Gender differences in the functional organization of the brain for working memory

Transcription

1 BRAIN IMAGING Gender differences in the functional organization of the brain for working memory Oliver Speck, 1,CA Thomas Ernst, 1,2 Jochen Braun, 3 Christoph Koch, 3 Eric Miller 4 and Linda Chang 1 Departments of 1 Neurology, 2 Radiology and 4 Psychiatry, UCLA School of Medicine, Harbor-UCLA Medical Center, 1124 W. Carson Street, Building B-4, Torrance, CA 90502; 3 Division of Computational Neural Systems, Department of Biology, California Institute of Technology, Pasadena, CA 91125, USA CA Corresponding Author Received 26 April 2000; accepted 5 June 2000 Gender differences in brain activation during working memory tasks were examined with fmri. Seventeen right-handed subjects (nine males, eight females) were studied with four different verbal working memory tasks of varying dif culty using whole brain echo-planar fmri. Consistent with prior studies, we observed activation of the lateral prefrontal cortices (LPFC), the parietal cortices (PC), and additionally, caudate activation in both sexes. The volume of activated brain tissue increased with increasing task dif culty. For all four tasks, the male subjects showed bilateral activation or rightsided dominance (LPFC, PC and caudate), whereas females showed activation predominantly in the left hemisphere. The task performance data demonstrated higher accuracy and slightly slower reaction times for the female subjects. Our results show a highly signi cant ( p, 0.001) gender differences in the functional organization of the brain for working memory. These gender-speci c differences in functional organization of the brain may be due to gender-differences in problem solving strategies or the neurodevelopment. Therefore, gender matching or strati cation is required for studies of brain function using imaging techniques. NeuroReport 11:2581±2585 & 2000 Lippincott Williams & Wilkins. Key words: fmri; functional brain organization; Gender differences; Lateralization; Task dif culty; Working memory INTRODUCTION Working memory (WM) is used for active temporary storage and manipulation of information within the time range of a few seconds, and is an important component for many higher cognitive functions. Early neuroimaging studies of WM were conducted using PET. Since the advent of fmri, a technique that can non-invasively detect brain activation during the performance of a de ned task, many studies using spatial and verbal working memory paradigms have been published [1]. The lateral prefrontal cortex (LPFC) and the parietal cortex (PC) in both hemispheres are consistently shown to be involved in WM tasks. Nevertheless, results vary as to what extent the hemispheres are involved in the execution of the tasks. None of the studies have evaluated gender differences related to WM tasks. Since gender related differences have been reported with other tasks (e.g. word processing [2]), we evaluated a group of healthy volunteers with fmri during the execution of verbal WM tasks. The tasks consisted of sequential letter (n-back) and sequential number (n-increment) paradigms. We hypothesized that gender differences exist and that they contribute to the variability in the results previously published. MATERIALS AND METHODS Subjects and tasks: Seventeen healthy, right-handed, non-smoking subjects (nine men: mean age years, education years; eight women: age years, education years) were scanned during the performance of a number of verbal WM tasks with varying degrees of dif culty. Neither age nor education level differed signi cantly between male and female subjects. Prior to enrolment, subjects were screened to exclude those with a history of any medical or neuropsychiatric illnesses, as well as anyone with a history of head injury or drug abuse. Subject motion was minimized using a bite-bar mounted to the head coil. In addition, extensive cushioning to provide a comfortable and relaxed head position further reduced head motion. Prior to the study, each subject signed a written consent approved by the Institutional Review Board at Harbor-UCLA Research and Education Institute. Four WM tasks were modeled after tasks from the California Computerized Assessment Package (CALCAP) [3]. For the two sequential letter tasks, random alphabetical letters were presented visually at a rate of 1/s. The subjects were instructed to respond whenever the current letter was & Lippincott Williams & Wilkins Vol 11 No 11 3 August

2 O. SPECK ET AL. the same as the one before (1-back) or two before (2-back). For the two sequential number tasks, randomly generated single digit numbers were sequentially displayed at a rate of 1/s. The subjects were instructed to respond whenever the current number was one or two higher (1-increment or 2-increment) than the previous number. During each task period of 30 s, six targets were presented at random time points. To respond, the subjects had to push a low resistance button that was designed to minimize responserelated movements. During the baseline periods (30 s), nonsense characters of the same size were randomly displayed at the same rate and luminance, and the subjects were instructed to maintain xation at the center cross on the screen. Each experimental run consisted of four blocks of 30 s stimulation each altered with a 30 s block of control condition. For each of the four different tasks, two experimental runs were performed resulting in eight activation and baseline periods. The performance data (reaction time and accuracy) were recorded during the scans. Tasks were generated from an HP workstation (Hewlett Packard, Palo Alto, CA) and displayed on a 20" monitor via a mirror mounted on the head coil inside the scanner. The monitor was placed at the end of the patient table and magnetically shielded to avoid geometric or chromatic distortions due to the magnetic eld. The entire battery of tasks was explained to each subject prior to the scans; subjects were allowed 1±2 practice trials outside of the scanner to ensure that each subject fully understood the tasks. MRI scans: Scans were performed on a 1.5 T system (GE SIGNA 5.7, Milwaukee, WI) with fast gradients (SR 120). A standard quadrature head coil was used for all scans. After obtaining a high resolution anatomic scan (fast inversion recovery, repetition time (TR) 4500 ms, echo time (TE) 32 ms, inversion time (IR) 120 ms, 36 slices, resolution mm), fmri was performed using a single shot gradient echo echo-planar-imaging sequence (EPI) (TR 2500 ms, TE 60 ms, 16 slices, resolution mm). In each subject, four periods of 30 s stimulation and 30 s baseline (4 min total) were acquired twice for each task. Each acquisition was preceded by 10 s of dummy scans to reach a steady state of the magnetization. The total scan time per subject was, 1h. Data processing: Data were processed on an SGI workstation (Silicon Graphics, Mountain View, CA) using the Statistical Parametric Mapping software (SPM96). The rst processing step was motion correction [4]. All data sets showed, 0.8 mm maximal displacement (25% of the voxel size) and, 1.08 rotation during an entire scan (before motion correction). Next, the EPI images were coregistered to the highresolution scans using a customized software package [5] and transformed into Talairach space for group analyses. Prior to the Talairach transformation, the brain was automatically segmented from the inversion recovery scan [5]. For the Talairach normalization, the data were matched to a standard reference brain of the same contrast using a full af ne transformation (12 degrees of freedom) and spatial cosine transformations. To account for residual inter-individual differences in cranial anatomy and to increase the signal to noise ratio, the data were smoothed with a Gaussian lter (10 mm FWHM). Activation maps were calculated for each paradigm and each group as contrasts between stimulation and baseline periods using the generalized linear model [6]. The design matrix was generated using a 6 s delayed boxcar reference function to account for the delay in hemodynamic response, a high pass lter with a threshold frequency of 1/ 120 Hz, and inter-subject scaling of the mean signal intensity. Activated areas were de ned as regions above a certain z-score threshold, and with an extent of > 0.5 ml. In addition, maps showing the interaction effect between stimulation and gender were calculated using the same model. For all activated brain regions a lateralization index L was de ned as L ˆ (V Right V Left )/(V Right V Left ), where V Left and V Right represent the activated brain volume in the left and right hemisphere. Thus, a lateralization index of L ˆ 1 represents an activation that is right-sided only while L ˆ 1 represents a completely left-sided activation. Since the lateralization index depends on the z-score threshold used, with a tendency towards higher degrees of lateralization at higher thresholds, the L-values were calculated using z-score thresholds adjusted to yield an overall activated brain volume of 50 ml. RESULTS As in prior studies, all subjects showed activation in the LPFC, the PC, and the supplementary motor area (SMA) during all the WM tasks. Additionally, we observed caudate activation with these WM tasks. The stereotactic coordinates of the local maxima of the activated brain regions, as detected with the Statistical Parametric Mapping software (SPM96), are given in Table 1. The total Table 1. The Talairach coordinates (x,y,z) of the foci of the activated brain regions for all different tasks and gender groups (N.D.: not detected) Task Gender LPFC PC Caudate SMA R L R L R L 1-back Female 38, 6,58 38, 18,58 38, 52,64 32, 54,62 N.D. N.D. 0, 10,72 Male 34, 2,60 40, 8,60 34, 64,56 24, 68,60 N.D. N.D. 2,2,62 1-inc. Female 38, 4,60 40, 10,54 36, 54,56 30, 60,62 N.D. N.D. 0, 2,70 Male 36,6,54 46,8,48 36, 58,54 34, 62,56 N.D. N.D. 0,10,62 2-inc. Female 34, 6,58 44, 4,52 32, 66,56 26, 70,62 14, 18,18 14, 18,24 0,8,66 Male 36, 6,56 44,0,46 36, 60,56 28, 72,58 12, 20,18 10, 20,20 2,6,62 2-back Female 32, 4,56 36, 12,60 30, 60,58 28, 56,60 18, 10,30 18, 10,24 2, 6,70 Male 38,6,58 42, 4,50 36, 64,56 26, 62,58 20, 2,22 12, 8,18 0,8, Vol 11 No 11 3 August 2000

3 GENDER DIFFERENCES IN WORKING MEMORY volume of brain activation (at a constant z-score threshold of 7.0) was smallest for the 1-back task, intermediate for the 1-increment and 2-increment tasks, and largest for the 2-back task (see Fig. 1). Since all subjects ranked the task dif culty in the order: 1-back, 1-increment 2-increment, 2-back, the activated volume paralleled the subjective task dif culty. There was a statistically signi cant difference in the activation pattern between male and female subjects: while men showed symmetric activation or right brain dominance, women predominantly activated the left hemisphere during all the tasks. Interestingly, the localizations of the activated areas do not differ between the groups or tasks. No gender difference in SMA activation (size or location) was observed for any of the tasks. As an example, the group activation patterns of men and women for the 2- increment task are shown in Fig. 2. The lateralization indices for all tasks and activated cortical areas are shown in Table 2, together with the z-score thresholds used to Activated brain volume (ml) back 1-increment 2-increment 2-back Fig. 1. Activated brain volume for the different working memory tasks ordered by their degree of dif culty (the 1-increment and 2-increment tasks were ranked similarly in their level of dif culty). The volume re ects the sum over all activated brain regions, using a xed z-score threshold of 7.0 for all the tasks. Table 2. Lateralization indices L for all activated regions in female and male subjects, and z-score thresholds used. The z-score thresholds were adjusted to result in a total activated brain volume of 50 ml. Positive values represent right hemisphere dominance. Negative values represent left dominance. (N.D.: not detected) Task Gender LPFC PC Caudate z-threshold 1-back Female N.D. 5.7 Male N.D inc. Female N.D. 6.5 Male N.D inc. Female Male back Female Male obtain 50 ml of activated volume. A graphical representation of the overall lateralization for the two gender groups during all the tasks is shown in Fig. 3. The brain areas that show statistically signi cantly different activation between male and female subjects were determined by calculating the interaction effect between gender and activation. As an example, the results of this group contrast for the 2-increment task are displayed in Fig. 4 (z-score threshold 4.0). With this task, as well as with the 1-increment and the 1-back tasks, men showed statistically signi cantly higher brain activity (volume and intensity) in the right hemisphere, whereas women showed higher activation in the left hemisphere. In contrast, the men showed stronger activation of all the detected brain regions in both hemispheres compared to the women with the 2-back task. The performance data, which were recorded during the fmri scans, showed an increase in reaction time with increasing task dif culty (F ˆ 20.7, p ˆ 2.2e-5, two-way ANOVA). For each task, the female subjects showed longer Fig. 2. The activated brain regions for the two gender groups during the 2-increment task rendered onto a standard brain surface (z-score threshold as shown in Table 2). Only activity within 3 cm below the brain surface is displayed. The activation is lateralized to the right for the male (a) and to the left for the female subjects (b). Vol 11 No 11 3 August

4 O. SPECK ET AL. 2-back 2-increment 1-increment 1-back Female Male lateralization index L Fig. 3. Lateralization indices L for the different tasks and the two gender groups. An index of L ˆ 1 corresponds to left brain activation only and an index of L ˆ 1 represents complete right brain dominance. While the female subjects showed a consistent left brain dominance throughout all the tasks, the male subjects showed bilateral activation or right brain dominance. Fig. 4. Regions of signi cantly stronger activation in the male (red to yellow) and the female (orange to green) subjects for the 2-increment task (gender-activation interaction effect, z-score threshold 4.0). Only regions in the right hemisphere were activated more strongly in men and only regions in the left hemisphere were activated more strongly in women. mean reaction times than the males, although the difference did not reach statistical signi cance. The accuracy decreased with increasing task dif culty (F ˆ 43.3, p ˆ 1e-7, two-way ANOVA) and was signi cantly higher in the female subjects (F ˆ 11.0, p ˆ 0.01). There was also an interaction effect between task dif culty and gender (F ˆ 3.8, p ˆ 0.04), in that the women maintained higher response accuracy with increasing task dif culty than the men. DISCUSSION The amount of brain activation, measured by the number of signi cantly activated volume elements (z-score. 7.0), in all brain regions increases with the work load of the WM tasks as has been observed earlier [7,8]. This implies a direct relationship between the cognitive load or activity and the strength and extent of the signi cant signal changes detected with fmri. Additionally, the increase of the volume of the activated regions with the WM load shows that these regions are activated by the WM task and not due to the imperfectly matched motor component between the task and the baseline condition. The four different WM tasks used in this study produced robust and consistent activation of the LPFC, the PC, the SMA, and the caudate across all subjects. The activation of the LPFC, the PC, and the SMA during WM tasks is in agreement with results of previous studies [1,9± 13], as are the Talairach-coordinates of the activated brain regions (see Table 1). However, to our knowledge, this is the rst report of caudate activation attributed to verbal WM tasks in humans [1]. In humans, caudate activation has been reported to occur with spatial WM tasks [14]. In non-human primates, caudate activation has been detected during spatial and non-spatial WM tasks [15]. The caudate activation reported in our results is likely to be related to the verbal WM component of the task since the activation increased with the task dif culty. In addition, the caudate activation showed the same gender-speci c lateralization as the LPFC and the PC. Our nding of gender differences in lateralization of brain activation during WM tasks (see Table 2) may be due to at least three possibilities: (1) problem solving strategies are different for men and women, even for these simple and straightforward tasks, (2) a difference in the underlying neural substrate, (3) gender-speci c mechanisms regulating the blood oxygenation level dependent (BOLD) effect that leads to the fmri signal change. The BOLD effect is mainly determined by activation-related localized changes in blood ow and to a lesser extent by the blood volume and the oxygen consumption. The resting state cerebral blood ow, for example, has been shown to be different in males and females [16]. However, this nding is unlikely to in uence the hemispheric dominance of brain activity, since the reported differences in resting blood ow are global. Gender differences have been reported in a number of activation studies using fmri. One of the earliest was a report on gender difference in the lateralization of word processing [2]. Contrary to our ndings, word processing is associated with left-lateralized activation of the Broca area in men and bilateral activation in women. Visual stimulation, however, elicited a gender difference in the activation strength, predominantly in the right hemisphere of the occipital cortex [17]. Experiments involving spatial navigation consistently showed recruitment of the right parietal and right prefrontal cortex in women while men activated the left hippocampus, and men were signi cantly faster than woman at the maze task [18]. In this WM study, however, women perform more accurately at the cost of 2584 Vol 11 No 11 3 August 2000

5 GENDER DIFFERENCES IN WORKING MEMORY longer reaction times. One may speculate that a better performance is attributed to a higher specialization of the cortical areas involved leading to a less widely distributed activation pattern and greater asymmetries. In the context of the published data and the results of this study, gender differences in lateralization of brain activation and in performance appear to be highly dependent on the type of task studied and how the performance is measured. Asymmetrical hemispheric recruitment without a distinct gender effect has been shown for various tasks, such as memory encoding and retrieval [19,20], auditory processing [21] and the processing of emotions [22]. However, the results of these studies are contradictory and do not establish a clear pattern of hemispheric differences. One factor contributing to the variability may be that hemispheric interaction improves performance for demanding tasks, as shown in computational simulations [23], and therefore might change with the same task at different levels of dif culty or for different tasks employed. In addition, different gender proportions in the subject populations may contribute to the inconsistencies between the studies. Many factors can in uence the outcome of an fmri study that is based on the BOLD effect. For example, the hemispheric dominance can be in uenced by the handedness of the subjects [24,25]. Additionally, the strength of the activation depends on the age [26,27], the physiological conditions, such as the CO 2 concentration in the blood (pco 2 ) [28], the use of drugs [29], or even the education level [30] of the subjects recruited [31,32]. Therefore, we have carefully matched our subjects with regard to age, handedness, and education, and excluded subjects on medications, with a history of drug abuse, or neurologic and psychiatric illnesses. Our data demonstrate that, in addition to these criteria, subjects participating in fmri studies should be matched or strati ed for gender in order to avoid additional confounds such as gender-speci c functional organization of cerebral lateralization. Gender differences in brain function are most likely determined by multiple biological and environmental factors, such as differences in the cultural environment, education, or exposure to different amounts of sex hormones in utero during early brain development [33,34]. Since sex hormones are present in varying amounts during different stages of brain development, and because the right hemisphere develops earlier than the left [35], gender-speci c developmental hemispheric differences in the neural substrate may also contribute to the gender effect for these working memory tasks. In animal studies the exposure to sex hormones during brain development has been shown to in uence gender-speci c brain organization. The effect may continue during postnatal brain development: castrated neonatal male rats produced female spatial learning patterns in adults while estradiol treatment of neonatal females produced male-typical spatial learning patterns [36]. While these developmental differences do not prove that the gender-speci c performance during certain tasks is the result of differences in the neural substrate, this explanation is consistent with previous observations and seems to be more likely than a change in cortical organization due to strategy differences induced by environmental factors. CONCLUSION Our data demonstrate strong gender differences in the pattern of brain activation associated with WM tasks. This indicates gender differences in problem solving strategies or differences in the underlying neural substrate. Our ndings also emphasize that careful gender matching or strati cation is necessary for brain activation studies. REFERENCES 1. Smith EE and Jonides J. Science 283, 1657±1661 (1999). 2. Shaywitz BA, Shaywitz SE, Pugh KR et al. Nature 373, 607±609 (1995). 3. Miller EN, Satz P and Visscher BR. Neurology 41, 1608±1616 (1991). 4. Friston KJ, Ashburner J, Poline JB et al. Hum Brain Mapp 2, 165±189 (1995). 5. Itti L, Chang L, Mangin JF et al. Hum Brain Mapp 5, 3±17 (1997). 6. Friston KJ, Worsley KJ, Frackowiak RSJ et al. Hum Brain Mapp 1, 214±220 (1994). 7. Braver TS, Cohen JD, Nystrom LE et al. Neuroimage 5, 49±62 (1997). 8. Manoach DS, Schlaug G, Siewert B et al. Neuroreport 8, 545±549 (1997). 9. Mellers JD, Bullmore E, Brammer M et al. Neuroreport 7, 109±112 (1995). 10. Cohen JD, Perlstein WM, Braver TS et al. Nature 386, 604±608 (1997). 11. Courtney SM, Petit L, Maisog JM et al. Science 279, 1347± D'Esposito M, Detre JA, Alsop DC et al. Nature 378, 279±281 (1995). 13. McCarthy G, Blamire AM, Purce A et al. Proc Natl Acad Sci USA 91, 8690±8694 (1994). 14. Postle BR and D'Esposito M. Brain Res Cogn Brain Res 8, 107±115 (1999). 15. Levy R, Friedman HR, Davachi L et al. J Neurosci 17, 3870±3882 (1997). 16. Gur RE and Gur RC. Schizophrenia Bull 16, 247±254 (1990). 17. Levin JM, Ross MH, Mendelson JH et al. Am J Psychiatry 155, 434± GroÈn G, Wunderlich AP, Spitzer M et al. Nature Neurosci 3, 404±408 (2000). 19. Tulving E, Kapur S, Craik KI et al. Proc Natl Acad Sci USA 91, 2016±2020 (1994). 20. Nolde SF, Johnson MK and D'Esposito M. Neuroreport 9, 3509±3514 (199son MK and D'Esposito M. Neuroreport 9, 3509± Belin P, Zilbovicius M, Crozier S et al. J Cogn Neurosci 10, 536± Canli T, Desmond JE, Zhao Z et al. Neuroreport 9, 3233± Belger A and Banich MT. Neuropsychology 12, 380± Dassonville P, Zhu XH, Ugurbil K et al. Proc Natl Acad Sci USA 94, 14015±14018 (1997). 25. Cerf B, Le Bihan D, van der Moortele PF et al. Ann N Y Acad Sci 855, 575± Ross MH, Yurgelun-Todd DA, Renshaw PF et al. Neurology 48, 173±176 (1997). 27. Yamada H, Sadato N, Konishi Y et al. Neuroreport 8, 3775±3778 (1997). 28. Weckesser M, Posse S, Olthoff U et al. Magn Reson Med 41, 213±216 (1999). 29. Levin JM, Ross MH, Mendelson JH et al. Psychiatry Res 82, 135± Miller EN, Bing EG, Selnes OA et al. J Clin Exp Neuropsychol 15, 66 (1993). 31. Mazaux JM, Dartigues JF, Letenneur L et al. J Clin Exp Neuropsychol 17, 71±81 (1995). 32. van Kammen DP, Kelley ME, Gilbertson MW et al. Am J Psychiatry 151, 372±378 (1994). 33. Hutchison JB and Beyer C. Psychoneuroendocrinology 19, 529±541 (1994). 34. Cooke B, Hegstorm CD, Villeneuve LS et al. Front Neuroendocrinol 19, 323± Bracha HS. Schizophr Bull 17, 551±553 (1991). 36. Williams CL, Meck WH. Psychoneuroendocrinology 16, 155±176 (1991). Acknowledgements: This work was supported in part by NIH (DA00280) and by the NSF-sponsored Neuromorphic Engineering Center at Caltech (JB). Vol 11 No 11 3 August