Striatal Patchwork of D1-like and D2-like Receptors Binding Densities in Rats with Genetic Audiogenic and Absence Epilepsies

Binding densities to dopamine D1-like and D2-like receptors (D1DR and D2DR) were studied in brain regions of animals with genetic generalized audiogenic (AGS) and/or absence (AbS) epilepsy (KM, WAG/Rij-AGS, and WAG/Rij rats, respectively) as compared to non-epileptic Wistar (WS) rats. Convulsive epilepsy (AGS) exerted a major effect on the striatal subregional binding densities for D1DR and D2DR. An increased binding density to D1DR was found in the dorsal striatal subregions of AGS-prone rats. Similar changes were seen for D2DR in the central and dorsal striatal territories. Subregions of the nucleus accumbens demonstrated consistent subregional decreases in the binding densities of D1DR and D2DR in epileptic animals, irrespective of epilepsy types. This was seen for D1DR in the dorsal core, dorsal, and ventrolateral shell; and for D2DR in the dorsal, dorsolateral, and ventrolateral shell. An increased density of D2DR was found in the motor cortex of AGS-prone rats. An AGS-related increase in binding densities to D1DR and D2DR in the dorsal striatum and motor cortex, areas responsible for motor activity, possibly reflects the activation of brain anticonvulsive loops. General epilepsy-related decreases in binding densities to D1DR and D2DR in the accumbal subregions might contribute to behavioral comorbidities of epilepsy.


Introduction
Epilepsy is a chronic brain disorder characterized by an enduring predisposition to generate seizures. Epileptic seizures may develop due to an imbalance of excitatory and inhibitory neurotransmitters. Glutamate and gamma-aminobutyric acid are the main neurotransmitters playing a crucial role in the pathophysiology of this balance. Accumulating evidence indicates that monoamines are another group to regulate seizure activity. Experimental studies have indicated that the basal ganglia and its major neurotransmitter dopamine (DA) are involved in seizure propagation [1] and play a role as a "remote control system" over seizures' duration [2] Importantly, epilepsy-related shifts in brain aminergic balance can be a substrate for a number of psychiatric and behavioral comorbidities [3,4] seen in epileptic patients.
Animal models are not replaceable for experimental approaches to studying brain pathologies [5]. One of the oldest animal models for generalized convulsive epilepsies is the Krushinsky-Molodkina (KM) rat, which is genetically prone to severe audiogenic seizures [6]. Rats of the KM strain demonstrate increased sound sensitivity, leading to generalized convulsive seizures in response to acoustic provocation. The main role in generating audiogenic seizures (AGS) belongs to the brain stem, including the corpora quadrigemina [7].

Measurements
The brain structures were identified according to Paxinos and Watson (2007), as described above (Section 2.1). Measurements were carried out in the dorsal and ventral striatum, cingulate (Cg), motor (M), and somatosensory (S) cortex. Somatosensory, motor, and cingulate areas of the neocortex were measured without division into the primary and secondary cortex. The dorsal and ventral striatum were divided into subregions ( Figure 2). The CPu was virtually divided into nine subregions, as shown in Figure 2. NAcb was divided into shell and core parts (abbreviated below as NAcbC and NAcbSh), further parcellated as shown on Figure 2. The obtained values of optical densities were converted into pmol/g of tissue by using the microscale standards (see Section 2.1). The obtained values of optical densities were converted into pmol/g of tissue by using the microscale standards (see Section 2.1).
Standards ® . After the development, all films were digitized, and t cessed in ImageJ.

Measurements
The brain structures were identified according to Paxinos and scribed above (Section 2.1). Measurements were carried out in the d atum, cingulate (Cg), motor (M), and somatosensory (S) cortex. So and cingulate areas of the neocortex were measured without divis and secondary cortex. The dorsal and ventral striatum were divided ure 2). The CPu was virtually divided into nine subregions, as show was divided into shell and core parts (abbreviated below as NAcbC a parcellated as shown on Figure 2. The obtained values of optical den into pmol/g of tissue by using the microscale standards (see Sectio values of optical densities were converted into pmol/g of tissue by standards (see Section 2.1).

Statistical Analyses
Statistical analysis was performed with Statistica14.0.0.15" (TIBCO Software Inc., Pao Alto, CA, USA). Values are shown as means ± SEM. Analysis of D1DR and D2DR binding densities in a defined brain region was performed for each anatomical level and region separately, using ANOVA GLM. Two measurements were taken within each site of each hemisphere and averaged. To account for sources of variation caused by local imperfections in the films, the local background levels were taken as continuous GLM predictors; their effects are not reported below. Anatomical levels were analyzed separately; anatomical subregions of a defined structure (i.e., parts of cortex, CPu, NAcbC, and NAcbSh) were taken as within-subjects factors, so a repeated measures analysis was used. ANOVA GLM for regional data was run with the epilepsy type (AGS and/or AbS, 2 × 2 design) as two between-subjects factors. This analysis provided information about the general effect of AGS/AbS susceptibilities, as previously done [24]. The putative effects of AGS were checked by comparing the pooled group of KM and WAG/Rij-AGS rats with the pooled group of AGS-unsusceptible rats (i.e., WS and WAG/Rij). The effects of AbS' proneness were estimated by comparing the pooled groups of WAG/Rij and WAG/Rij-AGS rats with the pooled groups of WS and KM rats. The general effect of epilepsy ("Epi") was assessed by comparing normal WS rats with the pooled data of the three groups of epileptic rats (i.e., KM, WAG/Rij, and WAG/Rij-AGS). General ANOVAs were followed by Fisher LSD post-hoc tests, if needed. The minimal level of significance was set at p = 0.05. Local striatal gradients in binding densities to D1DR and D2DR were assessed in normal Wistar rats by pairwise comparisons of dependent samples (Wilcoxon test; the minimal p level was set at 0.05). For the correlation part of the study [27], the striatal and accumbal sets of local binding densities to D1DR and D2DR were averaged, thus sets of the mean individual values of D1-and D2-like binding densities were generated for CPu and NAcb. Spearman's rank order for the inter-correlations between individuals' dorsal and ventral striatal values of binding densities was calculated for each group of rats separately. The minimal level of significance was set at p = 0.05.

Distribution of D1DR and D2DR Binding Densities in Normal WS Rats
Generally, the regional distribution of binding densities for D1DR and D2DR agrees with those reported earlier [23,28,29]. Representative autoradiographic images are shown on Figure 3b, together with the mean subregional values of D1DR and D2DR binding densities for normal WS rat brain ( Figure 3a).

Effects of Epilepsies
Recently, we have shown that there might be both epilepsy-type-specific changes in brain aminergic receptors and general (unspecific to seizure type) effects of epilepsy [27]. Below, we also map both epilepsy type-specific and type-unspecific effects (see the scheme for comparisons on Figure 1) on D1DR and D2DR binding densities in brain subregions of rats with genetic generalized epilepsies. The epilepsy-related changes composed a complex "patchwork" pattern, where sites of seizure-type-unspecific changes were surrounded by type-specific local alterations of binding densities. This pattern is similar to that seen for H3 histamine receptor binding densities in subregions of the prefrontal cortex [27] and putatively points to diffuse compensatory post-convulsive changes in brain aminergic activity.
x FOR PEER REVIEW 6 of 15

Effects of Epilepsies
Recently, we have shown that there might be both epilepsy-type-specific changes in

Dorsal Striatum Striatal Binding Densities to D1DR
The rat strain did not differ, and did not tend to differ in their local binding densities to D1DR measured within the CPu at the most rostral level (AP = +2.28). All four groups showed the same pattern of lateromedial and ventrodorsal gradients (Figure 4, left panels), with maximal values at the ventrolateral borders of the CPu. Namely, the strain effect was not significant, p > 0.10; within-subject repeated measures ANOVA GLM showed a tendency to subregions' effect, p = 0.09) These gradients were studied in detail in Wistar samples ( Figure 3). Pairwise comparisons showed that for D1DR, the most dorsomedial locus (number 3, Figure 2) expressed the minimal binding densities (red downward triangle on Figure 3a), being significantly lower than the lateral sites (all p's < 0.05 for comparisons with loci 1 and 6-9, level I; all p's < 0.05 for comparisons with loci 1-7, level II). The most ventrolateral subregion (site number 7, Figure 2) demonstrated the highest binding density to D1DR (green upward triangle on Figure 3a), being significantly higher than the central and median sites (all p's < 0.05 with loci 1-7, level I; all p's < 0.05 for comparisons with loci 1-9, level II). similar to that seen for H3 histamine receptor binding densities in subregions of the prefrontal cortex [27] and putatively points to diffuse compensatory post-convulsive changes in brain aminergic activity.

Dorsal Striatum
Striatal Binding Densities to D1DR The rat strain did not differ, and did not tend to differ in their local binding densities to D1DR measured within the CPu at the most rostral level (AP = +2.28). All four groups showed the same pattern of lateromedial and ventrodorsal gradients (Figure 4, left panels), with maximal values at the ventrolateral borders of the CPu. Namely, the strain effect was not significant, p > 0.10; within-subject repeated measures ANOVA GLM showed a tendency to subregions' effect, p = 0.09) These gradients were studied in detail in Wistar samples (Figure 3). Pairwise comparisons showed that for D1DR, the most dorsomedial locus (number 3, Figure2) expressed the minimal binding densities (red downward triangle on Figure 3a), being significantly lower than the lateral sites (all p's <0.05 for comparisons with loci 1 and 6-9, level I; all p's <0.05 for comparisons with loci 1-7, level II). The most ventrolateral subregion (site number 7, Figure 2) demonstrated the highest binding density to D1DR (green upward triangle on Figure 3a), being significantly higher than the central and median sites (all p's < 0.05 with loci 1-7, level I; all p's <0.05 for comparisons with loci 1-9, level II).  This lateromedial and ventrodorsal gradient is clearly visible (see Figure 4) and should be taken into account in experimental approaches as a source of internal variation. The next anatomical level (II, aimed at AP = +1.8) displayed a sharp AGS-related increase in D1DR binding densities, clearly localized at the dorsal border of the CPu. Namely, subregions were different {F(8.112 = 2.6), p = 0.01}, and strain*subregion interaction was significant as well {F(24.112 = 1.6), p = 0.04}. Closer analysis showed that the two AGS-susceptible groups were responsible for this main effect (AGS effect, F(1.14) = 4.9, p = 0.04), with a significant interaction of AGS * subregion (F(8.112 = 2.4, p = 0.02)). Post-hoc Fisher LSD tests were applied to estimate the local differences between the AGS-prone and AGS-resistant cohorts (Figure 1). The localized measurements (Figure 2) let us point out the subregions 1-3 as significantly elevated in AGS-prone subgroups (all p's < 0.05, deep blue for sites 1, 2, and 3 in Figures 2 and 4). The same tendency was seen in the neighborhood as well (all p's < 0.10, light blue for sites 4, 5, and 6 in Figures 2 and 4). The absence epilepsy cohort (i.e., WAG/Rij and WAG/Rij-AGS pooled together, see Figure 1) did not differ significantly from its controls (KM and WS rats pooled). Previously, we found that WAG/Rij rats displayed decreased D1DR binding densities in CPu of WAG/Rij rats, as compared to epilepsy-resistant ACI rats [23]. Here, we did not see the same effect; therefore, we couldn't confirm that absence-epilepsy-related processes evoke changes in D1DR of the dorsal striatum.

Striatal Binding Densities to D2DR
The rat strains did not differ significantly in their local binding densities to D2DR measured at the most rostral level (AP = +2.28). However, the subregions were highly heterogeneous (within-subject repeated measures, subregion effects, F(8.112) = 8.0 and 12.7 for levels I and II, respectively; both p's < 0.001), with significant strain*subregion interaction effect (F(24.112) = 4.2 and 4.6, for levels I and II, respectively; both p's < 0.001) These gradients were checked in Wistar rats (Figure 3). Pairwise comparisons showed that for D2DR, the most dorsomedial locus (number 3, Figure 2) expressed the minimal binding densities (red downward triangle on Figure 3a), being significantly lower than the lateral sites (all p's < 0.05 for comparisons with loci 1-9, level I; all p's < 0.05 for comparisons with loci 1-7, level II). The most ventrolateral subregion (site number 7, Figure 2) demonstrated the highest binding density to D2DR (green upward triangle on Figure 3a), being significantly higher than the central and median sites (all p's < 0.05 with loci 2-3 and 5-9, level I; all p's < 0.05 for comparisons with loci 1-3 and 5-9, level II). Closer analysis revealed that AGS-susceptible groups differed from their AGS-resistant controls ( Figure 1 explains the comparison), with a significant AGS * subregion interaction effect (F(8.112) = 6.6, p < 0.001). Post-hoc Fisher LSD tests pointed to the two central loci (numbers 3 and 6 on Figures 2 and 4) with an AGS-related increase in D2DR (both p's = 0.01 for comparisons with the AGS-resistant cohort). The next anatomical level (II, AP = +1.8) demonstrated a similar pattern, with the dorsomedial AGS-related increase in D2DR (Figure 4, right panels). The local striatal measurements differed from each other: since subregions showed a significant effect (F(8.112) = 12.7, p < 0.001), with a significant strain * subregion interaction (F(24.112) = 4.6, p < 0.001) and a major effect of AGS * subregion interaction (F(8.112) = 7.8, p < 0.001). To be more precise, subregions 3, 6 demonstrated the AGS-related increase (p's = 0.01 and 0.04, respectively; deep blue spots on Figure 4), neighbored by the same effect and the same tendency in subregions 5 and 2 (p's = 0.04, 0.05, respectively; deep and light blue spots in Figure 4). For audiogenic epilepsy, however, it is known that the hierarchical neuronal network, which determines this pathology, does not involve the basal ganglia [30]. It allows us to speculate that DA receptor density elevation in the dorsal striatum is a consequence of seizures and a possible reason for comorbid catalepsy. A previous study reported a reduced D2 receptor density in the isolated homogenized striatum of KM rats [31]. Therefore, this contradiction could be attributed to the heterogeneity of the striatum (see above) and to the different seizure status in the experimental groups: the cited paper describes seizure-naïve KM rats. Our rats had three sound provocations that resulted in seizure fits in the AGS-susceptible groups (see Methods, Section 2.1), to mimic an active (but not yet kindled) seizure state. It has been shown that the D2 receptor density in the striatum decreases after acute seizures, while antiepileptic treatment increases the D2DR density [32]. Therefore, the reported AGS-related increase in D2DR should be attributed to a post-convulsive activation of the DA receptor system. One important conclusion from our study is about the natural heterogeneity of the striatum, both dorsal and ventral. This should be taken into account during the interpretation of literature data and the planning of further experiments. The comparison of absence epilepsy-prone and absence epilepsy-free groups (Figure 1) did not reveal any significant difference in binding densities to D2DR. Thus, we did not confirm our previous finding on a decreased D2DR binding density related to absence epilepsy and judged by the comparison of WAG/Rij and ACI rats [23].

Ventral Striatum Accumbal Binding Densities to D1DR
The rat strains did not differ significantly in their binding densities to D1DR at the two anatomical levels studied. The accumbal subregions were highly heterogeneous, (F(3.45) = 7.2, p < 0.001) and (F(5.80) = 34.4, p < 0.001) for levels I and II, respectively. No epilepsy-related changes were seen at level I (AP = +2.28). More caudally, at level II (AP = +1.8), the pooled epileptic cohort (i.e., joined KM, WAG/Rij, and WAG/Rij-AGS groups compared to WS groups; see Figure 1) tended to be lower than the control rats (p = 0.10). A post-hoc test revealed that ventrolateral and dorsal aspects of the shell (p = 0.02 and p < 0.001 in Fisher LSD tests), as well as the dorsal core (p = 0.03) displayed epilepsy-related decrease in D1DR (magenta spots on Figure 5, left panels). This was accompanied by the same tendency to decrease, seen in the AGS-prone groups (medial shell, medial core, light blue spots on Figure 5).
x FOR PEER REVIEW 10 of 15 The rat strains did not differ significantly in their binding densities to D2DR at the two anatomical levels studied. No epilepsy-related changes were seen at level I (AP = Figure 5. Comparisons of the ventral striatum binding densities to D1DR and D2DR. The values are given as mean ± SEM, in pmol/g of tissue. For anatomical levels and abbreviations, see Figure 3. Differences are shown by stars: * for p ≤ 0.05, ** for p < 0.01, and 'tend' sign for tendency (0.05 < p < 0.10). The general effects of epilepsies (a difference between the non-epileptic group and the joint group of all epileptic rats) are marked by 'Epi' letters. Effects of audiogenic epilepsy are marked by the 'AGS' sign.
The results agree with the previous finding on a decreased D1DR binding density in the NAcbC of WAG/Rij rats [23]. As mentioned earlier, in the GAERS model, administrations of D1/D2 agonists into this particular structure were effective in modulating SWDs [17]. Therefore, we can hypothesize that the mediadorsal aspects of the NAcbC are engaged in the brain's anti-epileptic loop, which is responsible for the occurrence of SWDs. Our study extends the effect to convulsive epilepsy as well, since AGS-prone rats demonstrated the same pattern of decreased binding to D1DR seen in accumbal regions ( Figure 5 left panels).

Accumbal Binding Densities to D2-like Dopamine Receptors
The rat strains did not differ significantly in their binding densities to D2DR at the two anatomical levels studied. No epilepsy-related changes were seen at level I (AP = +2.28). More caudally, at level II (AP = +1.8), a complex pattern of epilepsy-related decreases was seen. Interactions between subregion * strains were significant {F(15.75) = 2.1, p = 0.02}. A seizure-type specific decrease was seen in a single structure: AGS-groups demonstrated a dramatic decrease in dorsolateral shell (p = 0.02; Figure 5 right panels), mostly at the expense of KM rats (deep blue column on a chart). A general effect of epilepsies was the decrease in binding densities to D2DR seen in the dorsal shell and ventrolateral shell (p= 0.04 and p = 0.01, respectively; magenta spots on the scheme, Figure 5), paralleled by the same tendency seen for the medial shell (p = 0.07).
The shell of NAcb was not a sensitive site to modulate SWDs in the GAERS model [17], and no difference in D2DR was seen earlier [23]. However, in epileptic patients, this particular region is reported as a putative site for deep brain stimulation in intractable partial epilepsy [33], and seizure-related neuronal degeneration was seen in the structure in patients with temporal lobe epilepsy [34]. Recently, it was shown that experimental temporal lobe seizures in mice induced by intra-amygdalar kainic acid administration increased the neuronal excitability of medium spine neurons of NAcbSh, both D1R-expressing and D2R ones [35]. Besides the fact that the pharmacology and the brain structures are completely different between temporal lobe epilepsy and absence epilepsy, it is also possible that the shell of the NAcb is not involved in ongoing seizure activity but is a constituent part of a slow anti-epileptic brain loop. Recently, behavioral characteristics were proposed as potential biomarkers for epileptogenesis in rats [36]. Therefore, DAergic transmission within NAcb is likely involved in shaping the behavioral comorbidities of epilepsies and thus might be an important target for pharmacological interventions.
An intriguing possibility would be to link the deficient D2DR binding densities ( Figure 5 right panels), seen in the ventrolateral shell of the nucleus accumbens of epileptic rats, profoundly in KM rats, with recently reported social behavioral deficits in KM rats [37]. In human patients, the volume of the nucleus accumbens positively correlates with the size of the individual social network [38]. Therefore, accumbal deficits and degeneration, related to epilepsies, might contribute to the increased frequency of social deficits in epileptic patients [39]. Behavioral depression and anhedonia, described in WAG/Rij rats [40,41], generally attributable to absence epilepsy-related shifts in tissue indices of brain aminergic turnover [42], also might be related to the observed deficient D2DR binding densities within medial and dorsal aspects of accumbal shell ( Figure 5, right panel). Further behavioral studies with local drug administration would shed more light on this important question.

Motor and Somatosensory Cortex
Rat strains did not differ significantly in D1DR and D2DR binding densities measured at all four levels. No epilepsy related changes were seen in the D1DR binding densities in the cortical areas' measures (joined primary and secondary motor cortices, somatosensory cortex).
In contrast, binding densities to D2-like DA receptors demonstrated a remarkable AGS-related elevation in the motor cortex ( Figure 6; level III, AP = −0.96). The AGS effect was significant (F(1.15) = 5.8, p = 0.02), as was the AGS * subregion interaction (F(1.15) = 4.7, p = 0.04). Post-hoc tests showed that the pooled AGS cohort had a higher Diagnostics 2023, 13, 587 11 of 14 D2DR binding density in the motor cortex (p < 0.01). No difference was seen for absenceepileptic rats, in contrast to what was reported earlier [23]. Previously, ACI rats were used as the control group and it cannot be excluded that the choice of a non-WS derived control might have contributed to the earlier obtained differences. Another possibility is that perhaps a more detailed analysis targeted specifically to the barrel cortex of WAG/Rij rats would bring more information about possible alterations of DA receptors around "the hot spot", the cortical site of origin of the SWDs. From the other side, it is also known that some WS rats have SWDs as well [43]. Therefore, the reason for the absence of a significant difference between WS and WAG/Rij rats could be the presence of SWDs at least in some WS rats or the choice of the non-epileptic control strain. For future experiments, it might be recommended to check not only the audiogenic seizures but also the presence of SWDs in all experimental groups.
In contrast, binding densities to D2-like DA receptors demonstrated a remarkable AGS-related elevation in the motor cortex ( Figure 6; level III, AP = −0.96). The AGS effect was significant (F(1.15) = 5.8, p = 0.02), as was the AGS * subregion interaction (F(1.15) = 4.7, p = 0.04). Post-hoc tests showed that the pooled AGS cohort had a higher D2DR binding density in the motor cortex (p < 0.01). No difference was seen for absence-epileptic rats, in contrast to what was reported earlier [23]. Previously, ACI rats were used as the control group and it cannot be excluded that the choice of a non-WS derived control might have contributed to the earlier obtained differences. Another possibility is that perhaps a more detailed analysis targeted specifically to the barrel cortex of WAG/Rij rats would bring more information about possible alterations of DA receptors around "the hot spot", the cortical site of origin of the SWDs. From the other side, it is also known that some WS rats have SWDs as well [43]. Therefore, the reason for the absence of a significant difference between WS and WAG/Rij rats could be the presence of SWDs at least in some WS rats or the choice of the non-epileptic control strain. For future experiments, it might be recommended to check not only the audiogenic seizures but also the presence of SWDs in all experimental groups.

Regional Correlations of D1DR and D2DR Binding Densities
Strong positive (Figure 7, all Rs >0.70) correlations between the mean local striatal and accumbal values of D1DR and D2DR binding densities (see Section 2.5) were seen in the control WS rats as well as in the WAG/Rij group, suggesting a highly functional coupling of local receptor systems. Putatively, a brainstem DA supply coordinates exposure and bindings to D1DR and D2DR within the projected regions, CPu and NAcb. Remarkably, this intimate relationship coupling was broken in the two AGS-prone groups ( Figure 6) where mean accumbal binding densities to D2DR receptors lost its strong and positive correlation with their D1DR areas in KM and WAG/Rij-AGS rats (Figure 7). Although correlations [44] provide only suggestive information on linkage between the correlated units, we might hypothesize that having AGS experience imbalances receptor response to the brainstem DA supply specifically within the nucleus accumbens. The difference between the dopaminergic reactivity of the nigrostriatal and mesolimbic systems is essential for the formation of the complex behavioral phenotype and should further contribute to epileptogenesis as well [45].
correlations [44] provide only suggestive information on linkage between the correlated units, we might hypothesize that having AGS experience imbalances receptor response to the brainstem DA supply specifically within the nucleus accumbens. The difference between the dopaminergic reactivity of the nigrostriatal and mesolimbic systems is essential for the formation of the complex behavioral phenotype and should further contribute to epileptogenesis as well [45]. 7. Brain regional binding densities to D1DR and D2DR correlate tightly in two non-AGS groups, but these local correlations are broken in the two AGS-prone cohorts. Axes are mean binding densities, in pmol/g. Each chart represents plots for the rat groups studied (Wistar, WAG/Rij, WAG/Rij-AGS, KM). Each data point represents mean regional D2DR NAcb density binding densities (pmol/g tissue) vs. D1DR or D2DR density in dorsal (blue diamonds for CPU, D1DR; red squares for CPU, D2DR) or ventral striatum parts (green triangles for NAcb, D1DR). In spearman rank order, Rs are given on the charts for significant local correlations (p < 0.05), "ns" stands for non-significant ones.

Conclusions
Figure 7. Brain regional binding densities to D1DR and D2DR correlate tightly in two non-AGS groups, but these local correlations are broken in the two AGS-prone cohorts. Axes are mean binding densities, in pmol/g. Each chart represents plots for the rat groups studied (Wistar, WAG/Rij, WAG/Rij-AGS, KM). Each data point represents mean regional D2DR NAcb density binding densities (pmol/g tissue) vs. D1DR or D2DR density in dorsal (blue diamonds for CPU, D1DR; red squares for CPU, D2DR) or ventral striatum parts (green triangles for NAcb, D1DR). In spearman rank order, Rs are given on the charts for significant local correlations (p < 0.05), "ns" stands for non-significant ones.

Conclusions
WAG/Rij and KM strains are both WS-derived [6,43]. It allows us to suggest that regional differences in D1DR and D2DR binding densities could be attributed to the difference between convulsive and non-convulsive forms of epilepsy or its specific comorbidities, since our strains differ in the presence or absence of particular forms of epilepsy.
Our results point to the essential heterogeneity of studied striatal territories in terms of local binding densities to D1DR and D2DR. This should be taken into account in the interpretation of data in the literature.
Decreased D1DR binding density in the dorsal aspect of the nucleus accumbens core is a consistent feature of epileptic brains reported in literature and confirmed in our study. The obtained results suggest that this effect is not specific for absence-epileptic brains, but expands also on convulsive epilepsy. This particular structure might be a part of the brains' anti-epileptic loop which in turn mediates behavioral comorbidities of epilepsies.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee of the IHNA and NPh of RAS (protocol 1 from 25 February 202l).