Neuroendocrine and behavioral measures of negative valence in male sign-tracker and goal-tracker rats

Cues, or stimuli in the environment, attain the ability to guide behavior via learned associations. As predictors, cues can elicit adaptive behavior and lead to valuable resources (e.g., food). For some individuals, however, cues are transformed into incentive stimuli and can elicit maladaptive behavior. The goal-tracker/sign-tracker animal model captures individual differences in cue-motivated behaviors, with reward-associated cues serving as predictors of reward for both goal-trackers and sign-trackers, but becoming incentive stimuli only for sign-trackers. While these distinct phenotypes are characterized based on Pavlovian conditioned approach behavior, they exhibit differences on a number of behaviors of relevance to psychopathology. To further characterize the neurobehavioral endophenotype associated with individual differences in cue-reward learning, we investigated neuroendocrine and behavioral profiles associated with negative valence in male goal-trackers, sign-trackers, and intermediate responders. We found that baseline corticosterone increases with Pavlovian learning, and that this increase is positively associated with the development of sign-tracking. We did not observe significant differences between goal-trackers and sign-trackers in behavior during an elevated plus maze or open field test, nor did we see differences in the corticosterone response to the open field test or physiological restraint. We did, however, find that sign-trackers have greater glucocorticoid receptor mRNA expression in the ventral hippocampus, with no phenotypic differences in the dorsal hippocampus. These findings suggest that goal-trackers and sign-trackers do not differ on indices of negative valence; rather, differences in neuroendocrine measures between these phenotypes can be attributed to distinct cue-reward learning styles. Significance Statement While the goal-tracker/ sign-tracker animal model derives from individual differences in Pavlovian conditioned approach behavior, other traits, including some of relevance to addiction and post-traumatic stress disorder, have been shown to co-exist with the propensity to sign-track. The extent to which this model encompasses differences in negative valence systems, however, remains largely unexplored. Here we show that behavioral and corticosterone response to paradigms associated with negative valence do not differ between goal-trackers and sign-trackers; but baseline corticosterone levels appear to be linked to the development of sign-tracking, as do differences in glucocorticoid receptor expression in the ventral hippocampus. These findings suggest that neuroendocrine measures typically associated with negative valence may, in fact, play an important role in positive valence systems.


Introduction
Through learned associations, environmental cues become predictors of biologically relevant stimuli. In turn, such cues elicit an adaptive response, facilitating behavior towards valuable resources. However, for some individuals, cues elicit complex emotional responses and prompt maladaptive behavior. For example, upon exposure to drug-associated cues, individuals with addiction report drug-craving and, consequently, often relapse (Ehrman et al., 1992).
Similarly, when exposed to trauma-related stimuli, individuals with post-traumatic stress disorder (PTSD) report hyperarousal and anxiety (Shin et al., 2004). Cues attain the ability to elicit extreme emotional states and aberrant behavior when they are attributed with excessive incentive motivational value, or incentive salience (Robinson & Berridge, 1993). The propensity to attribute incentive salience to environmental cues, thereby, may reflect a vulnerability trait for cue-motivated psychopathologies, like addiction and PTSD (Flagel et al., 2010;Morrow et al., 2011).
Individual variation in the propensity to attribute incentive salience to reward cues can be captured using a Pavlovian conditioned approach (PavCA) paradigm, consisting of a lever-cue paired with delivery of a food-reward . Upon lever-cue presentation, goal-trackers (GTs) direct their behavior towards the location of reward delivery, whereas signtrackers (STs) direct their behavior towards the cue itself. For both GTs and STs the cue attains predictive value, but for STs the cue also attains incentive value and is transformed into a "motivational magnet" (Berridge et al., 2009). Intermediate responders (IRs) vacillate between goal-and cue-directed behavior, without an innate preference for either cue-learning strategy.
Most of the research surrounding the GT/ST model has focused on "positive valence" and the associated neurobiology, and only a few studies have investigated indices of "negative valence" (e.g., Harb & Almeida, 2014;Morrow et al., 2011;Vanhille et al., 2015). The Research Domain Criteria (RDoc), put forth by the National Institute of Mental Health, defines "positive valence systems" as those responsible for responses to motivational situations, including reward learning; and "negative valence systems" as those responsible for responses to aversive situations, including fear and anxiety (RDoC). Per the RDoC, corticosterone (CORT), the final product of the hypothalamic-pituitary-adrenal (HPA) axis in rodents, is a molecular marker of negative valence. However, we know that the role of CORT extends beyond the "stress response" and into arenas of learning and memory (e.g., Sandi et al., 1997), reward learning (e.g., Tomie et al., 2002), and reinforcement (e.g., Piazza et al., 1993). Of particular relevance, CORT is involved in forming Pavlovian associations for both aversive (e.g., Marchand et al., 2007) and appetitive (e.g., Tomie et al., 2002) stimuli (for review see Lopez & Flagel, 2020).
With respect to the latter, relative to GTs, STs show a greater rise in CORT following an initial PavCA session, prior to the development of a conditioned response ). In the current study, we assessed the role of CORT in the propensity to attribute incentive value to reward cues by examining baseline levels before and after the acquisition of PavCA behavior (Experiment 1A). In addition, we probed negative valence systems within the context of the GT/ST animal model and assessed behavioral and CORT responses to environmental challenges, including the elevated plus maze, open field test, and acute physiological restraint (Experiment 1B). As the hippocampus is a central component of negative valence circuits (RDoC), and glucocorticoid receptors (GRs) within the hippocampus key regulators of HPA-axis activity (Akil, 2005;Herman et al., 1989), we also assessed hippocampal GR mRNA in GTs, STs, and IRs (Experiment 2). These studies allowed us to expand our characterization of the neurobehavioral endophenotype captured by the GT/ST model, and to begin to deconstruct the role of CORT beyond negative valence systems.

Animals
Sixty male Sprague-Dawley rats were obtained from Charles River Breeding Labs (Saint-Constant, Canada, Colony 72 (C72), and Raleigh, NC, USA, Colony 04 (R04)) for both Experiments 1A and Experiment 1B. Rats weighed between 225-275 g upon arrival and were pair-housed in standard acrylic cages (46 x 24 x 22 cm) in a temperature-controlled room under a 12-h light/dark cycle (lights on at 7:00). Food and water were available ad libitum for the duration of the study. Rats were allowed to acclimate to their colony room and remained undisturbed in their homecages for seven days after arrival. Rats were then briefly handled every day for five consecutive days before any experimental manipulation. During the last two days of handling, twenty-five 45-mg banana-flavored grain pellets (Bio-Serv, Flemington, NJ, USA) were placed inside the homecage, allowing rats to habituate to the food reward used during Pavlovian conditioned approach (PavCA) training. Behavioral testing occurred during the light cycle (between 10:00 and 14:00). All experimental procedures followed The Guide for the Care and Use of Laboratory Animals: Eight Edition (2011, National Academy of Sciences), and were approved by the University of Michigan Institutional Animal Care and Use Committee.

Pavlovian conditioned approach (PavCA) training
All PavCA training took place in standard behavioral testing chambers (MED Associates, St. Albans, VT, USA; 20.5 × 24.1 cm floor area, 29.2 cm high) located inside a room with red lighting. The chambers were enclosed in sound-attenuating boxes equipped with a ventilation fan that provided constant air circulation and served as white noise. Each chamber contained a foodcup centered on one of the walls and placed 6 cm above the grid floor. The food-cup was equipped with an infrared beam, and each beam break was recorded as a head entry.
Counterbalanced, right or left of the food-cup, was a retractable lever that illuminated upon presentation and was also placed 6 cm above the floor. A force of at least 10 g was necessary to deflect the lever; this deflection was recorded as a "lever contact." On the opposite wall, a white house light was placed 1 cm from the top of the chamber. House light illumination signaled the beginning of the session and remained on for the duration of the session.
Rats underwent a single pre-training session, where the food-cup was baited with threegrain pellets in order to direct the rat's attention to the location of the reward. Once placed in the chamber, the house light turned on after 5 min, signaling the beginning of the session. The pretraining session consisted of 25 trials during which the lever remained retracted, and pellets were delivered randomly into the food-cup; one pellet per trial on a variable interval 30 s schedule (range 0-60 s). The total session length was approximately 12.5 min.
Following pre-training, or twenty-four hours later, rats underwent a total of five consecutive PavCA training sessions. Each session consisted of 25 trials on a variable interval 90 s schedule (VI 90, range 30-150 s) during which an illuminated lever (conditioned stimulus, CS) was presented for a total of 8 s, and immediately upon its retraction, a food pellet (unconditioned stimulus, US) was delivered into the adjacent food-cup. Each session lasted approximately 40 min.
The following behavioral measures were recorded during each PavCA session: (1) number of lever contacts, (2) latency to contact the lever for the first time, (3) probability to contact the lever, (4) number of food-cup entries during presentation of the lever, (5) latency to first enter the food-cup during presentation of the lever, (6) probability of entering the food-cup during presentation of the lever, and (7) number of food-cup entries during the inter-trial interval.
These values were then used to calculate three measures of approach behavior that comprise the PavCA index: (1) response bias [(total lever pressestotal food-cup entries) ÷ (total lever presses + total food-cup entries)], (2) probability difference [probability to approach the leverthe probability to enter the food-cup], (3) latency difference [± (latency to approach the leverlatency to enter the food-cup) ÷ 8]. As previously described (Meyer et al., 2012), PavCA index score was calculated from the average of sessions 4 and 5 using this formula: [(response bias + probability difference + latency difference) ÷ 3]. Scores ranged from +1 to -1; a more positive score indicated a preference for sign-tracking behavior and a negative score for goal-tracking.
The cutoffs for phenotype classification were: ≤ -0.5 for a GT, ≥ 0.5 for a ST, and in between -0.5 and 0.5 for an IR, those that vacillate between the two conditioned responses.

Sample collection
To investigate plasma corticosterone (CORT) profiles, blood samples were collected via lateral tail nick at the time points indicated below for Experiment 1A and 1B (see also Figure   1A). A small (≤ 5 mm) nick was made with the tip of a razor blade, and blood was extracted via capillary action (~200 µL) into an EDTA-coated tube (Sarstedt, Nümbrecht, Germany). Samples were capped, inverted 2-3 times, and immediately placed onto ice where they remained until the last tail nick was performed (< 3 hr standing time). Samples were then separated by centrifugation (13,000 rpm for 10 min at 4 ºC), and plasma was extracted, flash-frozen on dry ice, and stored at -20 ºC until processed for radioimmunoassay.

Radioimmunoassay
Plasma CORT levels were measured using commercially available CORT I 125 Double Antibody Radioimmunoassay (RIA) kit (MP Biomedicals, Solon, OH) with a minimum detectable dose of 7.7 ng/mL. The manufacturer's protocol was followed verbatim. A range of 25-1000 ng/mL CORT calibrators was used to generate a standard logarithmic curve for every set of 76 test tubes (the centrifuge test tube capacity for one spin). For Experiments 1A and 1B, a total of 482 plasma samples (not including duplicates or calibration standards) were assayed using 19 centrifuge spins across 6 days, with no more than 4 sets (i.e., centrifuge spins) per day.
Gamma radiation counts per minute were averaged across duplicate samples and converted into CORT concentrations using the average standard curve generated from all sets that were run for each day of RIA. On average, the intra-assay coefficient of variation was 7.24%, while, the interassay coefficient of variation was 16.44%. Outliers were identified and removed if: 1) duplicates had a percent error greater than 10%, or 2) samples were identified as an extreme outlier (3x the interquartile range) by statistical software, or 3) notes confirmed an observable reaction to tail nicks (e.g., vocalization).

Sample collection
Samples were collected, as described above, under baseline conditions before Pavlovian conditioned approach training (Pre-PavCA) and following the development of a conditioned response to the lever-cue (Post-PavCA) (refer to the experimental timeline, Figure 1A). Pre-PavCA tail nicks were performed 24 h prior to the pre-training sessions (see Experiment 1 Behavioral Testing), while Post-PavCA tail nicks were performed 24 h after the last session (Session 5) of training. On days of collection, six rats were transported in their paired-housed homecages into a designated room (start 10:30), where all collection took place under white light. Tail nicks were performed one at a time (~ 60-90 s per collection). Each wave of six rats remained in the room together but on the opposite side of the room from the collection area.
After the last tail nick was performed, all rats were promptly returned to the colony room. Rats were left undisturbed for a total of ten-days before Experiment 1B began.

Sample collection
Plasma CORT levels induced by behavioral assays of negative valence and physiological restraint were captured using tail nick sampling procedures as described above. Collections took place 24 h before the open field test (time 0, or baseline) and 20, 40, 60, and 80 min post-onset of the test. For restraint-induced CORT profiles, collections took place immediately when rats were placed into the restraining device (time 0, or baseline) and 30 (before being released), 60, 90, and 120 min after the onset of restraint. Rats were transported into the designated room in a staggered fashion, one at a time to begin collections. Repeated nicks were performed on each rat to capture all of the time points. Up to 9 rats remained together in the designated collection room but were on the opposite side of the room from the collection area. Rats were returned to the colony room in a staggered fashion after their last sample was collected.

Elevated plus maze (EPM)
After a ten-day rest period following Experiment 1A, rats were exposed to an elevated plus maze (EPM) test (refer to experimental timeline, Figure 1A), considered to be a metric of anxiety-like and risk-taking behavior (Lister, 1987;Walf & Frye, 2007). The apparatus was constructed of four connected arms (each 70 cm from the floor, 45 cm long, and 12 cm wide) made of black Plexiglass and arranged in a cross shape. 45-cm high walls enclosed two opposite arms, and the remaining two were open platforms. A central square (12 x 12 cm) connected all four arms. The test room was dimly lit (40 lux) by a light fixture located above the maze. Prior to the test, rats were transported inside their homecage, along with their cage mate, into the testing room and left undisturbed to acclimate for 30 min. Upon starting the test, each rat was placed in the central platform facing an open arm and allowed to roam freely around the maze for a total of 5 min. The experimenter remained in the room but was distanced from the apparatus in order to be out of the rat's view. A video-tracking system (Noldus Ethovision 11.5, Leesburg, VA) using a live feed from a digital camera mounted on the ceiling directly above the center of the maze was used to detect and record: 1) latency to enter the open arms for the first time, 2) frequency to enter each arm, and 3) time spent in each arm. Additionally, universally used risk assessment behaviors (RABs, see Mikics et al., 2005;Rodgers et al., 1999) were scored manually by the experimenter viewing the live recording. Specifically, the number of times the rat exhibited a bout of grooming, rearing, and protected and unprotected head dips (i.e., head dips over the side of the maze while their body was inside an enclosed arm vs. their body being completely exposed on the open platforms) was quantified.

Open field test (OFT)
After a ten-day rest period following EPM testing, rats were exposed to an open field test (OFT; refer to the experimental timeline, Figure 1A), a metric of negative valence (see RDoC) and exploratory behavior (Walsh & Cummins, 1976). The OFT test occurred in the same room as the EPM test, and again, paired-housed rats were transported from the colony room to the dimly lit test room and allowed to acclimate for ~30 min before testing began. The open field apparatus was a 4-wall Plexiglass enclosure with an open top and plexiglass floor (100 x 100 x 50 cm). At the start of the test, rats were placed into the same corner (bottom left) of the arena and allowed to roam freely for 5 min. Behavior was video recorded with a digital camera mounted above the apparatus. Noldus Ethovision (11.5, Leesburg, VA) was used to detect: 1) the time spent in the center of the arena (a 50 x 50 cm square drawn in the center), 2) the time spent in the outer edge of the arena (25 cm wide border), 2) the number of entries into the center arena, 3) latency to enter the center of the arena for the first time, and 4) total distance traveled.

Restraint
After a ten-day rest period following the OFT, rats underwent a single session of physiological restraint. The restraining device consisted of a white 9 x 12 cm sleeve of flexible Teflon secured with two black Velcro straps attached to a 9 x 3 cm clear Plexiglas platform with a tail slit on one end and breathing holes on the other. Rats were transported in their homecage into the testing room, which was the same as that used for Experiment 1A and OFT time-course measures. Rats were placed into the restrainer and remained there for 30 min.

Experiment 2: Glucocorticoid receptor (GR) mRNA expression within the hippocampus of goal-trackers, sign-trackers, and intermediate responders Animals
An additional sixty male Sprague-Dawley rats were obtained from Charles River Breeding Labs (C72 and R04) for this experiment. Housing and testing conditions were identical to those described in Experiment 1, except that lights turned on and off at 6:00 and 18:00 h, respectively. Rats were exposed to 2 days of handling prior to behavioral testing, which occurred between 11:00 and 15:00 h.

Pavlovian conditioning approach (PavCA) training
PavCA training and classification of GT, ST, and IR were performed identically to that described above for Experiment 1.

Tissue collection
Twenty-four hours after completion of the 5 th PavCA training session (refer to the experimental timeline, Figure 1B), rats underwent live decapitation, and their brains were extracted and immediately flash frozen in 2-methyl butane (-30ºC). Brains were stored at -80ºC until further processing. Frozen brains were mounted perpendicularly to a metal cryostat chuck using Optimal Cutting Temperature compound (Fisher Healthcare, Thermo Fisher Scientific Kalamazoo, MI, USA) and coated with Shandon M-1 embedding matrix (Thermo Fisher Scientific, Kalamazoo, MI, USA) in preparation for sectioning. Whole brains were coronally sectioned at 10 μm on a cryostat at -20°C. Brain sections were collected, 4.68 to -7.08 mm from Bregma, and directly mounted onto Superfrost Plus microscope slides (Fischer Scientific, Pittsburg, PA, USA), with four sections per slide and ~200 µm between sections on a given slide. Slides were stored at -80°C in preparation for in situ hybridization.

Probe synthesis
Probes for in situ hybridization were synthesized in-house using rat mRNA sequences complementary to the RefSeq database number for Type II glucocorticoid receptor (M14053).
All cDNA segments were extracted using a Qiaquick Gel Extraction kit (Qiagen, Valencia, CA), of RNAse inhibitor, and 1μl of T7 RNA polymerase and incubated for 1.5 hours (37°C). Labeled probes were then purified using Micro Bio-Spin 6 Chromatography Column (BioRad, Berkeley, CA), and 1 µl of the probe was counted for subsequent radioactivity dilution calculations. Four to six labelings were used to reach the necessary volume and optimal radioactivity (1-2 million counts per minute/ slide). An additional 1μl of 1M DTT was also added to the labeled mRNA after purification, allowed to incubate at room temperature for 15 min, and stored at -20°C until further use.

In situ hybridization
The radioactive probe was diluted in hybridization buffer (50% formamide, 10% dextran sulfate, 3X saline-sodium citrate buffer, 50 mM sodium phosphate buffer, 1X Denhardt's solution, 0.1 mg/ml yeast tRNA, and 10 mM DTT) and the volume calculated based on the initial count in order to obtain roughly 1-2 x 10 6 radioactivity counts per 75 μL of the diluted probe in hybridization buffer. Slide-mounted brain tissue, -1.08 to -7.08 mm from Bregma, was fixed in 4% paraformaldehyde solution, washed in 2X saline-sodium citrate buffer (SSC), and incubated with 0.1 M triethanolamine (TEA) with 0.25% acetic anhydride. Slides were then dehydrated using ascending ethanol concentrations and air-dried for 1 hour. Hybridization buffer was warmed (37°C) and mixed with the calculated quantity of probe and 1 M DTT (~1% total HB volume). 75 µl of the diluted probe was then applied to coverslips, which were subsequently placed onto the tissue. Slides were then placed in humidity-maintained hybridization chambers soaked with formamide and incubated overnight (~16 hrs) at 55°C. The next day, coverslips were removed, and the slides were rinsed with 2X SSC. Slides were then incubated in RNase A solution (100 μg/mL RNase in Tris buffer with 0.5M NaCl, (37°C), washed in descending concentrations of SSC (2X, 1X, 0.5X), and incubated in 0.1X SSC (65°C). Next, sections were briefly rinsed in H2O, dehydrated using ascending ethanol concentrations, and air-dried for 1 hr.
Slides were then loaded into film cassettes and exposed in a dark room to 35 x 43cm Kodak BioMax MR film (Carestream Health Inc, Rochester, NY, USA) for seven weeks. Extra slides using spare experimental tissue were run concurrently to confirm optimal exposure time. The specificity of the probe was verified using sense strand controls similar to previous studies (Garcia-Fuster et al., 2012;Kabbaj et al., 2000).

Quantification
Films were developed using Microtek ScanMaker 1000XL (Fontana, CA, USA) and digitally scanned using SilverFast Lasersoft Imaging software (Sarasota, FL, USA). Signal expression was quantified using ImageJ (National Institutes of Health, Bethesda, MD), a computer-assisted optical densitometry software. The brush selection tool (size: 15 pixels) was used to trace the subregions of interest (CA1, CA2, CA3, and dentate gyrus) throughout the dorsal hippocampus (-2.64 to -4.56 mm from Bregma) and ventral hippocampus (-4.68 to -6.72 mm from Bregma), using the Rat Brain Atlas (Paxinos and Watson, 2007) for guidance. Area (total number of pixels), optical density (darkness of pixels) and integrated optical density (intensity and spread) measurements of the region of interest were taken using a macro that automatically enabled signal above background (3.5 x standard deviation) to be determined. The area (unit, 63 pixels/ 1 mm) and optical density (darkness) were calculated for each of the four hippocampal subregions across a range of 11-21 sections per rat that spanned the dorsal-ventral gradient of the hippocampus. A single value was calculated for each of the hippocampal subregions per rat, by averaging the values of both hemispheres across multiple sections. Further, given that the dorsal and ventral hippocampus are viewed as neuroanatomically and functionally distinct (see Fanselow & Dong, 2010), with the dorsal hippocampus considered to be more involved in cognitive function and the ventral hippocampus in stress and emotion, a single average value was used for dorsal vs. ventral subregions and data were graphed and analyzed separately (similar to Romeo et al., 2008). Sections with damaged tissue or unusual signal (e.g., dark artifacts) that distorted the region of interest were omitted from analyses. During quantification, the experimenter was blind to the phenotypes of each subject.

Experiments 1 and 2: Statistical analysis
Behavioral outcome measures (i.e., PavCA, EPM, OFT), plasma corticosterone concentrations, and in situ hybridization measures (i.e., area and optical density) were analyzed using the Statistical Package for the Social Sciences (SPSS) program version 24.0 (IBM, Armok, NY, USA). Linear mixed-effects models were performed for PavCA behavior and neuroendocrine measures (corticosterone and GR mRNA levels), using the best fit covariance structure with the lowest Akaike's information criterion for each set of data. Univariate analysis of variance was performed for EPM and OFT behavior and tested for normality using the Shapiro-Wilk test. When dependent variables failed to meet normality, log 10 or square root transformations were conducted, or a Kruskal-Wallis nonparametric test was performed (using StatView, version 5.0, SAS Institute Inc., Cary, NC, USA). Pearson correlations were performed to determine if there was a significant correlation between baseline CORT levels (pre-vs. post-PavCA) and baseline CORT levels and PavCA behavior. Statistical significance was set at p<0.05, and Bonferroni post-hoc analyses were conducted when significant interactions were detected. All figures were made using GraphPad Prism 7.

PavCA behavior
The following lever-directed (sign-tracking) and food cup-directed (goal-tracking) behaviors were assessed across five consecutive PavCA training sessions and compared between GTs (n=11), IRs (n=17), and STs (n=32): the probability to approach, the number of contacts, and the latency to approach the lever or food-cup during the presentation of the lever-CS ( Figure   2). Main Effects of Phenotype, Session, and Phenotype x Session interactions for all behavioral measures are reported in Table 1 (top). There was a significant Effect of Phenotype and Session for all behavioral measures. As expected, STs showed a significantly greater probability to approach the lever (Figure 2A), a greater number of lever contacts ( Figure 2C), and shorter latency to approach the lever ( Figure 2E), relative to IRs and GTs. These differences in leverdirected behaviors were apparent by the 2 nd PavCA training session (Figure 2, also Table 2 (top   left)). In contrast, relative to STs and IRs, GTs showed a significantly greater probability of approaching the food-cup ( Figure 2B), a greater number of food-cup entries (Figure 2D), and a shorter latency to enter the food-cup ( Figure 2F). These differences in food cup-directed behavior became apparent by the 3 rd PavCA training session (Figure 2, also Table 2 (top right)).

CORT levels
Overall, pre-and post-PavCA baseline plasma CORT levels did not significantly differ These data are in agreement with prior studies Tomie et al., 2000), demonstrating that pre-PavCA baseline plasma CORT levels do not significantly differ between Phenotypes, and extend these findings to show that baseline plasma CORT levels also do not differ between Phenotypes after the development of a conditioned response.

Correlations
To further investigate the relationship between baseline CORT levels and cue-motivated

Open field test
There were no significant differences between Phenotypes in their behavior on the OFT.  Figure 6B).

PavCA behavior
Similar to Experiment 1, there were significant Effects of Phenotype, Session, and Phenotype x Session interactions for all behavioral measures reported in Table 1 (bottom) (data are not shown in graphical format). Differences between Phenotypes were apparent for leverand food cup-directed behavior as early as the first PavCA training session (See Table 2 (bottom)).

GR mRNA expression
Dorsal hippocampus -There were no significant differences between Phenotypes in GR mRNA  Figure 7C). Given the anatomical variability in size between subregions (see schematic Figure 7A), significant differences in area

Discussion
The present studies allowed us to determine whether differences in behavioral and neuroendocrine measures of negative valence are included amongst the co-existing traits associated with the propensity to attribute incentive salience to reward cues. We report three main findings. First, across phenotypes, there is a general increase in baseline plasma CORT levels over the course of associative cue-learning, and this rise is positively correlated with the development of the sign-tracking response. Second, behavioral and CORT response to environmental challenges reflective of negative valence do not differ between goal-trackers and sign-trackers. Third, sign-trackers have greater expression of GR mRNA in the ventral hippocampus relative to goal-trackers and intermediate responders.
The basis of our understanding of CORT function in appetitive Pavlovian conditioning stems from the work of Tomie et al. (2000) who demonstrated that cue-food associations elicit an increase in plasma CORT. Subsequently, it was shown that, relative to GTs, STs exhibit a greater rise in plasma CORT following a single Pavlovian conditioning session; that is, before the development of a conditioned response . Further, prior to PavCA training, baseline CORT is similar across GTs, IRs, and STs ; and, based on work from Tomie et al. (2000), we would not expect there to be differences between phenotypes in baseline CORT later in training. To date, however, baseline plasma CORT had not been systematically assessed in the same rat to determine if CORT profiles change as a consequence of cue-learning. Thus, in Experiment 1A we compared, within the same rat, baseline CORT concentrations at a "naïve" state of learning (pre-PavCA) and once a conditioned response had been acquired (post-PavCA). There was a significant rise in baseline plasma CORT levels with the development of a conditioned response, and pre-PavCA CORT levels were predictive of post-PavCA CORT levels. However, CORT levels were not dependent on the innate cuelearning strategy that was employed, as CORT did not significantly differ between phenotypes.
Nonetheless, to determine whether CORT serves as a predictor of cue-learning, we assessed the relationship between baseline CORT profiles and behavior. Although baseline CORT levels at either time point did not significantly correlate with behavior, we found that the change in behavior, or the rate of learning between sessions 1 and 5 of PavCA training significantly correlated with baseline CORT levels once a conditioned response had been acquired. That is, there was a positive relationship between the degree to which sign-tracking is learned and post-PavCA CORT levels. Thus, while the relationship between baseline CORT and Pavlovian conditioned approach behavior is not apparent on a given day of training (early or late), the trajectory of CS-directed behavior (i.e., captured by ∆ PavCA index) is linked to CORT. This notion is supported by previous findings that demonstrated phenotype-dependent responses in CORT immediately following the first Pavlovian conditioning session ; and those by Tomie et al. (2000) that showed a positive relationship between the acquisition of Pavlovian conditioned approach behavior (average behavior of sessions 1-10) and the CORT response immediately following a later training session. While the relationship between CORT and Pavlovian conditioned approach behavior is complex, the current data are indicative of an association between the propensity to attribute incentive salience to a food-cue, or the emergence of a sign-tracking conditioned response, and baseline CORT levels after that conditioned response is established.
One of the primary roles of CORT is to act across the body and brain to broadly mediate the stress response (Herman et al., 2016). Thus, we wanted to determine whether differences in plasma CORT are present in goal-trackers vs. sign-trackers in contexts outside of Pavlovian conditioning and, explicitly, in response to paradigms reflective of negative valence. Experiment 1B showed no differences between phenotypes in CORT response to an open field test or physiological restraint. Further, goal-trackers and sign-trackers did not differ in their behavioral response to the open field test or elevated plus maze. These findings are consistent with those previously reported by Vanhille et al. (2015), who showed no pre-existing differences in behavior on the elevated plus maze test in rats that were later characterized as sign-trackers or goal-trackers; and those reported by Harb and Almeida (2014) who showed no differences in behavior on the open field test in mice characterized as sign-trackers or goal-trackers. In contrast to the present findings, however, Harb and Almeida (2014) did report a significant difference in peak CORT response following an acute stressor, with sign-tracker mice exhibiting a greater peak relative to goal-trackers or intermediate responders. These discrepant findings are likely due to differences in experimental procedures, including the species used and the nature and intensity of the stressor. In this regard, we note that the repeated testing implemented in the current study may have affected the CORT response in a manner that precluded observable differences (e.g. see Dallman et al., 2004). Indeed, it is possible that differences in the CORT profile in response to physiological restraint were not apparent because of a ceiling effect, as both baseline and peak CORT levels were high across all animals. Nonetheless, given that goaltrackers and sign-trackers behaved similarly on tests of negative valence, and showed no significant differences in CORT response to the open field test, we conclude that individual differences in the negative valence system are not captured by the goal-tracker/sign-tracker animal model.
The RDoC includes fear conditioning and the associated freezing response within the domain of negative valence, specifically within the construct of acute threat or "fear". Thus, it should be noted that goal-trackers and sign-trackers differ in their response to Pavlovian fear conditioning (Morrow et al., 2011;Morrow et al., 2015). Specifically, relative to goal-trackers, sign-trackers are more fearful of discrete cues that predict footshock (Morrow et al., 2011), and show exaggerated incubation of their fear response (Morrow et al., 2015). Importantly, however, goal-trackers exhibit greater contextual fear when placed back into a fear-conditioning context in the absence of discrete cues (Morrow et al., 2011). Thus, these differences seem to be specific to learning the value of the discrete cue, rather than differences in negative valence per se. While others have shown that CORT plays a critical role in fear conditioning, beyond the "stress" component (e.g., Marchand et al., 2007;Zorawski & Killcross, 2002), this has yet to be assessed within the context of the goal-tracker/sign-tracker animal model and will be the focus of future investigations.
The current findings and those of others (Harb & Almeida, 2014;Vanhille et al., 2015) demonstrate that goal-trackers and sign-trackers respond similarly to tests of negative valence.
Yet, exposure to "stress" has been shown to alter the propensity to attribute incentive salience to reward cues (Fitzpatrick et al., 2019;Hynes et al., 2018;Lomanowska et al., 2011). Rats exposed to stress early in life exhibit greater sign-tracking behavior in adulthood (Hynes et al., 2018;Lomanowska et al., 2011). In contrast, adult rats exposed to a single prolonged stressor show an attenuation of sign-tracking behavior (Fitzpatrick et al., 2019). Thus, the impact of "stress" on the propensity to sign-track appears to be dependent on the type of stressor and timing of exposure. In light of the current findings, we postulate that the neural processes underlying these reported stress-induced effects (Fitzpatrick et al., 2019;Hynes et al., 2018;Lomanowska et al., 2011) go beyond CORT and the HPA axis, and include components of the cortico-thalamicstriatal "motive" circuit, which is differentially engaged in sign-trackers vs. goal-trackers (Flagel, Cameron, et al., 2011; see also Kuhn et al., 2018).
One potential neural interface between "stress" and "motive" circuitry is the hippocampus (e.g., Maccari et al., 1991, for review see Barr et al., 2017). Glucocorticoid receptors are densely expressed within the hippocampus (Reul & de Kloet, 1985) and CORT-GR interactions within this brain region play a critical role in the negative feedback system that acts to maintain homeostatic levels of CORT in the face of physiological or environmental challenges (Herman et al., 2012). Specifically, greater GR mRNA expression in the hippocampus has been associated with more rapid negative feedback, or return to baseline CORT levels (Liberzon et al., 1999;Meaney et al., 1996). In the current study, we did not observe phenotypic differences in circulating CORT levels either under baseline conditions or in response to environmental challenges (i.e., open field test or physiological restraint). Yet, we did find that, relative to goal-trackers and intermediate responders, sign-trackers have significantly greater expression of GR mRNA in the ventral hippocampus. The fact that these differences in GR expression are specific to the ventral hippocampus and not apparent in the dorsal hippocampus may explain why we did not observe differences in circulating levels of CORT. Indeed, while the dorsal hippocampus has been shown to play a role in stress-induced negative feedback (Feldman & Weidenfeld, 1993, the ventral hippocampus has been reported to regulate tonic levels of CORT (Herman et al., 1992). Other findings suggest that the engagement of the ventral vs.
dorsal hippocampus is dependent on the type of stressor (e.g., Dorey et al., 2012;Herman et al., 2005;Maggio & Segal, 2007). Thus, additional studies are warranted to further investigate the role of GR expression in the ventral hippocampus within the context of the stress response and negative feedback regulation. Nonetheless, we propose that the phenotypic differences reported here in GR expression in the ventral hippocampus are directly related to motivated behavior and reward learning, rather than stress regulation and negative valence. In support, lesions of the ventral, but not the dorsal, hippocampus decrease the propensity to sign-track (Fitzpatrick et al., 2016). While it is remains to be determined whether these lesion effects are dependent on GR function, it should be noted that systemic administration of a GR antagonist similarly attenuates the acquisition of sign-tracking behavior (Rice et al., 2018;Rice et al., 2019). Taken together, GR function, and presumably that within the ventral hippocampus, appears to play an important role in incentive motivational processes.
In conclusion, these findings establish that the neurobehavioral endophenotype associated with the propensity to sign-track does not encompass differences in negative valence systems.
Further, we provide additional evidence that glucocorticoids, which have primarily been implicated in negative valence (but see Deroche et al., 1993;Piazza et al., 1993) are, in fact, involved in positive valence. Specifically, circulating levels of CORT appear to be linked to the propensity to attribute incentive salience to reward cues, or sign-tracking behavior. In addition, expression of glucocorticoid receptors in the ventral hippocampus appear to be related to inherent cue-reward learning strategies rather than the stress response. These studies provide a critical foundation for future work to further examine the mechanism by which glucocorticoids interact with other neural systems known to play a role in incentive motivation (for further discussion see Lopez & Flagel, 2020).        goal-tracking (i.e., food-cup-directed) behaviors. Effect of Phenotype, Session, and Phenotype x Session interactions were analyzed for Experiment 1 (top) and Experiment 2 (bottom).