مدولاسیون عاطفی آگاهی حرکتی در ناتوانی در ادراک بیماری برای همی پلژی: شواهد رفتاری و ضایعه

Abstract The possible role of emotion in anosognosia for hemiplegia (i.e., denial of motor deficits contralateral to a brain lesion), has long been debated between psychodynamic and neurocognitive theories. However, there are only a handful of case studies focussing on this topic, and the precise role of emotion in anosognosia for hemiplegia requires empirical investigation. In the present study, we aimed to investigate how negative and positive emotions influence motor awareness in anosognosia. Positive and negative emotions were induced under carefully-controlled experimental conditions in right-hemisphere stroke patients with anosognosia for hemiplegia (n = 11) and controls with clinically normal awareness (n = 10). Only the negative, emotion induction condition resulted in a significant improvement of motor awareness in anosognosic patients compared to controls; the positive emotion induction did not. Using lesion overlay and voxel-based lesion-symptom mapping approaches, we also investigated the brain lesions associated with the diagnosis of anosognosia, as well as with performance on the experimental task. Anatomical areas that are commonly damaged in AHP included the right-hemisphere motor and sensory cortices, the inferior frontal cortex, and the insula. Additionally, the insula, putamen and anterior periventricular white matter were associated with less awareness change following the negative emotion induction. This study suggests that motor unawareness and the observed lack of negative emotions about one's disabilities cannot be adequately explained by either purely motivational or neurocognitive accounts. Instead, we propose an integrative account in which insular and striatal lesions result in weak interoceptive and motivational signals. These deficits lead to faulty inferences about the self, involving a difficulty to personalise new sensorimotor information, and an abnormal adherence to premorbid beliefs about the body.

1. Introduction Neurological disturbances of body awareness provide a useful way of investigating the bodily self; a fundamental facet of self-consciousness (Gallagher, 2000). Anosognosia for hemiplegia (AHP; i.e., the denial of motor deficits contralateral to a brain lesion) is a prototypical example of a disturbance in body awareness. AHP occurs more frequently following right perisylvian lesions, and less often following left-hemisphere perisylvian lesions (Cocchini, Beschin, Cameron, Fotopoulou, & Della Sala, 2009). AHP can take various clinical forms, ranging from blatant denial of limb paralysis and associated delusional beliefs to milder forms of motor unawareness (see Fotopoulou, 2014, Jenkinson et al., 2011 and Marcel et al., 2004). Although the exact aetiology of AHP remains debated, the clinical variability of AHP suggests that it is a multifaceted and heterogeneous phenomenon (Fotopoulou, 2014, Marcel et al., 2004, Orfei et al., 2007 and Vocat et al., 2010). Accordingly, explanations have varied from selective deficits in motor planning, to multi-factorial accounts involving both basic sensorimotor and higher-order cognitive deficits (see Fotopoulou, 2014 and Jenkinson and Fotopoulou, 2010 for reviews). These cognitive deficits have been associated with either particular lesion sites such as the premotor cortex (Berti et al., 2005) and the insula (Karnath, Baier & Nagele, 2005), or involvement of a more varied pattern of cortical and subcortical regions and their connections (Fotopoulou et al., 2010, Moro et al., 2011 and Vocat et al., 2010). One facet of AHP that has received less empirical attention, despite a long history of clinical observations and theoretical debates (Bisiach and Geminiani, 1991 and Weinstein and Kahn, 1955), is the role of emotional factors. On clinical examination, patients typically manifest some degree of blunted affect or ‘indifference’ for their paralysis and its consequences. This indifference (anosodiaphoria, Babinski, 1914) can exist with or without concomitant explicit denial of deficits. On the contrary, depressive symptoms and ‘catastrophic reactions’ (sudden influx of strong, negative feelings and related behaviours; Goldstein, 1939) are encountered rarely. Moreover, there are some clinical indications that as unawareness decreases over time, depressive symptoms begin to emerge in patients who were previously emotionally unresponsive towards their paralysis (Besharati et al., 2014, Fotopoulou et al., 2009 and Kaplan-Solms and Solms, 2000). Exceptionally, some patients with or without explicit denial of deficits have been noted to show a strong hatred towards their paralysed limbs (misoplegia; Critchley, 1974), or a disproportionate exasperation with irrelevant, minor disappointments, despite their apparent indifference for their paralysis (Fotopoulou and Conway, 2004, Kaplan-Solms and Solms, 2000 and Weinstein and Kahn, 1950). Some authors have argued that this lack of affect, or misattribution of negative emotions, is caused by purely psychogenic ‘defence’ mechanisms. According to the now classic theory of Weinstein and colleagues (e.g. Weinstein, 1991 and Weinstein and Kahn, 1955), denial and related premorbid coping mechanisms prevent patients from explicitly acknowledging their paralysis, and self-attributing the associated negative emotions. Alternatively, this lack of emotional reactivity has been considered to be the direct consequence of damage to the right (frontal) hemisphere, regarded by some authors as specialised for the processing of negative, withdrawal-related emotions (Davidson, 2001; see Gainotti, 2012 for review). However, neither of these two approaches has been fully supported by empirical evidence. Specifically, the psychodynamic account of AHP fails to explain the relative neuroanatomical and behavioural specificity of anosognosic behaviours (Bisiach and Geminiani, 1991 and Heilman and Harciarek, 2010). The ‘valence’ hypothesis has similarly not been supported in the literature; although patients with AHP do typically score lower than control patients in self-report measures of depression and anxiety (e.g., Fotopoulou et al., 2010), more sensitive investigations have shown that they do not differ from controls groups in their ability to experience such emotions (Turnbull et al., 2005 and Vocat et al., 2010). They also show appropriate, negative emotional reactions to their deficits when the latter are evoked implicitly (Fotopoulou et al., 2010 and Nadrone et al., 2007). Thus, it appears that the relation between AHP and emotion is more complex than suggested by either the psychodynamic or the valence hypothesis. More generally, such rigid distinctions between purely psychodynamic and neurocognitive explanations have been challenged recently (Fotopoulou, 2012) and integrative accounts of AHP have been put forward (Fotopoulou, 2010, Turnbull et al., 2005, Turnbull and Solms, 2007 and Vuilleumier, 2004; see also Turnbull, Fotopoulou & Solms, 2014). According to such theories, complex imbalances between cognition and motivation may be caused directly by damage to insular, striatal, or limbic regions that have recently been found to be selectively associated with AHP (Fotopoulou et al., 2010, Moro et al., 2011 and Vocat et al., 2010). For example, Vuilleumier and colleagues have suggested that damage to the basal ganglia may obstruct the “discovery” of deficits, as patients have reduced affective drive to respond to errors and revise beliefs based on new perceptual evidence (Vocat et al., 2012, Vuilleumier, 2000 and Vuilleumier, 2004). Similarly, within a computational framework, Fotopoulou and colleagues have suggested that insular and basal ganglia damage may lead to weak and imprecise signals about the physiological condition of one's body. This leads to aberrant ‘top-down’ inferences about bodily states, and difficulties in affectively personalising new sensorimotor information (Fotopoulou, 2014). Taken together, these accounts suggest that the lack or misattribution of negative emotions in AHP relates to impairments in higher-order cognition, rather than to primary deficits in emotional processing. This ‘top-down’ perspective is consistent with a relatively neglected facet of AHP, namely, the fluctuations of awareness based on the emotional or social context in which awareness is probed. For instance, Kaplan-Solms and Solms (2000), see also Ross and Rush, 1981, Starkstein and Robinson, 1988 and Turnbull et al., 2002 have shown that when themes of loss are explored during psychotherapeutic sessions – particularly when such loss is apparently unrelated to their disabilities – transient awareness and depressive episodes can be experienced by patients that are otherwise stably anosognosic. Marcel et al. (2004) have further shown that awareness may increase in some patients when they are asked about their disabilities in an emotional, conspiratory manner, or from the perspective of the examiner (see also Fotopoulou, 2014 and Fotopoulou et al., 2009). Notwithstanding the theoretical interest of these observations, to our knowledge there is no systematic, experimental investigation of the moderating role of emotional and social context in AHP. Accordingly, we aimed to investigate the relation between emotion and motor awareness in AHP. To this end, we recruited right-hemisphere stroke patients with AHP and control patients without AHP, and assessed motor awareness before and after providing positive and negative feedback about performance on a standardised cognitive test (the Hayling Test; Burgess & Shallice, 1997). The task includes components of varied difficulty that we could match with the valence of the provided feedback to generate realistic conditions of positive and negative feedback. Moreover, it is unrelated to motor abilities so we could test the role of emotion on motor awareness, uncomplicated by ‘bottom-up’ sensorimotor signals and the patients' explicit or implicit feelings about their motor abilities. Based on the idea that patients with AHP have lost the ability to use signals from their own body to make related inferences about their current bodily state (Fotopoulou, 2014; see also above), our main aim was to test whether the ‘top-down’ experimental induction (by verbal, social feedback) of negative feelings about oneself could improve awareness of one's motor disabilities. We expected patients with AHP to show increased awareness of their deficits following negative feedback compared with positive feedback, while such effects were not expected in the control group. Furthermore, in order to ensure that the experimental feedback had induced the desired emotions in patients, we measured patients' self-reported emotional state following each condition of the main task. If patients with AHP were capable of experiencing negative emotions, we expected negative feedback to lead to more negative feelings than positive feedback in both patient groups. Lastly, we examined whether lesions to critical cortical (premotor and the insular cortex) and subcortical (basal ganglia and limbic structures) areas would be associated with increased unawareness scores, as in previous studies (Berti et al., 2005, Fotopoulou et al., 2010, Karnath et al., 2005 and Moro et al., 2011). Contrary to such lesion subtraction investigations, however, we used a voxel-based lesion-symptom mapping (VLSM) approach (Bates et al., 2003 and Rorden et al., 2007). This advanced method characterises the statistical relationship between tissue damage and behaviour on a voxel-by-voxel basis, regardless of the classification of patients into categorical groups, or implementing a cut-off for pathology (Bates et al., 2003). We also used this method to identify the brain regions associated with a change in motor awareness induced by our experimental task, which according to our hypothesis should include the insular cortex and basal ganglia structures (Fotopoulou, 2014; see also above). While the first clinico-anatomical correlation has been investigated before in the literature, to our knowledge, only two previous studies have investigated the association between behaviour on carefully-controlled experimental conditions and neuroanatomical data (Fotopoulou et al., 2010 and Moro et al., 2011), and no study has examined this association in relation to emotion.

3. Results 3.1. Demographic and neuropsychological results Patients' demographic characteristics and their performance on standardised neuropsychological tests are summarised in Table 1. The groups did not differ significantly in terms of age, education or symptom onset to assessment interval. As expected, there was a significant difference in awareness scores between the AHP and HP groups on both the Berti et al. (1996) interview [t(14) = 5.60, p = .00] and the Feinberg et al. (2000) scale [t(14) = 7.06, p = .00]. The groups showed similar sensory deficits, as well as similar impairments in general cognitive functioning, abstract thinking, reasoning abilities and neglect. Although both groups showed deficits in proprioception, the AHP group was significantly more impaired [t(12) = 2.33, p = .04]. The AHP group showed significantly lower scores for depression on the HADS when compared to controls [t(14) = 3.06, p = .01]. This difference was taken into account in subsequent analyses. Table 1. Groups' demographic characteristics and neuropsychological profile. AHP(n = 8) HP(n = 8) t-Test Mean SD Mean SD t df p Age (years) 71.63 16.18 64.75 12.14 .96 14.00 .35 Education (years) 11.88 1.81 12.63 1.92 .68 14.00 .51 Days from onset 11.13 11.26 14.38 10.56 .60 14.00 .56 MRC Left upper limb .25 .46 .38 .52 .51 14.00 .62 MRC left lower limb .63 .92 1.00 1.07 .75 14.00 .46 Premorbid IQ-WTAR 41.50 7.79 33.00 7.62 1.41 10.00 .19 Berti awareness interview 1.63 .52 .25 .46 5.60 14.00 .00* Feinberg awareness scale 6.31 2.17 .63 .69 7.06 14.00 .00* Orientation 2.88 .35 3.00 .00 1.00 7.00 .35 Digit span forwards 5.63 1.19 6.13 .99 .91 14.00 .38 Digit span backwards 2.88 .83 3.38 1.30 .91 14.00 .38 MOCA memory 3.75 .89 4.17 .98 .83 12.00 .42 MMSE 22.20 6.02 25.00 2.16 .88 7.00 .41 Visual fields 4.29a 1.89 3.57a 1.99 .69 12.00 .50 Somatosensory (max 6) 3.38a 1.41 3.00a 1.60 .50 14.00 .63 Proprioception (max 9) 3.71 2.21 6.57 2.37 2.33 12.00 .04* Comb/razor test left 4.75 4.13 5.25 2.60 .29 14.00 .78 Comb/razor test right 12.63 5.10 10.63 2.97 .96 14.00 .35 Comb/razor test ambiguous 5.88 1.96 4.13 2.42 1.59 14.00 .13 Bisiach one item test .75 .46 .38 .52 1.53 14.00 .15 Line crossing right 11.50 6.44 16.25 2.05 1.99 8.41 .08 Line crossing left 6.75a 8.14 10.00a 8.68 .77 14.00 .45 Star cancelation right (omissions) 13.75 6.11 11.00 6.19 .89 14.00 .39 Star cancelation left (omissions) 21.25a 10.43 18.88a 10.86 .45 14.00 .66 Copy .50a .76 1.00a 1.07 1.08 14.00 .30 Representational drawing .25 .46 .50 .53 1.00 14.00 .33 Line bisection right .43a .53 .38a .52 .20 13.00 .85 Line bisection centre .57a .53 .75a .46 .69 13.00 .50 Line bisection left .38a .52 .50a .53 .48 14.00 .64 Cognitive estimates 16.71a 4.86 15.50a 2.26 .56 11.00 .59 FAB total score 11.40a 2.70 13.50a 2.51 1.43 11.00 .18 HADS depression 2.88 2.70 8.00a 3.89 3.06 14.00 .01* HADS anxiety 5.13 3.00 7.25 4.89 1.05 14.00 .31 Berti awareness interview = Berti et al. (1996); Feinberg Awareness scale = Feinberg et al. (2000); MRC = Medical Research Council (Guarantors of Brain, 1986); MOCA = The Montreal Cognitive Assessment (Nasreddine et al., 2005); Comb/razor test = tests of personal neglect (MacIntoch, Brodie, & Beschin, 2000); Bisiach one item test = test of personal neglect; Visual fields and somatosensory = customary ‘confrontation’ technique = Bisiach, Vallar, & Perani (1986); line crossing, star cancellation, copy & representational drawing = conventional sub-tests of Behavioural Inattention Test (Wilson, Cockborn & Halligan, 1987); FAB = Frontal Assessment Battery (Dubois et al., 2000); HADS = Hospital Anxiety and Depression scale (Zigmond & Snaith, 1983). *Significant difference between groups, p < .05. a Scores below tests' cut-off points, or more than 1 SD below average mean. Table options 3.2. Main experimental results: awareness change A linear regression analysis revealed a significant main effect for the factor Group (b = 2.04, SE = −.45, p < .001, 95% CI = 1.16; 2.92), with the AHP group showing a greater change in awareness (marginal mean = .99) compared with the HP group (marginal mean = −.02). Also, a significant main effect of Emotion induction type (b = −1.07, SE = .46, p = .019, CI = −1.96; −.18) was observed, with awareness change being significantly greater following the negative (marginal mean = 1.6) compared with the positive emotional induction (marginal mean = −.57). The interaction between Emotion induction type and Group was also significant (b = −2.05, SE = .61, p = .001, CI: −3.26; −.84; see Fig. 1), with the AHP group (marginal mean = 2.55) showing a greater change in awareness compared with the HP group (marginal mean = .75) following the negative emotional induction only. Taking the HADS depression scores into account in this analysis did not change the pattern of these results. Marginal means and interquartile range (error bars) of the change in awareness ... Fig. 1. Marginal means and interquartile range (error bars) of the change in awareness for the AHP (dark grey bars) and HP (light grey bars) groups after the positive and negative emotional induction: *p < .05. The Y-axis indicates the change in awareness scores analysed by calculating the difference in awareness scores between each condition (post minus pre) for each group. Positive scores indicate an increase in awareness (i.e., less anosognosia) and negative scores indicate a decrease in awareness (i.e., more anosognosia). Figure options A qualitative example of the change in motor awareness observed as a result of the emotion induction is described here. During the pre-awareness assessment one patient stated “No, I have no weakness anywhere, no”, claiming that “I can move my arm, no problem” and was adamant that she raised her left arm and clapped her hands. Following the negative emotion induction, the same patient admitted that her left arm “is not as strong as before the stroke”, saying “I don't think I can move this arm now, it feels weak”. When asked if she can tie a knot, she replied “I'm not so sure now” and after attempting the action, she observed “no, I can't do that.” 3.3. Emotional state induction To investigate whether patients experienced a change in their emotional state following the positive and negative induction respectively, we examined the main effects of Emotion (positive vs negative feedback) and Group (AHP vs HP) on emotion ratings. The regression analysis confirmed a main effect of Emotion (b = 1.83, SE = .439, p < .001, CI: .97; 2.69) with patients giving significantly lower emotion ratings (i.e., reporting feeling less happy) following the negative emotional induction (marginal mean = 2.17) compared with the positive emotional induction (marginal mean = 3.83). The model also showed that the factor Group significantly predicted emotion ratings (b = .99, SE = .49, p = .046, CI: .019; 1.97), with AHP patients showing more positive emotion ratings (marginal mean = 3.41) compared with right-hemisphere controls (marginal means = 2.59). However, there was no significant interaction between the factors induction type and group (b = −.33, SE = .64, p = .6, CI: −1.59; .93; see Fig. 2). Marginal means and interquartile range (error bars) of emotion ratings for AHP ... Fig. 2. Marginal means and interquartile range (error bars) of emotion ratings for AHP (Dark grey bars) and HP (light grey bars) groups after positive and negative mood induction: *p < .05. The Y-axis indicates the patient's subjective mood ratings on a scale from zero to five (0 = very unhappy; 5 = very happy). Figure options 3.4. Baseline awareness scores A Wilcoxon Signed Rank Test revealed that there was no significant difference between pre-awareness scores of the positive (median = 2) and of the negative condition overall (median = 3, Z = −.27, p = .82, r = .067). This applied also to the AHP group (Z = −.9, p = .563, r = .23) and the HP group (Z = −.7, p = .75, r = .18), in respective, separate analyses. 3.5. Performance on the Hayling Test Analysis of the Hayling Sentence Completion Test using a Mann–Whitney U test showed no significant difference between total scaled scores of the AHP and HP groups (Z = −1.14, p = .28, r = .29). According to the tests norms, overall scaled scores indicated that the AHP group's performance was ‘low average’ (median = 4), while the HP group's performance was ‘moderate average’ (median = 5). Similarly, there was no difference found in Hayling part 1 (Z = −.9, p = .42, r = .23), with the scaled score for completion time being ‘low average’ for the AHP group (median = 4) and ‘moderate average’ for the HP group (median = 5). This again applied to Hayling part 2, with no difference found between groups in their total scaled score for completion time (Z = −.4, p = .8, r = .1) and response errors (Z = −1.1, p = .31, r = .28), with the AHP group performing ‘average’ for time (median = 6) and ‘abnormal’ for response errors (median = 1.5). Similarly, the HP group performed ‘average’ for time (median = 6) and ‘abnormal’ for responses errors (median = 2) (see Supplementary Materials). Therefore, the feedback given was realistic based on patients' actual performance, with both groups performing better on part 1 than on part 2, and showing no differences between groups on either part. 3.6. Reverse order control condition The two AHP patients who performed the experiment in the reverse order showed the same pattern of results as found in the main group analysis. After the negative emotion induction, both patients showed a greater improvement in awareness (AHP09: mean = 5, AHP10: mean = 3.5) compared to the control group (mean = .5; SD = .82; AHP09: t(7) = 5.13, p = .001, r = 5.49; AHP10: t(7) = 3,42, p = .007, r = 3.66). There was no difference between either AHP patient and the HP control group in awareness change following positive emotion induction (AHP09: t(7) = .45, p = .33, r = .48; and AHP10: t(7) = 1.7, p = .07, r = 1.81). 3.7. Specificity of effect control condition The three patients with right-hemisphere damage who performed this additional control experiment showed no change in personal neglect assessments, and a minor change in visuospatial neglect, with extrapersonal neglect becoming slightly worse following negative versus positive induction in two patients. Additionally, there was a non-mood specific improvement in awareness of neglect in one patient. The results are summarised in 3 case reports below (see Supplementary Materials Table S2 for a summary of results). Patient HP09 presented with no AHP, no personal neglect, no visuospatial neglect except on the ‘copy’ subtest, and mild unawareness of drawing neglect. There was no change in visuospatial and personal neglect, or awareness of drawing neglect following the positive and negative emotion induction condition. Patient HP10 presented with no AHP, mild personal neglect, visuospatial neglect and unawareness of drawing neglect. She showed no change in the line bisection subtest, personal neglect scores, and general questions for awareness of drawing neglect, but a small increase in visuospatial neglect following the positive and negative emotion induction conditions. There was also a small increase in awareness of drawing neglect following the negative emotion induction, but a much larger increase in awareness following positive induction. Lastly, patient AHP11 presented with AHP, personal neglect, visuospatial neglect and mild unawareness of drawing neglect. There was no change in her personal neglect and awareness of drawing neglect scores, and no change in her performance on the line bisection subtest following the negative and positive emotion inductions. There was a small increase in visuospatial neglect (star cancellation subtest) following the negative but not positive emotion induction.” 3.8. Follow-up awareness testing Wilcoxon signed rank tests showed that there was no significant difference in Feinberg awareness scores before and after the experiment, in either the AHP (Z = −.45, p = .66, r = .12) or HP group (Z = −1.63, p = .1, r = .42), suggesting that the observed awareness changes were temporary and experimental effects, rather than permanent, clinical changes. 3.9. Lesion analysis All lesions resulted from a first-ever unilateral stroke, mainly within the right middle cerebral artery territory. Group-level percentage lesion overlay for the AHP group (n = 8) identified the involvement of cortical and subcortical areas, comprising the inferior and superior frontal gyri, the pericentral cortex, the insula and insula ribbon, and the internal capsule (see Fig. 3A). In comparison, the lesion overlap map for the HP group (n = 7) revealed a more focal lesion pattern involving mainly subcortical regions (see Fig. 3B). Lesion volume (defined by number of voxels) was not significantly different between the AHP group (mean = 37132.5, SD = 43782.65) and the HP group (mean = 25997.14, SD = 33536.03; t (15) = .55, p = .594). The lesion subtraction map identified mainly the anterior and posterior insular ribbon, the posterior basal ganglia, and dorsal pericentral areas to differ between the groups (see Fig. 3C). Group-level lesion overlay maps for patients with anosognosia for hemiplegia ... Fig. 3. Group-level lesion overlay maps for patients with anosognosia for hemiplegia (AHP) and controls. A. Overlay of lesions in patients with anosognosia (AHP; n = 8); B. Overlay of patients without anosognosia (n = 7). C. Statistical analysis comparing the two populations of patients (AHP present-AHP absent; results are corrected for multiple comparisons, p < .05 for Z > 1.3). Figure options VLSM analysis using the continuous Feinberg awareness scores, revealed that voxels within the posterior insula, the supramarginal, the angular and superior temporal gyrus (SMG, AG and STG), internal capsule, pericentral gyri, and the inferior frontal gyrus (IFG) were significantly associated with differences in awareness (p < .05) (see Fig. 4A). Similar results were found when co-varying lesion size. Additionally, VLSM analysis, looking at the experimental change in awareness scores (i.e., differential scores following negative emotional induction only), without and with co-variation of lesion size, identified significant voxels (p < .05) within the anterior arm of the internal capsule, the anterior insula, the anterior lateral putamen with a lateral extension into the external capsule and an additional region in the dorsal anterior periventricular white matter (likely to contain limbic white matter connections) (see Fig. 4B). Voxel-based (topological) lesion-deficit analysis. A. Damaged MNI voxels ... Fig. 4. Voxel-based (topological) lesion-deficit analysis. A. Damaged MNI voxels predicting the severity of unawareness of symptom deficits when co-varying for lesion size (Feinberg scale, inverted, continuous measure; p < .05 for Z > 1.6449). B. Damaged MNI voxels predicting the change in awareness (differential scores, pre and post mood induction) when co-varying for lesion size (continuous measure; p < .05 for Z > 1.6449). PrC = precentral, PoC = postcentral, SMG = supramarginal, STG + superior temporal gyrus, IFG = inferior frontal gyrus, IC = internal capsule, MFG, middle frontal gyrus.