Functional and Anatomical Aspects of Prefrontal Pathology in Schizophrenia

Clinical and experimental research have provided anatomical, pharmacological, and behavioral evidence for a prominent prefrontal dysfunction in schizophrenia. Negative symptoms and behavioral disorganization in the disorder can be understood as a failure in the working memory functions of the prefrontal cortex by which information is updated on a momentto-moment basis or retrieved from long-term stores, held in mind, and used to guide behavior by ideas, concepts, and stored knowledge. This article recounts efforts to dissect the cellular and circuit basis of working memory with the goal of extending the insights gained from the study of normal brain organization in animal models to an understanding of the clinical disorder; it includes recent neuropathological findings that indicate that neural dystrophy rather than cell loss predominates in schizophrenia. Evidence from a variety of studies is accumulating to indicate that dopamine has a major role in regulating the excitability of the cortical neurons upon which the working memory function of the prefrontal cortex depends. Interactions between monoamines and a compromised cortical circuitry may hold the key to the salience of frontal lobe symptoms in schizophrenia, in spite of widespread pathological changes. We outline several direct and indirect intercellular mechanisms for modulating working memory function in the prefrontal cortex based on the localization of dopamine receptors on the distal dendrites and spines of glutamatergic pyramidal cells and on gamma-aminobutyric acid (GABA)ergic intemeurons in the prefrontal cortex. Understanding the interactions between the major cellular constituents of cortical circuits—pyramidal and nonpyramidal cells—is a necessary step in unraveling the receptor mechanisms, which could lead to an effective pharmacological treatment of negative and cognitive symptoms, as well as improved insight into the pathophysiological basis of the disorder. Schizophrenia Bulletin, 23(3):437-^*58,1997. Schizophrenia is commonly considered to be among the most intractable of mental illnesses and among the least comprehensible in terms of neurobiological mechanisms. Heterogeneity marks its phenotypic expression among different individuals, its multiple etiological paths, its fluctuations in different stages of the illness, and its responsiveness to treatment. It is unremarkable then that no one system of the brain nor singular dysfunction has been accepted as central to the pathophysiology of the disease. Yet, as more is learned about the brain and its integrative systems, there is every reason to seek unifying hypotheses that explicate the common threads among sufferers that classify them as having schizophrenia. It seems reasonable to proceed on the assumption that the major symptoms arise from disturbances in different brain areas and that explicit linkages exist between cellular changes in these areas and the information-processing failures that result in psychopathological signs and symptoms. Any full account of schizophrenia should meet the test of comportment with convergent evidence from neurophysiology, neuropsychology, and neuroanatomy. Previously, it has been argued that dysfunction in working memory is a fundamental deficit underlying the cognitive features of schizophrenia and, as such, invokes cellular mechanisms intrinsic to the prefrontal cortex (Goldman-Rakic 1987,1991, 1995a, 19956). The disorganized thought process in schizophrenia patients that manifests itself in idiosyncratic content may be reducible to an impairment of the neural mechanisms by which symbolic representations are both retrieved from long-term memory Reprint requests should be sent to Dr. P.S. Goldman-Rakic, Section of Neurobiology, Yale University School of Medicine, 333 Cedar St., New Haven, CT 06510. 437 at Penylvania State U niersity on Feruary 3, 2013 http://schizophletin.oxfordjournals.org/ D ow nladed from Schizophrenia Bulletin, Vol. 23, No. 3, 1997 P.S. Goldman-Rakic and L.D. Selemon and held "in mind" to guide behavior in the absence of instructive stimuli in the outside world. It is now becoming possible to further specify the cellular mechanisms that underlie working memory, the breakdown of which could lead to psychopathology. In particular, evidence is accumulating to suggest that the modulation of pyramidal cell firing, particularly in prefrontal cortex, may hold the key to an understanding of cognitive dysfunction. The purpose of this article is (1) to review briefly and selectively the association of working memory with prefrontal cortex and outline its relevance to schizophrenia, (2) to examine the neuropathological evidence for involvement of prefrontal cortex in the disorder, and (3) to elaborate on the neurobiological framework that we believe to be essential for further progress in understanding the breakdown of prefrontal circuits that may underlie diseases such as schizophrenia. A few caveats are in order, however, to dispel misconceptions about the working memory/frontal lobe hypothesis. First, cognitive neuroscience and neuropsychology have advanced to the stage where it is possible to be quite specific as to the areal and often the cellular level of analysis. When discussing the frontal lobe involvement in schizophrenia, it is important to designate which portion of this large territory and which of its ascribed functions is under consideration. Since the prefrontal cortex comprises several different areas, each defined by distinct anatomical connections, the range of affective and cognitive symptoms associated with schizophrenia may reflect the involvement to different degrees of one or more subsystems of the medial and posterior orbitofrontal as well as the dorsolateral regions that have recently received more attention. Although in this article we focus on the dorsolateral prefrontal cortex, the cellular findings reviewed below may apply to these other areas as well. Second, neither the working memory hypothesis nor the frontal lobe hypothesis generally requires an obvious or specific type of lesion in the prefrontal cortex, nor does each presume the basic cellular, pharmacological, developmental, or genetic defect in the disease. Finally, it should be very clear that a major role for the prefrontal cortex in schizophrenia necessarily implicates the basal ganglia, thalamus, brainstem, hippocampal formation, and other neocortical areas in the pathophysiology of the disorder. Indeed, volume losses that have been observed in medial and lateral temporal lobe structures have been correlated with frontal lobe impairment (Shenton et al. 1992; Bilder et al. 1995). We have consistently emphasized that the prefrontal cortex carries out its functions through interactions within a complex distributed array of anatomically and functionally related areas (Goldman-Rakic 1987, 1988a, 19886; Selemon and Goldman-Rakic 1988; Goldman-Rakic et al. 1993). Moreover, our recent postmortem study of prefrontal cortex in schizophrenia patients has revealed changes in regions as widespread and functionally diverse as the primary visual and prefrontal cortices (Selemon et al. 1995); these findings add to the considerable evidence for pathological changes in other regions of the schizophrenic brain (see below) and lead us to conclude that the pathological process in this disease does not respect regional boundaries. Thus, a major issue for the field and for advocates of a frontal lobe focus of psychopathology is to examine if and how "frontal" symptoms come to predominate in a condition that may involve widespread pathological changes. Obviously, this is the burden for any pathophysiological theory that assigns predominance to one basic function (e.g., sensory gating or attentional deficits) or brain area (e.g., anterior cingulate, posterior cingulate, superior temporal, hippocampal areas). In the face of such complexity, we rum to the basic anatomy, physiology, and functions of normal model systems as a starting point for approaching the neurobiology of mental disease and for pursuing a rational approach toward its diagnosis and treatments. Prefrontal Cortex, Working Memory, and Schizophrenia There are considerable grounds on which to conceptualize the disorganization in thought and behavior of schizophrenia patients in terms of profound information-processing deficits (Goldman-Rakic 1987, 1991, 1995a, 1995b; Goldman-Rakic and Chafee 1994). Among these, a working memory deficiency compels consideration as one explanation of the behavioral disorganization, features of negative symptoms, and perhaps even some features of positive symptomatology that typify schizophrenia. This section provides a brief review of working memory as a psychological concept and of its linkage to the basic functional and anatomical organization of the prefrontal cortex. Definition and Functional Architecture of Working Memory. The conception of working memory in this and previous articles follows, in most respects, that based on studies of normal human cognition; it encompasses both storage and processing functions that together support the distinctively human capacities of comprehension, computation, and planning (Atkinson and Shiffrin 1968; Baddeley 1986; Newell and Simon 1972; Carpenter and Just 1988). In its barest form, working memory serves as a computational arena or workspace for holding items of information in mind as they are recalled, manipulated, and associated to other ideas and incoming information. 438 at Penylvania State U niersity on Feruary 3, 2013 http://schizophletin.oxfordjournals.org/ D ow nladed from Functional and Anatomical Aspects of Prefrontal Pathology Schizophrenia Bulletin, Vol. 23, No. 3, 1997 Individuals with working memory problems are not amnestic, agnostic, or aphasic, and their sensory and gross motor capacities can be within the normal range. Rather, their problem lies in retrieving and processing information from past experience or keeping in mind a concept, idea, or schema to guide current behavior. Baddeley's (1986) metaphor of a mental "sketchpad" is appropriate for its implication both of limited capacity and of erasability. It is not difficult to imagine how a failure in the retrieval component, the transient storage component, or the erasure mechanism of working memory could result in either impoverished thought processes on the one hand or repetitive or perseverative thought processes on the other. Modular Architecture of Working Memory. Experimental studies in nonhuman primates indicate that there may be multiple working memory domains within the prefrontal cortex, each with its own specialized "central processor" and content-specific storage mechanisms that are organized in distinctly parallel anatomical networks (Goldman-Rakic 1987, 1988a; Selemon and GoldmanRakic 1988; Cavada and Goldman-Rakic 1989a, 19896; Wilson et al. 1993). Visuospatial processing as studied by delayed response tasks relies on the dorsolateral prefrontal convexity both in monkeys (Fuster and Alexander 1971; Kubota and Niki 1971; Friedman and GoldmanRakic 1988, 1994; Funahashi et aJ. 1989; MacAvoy et al. 1991) and in humans (Freedman and Oscar-Berman 1986; Verin et al. 1993), and these same areas are consistently activated as human subjects retrieve visuospatial information from long-term storage or immediate experience through representation-based action (McCarthy et al. 1994; Nichelli et al. 1994; Baker et al. 1996; Gold et al. 1996; Goldberg et al. 1996; Owen et al. 1996; Smith et al. 1996; Sweeney et al. 1996). In contrast, working memory for the features of objects or faces engages anatomically different, more lateral and inferior prefrontal regions in both species (Wilson et al. 1993; Cohen et al. 1994; Adcock et al. 1996; Courtney et al. 19%; McCarthy et al. 1996), and semantic encoding and retrieval as well as other verbal processes engage still more inferior, insular, and anterior prefrontal regions (Paulesu et al. 1993; Raichle et al. 1994; Demb et al. 1995; Fiez et al. 1996; Price et al. 1996). To date, noninvasive imaging of subjects performing working memory tasks has failed to identify one common locus of a central "executive processor" (Baddeley 1986) or contention scheduler (Shallice 1982) that would mediate any and all informational systems. On the contrary, it is now possible to reference specific neuropsychological deficits associated with schizophrenia to specific areas of the prefrontal cortex (for review, see Goldman-Rakic 1996). Nevertheless, one might not expect to find a circumscribed lesion in one and only one area of the prefrontal cortex in psychiatric disease (or any other endemic condition) that would result in a region-specific profile of impairment on the one hand or across-the-board processing deficits on the other. Prefrontal Dysfunction in Schizophrenia. The evidence for the direct dependence of intact cognitive function upon the integrity of the prefrontal cortex in schizophrenia is overwhelming and takes several forms. A persuasive line of support comes from positron emission tomography (PET) of deficient blood flow or metabolism in the prefrontal cortex of schizophrenia subjects, particularly during behavioral performance (Franzen and Ingvar 1975; Buchsbaum and Ingvar 1982; Ariel et al. 1983; Buchsbaum et al. 1984; Farkas et al. 1984; Kurachi et al. 1985; Berman et al. 1986; Chabrol et al. 1986; Guenther et al. 1986; Volkow et al. 1987; Mathew et al. 1988; Wolkin et al. 1988; Paulman et al. 1990; Sagawa et al. 1990; Andreasen et al. 1992; Cohen and O'Leary 1992; Liddle et al. 1992; Goldberg et al. 1996; Weinberger et al. 1996). An equally strong line of evidence derives from neuropsychological and neurophysiological observations that have repeatedly highlighted similarities of impairments observed in patients with schizophrenia and in those with frontal lobe damage (Morihisa et al. 1983; Knight 1984, 1992; Levin 1984; Frith and Done 1988; Williamson et al. 1989; Merriam et al. 1990; Goldman-Rakic 1991, 1995a, 1995ft; Liddle and Morris 1991; Park and Holzman 1992, 1993; Goldberg et al. 1993; Javitt et al. 1993). Patients compromised by frontal lobe lesions and schizophrenia patients are similarly impaired on the Continuous Performance Task (Rosvold 1956; Buchsbaum et al. 1990); on tests of categorization and flexibility, such as the Wisconsin Card Sort Test (Milner 1964; Kolb and Whishaw 1983; Weinberger et al. 1986; Seidman et al. 1991; Franke et al. 1992; Heaton et al. 1993), the Stroop Test (Stroop 1935; Abramczyk et al. 1983; Everett et al. 1989; Schooler et al. 1997), and the Tower of London task (Perret 1974; Shallice 1982; Andreasen et al. 1992); and on oculomotor delayed-response paradigms (Guitton et al. 1985; Fukushima et al. 1988; Hommer et al. 1991; Park and Holzman 1992; Currie et al. 1993; Pierrot-Deseilligny et al. 1993). Although these tasks are formally quite dissimilar, each requires working memory (e.g., keeping a running record of recent events or instructions), and it is this feature that makes them both vulnerable to prefrontal damage in humans and markers of prefrontal dysfunction in patients suffering from schizophrenia or other dementias. Questions for Future Research on Working Memory Dysfunction and Schizophrenia. At least three lines of 439 at Penylvania State U niersity on Feruary 3, 2013 http://schizophletin.oxfordjournals.org/ D ow nladed from Schizophrenia Bulletin, Vol. 23, No. 3, 1997 PS. Goldman-Rakic and L.D. Selcmon research have been opened up by the findings described above. First is the interest in applying PET or functional magnetic resonance imaging (fMRI) to determine the full range of prefrontal areas and prefrontal functions in which blood flow and metabolism may be abnormal. A useful approach would be to determine whether the several working memory domains (spatial, object, and verbal) are differentially vulnerable in schizophrenia and whether patients show diverse patterns of activation that are related to the expression of clinical symptoms. Although visuospatial working memory has been shown to be impaired in schizophrenia subjects, other working memory domains could also be compromised, possibly more severely. Strous et al. (1995) have recently shown that schizophrenia patients are impaired in the ability to match two tones after a 300-ms delay, but are unimpaired when there is no delay, thus providing evidence that schizophrenia patients have deficits in auditory processing when it involves memory guidance and therefore frontal lobe function. Such deficits may be causal for language disturbance. Context-dependent verbal working memory in schizophrenia patients has been correlated with volume loss in dorsolateral prefrontal cortex (Maher et al. 1995). Disturbance in echoic memory (Strous et al. 1995), semantic priming (Spitzer et al. 1993), and sentence completion (Salzinger et al. 1970) may also derive from an underlying deficit in auditory/linguistic working memory and prefrontal dysfunction, but these associations are still mainly conjectural. The use of verbal working memory, dual task paradigms, and other means of examining the temporal parameters of working memory within and across information-processing domains in patients and controls would be particularly illuminating. The recent studies of Currie et al. (1993) and Schooler et al. (1997) exemplify the departure from pure demonstration experiments to an analytic focus on the temporal dynamics, memory spans, and other indices of transient storage capacity in patients versus control populations. Second, research is needed on the relationship between the clinical symptoms expressed by patients in their daily lives and the variety of impairments exhibited in psychologically more delimited performance designed to test working memory processes. If, as has been argued, the prefrontal cortical areas are the places where internalized schema, symbolic representations, and ideas from long-term memory are brought to bear on ongoing events, it is not difficult to imagine that a defect in one of these domains or in any other of the nodes that feed into it could lead to loose associations, scrambled language, disorder thinking, and erratic or apathetic behavior; such a defect could also be highly correlated with formal thought disorder. At present, the evidence on this point is all too limited. The Tower of London task has been correlated with high scores for negative symptoms (Andreasen et al. 1992). Correlations between negative symptoms and MRI spectroscopic evidence of prefrontal pathology have also been reported (Williamson et al. 1991). If the working memory demand in neuropsychological tests and in human cognition generally can be shown to be the common nexus of vulnerability in a disease such as schizophrenia, this functional thread should lead to improvement in diagnosis and possibly to behavioral modification procedures that recognize the loss of "on line" processing ability. Finally, research is needed on the extent to which the cortical areas that are anatomically and functionally interconnected with the prefrontal cortex are functionally disconnected, as described by Liddle et al. (1992), Weinberger et al. (1992), and Friston et al. (1996). Studies on nonhuman primates have provided both anatomical (Goldman-Rakic et al. 1984; Cavada and Goldman-Rakic 1989ft) and functional (Friedman and Goldman-Rakic 1994; Goldman-Rakic and Chafee 1994) evidence for powerful prefrontal regulation of posterior cortical areas. Prefrontal Cortex and Neuropathology Because of the role of the prefrontal cortex in working memory functions and the evidence that working memory deficits are prominent in schizophrenia patients, some sign of compromise of the structural integrity of prefrontal cortex could be expected in this disorder. However, postmortem examination of brains of schizophrenia subjects has generally not revealed gross morphometric alteration in most neocortical areas examined, including the prefrontal cortex. The recent application of quantitative immunocytochemical and histochemical methods to well-documented case material has occasioned a resurgence of interest in postmortem studies, and many differences in cellular pathology between schizophrenic and normal prefrontal cortices have emerged (for recent reviews see Bogerts 1993; Shapiro 1993; Benes 1995). Additionally, modern methods of brain imaging, particularly MRI for analysis of volumetric changes, are providing a rich and invaluable complementary source of data in the brains of living subjects. Postmortem Studies of Prefrontal Cortex Reveal Subtle Pathological Changes. We have recently completed a detailed morphometric analysis of Brodmann's area 9 in the prefrontal cortex in schizophrenia patients and controls using a three-dimensional counting method (Selemon et al. 1995). The stereological method we used is one of several that avoid the counting errors to which 440 at Penylvania State U niersity on Feruary 3, 2013 http://schizophletin.oxfordjournals.org/ D ow nladed from Functional and Anatomical Aspects of Prefrontal Pathology Schizophrenia Bulletin, Vol. 23, No. 3, 1997 traditional methods are prone (Williams and Rakic 1988). Brodmann's area 9 was selected in this initial study because its location is particularly easy to track in different brains. With this new and more sensitive methodology, we observed an abnormally high neuronal density in conjunction with reduced cortical thickness. Layer V is the site of the largest increase in area 9, but neuronal density is also elevated in layers III through VI (Selemon et al. 1995). Similar increases in neuronal density have also been found in prefrontal area 46 (Selemon et al., unpublished observations) and in the primary visual cortex (Selemon et al. 1995). Thus, the morphometric abnormality we have uncovered is widespread and not confined to those regions of the cerebral mantle that have classically been labeled dysfunctional in schizophrenia. Further, the cells that appear most vulnerable in the disease are the larger pyramidal neurons of layer HI (Rajkowska et al. 1995), rather than the smaller cells associated with nonpyramidal morphologies that are compromised in the cingulate cortex (Benes et al. 1991). Neurons in layers III and V project to other cortical areas, whereas those in layer V project, in addition, to the caudate nucleus and superior colliculus. These findings point to a prominence of cortico-cortical circuitry in schizophrenia and are in agreement with structural findings obtained with MRI that report either a frontal lobe volume reduction (e.g., Andreasen et al. 1990a; Breier et al. 1992; Raine et al. 1992; Zipursky et al. 1992; Nopoulis et al. 1995) or a widespread cortical volume reduction that is relatively greater in the prefrontal cortex (Lim et al. 1995; Sullivan et al., in press). An increase in neuronal density in layer V of Brodmann's area 10 was also reported by Benes et al. (1991), and in agreement with our results, Daviss and Lewis (1995) have observed an increased density of calbindin-stained neurons in areas 9 and 46. Enlargement of the cerebral ventricles and of the anterior horns in particular (Brown et al. 1986; Andreasen et al. 1990fo; Shiraishi et al. 1990), and increased sulcal width (Weinberger et al. 1979; Shelton et al. 1988; Friedman et al. 1993) are all supportive of prefrontal volume reduction and indicative of possible cellular atrophy. On the basis of our findings of increased neuronal density and decreased cortical thickness, we have hypothesized that the intraneuronal neuropil—the space between cells—is reduced in the cortex of subjects with schizophrenia (figure 1) (Selemon et al. 1995). This space is packed with the dendritic processes of pyramidal and nonpyramidal neurons and the axonal processes of both intrinsic and extrinsic neurons; this denser cell packing indicates that the meshwork of connectivity in which cortical neurons are embedded may be underrepresented in the brain of schizophrenia subjects. Other constituents Figure 1. Higher cell-packing density as observed in postmortem prefrontal cortex of schizophrenia patients in the study of Selemon etal. (1995)