Temporal Lobectomy Evidence for the Role of the Amygdala in Early Emotional Face and Body Processing

Discussion

In this paper we investigated how two groups of patients with TLR comprising the amygdala and in the majority of instances the hippocampus and temporal pole encoded facial and bodily expressions using time-resolved EEG data. This approach yielded a number of key insights. For readability, we have included labels that pertain to significant temporal and spatial epoch clusters. First, the comparison of emotional facial expressions revealed significant differences in the neural processing of fearful and neutral faces for healthy controls from 140 to 170 ms over temporoparietal and occipital electrodes (N170) and then at 200–350 ms over temporoparietal and 240–350 ms over occipital electrodes (posterior P2). A similar but more restricted pattern of activity was observed for the clinical group that had undergone left TLR, with temporoparietal and occipital differences from 140 to 170 ms (N170) and occipital differences to emotional faces from 200 to 350 ms (posterior P2). In contrast, no significant neural differences were found for fearful compared with neutral faces in these time ranges for the right TLR group consistent with the findings from Framorando et al. (2021).

Second, the comparison of emotional bodily expressions revealed that, for healthy controls the first difference in the neural processing of fearful and neutral bodies was observed at 90–120 ms over parieto-occipital electrodes (P1), followed by anterior-central differences to fearful and neutral faces from 250 to 350 ms (anterior P2). Notably, the earliest emotion-based bodily difference seen in the control group, was absent in both TLR groups. Rather, the right TLR group showed a different emotion-based neural response to bodies at 200–250 ms over parieto-occipital electrodes and 300–350 ms over anterior central electrodes (posterior P2 and anterior P2). Meanwhile, the only difference observed for the left TLR group was at 200–250 ms over anterior central electrodes (anterior P2). This pattern of results indicates that, while the right amygdala is crucial for the initial processing of emotional faces, unilateral TLR impairs the earliest stages of emotional body perception, regardless of the hemisphere that is targeted in surgery.

While the MUA revealed that emotion-driven patterns appeared to differ between TLR and healthy groups, the typical ERP amplitude analysis did not indicate any between group interactions. The typical main effects of emotion were present for both face and body tasks (Stekelenburg and de Gelder, 2004; van Heijnsbergen et al., 2007; Hinojosa et al., 2015). However, the only between group differences evident were at mid-late processing stages for the facial task, between healthy and LTLR groups. While both groups showed emotion-related differences in amplitude, the overall amplitude to faces of both emotions for the LTLR group was dampened at 200–240 and 240–350 ms. We note, however, that these results are based on a small clinical group, and, thus there may not be sufficient power to detect meaningful interactions.

Considering the processing of emotional faces, for healthy controls the first emotion-based difference matched the time window and parieto-occipital region corresponding to the N170 (Hinojosa et al., 2015). The activation of this component is expected because the N170 is well established as the first face-specific component that is modulated by facial expression (Hinojosa et al., 2015). In the current data, this was followed by sustained emotion-based differential activation over temporoparietal and occipital cortices from ∼200 ms onward on the second positive peak. This temporo-posterior P2 has been linked to higher-stage cognitive processes (Campanella et al., 2002), such as stimulus evaluation and the mental representation of the emotional content which can contribute in a task-dependent fashion to adaptive behaviors. Therefore, the results from the control group are broadly consistent with the current scientific record and provide a healthy baseline to compare emotional face processing post-TLR.

Here, we show the complete absence of emotion-based activation in the right TLR group (from 90 to 350 ms). The absence of the earliest N170 ms deflection for the right TLR group is followed by the lack of any mid-stage differences. This indicates that the resected regions play a critical role in contributing to the initial encoding of relevant emotional information in faces. The amygdala responds rapidly to emotional faces (within ∼70 ms; Inagaki et al., 2023; Méndez-Bértolo et al., 2016; Weidner et al., 2022) and continually contributes via feedback and feedforward connections. The absence of an N170 modulation by emotion is likely a reflection of an absence of amygdala input to face-selective right hemispheric cortical regions. Other TLR research shows such impaired early emotion modulations following right resection (Rotshtein et al., 2010; Mielke et al., 2022; Kissler et al., 2023). The absence of a subsequent differential response to facial expressions builds on previous research showing that subcortical bodies modulate extrastriate cortical responses to fearful faces (Morris et al., 1996; Vuilleumier et al., 2004; Liu et al., 2022). Further, previous studies of patients with right TLR have indicated that the loss of the right amygdala reduces the response of the right fusiform lingual gyrus to faces (Reisch et al., 2022a) and the response of the visual cortex to affective scenes (Reisch et al., 2022b). By extension, the present findings indicate that the resected areas appear crucial for early affective face processing and uniquely contributes to mid-stage processing via cortical connections (Kapp et al., 1994; Armony et al., 1998; Cahill and McGaugh, 1998; Anderson and Phelps, 2001; Amaral et al., 2003; Hung et al., 2010; Vrticka et al., 2013; Kohno et al., 2015). Indeed, amygdala sclerosis is related to absent cortical activity to faces as typically seen in healthy populations (Vuilleumier et al., 2004). It is likely that later emotion-driven differences would emerge for the right TLR group, as retained emotion recognition is often reported (Reisch et al., 2022a; Kissler et al., 2023), and emotion differences comparable with controls are seen in ERPs from 400 ms onward when viewing affective pictures (Mielke et al., 2022).

The maintenance of early and mid-stage emotional-based differences in the left TLR, but absent in the right TLR population, highlights two interesting points: firstly, that there is a functional lateralization for emotional face processing, and secondly, that the initial stage of structural face encoding is seemingly necessary for the occurrence of mid-stage processes. Regarding the first point, this aligns with a large body of literature showing lateralized-right dominance in face processing (Kanwisher et al., 1997; Bukowski et al., 2013; Corrow et al., 2016). The present findings indicate that this hemispheric dominance extends to facial emotion processing (Gainotti, 2012) or that the structural encoding facilitated by the right amygdala is crucial for subsequent cortical emotion processing (Burton et al., 2003; Bertini et al., 2019). On the second point, the presence of later emotion differences in the left TLR group (posterior P2 from 200 ms onward) following the spared effect at N170 could indicate that facial emotion processing is hierarchical with later stages gated by earlier ones (Bruce and Young, 1986; Haxby et al., 2000).

It has been suggested that emotion and in particular threat are encoded automatically—a process posited to be reliant on the amygdala (Compton, 2003; Öhman, 2005; Phelps and LeDoux, 2005). However, task demands such as the relevance of the face or emotion could impact the extent to which the amygdala is recruited in visual perception. Intracranial EEG evidence has previously suggested that the amygdala is responsive to the emotional content of faces (from 130 ms) regardless of task relevance (Hadj-Bouziane et al., 2012). This has been corroborated by EEG studies showing task independent responses to emotional faces at early stages (Schupp et al., 2003, 2006; Pourtois et al., 2010; Schindler and Kissler, 2016; Schindler et al., 2020). In contrast, other electrophysiological recordings have suggested that mid-latency scalp (Weidner et al., 2022) and amygdala (Krolak-Salmon et al., 2004; Pourtois et al., 2010) responses to emotion are modulated by task relevance. A systematic review of the topic reported that task relevance did not alter emotion modulations at early ERP components such as EPN and N170 but did at later components like P3 and LPP (Schindler and Bublatzky, 2020; also see Krolak-Salmon et al., 2001). The implication is that bottom-up and top-down processes may interact with each other and that the amygdala could have bidirectional connections with attention-related cortical regions (Vuilleumier, 2005), as well as the ventral visual system (Morris et al., 1998; Pessoa et al., 2002; Krolak-Salmon et al., 2004). These bidirectional connections are particularly relevant to the present study because the face task was split into two subsamples, one an emotion-salient task (emotion recognition) and the other a perception/memory task (one-back). Both tasks necessitated sustained attention and perceptual encoding but varied in task demands and thus may have recruited the amygdala differently via different connections. Research has indicated that in a healthy sample, amygdala activity would be evoked to a greater degree during emotion recognition compared with a one-back task, due to the salience of the emotional content. However, this does not undermine our findings, as task type was matched across all groups (i.e., the same proportion of the LTLR, RTLR, and healthy samples completed both tasks). The combination of the two tasks in the present study therefore can be used to examine differences in affective processing postresection. Further research will be required to examine the impact of resection on the processing of emotional stimuli when task relevance varies.

Considering the processing of emotional bodies, for controls we observed the earliest emotion-based differences emerging over occipital electrodes at 90 ms (P1), followed by anterior-central activation from ∼250 to 350 ms (anterior P2). The occipital activation from 90 to 120 ms matches P100 visual activation (Batty and Taylor, 2003). P100 reflects visual encoding as well as aspects of attention. In healthy groups, P100 emotion modulations have been found elsewhere for bodies (van Heijnsbergen et al., 2007), followed by mid-late stage emotion modulated fronto-central negativity corresponding to our later emotion-related negative deflections on P2 (Stekelenburg and de Gelder, 2004). Importantly the earliest emotion-differential response was absent in both TLR groups, suggesting that the P100 is reflective of more than just the processing of low-level contributions. The absence of this early-stage difference in both groups suggests that the earliest neural response to bodies requires bilateral medial temporal activation, while contralateral activation or cortical activation is sufficient for later responses to bodies. Importantly, bodies have been shown to be represented in the primate temporal cortex—a site resected in the majority of the clinical group (Vogels, 2022). It is thus likely that temporal body-specific contributions are impaired in those cases. For both TLR groups later stage differences at P2 were present, which could reflect intact mid-late stage cognitive processes such as appraisal or judgment. This indicates that later stage processes that inform cognition do not rely on the initial neural response to bodily expressions.

Although studies of emotion frequently use static stimuli, it has been claimed that images of human figures may imply motion and as such might activate areas sensitive to biological motion in the brain. Without motion, the body processing network is thought to involve the STS and the extrastriate body area (de Gelder, 2006), with the processing of bodily expressions additionally activating the amygdala and fusiform gyrus (Hadjikhani and de Gelder, 2003). Emotional body expressions, in comparison with neutral representations, can indicate implied motion such as approach or avoidance. Implied motion activates medial temporal and medial superior temporal cortices (Kourtzi and Kanwisher, 2000). ERP research investigating lateralization of emotional body processing has found stronger right lateralized effects for emotional content, and stronger left effects for implied movement (Borhani et al., 2015), potentially explaining the required bilateral amygdala contribution for the early P1 difference.

The present results speak to the multiplexed contribution of the medial temporal areas to emotion processing. While it is evident that face and body emotion processing activate different networks (de Gelder et al., 2004; van de Riet et al., 2009), these networks have shared components, such as the amygdala (van de Riet et al., 2009). However, it has proven difficult to characterize the unique contributions of the amygdala to face and body emotion processing. We argue that by comparing the pattern of results from different groups of TLR patients, we can distil similarities and differences in the face and body processing networks. For example, our results demonstrate that medial temporal activation is necessary for early emotion-based electrophysiological differences for both categories.

While all clinical participants underwent amygdala resection, often adjacent areas were also affected. For the facial task, the additional resected sites included portions of the temporal lobe in 14 out of 18 clinical participants and the hippocampus in 14 out of 18 clinical participants. For the body task, the additional resected sites included portions of the temporal lobe in 14 out of 16 clinical participants, and the hippocampus in 8 out of 16 clinical participants. Temporal lobe epilepsy has been linked to deficits in fear-specific emotion recognition (Nineuil et al., 2023). However, these deficits are more pronounced than in cases of lateral temporal lobe epilepsy, suggesting that medial temporal structures like the amygdala play a unique role in fear recognition (Nineuil et al., 2023). Interestingly, hippocampal volume is positively associated with the speed of facial emotion identification (Szymkowicz et al., 2016). Therefore, it remains possible that these additional resections may have directly or indirectly impacted some of our results. One point of contention related to their pathologies is that individuals with epilepsy may be subject to reorganization of the neural networks in the medial temporal region involved in emotion processing. Indeed, recurrent seizures can lead to sclerosis or impaired development at the focal site and surrounding areas. Epilepsy-related sclerosis has been related to the degree of electrophysiological impairment in emotional face processing (Rotshtein et al., 2010). The age of seizure onset is related to the degree of damage, and impaired behavioral emotion recognition has been found for those with earlier but not later onset ages (McClelland et al., 2006). While amygdala activation is still evident for fearful faces in groups with medial temporal lobe epilepsy (Schacher et al., 2006), the lateralization of this activation can shift to the hemisphere contralateral to the focal seizure origin (Riley et al., 2015) and must be acknowledged when considering this clinical group. In this way, it is also possible that some differences observed between the face and body groups—which used different participant subsets—may have been due to variations in seizure-related damage or functional reorganization across individuals. For this reason, it is imperative that future research includes measures of the age of seizure onset as this could account for both functional variation between individuals and discrepancies in the lobectomy literature. Additionally, the assessment of both behavioral and neural measures pre- and postoperatively would provide an understanding of baseline processing and the functional outcomes related to surgical intervention. The present results shed new light on the complex role of medial temporal structures in emotion processing. Our results reveal the unique roles of these structures at the earliest emotion-sensitive epochs for face and body processing. Additionally, the findings demonstrate that activity in the unilateral right resected area is necessary for distinguishing different facial expressions, while contralateral activation is necessary for the earliest emotional body processing. Although these results alone cannot distinguish between the importance of different medial structures and temporal cortices, there have been previous indications that the amygdala is particularly important to these processes (Cecere et al., 2014; Reisch et al., 2022a). The difference in neural activation in response to faces and bodies following TLR indicates that while faces and bodies are processed by similar brain networks, there are key points of divergence that can function independently.