Dendritic spine density is increased on nucleus accumbens D2 neurons after chronic social defeat

Stress alters the structure and function of brain reward circuitry and is an important risk factor for developing depression. In the nucleus accumbens (NAc), structural and physiological plasticity of medium spiny neurons (MSNs) have been linked to increased stress-related and depression-like behaviors. NAc MSNs have opposing roles in driving stress-related behaviors that is dependent on their dopamine receptor expression. After chronic social defeat stress, NAc MSNs exhibit increased dendritic spine density. However, it remains unclear if the dendritic spine plasticity is MSN subtype specific. Here we use viral labeling to characterize dendritic spine morphology specifically in dopamine D2 receptor expressing MSNs (D2-MSNs). After chronic social defeat, D2-MSNs exhibit increased spine density that is correlated with enhanced social avoidance behavior. Together, our data indicate dendritic spine plasticity is MSN subtype specific, improving our understanding of structural plasticity after chronic stress.

www.nature.com/scientificreports/ The NAc also integrates excitatory input from several stress-sensitive afferents including thalamus, hippocampus, and medial prefrontal cortex [21][22][23] . After CSDS, excitatory input is strengthened onto NAc MSNs due in part to the creation of new dendritic spines 24 . However, changes in MSN physiology after CSDS are subtype specific. Excitatory input is only increased onto D2-MSNs, and is instead weakened onto D1-MSNs 20 . Given these opposing physiological changes in MSN subtypes, and an apparent lack of spine changes on D1-MSNs 9 , we hypothesized D2-MSNs have increased dendritic spine density after CSDS. To test this, we used viral labeling of D2-MSNs with the D2-MSN specific A2A-Cre line to characterize dendritic spines in mice subjected to CSDS and find density of specific D2-MSNs dendritic spine types correlates with social avoidance behavior.

Spine density is increased in D2-MSNs after chronic social defeat stress. Previous work indi-
cates spine density is increased after CSDS [24][25][26] , however work from our lab indicates spine density is largely unchanged in D1-MSNs after CSDS 9 . To test the hypothesis that spine density is altered in D2-MSNs, we infused a low titer Cre dependent eYFP or mCherry to sparsely label D2-MSNs in the NAc (timeline in Fig. 1a, schematic Fig. 1c) 8 . After CSDS, mice were tested for social avoidance behavior. Due to the small number of mice, we did not divide mice into susceptible or resilient groups 4 but instead report individual social interaction ratios. Consistent with previous work, most mice developed social avoidance after CSDS (SI ratio < 1, Fig. 1b). We assessed spine density in CSDS mice compared with unstressed controls irrespective of spine type (Fig. 2a 'Total Spines' , representative segments in Fig. 2b). The number of spines per 10 µm dendritic length was increased in CSDS mice relative to control mice (12.4 ± 2.2 vs 5.0 ± 0.4, Welch corrected t (7.418) = 3.297, p = 0.0121).
Spine density correlates with social avoidance. To strengthen the argument that increased spine density after CSDS is driven by D2-MSNs, we next sought to replicate correlations between social avoidance and spine density. Using Pearson's Correlation, we found a negative correlation between total spine density and social interaction ratio (Fig. 3, 'Total Spines' , r = − 0.6104, p = 0.0267). When we performed similar correlations based
In addition to overall spine density, individual dendritic spine morphology is important to consider since spine size and type are important determinants of synaptic strength 41,42 . While dendritic spine shape can change on a second-by-second basis 43 , changes to spine volume or type occur over days to months, and require several structural refinements 44 . Spines are classified into subtype based on the relative appearance of spine head and neck. "Thin" spines have long, thin necks and small heads, "mushroom" spines have thicker necks ending in large, bulbous heads, while "stubby" spines consist mainly of a spine head. Thin and stubby are considered immature spine types and are less stable as compared with the more mature mushroom type. Compared with thin and stubby, mushroom spines also form larger post synaptic densities that may translate to stronger synapses 45 . Here, we found significantly increased stubby spine density after CSDS that correlated with decreased social interaction behavior, consistent with non-subtype specific work 24,25 . Despite the smaller post synaptic density, stubby spines lack a defined neck and are more tightly coupled to their parent dendrite 46 . Thus, a change in stubby spine number may exert a proportionally greater influence on neuronal physiology. Excitatory input onto these spines may drive more action potentials in D2-MSNs, in turn driving stress-susceptibility 20 . We also found thin spine density was increased after CSDS that trended towards correlation with social avoidance (p = 0.055). The increased immature spine density after CSDS, together with the overall increase in spine density, indicates new spines are being formed on D2-MSNs-i.e. the increase in thin and stubby spine types is not due to the conversion of mushroom spines to another spine type. Indeed, mushroom spine density is trending towards an increase after CSDS, not a decrease, and we found a significant correlation between mushroom spine density and social avoidance. Thus, the most CSDS-susceptible mice may have more thin and stubby spines prior to stress that mature into mushroom spines during stress, in addition to newly created spines. As a corollary, CSDS-resilient mice may have lower spine density at baseline, leaving them less vulnerable to the negative consequence of new D2-MSN spines. Alternatively, resilient mice may upregulate synaptic pruning machinery to counteract the formation of immature dendritic structures, and/or decrease the synaptic strength of individual spines 47 . Regardless, decreasing overall D2-MSN spine strength and density is likely a key mechanism of CSDS-resilience. Future work tracking individual dendritic spine formation and stability in vivo will be critical for understanding specific mechanisms of chronic stress.
Aside from altered synaptic inputs, neuronal morphology changes are also driven by altered gene expression. For example, expression of cytoskeleton remodeling molecules is altered by CSDS 8,9,25 . While RhoA driven decreased D1-MSN dendritic complexity is sufficient to drive CSDS susceptibility 8,48 , the specific molecular mechanisms underlying increased spine formation on D2-MSNs remains unknown. It is tempting to speculate molecules with established, non-subtype specific roles, exert their effects preferentially in D2-MSNs. For example, both Dnmt3a overexpression 49 or constitutively active IκB kinase (IKK) 40 increase MSN spine density. Further work is needed to identify the specific molecular mechanisms for driving changes to dendritic spines specifically in D2-MSNs, and if spine changes alone are sufficient to confer susceptibility to CSDS.
In conclusion, we identified the specific MSN subtype that undergoes dendritic spine plasticity after CSDS. Since D1-and D2-MSNs undergo such opposing physiological and structural adaptations, this finding has important implications for interpreting non cell-type specific studies. Future work should continue to evaluate cellular and molecular changes after CSDS in the NAc with cell subtype in mind, which will ultimately increase our understanding of neurobiological processes underlying disrupted behaviors after stress, which has relevance for affective disorders such as depression. We imaged secondary dendrites from 3-4 cells per mouse from at least 2 separate slices in both NAc core and shell, including cells on the core/shell border (See Fig. 1a). Using Neuron Studio software 50 (Mt Sinai School of Medicine), a blinded experimenter selected secondary dendrites (> 20 µm long, > 40 µm from soma) that could be clearly resolved from neighboring dendrites (i.e. no overlapping branches) 24,25,[51][52][53] . This resulted in a different number of dendrites available for analysis in each cell, thus 3-4 of these dendrites per cell were randomly selected, yielding a total of 9-15 dendrites per mouse. We analyzed secondary dendrites due to the relative dearth of spines on the more proximal, primary dendrites of MSNs, and to be consistent with existing literature 24,25,40,52,54 . To reduce background noise, the Z-blur and median blur filters were applied to all images. All spines identified by Neuron Studio were visually confirmed in a 3D rendering by a blinded experimenter, and any misidentified spines (e.g. noise, dendritic segment) were manually removed. Spine densities were first averaged for a given cell, and then all cells from the mouse were averaged together to generate a 'grand average' for an individual mouse. If a sufficient number of dendrites could not be analyzed from a given mouse it was removed from the study. One control mouse was removed from the main figures with Grubb's Outlier Test (α = 0.05). Social defeat stress. Chronic social defeat stress (CSDS) was performed according to well-documented procedures 3 as previously in the lab [8][9][10]48,55 . The mice were placed in hamster cages with perforated plexiglass dividers containing a novel, aggressive CD1-resident. Mice were physically defeated by a new resident for 10 min, then housed opposite the resident for 24 h sensory interaction for 10 consecutive days, each day encountering a new aggressive resident. The unstressed control mice were housed 2/cage, separated by a perforated plexiglass divider. 24 h after the last defeat, social avoidance was assessed with videotracking software (CleverSys, Reston, VA, USA). Experimental mice were placed in an open field containing a perforated chamber. Time spent around the chamber ("interaction zone") was compared between two, 2.5 min trials during which the chamber was empty or contained a novel CD-1. Social interaction ratios were calculated by dividing time spent in the interaction zone with and without the novel mouse present.

Methods
Statistics. All statistics were performed using Graph Pad Prism software. Only mice for which we successfully analyzed 9-15 dendrites are included in this report. Unpaired two-tailed t-tests with Welch's correction for unequal variance were used to compare between control and CSDS spine densities. Pearson's correlation was used to compare the relationship between spine density and social interaction ratio.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.