A total of 29 men (mean age ± SD, 23.86 ± 2.47 years; range, 20–30) successfully completed the experiments. An additional eight subjects finished the experiment but were excluded from the analysis because they did not meet the criteria regarding sleep duration (n = 7) or learning performance (n = 1) (see below). None of the participants reported ongoing medication, health problems, current medical interventions, or a history of psychiatric, neurological, or sleep disorders. All participants passed a pre-experimental psychological diagnostic interview and a medical screening. Participants did not work on night shifts and did not have exam periods or other learning- or stress-intense occupations for at least 3 weeks prior to the experiment. On experimental days, daytime naps, extensive physical exercise, and the intake of alcohol or caffeine were prohibited. All subjects spent an adaptation night in the sleep laboratory to become accustomed to the experimental conditions. The study was approved by the local ethics committee of the medical faculty of the University Tübingen. All subjects gave written informed consent and were paid for participation.
Subjects were randomly assigned to one of two groups and spent two nights in the laboratory. Subjects in the physostigmine group (n = 15) received an intravenous physostigmine infusion in both nights, whereas subjects in the placebo group (n = 14) received a placebo infusion in both nights. During one of the nights, all subjects were presented with learning-associated odor cues during SWS; during the other night, they were presented with an odorless vehicle, in counter-balanced order (Fig. 1). Each experimental night started at 20:30 h with the placement of polysomnographic electrodes and the intravenous catheter to the participant’s non-dominant arm. At 22:00 h, all subjects learned a 2D object-location task in the presence of the experimental odor. Starting at around 23:15 h, subjects slept for about 40 min (min. 30 min, max. 90 min), including about 20 min of SWS (min. 15 min, max. 30 min). If a participant did not fall asleep within 60 min or did not reach at least 15 min of SWS within 90 min of sleep, the experiment was discontinued. Administration of physostigmine dissolved in saline solution (physostigmine group) or pure saline solution (placebo group) was started at sleep onset (defined as the first epoch of S1 followed by S2). The experimental odor or vehicle was presented for the entire duration of SWS in alternating on/off blocks of 30 s to reduce habituation. Participants were woken up after the max. of 30 min of odor/vehicle stimulation during SWS or if substantial arousals occurred. To allow potential aftereffects of the drug to fade out, all subjects watched a movie for 1.5 h after awakening. Participants then learned an interference object-location task to test for memory stability and were finally tested for recall of the original object-location task.
Odor/vehicle stimulation and physostigmine/placebo administration during sleep were performed in a double-blinded fashion. The odor was delivered by a computer-controlled olfactometer via a nasal mask. The experimental odor was isobutyraldehyde (99%) diluted in 1,2-propanediol at a concentration of 1:50. The intravenous infusion contained 0.25 ml Anticholium (containing physostigmine salicylate) in 17 ml saline solution, which was delivered across 40 min (starting at sleep onset) at a rate of 25.5 ml per hour. Physostigmine is an acetylcholine-esterase inhibitor, suppressing the enzymatic breakdown of acetylcholine, with an elimination half-life of about 20–30 min (Aquilonius and Hartvig 1986; Hartvig et al. 1986).
In a 2D object-location task, subjects learned the locations of 15 card pairs, resembling the game “concentration” (Rasch et al. 2007; Diekelmann et al. 2011 2012). The card pairs depicted animals and everyday objects and were presented on a computer screen in a 5 × 6 matrix. During learning, the locations of all 15 card pairs were presented twice. For each card pair, the first card was presented for 1 s, then both cards for 3 s, followed by a 3-s inter-trial break. During the subsequent immediate recall test, the first card of each pair was presented and the subjects had to indicate the second card location with the computer mouse. Independent of the subject’s response, the correct location of both cards was then shown again for 2 s. The immediate recall procedure was repeated until the subjects reached a criterion of 60% correct responses. Participants who did not reach the criterion after a maximum of six runs were dismissed from the study. The experimental odor was presented time-locked to the presentation of the cards throughout the learning session. For the two experimental nights, two parallel versions of the task were used showing different pictures at different locations.
Before the recall session, subjects learned an interference task to test for memory stability. The learning of the interference task was identical to learning of the original task, with the same 15 card pairs, but the second card of each pair being located at a different position (similar to an A–B, A–C interference paradigm, with A, B, and C referring to locations; Fig.1). Moreover, interference learning included only one immediate recall run to ensure comparable interference input for all subjects. About 30 min after interference learning, final recall of the original memory task was tested in only one recall run. The first card of each pair was presented and subjects had to indicate the location of the second card. No odor cues were presented during the interference task or final recall. Memory performance was calculated as the percentage of correctly recalled card locations at final recall relative to the number of correct card locations during the last immediate recall run at learning (i.e., criterion run).
Sleep recordings, physiological parameters, and control variables
During sleep, standard polysomnographic recordings were obtained, including electroencephalography (EEG, from positions F3, F4, C3, C4, P3, and P4) referenced to the averaged mastoids (M1, M2), electromyography, and electrooculography. An electrocardiogram was recorded to monitor the participants’ heart rate during drug application. Sleep recordings were scored online and offline according to standard criteria (Rechtschaffen and Kales 1968) as sleep stages S1–S4 (with stages S3 and S4 comprising SWS), REM sleep, and wakefulness.
Blood pressure and heart rate were assessed at five time points during each experimental night (before learning, before sleep, after waking, after the interference task, and after final recall). Blood samples were taken in order to determine cortisol concentration at three time points (before sleep, after waking, and between interference task and final recall) to assess individual stress levels. To control for general alertness, a vigilance test and a set of questionnaires were administered at three time points (before learning, after the interference task, and after final recall). In the vigilance task, subjects had to respond as quickly as possible to a red dot appearing every 2–10 s for a total of 10 min on the left or right side of a computer screen and reaction times (in ms) were assessed. Subjects rated their subjective sleepiness on the Stanford Sleepiness Scale (Hoddes et al. 1973), their general mood on the short version of the Multidimensional Mood State Questionnaire (Steyer et al. 1994), and the presence or absence of 27 potential side effects of physostigmine (e.g., sweating, blurred vision, headache). Before and after learning, an odor detection test was performed to ensure olfactory sensitivity in all participants.
EEG power spectral analysis and detection of spindles and slow oscillations
EEG signals were recorded at a sampling rate of 200 Hz and bandpass-filtered between 0.16–35 Hz. The power spectrum was estimated over subsequent 5-s segments (Hanning window with an overlap of 0.9 times the window length) for sleep stages S2 and SWS (i.e., combined sleep stages S3 and S4). Spindles and slow oscillations were detected in sleep stages S2–S4, using the open source toolbox SpiSOP (www.spisop.org; see Online Resource 1 for a more detailed description) and adopting the detection criteria from Mölle and colleagues (Mölle et al. 2011). Slow- and fast-spindle frequencies were defined individually for each dataset by peaks in the power spectrum of the averaged EEG channels F3/F4 between 9 and 12 Hz for slow spindles and of channels C3/C4 between 12 and 15 Hz for fast spindles. Slow oscillations were detected in the EEG recorded from F3/F4 after bandpass-filtering the signal between 0.3 and 3.5 Hz. For sleep spindles, the parameters density, duration, and maximal envelope were analyzed. For slow oscillations, duration, amplitude, up-to-down slope, and down-to-up slope were analyzed. In order to assess potential changes in these parameters due to the odor stimulation, we compared 10-s time windows before odor onset (“odor off”) and after odor onset (“odor on”).
Memory performance, physiological parameters, sleep data, and control variables were analyzed using analyses of variance (ANOVAs) with the between-subject factor “physostigmine/placebo” and the within-subject factor “odor/vehicle.” For the power analysis, we introduced the additional within-subject factor “S2/SWS.” For the analysis of spindle and slow-oscillation events, we introduced the additional within-subject factor “odor off/odor on.” For all electrophysiological data, we followed an interquartile range outlier rejection rule (lower threshold: Q1–2.2 × (Q3–Q1); upper threshold: Q3 + 2.2 × (Q3–Q1)) (Hoaglin and Iglewicz 1987). Effect sizes are provided as partial eta squared (η
2) or Cohen’s d. Control parameters were Bonferroni-corrected for the number of tested measures. A p value of ≤ 0.05 was considered significant.