Variations of Brain Functional Connectivity in Alcohol-Preferring and Non-Preferring Rats with Consecutive Alcohol Training or Acute Alcohol Administration

Alcohol addiction is regarded as a series of dynamic changes to neural circuitries. A comparison of the global network during different stages of alcohol addiction could provide an efficient way to understand the neurobiological basis of addiction. Two animal models (P-rats screened from an alcohol preference family, and NP-rats screened from an alcohol non-preference family) were trained for alcohol preference with a two-bottle free choice method for 4 weeks. To examine the changes in the neural response to alcohol during the development of alcohol preference and acute stimulation, different trials were studied with resting-state fMRI methods during different periods of alcohol preference. The correlation coefficients of 28 regions in the whole brain were calculated, and the results were compared for alcohol preference related to the genetic background/training association. The variety of coherence patterns was highly related to the state and development of alcohol preference. We observed significant special brain connectivity changes during alcohol preference in P-rats. The comparison between the P- and NP-rats highlighted the role of genetic background in alcohol preference. The results of this study support the alterations of the neural network connection during the formation of alcohol preference and confirm that alcohol preference is highly related to the genetic background. This study could provide an effective approach for understanding the neurobiological basis of alcohol addiction.


Introduction
As one of the most pervasive mental disorders or endogenous depression, alcoholism affects approximately 10% of Americans at some time in their lives [1]. First, drinking (Prof. Andrew John Lawrence) with permission from the School of Medicine, Indiana University. All the animals were bred in-house and acclimated to single housing, kept under a standard 12 h light-dark cycle, and given food ad libitum at an optimal environmental temperature (23~25 • C). Male P-rats (n = 12; Pre-: n = 12, Mid-: n = 9, Post-: n = 8 for fMRI analysis; two rats died during the fMRI data acquisition in the Mid-period, and three imaging data (Mid-: n = 1; Post-: n = 2) were discarded due to uncorrectable motion and noise) and male NP-rats (n = 9) (200-250 g at the time of training) were used for fMRI experiments.

Alcohol Preference Training
Single-housed P-and NP-rats (male,~200 g) had continuous access to one bottle of 5% alcohol (v/v) and one bottle of water in their home cage for four consecutive weeks according to a two-bottle free choice (2-BC) paradigm [23,24]. The mass of drinking liquid (5% alcohol and water) was monitored every day, and the bottles were randomly replaced every day. If the alcohol consumption ratio was higher than 80% (P-rats) for three consecutive days, then that animal was treated as successfully trained for alcohol preference. However, if this ratio was lower than 20% (NP-rats), then the animal was treated as alcohol-non-preferring.

Resting-State fMRI Experiments
At first, the whole-brain level MRI scans of P-and NP-rats were designed in three different periods (P-Pre: before alcohol training, P-Mid: 2 weeks after alcohol training; P-Post: 4 weeks after alcohol training) and two different states (RS: resting state; IS: intoxicated state/acute alcohol treatment). The MRI experiments were conducted using a Bruker Biospec 70/20 USR small animal MR system (Bruker BioSpin MRI, Ettlingen, Germany) operating at 7.0 T. The complete steps for the experimental procedure were as follows.
All the animals were initially anesthetized with isoflurane (3.0-5.0% for induction and set-up on the animal bed). They were kept in an MRI scanner breathing isoflurane (0.6-1.2% for light anesthesia) in oxygen-enriched air (20% O 2 and air). Breathing rates, heart rates, and blood oxygen saturation levels were monitored using a pulse oximeter positioned at the hind limb and a pressure-sensitive sensor under the abdomen (MR-compatible Small Animal Monitoring & Gating System, SA Instruments, Inc., New York, NY, USA). The body temperature of each rat was maintained at~37 • C using a warm water circuitry system. A needle connected to a PE-50 tube was implanted in the abdomen for the administration of the alcohol solution (0.76 g/kg i.p.) from outside the magnet. A planar receive coil surface coil with a diameter of 20 mm was placed on top of the skull and utilized in combination with a detunable partial volume transmit coil (Bruker BioSpin MRI, Ettlingen, Germany).
The experiment of the intoxicated state (IS) was defined as the period of 15 min~1 h after the alcohol injection (0.76 g/kg i.p.). The functional scans (RS and IS) were acquired per rat when the subject was under light anesthesia and immobility (estimated by the respiratory rate stabilized at 80-100 per minute).

Resting-State fMRI Analysis Data Analysis
The post-processing of all the functional images was performed by a single experienced observer who was unaware of to whom the scans belonged. The preprocessing and data analysis were performed using a homemade toolbox with MATLAB code (spmratIHEP) [25] in the statistical parametric mapping (SPM8) software (http: //www.fil.ion.ucl.ac.uk/spm (accessed on 8 April 2020), which comprised an fMRI rat brain template [19] and an atlas in the Paxinos and Watson space.
To calculate the whole-brain network of rsfMRI, the files of Bruker MR image data were first converted to the imaging analyzing format with the Micro Bruker2Anlyzer Converter (Bru2Anz) software. Then, the functional datasets of all the individuals were preprocessed in spmratIHEP using the following steps: (1) Data screening: The first 10 volumes of each scan were discarded to allow for magnetization equilibrium. (2) Slice timing: The differences in slice acquisition times of each scan were corrected using the slice timing approach. (3) Realignment: The temporal processed volumes of each subject were realigned to the first volume to remove the head motion, and a mean image was created over the 310 realigned volumes. All the participants had less than 1 mm of translation in the x-, y-, or z-axis and 1 • of rotation in each axis. (4) Spatial normalization: The realigned volumes were spatially standardized into the Paxinos and Watson space by normalizing with the EPI template of rat brain via their corresponding mean image. Then, all the normalized images were resliced by 1.0 × 1.5 × 1.0 mm 3 voxels. (5) Smooth: The normalized functional series were smoothed with a Gaussian kernel of 2 × 4 × 2 mm 3 FWHM (full width at half-maximum). Using DPARSF (http://rfmri.org/DPARSF (accessed on 8 June 2020), all smoothed images were then bandpass-filtered at 0.01-0.08 Hz and further corrected for the effect of head movement by regressing the translations and rotations of the head estimated during image realignment.
Considering the resolution of the fMRI data and brain regions related with alcohol preferences, the whole rat brain was automatically divided into 28 different regions of interest (ROIs) across both hemispheres based on physiological structures and functions using the brain template [19]. The functional connectivity was evaluated using seed-based correlational analyses with a similar former method [26] (Figure 1). The time courses from all the voxels within the individual seed regions were averaged and used as reference time courses. Pearson's cross-correlation coefficients among the reference time courses, as well as the time course of each voxel, were then calculated and used to quantify the strength of the functional connectivity. Finally, for each rat, a functional connectivity matrix was established.
n Sci. 2021, 11, x FOR PEER REVIEW 5 o Figure 1. Illustration of the 28 selected brain regions in the structure of the rat brain.

Statistical Analysis
The statistical analyses of the functional connectivity data of all the animals in diff ent groups (different states: RS and IS; different periods: P-Pre, P-Mid, and P-Post) w performed with MATLAB (MathWorks, Inc., Natick, MA, USA). Pearson's cross-corre tion coefficients were used to analyze the difference of the whole-brain networks amo the different groups. Student's t-test and ANOVA with a post-hoc test (LSD) were us to compare the two states and three periods. With equal variance not assumed, a nonp ametric test (Dunnett T3) was used to compare the differences between the groups. The values were adjusted for multiple testing with the "fdr" or its alias "BH" method. T results in the text are displayed as the means ± SD. In general, p-value of 0.05 was cons ered statistically significant, * p < 0.05; ** p < 0.01; **** p < 0.001.

Detection of Alcohol Preference Using Two-Bottle Free Choice
With training using the two-bottle free choice method, the P-and NP-rats show significantly different patterns of alcohol preference drinking ( Figure 2). For the NP-ra the alcohol drinking ratio was < 20% throughout the whole 4-week training period, wh was mostly caused by leaks from the bottles or limited alcohol drinking. However, for P-rats, the alcohol drinking ratio gradually increased following the training periods a ultimately reached > 80%, which was regarded as the development of alcohol preferen behavior. According to the drinking rate curve, three different periods were defined fr primary alcohol drinking to alcohol preference: P-Pre, P-Mid (2-week alcohol trainin and P-Post (4-week alcohol training).

Statistical Analysis
The statistical analyses of the functional connectivity data of all the animals in different groups (different states: RS and IS; different periods: P-Pre, P-Mid, and P-Post) were performed with MATLAB (MathWorks, Inc., Natick, MA, USA). Pearson's cross-correlation coefficients were used to analyze the difference of the whole-brain networks among the different groups. Student's t-test and ANOVA with a post-hoc test (LSD) were used to compare the two states and three periods. With equal variance not assumed, a nonparametric test (Dunnett T3) was used to compare the differences between the groups. The p-values were adjusted for multiple testing with the "fdr" or its alias "BH" method. The results in the text are displayed as the means ± SD. In general, p-value of 0.05 was considered statistically significant, * p < 0.05; ** p < 0.01; **** p < 0.001.

Detection of Alcohol Preference Using Two-Bottle Free Choice
With training using the two-bottle free choice method, the P-and NP-rats showed significantly different patterns of alcohol preference drinking ( Figure 2). For the NP-rats, the alcohol drinking ratio was <20% throughout the whole 4-week training period, which was mostly caused by leaks from the bottles or limited alcohol drinking. However, for the P-rats, the alcohol drinking ratio gradually increased following the training periods and ultimately reached >80%, which was regarded as the development of alcohol preference behavior. According to the drinking rate curve, three different periods were defined from primary alcohol drinking to alcohol preference: P-Pre, P-Mid (2-week alcohol training), and P-Post (4-week alcohol training). Brain Sci. 2021, 11, x FOR PEER REVIEW 6 of 16 Figure 2. Results of the two-bottle free choice of the alcohol drinking training for the alcohol-preferring rats (P-rats) and the alcohol-non-preferring rats (NP-rats).

Whole-Brain Network Analysis of the Resting State
To investigate the relationship of the brain network and different periods of alcohol preference (P-Pre, P-Mid, and P-Post) / the effects of alcohol stimulation (RS and IS), a series of EPI scans under different treatment periods and alcohol acute stimulation was performed ( Figure 3). The Pearson's cross-correlation coefficients across these different brain-wide ROIs were calculated, and the mean functional connectivity matrices of the P-and NP-rats between different periods and stages are illustrated in Figure 3A-D. From the global perspective, the patterns of the whole-brain connection showed massive agreements between different treatment periods and different stimulation states. The correlation coefficients of the functional connectivity ranged from ~0.84 to 0.96 (Table 1).

Figure 2.
Results of the two-bottle free choice of the alcohol drinking training for the alcoholpreferring rats (P-rats) and the alcohol-non-preferring rats (NP-rats).

Whole-Brain Network Analysis of the Resting State
To investigate the relationship of the brain network and different periods of alcohol preference (P-Pre, P-Mid, and P-Post) / the effects of alcohol stimulation (RS and IS), a series of EPI scans under different treatment periods and alcohol acute stimulation was performed ( Figure 3).
Brain Sci. 2021, 11, x FOR PEER REVIEW 6 of 16 Figure 2. Results of the two-bottle free choice of the alcohol drinking training for the alcohol-preferring rats (P-rats) and the alcohol-non-preferring rats (NP-rats).

Whole-Brain Network Analysis of the Resting State
To investigate the relationship of the brain network and different periods of alcohol preference (P-Pre, P-Mid, and P-Post) / the effects of alcohol stimulation (RS and IS), a series of EPI scans under different treatment periods and alcohol acute stimulation was performed ( Figure 3). The Pearson's cross-correlation coefficients across these different brain-wide ROIs were calculated, and the mean functional connectivity matrices of the P-and NP-rats between different periods and stages are illustrated in Figure 3A-D. From the global perspective, the patterns of the whole-brain connection showed massive agreements between different treatment periods and different stimulation states. The correlation coefficients of the functional connectivity ranged from ~0.84 to 0.96 (Table 1). The Pearson's cross-correlation coefficients across these different brain-wide ROIs were calculated, and the mean functional connectivity matrices of the P-and NP-rats between different periods and stages are illustrated in Figure 3A-D. From the global perspective, the patterns of the whole-brain connection showed massive agreements between different treatment periods and different stimulation states. The correlation coefficients of the functional connectivity ranged from~0.84 to 0.96 (Table 1). Table 1. The correlation coefficients of the whole-brain functional connectivity between different rat strains (alcohol preferring and alcohol non-preferring) and conditions (0-week, 2-week, and 4-week alcohol drinking training). To clearly show the effects of alcohol preference, two different kinds of animals (P-rats and NP-rats) under the state of RS were compared (R = 0.841, p < 0.005, Figure 4A), and another example from P-rats under different states (RS and IS) in the P-Pre group was also provided for comparison (R = 0.960, p < 0.005, Figure 4B). Among these coefficients, the different periods of alcohol preference and alcohol stimulation ranged from 0.910 to 0.960, which is much higher than the comparison between the NP-rats and these groups (from 0.841 to 0.878, p < 0.005).

RS, P-
alcohol drinking training). To clearly show the effects of alcohol preference, two different kinds of animals rats and NP-rats) under the state of RS were compared (R = 0.841, p < 0.005, Figure 4 and another example from P-rats under different states (RS and IS) in the P-Pre group also provided for comparison (R = 0.960, p < 0.005, Figure 4B). Among these coefficie the different periods of alcohol preference and alcohol stimulation ranged from 0.91 0.960, which is much higher than the comparison between the NP-rats and these gro (from 0.841 to 0.878, p < 0.005). To demonstrate the reliability of the coefficients and compare the connectivity of different brain regions clearly, the connectivity properties of several alcohol-preferen associated brain regions (habenula nucleus; accumbens nucleus; olfactory tubercle; p limbic cortex; piriform cortex; amygdala; septum; ventral tegmental area; visual cor with a well-known projection relationship were extracted and are shown in Figure 5. instance, relatively high correlation coefficients were found in the accumbens nucleus To demonstrate the reliability of the coefficients and compare the connectivity of the different brain regions clearly, the connectivity properties of several alcohol-preferenceassociated brain regions (habenula nucleus; accumbens nucleus; olfactory tubercle; prelimbic cortex; piriform cortex; amygdala; septum; ventral tegmental area; visual cortex) with a well-known projection relationship were extracted and are shown in Figure 5. For instance, relatively high correlation coefficients were found in the accumbens nucleus and hypothalamus (2), piriform cortex and entorhinal cortex (7), and ventral tegmental area and substantia nigra (16). hypothalamus (2), piriform cortex and entorhinal cortex (7), and ventral tegmental area and substantia nigra (16).

Variations of Functional Connectivity with Alcohol Stimulation and Alcohol Preference
To investigate the relationship between brain connectivity and alcohol stimulation or alcohol preference, the whole-brain connectivity of different brain regions (756 pairs in each state and period) under different states (periods and stimulation conditions) was compared.
At first, the changes of the brain functional connections under ethanol stimulation in different periods for P-rats (pre-preference and post-preference) were calculated and compared. With the stimulation of ethanol injection, the connectivity of 11 paired regions of the pre-preference periods and nine paired regions of the post-preference period had significant changes (p < 0.05, Table S1). The tendency of the changes in the brain connections of these regions was collected ( Figure 6A). There was a clear opposite tendency between different preference periods: mostly increasing (8/11) in the pre-preference period and decreasing (8/9) in the post-preference period. Although the effects of alcohol stimulation on the fMRI signals were complex (blood flow, blood volume, anesthetic effects of alcohol, etc.), we can see from the changes in the trend of the periods that the effects varied with the process of preference formation.

Variations of Functional Connectivity with Alcohol Stimulation and Alcohol Preference
To investigate the relationship between brain connectivity and alcohol stimulation or alcohol preference, the whole-brain connectivity of different brain regions (756 pairs in each state and period) under different states (periods and stimulation conditions) was compared.
At first, the changes of the brain functional connections under ethanol stimulation in different periods for P-rats (pre-preference and post-preference) were calculated and compared. With the stimulation of ethanol injection, the connectivity of 11 paired regions of the pre-preference periods and nine paired regions of the post-preference period had significant changes (p < 0.05, Table S1). The tendency of the changes in the brain connections of these regions was collected ( Figure 6A). There was a clear opposite tendency between different preference periods: mostly increasing (8/11) in the pre-preference period and decreasing (8/9) in the post-preference period. Although the effects of alcohol stimulation on the fMRI signals were complex (blood flow, blood volume, anesthetic effects of alcohol, etc.), we can see from the changes in the trend of the periods that the effects varied with the process of preference formation. Brain Sci. 2021, 11, x FOR PEER REVIEW 9 of 16

Different Patterns of Functional Connectivity Related to Various Alcohol Conditions
The ANOVA of the correlation coefficient revealed that there were significant effects of alcohol training on the connections between several brain regions. The whole comparison results are provided in Table S3. Among the different changes, there were three different variation patterns for functional connectivity. First, three pair regions (NAc-IC, OT-Hypo, DB-VLPO) showed significant increases in the P-Post period before the alcohol injection (RS) compared to the P-Pre period (NAc-IC: Pre-vs. Post-, p < 0.001, OT-Hypo: Prevs. Post-, p < 0.05, DB-VLPO: Pre-vs. Post-, p < 0.05) but decreased after alcohol injection stimulation (IS) (NAc-IC: Pre-vs. Post-, p < 0.05, OT-Hypo: Pre-vs. Post-, p < 0.05, DB-VLPO: Pre-vs. Post-, p < 0.05) ( Figure 6B). Second, there was no significant difference between the P-Pre and P-Post periods. However, the correlation coefficients were significantly different from the P-Mid and P-Pre/P-Post periods. The connection between the SNC (compacta of substantia nigra)-Tha (RS) and SNC-VTA (IS) was only observed to decrease and increase, respectively, in the P-Mid period (p < 0.05) ( Figure 6C). Thus, the region of the SNC should be related to the formation of alcohol preference. Lastly, changes of the functional connectivity were only observed after alcohol stimulation but were not

Different Patterns of Functional Connectivity Related to Various Alcohol Conditions
The ANOVA of the correlation coefficient revealed that there were significant effects of alcohol training on the connections between several brain regions. The whole comparison results are provided in Table S3. Among the different changes, there were three different variation patterns for functional connectivity. First, three pair regions (NAc-IC, OT-Hypo, DB-VLPO) showed significant increases in the P-Post period before the alcohol injection (RS) compared to the P-Pre period (NAc-IC: Pre-vs. Post-, p < 0.001, OT-Hypo: Pre-vs. Post-, p < 0.05, DB-VLPO: Pre-vs. Post-, p < 0.05) but decreased after alcohol injection stimulation (IS) (NAc-IC: Pre-vs. Post-, p < 0.05, OT-Hypo: Pre-vs. Post-, p < 0.05, DB-VLPO: Prevs. Post-, p < 0.05) ( Figure 6B). Second, there was no significant difference between the P-Pre and P-Post periods. However, the correlation coefficients were significantly different from the P-Mid and P-Pre/P-Post periods. The connection between the SNC (compacta of substantia nigra)-Tha (RS) and SNC-VTA (IS) was only observed to decrease and increase, respectively, in the P-Mid period (p < 0.05) ( Figure 6C). Thus, the region of the SNC should be related to the formation of alcohol preference. Lastly, changes of the functional connectivity were only observed after alcohol stimulation but were not observed in the resting state. The coefficient of the NAc-Hypo (IS) in the P-Post state increased significantly (p < 0.05) compared with the P-Pre state, while the coherence of the Au-Hipp (IS) decreased significantly from P-Pre/P-Mid to P-Post (p < 0.01~0.05) ( Figure 6D).

Functional Connectivity Related to Different Alcohol Preferences
Among the changes of the functional connectivity related to alcohol stimulation or alcohol preference periods, there was a significant difference between the alcohol-preferring rats and the alcohol-non-preferring rats. After the acute alcohol stimulation, the functional connectivity of the NAC-IC, OT-Hypo, DB-VLPO, and VTA-Hypo was observed in the alcohol-preferring rats, especially for the first three pairs (p < 0.001, Figure 7B). However, there was no difference among these pair regions in the alcohol-non-preferring rats ( Figure 7A). In addition, the opposite changes occurred with acute alcohol stimulation in the connectivity of the NAC-IC and OT-Hypo between different alcohol preference periods.
Brain Sci. 2021, 11, x FOR PEER REVIEW 10 of 16 observed in the resting state. The coefficient of the NAc-Hypo (IS) in the P-Post state increased significantly (p < 0.05) compared with the P-Pre state, while the coherence of the Au-Hipp (IS) decreased significantly from P-Pre/P-Mid to P-Post (p < 0.01~0.05) ( Figure  6D).

Functional Connectivity Related to Different Alcohol Preferences
Among the changes of the functional connectivity related to alcohol stimulation or alcohol preference periods, there was a significant difference between the alcohol-preferring rats and the alcohol-non-preferring rats. After the acute alcohol stimulation, the functional connectivity of the NAC-IC, OT-Hypo, DB-VLPO, and VTA-Hypo was observed in the alcohol-preferring rats, especially for the first three pairs (p < 0.001, Figure 7B). However, there was no difference among these pair regions in the alcohol-non-preferring rats ( Figure 7A). In addition, the opposite changes occurred with acute alcohol stimulation in the connectivity of the NAC-IC and OT-Hypo between different alcohol preference periods. Figure 7. Comparison of the functional connectivity between the NP-rats (A) and P-rats (B). The black arrow represents an upward trend, while the red one is a downward trend. The abbreviations of brain areas are the same as in the main text: NAc (nucleus of accumbens), OT (olfactory tubercle), VTA (ventral tegmental area), IC (insular cortex), Hypo (hypothalamus), DB (diagonal band), VLPO (ventrolateral preoptic nucleus); * p < 0.05, *** p < 0.005, **** p < 0.001.

Discussion
In this study, brain-wide rsfMRI was used to investigate the changes of the functional connectivity caused by different animal strains (P-and NP-rats) and the different periods during the development of alcohol preferences for P-rats. The results of the variation of the functional network might be related to the development of alcohol preference caused by persistent alcohol drinking or the related genetic background. To our knowledge, this is the first time that the dynamic changes of the whole-brain-wide functional connections during the development of alcohol preference (from the first alcohol drinking to alcohol preference or addiction) have been described. To summarize, the following regions were mainly involved: NAc, Tu, VTA, IC, Hypo, SNC, Au, DB, Hipp, and VLPO. These regions are highly related to the behaviors of reward, motivation, emotion, moods, and cognitive Figure 7. Comparison of the functional connectivity between the NP-rats (A) and P-rats (B). The black arrow represents an upward trend, while the red one is a downward trend. The abbreviations of brain areas are the same as in the main text: NAc (nucleus of accumbens), OT (olfactory tubercle), VTA (ventral tegmental area), IC (insular cortex), Hypo (hypothalamus), DB (diagonal band), VLPO (ventrolateral preoptic nucleus); * p < 0.05, *** p < 0.005, **** p < 0.001.

Discussion
In this study, brain-wide rsfMRI was used to investigate the changes of the functional connectivity caused by different animal strains (P-and NP-rats) and the different periods during the development of alcohol preferences for P-rats. The results of the variation of the functional network might be related to the development of alcohol preference caused by persistent alcohol drinking or the related genetic background. To our knowledge, this is the first time that the dynamic changes of the whole-brain-wide functional connections during the development of alcohol preference (from the first alcohol drinking to alcohol preference or addiction) have been described. To summarize, the following regions were mainly involved: NAc, Tu, VTA, IC, Hypo, SNC, Au, DB, Hipp, and VLPO. These regions are highly related to the behaviors of reward, motivation, emotion, moods, and cognitive circuits, which could independently or interactively lead to addictive or preference behaviors.

Influence of Alcohol Stimulation on the Brain Functional Connectivity
It is well-known that ethanol could serve as an anesthetic agent [27] and was used as such in ancient times. In addition, ethanol may have some unknown effects on the BOLD signal of the fMRI. To avoid the possibility that the changes of the brain connectivity are caused by ethanol's effects, the changes of brain connections were only investigated under different states of alcohol preferences.

Functional and Structural Connections of Brain Regions
Although the investigation of the formation of the BOLD signal remains unresolved and the fMRI method could not be used to calculate the structural connection among different brain regions, it was important to analyze the changes of the functional connectivity during different states of alcohol drinking and determine the key regions associated with alcohol preference or addiction.
First, it can be seen from the results that the functional connectivity was partly related with the structural connectivity, and the brain regions with high functional connectivity coefficients tended to be those with strong structural connectivity. For example, the connections of some typical and representative regions (from basal ganglia to neocortex) are investigated in this study. The seventh region, the piriform cortex, had high correlations with the amygdala, insular cortex, and entorhinal cortex (the 11th, 13th, and 19th regions, respectively). These regions collectively performed in many olfactory-related behaviors or neurodegenerative diseases and had direct projections with each other [42,43]. Furthermore, the connection between the regions VTA (region 16) and SN (region 14) consisted of strong interactions, as these two brain regions were rich in dopaminergic neurons and contributed to reward circuits or reward-seeking behavior (for the SN, the main region was the SNc) [44].
In addition to the direct nerve connections, there were also intensive functional connections (BOLD-signal-dependence coherence) arising in parts of some regions, which did not have directed structural input or output connections [45]. These functional connections indicate that there were some complex factors involved in the BOLD signal analysis, such as mini-circuits, neurotransmitters, and multiple neurotransmitter-specific neuroplasticity circuits. Furthermore, the BOLD signal was used to indirectly measure the neural activity and is affected by multiple factors, including the local neural activity, metabolic capacity, the blood vessel system (blood flow and volume), and the neuroanatomy of the examined regions. Therefore, the BOLD signal may vary from the signals of neural activity measured directly by means of other technologies.

Functional Connectivity and Alcohol Preference
In this study, the functional connectivity of brain regions NAc-IC, OT-Hypo, DB-VLPO, Au-Hipp, SNC-Tha, and SNC-VTA were highly relevant to alcohol preference during the alcohol training. Most of these regions have been mentioned in human alcoholics or animal models of alcohol dependence studies. For instance, NAc and its downstream ventral striatum, which is denoted as Hypo in this research, have been examined in studies of preference and drug abuse (dopamine, alcohol reward, or addiction) [46]. The thickness of the IC is altered in alcohol [47] and drug users [48]. DB is a brain region where acute or chronic ethanol administration increases the GABA-mediated inhibition of spontaneously active neurons (GABA receptor, alcohol-awake or asleep) [49]. VLPO, which mainly refers to sleep and motor activity, could be influenced by perinatal alcohol exposure (GABA receptor, alcohol-awake or asleep) [50].
Between the Pre-/Mid-and Post-periods, there is an opposite tendency with the RS and IS for the changes of the connectivity of NAc-IC, OT-Hypo, and DB-VLPO. In addition, the changes only occur during the state of stable preference or habituation (Poststate compared with the Pre-/Mid-state). This indicates that the synaptic plasticity of these regions, which is likely associated with alcohol expectation, changes during training (chronic alcohol use), and the craving would be fulfilled after alcohol drinking or injection. The changes of the connectivity coefficients from the resting state to the intoxicated state can be related to the changes in preference from a function level. Thus, the counter-connective trend of NAc-IC, OT-Hypo, and DB-VLPO in RS (P-Pre and P-Post) and IS (P-Pre and P-Post) suggests that alcohol induces changes in the neural circuits (stable rises or falls), as well as itself into the circuits (reverse connection mode). This is similar to the homeostasis neural circuit or the different meaning of alcohol at different periods. The coefficient decrease of Au-Hipp in the IS and Post-period characterizes the effect on the senses and memory of alcoholic intoxication or chronic alcohol use, as the auditory cortex and hippocampus are critical for episodic verbal and spatial memory, as well as cognition [51]. Furthermore, the transient action of alcohol itself on the brain as a depressant, analeptic, and tolerance should also be considered.
Among these significant different connections, it is also worthwhile noting that the changes to SNC-Tha and SNC-VTA (common reward and addiction-related brain regions) only happened in the Mid-term. There are direct neural connections that are relevant to the dopaminergic system between these two pairs [52]. Therefore, the changes might suggest that the pathway of the dopamine system among these regions may only take effect during the period when addiction is developing rather than when the addiction is being maintained.

Comparison of the Genetic and Alcohol Drinking Factors on the Alcohol Preference
It is well-known that alcoholism can be affected by social, psychological, and genetic factors. Previous research has indicated that the heritability of liability for alcohol addictions is estimated to be at 51-59% [53], and is mainly caused by familial resemblance. The inbred rat strains of P-and NP-rats were the best models to study the variations of functional networks on alcohol preference related to family history and alcohol training. Both the P-and NP-rats originated from a randomly bred Wistar colony [54]. These two kinds of animals have been used to study alcohol preference-related behaviors for many years [55]. According to the results of the correlation of the functional connectivity in different groups, there were significantly greater differences between different strains (P-rats vs. NP-rat) (p < 0.05) regardless of alcohol drinking training or alcohol stimulation. This provides possible evidence to support the fact that genetic factors play an important role in alcohol addiction or preference.
The difference in the brain connections could provide another important clue in investigating the underlying neurobiological mechanisms. Thus, the difference (∆r > 0.15 and r > 0.20) of functional connectivity among the P-rats before/after alcohol preference periods and the NP-rats were collected to study the molecular mechanism of alcohol addiction among difference strains (Table S2). Finally, the formation of alcohol addiction involves neuroplasticity and neuroadaptation with the molecular mechanism. To completely understand and cure this disease, the identification of the molecular and neurochemical relationships with the changes of connectivity of these relevant brain regions should be further investigated.

Limitations and Perspective
In this study, the changes of brain networks during the formation of alcohol preference were explored without focusing on gender differences and the influence of other psychiatric problems (such as stress). The cerebral functions for male and female animals were different, and we also found that the cerebral regional metabolic information for different genders of alcohol-preferring or alcohol-non-preferring rats was different [56]. However, only male animals were utilized in this study. This is the major limitation of this study. To explore the alcohol preferences of female animals, more experiments on female alcohol preferences should be studied in the future.
During the alcohol preference training, the alcohol drinking ratio for a single animal was monitored every day. Thus, the single-housed method was always utilized to screen the alcohol-preferring animals [57,58]. However, single-cage feeding is the common approach to model stress. We cannot strictly exclude the influence of external factors caused by single-cage feeding on the brain networks. Thus, all the animals in this study were housed using the same method.

Conclusions
To summarize, P-and NP-rats were used to analyze the variations of the whole-brain functional network during alcohol training with the fMRI technology. Here, the regions of the OT, NAC, IC, DB, VLPO, Hypo, VTA, and SNC were found to change at different stages of alcohol preference formation. Unlike previous clinical studies that focused on comparing alcohol addiction patients with healthy groups, this animal study focused more on the process of alcohol preference formation. Alcohol preference is highly correlated with many diseases, such as alcohol addiction. The brain regions we found hold promise as new research and therapeutic targets for alcohol abuse-related diseases. Furthermore, this study of P-and NP-rats of different family backgrounds also provides evidence that genetic factors are more influential in the development of alcohol preference than constant alcohol exposure. This evidence suggests that people with a genetic history of diseases such as alcohol addiction should be more careful about avoiding exposure to addictive substances. In conclusion, this study provides important implications for both the clinical prevention and early intervention of alcohol preference and addiction for the future clinical study of alcoholism therapy.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/brainsci11111474/s1, Table S1: Collection of significant changes of the functional connectivity after ethanol stimulation comparing with the rest state during the periods of Pre-0 and Post-4 weeks alcohol training; Table S2: Collection of the difference of functional connectivity among P-rats before/after alcohol preference periods and NP-rats (∆r > 0.15 and r > 0.20). Table S3: Results of the statistical analysis of the functional connectivity in during the different periods of alcohol training (File name: 'rs.xlsx') and acute alcohol administration (File name: 'is.xlsx').

Informed Consent Statement: Not applicable.
Data Availability Statement: The datasets generated to obtain the results presented in this article are available from the corresponding authors on reasonable request (jie.wang@wipm.ac.cn/fuqiang.xu@wipm.ac.cn).