Dorsomedial striatal neuroinflammation causes excessive goal-directed action control by disrupting astrocyte function

Compulsive acQons have been confusingly described as reflecQng both excessive habitual and excessive goal-directed acQon control. Here we sought to resolve this contradicQon by inducing the neuropathology commonly observed in individuals with compulsive disorders, specifically by causing neuroinflammaQon in the dorsomedial striatum of rats. We found that this caused rats to be excessively goal-directed, acquiring and maintaining goal-directed acQons under condiQons that would otherwise produce habits. Immunohistochemical findings suggested that these behaviours were a result of astrocyQc proliferaQon and its effects on neuronal acQvaQon. We therefore invesQgated the role of striatal astrocytes specifically, demonstraQng that chemogeneQcally acQvaQng the Gi-pathway in astrocytes altered the firing properQes of nearby medium spiny neurons and modulated goal-directed acQon control. Together, results suggest that striatal neuroinflammaQon is sufficient to cause excessive goal-directed acQon control through the dysregulaQon of astrocyte funcQon suggesQng that individuals with striatal neuroinflammaQon are excessively goal-directed rather than habitual, informaQon that could be used to direct future intervenQons and/or treatments. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 3 Individuals with compulsive disorders perform acQons repeQQvely, o^en against their desires and despite negaQve consequences. This, coupled with observaQons of aberrant neural acQvity in the corQco-striatal-thalamic loops that underlie acQon control, has led to widespread speculaQon that compulsions arise from a disrupQon in the balance between goal-directed acQons and habits. However, there is debate regarding the proposed direcQon of this disrupQon, with some researchers arguing that disorders like substance use disorder (SUD) and obsessive-compulsive disorder (OCD) are best characterised as an overreliance on habits (1, 2), whereas others argue that they reflect excessive goal-directed control (3, 4). Here we address this quesQon in a causal manner by inducing the neuropathological changes associated with disorders of compulsion (including SUD (5) and OCD (6), but also Parkinson’s and HunQngton’s diseases which o^en feature compulsivity (7, 8)) and examining the consequences for acQon selecQon. Specifically, we first invesQgated whether regionally-specific striatal neuroinflammaQon in rats disrupted the balance between goal- directed and habitual acQon control, then invesQgated the astrocyte-specific mechanism of that disrupQon. The neural circuit of goal-directed and habitual acQons has been extensively invesQgated over the last three decades and has been found to have considerable homology between rodents, primates, and humans (9–11). FoundaQonal studies (10, 11) revealed that goal-directed and habits acQons are controlled by disQnct but parallel circuits in the corQco- striatal network, with lesioning or inacQvaQng structures within one circuit producing a shi^ from one acQon control system to the other, mirroring the behavioural changes seen in individuals with compulsive disorders. However, the experimental approaches above do not adequately mimic the neural circuit disturbances in individuals with compulsive disorders, who do not exhibit widespread neuronal silencing or death (14, 15), and if they do, it usually .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 4 emerges long a^er symptoms appear (16). Therefore, it remains an open quesQon how shi^s in acQon control emerge in such individuals. InteresQngly, both post-mortem and neuroimaging studies have reported increased markers of neuroinflammaQon in the striatum and other regions in such individuals (6, 17). Accordingly, we aimed to mimic this neuropathology in rats so that we could examine the causal consequences of this treatment for goal-directed and habitual acQon control. NeuroinflammaQon was achieved by stereotacQcally injecQng lipopolysaccharide (LPS, 5mg/mL, n=16; Sham controls, n = 14), an endotoxin and neuroinflammatory mimeQc, into the striatum. We targeted the posterior dorsomedial striatum (pDMS, Fig. 1B-C) because it is the rodent homologue of the caudate, a region that is known to exhibit abnormal neural acQvity and elevated neuroinflammatory markers in individuals with substance use disorder (17, 18) and OCD (14, 19, 20). In rodents, pDMS is considered the neural locus of acQon control, as it is the only brain region known to be involved in both the learning and performance of goal-directed acQons (12, 21). Because we recently showed that neuroinflammaQon in the hippocampus of mice increased neuronal acQvaQon (22) we expected a similar effect in the pDMS. We further anQcipated that this increased acQvaQon would enhance the typical funcQon of pDMS to produce ‘excessive’ goal-directed acQon control, which we defined as animals exerQng goal-directed control when expected not to. Therefore, for the first series of experiments we adopted experimental parameters under which we expected our Sham animals to show impaired acQon selecQon, by feeding them a standard laboratory chow that is relaQvely high in fat and protein (see supplemental

and Table 1 for details) on a mild deprivaQon schedule (approx. 90% of their iniQal body weight) to induce low levels of hunger and arousal. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 5 We first tested acQon selecQon under the guidance of sQmuli using the specific Pavlovian-instrumental transfer design shown in Figure 1A. Rats were trained to associate two condiQoned sQmuli (CSs; a white noise or clicker) with two unique outcomes (sucrose soluQon or pellet), then to press a le^ and right lever for the same outcomes. On test, rats were presented with the clicker and noise together with the levers for the first Qme, but no food. We expected Sham animals to respond equally on each lever regardless of the sQmulus presented (due to the mild deprivaQon condiQons) but expected LPS animals to be excessively goal-directed and respond selecQvely on the lever that had earned the ‘same’ outcome as the sQmulus being presented. That is, the pellet-associated sQmulus should elicit presses on the pellet lever more than the sucrose lever, and the sucrose sQmulus should elicit presses on the sucrose lever: Same > Different. This predicQon was confirmed (Fig. 1D). That is, despite entries into the food magazine and lever press responses not differing between Sham and LPS groups during any phase of acquisiQon (largest F(1,28) = 1.085, p = 0.362, Supp. Fig. 1A-C), on test there was a group x transfer interacQon, F (1,28) = 5.710, p = 0.024, which was explained by a significant simple effect of transfer (Same > Different) for group LPS, F (1,28) = 15.996, p = 0.000, but not Shams (Same = Different), F < 1. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 6 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 7 Figure 1. Striatal neuroinflamma2on causes excessive goal-directed ac2on control in a regionally specific manner. (A) Pictorial representaQon of the experimental procedures for Pavlovian-instrumental transfer, outcome devaluaQon, and outcome-selecQve reinstatement, created with Biorender. (B) DistribuQon and locaQons of the lipopolysaccharide (LPS) injecQons in the posterior dorsomedial striatum (pDMS) included in the analysis. (C) pDMS image showing LPS placement as labelled with GFAP (glial fibrillary protein) and IBA1 (ionized calcium binding adapator molecule 1), (D-H) Individual data plots and mean lever presses during the (D) Pavlovian-instrumental transfer test, (E-F) outcome devaluaQon test, (G) consumpQon, and (H) outcome-selecQve reinstatement test under mild deprivaQon condiQons following pDMS LPS injecQons. (I-M) Individual data plots and mean lever presses during the (I) Pavlovian-instrumental transfer test, (J-K) outcome devaluaQon test, (L) consumpQon, and (M) outcome-selecQve reinstatement test under moderate (standard) deprivaQon condiQons following pDMS LPS injecQons. (N) Magazine entries per min (±SEM) during Pavlovian condiQoning. (O-Q) Individual data plots and mean lever presses during the (O-P) outcome devaluaQon test, and (Q) consumpQon following NAc core LPS injecQons. * denotes p < 0.05. (pDMS: n = 14 (SHAM), n = 16 (LPS), N = 30; NAc core: n = 14 (SHAM), n = 14 (LPS), N = 28). Next, rats were tested for goal-directed acQon control in the absence of sQmuli using the outcome devaluaQon design shown in Figure 1A. Rats were retrained to press the two levers for their disQnct outcomes, then given a choice test in which both levers were presented without any outcomes delivered. Immediately prior to this test, we reduced the value of either pellets or sucrose using sensory-specific saQety. If goal-directed, animals should prefer the lever associated with the sQll-valued outcome and avoid the lever associated with the devalued outcome, demonstraQng that their acQons were moQvated by a) the current value of the outcome, and b) the conQngency between the acQon and outcome: the two criteria of goal-direcQon (9, 23). Again, due to the mild deprivaQon condiQons we again predicted that the devaluaQon effect (Valued > Devalued) would be auenuated for group Sham and intact for group LPS. This predicQon was confirmed (Fig. 1E) indicated by a group x devaluaQon interacQon, F (1,28) = 4.878, p = 0.035, composed of a small but significant, simple effect for group Sham, F (1,28) = 7.445, p = 0.011, and a larger .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 8 effect for group LPS, F (1,28) = 31.060, p = 0.000. This difference was confined to the instrumental response because entries into the food magazine did not differ between groups, F < 1 (Fig. 1F) and nor did consumpQon during pre-feeding, F < 1 (Fig. 1G). Following a day of instrumental retraining, rats were tested for outcome-selecQve reinstatement as again illustrated in Figure 1A. For this test, responding on each lever was exQnguished for 15 min, followed by two unsignalled presentaQons of each outcome. These presentaQons typically reinstate responding selecQvely on the lever that had earned that outcome during training, i.e. sucrose presentaQons elicit presses on the sucrose lever and pellet presentaQons on the pellet lever (24, 25). Because selecQve reinstatement is a measure of acQon selecQon that is not goal-directed (24), we expected that it would remain unaffected by both the mild deprivaQon condiQons and pDMS neuroinflammaQon. This was confirmed, as reinstatement was intact (Reinstated > Nonreinstated) for both groups (Fig. 1H) supported by a main effect of reinstatement, F (1,28) = 67.951, p = 0.000, that did not interact with group, F < 1. To assess whether group differences persisted under moderate deprivaQon condiQons, we switched rats to a lower fat, lower protein laboratory chow and increased the deprivaQon schedule to maintain rats at approximately 85% of iniQal body weight (i.e. the standard approach for these kinds of studies, see supplemental methods for details). A^er brief retraining, we retested the rats using the same transfer, devaluaQon, and reinstatement tests. This Qme, performance did not differ between groups on any test. Indeed, on the transfer test both groups responded more on the Same relaQve to the Different lever: main effect, F (1,28) = 30.605, p = 0.000 (Figure 1I), interacQon with group, F < 1. Likewise, on devaluaQon both groups pressed the Valued more than the Devalued lever; main effect, F .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 9 (1,28) = 12.378, p = 0.000 (Figure 1J), which also did not interact with group, F NonReinstated); main effect (Reinstated > Nonreinstated), F (1,28) = 57.780, p = 0.000 (Figure 1M), and no interacQon with group, F < 1. We next tested whether these effects were regionally specific by injecQng LPS into the nucleus accumbens core (NAc core; n=14 LPS, n=14 Sham controls, a ventral striatal region involved in the performance of goal-directed acQons (26, 27)). This Qme, all training and tesQng occurred under the moderate deprivaQon/hunger schedule (lower fat/protein chow at 85% bodyweight) because a pilot study revealed that neuroinflammaQon in the NAc core was unlikely to produce the same increased propensity for goal-directed control. LPS in the NAc core did not affect instrumental responding during any stage of training or test (Supp. Fig. 1G, Fig. 1O,) but did increase the number of entries rats made into the food magazine during Pavlovian condiQoning, as revealed by a main effect of group, F (1,25) = 6.962, p = 0.014 (Fig. 1N), and during devaluaQon tesQng, as indicated by a group main effect, F (1,26) = 6.02, p = 0.021 (Fig. 1P). Again, this difference was not due to any influence of LPS on feeding or appeQte, because outcome consumpQon during prefeeding did not differ between groups (F < 1, Fig. 1Q). Pavlovian condiQoning and devaluaQon tesQng were the two phases of the experiment for which lever pressing was absent (during Pavlovian condiQoning) or low (during devaluaQon tesQng, due to rats being prefed to saQety). This pauern of results therefore suggests that NAc core neuroinflammaQon increased the propensity for Pavlovian responding, but that this effect was masked when response compeQQon from lever pressing was present (i.e. lever pressing and magazine entries compete because rats cannot perform .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 10 both at the same Qme). Because an enhanced propensity to respond to Pavlovian cues is also claimed to contribute to compulsive-like tendencies (28, 29), if translatable, these

suggest that differenQal distribuQons of neuroinflammaQon throughout the striatum could be a mulQfaceted source of compulsivity. Following confirmaQon that the effects of pDMS neuroinflammaQon on goal-directed acQon are region-specific, we aimed to determine whether it could also produce excessive goal-directed control under moderate deprivaQon condiQons, specifically when using experimental parameters that have been reliably shown to produce habits (Valued = Devalued) (30, 31). To achieve this, we trained a naïve cohort of rats (n=23 LPS, n=18 Sham) to press a single lever for sucrose on an interval schedule whilst being maintained at approximately 85% of their iniQal body weight on the low fat/low protein chow described. We then administered a progressive raQo test to determine whether pDMS neuroinflammaQon had altered moQvaQon per se, followed by another devaluaQon test which was performed a^er half of the animals were subjected to taste aversion training (shown in Fig. 2A). Specifically, the Devalued group received 3 x pairings of sucrose with lithium chloride injecQons to induce malaise, whereas the Valued group received saline injecQons. We expected Sham animals to show evidence of habits (Valued = Devalued) and group LPS to remain goal-directed (Valued > Devalued). Once again, there were no group differences during lever press acquisiQon, although the LPS group did appear to lever press marginally more than Shams (Fig. 2B), main effect of group F(1,39) = 3.36, p = 0.074. Importantly, however, the number of acQon-outcome pairings did not differ between groups, F < 1 (Fig. 2C). During progressive raQo tesQng, .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 11 animals with pDMS neuroinflammaQon took longer than Shams to reach ‘breakpoint’ (i.e. 5 minutes without responding) despite the number of lever presses required to earn a sucrose Figure 2. pDMS neuroinflamma2on prevents the forma2on of habits. (A) Pictorial representaQon of the outcome devaluaQon procedure designed to promote habits created with Biorender. (B) Lever pressing per min (±SEM) during instrumental condiQoning and (C) Number of acQon-outcome pairings (±SEM) during instrumental condiQoning, (D) Breakpoint (±SEM) obtained during the 2-h, 3-day Progressive RaQo tesQng schedule, (E) Lever presses during progressive raQo tesQng (±SEM) presented as a percentage of baseline responding, (F) Individual data plots and mean lever presses during the outcome devaluaQon ‘habit’ test. * denotes p < 0.05. (n = 18 (SHAM), n = 23 (LPS), N = 41) outcome increasing by 5 each Qme. There was a main effect of group, F (1,39) = 15.15, p = 0.0004 (Fig. 2D) that remained even a^er data was corrected for slightly higher lever presses in group LPS at baseline, F (1,39) = 6.243, p = 0.0168 (Fig. 2E). .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 12 As expected, when rats should have been habitual pDMS neuroinflammaQon produced goal-directed acQon control (Fig. 2F). There was a group x devaluaQon interacQon, F (1,37) = 4.373, p = 0.043 consisQng of intact devaluaQon in group LPS (Valued > Devalued), F (1,37) = 20.198, p = 0.000, and not Shams (Valued = Devalued), F (1,37) = 1.417, p = 0.241. These findings show that pDMS neuroinflammaQon increases moQvaQon and causes excessive goal-directed control, which is perhaps unsurprising given the high interdependence of the two processes (32). Most importantly, however, they reinforce the noQon that pDMS neuroinflammaQon produces goal-directed control that is excessive, because it prevented the formaQon of habits which, under non-pathological condiQons, are adapQve automated behaviours that allow for more rapid and efficient acQon execuQon. Indeed, the inability to form habits has been argued to underscore several cogniQve deficits in condiQons such as Parkinson’s disease (33). Our final aim was to invesQgate why pDMS neuroinflammaQon might cause excessive goal-directed control. We turned to the results of our immunohistochemical analyses for clues and found that, although intensity levels, proliferaQon, and various morphological measures of both the astrocyte marker glial acidic fibrillary protein (GFAP) and microglial marker ionised-calcium binding adaptor molecule (IBA1) were elevated in LPS rats relaQve to Shams (see Supp. Fig. 3 for full results), only levels of GFAP significantly correlated with acQon selecQon on the tests for which performance differed between groups (Fig. 3F-I, although breakpoint responding which also correlated with IBA1 levels, Fig. S3C, suggesQng that that more microglia were associated with moQvaQon but not the selecQvity of acQons). Moreover, and consistent with our findings in hippocampus (22), neuronal acQvaQon was increased by pDMS neuroinflammaQon, as evidenced by higher levels of the .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 13 acQvaQon marker and immediate early gene c-fos co-occurring with NeuN in LPS groups relaQve to Shams (Fig. 3N-O). These levels also significantly correlated selecQvely with acQon selecQon on tests for which group differences were detected (Fig. 3J-M). Figure 3. Injec2ons of lipopolysaccharide (LPS) into posterior dorsomedial striatal (pDMS) increased the counts of GFAP , IBA1, and the percentage of c-fos co-localised with NeuN. GFAP and cFos/NeuN co-localisa2on was correlated with excessive ac2on control. (A-C) RepresentaQve images of pDMS from a Sham (top panel) and LPS-injected rat immunostained for DAPI (bouom panel) and (A) GFAP , (B) IBA1, (C) c-fos-NeuN, final graphs show individual data points and mean values for quanQficaQon of each, (D-E) RepresentaQve images of pDMS iummunostained DAPI/GFAP/IBA1/NeuN merged from a Sham (D) and LPS- injected (E) rat, (F-K) CorrelaQons between GFAP , IBA1, and c-fos-NeuN intensity and behavioural performances. * denotes that the p < 0.05. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 14 These immunohistochemical results point to pDMS neuroinflammaQon causing excessive goal-directed acQon control by altering the interacQons between astrocytes and neurons. To gain a more detailed understanding of how this might have occurred, we used in vitro whole cell patch clamp electrophysiology to determine whether LPS injecQons in pDMS altered the firing properQes of medium spiny neurons (MSNs). For this study, we once again bilaterally injected LPS (5mg/mL, n = 4) or saline (n = 3) into the pDMS of rats, then removed recorded from the pDMS in acute brain slices 6 weeks a^er injecQon. Recordings were first taken at resQng membrane potenQal (RMP, data shown in Supplemental Fig. 4) then repeated while cells were voltage-clamped at -80mV, consistent with the reported in vivo RMP of MSNs (34). MSNs classificaQon was based on acQon potenQal profiles (see supplemental methods for details). LPS MSNs displayed a more depolarised acQon potenQal (AP) threshold following a depolarising current steps protocol when voltage-clamped at - 80mV (t20.49 = 2.46, p = 0.023, Fig. 4A). No changes were seen for average or instantaneous frequency or interspike interval (Fig. 4B- D). Analysis of the first AP showed changes in the AP profile with increased rise Qme (t37.34 = 3.21, p = 0.003, Fig. 4E, 4H), and decreased amplitude (t31.31 = 2.72, p = 0.011, Fig. 4F, 4H) in LPS MSNs. Furthermore, LPS MSNs showed a significantly more depolarised a^erhyperpolarizaQon (AHP) peak (t28.37 = 3.40, p = 0.002, Fig. 4G, 4I). No changes were seen in rheobase, latency to first spike, half-width or AHP posiQon (data not shown). Together, these firing pauerns suggest that LPS-affected cells were less likely to be acQvated overall. Taken together these data suggest that LPS produces changes in pDMS circuits that disrupts goal-direct acQon control; given that goal-directed control requires a precise mix of excitatory and inhibitory responding. For example, performance on an outcome devaluaQon .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 15 test requires animals to respond selecQvely on the valued lever whilst inhibiQng any previously-learned tendency to press the devalued lever. Figure 4. Electrophysiological changes to medium spiny neuron (MSN) ac2on poten2al (AP) profile and discharge characteris2cs in pDMS with neuroinflamma2on or following chemogene2c ac2va2on of the Gi -pathway in astrocytes. (A-I) Results of whole-cell patch clamp electrophysiology recordings from MSNs following LPS or sham injecQons into the pDMS. (A-D) Individual data points showing acQon potenQal (AP) threshold for each MSN voltage clamped at -80mV, (B) average frequency of AP discharge, (C) instantaneous frequency, or (D) interspike interval for each MSN. (E-G) Individual data points for changes to AP profile including (E) AP rise Qme, (F) AP amplitude, and (G) an a^erhyperpolarisaQon .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 16 (AHP) peak for each MSN. (H) Example cell average trace show the AP profile characterisQcs of rise Qme and amplitude (LPS = green, saline = grey). (I-L) Results of whole-cell patch clamp electrophysiology recordings from MSNs following the applicaQon of arQficial cerebrospinal fluid (ACSF) then designer receptors exclusively acQvated by designer drugs (DREADD) agonist deschloroclozapine (DCZ) to astrocytes transfected with hM4Di DREADDs. (I-K) Individual data points showing (I) resQng membrane potenQal (RMP), (J) AP threshold, and (K) rheobase for each MSN. (L) Example cell rheobase traces (ASCF = grey, DCZ = orange). LPS vs saline; LPS at RMP n = 33 cells and at -80 voltage clamp n = 32 cells, from n = 4 animals; saline; n = 15 cells from n = 3 animals. GFAP-HM4Di n = 7 cells from n = 2 animals tested with ACSF then DCZ. We next turned our auenQon to astrocytes. Consistent with increased GFAP and c- Fos/Neun expression (Fig. 3) , a previous study that chemogeneQcally acQvated hM3Dq designer receptors exclusively acQvated by designer drugs (DREADDs) (35) on astrocytes in DMS found that this led to excessive goal-directed control (albeit in mice and using slightly different behavioural procedures). Thus, we sought to extend these findings by employing a procedure that would reveal informaQon about how pDMS astrocytes might contribute to goal-directed control when in their homeostaQc form. Based on evidence that Gi g-protein- coupled receptors (GPCR)s are highly expressed on striatal astrocytes (36), and that acQvaQng these receptors has been shown to ‘correct’ a number of HunQngton-like (37) and compulsion-like (38) deficits in mice, we employed astrocyte-specific hM4Di DREADDs to invesQgate the consequences of acQvaQng the Gi pathway on neuronal firing and on acQon selecQon. Although prior studies have invesQgated the acQvaQon of astrocyQc Gi-GPCRs in the striatum (36, 37), they have primarily focused on the dorsolateral striatum. Our brief fiber photometry study highlighted the importance of regional specificity to astrocyte funcQon within the striatum (39), because it showed that the acQvaQon of astrocyQc hM4Di- DREADDs in the pDMS produced a different profile of calcium transient acQvity than it did in the NAc core, as measured by relaQve increase in fluorescence. Specifically, in pDMS, i.p .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 17 injecQons of the DREADD agonist deschloroclozapine (DCZ) increased the number but not amplitude of calcium peaks (Supp. Fig. 4A), whereas it increased the amplitude of these peaks but not their frequency in the NAc core (Supp. Fig. 4B). Thus, to establish the effects of astrocyQc hM4Di acQvaQon on neuronal firing properQes, we again employed in vitro whole-cell patch clamp electrophysiology. A^er bilateral injecQon of GFAP-hM4Di-DREADD into the pDMS, with recordings for each cell (n=7 cells from n=2 animals) taken firstly in arQficial cerebrospinal fluid (ACSF) and then repeated following bath applicaQon of hM4DI-DREADD agonist DCZ (1µM). Following the applicaQon of DCZ, RMP was significantly more depolarised (t6.00 = 4.14, p = 0.0118, Fig. 4J) shi^ing cells closer to AP threshold. Then, following the same depolarising current steps protocol used in LPS electrophysiology experiments, AP threshold was lower at RMP (t6.00 = 3.57, p = 0.012, Fig. 4K) further narrowing the range between RMP and AP threshold. Rheobase was also significantly reduced following DCZ applicaQon (t6.00 = 4.86, p = 0.003 Fig. 4L). No other changes were seen in AP profile (Supp. Fig. 4E-L) or firing properQes (data not shown) with DCZ applicaQon at either RMP or while cells were voltage-clamped at -80mV. This profile of MSN firing is in direct contrast to that produced by LPS injecQon, and also contrasts with the effects of hM3Dq-transfected astrocytes on MSNs observed by Kang et al., (35), which reduced EPSPs and IPSPs in both direct and indirect pathway MSNs respecQvely. Thus, given that both LPS and hM3Dq acQvaQon on astrocytes facilitated goal- directed acQon control whilst producing an opposite profile of neuronal firing, we hypothesised that the acQvaQon of Gi receptors on pDMS astrocytes would have the opposite effect and abolish it. Although this predicQon may seem counterintuiQve, given prior findings that lesioning or otherwise inacQvaQng this structure also abolishes acQon .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 18 control (12, 21), recent findings paint a more nuanced picture of the precise condiQons necessary for goal-directed acQons (40, 41). In parQcular, these studies have suggested that spaQal-based neuronal ensembles within the striatum must behave in a precise and complementary manner (so-called “behavioural syllables”) to produce accurate acQon selecQon (36). Our electrophysiology results suggest that acQvaQng the Gi pathway in pDMS astrocytes disrupts this precision, an effect that would be expected to disrupt the behavioural selecQvity necessary for goal-directed acQon control. Behavioural experiments employing chemogeneQcs were conducted under moderate (standard) deprivaQon condiQons. Figure 5A and the bouom le^ panel of Figure 5B show the representaQve placements of AAV transfecQon in the pDMS. The bouom panel of Figure 5C shows a high degree of co-localisaQon of GFAP and AAV-hM4Di-GFAP-mCherry, and the top panel shows lack of overlap with NeuN, suggesQng specificity of transfecQon for astrocytes. Pavlovian and instrumental training were conducted in the absence of DCZ administraQons and proceeded without incident (Supp. Fig. 5A-C, all Fs < 1). Animals did, however, receive vehicle or DCZ injecQons 25-30 mins prior to each test. AcQvaQon of the astrocyQc Gi pathway in pDMS prevented transfer, which was impaired (Same = Different) for animals that received both the acQve virus and DCZ (hM4Di+DCZ; n=12) but was intact (Same > Different) for both vehicle (mCherry or hM4Di+Veh; n=8) and DCZ-only (mCherry+DCZ; n=11) controls (Fig. 5D). There was a group x transfer effect, F (1,28) = 4.947, p = 0.034, consisQng of intact transfer simple effects for groups mCherry+DCZ, F (1,28) = 5.995, p = 0.021, and hM4Di+Veh, F (1,28) = 11.731, p = 0.002, but not group hM4Di + DCZ, F Devalued) for controls, and abolished (Valued = Devalued) for group hM4Di+DCZ (Fig. 5D). The group x devaluaQon interacQon, F(1,28) = 5.494, p = 0.026, consisted of intact devaluaQon simple effects for groups hM4Di + Veh, .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 19 F(1,28) = 16.464, p = 0.000, and a marginal simple effect for mCherry + DCZ, F(1,28) = 4.063, p = 0.054, and no effect for group hM4Di + DCZ F nonReinstated) F(1,28) = 67.965, p = 0.000, interacQon F < 1 (Fig. 5E). This lauer finding demonstrates that the acQvaQon of the Gi pathway in pDMS astrocytes doesn’t simply replicate the behavioural results observed following a pDMS lesion or inacQvaQon, which have been shown to abolish reinstatement (12) as well as devaluaQon (12, 21) and transfer (42). Rather, these findings demonstrate a disQnct role for astrocytes in moderaQng the neuronal acQvity that is necessary for intact goal-directed control. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 20 Figure 5. Chemogene2c ac2va2on of the Gi -pathway in pDMS astrocytes abolished goal- directed ac2on control. (A) DiagrammaQc representaQon of the distribuQon and locaQons of the viral expressions in the pDMS included in the analysis. (B) Histological verificaQon of the GFAP virus expression in pDMS. (C) RepresentaQve images showing lack of colocalizaQon with NeuN (top panel) and colocalizaQon of mCherry from GFAP-hM4D-Gi-DREADD virus with the GFAP (bouom panel) scale bars = 45 µm. (D-F) Individual data plots and mean lever presses during the (D) Pavlovian-instrumental transfer test, (E) outcome devaluaQon test, and (F) outcome-selecQve reinstatement test. * denotes that the p < 0.05, # denotes p < .055. (n = 8 (M4/mCherry + VEH), n = 11 (mCherry + DCZ), n = 12 (M4 GFAP + DCZ), N = 31) Altogether, the results presented here show that pDMS neuroinflammaQon in the posterior dorsomedial striatum is sufficient to cause excessive goal-directed control, and that it does so because it disrupts the homeostaQc funcQon of astrocytes. If translatable, this implies that individuals with neuroinflammaQon in this brain region, including individuals with compulsive disorders, Parkinson’s disease, and HunQngton’s disease, might struggle to exert adequate control over their acQons due to an overreliance on goal-directed control. The specificity of this effect to the pDMS does, however, raise the possibility that individuals with more neuroinflammaQon in their ventromedial striatum (specifically nucleus accumbens core) might struggle to exert control over their acQons due to enhanced responding to Pavlovian cues, rather than effects on goal-directed control per se. It is also possible that individuals may have neuroinflammaQon in both regions, leading to either type of aberrant acQon control depending on the relaQve excitaQon of each based on the pauern of corQcal/thalamic inputs. Overall, then, findings indicate that the alteraQons to acQon control experienced by individuals with compulsive disorders cannot be reduced to a single mechanism (43), but rather are mulQfactorial (44) and likely operaQng differently between individuals, or possibly even within the same individual at different Qmes. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 21

We thank Tom Burton (UNSW) for his help analysing the fibre photometry data. We thank the technical staff at the Garvan Biological TesQng Facility at the Garvan InsQtute of Medical Research and the technical staff at the Ernst Facility at the University of Technology Sydney, and the Bioresearch Facility Staff at the University of Newcastle for technical support. Figures 1A and 2A were created with BioRender.com. Funding: This work was supported by the Australian Research Council (ARC) discovery project DP200102445 awarded to L.A.B, and the NaQonal Health and Medical Research Council grants GNT2003346 awarded to L.A.B, GNT2028533 awarded to L.A.B. and K.T., GNT1147207 awarded to B.W.B. and C.V.D., and GNT2020768 awarded to C.V.D. and E.M. Author contribu2ons: A.R.A., J.G., S.B., J.A.I., H.R.D., E.E.M., C.V.D., and L.A.B., conceptualised and designed the research, A.R.A., J.G., J.A.I., H.R.D., A.D., K.G., K.T., and K.C., performed the research (i.e. data curaQon), A.R.A., J.G., J.A.I., H.R.D., E.E.M., C.V.D., K.T., M.D.K., C.N., L.C., analysed the data, L.A.B., K.T., B.W.B., E.E.M., and C.V.D., acquired the funding, A.R.A., J.G., J.A.I., H.R.D., E.E.M., C.V.D., M.D.K., B.W.B., A.C., and L.A.B. wrote the paper (original dra^ – A.R.A., J.G., and L.A.B., review and ediQng - A.R.A., J.G., J.A.I., H.R.D., E.E.M., C.V.D., M.D.K., B.W.B., A.C., and L.A.B.). Compe2ng Interests: None Data and materials availability: DOI 10.17605/OSF.IO/297ZU License informa2on: N/A .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 22

1. C. M. Gillan, T. W. Robbins, B. J. Sahakian, O. A. Van Den Heuvel, G. Van Wingen, The role of habit in compulsivity. European Neuropsychopharmacology 26, 828–840 (2016). 2. B. J. Everiu, T. W. Robbins, Neural systems of reinforcement for drug addicQon: from acQons to habits to compulsion. Nat Neurosci 8, 1481–1489 (2005). 3. L. Hogarth, AddicQon is driven by excessive goal-directed drug choice under negaQve affect: translaQonal criQque of habit and compulsion theory. Neuropsychopharmacol. 45, 720–735 (2020). 4. S. C. Piantadosi, S. E. Ahmari, Using OptogeneQcs to Dissect the Neural Circuits Underlying OCD and Related Disorders. Curr Treat OpXons Psychiatry 2, 297–311 (2015). 5. M. Kohno, J. Link, L. E. Dennis, H. McCready, M. Huckans, W. F. Hoffman, J. M. Lo^is, NeuroinflammaQon in addicQon: A review of neuroimaging studies and potenQal immunotherapies. Pharmacology Biochemistry and Behavior 179, 34–42 (2019). 6. S. Auwells, E. SeQawan, A. A. Wilson, P . M. Rusjan, R. Mizrahi, L. Miler, C. Xu, M. A. Richter, A. Kahn, S. J. Kish, S. Houle, L. Ravindran, J. H. Meyer, InflammaQon in the Neurocircuitry of Obsessive-Compulsive Disorder. JAMA Psychiatry 74, 833–840 (2017). 7. M. Alegret, Obsessive-compulsive symptoms in Parkinson’s disease. Journal of Neurology, Neurosurgery & Psychiatry 70, 394–396 (2001). 8. K. Dewhurst, J. Oliver, K. L. K. Trick, A. L. McKnight, Neuro-psychiatric Aspects of HunQngton’s Disease. Stereotact Funct Neurosurg 31, 258–268 (1969). 9. B. W. Balleine, The Meaning of Behavior: DiscriminaQng Reflex and VoliQon in the Brain. Neuron 104, 47–62 (2019). 10. B. W. Balleine, J. P . O’Doherty, Human and Rodent Homologies in AcQon Control: CorQcostriatal Determinants of Goal-Directed and Habitual AcQon. Neuropsychopharmacol 35, 48–69 (2010). 11. L. A. Bradfield, B. W. Balleine, “The Learning and MoQvaQonal Processes Controlling Goal-Directed AcQon and Their Neural Bases” in Decision Neuroscience (Elsevier, 2017; hups://linkinghub.elsevier.com/retrieve/pii/B9780128053089000063), pp. 71–80. 12. H. H. Yin, S. B. Ostlund, B. J. Knowlton, B. W. Balleine, The role of the dorsomedial striatum in instrumental condiQoning. Eur J Neurosci 22, 513–523 (2005). 13. H. H. Yin, B. J. Knowlton, B. W. Balleine, Lesions of dorsolateral striatum preserve outcome expectancy but disrupt habit formaQon in instrumental learning. Eur J Neurosci 19, 181–189 (2004). .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 23 14. T. V. Maia, R. E. Cooney, B. S. Peterson, The neural bases of obsessive-compulsive disorder in children and adults. Dev Psychopathol 20, 1251–1283 (2008). 15. V. Pando-Naude, S. Toxto, S. Fernandez-Lozano, C. E. Parsons, S. Alcauter, E. A. Garza- Villarreal, Gray and white mauer morphology in substance use disorders: a neuroimaging systemaQc review and meta-analysis. Transl Psychiatry 11, 29 (2021). 16. S. M. De La Monte, J. J. Kril, Human alcohol-related neuropathology. Acta Neuropathol 127, 71–90 (2014). 17. P . Mews, A. M. Cunningham, J. Scarpa, A. Ramakrishnan, E. M. Hicks, S. Bolnick, S. Garamszegi, L. Shen, D. C. Mash, E. J. Nestler, Convergent abnormaliQes in striatal gene networks in human cocaine use disorder and mouse cocaine administraQon models. Sci. Adv. 9, eadd8946 (2023). 18. J. L. Cadet, V. Bisagno, C. M. Milroy, Neuropathology of substance use disorders. Acta Neuropathol 127, 91–107 (2014). 19. S. Saxena, S. L. Rauch, FUNCTIONAL NEUROIMAGING AND THE NEUROANATOMY OF OBSESSIVE-COMPULSIVE DISORDER. Psychiatric Clinics of North America 23, 563–586 (2000). 20. S. C. Piantadosi, E. E. Manning, B. L. Chamberlain, J. Hyde, Z. LaPalombara, N. M. Bannon, J. L. Pierson, V. M. K Namboodiri, S. E. Ahmari, HyperacQvity of indirect pathway-projecQng spiny projecQon neurons promotes compulsive behavior. Nat Commun 15, 4434 (2024). 21. H. H. Yin, B. J. Knowlton, B. W. Balleine, Blockade of NMDA receptors in the dorsomedial striatum prevents acQon-outcome learning in instrumental condiQoning. Eur J Neurosci 22, 505–512 (2005). 22. K. Ganesan, S. Ghorbanpour, W. Kendall, S. T. Broome, J. M. Gladding, A. Dhungana, A. R. Abiero, M. Mahmoudi, A. Castorina, M. D. Kendig, S. Becchi, V. Valova, L. Cole, L. A. Bradfield, Hippocampal neuroinflammaQon causes sex-specific disrupQons in acQon selecQon, food approach memories, and neuronal excitaQon. [Preprint] (2024). hups://doi.org/10.1101/2024.05.19.594460. 23. B. W. Balleine, A. Dickinson, Goal-directed instrumental acQon: conQngency and incenQve learning and their corQcal substrates. Neuropharmacology 37, 407–419 (1998). 24. S. B. Ostlund, B. W. Balleine, SelecQve reinstatement of instrumental performance depends on the discriminaQve sQmulus properQes of the mediaQng outcome. Animal Learning & Behavior 35, 43–52 (2007). 25. A. R. Abiero, Z. Ali, B. Vissel, L. A. Bradfield, Outcome-selecQve reinstatement is predominantly context-independent, and associated with c-Fos acQvaQon in the posterior dorsomedial striatum. Neurobiol Learn Mem 187, 107556 (2022). .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 24 26. L. H. Corbit, J. L. Muir, B. W. Balleine, The Role of the Nucleus Accumbens in Instrumental CondiQoning: Evidence of a FuncQonal DissociaQon between Accumbens Core and Shell. J. Neurosci. 21, 3251–3260 (2001). 27. G. Hart, L. A. Bradfield, S. Y . Fok, B. Chieng, B. W. Balleine, The Bilateral Prefronto- striatal Pathway Is Necessary for Learning New Goal-Directed AcQons. Curr Biol 28, 2218-2229.e7 (2018). 28. T. Robinson, The neural basis of drug craving: An incenQve-sensiQzaQon theory of addicQon. Brain Research Reviews 18, 247–291 (1993). 29. Bradfield, Laura A, Morris, Richard, B. W. Balleine, “Obsessive-compulsive disorder as a failure to integrate goal-directed and habitual acQon control” in Obsessive-Compulsive Disorder: Phenomenology, Pathophysiology, and Treatment (Oxford University Press, 2017), p. 343. 30. C. D. Adams, A. Dickinson, Instrumental Responding following Reinforcer DevaluaQon. The Quarterly Journal of Experimental Psychology SecXon B 33, 109–121 (1981). 31. N. W. Lingawi, B. W. Balleine, Amygdala Central Nucleus Interacts with Dorsolateral Striatum to Regulate the AcquisiQon of Habits. J. Neurosci. 32, 1073–1081 (2012). 32. A. Dickinson, B. Balleine, MoQvaQonal control of goal-directed acQon. Animal Learning & Behavior 22, 1–18 (1994). 33. P . Redgrave, M. Rodriguez, Y . Smith, M. C. Rodriguez-Oroz, S. Lehericy, H. Bergman, Y . Agid, M. R. DeLong, J. A. Obeso, Goal-directed and habitual control in the basal ganglia: implicaQons for Parkinson’s disease. Nat Rev Neurosci 11, 760–772 (2010). 34. T. S. Gertler, C. S. Chan, D. J. Surmeier, Dichotomous anatomical properQes of adult striatal medium spiny neurons. J Neurosci 28, 10814–10824 (2008). 35. S. Kang, S.-I. Hong, J. Lee, L. Peyton, M. Baker, S. Choi, H. Kim, S.-Y . Chang, D.-S. Choi, AcQvaQon of Astrocytes in the Dorsomedial Striatum Facilitates TransiQon From Habitual to Goal-Directed Reward-Seeking Behavior. Biol Psychiatry 88, 797–808 (2020). 36. B. S. Khakh, Astrocyte-Neuron InteracQons in the Striatum: Insights on IdenQty, Form, and FuncQon. Trends Neurosci 42, 617–630 (2019). 37. X. Yu, J. Nagai, M. MarQ-Solano, J. S. Soto, G. Coppola, M. M. Babu, B. S. Khakh, Context-Specific Striatal Astrocyte Molecular Responses Are Phenotypically Exploitable. Neuron 108, 1146-1162.e10 (2020). 38. J. S. Soto, C. Neupane, M. Kaur, V. Pandey, J. A. Wohlschlegel, B. S. Khakh, Astrocyte Gi- GPCR signaling corrects compulsive-like grooming and anxiety-related behaviors in Sapap3 knockout mice. Neuron, S0896627324005415 (2024). .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 25 39. H. Chai, B. Diaz-Castro, E. Shigetomi, E. Monte, J. C. Octeau, X. Yu, W. Cohn, P . S. Rajendran, T. M. Vondriska, J. P . Whitelegge, G. Coppola, B. S. Khakh, Neural Circuit- Specialized Astrocytes: Transcriptomic, Proteomic, Morphological, and FuncQonal Evidence. Neuron 95, 531-549.e9 (2017). 40. J. Peak, B. Chieng, G. Hart, B. W. Balleine, Striatal direct and indirect pathway neurons differenQally control the encoding and updaQng of goal-directed learning. eLife 9, e58544 (2020). 41. M. Matamales, A. E. McGovern, J. D. Mi, S. B. Mazzone, B. W. Balleine, J. Bertran- Gonzalez, Local D2- to D1-neuron transmodulaQon updates goal-directed learning in the striatum. Science 367, 549–555 (2020). 42. L. H. Corbit, P . H. Janak, Posterior dorsomedial striatum is criQcal for both selecQve instrumental and Pavlovian reward learning. Eur J Neurosci 31, 1312–1321 (2010). 43. Y . Vandaele, S. H. Ahmed, Habit, choice, and addicQon. Neuropsychopharmacol. 46, 689–698 (2021). 44. G. P . McNally, P . Jean-Richard-dit-Bressel, E. Z. Millan, A. J. Lawrence, Pathways to the persistence of drug use despite its adverse consequences. Mol Psychiatry 28, 2228– 2237 (2023). 45. Hays, W. L., StaXsXcs for the Social Sciences. (New York, Holt, Rinehart, & Winston, 1973). .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 26 SUPPLEMENTARY MATERIAL

Experiment 1 and 2: Effects of neuroinflamma7on in pDMS and NAc core on goal -directed ac7on control Animals and housing condiXons For behavioural experiments, a total of 158 Long-Evans rats [34 for Experiment 1 (15 male and 19 female), 31 for Experiment 2 (15 male, 16 female), 54 for Experiment 3 (27 male and 27 female), and 42 for Experiment 4 (20 male and 22 female)], weighing 180 –350 g, 8 -10 weeks of age at the beginning of the experiment were purchased from the Australian Research Centre, Perth, Australia, and were housed in groups of 2-3 in transparent amber plasQc boxes located in a temperature- and humidity-controlled room with a 12-h light/dark (07:00–19:00 h light) schedule. Experiments were conducted during the light cycle. Before the experiments, all animals were habituated to the laboratory se‚ngs for a week with full access to food and water and environmental enrichment which include plasQc tunnel, shreds of paper, and wooden object to gnaw. Throughout the training and actual experiment, ani mals were maintained at ~85% of their free-feeding body weight by restricQng their food intake to 8-14g of their maintenance diet per day. All procedures were approved by the Ethics Commiuees of the Garvan InsQtute of Medical Research Sydney (AEC 18.34) , and Faculty of Science, University of Technology Sydney (ETH21-6657), and the University of Newcastle (A-2020-018). Surgery .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 27 Animals were anaestheQzed with isoflurane (5% inducQon, 2 –3% maintenance) and posiQoned in a stereotaxic frame (Kopf Instruments). An incision was made into the scalp to expose the skull surface and the incisor bar was adjusted to align bregma and lambda on the same horizontal plane. Small holes were drilled into the skull above the appropriate targeted region and animals received bilateral injecQons by infusing 1 µl per hemisphere of LPS (5ug/ µl) via a 1-µl glass syringe (Hamilton Company) connected to an infusion pump (Pump 11 Elite Nanomite, Harvard Apparatus) into the pDMS (anteroposterior, −0.2mm; mediolateral, ±2.4mm (male), ±2.3mm (female); and dorsoventral, −4.5mm, relaQve to bregma) and another cohort of animals received LPS injected into their NAc core (anteroposterior, 1.4mm; mediolateral, ±2.2mm; and dorsoventral, −7.5mm, relaQve to bregma) . The infusion was conducted at a rate of 0.15 µl/min, and injectors were le^ in place for an addiQonal 5 min to ensure adequate diffusion and to minimize LPS spread along the injector tract. The remaining control animals underwent idenQcal procedures but with injecQon of sterile saline rather than LPS. A nonsteroidal anQ-inflammatory/anQbioQc agent were administered preoperaQvely and postoperaQvely to minimize pain and discomfort. Animals were allowed to recover for 7 days before the onset of any behavioural training. Apparatus All behavioural procedures took place in twelve idenQcal sound auenuaQng operant chambers (Med Associates, Inc.,) and these chambers were located within individual cubicles. The ceiling, back wall, and hinged front door of the operant chambers were made of a clear Plexiglas and the side wall were made of grey aluminium. The floor was made of stainless steel grids. Each chamber was equipped with a recessed food magazine, located at the base of one end wall, through which 20% sucrose-10% polycose soluQon (0.2 ml) and food pellets (45 mg; .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 28 Bio-Serve, Frenchtown, NJ) could be delivered using a syringe pump and pellet dispenser, into separate compartments respecQvely. Two retractable levers could be inserted individually on the le^ and right sides of the magazine. An infrared light situated at the magazine opening was used to detect head entries. IlluminaQon was provided by a 3-W, 24-V houselight situated at the top-centred on the le^ end wall opposite the magazine provided constant illuminaQon, and an electric fan fixed in the shell enclosure provided background noise (≈70 dB) throughout training and tesQng. The apparatus was controlled, and the data were recorded using Med-PC IV computer so^ware (Med Associates, Inc.). The boxes also contained a white -noise generator, a sonalert that delivered a 3 kHz tone, and a solanoid that, when acQvated, delivered a 5 Hz clicker sQmulus. All sQmuli were adjusted to 80 dB in the presence of

noise of 60 dB provided by a venQlaQon fan. Outcome devaluaQon procedures took place in transparent plasQc tubs that were smaller, but otherwise idenQcal to the cages in which rats were housed. Food restricXon and Chow maintenance One week following recovery from surgery, animals underwent 3 days of food restricQon before the onset of lever press training. During this Qme animals received 10-14g of chow per day, and their weight was monitored daily to ensure it remained at ~85% of their pre-surgery body weight. For the iniQal Pavlovian and instrumental training, as well as the first round of tesQng for sPIT, devaluaQon, and reinstatement, the chow that rats were maintained on the higher-fat, higher -protein Gordons Specialty Feed (see Table 1). Following reinst atement tesQng, animals were switched to a lower fat, lower protein Irradiated Specialty Feed’s chow (see Table 1) and re-trained and re-tested. Pavlovian training .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 29 For the first 8 days, animals were placed in operant chambers for 60 min during which they received eight 2 min presentaQons of two condiQoned sQmuli (CS; white noise or clicker) paired with one of two outcomes (sucrose soluQon or pellet) presented on a random Qme schedule around an average of 30 s throughout each CS presentaQon. Each CS was presented 4 Qmes, with a variable intertrial interval (ITI) that averaged to 5 min. For half the subjects, tone was paired with sucrose and noise with pellets, with the other half receiving the opposite arrangement. Magazine entries throughout the session were recorded and reported for the 2 min prior to each CS presentaQon (PreCS) and the 2 min during each CS presentaQon. Lever press training Following Pavlovian training, animals were trained to press a le^ and right lever over 8 days which earned the same sucrose and grain pellet outcomes. Specifically, for half of the animals, the le^ lever earned pellets and the right lever earned sucrose, and the other half received the opposite arrangement (counterbalanced). Each session lasted for 50 minutes and consisted of two 10 minutes sessions on each lever (i.e., four x 10 minutes sessions in total) separated by a 2.5 minutes Qme -out period in whi ch the levers were retracted and the houselight was switched off. Animals could earn a maximum of 40 sucrose and 40 pellets deliveries within the session. For the first 2 days, animals were trained on a conQnuous reinforcement schedule (CRF) in which each lever press produced a single outcome. Animals were then shi^ed to a random raQo-5 schedule for the next 3 days (i.e. each acQon delivered an outcome with a probability of 0.2), then to a RR -10 schedule (or a probability of 0.1) for the final 3 days. A ^er 40 sucrose soluQons and 40 pellets were delivered or 50 minutes had elapsed, whichever came first, the session was terminated, levers were retracted, and house lights switched off. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 30 Pavlovian Instrumental Transfer (Specific PIT) test One day a^er the end of instrumental training, rats were tested for sPIT performance. For this test, responding on both levers was first exQnguished for 8 min to reduce baseline performance. Subsequently, each CS was presented four Qmes over the next 40 min in the following order: clicker -noise-noise-clicker-noise-clicker-clicker-noise. Each CS lasted 2 min and had a fixed ITI of 3 min. Magazine entries and lever pressing rates were recorded throughout the session and responses were separated into PreCS and CS periods (2 min each). Lever presses were recorded, but not reinforced. Outcome DevaluaXon One day a^er sPIT tesQng, rats were given 1 day of instrumental retraining on RR-10 in the manner previously described. On the following day, animals were given free access to either the pellets (20 g placed in a bowl) or the sucrose soluQon (100 ml in a drinking boule) for 1 hr. The amount of pellets and sucrose soluQon consumed each day was measured. Animals were then placed in the operant chamber for a 10 min choice exQncQon test. During this test, both levers were extended and lever presses recorded, but no outcomes were delivered. The next day, a second devaluaQon test was administered with the o pposite outcome (i.e. if animals were prefed on pellets the previous day they were now prefed on sucrose, and vice versa). Following pre-freeding animals were again placed into the operant chambers for a second 10 min choice exQncQon test. All test results are reported as averaged across these two tests. Outcome SelecXve Reinstatement Test .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 31 A^er devaluaQon tesQng, rats received one day of instrumental retraining on an RR -10 schedule for 1 day. The next day, animals were tested for outcome-selecQve reinstatement in which rats received a 15 min period of exQncQon to reduce baseline performance. They then received four reinstatement trials separated by 4 min each as before, and each reinstatement trial consisted of a single free delivery of either the sucrose soluQon or the grain pellet presented in the following order: sucrose, pellet, pe llet, and sucrose. Responding was measured during the 2 min periods immediately before (pre) and a^er (post) each delivery. Switching maintenance chow, re-training and re-tesXng As noted, I iniQally used a highly palatable home chow (high-fat/high-protein lab chow) for Experiment 2 which reduced performance in Sham controls. Following training and tesQng during which animals were given this chow, I switched to a smaller amount (6 -8g) of less palatable home chow (lower-fat/lower-protein lab chow) to increase hunger and moQvaQon to lever press for food, with the aim of improving test performance in group Sham. Then I sought to answer whether pDMS neuroinflammaQon sQll facilitated goal-directed acQon when Sham performance was improved. Following the switch from Gordon’s to Specialty feeds chow, rats were given an addiQonal 4 days of Pavlovian training, and an addiQonal 4 days of intrumental training, then tested for performance on sPIT, outcome devaluaQon, and outcome-selecQve reinstatement as before. Tissue preparaXon One day a^er the outcome-selecQve reinstatement test, animals were sacrificed via CO2 inhalaQon and perfused transcardially with cold 4% paraformaldehyde in 0.1 M phosphate buffer saline (PBS; pH 7.3-7.5). Brains were rapidly and carefully removed and pos†ixed in 4% .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 32 paraformaldehyde overnight and then placed in 30% sucrose. Brains were secQoned coronally at 40 µm through the pDMS and NAc core defined by Paxinos and Watson (2014) using a cryostat (CM3050S, Leica Microsystems) maintained at approximately -20˚ Celsius. The secQoned slices were immediately immersed in cryoprotectant soluQon and stored in the - 20°C freezer. Later, five representaQve secQons from pDMS and NAc core were selected for each rat. SecQons were first washed three Qmes (10 minutes per wash) in PBS to remove any exogenous substances. The secQons were then incubated in a blocking soluQon comprising of 3% Bovine Serum Albumin (BSA) + 0.25% TritonX-100 in 1 x PBS for one hour to permeabilize Qssue and block any non -specific binding. SecQons were then incubated in anQ -GFAP mouse primary anQbody (1:300, Cell Signalling Technology Catalog #3670), anQ-IBA1 rabbit primary anQbody (1:500, FUJIFILM Wako Chemicals U.S.A. CorporaQon), and anQ -NeuN chicken primary anQbody (1:1000, GeneTex Catalog #GTX00837) diluted in blocking soluQon for 72 h at 4°C. SecQons were then washed 3 Qmes in 1 × PBS and incubated overnight at 4°C in goat anQ - mouse AlexaFluor-488 secondary anQbody (1:250, ThermoFisher Catalog #A-11001), donkey anQ-rabbit AlexaFluor-568 secondary anQbody (1:250, ThermoFisher Catalog #A10042), and goat anQ -chicken AlexaFluor-647 secondary anQbody (1:250, ThermoFisher Catalog #A - 21449), fo llowed by a counterstain with 4ʹ,6 -diamidino-2-phenylindole (DAPI; Thermo ScienQfic; 1:1000, diluted in 1x PBS). Finally, every secQon was mounted onto Superfrost microscope slides (Fisher ScienQfic) and were coverslipped (Menzel -Glaser) using the mounQng agent Vectashield and le^ to dry overnight in darkness. Separate brain secQons of the pDMS and NAc core were also processed and incubated in anQ-c-fos primary anQbody (1:500, SynapQc Systems Catalog #226 003) and anQ-NeuN chicken .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 33 primary anQbody (1:500, GeneTex Catalog #GTX00837) diluted in blocking soluQon for 72 h at 4°C to see how much acQvaQon during neuroinflammaQon. SecQons were then washed 3 Qmes in 1 × PBS and incubated overnight at 4°C in donkey anQ-rabbit AlexaFluor-568 secondary anQbody (1:500, ThermoFisher Catalog #A10042) and goat anQ-chicken AlexaFluor- 647 secondary anQbody (1:500, ThermoFisher Catalog #A-21449), followed by a counterstain with DAPI (Thermo ScienQfic; 1:1000, dilu ted in 1x PBS). SecQons we re mounted and quanQfied using the same procedures described above. Imaging and immunofluorescence analysis For quanQficaQon of GFAP , IBA1, NeuN, and c-fos, a single image was taken of the pDMS and NAc core per hemisphere of each slice (6-10 images in total per brain region of each rat) on a Nikon TiE2 microscope using a 10x objecQve and Leica STELLARIS 20x air objecQve for representaQve images. Microscopy: Images were quanQfied using imaging so^ware (ImageJ, Fiji Cell Counter), whereby each fluorescent channel was split to isolate and count the cells of interest. Z-stacks were used instead of simply a single image plane. Briefly, the image was adjusted to 8-bit and

subtracQon was applied to remove background noise. Thresholding was used to isolate posiQve stained cells and the threshold for contrast and brightness was adjusted for all images unQl consistent between images (maximum: 255, minimum: 0). Images were then converted to binary and finally, the Analyze ParQcles tool was used to quanQfy the number of cells based on a minimum parQcle size of 16. ImageJ counted each cell between our parameters and presented it as a “count.” This was followed by intensity measurement, which is represented as Mean grey value (MGV) and background intensity subtracted from the reported MGV. SpaQal co-occurrence was measured for c -fos and NeuN only to determine .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 34 levels of neuronal acQvaQon. Stacked images of each channel were converted into RGB colour images and made composite. The colour threshold was adjusted to select all the red signal (which was the co -occurrence of c-fos and NeuN), and the percentage area of the selected signal was measured and averaged for each brain using ImageJ. Data and StaXsXcal analysis Data were collected automaQcally by Med-PC and uploaded to Microso^ Excel using Med- PC to Excel so^ware. Pavlovian condiQoning and lever press acquisiQon data was analysed using two-way repeated measures ANOVAs controlling the per-family error rate at α=0.05. If condiQons for sphericity were not met, the Greenhouse -Geisser correcQon was applied. To allow for a more fine-grained analysis of test data, all data for sPIT, outcome devaluaQon, and outcome-selecQve reinstatement were analysed using compl ex orthogonal contrasts controlling the per-contrast error rate at α=0.05 according to the procedure described by Hays (45). AcquisiQon data were expressed as mean ± standard error of the mean (SEM) averaged across counterbalanced condiQons. Test data were expressed as individual data points with means. If interacQons were detected, follow -up simple effects analyses ( α=0.05) were calculated to determine the source of the interacQon. For immunohistochemical analysis, counts and intensity were compared between LPS and Sham groups using two tailed t -tests and correlated using GraphPad. Test behaviours were correlated with GFAP, IBA1, NeuN, and c-fos expression, as well as c-fos-NeuN intensity using the immunohistochemical results from Figure 3. For correlaQons with behaviour we used a “PIT score”, a “devaluaQon score”, or a “reinstatement score” that were calculated in such a way as to ensure that any associaQon detected was not driven by baseline differences in lever press responding per se, but rather by the animals selecQvity of responding for one or the other levers. For these scores, we first .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 35 calculated suppression raQo (SR) scores on each of the levers individually (i.e., the same and different levers for sPIT, the valued and devalued levers for devaluaQon, and the reinstated and non-reinstated levers for outcome selecQve reinstatement) according to EquaQon 1: 1) 𝑆𝑐𝑜𝑟𝑒 = !"#"$ &$"'' $()" *+ ,"') !"#"$ &$"'' $()" *+ ,"') - .('"/0+" !"#"$ &$"'' 1()" In this equaQon, “baseline lever press rate ” was taken as the average press rate on each lever across the last two days of training prior to test. We then calculated the PIT score by subtracQng the normalised scores on the different lever from the normalised scores on the same lever (i.e. Same – Different), such that a higher score indicated beuer sPIT performance. Likewise, for devaluaQon we subtracted the normalised scores on the devalued from those on the valued lever (i.e. Valued – Devalued), such that a higher score indicated beuer devaluaQon performance, and did the same thing for reinstatement, this Qme subtracQng scores on the nonreinstated from scores on the reinstated lever (i.e. Reinstated – NonReinstated) such that a high score indicated beuer reinstatement performance. Each of these scores were then separately correlated with GFAP , IBA1, c-fos-NeuN intensity (correlaQons with GFAP, IBA1, and c-fos count can be found in Supplemental Figure 3C-D. Values of p < 0.05 were considered staQsQcally significant. The staQsQcal so^ware GraphPad Prism, SPSS, and PSY were used to carry out these analyses. Experiment 3: Effects pDMS neuroinflamma7on on overtraining-induced habits Surgery All surgical procedures were conducted idenQcally to that described for Experiment 1. Food restricXon and Chow maintenance .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 36 For this experiment, animals received only 6-8g of the Irradiated Specialty Feeds chow per day to maintain high moQvaQon condiQons. They did not receive Gordon’s chow at any point. Apparatus All Apparatus were as described for Experiments 1 and 2. Magazine Training Following recovery from surgery to inject LPS or saline into the pDMS, animals received 3 days of food deprivaQon and were then given two sessions of magazine training. For these sessions, the house light was turned on at the start of the session and turn ed off when the session was terminated. No levers were extended. Sucrose soluQon was delivered at random 60 s intervals for 30 outcomes per session. The session terminated a^er 45 min or a^er 30 outcomes had been delivered, whichever came first. Lever Press Training Following magazine training, animals then received 8 days of instrumental training (two sessions per day) to press a single lever for sucrose soluQon delivery. Animals received three sessions of conQnuous reinforcement, four sessions of random interval of 15 s (RI -15), four sessions of RI -30, and four sessions of RI -60. Right and le^ lever assignment was counterbalanced across animals. Sessions ended, levers retracted and the houselight terminated when 30 reinforcements were earned or a^er 60 min, which ever came first. Progressive raXo test Following lever press training, animals underwent 2-h of progressive raQo (PR) tesQng each day for 3 days. A progressive raQo schedule requires the subject to perform an increasing number of lever presses for the next presentaQon of a reinforcer (Hodos, 1961). For the .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 37 current study, the PR was set at n+5. This meant that animals iniQally received a sucrose reward for a single lever press, then for 5 lever presses, then n+5 lever presses unQl breakpoint – with breakpoint defined as 5 min of no lever pressing. The number of responses required to obtain each successive delivery of the sucrose reward was collected automaQcally by Med-PC. Outcome devaluaXon The day a^er progressive raQo tesQng, animals were given 2 days of instrumental retraining on an RI-60 schedule in the manner previously described. The following day , the sucrose soluQon was devalued using condiQoned taste aversion method for half of the animals. That is, all animals were given ad libitum access to sucrose soluQon in clear plasQc tubs for 30 min each day for 3 days. Immediately a^er the 30 mins, half of each type of lesion group received an intraperitoneal injecQon of lithium chloride (0.15 M LiCl, 20 ml/kg) to induce illness which the rat will associate with the outcome, effecQvely devaluing it, a^er which they placed back in their home cages. The remaining rats received 0.9% purified saline injecQons (20 ml/kg) and these animals comprised the valued groups. In total this manipulaQon yielded 4 groups: Sham- Valued, Sham -Devalued, LPS -Valued, LPS -Devalued. The amount of sucrose soluQon consumed each day was measured. ExXncXon test The day following the last day of LiCl pairings, all animals received a 5 min exQncQon test. The test began with the inserQon of the same lever used during training and ended with the retracQon of the lever. Lever presses were recorded, and no sucrose reward was delivered. Tissue Processing and Fluorescent Microscopy .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 38 All Qssue processing and microscopy were conducted idenQcally to that described for Experiments 1 and 2. StaXsXcal analysis Lever press and magazine entry data were collected automaQcally by Med-PC (version 5) and uploaded directly to Microso^ Excel using Med -PC to Excel so^ware. Lever press acquisiQon and progressive raQo data were analysed using repeated measures (Group x Session) ANOVA controlling the per -family error rate at α=0.05. To allow for a more fine - grained analysis of test data, we used planned, complex orthogonal contrasts controlling the per-contrast error rate at α=0.05 for analyzing the outcome devaluaQon according to the procedure described by Hays (1973). The amount of sucrose consumed was analysed using Three-Way ANOVA repeated measures. If condiQons for sphericity were not met, the Greenhouse-Geisser correcQon was used. Data analysis was conducted in the manner described for Experiment 3. Patch-clamp electrophysiology Acute brain slice preparaXon To prepare brain slices animals were deeply anestheQsed using ketamine injecQon (100mg/kg, i.p.) and then rapidly decapitated. Following this, brains were rapidly extracted and immersed in ice -cold sucrose subsQtuted arQficial cerebrospinal fluid (ACSF) containing (in mM): 236 sucrose, 25 NaHCO3, 11 glucose, 2.5 KCL, 1 NaH2PO4, 1 MgCl2, and 2.5 CaCl2. Coronal slices (300µm) of the pDMS were made using a vibraQng microtome (VT1200s, Leica, Nussloch, Germany). Slices were then transferred to an incubaQon chamber containing .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 39 oxygenated ACSF (120mM NaCl subsQtuted for sucrose) and allowed to equilibrate for 1-hour at room temperature (22-24°C) prior to recording. Patch-clamp electrophysiology Slices were transferred to a recording chamber and conQnuously perfused at a rate of 4-6 bath volumes/min with ASCF constantly bubbled with Carbonox (95% O2, 5% CO2) to achieve a final pH of 7.3 -7.4. All recordings were obtained at room temperature (22 -24°C), with neurons visualized using near -infrared differenQal interference contrast opQcs (IR -DIC). Recordings were restricted to the pDMS in both LPS and hM4Di -DREADD studies, and taken using patch pipeues (4-8 MW, Harvard Glass) filled with a potassium gluconate based internal soluQon containing (in mM): 135 C6H11KO7, 8 NaCl, 2 Mg 2-ATP , 10 HEPES, 0.1 EGTA, and 0.3 Na3GTP , pH 7.3 (with KOH). Recordings were collected using a MulQclamp 700B amplifier (Molecular Devices, Sunnyvale CA). Signals were sampled at 20kHz, filtered at 10kHz and digiQsed using an InstraTECH ITC -18 A/D board (HEKA Instruments, Belmore, New York) , acquired using Axograph X so^ware (Axograph X, Sydney, Australia). PutaQve MSN cell selecQon was based on MSN cell morphology and post-hoc confirmaQon of MSN delayed firing AP profile, excluding cells without this profile from analysis. Once whole-cell recording was iniQated, series and input resistance were calculated based on the response to a -5 mV voltage step from a holding potenQal of -70 mV. These measurements were repeated throughout and at the end of all recordings, and data were rejected if this changed by >20 % for an individual cel l. AP discharge was invesQgated in current clamp mode, firstly at RMP , and subsequently voltage clamped at -80mV by injecQon of current when necessary. Depolarising current steps used to evoke AP discharge were .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 40 increased in 20pA increments for a duraQon of 1 second and features relaQng to AP profile were extracted from this data. For recordings using hM4Di -DREADD agonist DCZ (1mM) the above recordings were firstly taken in ACSF and then repeated following bath applicaQon of DCZ. Data were analysed offline using Axograph X and Igor Pro 9 (Wavemetrics, Portland, OR) so^ware. AP threshold was taken from rheobase response and AP characterisQcs were extracted from this including latency to rheobase AP rise Qme, AP amplitude, AP half-width, AHP peak and AHP posiQon. AP discharge properQes were then calculated from rheobase +20pA to determine frequency (mean and instantaneous) and interspike interval. StaXsXcs Data is presented as mean ± SEM. Unpaired t -tests with Welch’s correcQon were used to compare LPS and sham affected MSN populaQons. Paired t-tests were used when comparing unaffected and DCZ treated MSNs. Experiment 4: Chemogene7c ac7va7on of Gi-protein-coupled receptors in astrocytes ChemogeneXcs The DREADD agonist deschloroclozapine (DCZ) dihydrochloride (NIMH D-925) was acquired from NaQonal InsQtute of Mental Health (NIMH) through the NIMH Chemical Synthesis and Drug Supply Program. DCZ was diluted with normal saline (SAL) (0.9% w/v NaCl) to a final injectable concentraQon of 0.1 mg/kg (at a volume of 1ml/kg). DCZ was always handled in dim/low light condiQons (i.e. a single lamp in a darkened room) and freshly prepared on the morning of each test day. Surgery .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 41 All surgical procedures were conducted idenQcally to that described for Experiment 1, except that animals received bilateral injecQons of 1 µl per hemisphere of AAV-GFAP-hM4Di- mCherry (Addgene, item ID 50479-AAV5, Qter 7×10¹² vg/mL). The infusion was conducted at a rate of 0.2 µl/min, and injectors were le^ in place for an addiQonal 5 min to ensure adequate diffusion and to minimize DREADDs spread along the injector tract. The remaining control animals underwent idenQ cal procedures but with injecQon of AAV -GFAP104-mCherry (Addgene, item ID 58909-AAV5, Qter 1×10¹³ vg/mL) as control group. Apparatus and Behavioural Procedures All apparatus and behavioural procedures were conducted idenQcally to that described for Experiment 1, except for outcome devaluaQon (specific saQety) where animals were given free access to either the pellets or the sucrose soluQon for 45 mins instead of 1 hr, a^er which DCZ was administered intraperitoneally (i.p) and rats returned to their home cage for 25 -30 min prior to behavioural tesQng. Tissue Processing and Fluorescent Microscopy The extent of the expression was determined using the boundaries defined by Paxinos and Watson (2014). SecQons were then stained with Living Colors® DsRed Polyclonal AnQbody (1:500, Takara Bio USA, Inc. Catalog #632496) to recognize the mCherry DREADDs expression, anQ-GFAP mouse primary anQbody (1:300, Cell Signalling Technology Catalog #3670) to check the co-localizaQon, diluted in blocking soluQon for 72 h at 4°C. SecQons were then washed 3 Qmes in 1 × PBS and incubated overnight at 4°C in donkey anQ-rabbit AlexaFluor-568 secondary anQbody (1:500, ThermoFisher Catalog #A10042), goat anQ-mouse AlexaFluor-488 secondary anQbody (1:500, ThermoFisher Catalog #A11001), followed by a counterstain with .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 42 DAPI (Thermo ScienQfic; 1:1000, diluted in 1x PBS). SecQons were mounted and quanQfied using procedures idenQcal to those described above. StaXsXcal analysis All statsiQcal analysis was conducted idenQcally as described for Experiment 1. Tables Gordons Specialty Feed Irradiated Specialty Feed Protein 23% 19% Saturated Fat 21.3% 0.78% Mono-unsaturated Fat 42.9% 2.06% Poly-unsaturated Fat 30.7% 1.88% Crude Fibre 5% 5.20% Table 1. Nutri7onal informa7on of the lab chow used during the experiment .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 43 Table 2: Full results for firing proper2es of medium spiny neurons in posterior dorsomedial striatum injected with 1ul lipopolysaccharide (LPS, 5mg/mL) at res2ng membrane poten2al. Measurement LPS Saline Significant Difference Amplitude (mV) 67.37 71.09 ns (0.1989) Rise (ms) 0.8204 0.7286 * (0.0489) Width (ms) 2.424 2.309 ns (0.3305) AHP Peak (mV) -17.33 -15.48 ns (0.0577) AHP Position (ms) 16.85 17.33 ns (0.8014) AHP Threshold (mV) -35.79 -37.82 ns (0.1562) Rheobase (pA) 196.7 222.0 ns (0.4996) Latency Rheobase (ms) 608.3 621.5 ns (0.8741) Resting Membrane Potential (mV) -73.30 -72.81 ns (0.8514) Instantaneous Frequency (Hz) 6.052 4.760 * (0.0359) Average Frequency (Hz) 3.091 3.200 ns (0.8142) Interspike Interval (ms) 177.8 265.0 * (0.0300) .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 44 Table 3: Full results for firing proper2es of medium spiny neurons in posterior dorsomedial striatum injected with 1ul lipopolysaccharide (LPS, 5mg/mL) at -80mV. Measurement LPS Saline Significant Difference Amplitude (mV) 64.54 72.80 * (0.0105) Rise (ms) 0.8743 0.7086 ** (0.0027) Width (ms) 2.552 2.389 ns (0.2822) AHP Peak (mV) -17.24 -14.26 ** (0.0020) AHP Position (ms) 17.61 17.95 ns (0.8497) AHP Threshold (mV) -36.16 -40.67 * (0.0228) Rheobase (pA) 256.3 272.7 ns (0.6721) Latency Rheobase (ms) 524.1 661.4 ns (0.0778) Instantaneous Frequency (Hz) 6.340 5.422 ns (0.3581) Average Frequency (Hz) 2.545 3.067 ns (0.3126) Interspike Interval (ms) 238.6 230.1 ns (0.8289) .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 45 Measurement ACSF DREADDs Significant Difference Amplitude (mV) 75.39 75.30 ns (0.9809) Rise (ms) 0.7501 0.7640 ns (0.8687) Width (ms) 2.764 3.074 ns (0.3844) AHP Peak (mV) -14.95 -12.07 ns (0.0879) AHP Position (ms) 23.59 22.64 ns (0.5459) AP Threshold (mV) -40.39 -45.04 * 0.0118 Rheobase (pA) 185.7 125.7 ** 0.0028 Latency Rheobase (ms) 640.5 441.5 ns (0.2462) Resting Membrane Potential (mV) -76.19 -64.67 ** (0.0018) Instantaneous Frequency (Hz) 4.762 4.749 ns (0.9910) Average Frequency (Hz) 3.429 2.429 ns (0.1563) Interspike Interval (ms) 227.5 259.5 ns (0.5656) Table 4: Full results for firing proper2es of medium spiny neurons adjacent to hM4Di - expressing astrocytes in posterior dorsomedial striatum at res2ng membrane poten2al. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 46 Measurement ACSF DREADDs Significant Difference Amplitude (mV) 79.78 79.49 ns (0.9293) Rise (ms) 0.6602 0.6550 ns (0.8425) Width (ms) 2.552 2.937 ns (0.1000) AHP Peak (mV) -13.77 -11.19 ns (0.4693) AHP Position (ms) 21.98 19.39 ns (0.1974) AP Threshold (mV) -43.85 -48.04 ns (0.3269) Rheobase (pA) 203.3 220.0 ns (0.4859) Latency Rheobase (ms) 611.2 604.2 ns (0.9769) Instantaneous Frequency (Hz) 4.478 3.664 ns (0.3537) Average Frequency (Hz) 3.400 2.200 ns (0.0705) Interspike Interval (ms) 245.2 367.8 ns (0.4086) Table 5: Full results for firing proper2es of medium spiny neurons adjacent to hM4Di - expressing astrocytes in posterior dorsomedial striatum at -80mV. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 47 Supplemental Figures and Results Supplemental Figure 1, relates to Figure 1. Supplemental behaviour results. There were no significant Sham/LPS differences at any stage of acquisiXon for pDMS experiment (Experiment 1) . (A-C) AcquisiXon under mild deprivaXon condiXons, (A) Magazine entries per min (±SEM) during Pavlovian condiXoning , F (7,196) = 0.669, p = 0.698, for CS x group x session interacXon, (B) Lever presses per min (±SEM) during instrumental condiXoning, main effect of day F (7,196) = 53.28, p = 0.000, no main effect of group and no day x group interacXon, all Fs < 1,(C) Magazine entries per min (±SEM) during instrumental condiXoning, main effect of day F (7,196) = 15.282, p = 0.000, no main effect of group and no day x group interacXon, all Fs < 1, (D-E) AcquisiXon under moderate deprivaXon condiXons, (D) Magazine entries per min (±SEM) during Pavlovian condiXoning, F (3,84) = 2.15, p = 0.111, for CS x group x session interacXon, (E) Lever presses per min (±SEM) during instrumental condiXoning, main effect of day F (3,84) = 15.87, p = 0.000, no main effect of group and no day x group interacXon, all Fs < 1, (F) Magazine entries per min (±SEM) during instrumental .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 48 condiXoning, main effect of day F (3,84) = 2.865, p = 0.041, no main effect of group and no day x group interacXon, all Fs < 1. (G-J) Supplemental behavioural results from NA c core neuroinflammaXon study, there were no significant Sham/LPS differences for this experiment, (G) Lever presses per min (±SEM) during instrumental condiXoning, main effect of day F (7,182) = 27.711, p = 0.000, no main effect of group and no day x group interacXon, Fs < 1, (H) Magazine entries per min (±SEM) during instrumental condiXoning, main effect of day F (7,182) = 9.692, p = 0.000, no main effect of group and no day x group interacXon, Fs < 1, (I) Individual data points and mean magazine entries per min during Pavlovian instrumental transfer tesXng, a main effect of sPIT, F (1,26) = 14.349, p = .001, that did not interact with the group, F (1,26) = 1.118, p = 0.30. A significant simple effect for the Sham group (Same > Different), F (1,26) = 11.739, p = 0.002, but no such effect (or a marginal simple effect) for the LPS group (Same = Different), F (1,26) = 3.728, p = 0 .064, (J) Individual data points and mean magazine entries per min during outcome selecXve reinstatement tesXng, a main effect of reinstatement (Reinstated > Nonreinstated) F (1,26) = 19.278, p = 0.000, which did not interact with any group differences, all Fs < 1.* denotes p < 0.05. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 49 Supplemental Figure 2, relates to Figure 2. pDMS neuroinflamma2on did not alter magazine entries during lever press acquisi2on, nor sucrose devalua2on by condi2oned taste aversion. (A) Magazine entries per min (±SEM) during instrumental condiXoning , main effect of day F (14,546) = 51.27, p = 0.000, no Sham/LPS differenc e, F (1,39) = 0.002, p = 0.9629, and no day x group interacXon, F < 1, (B) Millilitres of sucrose consumed (±SEM) during condiXoned taste aversion training, session x devaluaXon interacXon, F (1.64,60.9) = 90.44, p = 0.000 that did not interact with group, F < 1. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 50 Supplemental Figure 3. Relates to Figure 3. Supplemental immunohistochemical results following injec2ons of lipopolysaccharide (LPS) into posterior dorsomedial striatal (pDMS) in Experiments 1&3. (A-B) Individual data points and mean values for quanXficaXon of, from lem to right, cell counts, mean gray value, circularity (lem y axis) and perimeter (right y axis) of GFAP , IBA1, and NeuN from rats in (A) Experiment 1 and (B) Experiment 3. For Experiment 1: cell counts were significantly higher in LPS Xssue relaXve to Sham for GFAP , t(28) = 6.255, p = 0.000, and IBA1, t(28) = 8.74, p = 0.000, but not for NeuN, t(28) = 1.9, p = .068, mean gray .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 51 value was likewise significantly higher in LPS Xssue relaXve to Shams for GFAP , t(28) = 4.046, p = 0.0004, IBA1, t(28) = 2.799, p = 0.0092, but not for NeuN, t(28) = 0.7616, p = 0.4527, circularity and perimeter of cells did not differ for any GFAP-posiXve or IBA1-posiXve cells in Experiment 1 , closest t(28) = 0.77, p = 0.443, for IBA1 perimeter. For Experiment 3: cell counts were significantly higher in LPS Xssue relaXve to Sham for GFAP , t(39) = 5.38, p = 0.000, and IBA1, t(39) = 10.27, p = 0.000, but not for NeuN, t(39) = 0.336, p = 0.074, mean gray value was likewise significantly higher in LPS Xssue relaXve to Shams for GFAP , t(39) = 2.072, p = 0.045, IBA1, t(39) = 2.088, p = 0.043, but not for NeuN, t(39) = 0.8789, p = 0.3848, GFAP- posiXve cells were significantly more circular for this experiment , t(39) = 2.152, p = 0.0378, as were IBA1-posiXve cells, t(39) = 2.109, p = 0.0414, whereas perimeter of GFAP-posiXve cells did not differ between groups, t(39) = 0.7562, p = 0.4541, but was significantly lower in IBA1- posiXve cells for LPS animals, t(39) = 2.665, p = 0.0113. (C-D) CorrelaXons between (C) IBA1, and (D) GFAP and c-fos-NeuN percentage colocalizaXon and behavioural performances , r and p values displayed on graphs. (E-F) Results of analyses of immunohistochemical labelling of GFAP , IBA1, and NeuN following NAc core neuroinflammaXon, (E) Individual data points for cell counts (open shapes and lem y axis) and Mean Gray Value (closed shapes and right y axis) values for GFAP (increased counts, t(26) = 4.886, p = 0.000, but not intensity, t(26) = 0.3273, IBA1 (increased intensity, t(26) = 5.110, p = 0.000, but not counts, t(26) = 0.9705, p = 0.5238, and NeuN-posiXve cells (did not differ between groups), (F) Individual data points for circularity (open shapes and lem y axis) and perimeter (closed shapes and right y axis) for GFAP (significant decrease in GFAP circularity, t(26) = 2.412, p = 0.047; perimeter was unchanged, t(26) = 1.955, p = 0.1209) and IBA1 (significant increase in IBA1 circularity, t(26) = 3.829, p = 0.0024; significantly smaller perimeter, t (26) = 3.502, p = 0.0055). * denotes that the p < 0.05. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 52 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 53 Supplement Figure 4. Relates to Figure 4. Supplemental data from fibre photometry (A) and whole-cell patch clamp electrophysiology recordings of MSNs following (B-E) LPS or sham injecXons into the pDMS or (F-I) following the applicaXon of hM4Di-DREADD agonist deschloroclozapine (DCZ) to transfected astrocytes. (B-D) Individual data points and means showing (B) increased rise Xme (t36.26 = 2.038, p = 0.0489), (C) increased instantaneous frequency (t20.47 = 2.245, p = 0.0359), and (D) reduced interspike interval (t16.23 = 2.417, p = 0.0278). Rise Xme changes are further reflected in ( E) example cell average traces for the AP profile characterisXcs (LPS = green, saline = grey). Individual data points and means following DCZ applicaXon with MSNs at (F-I) RMP or (J-M) voltage clamped at -80mV. Data reflected in example cell profile traces presented in (I and M; ASCF = grey, DCZ = black) show AP profile characterisXcs of rise Xme and amplitude in the top traces and AHP profile in the boqom traces. LPS vs saline; LPS at RMP n = 33 cells and at -80 voltage clamp n = 32 cells, from n = 4 animals; saline; n = 15 cells from n = 3 animals. GFAP-HM4Di n = 7 cells from n = 2 animals test ed with ACSF then DCZ. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint 54 Supplemental Figure 5. Relates to Figure 5. Supplemental behavioural results. (A) Magazine entries per min (±SEM) during Pavlovian condiXoning , supported by a main effect of CS period (preCS vs CS) F (1,28) = 742.205, p = 0.000, and of Day F (7,196) = 27.685, p = 0.000, and a Day x CS period interacXon (preCS vs CS) F (7,196) = 48.789, p = 0.000. No main effect of group or any interacXons with group has been detected, Fs < 1, (B) Lever presses per min (±SEM) during instrumental condiXoning, supported by a main effect of day F (7, 196) = 67.262, p = 0.000, no main effect of group (F (2, 28) = 1.803, p = 0.183) and no day x group interacXon (F (14, 196) = 1.281, p = 0.222), largest F (5.66, 79.3) = 1.281, p = 0.277, for group x session interacXon, (C) Magazine entries per min (±SEM) during instrumental condiXoning , supported by a main effect of day F (7, 196) = 8.831, p = 0.000, no main effect of group (F (2, 28) = 1.827, p = 0.180) and no day x group interacXon (F (14, 196) = 0.592, p = 0.870), largest F (10.18, 142.47) = 0.592, p = 0.821, for group x session interacXon, (D) Individual data points and mean magazine entries per min during outcome devaluaXon tesXng , no group differences in magazine entries on test, F (2, 28) = 0.8223, p = 0.4497. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted October 26, 2024. ; https://doi.org/10.1101/2024.10.24.620154doi: bioRxiv preprint