Unravelling the neurocognitive mechanisms underlying counterconditioning in humans

Fear conditioning, counterconditioning, reward, ventral striatum, ventromedial prefrontal 20 cortex 21

200 words 22 Main text (excl. abstract, methods, references, figure legends): 5517 23 N Figures: 6 24 N Tables: 1 25 Supplementary Information 26 27 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 2

28 Counterconditioning (CC) aims to enhance extinction of threat memories by establishing new 29 associations of opposite valence. While its underlying neurocognitive mechanisms remain largely 30 unexplored, previous studies suggest qualitatively different mechanisms from regular extinction. In 31 this functional MRI study, participants underwent categorical threat conditioning (CS+/CS-: images of 32 animals/tools), followed by either CC ( CS+ images reinforced with monetary rewards , n=24) or 33 regular extinction (n=24). The following day, we assessed spontaneous recovery of threat responses 34 and episodic memory for CS+ and CS - category exemplars. While the ventromedial prefrontal cortex 35 (vmPFC) was activated during regular extinction, participants undergoing CC showed persistent CS+ -36 specific deactivation of the vmPFC and hippocampus, and CS+ -specific activation of the nucleus 37 accumbens (NAcc). The following day, p hysiological threat responses returned in the regular 38 extinction group, but not in the CC group. Counterconditioning furthermore strengthened episodic 39 memory for CS+ exemplars presented during CC , and retroactively also for CS+ exemplars presented 40 during the threat conditioning phase. Our findings confirm that CC leads to more persistent 41 extinction of threat memories, as well as a ltered consolidation of the threat conditioning episode. 42 Crucially, we show a q ualitatively different activation pattern during CC versus regular extinction, 43 with a shift away from the vmPFC and towards the NAcc. 44 45 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 3

46 Trauma-related disorders are prevalent and highly detrimental to the individual’s quality of life 1. To 47 treat these disorders, patients undergo exposure therapy in a safe therapeutic environment , causing 48 threat responses to fade away 2. Although exposure therapy may be successful initially, relapse often 49 occurs and is the most pr evalent remaining challenge in optimizing treatment efficacy . Research 50 suggests that exposure therapy creates a safety memory that competes for expression with the 51 original threat memory 3,4, suggesting that relapse may occur because of relatively weak learning and 52 retention of the safety memory . Therefore, identifying mec hanisms that can be used to strengthen 53 safety learning is a key step in advancing treatment for trauma -related disorders. A promising 54 approach to strengthen safety learning is to create a new, positive association with the event that 55 was previously linked to an aversive outcome. However, while there are indications that establishing 56 positive associations can prevent relapse, the underlying mechanism s are poorly understood (for a 57 review, see 5). 58 To study threat responses in a controlled setting, aversive Pavlovian conditioning is typically used. A 59 neutral stimulus (conditioned stimulus, CS; e.g., a picture) is coupled with a biologically aversive 60 unconditioned stimulus (US; e.g., an electrical shock), after which the CS alone also elicits a 61 conditioned threat response. Conditioned threat responses to the C S can be attenuated using 62 extinction, during which the CS is repeatedly presented in absence of the US . However, early theories 63 have suggested that threat responses may more easily be inhibited by engaging appetitive systems 6,7. 64 Indeed, experiments provide evidence that coupling a CS to a p ositive US after threat conditioning, a 65 process known as aversive -to-appetitive counterconditioning (CC), may be superior to regular 66 extinction. Specifically, CC compared to regular extinction was associated with a faster attenuation of 67 learned threat responses6,8, stronger decreases in threat expectan cy9,10, and more positive valence 68 ratings of the CS11-13 immediately post-CC. 69 Tests for spontaneous recovery, reinstatement, and renewal can subsequently be used to evaluate 70 the return of threat responses over time, after unsignaled presentation of the US, or in a novel 71 context, respectively 3,14. Thereby , one can investigate whether CC persistently attenuates threat 72 responses. While early rodent studies showed that CC may be prone to the same relapse as 73 extinction15,16, recent neurobiological work in rodents showed that CC can enhance the activation of 74 an amygdala-striatal pathway, which is also recruited during extinction – albeit to a lesser degree –, 75 and that CC compared to regular extinction can reduce the return of threat responses17. Recent 76 studies suggest that CC may diminish the return of threat responses in humans as well. Specifically, it 77 was shown that CC compared to regular extinction reduced renewal of previously learned food -78 allergy associations when presented in a novel context one day later 18. Counterconditioning 79 compared to regular extinction was also associated with reduced recovery of arousal and shock 80 expectancy the following day9,19, as well as reduced reinstatement9. 81 Extinction learning appears to be mediated by activation of the ventromedial prefrontal cortex 82 (vmPFC), which inhibits the expression of threat responses by suppressing amygdala activity 20-23. 83 When extinc tion is enhanced by replacing aversive with novel, neutral outcomes , the vmPFC was 84 found to be engaged more effectively than during standard extinction 24. When extinction is 85 enhanced by replacing aversive outcomes with a reward (counterconditioning), evidence in rodents 86 suggests stronger engagement of the ventral striatum , a region known to be involved in the 87 anticipation and receipt of reward25. However, human studies only provide indirect evidence for such 88 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 4 a mechanism since involvement of the ventral striatum could only be shown during spontaneous 89 recovery19 or during reinstatement 26, but not during CC itself. Although it was observed that brain 90 areas of the fear network are reduced during CC versus regular extinction in humans 19, it is unclear 91 how this difference is achieved. Therefore, although evidence suggests that CC is more effective than 92 regular extinction in preventing the return of threat responses the neural mechanisms are not well 93 understood yet. It remains unclear whether CC is a form of enhanced extinction that is mediated by 94 enhanced engagement of extinction networks , including the vmPFC , or whether it is driven by 95 engagement of reward networks. 96 To investigate the qualitative differences between CC versus regular extinction further, category 97 conditioning can be used, a procedure in which conditioned threat responses are learnt by coupling a 98 US to conceptually linked exemplars that together form a category (e.g. , pictures of animals )27. It 99 allows for the typical measures of threat condition ing, but also provides the opportunity to probe 100 episodic memory for the CS category exemplars 28. When episodic memory was probed 24h after CC 101 and extinction, it was shown that memory for CS+ stimuli that had undergone CC was stronger than 102 memory for CS+ stimuli that had undergone regular extinction 29. This suggests that compared to 103 regular extinction, CC can enhance episodic memory consolidation and potentially provide stronger 104 retrieval competition against a threat memory. 105 To investigate the neural mechanisms that distinguish CC from regular extinction and to establish 106 whether CC is indeed associated with a memory that is qualitatively different from the safety 107 memory established during regular extinction , we performed a two -day fMRI study comparing CC 108 versus regular extinction in a between -subjects design (Figure 1A). Participants underwent category 109 conditioning and subsequently underwent either aversive-to-appetitive CC (CC group) or regular 110 extinction (Ext group ; Figure 1 B-C). During the CC task, participants in the CC group obtain ed 111 monetary rewards depending on how quickly they responded to a cue superimposed on novel 112 category exemplars from t he CS+ category, a procedure similar to the monetary incentive delay 113 (MID) task30. To maximize task similarity between tasks and groups, the cued-response element was 114 kept consistent in all tasks (acquisition, CC/extinction, spontaneous recovery, reinstatement), but 115 response-time conting ent monetary rewards were only present during the CC task (Figure 1F ). To 116 assess the potential of CC v ersus regular extinction in persistently attenuating the expression of 117 threat response , we tested retrieval of the threat memory and reinstatement of threat responses 118 one day later (Figure 1D). Episodic memory for exemplars of the CS categories that were presented 119 during threat conditioning and CC/extinction was assessed by means of a surprise memory test. To 120 characterize pupil dilation responses (PDRs) and skin conductance responses ( SCRs) during the 121 anticipation of shock - and reward-reinforcement independently from prior conditioning , a separate 122 valence-specific response characterization task was included at the end of the experiment ( Figure 123 1E). 124 In line with previous results 9,19, w e hypothesized that CC compared to extinction would lead to a 125 more persistent attenuation of threat responses . As indicated above, t his could be mediated by two 126 possible neural mechanisms: either through enhanced engagement of extinction networks, reflected 127 by increased engagement of the vmPFC , or through a shift towards reward networks, reflected by 128 activation of the ventral striatum. Based on previous results 29, we expect ed stronger episodic 129 memory for CS exemplars presented during CC, whereas regular extinction would not show such a 130 strengthening effect. 131 132 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 5 133 134 Figure 1. Overview of the experimental design . (A). Participants were assigned to the counterconditioning (CC) or 135 extinction (Ext) group. On day 1, participants performed two blocks of acquisition of category -conditioned threat responses 136 separated by a 30 second break, followed by CC or extinction. Day 2 consisted of a spontaneous recovery test, a 137 reinstatement procedure and test, an item memory test and a valence -specific response characterization. Valence and 138 arousal ratings for the different categories were taken before or after the tasks as indicated by ‘V+A Rating’. All tasks were 139 performed in an MRI scanner. (B) During acquisition, participants viewed trial -unique exemplars of objects and animals. 140 Exemplars of one category (CS+ animals or objects counterbalanced) were paired with a shock in 50% of trials . CS- trials 141 were not reinforced. (C) Participants in the CC group could earn a monetary reward if they responded quickly enough to 142 exemplars in the CS+ category. (D) Participants in the Ext group underwent extinction. During the ex tinction task, recovery 143 test and reinstatement test, neither CS+ nor CS - exemplars were paired with a shock . (E) In the valence -specific response 144 characterization task, participants viewed three different coloured squares. One colour was associated with sh ock (CS+S), 145 one colour with reward (CS+R) and one colour served as CS -. The trial structure was otherwise identical to the acquisition 146 and CC tasks. (F) In all Pavlovian tasks, trial onset was marked by presentation of a unique category exemplar. After a 147 variable interval, a ring appeared, to which participants were instructed to respond as quickly as possible. Upon response, 148 the ring shifted in colour as response confirmation . In the acquisition task, shocks could occur 0.5-1.5-seconds after the 149 response w indow had elapsed (indicated as ‘pre -shock). The category exemplar and cue remained visible 1 second after 150 potential shock administration (indicated as ‘post -shock’). During CC, participants received visual feedback for 2 seconds 151 (+€0.50 approximately the fastest 70% of trials, +€0.00 on other trials). During the other tasks, participants viewed neutral 152 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 6 feedback (three dots). Trials were separated by an 8-10 s intertrial interval, during which a fixation cross is displayed in the 153 centre of the screen. 154

155 In the valence-specific response characterization task, we observed that both threat and reward-156 anticipation induced strong arousal -related PDRs and SCRs (see Supplementary Information ). 157 However, PDRs allowed for a better differentiation of the two compared to the CS- (Supplementary 158 Figure 1A). Therefore, we focus ed on PDRs in all analyses and refer to the Supplementary 159 Information for details on the analysis of SCRs. During the acquisition task, both groups showed 160 comparable and successful acquisition of differential conditioned threat responses ( PDR means ± SD: 161 CC CS+=1.085±.030, CC CS -=1.054±.033, Ext CS+ =1.084±.050, Ext CS -=1.050±.035; for PDR, SCR and 162 fMRI results see Supplementary Information). 163 Extinction and aversive-to-appetitive counterconditioning 164 After threat acquisition, participants in the CC group underwent CC, while participants in the Ext 165 group underwent regular extinction. Across both groups and pha ses (early vs. late), we observed 166 retention of conditioned differential PDRs (CS-type (CS+, CS -) x Phase (Early, Late) x Group (CC, Ext) 167 rmANOVA, main effect CS-type: F(1,34)=15.393, p<0.001, η²=0.312, Figure 2A), as well as a decrease in 168 PDRs over the course of the task (main effect phase: F( 1,34)=10.121, p=0.003, η²=0.229). These 169 findings are in contrast to our expectation of a CS-type x Phase x Group interaction. Specifically, we 170 expected differential PDRs to become extinguished in the Ext group , while being sustained in the CC 171 group, potentially due to increased reward anticipation. Extinction in the Ext group however already 172 occurred during the early phase (paired t -test, early CS+ vs. CS -, p=0.233), and differential responses 173 did not change to wards the late phase (p=0.979) . As a result, we found distinct differential 174 conditioned PDRs throughout the CC/extinction task between groups ( CS-type x Group interaction : 175 F(1,34)=6.053, p=0.019, η²=0.151) , with participants undergoing CC showing stronger PDRs to CS+ vs. 176 CS- category exemplars (paired t-test average CS+ vs. CS -, t(20)=3.602, p=0.002, CS+: 1.07 ±0.04, CS-: 177 1.04±0.04), whereas differential PDRs were extinguished in participants undergoing extinction 178 (paired t-test average CS+ vs. CS -, p=0.246, CS+: 1.05±0.04, CS-: 1.04±0.04). Results of t he valence-179 specific response characterization task showed that differential PDRs can also be indicative of 180 anticipation of reward (Supplementary Figure 1A). Thus, while PDRs in the Ext group indicated that 181 differential conditioned threat responses were successfully extinguished, differential PDRs persist ed 182 in the CC group, likely reflecting reward anticipation. Differential SCRs persisted during the late phase 183 Figure 2. Differential PDRs during CC/extinction and explicit ratings of arousal and valence provided after the counterconditioning or extinction phase. (A) Differential PDRs for the early (light red) and late (dark red) phase of counterconditioning (CC, solid bars) or extinction (EXT, open bars). Participants undergoing CC showed increased differential PDRs as compared to participants undergoing extinction. (B) Arousal and (C) valence ratings displayed separately for part icipants assigned to the counterconditioning (CC, solid bars) and extinction (EXT, open bars) groups. Participants that had undergone CC gave stronger differential arousal scorings than participants that had undergone extinction. In addition, participants that underwent CC showed flipped differential valence ratings: while valence differential valence ratings were negative after extinction, the direction reversed to positive differential ratings after CC. Error bars represent ± standard error of the mean . *=p<0.05, **=p<0.01, ***=p<0.001, ≠ indicates that the bar is significantly different from 0. .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 7 of both CC and extinction but were no longer detectable in the last two trials and were comparable 184 between groups (see Supplementary Information). 185 Valence and arousal ratings provide further support for the extinction of differential responses in the 186 Ext group and positive, reward -induced arousal for CS+ items in the CC group (Figure 2B-C). 187 Differential valence ratings for the CS+ and CS - differed between groups after the CC/extinction task 188 (CS-type (CS+, CS-) x Group (CC, Ext) rmANOVA, CS-type x Group interaction: F(1,44)=12.054, p=0.001, 189 η²=0.215). Participants in the CC group rated CS+ stimuli more positive than CS - stimuli (t(21)=3.469, 190 p=0.002, CS+: 7.5 ±0.30, CS -: 5.41±0.38) , while participants in the Ext group gave both categories 191 similar valence ratings (p=0.245, CS+: 5.63±0.32, CS -: 6.21±0.28). Differential arousal ratings for the 192 CS+ and CS- also differed between groups (CS-type (CS+, CS-) x Group (CC, Ext) rmANOVA, CS-type x 193 group interaction : (F( 1,44)=20.862, p<0.001, η²=0.322). Participants in the CC group reported higher 194 arousal levels for the CS+ category than for the CS - category (t(21)=6.370, p<0.001, CS+: 6.64 ±0.20, 195 CS-: 3.45±0.38) while participants in the Ext group gave similar arousal ratings for the CS+ and CS - 196 categories (p=0.290, CS+: 4.21±0.43, CS -: 3.80±0.40). Taken together, more positive valence and 197 higher arousal ratings for the CS+ in the CC group as compared to the Ext group further support the 198 interpretation of increased differential PDRs reflecting arousal induced by reward anticipation. 199 CC prevents differential spontaneous recovery 200 To investigate whether CC prevented the spontaneous recovery of differential conditioned threat 201 responses, we compared PDRs in the last two trials of the CC/extinction phase and the first two trials 202 of the spontaneous recovery test in a CS -type (CS+, CS -) x Group (CC, Ext) x Phase (last two trials of 203 CC/extinction, first two trials of the spontaneous recovery test) rmANOVA. We exp ected the E xt 204 group to show an increase in PDRs from the extinction task to the spontaneous recovery task, while 205 we expected PDRs for the CC group to remain stable or decrease . Critically, differential spontaneous 206 recovery of PDRs differed between groups ( Group x CS-type x Phase interaction : F( 1,28)=6.329, 207 p<0.018, η²=0.184, Figure 3). While the CC group showed a decrease in differential PDRs from CC to 208 spontaneous recovery (t(14)= -1.807, p=0.046, one -tailed, CC: 0.34 ±0.2, spon taneous recovery : -209 0.01±0.18), the Ext group show ed an increase in differential PDRs (t(14)=1.850, p=0.043, one -tailed 210 significance, extinction: 0.11±0.01, spon taneous recovery : 0.04±0.02). To conclude , while we 211 observed differential spontaneous recovery in the Ext group, we did not find evidence for differential 212 spontaneous recovery in the CC group , suggesting that CC attenuated the recovery of threat-213 responses compared to regular extinction. 214 However, since participants undergoing CC showed persistent differential PDRs during the last two 215 trials of the CC phase, while participants undergoi ng extinction did not, we additionally explored 216 whether there was differential responding during the first two trials of the spontaneous recovery 217 test. During the first two trials of the spontaneous recovery test, participants in the CC group showed 218 decreased differential PDRs as compared to the Ext group (CS-type (CS+, CS -) x Group (CC, Ext) 219 rmANOVA, CS-type x Group interaction: F(1,29)=3.901, p=0.029, one -tailed, η²=0.119). Further 220 exploration within the groups confirmed that participants in the CC group did not show retention of 221 differential responses ( paired t-test, CS+ and CS - responses during the first two trials of the 222 spontaneous recovery test, p=0.219, one-tailed), while the Ext group did show increased responses 223 to the CS+ as compared to the CS - (t(14)=1.958, p=0. 035, one -tailed). Thus, both the differential 224 spontaneous recovery of PDRs between sessions, and differential responding within the firs t two 225 trials of the spontaneous recovery test suggest ed that CC prevented spontaneous recovery of 226 differential responses compared to extinction. SCRs did not show differential recovery and were 227 comparable between groups (see Supplementary Information). 228 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 8 229 Figure 3. Differential PDRs during the last two trials of extinction (grey) and the first two trials of the spontaneous 230 recovery test (dark red). Differential PDRs show selective spontaneous recovery after extinction (Ext group , open bars) but 231 not after CC (CC group , solid bars ). During the first two trials of the spontaneous recovery test , differential PDRs are 232 increased in the Ext group as compared to the CC group. Insets show PDRs to the CS+ (red) and CS- (blue) during the last 233 two trials of extinction an d the first two trials of the spontaneous recovery test. While the Ext group shows differential 234 responding during the spontaneous recovery test, the CC group does not. Error bars represent ± standard error of the 235 mean. *=p<0.05, #=p<0.05 one-tailed significance. 236 CC also appeared to have lasting beneficial effects on valence ratings compared to extinction. At the 237 start of the second testing day, differential valence ratings continued to differ between groups (CS -238 type (CS+, CS -) x Group (CC, Ext) rmANOVA, CS -type x Group interaction: F( 1,44)=5.160, p=0.028, 239 η²=0.105). While participants in the CC group gave similar valence ratings to both categories 240 (p=0.179, CS+: 6.3 ±0.34, CS -: 5.4±0.35), participants in the Ext group gave more negative valence 241 ratings t o the CS+ category than to the CS - category (t(23)= -1.964, p=0.031 one -tailed test, CS+: 242 5.5±0.30, CS-: 6.3±0.24), also illustrative of relapse of threat associations. 243 Surprisingly, while participants in the CC group showed heightened differential arousal ratings 244 immediately after CC as compared to ratings from participants who had undergone extinction (Figure 245 2B), p articipants in both groups gave comparable differential arousal ratings at the start of the 246 second day immediately before the spontaneous recovery test (CS-type (CS+, CS -) x Group (CC, Ext) 247 rmANOVA, main effect of CS -type: F( 1,44)=10.932, p=0.002, η²=0.022, CS+: 4.8±0.28, CS -: 3.9±0.24). 248 Likewise, response times to the CS+ and CS - during the first two trials of the spontaneous recovery 249 task were similar across both groups (all p’s>0.2). These findings may suggest that differential arousal 250 evoked by the categories was similar in both groups immediately before and during the spontaneous 251 recovery test. 252 The spontaneous recovery test was followed by a reinstatement procedure, consisting of three 253 unsignaled shocks, and a reinstatement test. However, mean PDRs decreased from spontaneous 254 recovery to reinstatement (t(30)=3.063, p=0.005, last two trials of spontaneous recovery: 1.04 ±0.01, 255 first two trials of reinstatement: 1.01±0.01). Given that we did not observe successful reinstatement 256 in either group, our reinstatement test was not informative on whether CC can lead to a more 257 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 9 persistent attenuation of threat responses as compared to regular extinction. A full description of 258 PDR and SCR results of the reinstatement test can be found in the Supplementary Information. 259 Distinct CS-type specific activation for extinction and appetitive counterconditioning 260 After acquisition, the CC group underwent appetitive CC, while the Ext group underwent regular 261 extinction. Whole brain analysis revealed that over the course of this task, CS-type specific activation 262 changed differe ntially between the two groups in a large cluster encompassing multiple regions in 263 the medial temporal lobe (Group x CS -type x Phase interaction, cluster size = 1760 mm3, p=0.034, 264 whole-brain FWE-corrected, Figure 4B and Table 1). We further investigated the anatomical location 265 of the cluster using our ROIs to probe for activity and fou nd that the effect encompassed the 266 amygdala. To further investigate the interaction effect in the amygdala, we extracted parameter 267 estimates from the complete bilateral amygdalae (Automated Anatomic Labelling, AAL, atlas in the 268 WFU PickAtlas toolbox in MN152 space) and performed post -hoc comparisons. In the early phase, 269 CS-type specific responses differed between the groups (t(1,44)=2.173, p=0.035, CC: 0.18±0.08, Ext: -270 0.073±0.08). Specifically, the CC group showed increased amygdala activation to the CS + as 271 compared to the CS- (t(23)=2.210, p=0.037) while that was not the case in the Ext group (p=0.39 0). In 272 the late phase, differential responses were comparable between the groups (p=0.503). 273 Whole-brain analysis further revealed a number of clusters showing distinct CS-specific activations 274 between groups throughout the task, including the anterior cingulate, cuneus, nucleus accumbens, 275 caudate, thalamus and inferior frontal gyrus ( Figure 4A, Table 1 ). The group and stimulus -specific 276 activation of the NAcc w as in line with a priori expectations for the CC phase (Figure 4C). To further 277 explore this effect, averaged parameter estimates from the bilateral NAcc ROI (mask acquired from 278 the IBASPM 71 atlas in the WFU PickAtlas toolbox in MNI152 space) w ere extracted. Across the 279 bilateral NAcc, differential activation was increased in the CC as compared to the Ext group 280 (t(44)=2.731, p=0.009, CC: 0.37±0.10, Ext: 0.04±0.06) , with the CC showing increased NAcc activation 281 to the CS+ compared to the CS - (t(23)=6.194, p<0.001, CS+: 0.59±1.12, CS -: 0.16±0.09) wh ereas the 282 Ext group did not (p=0.574). 283 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 10 284 Figure 4. Stimulus-type specific activation differs between participants undergoing CC versus extinction. A. Whole-brain 285 Group x CS-type interaction effects revealed distinc t stimulus-specific activation of regions including the anterior cingulate, 286 cuneus, nucleus accumbens, caudate, thalamus and inferior frontal gyrus during the counterconditioning vs. extinction 287 phase. Panel A displays g roup F -images (see Table 1 for directions) FWE-corrected at p<0.05, cluster -forming threshold 288 p=0.001. B. The right amygdala showed a Group x CS-type x Phase interaction during the CC/extinction task, indicating that 289 CC compared to extinction is associated with decreased activation of the amygdala. C. The bilateral NAcc showed a Group x 290 CS-type interaction during the CC/extinction task, revealing increased NAcc activation in response to the CS+ compared to 291 the CS - in the CC but not in the Ext group. Panel B and C display g roup F -images FWE -SVC at p<0.05, cluster -forming 292 threshold p=0.001, along with post-hoc tests on mean parameter estimates from the complete ROI included in the analyses. 293 ** p<0.01, * p<0.05, ≠ indicates that the value is significantly different from 0. 294 295 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 11 Table 1. Whole-brain main effects of group (CC, Ext), CS type (CS+, CS-) and phase (early, late) and interactions, during 296 the counterconditioning/extinction task. Cluster-forming threshold p=0.001, FWE-corrected at p<0.05, clusters were 297 labelled using the Talairach Daemon atlas and the AAL atlas for ROIs. For each cluster, the peak voxel coordinates (MNI 298 space) and regions are reported, and additional regions contained within the cluster are added in italics . See 299 Supplementary Table 1 for main effects of CS-type. 300 Peak MNI coordinate Region Cluster x y z Size (mm3) pFWE (cluster) Peak F- value Direction Group x CS-type x phase Parahippocampal Gyrus BA34R Parahippocampal Gyrus Amygdala, Uncus BA34R 1 18 -8 -20 1760 0.034 23.40 CS+>CS- difference increases from early to late phase for CC, not for Ext Group x CS-type Lateral Geniculum Body LR, Caudate Head LR, Thalamus LR, Lentiform Nucleus LR 1 2 -26 -18 29920 CS-) > (Ext CS+>CS-) Cuneus L Lingual Gyrus BA17/BA18 LR, Posterior Cingulate LR, Cuneus BA18R, Cuneus BA30L Declive R 2 -6 -96 2 23272 <0.001 43.50 Inferior Frontal Gyrus BA47L Insula BA13 L 3 -36 18 -6 4504 0.009 30.62 Extra-Nucleus R 4 30 26 2 3136 0.016 37.67 Superior Temporal Gyrus L Superior Temporal Gyrus BA41 L, Transverse Temporal Gyrus L 5 -60 -44 14 9088 0.002 43.56 Transverse Temporal Gyrus BA41 R Superior Temporal Gyrus R, Superior Temporal Gyrus BA42/BA22R 6 44 -22 12 7784 0.003 42.17 Anterior Cingulate BA32R Anterior Cingulate BA32L, Cingulate Gyrus R 7 6 30 26 8880 0.002 27.90 Precentral Gyrus L Inferior Frontal Gyrus L 8 -36 0 30 3624 0.014 30.10 Precentral Gyrus R Sub-Gyral R 9 40 2 32 4056 0.011 40.64 Precentral Gyrus BA6L Middle Frontal Gyrus BA6L 10 -44 -6 52 2184 0.028 24.34 (CC CS+>CS-) > (Ext CS+>CS-) Angular Gyrus R Supramarginal Gyrus R 11 54 -60 36 1944 0.032 24.18 Group x Phase No significant clusters CS-type x Phase No significant clusters Group No significant clusters Phase Inferior Frontal Gyrus R Inferior Frontal Gyrus BA45 R 1 30 26 8 4848 0.006 40.27 Early Phase > Late Phase Insula L Superior Temporal Gyrus BA22, Precentral Gyrus L 2 -28 26 0 4368 0.007 38.41 Postcentral Gyrus L 3 -54 -24 22 1768 0.031 23.75 301 Contrast estimates in further a priori defined ROIs during the CC/Ext task were submitted to a Group 302 (CC, Ext) x CS-type (CS+, CS-) x Phase (early, late) rmANOVA (Figure 5). The bilateral hippocampi (right 303 hippocampus cluster size: 664 mm 3, p=0. 001, FWE -SVC, left hippocampus cluster size: 112 mm 3, 304 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 12 p=0.024, FWE-SVC) and the left vmPFC (mask defined as bilateral gyrus rectus and medial orbital gyri, 305 cluster size = 160 mm3, p=0.01 3, FWE-SVC) showed differentially changing CS -type-specific 306 activations between the groups (Group x CS-type x Phase interaction). While CS+-specific suppression 307 of these regions appeared to increase during the CC task, this was not the case during the extinction 308 task. Post-hoc comparisons on averaged parameter estimates in the bilateral hippocampi confirmed 309 that stimulus -specific suppression increased during the course of the task in the CC group 310 (t(23)=3.280, p=0.003, early CS+ -CS-: 0.054 ±0.07, late: -0.150±0.07), but not in the Ext group 311 (p=0.266). Post -hoc comparisons across the vmPFC ROI also revealed increased CS+ -specific 312 suppression in the CC group compared to the Ext group (t(44)=2. 221, p=0.032, CC: -0.189±0.06, Ext: -313 0.070±0.10). While the extinction group showed increased CS+ -specific activation from the early to 314 the late phase of the extinction task (t(21)=2.235, p=0.036, early CS+: -0.149±0.08, late CS+: 315 0.040±0.09), the CC group did not (p=0.120). During the late phase , the CC group showed increased 316 vmPFC deactivation to the CS+ compared to the CS- (t(23)=3.174, p=0.004, late CS+: -0.284±0.06, late 317 CS-: -0.095±0.05), while the Ext group did not (p=0.503). Thus, across both the hippocampus and the 318 vmPFC, CC induced increased stimulus-specific suppression. 319 320 Figure 5. ROI analyses during the CC/extinction task reveal distinct activity in the hippocampus and left vmPFC. During 321 the CC/extinction task, stimulus-specific activation of the hippocampus (C) and left vmPFC ( D) changes differently between 322 groups. ** p<0.01, * p<0.05, ≠ indicates that the bar is significantly different from 0. Group F-images FWE-SVC at p<0.05, 323 cluster-forming threshold p=0.001, along with post-hoc tests on mean parameter estimates from complete ROI included in 324 the analyses. 325 During the spontaneous recovery task, a priori defined regions of interest did not reveal any effects 326 (see Supplementary Information). 327 Counterconditioning retroactively enhances item recognition for conditioned exemplars 328 Following the reinstatement test and re-extinction, participants completed a surprise item 329 recognition test approximately 24 hours after acquisition an d the CC/extinction task. One outlier was 330 excluded from this analysis (CS - false alarm rate = 0.91). Threat conditioning has previously been 331 shown to enhance 24-hour item recognition for category exemplars presented during the acquisition 332 phase27. However, th is enhancement for CS+ items did not extend to items presented during an 333 extinction session separated from the acquisition phase by a short break 31. We therefore analysed 334 item recognition for the CS+ and CS - during acquisition and the CC/extinction phase separately to 335 examine whether the groups differed in recognition memory performance (Figure 6). 336 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 13 337 Figure 6. Twenty-four hour recognition memory results. During acquisition and extinction on the first day of the 338 experiment, participants viewed trial-unique exemplars from two semantic categories (objects, animals) that served as CS+ 339 and CS-. The next day, participants completed a surprise memory test for these items, mixed with an equal number of novel 340 exemplars. Participants recognized relatively more items from the CS+ category, and participants that underwent CC 341 showed improved item recognition comp ared to participants in the Ext group. Error bars represent ± standard error of the 342 mean. *=p<0.05. 343 Corrected recognition scores (hits probability-false alarms probability) were subjected to a task 344 (acquisition, CC/extinction task) x CS -type (CS+, CS -) x Group ( CC, Ext) rmANOVA , including CS+ -345 category (animals, tools) as covariate. Overall, participants showed better memory for items from 346 the CS+ category (main effect of CS -type: (F( 1,42)=10.615, p=0.00 2, η²=0.2 02) and participants who 347 underwent CC showed better memory as compared to participants who underwent extinction (main 348 effect of Group: (F( 1,42)=4.963, p=0.0 31, η²=0.1 06). Stimulus -type specific item recognition differed 349 between the CC and Ext groups (CS -type x Group interaction: F(1,42)=4.535, p=0.039, η²=0.094). While 350 participants in the CC group showed better recognition memory for the CS+ category compared to 351 the CS- category (t(22)=2.531, p=0.019, means ± SD: CS+ 0.39±0.17, CS- 0.31±0.10), this was not the 352 case for participants in the Ext group ( t(23)=0.889, p=0.384, means ± SD: CS+ 0.30±0.13, CS - 353 0.28±0.11). Although the effect of stimulus -type was stronger for tools as CS+, this was not different 354 between groups (see Supplementary Information ). Thus, across the acquisition and CC/extinction 355 phase, participants who underwent CC showed a stronger enhancement of CS+ memory compared to 356 the participants that underwent extinction. 357 To further investigate to what extent CC retroactively affected memory for items presented during 358 the acquisition task, we examined item recognition during acquisition and the CC/extinction tasks 359 separately. While threat conditioning increased memory for CS+ items presented during the 360 acquisition task across both groups (main effect CS-type: (F( 1,42)=18.147, p=<0.001, η²=0.30 2), 361 subsequent CC enhanced this effect (Group x CS -type interaction: (F( 1,42)=5.112, p=0.029, η²=0.109). 362 Post-hoc tests reveal ed increased item memory for the CS+ category compared to the CS - category 363 presented during acquisition in the CC group (t(2 2)=2.341, p=0.029, means ± SD: CS+ 0. 40±0.21, CS- 364 0.31±0.12) but not in the Ext group ( t(23)=0.818, p=0. 422, means ± SD: CS+ 0.3 3±0.16, CS - 365 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 14 0.30±0.13). Again, although the effect of stimulus-type was stronger for tools as CS+, this was not 366 different between groups (see Supplementary Information ). As the acquisition task was identical 367 between groups, it appears that CC in comparison to extinction retroactively enhanced memory for 368 CS+ items. For items presented during the CC/extinction task, overall item recognition was better in 369 the CC group compared to the Ext group (main effect group: F( 1,42)=8.706, p=0.005, η²=0.172, means 370 ± SD: CC 0.35±0.12, Ext 0.26±0.09). Thus, compared to regular extinction, CC enhanced recognition of 371 items presented during CC, but interestingly also strengthened the emotional memory enhancement 372 of CS+ exemplars presented during acquisition, suggesting that immediate CC may alter consolidation 373 of a prior threat learning episode. 374 Following previous work 29,31,32, we explored stimulus-type specific decreases in item recognition 375 between tasks, as well as within -phase differences between item recognition for the CS+ and CS - 376 within each group. As expected, a post-hoc paired samples t -test showed that participants in the Ext 377 group remember ed significantly more CS+ items from the acquisition phase as compared to the 378 extinction phase (t(2 3)=2.238, p=0.036, means ± SD: acquisition 0.33±0.16, extinction 0.27±0. 13). In 379 contrast, p articipants who had undergone CC remembered CS+ items presented during acquisition 380 and CC equally well ( t(22)=0.390, p=0.701, means ± SD: acquisition 0.40±0.21, CC 0.38±0.16). Thus, 381 while recognition memory for items encoded during the extinction task was substantially weaker 382 than memory for items from the acquisition task, this was not the cas e for items presented during 383 CC. 384

385 This study aimed to test whether CC compared to regular extinction can lead to a more persistent 386 attenuation of threat responses, and to investigate whether this is mediated by neural mechanisms 387 reflecting extinction-related enhanced engagement of the vmPFC or engagement of reward -focused 388 networks. We found that CC prevented differential spontaneous recovery of PDRs compared to 389 regular extinction, suggesting that CC reduces the recovery of threat responses. Our fMRI results 390 suggested that CC engages different neural mechanisms compared to extinction. Most notably, while 391 the extinction group showed an increase in CS+ -specific vmPFC activation during extinction, the CC 392 group showed CS+-specific deactivation of the vmPFC that persisted throughout the late phase of CC. 393 Furthermore, CC led to increased NAcc activation for the CS+ compared to the CS -, whereas this was 394 not the case for extinction. Lastly, phase - and stimulus -specific activation of the hippocampus and 395 the amygdala differed between extinction and CC. Compared to extinction, CC led to increased 396 activation of the amygdala in the early phase, and increasing stimulus -specific deactivation of the 397 hippocampus over the course of the early and late phases. In addition, CC retroactively enhanced 398 item recognition for conditioned exemplars presented during acquisition and strengthened memory 399 for conditioned exemplars presented during CC compared to extinction. 400 The mechanism underlying CC appears to be qualitatively different from the mechanism underlying 401 regular extinction. Regular extinction is associated with activation of the vmPFC 23,33, which is thought 402 to inhibit the expression of threat responses by suppressing amygdala activity 20-23. In comparison to 403 regular extinction, n ovelty facilitated extinction, a form of enhanced extinction , in which aversive 404 events are replaced with novel neutral outcomes, has shown stronger CS+ -specific vmPFC 405 activation24. If CC was similarly mediated by enhanced recruitment of extinction networks, we would 406 have expected increased activation of the vmPFC, yet we observed a CS+-specific deactivation of the 407 vmPFC during CC, disproving this hypothesis. Interestingly, deactivation of the vmPFC during CC was 408 also found in studies investigating a form of counterconditioning induced by means of real-time fMRI 409 decoded neurofeedback34,35. During neurofeedback CC, participants implicitly learned to obtain 410 monetary rewards by generating a representation of the target CS+ in the visua l cortex 34. After 411 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 15 neurofeedback CC, reductions in threat responses were stronger in participants showing stronger 412 vmPFC deactivation, suggesting that vmPFC disengagement may be associated with fear reductions34. 413 Taken together, both our findings and previous neurofeedback studies suggest that in contrast to 414 enhanced extinction, CC disengages the vmPFC. Given that we replicate this finding using a different 415 approach that includes direct exposure to the CS+, vmPFC disengagement may be a distinguishing 416 characteristic of CC. The observed pattern of activity, including vmPFC deactivation further bears 417 resemblance to activity patterns observed during goal-directed eye movements in an experimental 418 model of eye -movement desensitization and reprocessing (EMDR) , which has also been shown to 419 improve extinction learning 36. A similar activity pattern and effect has also been found for working 420 memory-like tasks, such as a game of Tetris 37-39. Given that the above-mentioned tasks associated 421 with vmPFC deactivation share their strong engage ment of working memory and/or endogenous 422 attention mechanisms, th ereby engaging the executive control -network, deactivation of the vmPFC 423 and hippocampus could be the result of a deactivated default mode network due to competition 424 between activation of large scale brain networks40-42. 425 The CC procedure led to clear CS+-specific activation of the NAcc, which is in line with expectations 426 for reward anticipation in tasks with a monetary incentive delay aspect 43. Activation of the ventral 427 striatum has also been reported for active avoidance, and may be generally associated with 428 instrumental actions as opposed to passive delivery of an outcome 44,45. In line with studies on active 429 avoidance, delivery of a reward contingent on instrumental action s has been shown to yield CC that 430 is more resistant to renewal 46. CS+-specific activation of the NAcc was not seen in participants 431 undergoing extinction, suggesting that this activation is specific to CC. However, p revious work in 432 rodents revealed a n amygdala-ventral striatum (NAcc) pathway that is activated during extinction 433 training17. The recruitment of this pathway was shown to be enhanced during CC, and reduced the 434 return of fear 17, suggesting that CC may in fact enhance activation of reward-related networks that 435 are weakly activated by extinction. Indeed, fMRI studies in humans that modelled prediction error for 436 omitted aversive outcomes during extinction training (i.e. outcomes “better-than-expected”) showed 437 involvement of the NAcc 47-49. Possibly, activation of the NAcc during extinction is limited to early 438 extinction trials generat ing prediction error s. Nevertheless, b ased on our findings, it appears that 439 sustained CS+ -specific activation o f the NAcc is a distinct mechanism underlying CC but not 440 extinction, which is potentially associated with instrumental actions. 441 A recent neuroimaging study suggests that the neural differences between regular extinction and CC 442 may be maintained over time 19. In their within-subject study, two CS+ categories (animals, objects) 443 were used during threat conditioning. Subsequently one of the CS+ categories was used for regular 444 extinction, whereas the other was used for CC. During CC, CS+ exemplars were paired with positively 445 valenced picture s. During a spontaneous recovery task the following day, it was shown that 446 involvement of the vmPFC (amygdala -vmPFC functional connectivity) was stronger for regular 447 extinction compared to CC. In contrast, CS+-specific increases in functional connectivity between the 448 amygdala and the ventral striatum (NAcc) were only observed in the CC condition during a 449 spontaneous recovery task . Both findings are in line with the CC-associated vmPFC deactivation and 450 NAcc activation that we observed and suggest that differences in the neural mechanisms of regular 451 extinction and CC may be maintained during threat retrieval. 452 CC compared to regular extinction also strengthened item memory for the conditioned category. 453 While both reward and threat conditioning can enhance item recognition for the CS+ category27,50, 454 recognition of CS+ exemplars presented during extinction was shown to drop compared to 455 acquisition31. In contrast to extinction, within-session CC was previously shown to enhance memory, 456 suggesting that CC has a unique, strengthening effect on memory29. In the current study, we replicate 457 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 16 this finding, showing strengthened memory after CC compared to extinction. While enhanced 458 recognition of items presented during CC could be mediated by attentional prioritization 51, CC also 459 retrospectively strengthened memory for items presented during acquisition, suggesting that CC may 460 alter the consolidation of a prior threat conditioning episode. Retroactive enhancement of memory 461 consolidation for related items has previously been shown for conceptually related neutral items 462 presented prior to threat conditioning 32 and reward conditioning 50. At a neurobiological level, these 463 findings have been related to the synaptic tagging-and-capture hypothesis postulating that memories 464 for neutral events can be strengthened if they are followed by salient events, due to an initially short-465 lived synaptic “tag” that allows later events to stabilize the memory 32,52,53. At a systems level , 466 retroactive memory strengthening has been linked to reverse replay 54. Specifically, animal research 467 indicates that reward s increase reverse replay 55-57, and reward-induced reverse replay occurs 468 concurrently with firing of midbrain dopamine neurons 58. Interestingly, spontaneous replay is also 469 involved in regular extinction, in which unexpected omission of the US drives spontaneous 470 reactivation of activity patterns in the vmPFC . This spontaneous reactivatio n was shown to be 471 predictive of extinction recall and c ould be amplified through pharmacological enhancement of 472 dopaminergic activity 59. Yet while physiological dopaminergic modulation during extinction may be 473 limited to prediction error signals during the early phase 47-49, dopaminergic modulation may be 474 sustained throughout the MID -based CC task applied in this study. While we did not measure 475 dopaminergic activity directly, activation of the N Acc during reward anticipation is predictive of 476 dopamine release within the NAcc60-63. Given the increased stimulus-specific activation of the NAcc in 477 the CC group, it is likely that dopaminergic activity was enhanced during CC compared to regular 478 extinction. The enhanced dopaminergic modulation could strengthen memories through replay 55,64, 479 or may increase synaptic plasticity directly, potentially explaining enhanced item recognition after CC 480 compared to regular extinction 54,65,66. In line with these findings, research in humans shows that 481 reward systematically modulates memory for neutral objects in a retroactive manner, with objects 482 closest to the reward being prioritized54. It could be that reward-conditioning during CC similarly 483 drives reward -driven reverse replay, enhancing episodic memory for conceptually related items 484 presented during the preceding acquisition task. 485 Several limitations of the current study are worth considering. First, while the monetary incentive 486 aspect during CC clearly induced positive valence, it also increased physiological arousal, making it 487 difficult to isolate the individual effects of positiv e valence and reward-induced arousal. While the 488 current results are in line with previous work in CC using low -arousal, positive-valence pictures29, we 489 cannot exclude the possibility that the current findings (in part) reflect differences in task 490 engagement between participants due to active instead of passive reward delivery . However, it is 491 questionable whether it is meaningful to tease individual effects of valence and arousal apart since 492 arousal may facilitate reward processing. Indeed, striatal responses to obtained monetary rewards 493 are dependent on salience and are increased when rewards are dependent on active re sponses 494 compared to passive delivery 67. Second, although we included a reinstatement procedure in the 495 experiment, neither the Ext nor the CC group showed differential reinstatement. It is worth noting 496 however that reinstatement paradigms in humans may not reliably produce differential 497 reinstatement after extinction 68. Third, it is important to note that CC/extinction was carried out 498 within minutes after the acquisition phase, and the effects of CC and extinction may differ when 499 carried out after the acqui sition memory has been consolidated 69-72. Fourth, whole-brain analysis of 500 the CS -specific activation during the spontaneous recovery test in the Ext group did not yield any 501 clusters above threshold , while physiological results indicated spontaneous recovery of differential 502 threat responses. Given that recovered threat responses are often quick to extinguish and fMRI 503 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 17 analyses require averaging across multiple trials to achieve sufficient signal- to-noise ratio, threat-504 evoked neural activity may have been too brief to be detected. 505 In conclusion, our findings show that appetitive CC improves the retention of safety memory over 506 standard extinction. Strikingly, in cont rast to activation of the vmPFC during extinction, CC was 507 associated with stimulus-specific deactivation of the vmPFC. These findings may inform development 508 of future treatments for fear - and anxiety disorders. While a large body of research focuses on 509 enhancing regular extinction, this study i ndicates that another promising and potentially longer -510 lasting approach may be to engage reward -circuits. Although further work is needed, a major 511 advantage of CC -based interventions over extinction -based interventions may be that CC could be 512 more tolerable as it may shift attention away from the experience of fear. 513

514 Participants 515 Forty-eight healthy right-handed volunteers (15 males, 33 females; age [22.71±0.44]) with no 516 neurological or psychiatric history, and with normal hearing and normal or corrected -to-normal 517 vision completed the study. Exclusion criteria were pregnancy, disorders of the autonomic system, 518 heart conditions, recreational drug use and any contraindications for MRI. Participants provided 519 written informed consent and were paid 55 euros for their participation. Participants in the CC group 520 were able to earn an additional 14 euros. This study was approved by the local ethical review board 521 (METC Oost -Nederland and CMO Radboudumc). Participants were excluded from the threat 522 acquisition, CC /Extinction, spontaneous recovery, and reinstatement analyses if there was no 523 evidence for successful threat acquisition (mean CS->CS+ or CS+=CS-). For SCRs this was the case for 524 three participants, for PDR this was the case for two participants. Additional participants were 525 excluded in case of missing data due to technical failure. 526 Design and procedure 527 This study was a two-day between-subjects experiment carried out in the fMRI scanner (see Figure 1 528 for an overview of the design). Participants were assigned to either the CC or Ext group according to 529 a predetermined allocation sequence. At the start of each session, two Ag/AgCl electrodes were 530 attached to the medial phalanges of the second and third d igit of the left hand, a pulse oximeter was 531 attached to the first digit of the left hand to measure finger pulse and a respiration belt was placed 532 around the abdomen to measure respiration. All measures were taken using a BrainAmp MR system 533 and recorded us ing the BrainVision Recorder software (Brain Products GmbH, Munich, Germany). 534 The first day consisted of individual adjustment of the electrical shock followed by a single fMRI 535 session that included the following tasks: an object localizer task ( 17 min), a category threat 536 conditioning task ( 23 min) and a CC or extinction task ( 23 min). The second session took place the 537 following day and consisted of three runs: the spontaneous recovery and reinstatement test (12 538 min), item recognition test (29 min) and the valence-specific response characterization task (17 min). 539 Pavlovian conditioning paradigm 540 Note that CC included an instrumental and not Pavlovian conditioning procedure. This was done 541 because of pragmatic constraints in studies with humans. For example, w e cannot food deprive 542 humans to make an appetitive reward truly reinforcing and make participants anticipate the reward. 543 Previous work 50,67 and our pilot studies indicated that to maximize reward anticipation and evoke 544 conditioned responses, the reward conditioning needed to be instrumental. 545 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 18 The acquisition, counterconditioning, extinction, spontaneous recovery and reinstatement tasks 546 consisted of a categorical differential delay threat conditioning paradigm 27 with elements of the 547 monetary incentive delay task 30. Participants viewed trial-unique exemplars of pictures from two 548 categories (animals or objects, see Figure 1). In a counter -balanced manner, exemplars from one 549 category served as CS+ (reinforced) stimuli, while exemplars form the other category served as CS - 550 (unreinforced stimuli). Each trial start ed with the presentation of a stimulus. After a variable delay of 551 2.5-4s, a cue appeared to which participants were instructed to respond as quickly as possible with a 552 button press. After the button press, or when a 1s response window ha d elapsed, the colour of the 553 cue shift ed from black to blue. 0.5 -1.5s after the response window elapsed, CS+ items presented 554 during the acquisition phase could be reinforced with a shock. During the acquisition phase, 50% of 555 the CS+ pictures w ere followed by a shock. After 1s, the stimulus was replaced by neutral feedback 556 during the acquisition, extinction, and recovery tasks. During the CC phase, neutral feedback was 557 replaced by monetary feedback. During the CC phase, participants could obtain a €0.50 reward for 558 their quickest responses to the cues presented on top of CS+ stimuli. The response time target was 559 dynamically adjusted to achieve a reward reinforcement rate of approximately 70%. Reward was 560 withheld during the first three CS+ trials during the CC phase to make the tran sition from the 561 acquisition to the CC phase more gradual. The inter-trial interval (ITI) varied randomly between 8 and 562 10s. Pictures were presented in a pseudorandom order with no more than 3 consecutive 563 presentations of items from the same category and CC blocks consisted of 40 CS+ and 40 CS - 564 presentations each. The spontaneous recovery block consisted of 15 CS+ and 15 CS+ presentations, 565 and the reinstatement test consisted of 5 CS+ and 5 CS- presentations. 566 Item recognition memory test 567 Participants carried out a surprise recognition memory test compromised of 160 pictures (80 CS+, 80 568 CS-) shown during the acquisition and CC/extinction phases , as well as 160 category-matched new 569 items (80 CS+, 80 CS-). Participants rated on a 6 -point scale whether the picture was ‘definitely old’, 570 ‘probably old’, ‘maybe old’, ‘maybe new’, ‘probably new’, ‘definitely new’. 571 Valence-specific response characterization 572 The valence-specific response characterization task consist ed of an adapted version of the 573 conditioning paradigm used during the acquisition phase. Instead of category items, participants 574 were presented with squares in three different colours. One of the stimuli was reinforced with 575 shocks (CS+ -shock, 50% reinforceme nt rate), one stimulus was reinforced with monetary rewards 576 (CS+-reward, approximately 70% reinforcement rate, response time target adjusted dynamically) and 577 the last stimulus was not reinforced (CS-). Each stimulus was presented 40 times in a pseudorandom 578 order with no more than three repetition s of each stimulus. Colours and reinforcement (shocks vs. 579 rewards) were counterbalanced across participants. 580 Peripheral stimulation 581 Electrical shocks were deliver ed using two Ag/AgCl electrodes attached to the medial phalanges of 582 the second and third digit of the right hand using a MAXTENS 2000 (Bio -Protech) device. Shock 583 intensity varied in 10 intensity steps between 0 to 40 V and 0 to 80 mA. Shock duration was 200 ms. 584 In line with prior threat conditioning protocol s, s hock intensity was calibrated using an ascending 585 staircase procedure starting with a low voltage setting near a perceptible threshold and increasing to 586 a level deemed “maximally uncomfortable but not painful” by the participant32,73,74. 587 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 19 Arousal and valence ratings 588 Arousal and valence ratings were acquired using self-assessment manikin scales. The arousal scale 589 ranged from 1 (=extremely calm) to 10 ( =extremely excited). The valence scale ranged from 1 590 (=extremely negative) to 10 (=extremely positive). The valence and arousal ratings were collected for 591 the two categories (animals and tools) after the acquisition phase, after the CC/extinction phase, at 592 the start of day 2 immediately before the spontaneous recovery test and after the reinstatement 593 test. For the stimuli used in the valence-specific response characterization task , valence and arousal 594 ratings were collected immediately after the task. 595 SCR pre-processing and analysis 596 Electrodermal activity data were pre-processed using in-house software; radio frequency (RF) 597 artefacts were removed and a low-pass filter was applied75,76. Skin conductance responses (SCR) were 598 automatically scored with additional, blinded, manual supervision using Autonomate 77. SCR 599 amplitudes (measured in μSiem) were determined for each trial as the maximum response with an 600 onset between 0.5 and 7.5s after stimulus onset and maximum rise time of 14.5s. Shock- and reward- 601 reinforced trials were excluded from analysis. All response amplitudes were square-root transformed 602 and normalized according to each participant’s mean UCS response prior to statistical analysis. The 603 average SCRs were computed per CS-type, task, phase (early, late), and participant. 604 PDR pre-processing and analysis 605 Pupil dilation was measured with a MR-compatible eye-tracker from SensoMotoric Instrument (MEye 606 Track-LR camera unit, SMI, SensoMotoric Instruments) and sampled at a rate of 50 Hz. Data were 607 analysed using in -house software 78 implemented in Matlab R2018b (MathWorks), based on 608 previously described methods 79. Eyeblink artifact s were identified and linearly interpolated 100 ms 609 before and 100 ms after each identified blink. Data from scan runs missing 50% time points or more 610 were excluded. After interpolating missing values, time series were band-pass filtered at 0.05 to 5 Hz 611 (by subtracting the mean and dividing by the standard deviation) within each participant and run to 612 account for between -subjects variance in overall pupil size. Event -related pupil diameter responses 613 were calculated by averaging pupil diameter during 3.5 to 7 sec period after stimulus onset, divided 614 by the 1 sec pre-stimulus pupil diameter (-1 to 0 sec). The average PDR s were computed per CS-type, 615 task, phase (early, late), and participant. 616 MRI data acquisition 617 MRI scans were acquired using a Siemens (Erlangen, Germany) 3T MAGNETOM PrismaFit MR scanner 618 equipped with 32 -channel transmit -receiver head coil. The manufacturer’s automatic 3D -shimming 619 procedure was performed at the beginning of each experiment. Participants were placed in a light 620 head restraint withi n the scanner to limit head movements during acquisition. Functional images 621 were acquired with multi -band multi-echo gradient echo -planar (EPI) sequence [5 1 oblique 622 transverse slices; slice thickness, 2.5mm; TR, 1.5s; flip angle, 75°; echo times, 13.4, 34.8, and 56.2 ms; 623 FOV, 210 x 210 mm2 ; matrix size 84x84x64, fat suppression ]. To account for regional variation in 624 susceptibility-induced signal drop out, voxel -wise weighted sums of all echoes were calculated based 625 on local contrast -to-noise ratio after which echo-series are integrated using PAID weighting 80. Field 626 maps were acquired (51 oblique transverse slices; slice thickness, 2.5mm; TR, 0.49 s; TE, 4.92 ms and 627 7.48 ms; flip angle, 60°; FOV, 210 x 210 mm2; matrix size 84x84x64 ) at the start of each session to 628 allow for correction of distortions due to magnetic field in homogeneity. A high-resolution structural 629 image (1mm isotropic) was acquired using a T1 -weighted 3D magnetizatio n-prepared rapid gradient 630 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 20 echo sequence [MP-RAGE; TR, 2300 ms; TE, 3.03 ms; flip angle, 8°; 192 contiguous 1 mm slices; FOV = 631 256 x 256 mm2]. 632 fMRI analysis 633 Anatomical and functional data were pre-processed using fMRIPrep 20.0.6 81. The complete 634 boilerplate can be found in Supplementary Information 1. In brief, MRI data were pr e-processed in 635 standard stereotactic (MNI152) space. Pulse and respiration data were processed offline using in -636 house software and visually inspected to remove artefacts and correct peak detection , and corrected 637 pulse and respiration data were used for retrospective image -based correction (RETROICORplus) of 638 physiological noise artefacts in BOLD -fMRI data 82. Identical transformations were applied to all 639 functional images, which were resliced into 2 mm isotropic voxels. After pre -processing in fMRIPrep, 640 functional images were smoothed with a 6 mm FWHM Gaussian kernel (using SPM12; 641 http://www.fil.ion.ucl.ac.uk/spm; Wellcome Department of Imaging Neuroscience, London, UK). 642 For the acquisition, extinction/cc and spontaneous recovery phases, BOLD responses to CS+, and CS- 643 during the early phase (first half of the trials) and late phase (second half of the trials) were modelled 644 in 4 separate regressors using box -car functions. Additionally, during all these phases, target 645 presentation, button pres s and shocks were modelled using stick functions, and feedback 646 presentation and breaks were modelled using box-car functions and included as nuisance regressors. 647 For the category localizer, BOLD responses to animals, objects, and phase-scrambled blocks were 648 modelled in 3 separate regressors using box functions. All first-level models also included six 649 movement parameter regressors (3 translations, 3 rotations) derived from rigid -body motion 650 correction, 2 5 RETROICOR physiolog ical noise regressors , high-pass filtering (1/128 Hz cut -off), and 651 AR(1) serial correlations correction. First-level contrast s were calculated for early and late CS+ and 652 CS- separately for the acquisition, CC/extinction, and spontaneous recovery phases. 653 For the acquisition and CC/extinction, first-level contrast were entered into a second-level Group 654 (extinction, cc) x CS -type ( CS+, CS -) x Phase (early, late) mixed factorial model using the Multilevel 655 and Repeated Measures (MRM) toolbox 83. For the spontaneous recovery test, BOLD-responses from 656 the early phase were entered into a second -level Group (extinction, cc) x CS -type (CS+, CS -) mixed 657 factorial model. Thresholding was achieved using nonparametric permutation testing (5,000 658 iterations), with a cluster -setting threshold of p <.001 for whole -brain analysis and familywise error 659 (FWE) correction at p<0.05 at cluster -level for whole -brain analysis and voxel -level for ROI -analysis 660 (Amygdala, Hippocampus, vmPFC, NAcc). Activations are displayed on the single -subject high-661 resolution T1 volume provided by the Montreal Neurological Institute (MNI). 662 Region of interest definition 663 Based on a priori hypotheses, results for the amygdala, NAcc, hippocampus and the ventromedial 664 prefrontal cortex are corrected for reduced search volumes using small volume. Masks were created 665 using the WFU PickAtlas toolbox84 in combination with the Automated Anatomical Labeling atlas 85 for 666 the bilateral amygdala , bilateral hippocampus and vmPFC ( Frontal_Med_orb_L&R and Rectus L&R ). 667 The IBASPM 71 anatomical atlas toolbox86 as used to create a mask for the bilateral NAcc. 668 Statistical testing 669 Statistical analyses of behavioural and physiological variables were performed in SPSS (IBM SPSS 670 Statistics Inc.). Dependent measures were submitted to repeated measure ANOVAs and statistics 671 were Greenhouse -Geisser or Huyn -Feldt corrected for non -sphericity when appropriate. Significant 672 findings from ANOVAs were followed -up by paired - and independent samples t -tests. We report 673 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 21 partial eta-square as measure of effect size. Means ± s.e.m are provided where relevant unless 674 otherwise indicated. 675 Deviations from the pre-registration 676 The preregistration for this project can be found on OSF (https://osf.io/fbz6n). We pre-registered to 677 sample SCRS in a 0.75 and 3.15 s window after stimulus onset. However, visual inspection of SCR 678 responses during the acquisition phase indicated that response latencies shifted towards the late 679 phase of the trial. We therefore opted to use a longer window (0.5s to 7.5s for stimulus onset) and 680 exclude reinforced trials. The pre -registration erroneously stated that pupil -dilation data would be z -681 scored and later divided by the pre -stimulus average. PDR data were not z -scored but were only 682 normalized to a 1-sec pre-stimulus baseline. In line with the SCR data, response onset latencies were 683 later than expected. Based on visual inspection of the data from the acquisition phase, we decided to 684 use a window around the expected shock onset: 3.5 -7s after stimulus onset. Reinforced trials were 685 excluded. Results for SCR, retrospective reinforcement estimations and the reinstatement test can be 686 found in the Supplementary Information. Due to an error in the scripts for the item recognition test, 687 trial-by-trial data w ere not recorded for the first 12 participants. Therefore, analysis of the me mory 688 data focused on averaged data for the early and late phase of acquisition and CC/extinction, leaving 689 out planned change point analyses on bins of 4 trials. 690 While we planned to extract a vmPFC mask for ROI analysis based on a [CS- > CS+ shock] contrast of 691 BOLD responses during the valence -specific response characterization task to identify “extinction 692 regions”, this did not yield ventromedial prefrontal clusters that survived correction. Instead, in line 693 with our other ROIs, we opted to create a mask based on the AAL atlas. Due to time constraints, 694 native-space and functional connectivity analyses were not carried out for this manuscript. 695 696 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 22 Data Availability 697 Pseudonymized data generated in this study are available upon request from the Radboud Data 698 Repository at https://data.ru.nl/. Raw MR images are not publicly available due to privacy or ethical 699 restrictions. 700

701 1 Kessler, R. C. et al. Lifetime Prevalence and Age-of -Onset Distributions of. Arch Gen 702 Psychiatry 62, 593-602 (2005). 703 2 Scheveneels, S., Boddez, Y., Vervliet, B. & Hermans, D. The validity of laboratory-based 704 treatment research: Bridging the gap between fear extinction and exposure treatment. 705 Behaviour Research and Therapy 86, 87-94, doi:10.1016/j.brat.2016.08.015 (2016). 706 3 Bouton, M. E. Context and behavioral processes in extinction. Learning and Memory 11, 485-707 494, doi:10.1101/lm.78804 (2004). 708 4 Myers, K. M. & Davis, M. Behavioral and Neural Analysis of Extinction. Neuron 36, 567-584 709 (2002). 710 5 Keller, N. E., Hennings, A. C. & Dunsmoor, J. E. Behavioral and neural processes in 711 counterconditioning: Past and future directions. Behaviour Research and Therapy 125, 712 doi:10.1016/j.brat.2019.103532 (2020). 713 6 Dickinson, A. & Pearce, J. M. Inhibitory interactions between appetitive and aversive stimuli. 714 Psychological Bulletin 84, 690-711 (1977). 715 7 Rescorla, R. A. & Solomon, R. L. Two-process learning theory: Relationships between 716 Pavlovian conditioning and instrumental learning. Psychological Review1 74, 151-182 (1967). 717 8 Pearce, J. M. & Dickinson, A. Pavlovian countercondition: Changing the suppressive 718 properties of shock by association with food. Animal Behavior Processes 1, 170-177 (1975). 719 9 Kang, S., Vervliet, B., Engelhard, I. M., van Dis, E. A. M. & Hagenaars, M. A. Reduced return of 720 threat expectancy after counterconditio ning verus extinction. Behaviour Research and 721 Therapy 108, 78-84, doi:10.1016/j.brat.2018.06.009 (2018). 722 10 Newall, C., Watson, T., Grant, K. A. & Richardson, R. The relative effectiveness of extinction 723 and counter-conditioning in diminishing children's fe ar. Behaviour Research and Therapy 95, 724 42-49, doi:10.1016/j.brat.2017.05.006 (2017). 725 11 Luck, C. C. & Lipp, O. V. Verbal instructions targeting valence alter negative conditional 726 stimulus evaluations (but do not affect reinstatement rates). Cognition and E motion 32, 61-727 80, doi:10.1080/02699931.2017.1280449 (2018). 728 12 van Dis, E. A. M., Hagenaars, M. A., Bockting, C. L. H. & Engelhard, I. M. Reducing negative 729 stimulus valence does not attenuate the return of fear: Two counterconditioning 730 experiments. Behaviour Research and Therapy 120, 103416, doi:10.1016/j.brat.2019.103416 731 (2019). 732 13 Jozefowiez, J. et al. Retroactive interference: Counterconditioning and extinction with and 733 without biologically significant outcomes. Journal of Experimental Psychology: Animal 734 Learning and Cognition 46, 443-459, doi:10.1037/xan0000272 (2020). 735 14 Bouton, M. E. Context, ambiguity, and unlearning: sources of relapse after behavioral 736 extinction. Biological psychiatry 52, 976-986, doi:10.1016/s0006-3223(02)01546-9 (2002). 737 15 Bouton, M. E. & Peck, C. A. Spontaneous recovery in cross-motivational transfer ( 738 counterconditioning ). Animal Learning & Behavior 20, 313-321 (1992). 739 16 Brooks, D. C., Hale, B., Nelson, J. B. & Bouton, M. E. Reinstatement after counterconditioning. 740 Animal Learning & Behavior 23, 383-390, doi:10.3758/BF03198938 (1995). 741 17 Correia, S. S., McGrath, A. G., Lee, A., Graybiel, A. M. & Goosens, K. A. Amygdala-ventral 742 striatum circuit activation decreases long -term fear. eLife 5, 1 -25, doi:10.7554/eLife.12669 743 (2016). 744 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 23 18 Keller, N. E., Cooper, S. E., McClay, M. & Dunsmoor, J. E. Counterconditioning reduces 745 contextual renewal in a novel context but not in the acquisition context. Neurobiol Learn 746 Mem 201, 107749, doi:10.1016/j.nlm.2023.107749 (2023). 747 19 Keller, N. E., Hennings, A. C., Leiker, E. K., Lewis-Peacock, J. A. & Dunsmoor, J. E. Rewarded 748 Extinction Increases Amygdalar Connectivity and Stabilizes Long -Term Memory Traces in the 749 vmPFC. J Neurosci 42, 5717-5729, doi:10.1523/JNEUROSCI.0075-22.2022 (2022). 750 20 Quirk, G. J., Russo, G. K., Barron, J. L. & Lebron, K. The role of ventromedial prefrontal cortex 751 in the recovery of extinguished fear. Journal of Neuroscience 20, 6225 -6231, 752 doi:10.1523/jneurosci.20-16-06225.2000 (2000). 753 21 Morgan, M. A., Romanski, L. M. & LeDoux, J. E. Extinction of emotional learning: Contribution 754 of medial prefrontal cortex. Neuroscience Letters 163, 109 -113, doi:10.1016/0304 -755 3940(93)90241-C (1993). 756 22 Quirk, G. J., Likhtik, E., Pelletier, J. G. & Paré, D. Stimulation of medial prefrontal cortex 757 decreases the responsiveness of central amygdala output neurons. Journal of Neuroscience 758 23, 8800-8807, doi:10.1523/jneurosci.23-25-08800.2003 (2003). 759 23 Phelps, E. A., Delgado, M. R., Nearing, K. I. & Ledoux, J. E. Extinction learning in humans: Role 760 of the amygdala and vmPFC. Neuron 43, 897-905, doi:10.1016/j.neuron.2004.08.042 (2004). 761 24 Dunsmoor, J. E. et al. Role of human ventromedial prefrontal cortex in learning and recall of 762 enhanced extinction. The Journal of Neuroscience 39, 2713 -2718, 763 doi:10.1523/JNEUROSCI.2713-18.2019 (2019). 764 25 Diekhof, E. K., Kaps, L., Falkai, P. & Gruber, O. The role of the human ventral striatum and the 765 medial orbitofrontal cortex in the representation of reward magnitude - An activation 766 likelihood estimation meta -analysis of neuroimaging studies of passive reward expectancy 767 and outcome processing. Neuropsychologia 50, 1252 -1266, 768 doi:10.1016/j.neuropsychologia.2012.02.007 (2012). 769 26 Bulganin, L., Bach, D. R. & Wittmann, B. C. Prior fear conditioning and reward learning 770 interact in fear and reward networks. Frontiers in Behavioral Neuroscience 8, 1 -11, 771 doi:10.3389/fnbeh.2014.00067 (2014). 772 27 Dunsmoor, J. E., Martin, A. & LaBar, K. S. Role of conceptual knowledge in learning and 773 retention of conditioned fear. Biological Psychology 89, 300 -305, 774 doi:10.1016/j.biopsycho.2011.11.002 (2012). 775 28 Dunsmoor, J. E. & Kroes, M. C. Episodic memory and Pavlovian conditioning: ships passing in 776 the night. Current Opinion in Behavioral Scienc es 26, 32 -39, 777 doi:10.1016/j.cobeha.2018.09.019 (2019). 778 29 Keller, N. E. & Dunsmoor, J. E. The effects of aversive-to -appetitive counterconditioning on 779 implicit and explicit fear memory. Learning & memory (Cold Spring Harbor, N.Y.) 27, 12-19, 780 doi:10.1101/lm.050740.119 (2020). 781 30 Knutson, B., Westdorp, A., Kaiser, E. & Hommer, D. FMRI visualization of brain activity during 782 a monetary incentive delay task. NeuroImage 12, 20-27, doi:10.1006/nimg.2000.0593 (2000). 783 31 Dunsmoor, J. E. et al. Event segmentation protects emotional memories from competing 784 experiences encoded close in time. Nature Human Behaviour (2018). 785 32 Dunsmoor, J. E., Murty, V. P., Davachi, L. & Phelps, E. A. Emotional learning selectively and 786 retroactively strengthens memo ries for related events. Nature 520, 345 -348, 787 doi:10.1038/nature14106 (2015). 788 33 Milad, M. R. et al. Recall of Fear Extinction in Humans Activates the Ventromedial Prefrontal 789 Cortex and Hippocampus in Concert. Biological Psychiatry 62, 446 -454, 790 doi:10.1016/j.biopsych.2006.10.011 (2007). 791 34 Koizumi, A. et al. Fear reduction without fear through reinforcement of neural activity that 792 bypasses conscious exposure. Nature Human Behaviour 1, 1-7, doi:10.1038/s41562-016-0006 793 (2017). 794 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 24 35 Taschereau-Dumouchel, V. et al. Towards an unconscious neural reinforcement intervention 795 for common fears. Proceedings of the National Academy of Sciences of the United States of 796 America 115, 3470-3475, doi:10.1073/pnas.1721572115 (2018). 797 36 de Voogd, L. D. et al. Eye-movement intervention enhances extinction via amygdala 798 deactivation. Journal of Neuroscience 38, 8694-8706, doi:10.1523/JNEUROSCI.0703-18.2018 799 (2018). 800 37 Holmes, E. A., James, E. L., Coode-Bate, T. & Deeprose, C. in PLoS ONE Vol. 41 (2009). 801 38 James, E. L. et al. Computer Game Play Reduces Intrusive Memories of Experimental Trauma 802 via Reconsolidation -Update Mechanisms. Psychological Science 26, 1201 -1215, 803 doi:10.1177/0956797615583071 (2015). 804 39 Price, R. B., Paul, B., Schneider, W. & Siegle, G. J. Neural correlates of three neurocognitive 805 intervention strategies: A preliminary step towards personalized treatment for psychological 806 disorders. Cognitive Therapy and Research 37, 657 -672, doi:10.1007/s10608 -012-9508-x 807 (2013). 808 40 Qin, S., Hermans, E. J., van Marle, H. J. F., Luo, J. & Fernández, G. Acute Psychological Stress 809 Reduces Working Memory -Related Activity in the Dorsolateral Prefrontal Cortex. Biological 810 Psychiatry 66, 25-32, doi:10.1016/j.biopsych.2009.03.006 (2009). 811 41 Seeley, W. W. et al. Dissociable intrinsic connectivity networks for salience processing and 812 executive control. Journal of Neuroscience 27, 2349 -2356, doi:10.1523/JNEUROSCI.5587 -813 06.2007 (2007). 814 42 Liang, X., Zou, Q., He, Y. & Yang, Y. Topologically Reorganized Connectivity Architecture of 815 Default-Mode, Executive -Control, and Salience Networks across Working Memory Task 816 Loads. Cerebral Cortex 26, 1501-1511, doi:10.1093/cercor/bhu316 (2016). 817 43 Knutson, B. & Cooper, J. C. Functional magnetic resonance imaging of reward prediction. 818 Current Opinion in Neurology 18, 411-417, doi:10.1097/01.wco.0000173463.24758.f6 (2005). 819 44 Boeke, E. A., Moscarello, J. M., LeDoux, J. E., Phelps, E. A. & Hartley, C. A. Active avoidance: 820 Neural mechanisms and attenuation of pavlovian conditioned responding. Journal of 821 Neuroscience 37, 4808-4818, doi:10.1523/JNEUROSCI.3261-16.2017 (2017). 822 45 Delgado, M. R., Jou, R. L., LeDoux, J. E. & Phelps, E. A. Avoiding negative outcomes: Tracking 823 the mechanisms of avoidance learning in humans during fear conditioning. Frontiers in 824 Behavioral Neuroscience 3, 1-9, doi:10.3389/neuro.08.033.2009 (2009). 825 46 Thomas, B. L., Cutler, M. & Novak, C. A modified counterconditioning procedure prevents the 826 renewal of conditioned fear in rats. Learning and Motivation 43, 24 -34, 827 doi:10.1016/j.lmot.2012.01.001 (2012). 828 47 Raczka, K. A. et al. Empirical support for an involvement of the mesostriatal dopamine 829 system in human fear extinction. Translational Psychiatry 1, 1 -8, doi:10.1038/tp.2011.10 830 (2011). 831 48 Thiele, M., Yuen, K. S. L., Gerlicher, A. V. M. & Kalisch, R. A ventral striatal prediction error 832 signal in human fear extinction learning. NeuroImage 229, 117709, 833 doi:10.1016/j.neuroimage.2020.117709 (2021). 834 49 Esser, R., Korn, C. W., Ganzer, F. & Haaker, J. L-DOPA modulates activity in the vmPFC , 835 nucleus accumbens , and VTA during threat extinction learning in humans. eLife, 1-21 (2021). 836 50 Patil, A., Murty, V. P., Dunsmoor, J. E., Phelps, E. A. & Davachi, L. Reward retroactively 837 enhances memory consolidation for related items. Learning and Memory 24, 65 -69, 838 doi:10.1101/lm.042978.116 (2017). 839 51 Talmi, D., Anderson, A. K., Riggs, L., Caplan, J. B. & Moscovitch, M. Immediate memory 840 consequences of the effect of emotion on attention to pictures. Learning and Memory 15, 841 172-182, doi:10.1101/lm.722908 (2008). 842 52 Ballarini, F., Moncada, D., Martinez, M. C., Alen, N. & Viola, H. Behavioral tagging is a general 843 mechanism of long -term memory formation. Proceedings of the National Academy of 844 Sciences of the United States of America 106, 14599 -14604, doi:10.1073/pnas. 0907078106 845 (2009). 846 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 25 53 Frey, U. & Morris, R. G. M. Synaptic tagging and long-term potentiation. Nature 385, 533-536 847 (1997). 848 54 Braun, E. K., Wimmer, G. E. & Shohamy, D. Retroactive and graded prioritization of memory 849 by reward. Nature Communications 9, 1-12, doi:10.1038/s41467-018-07280-0 (2018). 850 55 Ambrose, R. E., Pfeiffer, B. E. & Foster, D. J. Reverse Replay of Hippocampal Cells is Uniquely 851 Modulated By Changing Reward. Neuron 91, 1124 -1136, 852 doi:10.1016/j.neuron.2016.07.047.Reverse (2018). 853 56 Foster, D. J. & Wilson, M. A. Reverse replay of behavioural sequences in hippocampal place 854 cells during the awake state. Nature 440, 680-683, doi:10.1038/nature04587 (2006). 855 57 Diba, K. & Buzsáki, G. Forward and reverse hippocampal place-cell sequences during ripples. 856 Nature Neuroscience 10, 1241-1242, doi:10.1038/nn1961 (2007). 857 58 Gomperts, S. N., Kloosterman, F. & Wilson, M. A. VTA neurons coordinate with the 858 hippocampal reactivation of spatial experience. eLife 4, 1-22, doi:10.7554/elife.05360 (2015). 859 59 Gerlicher, A. M. V., Tüscher, O. & Kalisch, R. Dopamine-dependent prefrontal reactivations 860 explain long-term benefit of fear extinction. Nature Communications 9, doi:10.1038/s41467-861 018-06785-y (2018). 862 60 Weiland, B. J., Zucker, R. A., Zubieta, J. K. & Heitzeg, M. M. Striatal dopaminergic reward 863 response relates to age of first drunkenness and feedback response in a t-risk youth. 864 Addiction Biology 22, 502-512, doi:10.1111/adb.12341 (2017). 865 61 Weiland, B. J. et al. Relationship between impulsivity, prefrontal anticipatory activation, and 866 striatal dopamine release during rewarded task performance. Psychiatry Research - 867 Neuroimaging 223, 244-252, doi:10.1016/j.pscychresns.2014.05.015 (2014). 868 62 Schott, B. H. et al. Mesolimbic functional magnetic resonance imaging activations during 869 reward anticipation correlate with reward -related ventral striatal dopamine release. Journal 870 of Neuroscience 28, 14311-14319, doi:10.1523/JNEUROSCI.2058-08.2008 (2008). 871 63 Buckholtz, J. W. et al. Mesolimbic dopamine reward system hypersensitivity in individuals 872 with psychopathic traits. Nature Neuroscience 13, 419-421, doi:10.1038/nn.2510 (2010). 873 64 Singer, A. C. & Frank, L. M. Rewarded Outcomes Enhance Reactivation of Experience in the 874 Hippocampus. Neuron 64, 910-921, doi:10.1016/j.neuron.2009.11.016 (2009). 875 65 Atherton, L. A., Dupret, D. & Mellor, J. R. Memory trace replay: The shaping of memor y 876 consolidation by neuromodulation. Trends in Neurosciences 38, 560 -570, 877 doi:10.1016/j.tins.2015.07.004 (2015). 878 66 Brzosko, Z., Schultz, W. & Paulsen, O. Retroactive modulation of spike timingdependent 879 plasticity by dopamine. eLife 4, 1-13, doi:10.7554/eLife.09685 (2015). 880 67 Zink, C. F., Pagnoni, G., Martin-skurski, M. E., Chappelow, J. C. & Berns, G. S. Human Striatal 881 Responses to Monetary Reward Depend on Saliency. Neuron 42, 509-517 (2004). 882 68 Haaker, J., Golkar, A., Hermans, D. & Lonsdorf, T. B. A review on human reinstatement 883 studies: An overview and methodological challenges. Learning and Memory 21, 424 -440, 884 doi:10.1101/lm.036053.114 (2014). 885 69 Chang, C. h. & Maren, S. Early extinction after fear conditioning yields a context-independent 886 and short-term suppression of conditional freezing in rats. Learning & memory (Cold Spring 887 Harbor, N.Y.) 16, 62-68, doi:10.1101/lm.1085009 (2009). 888 70 Devenport, L. D. Spontaneous recovery without interference: Why remembering is adaptive. 889 Animal Learning and Behavior 26, 172-181, doi:10.3758/BF03199210 (1998). 890 71 Maren, S. Nature and causes of the immediate extinction deficit: A brief review. 891 Neurobiology of Learning and Memory 113, 19-24, doi:10.1016/j.nlm.2013.10.012 (2014). 892 72 Myers, K. M., Ressler, K. J. & Davis, M. Different mechanisms of fear extinction dependent on 893 length of time since fear acquisition. Learning and Memory 13, 216 -223, 894 doi:10.1101/lm.119806 (2006). 895 73 Kroes, M. C. W., Dunsmoor, J. E., Mackey, W. E., McClay, M. & Phelps, E. A. Context 896 conditioning in humans using commercially available immersive Virtual Reality. Scientific 897 Reports 7, 8640, doi:10.1038/s41598-017-08184-7 (2017). 898 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 26 74 LaBar, K. S., Gatenby, J. C., Gore, J. C., LeDoux, J. E. & Phelps, E. A. Human amygdala 899 activation during conditioned fear acquisition and extinction: A mixed -trial fMRI study. 900 Neuron 20, 937-945, doi:10.1016/S0896-6273(00)80475-4 (1998). 901 75 de Voogd, L. D., Fernández, G. & Hermans, E. J. Disentangling the roles of arousal and 902 amygdala ac tivation in emotional declarative memory. Social Cognitive and Affective 903 Neuroscience 11, 1471-1480, doi:10.1093/scan/nsw055 (2016). 904 76 de Voogd, L. D., Fernández, G. & Hermans, E. J. Awake reactivation of emotional memory 905 traces through hippocampal -neocortical interactions. NeuroImage 134, 563 -572, 906 doi:10.1016/j.neuroimage.2016.04.026 (2016). 907 77 Green, S. R., Kragel, P. A., Fecteau, M. E. & LaBar, K. S. Development and validation of an 908 unsupervised scoring system (Autonomate) for skin conductance response analysis. 909 International Journal of Psychophysiology 91, 186 -193, doi:10.1016/j.ijpsycho.2013.10.015 910 (2014). 911 78 Hermans, E. J., Henckens, M. J. A. G., Roelofs, K. & Fernández, G. Fear bradycardia and 912 activation of the human periaqueductal grey. NeuroImage 66, 278 -287, 913 doi:10.1016/j.neuroimage.2012.10.063 (2013). 914 79 Siegle, G. J., Steinhauer, S. R., Stenger, V. A., Konecky, R. & Carter, C. S. Use of concurrent 915 pupil dilation assessment to inform interpretation and analysis of fMRI data. NeuroImage 20, 916 114-124, doi:10.1016/S1053-8119(03)00298-2 (2003). 917 80 Poser, B. A., Versluis, M. J., Hoogduin, J. M. & Norris, D. G. BOLD contrast sensitivity 918 enhancement and artifact reduction with multiecho EPI: Parallel -acquired inhomogeneity -919 desensitized fMRI. Magnetic Resonance in Medicine 55, 1227-1235, doi:10.1002/mrm.20900 920 (2006). 921 81 Esteban, O. et al. fMRIPrep: a robust preprocessing pipeline for functional MRI. Nature 922

16, 111-116, doi:10.1038/s41592-018-0235-4 (2019). 923 82 Glover, G. H., Li, T. Q. & Ress, D. Image-based method for retrospective correction of 924 physiological motion effects in fMRI: RETROICOR. Magnetic Resonance in Medicine 44, 162-925 167, doi:10.1002/1522-2594(200007)44:13.0.CO;2-E (2000). 926 83 McFarquhar, M. et al. Multivariate and repeated measures (MRM): A new toolbox for 927 dependent and multimodal group -level neuroimaging data. NeuroImage 132, 373 -389, 928 doi:10.1016/j.neuroimage.2016.02.053 (2016). 929 84 Maldjian, J. A., Laurienti, P. J., Kraft, R. A. & Burdette, J. H. An automated method for 930 neuroanatomic and cytoarchitectonic atlas -based interrogation of fMRI data sets. 931 NeuroImage 19, 1233-1239, doi:10.1016/S1053-8119(03)00169-1 (2003). 932 85 Tzourio-Mazoyer, N. et al. Automated anatomical labeling of activations in SPM using a 933 macroscopic anatom ical parcellation of the MNI MRI single -subject brain. NeuroImage 15, 934 273-289, doi:10.1006/nimg.2001.0978 (2002). 935 86 Alemán-Gómez, Y., Melie-Garcia, L. & Valdés-Hernández, P. A. in Proceedings of the 12th 936 Annual Meeting of the Organization for Human Brain Mapping (2006). 937

938 This work was supported by the European Research Council (ERC-2015-CoG 682591). 939 Author contributions 940 M.C.H., J.E.D., J.H., M.J.A.G.H. and E.H. designed the study. M.C.H. and J.V. implemented and 941 conducted the experiment. M.C.H., L.W. and J.V. analysed the results with support of E.H.. L.W. and 942 M.C.H. wrote the paper with contributions from all authors. 943 Competing interests 944 The authors declare no competing interests. 945 946 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 27 Supplementary Information 947 Valence-specific response characterization 948 At the end of the experiment, participants underwent a simplified version of the main experimental 949 task, in which category exemplars were replaced by colored squares. This task was used to 950 investigate to what extent sin conductance responses (SCRs) and pupil dilation responses (PDRs) can 951 be used to disentangle anticipation of shock and reward. Participants viewed three different 952 coloured squares and learned that one colour was associated with shocks (CS+S), one colour with 953 rewards ( CS+R) and one colour served as CS -. The trial structure was otherwise identical to 954 comparable trials from the acquisition and CC phases. At the end of the task, participants were asked 955 to rate the three stimuli on valence and arousal self -assessment maniki n scales (Bradley & Lang, 956 1994). 957 During this valence-specific response characterization task, we observed habituation in SCRs over the 958 course of the task (CS -type (CS+ S, CS+ R, CS -) x Phase (early, late) x Group (CC, Ext) rmANOVA, main 959 effect phase: F( 1,38)=13.921, p=0.001, η²=0.268) and different SCR magnitudes for the three different 960 CS-types (main effect CS -type (CS+S, CS+R, CS -): F( 2,76)=78.460, p<0.001, η²=0.674). In addition, 961 habituation depended on CS -type (CS -type x Phase interaction: F( 2,76)=6.825, p=0.002, η²=0.152). 962 During the early phase, SCRs in response to the CS+R and the CS - were not distinguishable 963 (t(40)=0.115, p=0.909, CS+R: 0.32 ±0.03, CS -:0.32 ±0.03), while during the late phase, SCRs to the 964 CS+R were larger than the CS - (t(40)=4.993, p< 0.001, CS+R: 0.29±0.03, CS -:0.19±0.02). SCRs to the 965 CS+S were consistently larger than SCRs to the CS+R (early: t(41)=9.345, CS+S: 0.62±0.04, p<0.001, 966 late: t(40)=5.952, p<0.001, CS+S: 0.56±0.04) and the CS - (early: t(40)=10.020, p<0.001, late: 967 t(4)=10.122, p<0.001). Thus, anticipation of aversive reinforcement (CS+S) led to increased SCRs 968 compared to anticipation of reward (CS+R) and CS - presentation throughout the task. Due to the fact 969 that SCRs performed less well in differentiation between CS+R and CS -, we focused our analyses on 970 PDRs, but report SCR results here as well. 971 We also observed CS-type dependent differences in PDRs (CS-type (CS+ S, CS+ R, CS-) x Phase (early, 972 late) x Group (CC, Ext) rmANOVA, main effect CS -type (CS+S, CS+R, CS -): F( 2,68=19.783, p<0.001, 973 η²=0.368). In comparison to the neutral CS -, both the shock-reinforced CS+ (CS+S) and reward -974 reinforced CS+ (CS+R) evoked larger PDRs (Supplementary Figure 1A, t(36)=7.071, p<0.001 and 975 t(26)=4.900, p<0.001 respectively, CS+S: 1.05 ±0.03, CS+R: 1.04 ±0.04, CS -: 1.01±0.02). However, 976 reward- and shock anticipation-induced PDRs did not differ statistically (t(36)=1.146, p=0.259). While 977 both shock anticipation and reward anticipation led to similar increases in PDRs as compared to the 978 neutral con dition, valence and arousal ratings indicated that participants experienced shock and 979 reward trials differently . Specifically, the CS+R was rated more positive than the CS - (t(47)=9.046, 980 p<0.001, CS+R: 7.79±0.14, CS-: 5.96±0.16, Supplementary Supplementary Figure 1C), while the CS+S 981 was rated less positive than the CS - (t(47)=-10.337, p<0.001, CS+S: 2.96±0.25). Participants reported 982 increased arousal to both the CS+S and CS+R as compared to the CS - (t(47)=4.666, p<0.001 and 983 t(47)=8.897, p<0.001 respectively, CS+S: 5.42±0.35, CS+R: 6.31±0.21, CS -: 3.33±0.30, Supplementary 984 Figure 1B). While it was not possible to distinguish PDRs to the CS+S and CS+R, explicit ratings of 985 arousal were marginally increased for the CS+R as compared to the CS+S (t(47)= -2.100, p=0.041). In 986 conclusion, the response characterization task showed that while anticipa tion of reward and shock 987 both generate increased PDRs as compared to the CS -, distinct retrospective valence ratings show 988 the expected directions. 989 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 28 Threat acquisition 990 Physiological and behavioural evidence for acquisition of conditioned threat responses 991 Participants pre-assigned to the CC and Ext groups underwent an identical threat acquisition 992 procedure. To verify that participants pre -assigned to both groups acquired conditioned threat 993 responses of comparable strength, we compared PDRs, explicit valence and arousal ratings , and 994 response times between groups. During the acquisition task, participants pre -assigned to both 995 groups showed stable and comparab le differential conditioned threat responses as measured by 996 PDRs (Supplementary Supplementary Figure A, CS-type (CS+, CS-) x Phase (Early, Late) x Group (CC, 997 Ext) rmANOVA, main effect CS -type: F( 1,37)=41.172, p0.2). Both groups also acquired comparable differential SCRs (main effect CS -999 type: F( 1,42)=58.633, p<0.001, η²=0.583 ), although SCRs showed habit uation over the course of the 1000 task (main effect phase: F(1,42)=66.907, p0.3 ). Thus, both SCRs and 1001 PDRs demonstrated comparable acquisition of conditioned threat responses between groups. 1002 Successful threat acquisition was further confirmed by valence and arousal ratings for the CS+ and 1003 CS- categories at the end of the acquisition task. Arousal ratings for the CS+ category exceeded 1004 arousal ratings for the CS - category (Supplementary Supplementary Figure B, CS-type (CS+, CS-) x 1005 Group (CC, Ext) rmANOVA, main effect CS -type: F(1,44)=27.573, p0.2). Similarly, the CS+ category was given lower valence (less positive) 1007 ratings than the CS - category (Supplementary Supplementary Figure C, CS-type (CS+, CS-) x Group 1008 (CC, Ext) rmANOVA , main effect CS -type: F( 1,44)=12.626, p0.7), the effect of CS -category unexpectedly differed 1010 between the CC and Ext group (CS -type x Group interaction: F( 1,44)=4.512, p=0.039, η²=0.093), due to 1011 more positive ratings to the CS - category in the Ext group (CC: 5.8±0.4, Ext: 6.9±0.3, t(44)=2.156, 1012 p=0.037). Nevertheless, v alence ratings for the CS+ category were comparable between groups (CC: 1013 5.1±0.5, Ext: 4.2 ±0.4, p>0.1), suggesting that the s trength of the acquired threat responses is likely 1014 similar between groups. 1015 Supplementary Figure 1. Pupil dilation responses (PDRs), explicit arousal and valence rating for the diffe rent CSs presented during the valence-specific response characterisation task. (A) PDRs to the shock reinforced (CS+S), reward reinforced (CS+R) and CS - stimuli, averaged across the task and all participants. PDRs were increased for the CS+S and CS+R as compared to the CS- (B) Explicit ratings of arousal and (C) valence provided immediately after the task. Explicit ratings of arousal for the CS+S exceeded ratings for the CS -, and the CS+R was rated higher in arousal than the CS+S. Valence ratings (1=extremely negative, 10=extremely positive) for the CS+R were more positive than ratings for the CS -, while ratings for the CS+S were more negative than for the CS - and CS+R. Error bars represent ± standard error of the mean *=p<0.05, ***=p<0.001 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 29 1016 To keep all experimental tasks similar between groups, participants in both groups were asked to 1017 respond to targets that were superimposed on the stimuli as quickly as possible. T o verify that both 1018 groups performed similarly on this task, we compared response times for the different stimuli 1019 between the groups. During the acquisition task, participants responded faster to targets in CS+ trials 1020 compared to CS - trials ( CS-type (CS+, C S-) x Group (CC, Ext) rmANOVA , main effect CS -type: 1021 F(1,45)=10.839, p=0.002, η²=0.194), with no differences between groups (all p’s>0.058). 1022 After the acquisition task, participants in both groups reported higher estimated reinforcement rates 1023 for the CS+ ca tegory as compared to the CS - category (CS-type (CS+, CS -) x Group rmANOVA, main 1024 effect CS -type: F( 1,45)=82.176, p0.3). 1026 Brain activation supports successful acquisition of conditioned threat responses 1027 The acquisition of conditioned fear on the first day reliably activated networks associated with fear 1028 conditioning. Whole -brain analysis identified regions that were more responsive to the CS+ versus 1029 the CS- category (Supplementary Figure 3, see Supplementary Table 1 for a complete overview of 1030 findings). We observed differential BOLD responses in a large number of brain areas, including the 1031 bilateral insula, posterior and anterior cingulate, thalamus, precuneus ( undirected test, cluster size = 1032 425400 mm 3, p<0.001, whole-brain FWE-corrected) and the bilateral amygdala (right cluster size = 1033 1088 mm3, p<0.001, FWE-SVC, left cluster size = 736 mm3, p<0.001, FWE-SVC). 1034 Supplementary Figure 2. Differential PDRs during acquisition and explicit ratings of arousal and valence provided after acquisition. (A) Differential PDRs for the early (light red) and late (dark red) phase of the acquisition task, (B) arousal and (C) valence ratings, displayed separately for participants assigned to the counterconditioning (CC, solid bars) and extinction (EXT, open bars) groups. Both groups showed comparable di fferential PDRs and arousal ratings during the acquisition task. Participants in both groups showed negative differential valence ratings (stronger negative valence for CS+ vs. CS -), although this effect was stronger in the Ext group. Error bars represent ± standard error of the mean . *, p<0.05. ≠. Significantly different from 0 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 30 1035 Supplementary Figure 3. Differential threat responses during acquisition revealed CS-specific activation of clusters 1036 encompassing a range of regions including the bilateral insula, thalamus, precuneus, anterior cingulate and midbrain . 1037 Group F -image of the effect of CS type, thresholded at cluster -level FWE-corrected p<0.05, cluster -forming threshold 1038 p=0.001, displayed on the single-subject high-resolution T1 volume provided by the Montreal Neurological Institute (MNI). 1039 Supplementary Table 1. Whole-brain main effects of group (CC, Ext), CS type (CS+, CS-) and phase (early, late) and 1040 interactions, during the acquisition task. Cluster-forming threshold p=0.001, FWE-corrected at p<0.05, clusters were 1041 labelled using the Talairach Deamon atlas and the AAL atlas for ROIs. For each cluster, the peak voxel coordinates (MNI 1042 space) and regions are reported, and additional regions contained within the cluster are added in italics. 1043 Peak MNI coordinates Region Cluster x y z Size (mm3) pFWE (cluster) Peak F-value Direction CS-type x phase Parahippocampa Gyrus L Insula L, Parahippocamal Gyrus Hippocampus L, Claustrum L, Lentiform Nucleus Putamin L, Uncus L, Postcentral Gyrus BA43 L 1 -18 -10 -16 6656 0.005 25.86 Early CS+ > Late CS+ Parahippocampa Gyrus Amygdala R Inferior Frontal Gyrus R, Subcallosal Gyrus BA34R 2 20 -4 -22 2116 0.033 25.34 Culmen L Declive L, Lingual Gyrus L 3 -8 -54 -16 1720 0.027 20.57 Parahippocampal Gyrus L Parahippocampal gyrus BA36L/BA30L, Culmen L 4 -20 -42 -2 5236 0.006 31.65 Medial Frontal Gyrus BA11 L Anterior Cingulate BA32L, Medial Frontal Gyrus BA10 R, BA11 R 5 -4 38 -14 1104 0.046 17.23 Superior Temporal Gyrus L Middle Temporal Gyrus BA21/BA22 L 6 -52 10 -14 3056 0.015 28.06 Lingual Gyrus BA18/BA19 R 7 -6 -68 -2 2584 0.018 35.21 Insula R Inferior Parietal Lobule R, Superior Temporal Gyrus BA22 R, Postcentral gyrus BA3 R, Superior Temporal Gyrus BA22 R, Precentral Gyrus BA4/BA6, Inferior Parietal Lobule BA40, Middle temporal gyrus, Superior temporal gyrus BA42 8 38 -6 18 24488 0.001 27.62 Parahippocampal Gyrus R 9 24 -36 -4 1432 0.035 22.26 Inferior Frontal Gyrus BA45 R 10 52 14 14 1392 0.036 21.37 Precentral Gyrus L 11 -60 -8 32 5640 0.006 22.09 Inferior Parietal Lobule BA40L Postcentral gyrus BA2L 12 -56 -36 42 1104 0.046 20.42 Precuneus L Postcentral gyrus L, cingulate gyrus L 13 -14 -42 54 1496 0.034 19.85 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 31 Precuneus R Paracentral Lobule Ba7 R, Precuneus R, Cingulate gyrus R, Superior Parietal Lobule BA7 R 14 20 -52 54 5528 0.006 24.86 Medial Frontal gyrus L (23) Medial frontal gyrus BA6LR, Paracentral Lobule L 15 -6 -20 64 1840 0.025 21.47 CS-type Postcentral Gyrus L Inferior Parietal Lobule LR, Insula LR, Postcentral gyrus R, Cingulate Gyrus LR, Thalamus LR, Caudate LR, Inferior- Middle- and Superior Frontal Gyrus LR, Posterior Cingulate R, Precentral Gyrus LR, Precuneus L, Delice R, Culmen R, Cuneus L, Superior Temporal Gyrus LR, Anterior Cingulate LR, Parahippocampal Gyrus BA27 R, Lentiform nucleus LR 1 -50 -20 16 425400 CS- Posterior Cingulate BA31 L Precuneus M 2 -4 -56 24 2816 0.021 26.17 CS+<CS- Corpus Callosum M Corpus Callosum R 3 0 0 22 1296 0.049 35.45 Angular Gyrus R Angular Gyrus BA39 R, Precuneus R 4 56 -66 30 2432 0.024 31.42 Angular Gyrus BA39L 5 -54 -68 30 5584 0.010 36.02 Superior Frontal Gyrus BA9L Superior Frontal Gyrus BA8L, Middle frontal gyrus BA6L 6 -18 40 42 7200 0.007 33.18 Phase Superior Temporal Gyrus LR, Inferior Parietal Lobule R, Middle Temporal Gyrus LR, Inferior- Middle- and Superior Frontal Gyrus LR, Caudate LR, Middle Occipital Gyrus LR, Cingulate Gyrus LR, Anterior Cingulate LR, Declive LR, Precuneus LR, Insula LR, Culmen LR, Superior Temporal Gyrus LR, Lingual Gyrus LR, Fusiform Gyrus LR, Angular Gyrus R, Claustrum LR, Thalamus LR, Parahippocampal Gyrus LR, Cuneus LR 1 -64 -38 12 784632 Late Counterconditioning and extinction 1044 Skin conductance responses 1045 Differential SCRs were still apparent during the CC/extinction task (CS-type (CS+, CS-) x Phase (Early, 1046 Late) x Group (CC, Ext) rmANOVA, main effect CS -type: F( 1,40)=17.609, p<0.001, η²=0.306). To verify 1047 that successful extinction was reached by the end of the task, we explored SCRs in the late phase 1048 separately, but found that differential SCRs were still apparent in that phase (F( 1,41)=12.166, p=0.001, 1049 η²=0.229). Finally, we explored whether the last two trials of the extinction task showed evidence of 1050 residual differential SCRs. In the last two trials of extinction, across both groups, there was no 1051 evidence for differential SCRs (all p’s>0.2). Thus, while differential SCRs persisted during the late 1052 phase of the extinction task, differential responses were no longer apparent in the last two trials. 1053 Throughout the CC/extinction, there was no evidence for different SCRs between groups, suggesting 1054 that participants in both groups underwent comparable but slow extinction of differential SCRs. 1055 Overlapping stimulus-specific activation during counterconditioning and extinction 1056 A number of clusters showed comparable stimulus-specific activations during CC and extinction 1057 (Supplementary Table 2). 1058 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 32 Supplementary Table 2. Whole-brain main effect of CS-type during the CC/extinction task . Cluster-forming threshold 1059 p=0.001, FWE-corrected at p<0.05, clusters were labelled using the Talairach Daemon atlas and the AAL atlas for ROIs. For 1060 each cluster, the peak voxel coordinates (MNI space) and regions are reported, and additional regions contained within the 1061 cluster are added in italics. 1062 Peak MNI coordinate Region Cluster x y z Size (mm3) pFWE (cluster) Peak F- value Direction CS-type Caudate Head L Thalamus LR, Caudate Head R, Substantia Nigra LR 1 -10 10 -2 25136 0.001 60.98 CS+>CS- Insula R Inferior Frontal Gyrus R, Precentral Gyrus BA44 R, Inferior Frontal Gyrus BA45 R 3 28 26 0 22800 0.001 89.75 Inferior Frontal Gyrus L Insula BA13 L 4 -32 28 0 7808 0.004 52.01 Lingual Gyrus L Inferior Occipital Gyrus L, Cuneus L, Middle Occipital Gyrus L 5 -24 -80 -12 5696 0.006 32.81 Superior Temporal Gyrus R Transverse Temporal Gyrus R 6 50 -18 6 3744 0.012 30.65 Lingual Gyrus R Cuneus R 7 8 -94 6 3688 0.012 27.81 Superior Temporal Gyrus L Transverse temporal Gyrus L 8 -44 -24 8 3864 0.011 44.50 Anterior Cingulate BA32 R Medial Frontal Gyrus BA8 R, Anterior Cingular LR, Cingulate Gyrus BA32 R 9 4 38 20 7064 0.004 26.61 Superior Temporal Gyrus R Supramarginal Gyrus R, Inferior Parietal Lobule BA40R 10 64 -34 14 3720 0.012 25.88 Superior Temporal Gyrus L 11 -60 -46 16 2504 0.023 36.62 Cingulate Gyrus L Posterior Cingulate BA23R, Posterior Cingulate L 13 -6 -20 30 3928 0.011 38.33 Angular Gyrus L Middle Temporal Gyrus L, Angular Gyrus BA39 L 12 -44 -64 32 4104 0.010 23.58 CS->CS+ Inferior Temporal Gyrus BA21 L, Middle Temporal Gyrus BA21 L 2 -64 -10 -22 1696 0.039 27.05 Angular Gyrus R Supramarginal Gyrus R 14 44 -66 34 1392 0.050 18.35 Postcentral Gyrus BA40R Precentral Gyrus Ba4/BA3 R 15 34 -40 58 1704 0.038 19.84 Middle Frontal Gyrus BA8/BA6 L 16 -24 16 48 4576 0.008 34.03 Spontaneous recovery test 1063 Skin conductance responses 1064 To investigate whether CC can prevent spontaneous recovery of differential SCRs, SCRs during the 1065 last two trials of extinction and the first two trials of the spontaneous recovery test were submitted 1066 to a CS -type (CS+, CS -) x Phase (last 2 trials of the CC/extinction phase, first two trials of the 1067 spontaneous recovery test) x Group (CC , Ext) rmANOVA. SCRs showed a generalized increase from 1068 the last two trials of extinction to the first two trials of the spontaneous recovery test (main effect 1069 phase: F(1,38)=32.392, p<0.001, η²=0.460)). There was evidence for differential SCRs across both 1070 phases (main effect CS -type: (F(1,38)=9.560, p=0.004, η²=0.201), as CS+ stimuli evoked higher SCRs 1071 than CS - stimuli (t(43)=2.518,p=0.016, CS+:0.41 ±0.03, CS -:0.35±0.03), yet we did not find evidence 1072 for CS+ -specific spontaneous recovery or effect of group (all p’s>0.4). Thus, although there was a 1073 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 33 generalized increase in responding from the end of extinction to the start of the spontaneous 1074 recovery test, SCRs did not show differential recovery and were comparable between groups. 1075 Brain activation 1076 During the spontaneous recovery test, CS-specific activation differed between groups in the inferior 1077 temporal gyrus (cluster size = 2008 mm3, p=0.020, FWE-corrected, Supplementary Figure 4A) and the 1078 inferior frontal gyrus (cluster size = 1920 mm 3, p=0.022, FWE-corrected, Supplementary Figure 4B). 1079 Separate analysis of the spontaneous recovery phase within each group did not reveal any 1080 suprathreshold clusters in the Ext group, while a number of clusters showed stimulus -specific 1081 activation i n the CC grou p. Spe cifically, the CC group showed stimulus -specific activation in the 1082 bilateral fusiform gyri, superior parietal lobes and inferior frontal gyri, and in the right thalamus, 1083 caudate, middle frontal gyrus, and angular gyrus (see Supplementary T able 3). A priori defined 1084 regions of interest (ROIs) during the spontaneous recovery task were submitted to a Group (CC, Ext) x 1085 CS-type (CS+, CS-) x Phase (early, late) rmANOVA but did not reveal any effects. 1086 1087 Supplementary Figure 4. During the spontaneous recovery test, stimulus type-specific activation of the inferior temporal 1088 and frontal gyri differed between groups. The inferior temporal Gyrus (A) and Inferior frontal gyrus (B) show increased 1089 CS+-specific activation in the CC gro up as compared to the Ext group. Group F -images thresholded at FWE -corrected 1090 p<0.05, cluster -forming threshold p=0.001, displayed on the single-subject high-resolution T1 volume provided by the 1091 Montreal Neurological Institute (MNI) and parameters estimates from peak voxels. 1092 Supplementary Table 3. Peak voxel coordinates and statistics of activations during the spontaneous recovery phase in 1093 the CC group . Clusters were labelled using the AAL atlas. For each cluster, the peak voxel coordinates and regions are 1094 reported, and additional regions contained within the cluster are added in italics. Clusters are whole-brain FWE-corrected at 1095 p<0.05. 1096 Peak MNI coordinate Region Cluster x y z Size (mm3) pFWE (cluster) Peak T- value Direction CS-type Thalamus R Parahippocampal Gyrus R 1 10 -22 -4 2160 CS- Inferior temporal Gyrus R Fusiform gyrus R 2 46 -52 -2 4856 <0.001 6.56 Inferior frontal gyrus, triangular R 3 58 26 26 4992 <0.001 5.93 Superior parietal lobe R Angular Gyrus R 4 26 -60 50 2048 <0.001 5.36 Inferior frontal gyrus, orbital part L 5 50 26 -6 1120 0.019 5.31 Fusiform gyrus L Lingual gyrus 6 -38 -80 -18 1176 0.015 5.18 Caudate R 7 6 0 10 1952 <0.001 4.87 Middle Frontal gyrus R 8 37 0 44 992 0.03 4.86 Superior parietal lobule L Angular gyrus L 9 -32 -58 58 1480 0.004 4.53 1097 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint 34 Reinstatement test 1098 SCRs showed a generalized increase from the spontaneous recovery phase to the reinstatement test 1099 (CS-type (CS+, CS -) x Group (CC, Ext) x Phase (spontaneous recovery test, reinstatement test) 1100 rmANOVA, main effect Phase: F( 1,39)=25.758, p<0.001, η²=0.398, last two trials of spontaneous 1101 recovery: 0.22 ±0.04, first two trials of reinstatement: 0.38±0.03 ). Across the last two trials of the 1102 spontaneous recovery test and the first two trials of the reinstatement test, differential SCRs differ 1103 between the counterconditioning and extinction group (interaction effect of stimulus type and 1104 group: F( 1,39)=4.967, p=0.032, η²=0.113). Yet, there is no evidence for differential reinstatement 1105 between groups (no CS-type x Phase x Group interaction, p=0.218). Moreover, mean SCRs to CS+ and 1106 CS- stimuli do not differ within either group (all p’s>0.12). 1107 PDRs showed a generalized decrease from the spontaneous recovery phase to the reinstatement test 1108 (CS-type (CS+, CS -) x Group (CC, Ext) x Phase (spontaneous recovery test, reinstatement test) 1109 rmANOVA, main effect of Phase (F( 1,29)=9.104, p=0.005, η²=239)). Mean PDRs decreased from 1110 spontaneous recovery to reinstatement (t(30)=3.063, p=0.005, last two tr ials of spontaneous 1111 recovery: 1.04±0.01, first two trials of reinstatement: 1.01±0.01). Given that we did not observe 1112 successful reinstatement in either group, our reinstatement test does not inform us about whether 1113 CC can lead to a more persistent attenuation of fear as compared to classic extinction. 1114 CS+-specific enhancement of recognition memory depends on CS+ category 1115 Corrected recognition scores (pHits – pFA) were subjected to a task (acquisition, CC/extinction task) x 1116 CS-type (CS+, CS -) x Group (CC, E xt) rmANOVA, including CS+-category (animals, tools) as covariate. 1117 Although the effect of CS -type differed depending on the category used as CS+ (CS -type x CS+ -1118 category interaction: F( 1,42)=19.400, p<0.001, η²=0.316) and task (CS-type x CS+ -category x task 1119 interaction: F(1,43)=5.375, p=0.025, η²=0.113), showing a stronger effect for tools as CS+, this did not 1120 differ between our experimental groups (CS-type x CS+ -category x Group interaction: F(1,41)=0.050, 1121 p=0.824; CS-type x task x CS+-category x Group interaction: F(1,41)=0.005, p=0.946). 1122 To further investigate to what extent CC retroactively affected memory for items presented during 1123 the acquisition task, we examined recognition of items presented during acquisition and the 1124 CC/extinction tasks separately. Retrospective memory enhancement for the CS+ items compared to 1125 CS- items differed depending on the CS+ category during both the acquisition task (CS-type x CS+ 1126 category interaction: F( 1,42)=29.730, p<0.001, η²=0.414, CS+ category main effect: F( 1,42)=5.346, 1127 p=0.026, η²=0.113) and the CC/extinction task (CS-type x CS+ category interaction: F( 1,42)=5.121, 1128 p=0.029, η²=0.1 09), showing a stronger effect for tools as CS+. Importantly, this effect was 1129 comparable between groups (CS-type x CS+ category x Group interaction: acquisition task 1130 F(1,41)=0.016, p=0.806; CC/extinction task F(1,41)=0.019, p=0.890). 1131 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted July 30, 2024. ; https://doi.org/10.1101/2024.07.29.605706doi: bioRxiv preprint