Materials and Methods

Animals
A total of 63 adult male Sprague-Dawley rats aged 3 months (250-300 g) were obtained from the Animal Laboratory Center of the University. The animal trial was approved by the Ethics Committee of the University (approval No. 94-7624) and all manipulations were done under the regular instructions of the Animal Ethics Committee. First, the rats were arbitrarily divided into nine groups each containing 7 animals. The animals were then kept in a room with normal conditions and temperature (22-24 °C) and had free access to water and food. Through daily gavage feeding, the groups received the following:
Group I: (Control) 1.5 ml of distilled water
Group II: Curcumin (100 mg/kg/day) dissolved in olive oil
Group III: 1.5 ml of olive oil
Group IV: Distilled water
Group V: Curcumin
Group VI: Olive oil
Groups VII: Distilled water
Group VIII: Curcumin
Group IX: Olive oil
It should also be noted that groups IV, V and VI experienced SD and the animals in the last three groups were placed on cages with grid floors. After a period of 21 days for the SD protocol, the rats underwent the eight-arm radial maze test and were then sacrificed.

Sleep deprivation procedure
The setup used in this study was a modified multi-platform container made of plastic that have been explained earlier (Kamali et al., 2016[12]). The container was then filled with water. When a rat entered the sleep, it touched the water by tilting its head downwards or fell into the water due to muscle atonia. The SD period continued for 21 days during which the rats were kept in the multiple-platform box for 18 hours from 6:00 P.M. to 12:00 noon on the subsequent day. During the next 6 hours, the rats were permitted to fall asleep (12:00 noon to 6:00 pm). The light schedule of the SD-boxes, on from 6:00 A.M. until 6:00 P.M, was in accordance with the light/dark cycle similar to the non-SD animals (Wang et al., 2014[27]). In the grid groups (VII, VIII, IX), the rats were located on wire-mesh grids in the deprivation container. The grid was made of stainless steel. This was incorporated as a control group for SD. Finally, the rats in these groups also underwent the eight-arm radial maze test.

Assessment of behavior in the eight-arm radial maze
To assess spatial learning and memory, the eight-arm radial maze test was carried out according to the previous study (Karkada et al., 2012[13]). The first stage (also known as the adaptation phase) in which the rats were allowed to discover the baited arms of the maze for 10 min lasted for two consecutive days. In the second phase (also known as the acquisition phase), the rats were given two five-minute trials per day until they reached the learning standards which was defined as reaching 80 % correct choices; i.e., at least four correct entries out of five. This period lasted for eight to fifteen days. The third stage (also known as the retention period) started ten days after the acquisition phase. The average of the two trials and the percentage of correct choices and errors were used for analysis.

Tissue preparation
The animals were anesthetized and their brains were removed. Then, the right hemispheres were processed, coronally sectioned (26 μm thick) and stained using Giemsa and were used to estimate the number of neurons and glial cells as well as the volume of the mPFC. Besides, the slabs of the left hemispheres were processed and stained with Golgi impregnation method (Figure 1(Fig. 1)).

Estimation of the volume
Using a stereomicroscope, the live image of each brain section was evaluated according to the rat brain atlas. The volume "V(mPFC)" was estimated using the Cavalieri method (Kristiansen and Nyengaard, 2012[14]). The sum of the area of the favored structure (ΣA) (mPFC, ACC, PLC, and ILC) was estimated using the software designed at the University. The distance between the sampled sections was defined as “d”. Finally, the volume of each structure was estimated by the following formula:

Estimation of the number
A videomicroscopy system along with an oil immersion lens (40×, numerical aperture: 1.3) were used in order to estimate the total number of the mPFC neurons and glial cells according to the optical disector method. To analyze the appropriate guard zone and the height of the disector (h), Z-axis distribution of the nuclei was plotted (Figure 1(Fig. 1)). The total number of the neurons was estimated by multiplying the numerical density”
where "ΣQ-" is the number of the nuclei coming into focus during the scanning of “h” (the height of the disector). "ΣP" is the total number of counting frames in all fields, "a/f" is the frame area, “t” is the mean section thickness measured in every sampled field using the microcator (20 µm on average), and “BA” is the block advance of the microtome (Rubinow and Juraska, 2009[24]).

Estimation of the coefficient of error (CE)
After the cross-sectional areas “ΣA” were estimated by the software, CE (V) was calculated using the following formula:
The CE for the estimate of the total number of neurons and glial cells, CE (N), was derived from CE(V) and CE(Nv) using the following formula (Gundersen et al.,1999[9]):

Estimation of the length of dendrites and spine morphology
Length estimation was performed on 9-10 vertical uniform random cylinders (Figure 1(Fig. 1)). 100 μm thickness slabs were obtained and stained with Golgi method (De Ruiter and Uylings, 1987[6]). The mean dendritic length per neuron was calculated by dividing the total length by the total number of neurons. To estimate the length, a cycloid grid and a counting frame were mounted on the live images of the mPFC parallel to the vertical axis of the cylinder. Using a videomicroscopy system equipped with a high numerical aperture objective lens, the slab height was scanned. Finally the following formula was used:
where "a/l" is the test area per cycloid test length, "asf" is the area associated with the cycloid grid divided by the area of the counting frame, and “M” is the final magnification at ×4000, "ΣI" is the total number of intersections, "ΣQ-" is the total number of neuron bodies. To estimate the density and morphology of the spines, the above-mentioned dendrites were examined. Dendritic spines were identified and classified as stubby, thin or mushroom spines (Chapleau et al., 2008[4]).

Statistical analysis
The data were analyzed using Kruskall-Wallis, Mann-Whitney U-test, and either two-way or one-way ANOVA followed by Tukey's post-hoc test. Besides, p≤0.05 was considered to be statistically significant.