+ background of the GCL. No hilar cell loss was visible. (E) The dentate gyrus of the kainate-injected hippocampus (KA inj) showed massive granule cell dispersion (GCD). The usually sharply blue and red stained GCL was wider and less intensely stained. In the hilus nearly all neurons were absent. The GluR2/3 immunoreactivity of the IML and outer molecular layer was weaker than in control mice, and the borders of these structures were completely displaced through the GCD. The supragranular band, which appeared clearly yellow in the control- and p35−/− mouse, was green, indicating a weaker immunoreactivity against GluR2/3, probably through lost innervation of deceased GluR2/3+ hilar neurons. (F) In the dentate gyrus of the noninjected hippocampus (KA non-inj), the red GluR2/3+ of the GCL was lower than in control mice, making the DAPI stain more prominent. There was a major loss of GluR2/3+ and GluR2/3+CR+ hilar neurons, but there was an increase of CR+ hilar neurons (arrowhead) in the subgranular zone. Scale bars A, B 10 μm, CF 100 μm.

Stereologic analysis

Stereologic fractionator and optical disector were used according to Gundersen et al. (1988) applied with StereoInvestigator software (Micro Bright Field Bioscience, Williston, VT, U.S.A.); Leica DMRB microscope, Wetzlar, Germany; Heidenhain stage controller, Traunreut, Germany; and DXC950P3-CCD camera, Sony Corporation, Tokyo, Japan. The total number of neurons (N) was estimated using the equation: N = Q*(1/ssf)*(1/asf)*(1/hsf), Q was the number of neurons counted, ssf the section sampling fraction (every fifth slice), asf the area sampling fraction (asf = A(f)/(Δx*Δy)), A(f) the dissector area (40*40 μm), Δx*Δy the counting grid intervals (100*100 μm), hsf the height sampling fraction (hsf = h/T), h the dissector height (10 μm), and T the measured slice thickness after staining-induced slice shrinkage (Fig. 1). The intraanimal variance was indicated by the coefficient of error (CE), which was calculated as the standard error of the mean (SEM) of repeated estimates to the mean of the estimates (CE = SEM/mean N, Table 1). Sampling was considered sufficient with CE value below coefficient of variation (CV = SD/mean N, Table 1), where SD is the standard deviation of the estimates (West & Gundersen, 1990). Group differences were analyzed using mean, CI95 and U-test with p < 0.05.

Table 1.   Stereologic parameters for the estimation of total hilar neuron numbers
ParameterWild-typep35−/−KA injKA non-inj
  1. Mouse models were listed as follows: control (CTRL)-, p35 knockout (p35−/−) mouse, unilateral intrahippocampal kainate-injected mouse, ipsilateral injected hippocampus (KA inj), contralateral noninjected hippocampus (KA non-inj). The parameters of the disector and fractionator were followed by the counted disectors and the cell types: GluR2/3+ hilar neurons (presumptive MCs without immunoreactivity against CR), GluR2/3+CR+ (presumptive MCs with double immunoreactivity against GluR2/3 and CR) and CR+ hilar neurons (presumptive inhibitory interneurons in control and p35−/− mice but in KA non-inj additionally neurons of the subgranular zone).

  2. CI95, 95% confidence intervals; SEM, standard error of the mean; SD, standard deviation; CE, coefficient of error; CV, coefficient of variation. The level of precision of an estimate is indicated by CE = SEM/mean N and is indicative of the intraanimal variance. The sampling is optimized when the CE value is below the coefficient of variation (CV = SD/mean N).

Animal N10101010
Counting frame area (μm)40*4040*4040*4040*40
Counting frame height (μm)10101010
Guard zone (μm)2222
Grid point area (μm)100*100100*100100*100100*100
Disectors mean numbers101.290.2220.5180.5
Number of GluR2/3+ hilar neurons3,7393,7547701,798
 CI953,030, 4,4473,471, 4,037609, 9321,583, 2,013
Number of GluR2/3+CR+ hilar neurons1,6011,5412381,288
 CI951,022, 2,1791,195, 1,888140, 3351,022, 1,554
Number of CR+ hilar neurons19026171483
 CI9586, 294165, 35724, 143221, 744

The intrahilar cell types were differentiated as follows: (1) GluR2/3+ hilar neuron (Fig. 2A): GluR2/3-immunoreactivity, large multipolar cell body, and large nucleus; (2) GluR2/3+CR+ hilar neuron (Fig. 2B): additional CR-immunoreactivity; (3) CR+ hilar interneuron (Fig. 2A): CR-immunoreactivity, GluR2/3-negative, large cell body, often not multipolar. We did not count GC (GluR2/3-immunoreactivity, CR-negative, small mono- or bipolar cell body, small nucleus, usually located inside the GCL, rarely located at the subgranular margin of the hilus) and triangle-shaped CA3 pyramidal cells that were located outside the marked hilus.


Hilar morphology

Control mice

We assumed that GluR2/3+ and GluR2/3+CR+ hilar neurons were MCs (Leranth et al., 1996; Sloviter et al., 2001, 2003; Bender et al., 2003), without and with additional immunoreactivity against CR (Fujise et al., 1998; Fujise & Kosaka, 1999) and that CR+ hilar neurons were inhibitory interneurons. The GluR2/3+ background immunoreactivity was low in the hilus, getting more intense toward CA3 (Fig. 2C). The GCL appeared as a dense red-blue dotted layer, consisting of intensely GluR2/3+-labeled GCs and DAPI-blue labeled nuclei. The inner molecular layer (IML) appeared intensely yellow caused by the overlay of red GluR2/3+ and green CR+, indicating the strong innervation of GC dendrites by MCs. The outer molecular layer showed red GluR2/3+ exclusively. The SPO+ MFs were confined to the hilus and showed no sprouting into the GCL (Fig. 3A), whose mean diameter amounted to 67 μm (CI95 = 63, 70 μm, n = 10, Fig. 4).

Figure 3.

Mossy fibers (MFs) labeled by synaptoporin (SPO) immunohistochemistry. (A) In control mice SPO+ fibers were seen as pearl chain-like structures (arrow) in the hilus (H) only. There is a slight background of SPO-immunoreactivity (asterisk) in the IML. (B) In p35−/− mice, SPO+ fibers (arrow) sprouted through the slightly dispersed GCL toward the IML; their synaptic contacts were visible as spots (arrowhead) and could be distinguished from the elevated SPO+ background (asterisk). (C) In the kainate-injected hippocampus, a massive sprouting of SPO+ fibers (arrow) was visible, extending into the IML (arrowhead) through the massively dispersed GCL. (D) On the noninjected hippocampus, no SPO+ fiber (arrow) sprouting but only GCD and an elevated SPO+ background in the IML (asterisk) was observed. Scale bars 50 μm.

Figure 4.

Measurement of GCL diameter. Error bars were shown as 95% confidence intervals (CI95); asterisks in the graph indicated significant differences. The mean GCL diameter amounted to 67 μm in control mice (CTRL), 75 μm in p35−/− mice, and 87 μm in the noninjected hippocampus of the kainate-injected mice (KA non-inj). All differences were highly significant. The GCD in the injected hippocampus amounted to over 200 μm. Because this result is trivially significant and for better clarity of the smaller values of the other diameters this bar is not shown.

p35 Knockout mice

In contrast to C57BL/6 control mice, all hippocampal cell layers showed a slight heterogeneity and dispersion in p35 knockout mice. The dispersion was significant and amounted to 75 μm (CI95 = [70, 80 μm], n = 10, p = 0.05, Fig. 4). The IML was not as clearly demarcated and showed some dissolution with the GluR2/3+ background of the GCL (Fig. 2D). The immunohistochemistry for SPO revealed MF sprouting through the GCL into the IML (Fig. 3B).

Mice with unilateral intrahippocampal kainate injection

In the hilus of the injected hippocampus, the number of GluR2/3+ and GluR2/3+CR+ hilar neurons was reduced significantly, as was the number of CR+ interneurons. The GCL was in the septal parts of the hippocampus dispersed, amounting to 200 μm (Fig. 2E). As a consequence of the massive GCD the hilus was smaller in size and the IML border harder to identify. The IML showed a loss of GluR2/3+ immunoreactivity, most likely due to degeneration of GluR2/3+ fibers originating from deceased MCs. The outer molecular layer was also less densely stained against GluR2/3. The SPO+ immunolabeling revealed a massive sprouting of MFs into the GCL and IML (Fig. 3C).

In the hilus of the contralateral, noninjected hippocampus, a moderate loss of GluR2/3+ and GluR2/3+CR+ neurons was observed (Fig. 2F). In contrast, the number of CR+ neurons was increased. These CR+ neurons were predominately located close to the subgranular border. The GCL showed reduced GluR2/3-immunostaining. No MF sprouting through the GCL was found (Fig. 3D). Significant GCD with a mean diameter of 87 μm (CI95 = 84, 91 μm, n = 10, p < 0.01, Fig. 4) was observed.

Stereologic estimations

In C57BL/6 control mice we estimated 3,739 GluR2/3+ hilar neurons (Fig. 5, Table 1), of which 43% were GluR2/3+CR+ (in absolute numbers: 1,601). In addition, 190 CR+ hilar interneurons were estimated. There were no significant differences in cell numbers between p35−/− mice and control mice. In p35−/− mice we estimated 3,754 GluR2/3+ hilar neurons, among which 41% were GluR2/3+CR+ (in absolute numbers: 1,541) as well as 261 CR+ interneurons. In kainate-injected mice, only 770 GluR2/3+ hilar neurons were estimated, amounting to an 80% (p < 0.001) loss of these cells in the injected hippocampus when compared to control mice. Thirty-one percent of the remaining MCs were GluR2/3+CR+ (in absolute numbers 238 neurons), which means that GluR2/3+CR+ hilar neurons sustained an 85% loss (p < 0.001). The numbers of CR+ interneurons were reduced by 63% (p < 0.01); only 71 were estimated. In the contralateral hippocampus, there was a 52% loss of GluR2/3+ hilar neurons (in absolute numbers: 1,792, p < 0.001). The number of GluR2/3+CR+ hilar neurons was reduced by only 20%, shifting the ratio of these GluR2/3+CR+ cells from 41% in control mice to 72% in the noninjected hippocampus (in absolute numbers: 1,288, p < 0.001). The number of CR+ interneurons was increased to 254% (in absolute numbers: 483 CR+ interneurons, p < 0.01). For all stereologically estimated cells the CE value was below the CV value (Table 1).

Figure 5.

Total number of hilar cells in hippocampi of control (CTRL), p35−/−, and mice with unilateral intrahippocampal kainate injection, in the injected hippocampus (KA inj), and in the contralateral non-injected hippocampus (KA non-inj). Color codes indicate GluR2/3+ hilar neurons in red, GluR2/3+CR+ hilar neurons in yellow, and CR+ hilar neurons in green. Error bars were shown as 95% confidence intervals (CI95, Table 1); asterisks indicate significant differences in neuron numbers. No difference was observed between control- and p35−/− mice (n.s.). On the kainate-injected side, there was a significant loss of all three estimated neuron types compared to control mice. On the noninjected side there was also a significant reduction of GluR2/3+ and GluR2/3+CR+ hilar neurons compared to control mice as well as compared to the injected side. There was a 2.5-fold increase of CR+ hilar neurons presumably caused by GCs with the immaturity marker calretinin.

Spatial distribution of hilar neurons along the septotemporal hippocampal axis

In general, more neurons were found in sections from the ventral hippocampus due to the enlarged area of the hilus. In the septal hippocampus of control mice, almost all hilar neurons were GluR2/3+ and only a few GluR2/3+CR+ neurons were counted (Fig. 6). At the middle level about one third of GluR2/3+ hilar neurons showed co-labeling with CR+ (e.g., GluR2/3+CR+). In the temporal hippocampus almost all GluR2/3+ hilar neurons were also GluR2/3+CR+. The number of CR+ interneurons was similar in each section throughout the entire control hippocampus. In p35−/− mice, the neuron numbers per hippocampal section was similar; however, in the temporal tail not all but only two-thirds of the GluR2/3+ hilar neurons were GluR2/3+CR+. In the kainate-injected hippocampi, GluR2/3+ and GluR2/3+CR+ hilar neurons were equally reduced throughout the entire hippocampus. On the contralateral side, GluR2/3+ and GluR2/3+CR+ hilar neurons were reduced in proportion mostly in the septal part. However, because most of the neurons were located in the temporal part, the loss of neurons in absolute numbers was the highest there.

Figure 6.

Spatial distribution of GluR2/3+ (red), GluR2/3+CR+ (yellow), and CR+ (green) hilar neurons along the septotemporal hippocampal axis. In the septal portion (section 1–16) of the control hippocampus nearly all presumptive MCs were GluR2/3+, whereas in the temporal portion (sections 36–46/51) nearly all MCs were GluR2/3+CR+. In p35−/− mice, there was no significant difference concerning MC distribution. In the kainate-injected hippocampi, the numbers of GluR2/3+ and GluR2/3+CR+ hilar neurons were reduced equally. On the contralateral side, cell death affected GluR2/3+ more severely than GluR2/3+CR+ hilar neurons. The loss of GluR2/3+ and GluR2/3+CR+ neuronal population was most severe in the septal portions, when evaluated as percentage of cell loss. However, because in the temporal hippocampus neuron numbers are higher, the cell loss in absolute numbers was the highest there.


Methodologic considerations

The most reliable way to estimate neuron numbers is stereologic sampling with fractionator and optical dissector (Gundersen et al., 1988; West & Gundersen, 1990; West, 1999). For the stereologic estimation of MCs it was necessary to label all MCs. Most of the previously published data were based on indirect evidence of MCs, such as calculating hilar cell loss from Nissl stains or studying changes in the number of cells that lack expression of glutamate decarboxylase (GAD65/67) (Frotscher, 1992; Freund & Magloczky, 1993; Soriano & Frotscher, 1994; Blasco-Ibanez & Freund, 1997; Buckmaster & Jongen-Relo, 1999). We chose fluorescent GluR2/3-immunolabeling, since it has been shown that GluR2/3 is expressed by all telencephalic glutamatergic neurons, whereas inhibitory interneurons express it to a much lesser extent (Leranth et al., 1996), and nearly all glutamatergic neurons in the hilus are MCs. Sloviter et al. (2001) reported that <1% of the GluR2/3+ hilar neurons are immunoreactive for GABA. Bender et al. (2003) demonstrated that GluR2/3+ neurons, presumably MCs, and GAD67 mRNA positive neurons, presumably inhibitory interneurons, accounted for 98.4% of all hilar neurons. Fujise et al. (1998) (Fujise & Kosaka, 1999) demonstrated that almost all GluR2/3+ hilar neurons showed morphologic characteristics of MCs, using Golgi impregnations, Lucifer yellow injections, and retrograde labeling. These studies identified GluR2/3 as a reliable, albeit not perfect, marker for hilar MCs. By defining the dentate gyrus as our stereologic region of interest, using the inner border of the GCL, MCs were clearly distinguishable from other GluR2/3+ neurons like GCs because of their larger diameter and characteristic morphology, even if these are positioned in the hilus as described by Jiao and Nadler (2007).

Supragranular MF sprouting has been primarily reported in rats with experimentally induced AHS (Sutula et al., 1988) and in TLE patients (Sutula et al., 1989) using Timm staining for detection. Because we aimed at identification of MCs and MFs in the same animal, we used immunolabeling for synaptoporin, an established marker of MFs (Grabs et al., 1994; Singec et al., 2002; Jinno & Kosaka, 2003).

No mossy cell death but mossy fiber sprouting in p35 knockout mice and bilateral mossy cell death in unilaterally kainate-injected mice

p35−/− Mice show bilateral, generalized, and limbic seizures on EEG recordings; they exhibit GCD and MF sprouting (Wenzel et al., 2001; Patel et al., 2004). However, MC loss was not investigated. Patel et al. (2004) demonstrated increased excitability of GCs, using biocytin-labeling and antidromic stimulation of MFs, likely through MF sprouting. Therefore, we expected a mild MC loss, which would trigger MF sprouting. However, we found no GluR2/3+ neuron loss, despite MF sprouting and GCD.

In the kainate-injected hippocampus we observed an 80% loss of GluR2/3+ hilar neurons. Surprisingly, we also found a 52% GluR2/3+ neuron reduction in the contralateral, noninjected hippocampus. Until now only marginal histologic changes have been described in the contralateral hippocampus, e.g., reduced calbindin-immunoreactivity in CA1 pyramidal cells or increased cholecystokinin- and neuropeptide Y-immunoreactivity in the dentate gyrus (Arabadzisz et al., 2005; Meier et al., 2007). To our knowledge, this is the first description of a significant contralateral hilar cell loss in this model. Because contralateral hilar cell death was not described previously, we critically considered a potential effect of the injected kainate to the contralateral hippocampus through direct injection or transventricular spread. However, this seemed unlikely, since the lesion channel of the cannula did not touch the ventricle or contralateral hippocampus. Alternatively, loss of GluR2/3+ and CR+ cells could be explained by reduced expression of GluR2/3 or CR, caused by kainate toxicity or seizures. However, hilar cell loss was clearly visible in DAPI stains. A reduction of protein expression as cause of the observed loss of GluR2/3 and CR immunoreactivity seems unlikely, since downregulation of GluR2/3 to an extent undetectable by immunocytochemistry would probably kill neurons (Hollmann et al., 1991; Yin & Weiss, 1995). An explanation for contralateral cell loss could be excitotoxicity from commissural MC projections (Frotscher et al., 1991; Blasco-Ibanez & Freund, 1997). Magloczky and Freund (1993, 1995) demonstrated that MCs were decreased ipsilaterally and contralaterally after injecting kainate into the CA3 region.

Unilateral mossy fiber sprouting despite bilateral cell death of GluR2/3+ hilar neurons in unilaterally kainate-injected mice

Discovering a significant loss of GluR2/3+ hilar neurons (presumptive MCs) in the contralateral hippocampus of kainate-injected mice without MF sprouting was surprising, since it has been postulated that MF sprouting follows MC death (Blümcke et al., 2000). Bouilleret et al. (1999) observed a very discrete MF sprouting in the contralateral hippocampus 30 days after kainate injection in one single animal (of four). After 120 days more sprouting was visible in two animals, but it was still less pronounced than on the injected side. We examined sprouting 28 days after kainate injection, which could possibly explain our different results. However, Bouilleret et al. (1999) did not investigate MC death and, therefore, did not correlate MC death with MF sprouting. Jiao and Nadler (2007) reported a 95% loss of GluR2/3+ hilar neurons in rats treated with intraperitoneal pilocarpine injection. This cell death was reduced to 74% by phenobarbital administration, as was MF sprouting. However, systemic pilocarpine injection is often lethal and causes severe cellular damage throughout the brain, including both hippocampi (Mello et al., 1993) and is characterized by severe generalized seizures in the chronic stage (Liu et al., 1994). In contrast, intrahippocampal kainate injection in mice causes only a mild initial status epilepticus; in the chronic stage, animals show focal limbic but no generalized seizures (Bouilleret et al., 1999; Riban et al., 2002). Therefore, the invasiveness of the pilocarpine TLE model could have been causative for the more severe neuron loss and more pronounced MF sprouting.

In the present study, MFs were visualized by immunolabeling for SPO, whereas Bouilleret et al. (1999) and Jiao and Nadler (2007) used Timm staining. Interestingly, Jinde et al. (2008) described a MC deletion mouse, which exhibits 90% MC loss without MF sprouting expressing no obvious epilepsy-like discharges in local field potential recordings. In our stereologic study, we observed no correlation of the two phenomena, namely death of GluR2/3+ hilar neurons (presumptive MCs) and SPO+ fiber sprouting (presumptive MFs). In p35−/− mice we observed no GluR2/3+ hilar neuronal death but sprouting of SPO+ MFs. In the contralateral hippocampus of kainate-injected mice we found a 50% loss GluR2/3+ neurons but no sprouting of SPO+ MFs. Nevertheless, a causative explanation of whether MC death is obligatory followed by MF sprouting is still pending.

It is still unclear whether MF sprouting is causative for limbic seizures in TLE or only a concomitant phenomenon. Bouilleret et al. (1999) and Meier et al. (2007) reported that the injected hippocampus, but not the contralateral hippocampus, was the source of limbic seizures. These results might be explained by our finding, that SPO+ MFs, which sprouted in the injected but not the contralateral hippocampus, may have a proconvulsive role as previously suggested by other authors (Sutula et al., 1989; Cavazos et al., 1994; Nadler, 2003). In contrast, Elmer et al. (1997) observed that kindled rats developed seizures without MF sprouting. Gorter et al. (2001) showed in a status epilepticus model that spontaneous seizures occur only in hippocampi, which show hilar cell death and MF sprouting, whereas very mild MF sprouting alone was observed in the animals without spontaneous seizures.

Calretinin and GluR2/3 positive hilar neurons were preserved only in the noninjected hippocampus

Due to their buffering properties, Ca2+-binding proteins affect intracellular Ca2+ homeostasis and to some of the family members, including CR, a neuroprotective role during excitotoxicity has been attributed (Lukas & Jones, 1994; D’Orlando et al., 2001, 2002). The CR immunoreactivity of MCs is species-dependent. Whereas human and rat MCs do not express CR, mice, gerbil, and hamster MCs contain CR (Seress et al., 2008). We demonstrated that the GluR2/3+CR+ subpopulation of GluR2/3+ hilar neurons was located predominantly in the temporal hippocampus, a result previously shown by Fujise et al. (1998) and Fujise and Kosaka (1999) using semiquantitative methods. Kotti et al. (1996) observed that CR+ MCs in gerbils were more resistant to excitotoxic neuronal death than CR-negative MCs in rats. Blasco-Ibanez and Freund (1997) found in mice that CR+ MCs in the temporal part receive more input via contralateral commissural projections than MCs in the septal part. They postulated that this projection leads to an increased susceptibility for pathologic neuronal loss. In our study MC loss affected both subgroups (GluR2/3+ and GluR2/3+CR+ hilar neurons) on the injected side equally; however, on the noninjected side GluR2/3+CR+ hilar neurons were preserved more often. When we evaluated the cell loss of GluR2/3+ and GluR2/3+CR+ hilar neurons along the hippocampal axis, the relative cell loss was highest in the septal parts of the noninjected hippocampus; however, because most of the neurons were located in the temporal part, the loss of neurons in absolute numbers was the highest there.

Calretinin-positive hilar neurons were increased in number in the noninjected hippocampus

In our study we report an elevated number of CR+ hilar neurons in the subgranular zone in the noninjected hippocampus in addition to a reduction of GluR2/3+ hilar neurons. It was shown by Kralic et al. (2005) in the same animal model that neurons in the contralateral subgranular zone showed an uptake of 5-bromo-2-deoxyuridine (BrdU), which was considered as a sign for neurogenesis. In other TLE models, for example, systemic pilocarpine administration, neurons in the subgranular zone showed a transient increase of CR and the polysialylated isoform of the neural cell adhesion molecule (PSA-NCAM)–immunoreactivity, which was interpreted as reexpression of antigenic markers of immature neurons by GCs (Dominguez et al., 2003). This expression of CR was observed after the loss of MCs and remained detectable for a month, until sprouting of MFs substituted the lost projection of the MCs (Dominguez et al., 2003; Marques-Mari et al., 2007). However, our study did not address the question of neurogenesis. We found this transient increase of CR+ hilar neurons, which lasted 4 weeks after kainate injection, but no MF sprouting in the contralateral hippocampus.


The authors thank Friederike Moos for excellent technical assistance and Ute Haeussler, Ph.D. and Martin Mueller, Ph.D. for help with kainate injections and stereologic methods. This study was supported by the German Research Council (Transregional Collaborative Research Center TR-3 D7 to T.M.F. and C.A.H. and Grant BO 1806/2-1 to H.H.B.) and the Research Trust of the German Society for Neurosurgery (annual grant 2010 to T.M.F.).


None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.