All methods and experiments have been approved by The Ethics Committee of National University for Theatre and Film I.L Caragiale Bucharest, and followed the guidelines of the Declaration of Helsinki. All participants provided written informed consent for their participation. Subjects: An experimental group of 29 subjects (25 women and 4 men) with a mean age of 34.6 years and a control group of 30 subjects (24 women and 6 men) with a mean age of 32.5 years. Subjects were randomly assigned to the two groups. All subjects were volunteers selected from among the students. Subject inclusion criteria. A complete blood count and C-reactive protein (CRP) measurement were used to check for the presence of an infection / inflammation. Only subjects without signs of infection or inflammation were included. From the same blood samples collected from them, the TNF-α levels from lymphocytes has been measured with a high sensitivity ELISA kit.
Exclusion criteria: rhinitis (or other medical problems that lead to impaired smell), depression, anxiety, chronic diseases that cause infection / inflammation, eyeglasses, metal implants, cardiac pacemaker, claustrophobia. The IQ of the subjects were not measured because their quality of being college students excludes a possible mental disability.
Materials: Fifteen odors were used: coffee, vinegar, vanilla, cocoa, wine, onion, fresh apples, cinnamon, orange, sanitary alcohol, paint, tobacco, diesel oil, jasmine fragrance and chamomile. The odors were selected and adapted from the stimuli used in previous studies (Chu and Downes, 2002; Gardner, et al., 2012). The odors were presented individually from small containers with perforated lid. A second questionnaire containing 2 seven point Likert scales was used for measuring subjective effect upon memory after one month of training. One scale asks to what extent did the subject noticed the onset of spontaneous memories during the day (outside of the experiment) (where1 means “none” and 7 “to a very large extent”). The second scale asks if the subject noticed a greater ease of voluntarily accessing memories, (where 1 means “none” and 7 means “very easy”).
Pre-training session. All the subjects have been exposed to an odor-triggered retrieval session and the subjects have been video monitored during the procedure. The procedure took 30 minutes. After this session, all the subjects have been scanned using resting-state functional connectivity fMRI procedure.
Training session. After the Pre-training session, each subject from the experimental group underwent an autobiographical reminder training for one hour, 2 times / week, for 4 weeks. The experimental group was stimulated to recall autobiographic memories using 15 odors. The participant took each container in his/her hand and smelled its contents through the holes in the lid. He/she waited for a maximum of 20 seconds to see if a memory triggered by that smell appears. If a memory appeared, she/he described it in as much detail as possible. If not, he/she moved on to the next container. Subjects were encouraged to detail the memories as much as possible, insisting on the description of sensory, social and emotional details. After the Pre-training session, each subject from the control group watched 2 short movies for 45 minutes, 2 times / week, for 4 weeks.
Post-training session. After 4 weeks, all subjects have been exposed to the following assessments: A complete blood count and C-reactive protein (CRP) measurement in order to check for the presence of an infection / inflammation, and also for the serum level of TNF-α. All subjects have been exposed to an odor-evoked autobiographical memory recall session, and during this session they have been video monitored. In addition, they completed 3 Likert scales regarding the changes they observed after one month of training (ease of voluntarily accessing memories, the onset of spontaneous memories during the day, and the ease of remembering her/his dreams). The procedure took 30 minutes. After this session, all subjects were scanned using resting-state functional MRI procedure.
Imaging: A 3T Siemens Skyra-MR scanner was used to acquire a resting state functional acquisitions with 281 axial volumes, by means of a 2-dimensional multi-slice echo-planar imaging sequence (TR=2500 ms, TE=30ms, FA=900, matrix size=94x94, voxel size=4x4x4.3mm, 281 volumes of 40 axial images each). Each functional acquisition duration was 11min42s. Additionally, anatomical images were acquired (T1-weighted MP-RAGE, TRTR=2200, TE=2.51 ms, matrix size=256x256, voxel size 0.9x0.9x0.9 mm). The first 5 volumes, acquired to allow longitudinal magnetization to reach a steady state, were discarded.
Data analysis was performed using FMRIB Software Library (FSL) package (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/). Head motion in the fMRI data was corrected using multi-resolution rigid body co-registration of volumes with 12-DOF, as implemented in the MCFLIRT software. brain extraction with BET, spatial smoothing (Gaussian kernel FWHM 5mm) and denoising using nonlinear filtering (SUSAN), ,For one experimental and one control subjects, the movement was too substantial to be corrected (mean scan to scan displacement larger than 0.2 mm, maximum displacement larger than 2 mm), and data from these subjects was excluded from the rest of the analysis. Brain image extraction was carried out for motion corrected BOLD volumes with optimization of the deforming smooth surface model, as implemented in the BET software. Rigid body registration as implemented in the FLIRT software was used to co-register fMRI volumes to T1-MPRAGE (brain-extracted) volumes of the corresponding subjects and subsequently, to the MNI152 standard space. The images were smoothed with a Gaussian kernel FWHM of 5 mm. Image denoising was performed using nonlinear filtering (SUSAN), and a temporal high-pass filtering (with a cutoff frequency of 0.01 Hz) was applied,
Resting state acquisition: Independent Component Analysis (ICA) - the Multivariate Exploratory Linear Decomposition into Independent Components (MELODIC) tool was used to perform spatial group-ICA using multisession temporal concatenation to produce 50 independent component maps (IC maps) representing average resting state networks. Previous studies undertook different ICA dimensionality and found that the number of independent components do not affect these maps for numbers>40 (Dimensionality of ICA in Resting-State fMRI Investigated by Feature Optimized Classification of Independent Components with SVM, May 2015, Frontiers in Human Neuroscience 9(259), Yanlu Wang and Tie-Qiang Li), suggesting 50 as the minimum number of ICA components for obtaining separate known networks (Functional connectivity in the basal ganglia network differentiates PD patients from controls, Konrad Szewczyk-Krolikowski et al,).
,Resting state networks were identified by calculating spatial correlation coefficients between our group ICA maps and the 20-dimensional ICA, Resting-FMRI components from Functional Magnetic Resonance Imaging of the Brain (FMRIB) Laboratory (S.M. Smith, P.T. Fox, K.L. Miller, D.C. Glahn, P.M. Fox, C.E. Mackay, N. Filippini, K.E. Watkins, R. Toro, A.R. Laird, and C.F. Beckmann. Correspondence of the brain's functional architecture during activation and rest. Proc Natl Acad Sci USA (PNAS), 106(31):13040-13045, 2009). We used for it the FSL tool: fslcc which allow running cross-correlations between every volume in one 4D data set with every volume in another, for investigating similarities in ICA outputs. The ICA maps with the greatest correlation coefficient were selected for further analysis. The ICA maps associated with motion or which were localized primarily in the white matter or CSF spaces were classified using criteria suggested by Kelly et al. (2010) and excluded from further study. We also took into account ICA prominent low-frequency power of Fast Fourier Transformation (FFT) spectra and slow fluctuation in time courses. The remaining 15 networks were identified as classical ICA maps as previously reported (Smith et al., 2009; Zuo et al., 2010). These networks are: motor, attention, posterior default mode network (pDMN), higher visual, anterior default mode network (aDMN), fronto-parietal left, primary visual, temporal, fronto-parietal right, executive, somatosensory, basal-ganglia, anterior salience, hippocampal, pontine). The Juelich histological atlas and Harvard-Oxford cortical and subcortical atlases (Harvard Center or Morphometric Analysis) were used to identify the anatomical location, and NeuroSynth 100 top terms atlas (http://neurosynth.org) was used to identify the functional components of the resulting ICA maps.
An intra-network connectivity analysis was performed. This analysis involves comparing the subject-specific spatial maps between experimental and control conditions. To determine subject-specific spatial maps, dual regression analysis was performed on the obtained neural networks using variance normalization (with variance normalization the dual regression reflects differences in both activity and spatial spread of the resting-state networks), similar to previous studies (Emerson et al., 2016; Onu et al., 2015). We performed the following statistical analysis for each of the 15 ICA maps. For the paired two-group difference (two-sample paired t-test), the different component maps were collected across subjects into single 4D files (1 per original ICA map) and tested voxel-wise by nonparametric permutation using the FSL randomize tool (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise) with 5000 permutations and a threshold-free cluster enhanced (TFCE) technique to control for multiple comparisons. As we tested a multitude of resting state networks, we addressed the issue of multiple testing correction by controlling the false discovery rate (FDR) at p<0.05.
To evaluate the relationship between intranetwork connectivity changes as effect of the training program and cognitive performance, mean z-corrected parameters estimates extracted from the clusters of significant experimental versus control conditions differences, were correlated with behavioral variables. The subject-specific z-corrected parameter estimates spatial maps are outputs of stage 2 of the dual-regression.
The intranetwork connectivity quantified indices (mean z-corrected parameter estimates scores) calculated as a result of these procedures were then analyzed in GraphPad Prism.
Biochemistry: TNFα levels were measured from lymphocytes. Blood samples were obtained by venipuncture using EDTA-coated tubes. 2.5 ml fasting venous blood were used to obtain lymphocytes, which were separated by density gradient centrifugation (Biocoll separating solution, Biochrom GmbH). After separation, the lymphocytes were resuspended in 1ml RPMI culture media (Biochrom GmbH) and ultrasonicated. The supernatant was then aliquoted and stored at -20ºC. Due to technical problems, many of the stored probes were compromised. We were able to use PRE and POST probes from 9 subjects that underwent training and from 5 subjects in the control group. TNFα was measured in these samples using a high sensitivity ELISA kit (IBL International GmbH) with the detection limit of 0.13 pg/ml. The calculated intra-assay coefficient of variation was 8.5% and the inter-assay coefficient of variation was 9.8%. TNFα concentrations were measured using the Tecan Reader, with Magellan Reader software (Tecan Group, Ltd, Switzerland). For the calculation of results we used a 4-parameter curve.