Behavioral deficits in WM have been observed in both schizophrenia and bipolar disorder, though the disorders maintain distinct profiles with regard to other cognitive and clinical symptoms (Green 2006
). To our knowledge, no prior imaging studies have examined both populations simultaneously to determine whether there are neurophysiological differences in brain activation between groups while performing a WM task. We found that despite similar WM task performance, controls showed increased BOLD activation in DLPFC compared to both schizophrenia patients and euthymic bipolar patients. The difference in activation was significant between the controls and schizophrenic subjects. Bipolar subjects fell intermediate to both groups in level of DLPFC activation, and were not significantly different from either.
In accordance with our hypothesis, we found increased activation in the DLPFC and proximal frontal regions in controls compared to schizophrenia patients. The DLPFC is an integral part of the frontostriatal circuitry thought to participate in WM (Manoach, et al. 2000
), and is involved in prefrontal networks thought to be the basis of the “central executive” component of WM. Researchers report both hyperactivation (Callicott, et al. 2000
; Manoach, et al. 2000
) and hypoactivation (Ragland, et al. 2007
; Scheuerecker, et al. 2008
) of this region in schizophrenia patients during WM tasks. Though visual WM is generally associated with activation in the right DLPFC, our findings of overall bilateral DLPFC activation, with left hemisphere lateralization in the group contrast, may be related to symbolic or linguistic rather than image-based encoding of stimuli in this task (Ungerleider, et al. 1998
In both the within and between group analyses, bipolar subjects had an average activation pattern of DLPFC in between the schizophrenic and control groups ( and ). In contrast to the schizophrenia group, the bipolar group did not differ significantly from controls with regard to activation in DLPFC or other prefrontal regions, but showed hypoactivation of some spatially-diffuse areas of visual cortex compared to controls. This could indicate differences between the groups in the encoding and maintenance strategies used in the “visuospatial sketchpad” component of WM, which is consistent with reports of bipolar-related behavioral impairments in visuospatial WM tasks (Bearden, et al. 2001
). Reductions in BOLD activation in the visual cortices of bipolar patients have also been seen in non-WM related studies (Pavuluri, et al. 2007
), though differences in stimulus conditions and subject populations make it difficult to compare results. We predicted less pronounced prefrontal hypoactivation in the bipolar group, and thus may have been underpowered to detect more subtle effects in PFC. Sub-threshold reductions in prefrontal activation were observed in the bipolar group average activation, and overall magnitude and pattern of activation was more similar to the schizophrenia group than to the control group (). Importantly, another (unpublished) study investigating working memory using an n
-back task in a sample of bipolar patients overlapping with those in the current study found decreased activation in bipolar patients compared to controls in right DLPFC specifically, regardless of mood state (Townsend, et al. 2008
), which suggests that differences may be both subtle and task-specific.
We observed unexpected increases of brain activation in left motor regions in schizophrenia compared to bipolar subjects. This finding, however, does not appear suggestive of group differences in WM, since effects appeared restricted to primary motor areas. Instead, this effect could reflect group differences related to other task demands. For example, regional changes identified in motor cortices may reflect differences in information processes associated with the motor responses for memory or baseline condition stimuli, or could relate to degree of dextrality—though the latter explanation may be unlikely given handedness scores were similar across groups. Finally, manipulation of the button box may have influenced results, despite the inclusion only of right-handed subjects and attempts by the researchers to control for the hand/fingers used to operate the button box.
Discrepancies in the directionality of DLPFC findings might result from differences in performance and/or training on the task, the type of WM task, and whether visual or spatial WM is being tested. For example, schizophrenia and bipolar patients perform significantly worse than controls on the Wisconsin Card Sort Test (WCST) (Altshuler, et al. 2004
; Heinrichs and Zakzanis 1998
), but because this test incorporates task switching and requires sustained attention from the participant, discrepancies in performance or functional activation cannot be attributed to impairments in WM alone (Manoach 2003
). Interpretations must account for the type of task used and additional requirements for task execution including attention, manipulation, and maintenance. Manoach et al. (2003)
also suggested that greater spatial heterogeneity of DLPFC activation in schizophrenia may contribute to increased findings of hypoactivation in this region, especially in studies averaging across subjects. This hypothesis was corroborated by Anticevic and colleagues (2008)
, who showed greater WM-related signal preservation in schizophrenia patients relative to controls after employing surface-based registrations of structural and functional data that control for heterogeneity in cortical folding, which is predicted to be greater in patients. Thus, it is possible that individual differences in brain structure may manifest as functional variation and reduce statistical power, especially in diseased populations known to possess alterations in brain structure. Larger structural variation in schizophrenia, which may be less pronounced in bipolar disorder, could similarly have influenced the patterns of group-related DLPFC activation observed in this study.
In general, DLPFC hypoactivation appears most prevalent in fMRI studies, but it is also suggested that fMRI signal in the DLPFC relates to WM load in an inverted-U shaped curve, with hypoactivation occurring when WM load has exceeded capacity (Callicott, et al. 1999
). Models of WM deficits in schizophrenia in nonhuman primates have shown similar inverted-U shaped functions relating performance to dopamine D1 receptor stimulation, where dysregulation of dopamine in the PFC may relate to observed symptoms (Robbins 2005
). Schizophrenia-related differences of DLPFC activation in either direction may also be explained by an inefficient recruitment of this cortical network. Potkin et al. (2009)
suggest that hypofrontality may reflect poorer performance in schizophrenic subjects, or could reflect that WM capacity has been exceeded for high load tasks. This hypothesis is supported by observations of schizophrenia-related DLPFC hyperactivation during a Sternberg Item Recognition WM task, although schizophrenia patients also exhibit peak DLPFC activation at a lower WM load than controls. A meta-analysis by Van Snellenberg, et al. (2006)
further supports the notion that WM capacity may be reached earlier in schizophrenia. Specifically, investigators showed that there was a shift in the inverted-U relationship between WM performance and DLPFC activation in schizophrenia (as suggested by Callicott et al. 1999
), where schizophrenia patients reached peak activation at lower task difficulties compared to controls. They do note, however, that in cases of comparable performance in schizophrenia patients and controls, it is difficult to interpret differences in DLPFC activation. Consequently, they suggest that hypofrontality in patients may be related to the use of non-DLPFC cortical networks to perform WM tasks. Since task load was not manipulated in the current investigation, we were unable to address whether WM capacity may differ in bipolar and schizophrenia patient groups, although this may be a focus of future studies. Still, it is clear that dysfunctions in either direction (hypo- and hyperfrontality) are detrimental to performance.
Despite significant alterations in BOLD activation during the WM task, accuracy was comparable between groups. This was expected, given the practice requirements. Furthermore, the difficulty load allowed more unambiguous interpretation group differences, as there was no “ceiling effect” to suppress observable differences in accuracy. Alterations in BOLD activation were seen irrespective of performance, though variations in accuracy within groups could mask subtle differences across individuals.
Imaging studies in psychiatric populations are sensitive to several unique confounds, including global deficits in attention or IQ, the modulating factors of antipsychotic drugs, and potential increases of head motion. We excluded subjects with head movement artifacts (>1mm), and carefully registered each brain to a standard template. Prior behavioral and imaging research (Manoach 2003
; Phillips, et al. 2008
; Scheuerecker, et al. 2008
) suggests antipsychotic medication has little measurable effect on WM in bipolar disorder and schizophrenia. It remains possible, however, that medications could have contributed to increased variance in our sample. It is clear that active psychosis is not required for the manifestation of cognitive deficits in bipolar disorder (Robinson, et al. 2006
). We cannot rule out the contribution of other agents not included in our screening (e.g. nicotine, caffeine, or alcohol) to observed changes in the BOLD signal. In addition, although we expect that the within-group activations observed in visual cortices are attributable to the use of less visually complex stimuli for the control task, we cannot rule out that this aspect of the study design might also affect brain activation outside of visual cortices.
Although our sample sizes were larger than those typically included in other schizophrenia/bipolar functional imaging studies (Abler, et al. 2004
; Lagopoulos, et al. 2007
; Manoach, et al. 2000
), it is possible we may have been underpowered to detect subtle prefrontal deficits in the bipolar group. Though the percent signal change in the DLPFC in the bipolar group appeared to be intermediate to the control and schizophrenia groups (), it was not significantly different from either of the other two groups. Thus, our study does not provide direct support for prefrontal dysfunction in the role of WM impairments in bipolar disorder, although clearly indicates that schizophrenia patients exhibit more severe dysfunction with respect to controls in cortical networks serving WM. Finally, discrepancies in our findings and the findings of others may be due in part to testing different WM sensory modalities, testing manipulation versus retention, and other variables regarding the behavioral testing itself. The ICBM WM task shows robust and reliable activation in areas related to visual WM, making this paradigm well suited to study differences in WM function between diagnostic groups.