Neuronal volume of the hippocampal regions in ageing

Abstract

The hippocampal formation (HF) has an important role in different human capacities, such as memory processing and emotional expression. Both extensive changes and limited variations of its components can cause clinically expressed dysfunctions. Although there remains no effective treatment for diseases caused by pathological changes in this brain region, detection of these changes, even minimally, could allow us to develop early interventions and establish corrective measures. This study analysed the neuronal islands of layer II of the entorhinal cortex (EC), the neuronal clumps of the external principal layer of the presubiculum (PrS) and the dentate granule cells of the dentate gyrus (DG), which represent the prominent structural regions within the HF circuit. Subjects from two age groups (younger or older than 65 years) were studied and their neuronal size assessed by the point‐sampled intercepts stereological method. The quantitative ${\overline{v}}_v(soma)$ estimate was a volume of roughly 8,500 µm³ for EC layer II neurons, and DG granule neurons and presubicular neurons were five and 10 times smaller, respectively. The older age group showed a ${\overline{v}}_v(soma)$ increase of 2%, 18% and 28% with respect to the younger group in the PrS, DG and EC regions, respectively. None of these regions showed interhemispheric differences. This quantitative estimation is relevant because the observed variance in the ${\overline{v}}_v(soma)$ estimates suggests that biological variation is the main contributory factor, with intercepts and measurements having a smaller impact. Therefore, we suggest that age has a limited influence on neuronal volume variation in these HF regions, which needs to be compared with similar measurements in neurodegenerative disorders such as Alzheimer’s.

The ${\overline{v}}_v(soma)$ in the PrS and DG hippocampal regions was not significantly influenced by age, but the neuronal volume of the islands of layer II of the EC was greater in the older age group. The higher variance in the ${\overline{v}}_v(soma)$ estimates (with a coefficient of variation of 19%, 33%, and 18% for EC, PrS and DG, respectively) was mainly due to biological variation.

1. INTRODUCTION

The human temporal lobe (HTL) is a heterogeneous region consisting of cortical and subcortical areas which differ from cytoarchitectonic and phylogenetic points of view. In its medial aspect lies the hippocampal formation (HF), which includes the dentate gyrus (DG), the cornu ammonis fields (CA1, CA2 and CA3), the subiculum (Sb), presubiculum (PrS), parasubiculum (PaS) and the entorhinal cortex (EC) (Van Hoesen, 1995; Amaral and Lavenex, 2007; Insausti and Amaral, 2012; Delgado‐González et al., 2015). This region is essential for declarative memory processing, facilitating its long‐term storage through a wide network of connections (Figure 1). Axons of neurons in layer II of the EC project towards the molecular layer of the DG, the latter sending projections to the CA3 cells, which transfer the information to the CA1 neurons across CA2 and, in turn, through the Sb, PrS and PaS, send projections to deeper layers of the EC (Schultz and Engelhardt, 2014). Finally, the information output is passed along to various neocortical regions (Buzsaki, 1996; Muñoz and Insausti, 2005; Amaral and Lavenex, 2007).

FIGURE 1.

Coronal section of the HF corresponding to a case under study. All regions and main circuits are represented: DG, dentate gyrus; Cornu Ammonis CA3, CA2 and CA1; Sb, PrS, and PaS, subiculum, presubiculum and parasubiculum; EC, entorhinal cortex; LV, lateral ventricle; cs, collateral sulcus; hf, hippocampal fissure; f, fimbria

The information channelled through the HF can be modified or altered at any point along this route. As a consequence, neurons may respond with morphological changes, and neuronal changes can cause dysfunction in the hippocampal circuit (Braak and Braak, 1995, 1997). Both contingencies result in variations of neuronal shape and size configuration that can be revealed through quantitative studies. Neurodegenerative diseases lead not only to damage in the HF generating a reduced thickness of the region, but also, more significantly, to an important decrease in neuronal populations (Price et al., 2001; Artacho‐Pérula and Insausti, 2007). These neuronal changes could be present in normal ageing or in the initial phases of neurodegenerative processes, while showing little clinical symptomatology. However, the geometry of the hippocampal neurons is affected and this may be detected by precise quantitative analyses that have been widely used in the neurosciences (West and Gundersen, 1990; Rapp et al., 1996; West and Slomianka, 1998; Slomianka and West, 2005) in the successful assessment of different neurodegenerative processes.

The neuronal islands of layer II of the entorhinal cortex (EC), the neuronal clumps of layer II of the presubiculum (PrS) and the granule cells of the dentate gyrus (DG) are very prominent and distinct fields in the HF, configuring zones with a detectable cytoarchitecture. Our study aimed at analysing neuron body size using stereology in young (< 65 years) and older age groups (> 65 years). The study assessed 11 cases (six under 65 years, and five over 65 years), none of which had a record of neurological or psychiatric disorders. This simple stereological methodology could be of help in different studies, complementing other methods of analysis.

2. MATERIALS AND METHODS

2.1. Materials

This study used brain samples (n = 11) with no clinical record of neurological or psychiatric diseases, taken from the archives of the Human Neuroanatomy Laboratory (HNL) of the University of Castilla‐La Mancha. All case studies were approved by the Ethical Committee of the Albacete University Hospital. The material was divided into two groups according to age (LT65y, < 65 years, n = 6; and GT65y, > 65 years, n = 5). A summary of the demographic data is shown in Table 1.

TABLE 1.

Cases included in the study

Group	Case	Age (y)	Sex	Cause of death
LT65y	HN_1	23	Male	Drop and cardiorespiratory arrest
	HN_2	40	Male	–
	HN_3	40	Female	Diabetes mellitus
	HN_7	38	Male	Suicide
	HN_9	24	Male	Sudden death
	HN_11	58	Female	Coeliac disease, diffuse intestinal infiltration
GT65y	HN_4	90	Male	Lymphoma
	HN_5	83	Female	Colon cancer
	HN_6	84	Male	Pulmonary thromboembolism
	HN_8	82	Male	Heart disease, acute bacteraemia
	HN_10	78	Female	Sepsis

The postmortem period before brain fixation was, on average, 10 hr, ranging between 4 and 14 hr according to the Pathology Department protocol. The brain samples were fixed in 10% formalin for a variable time period (between 3 and 4 weeks), after which, upon arrival at the HNL, they were transferred to a 4% paraformaldehyde in phosphate buffer (PB, 0.1 M, pH 7.4) for at least 1 month. Subsequently, each brain hemisphere was cut into several slabs ~ 1 cm thick, orthogonal to the anterior‐posterior intercommissural line (AC‐PC) (Figure 2). The anterior and posterior surfaces of each slab were photographed prior to the cutting phase. Slabs were immersed in a cryoprotectant solution of 10% glycerol and 2% dimethylsulphoxide in phosphate buffer (pH 7.2) for 3 days, and then in 20% glycerol an additional 5 days. They were rostrocaudally arranged and each slab containing the regions of interest was coronally sliced (50 µm thick) using a sliding microtome (Micron® HM400) coupled to a freezing unit. Every 10th section (with a spacing of 500 µm) was Nissl‐stained with thionin for histological examination after several initial steps that included tissue defatting and dehydrating, before coverslipping with DPX mounting medium. The remaining sections were kept in tissue‐collecting solution for future studies. All sections were scanned to register the HF regions after cytoarchitectonic evaluation (Figure 3). All slab surfaces and the first or last full sections of the corresponding thionin‐stained slabs were analysed to evaluate shrinkage (measuring and comparing the x‐ and y‐axes and the total area of each slab and section) (Figure 4).

FIGURE 2.

Medial aspect of a brain hemisphere: The line drawn corresponds to the anterior‐posterior intercommissural line (AC‐PC) as well as the cutting plane for medial temporal lobe sections (1 cm thick). The slabs for both hemispheres were rostrocaudally arranged and digitalised. Hp, hippocampus; cs, collateral sulcus; rs, rhinal sulcus

FIGURE 3.

Medial temporal lobe section with HF regions indicated. On the right, systematic sampling of the different HF regions for selection of fields of vision (40× objective). Field sampling in islands of layer II of the EC, neuronal clumps of layer II of the PrS, and granule neurons of the DG, are showed in grey (enlarged fields are shown at the bottom)

FIGURE 4.

Top: Block #3, posterior surface, case HN3201, shows simple measurements (x‐ and y‐planes and total area of the slab) after calibration (ruler showed). Bottom: Same case and block (section 26 of 28) with very similar measurements after calibration (millimetre ruler is showed). imagej® software was used for measurements

2.2. Stereological study

The neuronal volume of the EC, PrS and DG regions was estimated on four histological preparations systematically selected along the HF rostrocaudal axis (on average, 1 in every 15 sections). Fields of vision were systematically sampled in each section (Figure 3). Neuronal volume measurements were made in a plane located approximately at the middle of the thickness section (the z‐axis equator). All fields of vision chosen for sampling were digitised (40× objective, using Olympus NewCAST®), with a final magnification of 1000×. The soma size of the neurons ${\overline{v}}_v(soma)$ (islands of layer II of the EC, neuronal clumps of layer II of the PrS, and the dentate granule cells of the DG) was estimated using the unbiased point‐sampled intercepts method (PSI; Gundersen and Jensen, 1985) to obtain the volume‐weighted mean neuronal volume (Figure 5). On each microscopic field of vision, a test system composed of regularly spaced test points associated with test lines was randomly superimposed. The application of this quantitative procedure to our samples was limited because the rostrocaudal sectional plane was preserved and therefore it was impossible to rotate sections in 3D. We maintained this plane of section in accordance with the common and routine rules for preparing brain tissue. In addition, the use of different rotation angles in 3D, especially when the coronal plane is lost, entails great difficulties in obtaining accurate identifications in all hippocampal regions.

FIGURE 5.

A sampled field of vision (EC) illustrating the volume‐weighted mean neuronal volume estimation method. Each test point hitting a neuronal soma (black points) is used to measure the length of the intercepts (l); if two or more points hit the same soma, the measurement is made for each test point. Finally, the stereological formula used for estimating the ${\overline{v}}_v(soma)$ is shown. The image to the right shows the NewCAST® equipment for stereological quantitation used, including an Olympus BX51 microscope, a DP72 camera, a set of motors and a microcator

2.3. Statistical analysis

A descriptive analysis (mean ± SD, SEM and coefficient of variation) was calculated for both groups and in both hemispheres. The contribution to the overall observed variance of ${\overline{v}}_v(soma)$ estimates from each level of sampling was obtained by the method of nested analysis of variance (Gundersen and Osterby, 1981). Data between brain hemispheres and aged groups were analysed using Student’s t‐test (paired and independent data, respectively). Also, the ${\overline{v}}_v(soma)$ was compared between EC, PrS and DG using ANOVA and the exact Friedman non‐parametric test. Data analyses were carried out using the SPSS/PC programme, v.23.

3. RESULTS

The results of the study showed a variation in the neuronal volume estimates ${\overline{v}}_v(soma)$ in the different HF regions analysed. The neuronal volume of the islands of layer II of the EC was roughly five times greater than that of the granule neurons of the DG. Also, granule cells of the DG are twice the neuronal volume of the granule cells of the PrS lamina principalis externa. The statistical analysis, as expected, indicated significant differences in neuronal size between EC layer II neurons and those in the PrS and DG regions (Figure 6, Table 2).

FIGURE 6.

Plot of the ${\overline{v}}_v(soma)$ for both groups (LT65y and GT65y) for each region of the HF (EC, PrS and DG). Data represent each case and the mean ± SD. (The y‐axis is at the same value to establish comparisons between hippocampal regions.) Statistics between LT65y and GT65y: *p < .05, NS, not significant

TABLE 2.

Results of the ${\overline{v}}_v(soma)$

Region	Group	Hemisphere	Neuronal volume (µm3)
EC	LT65y	Left	8,069 ± 2,015 (823)
		Right	6,641 ± 1,519 (620)
	GT65y	Left	9,330 ± 1,530 (684)
		Right	9,490 ± 983 (439)
PrS	LT65y	Left	1,021 ± 432 (176)
		Right	948 ± 276 (113)
	GT65y	Left	950 ± 180 (81)
		Right	1,050 ± 452 (202)
DG	LT65y	Left	1,438 ± 172 (70)
		Right	1,745 ± 196 (80)
	GT65y	Left	1,877 ± 436 (195)
		Right	1,894 ± 506 (226)

Data are mean ± SD (SEM).

The ${\overline{v}}_v(soma)$ values in the GT65y cases increased 28%, 2% and 18% with respect to the LT65y in the EC, PrS and DG regions, respectively. The only significant differences found were in the EC region (t = 2.41; p < .05) (Figure 6). In contrast, hemispheric differences had no effect on the volume‐weighted mean neuronal volume estimates in the three regions analysed (Figure 7).

FIGURE 7.

Plot of the ${\overline{v}}_v(soma)$ for brain hemispheres (HD, right; HI, left) for each HF region. Lines represent corresponding hemispheres for each analysed case (mean ± SD are indicated)

Figure 8 shows the contribution from different levels of sampling to the total observed variance associated with ${\overline{v}}_v(soma)$ estimates between the different regions of the HF and between the two age groups. The greatest contribution was attributed to differences between individuals (i.e. biological differences), with an average value of 78%. The contribution due to intercepts and measurements was negligible (3.8%). The variance contribution of sections increased from 1.5% to 5.7% to 8.9% for the EC, PrS and DG regions, whereas the fields of vision contribution fluctuated between 9.2% and 15.7%.

FIGURE 8.

Sources of variation in the volume‐weighted mean volume of the neuronal somas. Top: Contributions for each region of the HF. Bottom: Contributions for each age group. Data show items and percentages. In all cases, biological variation is the main contributor to the variance of ${\overline{v}}_v(soma)$

4. DISCUSSION

The relevance of the HF in neurodegenerative processes characterised by memory deficits is beyond doubt (Andersen et al., 2007; Hannula and Duff, 2017). Thus, the quantitative features of neurons in these allocortical and periallocortical regions and their association with ageing and with neurodegenerative lesions are extremely valuable and can be measured using objective quantitative tools. These tools can identify small changes in the structure of HF regions in the medial temporal lobe. In this regard, numerous studies have carried out morphometric and stereological estimations of the HF, revealing a non‐significant reduction in hippocampal volume with ageing (with only the 76‐ to 85‐year‐old group showing a marked decrease in volume) (Lupien et al., 2007). Meanwhile, Simic et al. (1997) found an inverse relationship between age and the number of neurons in CA1 and subiculum, but not in other subdivisions. Anatomically, the head and body of the hippocampus show significant volume reductions in older age, but this is not the case for the hippocampal tail (Rajah et al., 2010). Frisoni et al. (2010) have suggested that hippocampal atrophy is a sensitive marker of Alzheimer’s, whereas West et al. (1994), Rossler et al. (2002) and Pakkenberg et al. (2003) reported that neuron loss during ageing is a weak contributor to tissue atrophy. When studying the EC, Yanpei et al. (2019) observed that volume, thickness and surface area initially increased with age, reaching a peak at about 30–50 years of age, after which they started to decrease.

However, few studies have assessed neuronal volume in the hippocampus. Simic et al. (1997) reported a negative correlation of HF neuronal volume (without EC) with age in normal subjects, and a notable reduction (32%) in Alzheimer’s cases compared with controls. Additionally, Gemmell et al. (2012) have suggested that dementia produces hippocampal neuron atrophy in both cerebrovascular and neurodegenerative diseases. Lavenex et al. (2009) and Hosseini‐Sharifabad and Nyengaard (2007) studied DG granule cells in primates and rats, respectively; their data, obtained using the nucleator method, yielded values of around 700 µm³ (slightly lower in primates). In rhesus monkeys, Piguet et al. (2018) reported values of roughly 2,500 µm³ for soma size in EC layer II neurons. Also, Wegiel et al. (2014) described small variations in neuronal volume in the EC, as well as in other brain regions, associated with age. Wegiel et al. (2015) examined seven young humans (average age 22 years) and reported a slightly lower volume in EC layer II neurons than in our study; their data indicated values of 5,116 ± 1,960 µm³. These variations in neuronal size could be due to the age differences of the samples but also to the methodology used, as our study used volume‐weighted estimates that have a tendency to overestimate volume as compared with number‐weighted methods. In other words, in volume‐weighted estimates, larger particles (neurons) are more likely to be cut during tissue preparation than are those of smaller size and therefore have a higher probability of being sampled. Other studies have used the nucleator stereological method (Gundersen, 1988; Gundersen et al., 1988; Delgado‐González et al., 2015) for estimating neuronal volume in those cases where there was a previous selection of neurons, thus limiting the probability of sampling larger‐sized ones (number‐weighted); these studies used the disector technique to count the total number of particles in a precise volume (Sterio, 1984; Gundersen et al., 1988). Thus, volume‐weighted estimates ${\overline{v}}_v(soma)$ are larger than number‐weighted ones ${\overline{v}}_N(soma)$. On the other hand, our study reveals few changes in neuronal volume during ageing and only a slight increase in EC neurons in the older group. Several studies (Merrill et al., 2000; Wegiel et al., 2015) on primates and humans have reported that age has no influence in estimations of neuronal size. In contrast, the consequence of neurodegenerative processes such as Alzheimer’s is a notable (28%) decrease in neuronal volume (Artacho‐Pérula and Insausti, 2007).

Stereological studies can be influenced by tissue shrinkage or deformation when both number and object size are assessed. Many studies have dealt with this issue (Haug et al., 1981, 1984; Andersen and Gundersen, 1999; Dorph‐Petersen et al., 2001; West, 2013; Matenaers et al., 2018). Tissue shrinkage or deformation (in our study, brain tissue) is a difficult problem to resolve and predict. Delays in postmortem tissue collection, formalin fixation, cutting and embedding material, and dehydration or mounting methods all can introduce tissue changes, and other factors such as age could also have an effect. Also, it is very difficult to establish the response of each tissue component, as each one could have responded differently. Dorph‐Petersen et al. (2001) and West (2013) explored ways of reducing the impact of these changes. Both studies evaluated deformation in the x‐ and y‐planes, and z‐axis shrinkage. In our material, remarkable thickness shrinkage after histological processing was found (about 40%–60%, E. Artacho‐Pérula, unpubl. data). The use of glycol methacrylate embedding is a possible solution, as shrinkage is minimal (Gerrits et al., 1987; Gerrits and Horobin, 1996). However, the procedure is more complex than routine processing, including cutting (it requires more specific microtomes) and changes in staining techniques that are troublesome to use. Human material is scarce and, therefore, each sample can be used in different studies with notable differences in objectives, methodologies or treatment. Also, the material used in this study was obtained from Pathology Services, where they are required to maintain this processing and they sometimes remove small blocks to perform pathological diagnosis. All these facts affect the study of human material. Also, to control the presence (or absence) of changes due to other factors (age, sex or neurodegenerative diseases, among others), all samples must follow similar processing. Our study follows the suggestions of Dorph‐Petersen et al. (2001) and West (2013) for size estimation, taking measurements in the equator of the section plane (z‐axis), and using mounted frozen sections to minimise distortion.

The point‐sampled intercepts method to estimate particle volume has been widely used in biomedicine (Sorensen and Ottosen, 1991; Artacho‐Pérula and Roldán‐Villalobos, 1997; Leze et al., 2014) to differentiate between benign and malignant pathologies as well as for disorder classification and grading. Its unbiased nature, good reproducibility and low estimation requirement should be emphasised (Sorensen and Ottosen, 1991), although the method requires either isotropic uniform random (IUR) tissue sections or vertical sections. In our study, the ${\overline{v}}_v(soma)$ estimates are highly efficient and precise, as demonstrated by analysis of the total observed variance, where the most important contributory factor is biological variation (i.e. differences between specimens, near 80%), whereas the contribution of intercepts and measurements, with an average of 65 measurements (range 51–78), is negligible. Our study used sections with a non‐random orientation of neurons (at least in a 3D plane), which generates a source of error; however, as this limitation was taken into account for all estimates, we concluded that it would have affected data in all different groups equally. A general solution to this problem does not exist, although in the future it may be necessary to carry out a pilot study to assess the difference between estimates obtained from isotropic sections and those obtained from conventional sections (in our case, coronal sections).

In short, our study quantitatively determined ${\overline{v}}_v(soma)$ in neuronal groups within the HF, and our results indicated important differences between regions and a smaller variation between brain hemispheres. In addition, ${\overline{v}}_v(soma)$ in the PrS and DG hippocampal regions was not significantly influenced by age, but the neuronal volume of the islands of layer II of the EC was greater in the older age group. The higher variance in the ${\overline{v}}_v(soma)$ estimates (with a coefficient of variation of 19%, 33% and 18% for EC, PrS and DG, respectively) was mainly due to biological variation, while the methodology (number of sections, fields of vision and intercepts) had a smaller impact. These results will allow us to assess changes in neuronal size in neurodegenerative diseases such as Alzheimer’s.

Delgado‐González JC, de la Rosa‐Prieto C, Tarruella‐Hernández DL, et al. Neuronal volume of the hippocampal regions in ageing. J. Anat. 2020;237:301–310. 10.1111/joa.13189

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.