Fig 1.
Overview of the experimental design.
In Experiment 1, the rats underwent pre-training dHPC lesions and, after recovery, a CFC session. Half the sample underwent fear memory test 48 h later, whereas the other half was perfused 3 hours after CFC and had their brains processed and stained for pCREB protein. Thirty regions had their pCREB expression quantified and their pCREB inter-regional correlations computed. After thresholding the correlations, we analyzed the networks properties and compared them between the groups. Following network analysis, Experiments 2 and 3 employed double lesions to test if the network differences observed could be empirically supported.
Fig 2.
Regions included in Experiment 1.
The columns show the name of each region, the abbreviations, the source of the anatomical definition adopted, and the color code used in the following figures for each group of regions. Color code—Red: basolateral complex of the amygdala; Dark Red: central amygdala nuclei; Blue: dorsal hippocampus; Light Blue: ventral hippocampus; Green: parahippocampal regions; Purple: prefrontal cortices; Magenta: retrosplenial cortices. A) 3D diagram of a rat brain showing anatomical localization of the included regions. Representative photomicrograph of a pCREB immunolabelled brain slice before (B) and after (C) nuclei quantification and region parcellation by Cellprofiler. The scale bars indicate 200 μm.
Fig 3.
dHPC lesion does not impair CFC learning and memory.
(A) Schematic diagram showing the distribution of the lesions in the dHPC group. (B) Mean (black line) and bootstrapped 95% CI of the Total Freezing Time during the five min context fear memory test of dHPC (N = 12), SHAM (N = 12) and Imm (N = 8) groups. The open circles show data distribution in each group. (C) Cumulative distribution of the sample as a function of Freezing Time showing the sample distributions. The “*” shows a significant difference from Imm at level of p<0.05. (D) Mean (black line) and Bootstrapped 95% CI of the mean (boxplots) of the pCREB-positive nuclei density in each region and each group. The black dots show the data point distributions.
Fig 4.
Generation of connectivity networks in each group.
After computing the inter-regional correlations (left), three thresholds were applied (p < 0.05, 0.025 and 0.01) and the most robust correlation coefficients (center) composed the networks (right). Networks were generated for SHAM (top), SHAM-nH (middle) and dHPC (bottom) matrices. In the matrices, colors reflect correlation strength (scale, right). In the network, the colors of the nodes are coded according to the Fig 2, and the sizes of the nodes represent their degree (number of connections).
Fig 5.
dHPC damage does not alter the CFC learning network small-worldness.
Boxplots showing mean, lower and upper quartiles, and 95% CIs of the Empirical/Random ratio of Geff and mean Leff for the SHAM (left), SHAM-nH (center) and dHPC (right) networks and on the 0.05 (bottom), 0.025 (center) and 0.01 (top) thresholds. Small-world networks are expected to have Geff ratios around 1 (empirical and randomized networks have roughly the same values) and higher mean Leff ratios (higher empirical values than those of the randomized networks).
Fig 6.
Hub identification of the networks.
(A) The rankings of each centrality are shown for the dHPC network under the 0.05 threshold. The colored nodes are the upper 25% most central in each metric. (B) The intersections of the upper 25% most central regions of each metric are shown for the dHPC network under the 0.05 threshold. Any region within the overlapping area of at least three metrics was considered a network hub (inside the red perimeter). The hubs were identified in the networks with 0.025 and 0.01 thresholds as well (S2–S4 Figs), and the hubs of each threshold were intersected (C) to identify stable hubs across the thresholds in each network. Nodes are colored according to the code in Fig 2.
Fig 7.
Centrality comparison between SHAM-nH and dHPC networks.
The comparison was done for each region, centrality metric and threshold. Values in each cell show ranks of the amount of significant differences (p < 0.05 in the permutation test) across the thresholds in each region and metric, ranging from 0 to 3. The hub score column shows the amount of thresholds in which a region was considered a hub. Significantly higher centrality values and hub scores in the dHPC and SHAM networks are shown in shades of blue and green, respectively. If a region was hub in both networks, a stripped blue-green cell is shown with their respective gradient. It is possible to observe which hub scores are associated with significant differences in the centrality comparisons. The thresholds in which each difference and hub occurred is shown in S5 Fig. Wdg: Weighted Degree; Evc: Eigenvector; Clo: Closeness; Bet: Betweenness.
Fig 8.
Connectivity change in dHPC network.
(A) Cumulative distributions of the Fisher’s Z transformed correlation coefficients from the SHAM-nH and dHPC matrices. The “#” indicates that these distributions are significantly different (Kolgomorov-Smirnov test, p<0.05). (B) Cumulative distribution showing the z-score of the correlation coefficient differences between the groups. The dashed line shows the absolute Z-score of 2, revealing the values considered significant (beyond it) at the level of α = 0.05. (C) The significantly different coefficients were plotted in each network, showing the network and nodes to which it belonged. The same procedure was performed in the 0.025 and 0.01 threshold networks and showed similar results S6 Fig.
Fig 9.
Per and dHPC-Per lesions on CFC learning.
(A) Histological diagrams showing the distribution of areas damaged in dHPC, Per and dHPC-Per groups. The more overlapped the damaged areas across subjects, the darker the area. (B) Mean and bootstrapped 95% IC of the total freezing time in SHAM, dHPC, Per and dHPC-Per groups during 5 min of CFC memory test. Dots show the sample distribution of each group. (C) Cumulative distribution of the total freezing time in each group in the same CFC memory test. The bottom table shows all the statistical tests performed and the corrected p-value for each comparison.
Fig 10.
RSC and dHPC-RSC lesions on CFC learning.
(A) Histological diagrams showing the distribution of areas damaged in dHPC, RSC and dHPC-RSC groups. The more overlapped the damaged areas across subjects, the darker the area. (B) Mean and bootstrapped 95% IC of the total freezing time in SHAM, dHPC, RSC and dHPC-RSC groups during 5 min of CFC memory test. Dots show the sample distribution of each group. (C) Cumulative distribution of the total freezing time in each group in the same CFC memory test. The bottom table shows all the statistical tests performed and the corrected p-value for each comparison. “*” shows significant differences relative to SHAM group (corrected-p <0.05).