پروپرانولول حواس پرتی عاطفی در حافظه کاری را کاهش می دهد: نقش واسطه جزئی پروپرانولول ناشی از کورتیزول را افزایش می دهد؟

Abstract Noradrenalin modulates prefrontal function, such as working memory (WM), and is associated with enhanced distractibility, and enhanced memory for emotional events and stimuli. The beta-blocker propranolol has been shown to reduce memory for emotional stimuli. Herein we describe investigations aimed at assessing whether the administration of propranolol would reduce the interference by emotional distractions during WM performance. In a between-subjects design, 48 young, healthy men received 80 mg propranolol (n = 25) or placebo (n = 23), before performing an “emotional Sternberg task” with neutral and negatively arousing distracters. Compared to placebo, propranolol impaired WM at low load, however, it also reduced the interference by emotional distracters at high load. Furthermore, an explorative moderated-mediation analysis indicated that the observed propranolol effects on emotional distraction were partially mediated by cortisol. In future non-clinical and clinical memory studies using propranolol administration, cortisol elevations should be monitored to further investigate the potential mediating role of cortisol.

. Introduction When stressed, one of the neurohormonal systems that is activated is the locus coeruleus-noradrenergic system (Berridge & Waterhouse, 2003). This system plays a key modulatory role in prefrontal function (Minzenberg et al., 2008, Ramos and Arnsten, 2007 and Berridge and Waterhouse, 2003), and is critically involved in emotional memory (McGaugh and Roozendaal, 2002 and Roozendaal et al., 2008). Optimal levels of noradrenalin (NA) can improve functioning of the prefrontal cortex (PFC), whereas excessive NA or a depletion of NA impairs PFC function (Ramos & Arnsten, 2007; Arnsten, 2009). Stress-induced elevated NA is thought to take the reflective PFC “of-line” in favor of other more posterior brain areas, such as amygdala, hippocampus, and sensory- and motor areas, to allow for rapid emotional, or more habitual and reflexive behaviors (Ramos & Arnsten, 2007; Arnsten, 1997). Given the importance of the PFC in working memory (WM) performance (Kane and Engle, 2002 and Ranganath et al., 2003), it is of no surprise that high levels of NA have also been found to be associated with impaired WM performance (Arnsten et al., 1999, Birnbaum et al., 1999 and Mao et al., 1999). WM can be defined as the capacity to maintain relevant information and to suppress irrelevant information. Patients with stress-related psychiatric disorders such as PTSD and depression, show poor WM performance and stronger interference from irrelevant negative emotional material (Joormann & Gotlib, 2008; Morey et al., 2009). Typically, in PTSD patients, pharmacological challenge tests or exposure to traumatic reminders are associated with increased noradrenergic responsiveness (Bremner, Krystal, Southwick, & Charney, 1996), and hypoactive responding in medial PFC, along with a hyperactive amygdala (Elzinga and Bremner, 2002, Etkin and Wager, 2007, Liberzon and Sripada, 2008 and Shin et al., 2006). When instructed to ignore emotional images shown during a WM task, PTSD patients displayed a similar pattern of decreased activity in dorsal areas, associated with WM and attention, and an enhanced neural activity in ventral areas (including the amygdala) associated with emotion processing relative to the trauma-exposed non-PTSD control group (Morey et al., 2009). These observations may be described as an exaggerated form of a “normal” response to emotional distractions during WM. That is, healthy individuals also pay more attention to emotional stimuli than neutral ones, because of their salience and significance for survival even when these are deemed irrelevant, for example, in a context of an ‘emotional WM task’, where emotional stimuli are used as distracters (Kensinger & Corkin, 2003). As a result, WM performance slows down during the emotional distraction trials (Dolcos & McCarthy, 2006; Kensinger & Corkin, 2003). The response to emotional stimuli by the amygdala is mediated by NA (Berridge and Waterhouse, 2003, van Stegeren, 2008 and van Stegeren et al., 2005). Elevated NA enhances amygdala response (Onur et al., 2009) and enhances the attention for emotional stimuli (DeMartino, Strange, & Dolan, 2008). Imaging studies have shown that administration of propranolol, a highly lipophilic non-selective beta-adrenergic receptor blocker, that blocks the action of adrenalin on both beta1 and beta2 adrenergic receptors, reduces the activity in the amygdala during emotional processing (Strange and Dolan, 2004 and van Stegeren et al., 2005). A number of studies aimed at elucidating the role of NA in emotional memory, have further shown that propranolol generally reduces memory for emotional events and stimuli (see for a review Chamberlain, Muller, Blackwell, Robbins, & Sahakian, 2006), when encoding takes place after propranolol administration (Cahill et al., 1994, Cahill and van Stegeren, 2003 and van Stegeren et al., 2005). Taken together, these findings suggest that propranolol might improve emotional WM performance, owing to the diminished interference of emotional distractions. The main aim of the present study was to investigate whether propranolol would improve emotional WM performance in young healthy men, by reducing the impact of emotionally negative distracters. Furthermore, we also performed an explorative analysis to investigate whether the stress hormone cortisol might mediate the effects of propranolol on emotional WM performance. There were two indicators that point towards a possible mediating role of cortisol in this regard: First, propranolol administration had been previously shown to elevate the levels of cortisol in the present sample (Tollenaar, Elzinga, Spinhoven, & Everaerd, 2009), as well as in other memory studies in which propranolol was administered (Maheu et al., 2004 and Maheu et al., 2005). Secondly, as part of the present study, we have also found that cortisol administration leads to enhanced performance on the present emotional WM memory task (Oei, Tollenaar, Spinhoven, & Elzinga, 2009).

. Results WM data of two participants from the placebo group and one from the propranolol group were not recorded because of a computer failure. Three participants (two from the placebo group and one from the propranolol group) had to be excluded from further analyses because of extreme numbers of errors (>25%), leading to missing data in at least one category. A total of 48 participants, 25 participants in the propranolol group and 23 in the placebo group were left for further analysis. Participants were not able to tell whether they had received placebo or propranolol (Chi-square = 4.70, df = 4, p = .32): Four participants correctly indicated noticing an effect of propranolol, while five participant, who received placebo, erroneously indicated that they had ingested propranolol 3.1. Physiological measurements 3.1.1. Heart rate See Fig. 1 for means and standard errors. Separate t-tests showed that heart rate was significantly lower in the propranolol group compared to placebo, at t110 (t46 = 4.31, p < .0005) and t135 (t46 = 4.71, p < .0005), but not at baseline, t1 (t46 = 0.18, p = .86). Mean heart rate and standard errors in the propranolol- and placebo group. ... Fig. 1. Mean heart rate and standard errors in the propranolol- and placebo group. ***Significant difference between groups at p < .0005. Figure options 3.1.2. Blood pressure Systolic blood pressure was significantly lower in the propranolol group after pill administration (systolic blood pressure at t1, t46 = 0.41, p = .68; t110, t46 = 2.83, p = .007; t135, t46 = 3.99, p < .0005; diastolic blood pressure showed a trend at t135, t46 = 1.91, p = .06. At both other time points, ps > .36, see Fig. 2 for means and standard errors). Mean blood pressure and standard errors in the propranolol- and placebo group. ... Fig. 2. Mean blood pressure and standard errors in the propranolol- and placebo group. SBP = systolic blood pressure; DBP = diastolic blood pressure. **Significant difference between groups at p < .01. ***Significant difference between groups at p < .0005. Figure options 3.1.3. Alpha-amylase Amylase data were normalized using log-transformations (see Fig. 3 for untransformed means and standard errors). There were no differences between groups at baseline just before pill administration (t0, t45 = 0.85, p = .40). The propranolol group had trend-level lower amylase levels just before WM testing (t110, t45 = 1.75, p = .08) and significantly lower values after testing (t135, t46 = 2.29, p = .03) as compared with the placebo group. Mean alpha-amylase and standard errors in the propranolol- and placebo group. ... Fig. 3. Mean alpha-amylase and standard errors in the propranolol- and placebo group. *Significant difference between groups at p < .05. Figure options 3.1.4. Cortisol Mean cortisol levels and standard errors at the three time points are depicted in Fig. 4. At t1, log transformed cortisol levels did not differ between groups (t30.66 = 0.27, p = .79). The placebo group had lower cortisol levels at t110 (t46 = −3.29, p = .002), and t135 (t46 = −3.45, p = .001). Mean cortisol levels and standard errors in the propranolol- and placebo group. ... Fig. 4. Mean cortisol levels and standard errors in the propranolol- and placebo group. ***Significant difference between groups at p < .005. Figure options 3.2. Present-target trials 3.2.1. Reaction times Mean reaction times and standard errors are shown in Table 1. At present-target trials, significant within-subjects effects for Load (F[1, 46] = 293.51) and Distracter (F[1, 46] = 25.79) were found, with shorter RTs at low load compared to high load, and shorter RTs when distracters were neutral compared to emotional (both ps < .0005). Also, a significant Group by Load by Distracter effect was revealed (F[1, 46] = 5.49, p = .02), which indicated that in contrast to the propranolol group, the placebo group was significantly slower during emotional trials than neutral trials at high load (see Fig. 5). There was, however, no significant overall between-groups effect (F[1, 46] = 0.62, p = .43). Table 1. Means and standard errors of the reaction times in the propranolol- and placebo group. Target Group Propranolol Placebo Load Low High Low High Distracter M ± SE M ± SE M ± SE M ± SE Present Emotional 793.87 ± 29.84 1120.18 ± 41.34 819.94 ± 31.11 1192.96 ± 43.09 Neutral 739.16 ± 26.28 1082.95 ± 34.27 785.13 ± 27.40 1067.03 ± 35.75 Total 766.52 ± 26.85 1101.57 ± 35.14 802.54 ± 27.99 1129.99 ± 36.63 Absent Emotional 848.79 ± 28.77 1261.38 ± 47.08 849.68 ± 29.99 1356.52 ± 49.08 Neutral 808.03 ± 29.87 1196.74 ± 49.41 824.97 ± 31.15 1337.57 ± 51.52 Total 828.41 ± 27.88 1229.06 ± 44.89 837.32 ± 29.06 1347.04 ± 46.81 Table options The group by load by distracter interaction. **Difference between emotional and ... Fig. 5. The group by load by distracter interaction. **Difference between emotional and neutral trials within placebo group, t(22) = 3.55, p < .005. Figure options 3.2.2. Errors Mean errors (and SE) are shown in Table 2. Analysis of errors indicated that the significant triple interaction in RTs during present-target trials was not the result of a speed/accuracy trade-off (Group × Load × Distracter, F[1, 46] = 0.44, p = .51). During present-target trials significantly less errors were made at low load (M ± SE, 0.73 ± 0.08), than at high load (2.49 ± 0.19). There was an interaction between Group and Load (F(1, 46) = 5.19, p = .03), which revealed that the propranolol group made more errors at low load (0.90 ± 0.71) than the placebo group (0.57 ± 0.41) (t38.92 = −2.03, p = .05) but not at high load (placebo group: 2.78 ± 0.35; propranolol group: 2.20 ± 0.18, t32.61 = 1.47, p = .15). Furthermore, there was a within-subjects interaction between Load and Distracter F(1, 46) = 4.28, p = .02. At high load, more errors were made when distracters were emotional than when they were neutral. There were no other significant effects (all ps > .18). Table 2. Means and standard errors of the Error rates in the propranolol- and placebo group. Target Group Propranolol Placebo Load Low High Low High Distracter M ± SE M ± SE M ± SE M ± SE Present Emotional 0.80 ± 0.16 2.48 ± 0.32 0.57 ± 0.16 3.00 ± 0.33 Neutral 1.00 ± 0.1 1.92 ± 0.30 0.57 ± 0.17 2.57 ± 0.31 Total 0.90 ± 0.12 2.20 ± .27 0.57 ± 0.12 2.78 ± 0.28 Absent Emotional 0.56 ± 0.14 0.68 ± 0.19 0.44 ± 0.15 0.48 ± 0.19 Neutral 0.52 ± 0.13 0.48 ± 0.18 0.44 ± 0.14 0.70 ± 0.18 Total 0.54 ± 0.11 0.58 ± 0.14 0.44 ± 0.11 0.59 ± 0.15 Table options 3.3. Absent-target trials 3.3.1. Reaction times See Table 1 for means and standard errors during absent-target trials. During absent-target trials, the within subjects factors Load (F[1, 46] = 285.02, p < .0005) and Distracter (F[1, 46] = 6.46, p = .01) were significantly different, with faster RTs at low load (M ± SE: Low load, 832.87 ± 20.14; High load, 1288.05 ± 32.43), and faster RTs when distracters were neutral (M ± SE: emotional trials, 1079.09 ± 23.89; neutral trials, 1041 ± 25.10). There was a significant interaction between Group and Load (F[2, 46] = 4.09, p = .05), with a trend for faster RTs in the propranolol group at high load (M ± SE: 1229.06 ± 44.89) compared to placebo (M ± SE: 1347.04 ± 46.81) (t46 = 1.82, p = .08), but not at low load (t46 = 0.22, p = .83). There was no significant overall difference between groups (F[1, 46] = 1.84, p = .18), nor other interactions (all ps > .30). 3.3.2. Errors Analysis of the errors during absent-target trials, revealed no significant within- or between-subjects effect, nor any interaction effects (all Fs < 1.59, all ps > .21) (see Table 2 for means and standard errors). 3.4. Explorative moderated-mediation analysis There were differences in cortisol levels between the propranolol and placebo group, with significantly decreasing cortisol levels over time in the placebo group only (see Tollenaar et al., 2009). Inspection of the data revealed that in the placebo group all – but one participant – showed decreased cortisol levels compared to baseline. In contrast, in the propranolol group, apart from three participants, there was an absence of decrease and in half of the group (n = 12) cortisol levels even increased between baseline and the start of the WM task (maximum increase (t110–t0) = 16.59 nmol/L). To explore whether the enhancing effects of propranolol on the emotional trials of the WM task were mediated by cortisol, we first converted the RTs of the present-target trials into a single difference score (WMDiff), by subtracting the difference in RTs between High load emotional trials and neutral trials, from the difference between Low load emotional and neutral trials. This way, high scores represented a load-dependent difference between emotional and neutral trials, while low scores represented smaller differences in load and distracter type (M ± SE: propranolol group, −12.13 ± 25.49, placebo group, 85.95 ± 35.99, F(1, 47) = 5.07, p = .03). Furthermore, a cortisol-difference score was calculated (cortisol level just before testing WM minus baseline cortisol level, just before ingesting propranolol or placebo) (M ± SE: propranolol group, 0.82 ± 1.18, placebo group, −4.38 ± 0.96, F(1, 47) = 11.42, p = .001). These two new variables were then checked for outliers. In both groups two outliers were detected which were subsequently removed (all outliers had extreme values regarding cortisol-difference scores: placebo group outliers were due to extremes in baseline cortisol levels >18 nmol/L, the propranolol group outliers were the two participants with the highest increase in cortisol level after propranolol administration: increase >12 nmol/L, leading to cortisol levels >16 nmol/L). As a next step, the dependent variable was entered into a moderated mediation model (see Preacher, Rucker, & Hayes, 2007) with Group as independent variable, and cortisol-difference score level as mediator variable, and Group as moderator variable. Group was added as moderator variable to be able to differentiate the conditional indirect effect of propranolol on the dependent variable, because the influence of placebo on cortisol is likely to be zero (Model 1, Preacher et al., 2007). The SPSS macro used, was provided by Dr. A. Hayes (http://www.comm.ohiostate.edu/ahayes/SPSS%20programs/modmed.htm). It calculates the Sobel test for the conditional indirect effects as well as its percentile-based, bias-corrected, and bias-corrected and accelerated bootstrap confidence intervals, which is recommended for small samples (Preacher and Hayes, 2004 and Shrout and Bolger, 2002). Estimates of all paths were calculated using ordinary least squares regression. The results of these analyses are displayed in Table 3. Table 3. Regression results of the moderated mediation model. Predictor B SE t p Mediator variable model Constant −3.53 0.89 −3.94 .0003 Group 3.14 1.23 2.54 .015 Dependent variable model Constant 113.44 42.81 2.65 .01 Cortisol 12.97 8.35 1.55 .13 Group −129.31 52.19 −2.28 .02 Cortisol × Group −26.43 10.19 −2.42 .02 Table options A statistically significant interaction was found between Cortisol and Group, indicating that the indirect effect of Cortisol on WMdiff was moderated by group, with smaller WMdiff scores as cortisol levels increased. None of the bootstrap confidence intervals when Group was propranolol contained 0, which confirmed that the conditional indirect effect of propranolol was significant at α = .05 (see Table 4). Table 4. Bootstrapped conditional indirect effects of Group on WM score via Cortisol at specific values of the moderator (Group). Group Point estimate SE Z Bootstrapping Percentile 95 CI BC 95% CI BCA 95% CI Lower Upper Lower Upper Lower Upper Placebo 40.88 36.47 1.12 −20.40 126.01 −9.91 143.55 −11.75 140.74 Propranolol −41.59 24.29 −1.71 −97.66 −2.43 −106.66 −6.85 −107.05 −6.97 Note. CI = confidence interval; BC = Bias-corrected; BCA = Bias-corrected and accelerated; 5.000 bootstrap samples. Table options Finally, we investigated the possibility that propranolol-induced cortisol elevations were markers for inter-individual propranolol efficacy. An increased propranolol action along with cortisol elevations could mean that propranolol reduced emotional distracter interference due to a stronger reduction in adrenergic activation, rather then through cortisol release. However, post hoc ANOVAs with the propranolol group divided into a high cortisol elevation (n = 12) and a no elevation group (n = 13), did not show any differences in change in levels of alpha-amylase, heart rate, or blood pressure, nor any significant correlation between cortisol and adrenergic measures (all ps > .05).