Accelerated Aging and Motor Control Deficits Are Related to Regional Deformation of Central Cerebellar White Matter in Alcohol Use Disorder

Abstract

The World Health Organization estimates a 12-month prevalence rate of 8+% for an Alcohol Use Disorder (AUD) diagnosis in people age 15 years and older in the U.S. and Europe, presenting significant health risks that have the potential of accelerating age-related functional decline. According to neuropathological studies, white matter systems of the cerebellum are vulnerable to chronic alcohol dependence. To pursue the effect of AUD on white matter structure and functions in vivo, this study used T1-weighted, Magnetic Resonance Imaging (MRI) to quantify the total corpus medullare of the cerebellum and a finely-grained analysis of its surface in 135 men and women with AUD (mean duration of abstinence: 248 days) and 128 age- and sex-matched control participants; subsets of these participants completed motor testing. We identified an AUD-related volume deficit and accelerated aging in the total corpus medullare. Novel deformation-based surface morphometry revealed regional shrinkage of surfaces adjacent to lobules I-V, lobule IX, and vermian lobule X. In addition, accelerated aging was detected in the regional surface areas adjacent to lobules I-V, lobule VI, lobule VIIB, and lobules VIII, IX, and X. Sex differences were not identified for any measure. For both volume-based and surface-based analysis, poorer performance in gait and balance, manual dexterity, and grip strength were linked to greater regional white matter structural deficits. Our results suggest that local deformation of the corpus medullare has the potential of identifying structurally and functionally segregated networks affected in AUD.

INTRODUCTION

The cerebellum is a principal target of CNS damage related to excessive and chronic consumption of alcohol. Neuropathologically, atrophy is detectable even when cell loss is not; in those cases, cerebellar white matter can exhibit degradation of fiber cores (reviewed by de la Monte & Kril, 2014; Sutherland, Sheedy & Kril, 2014). These signs of disrupted fiber integrity have been identified in cases of alcoholism uncomplicated by common concomitants, such as liver cirrhosis or nutritional insufficiencies typically from thiamine depletion causing Wernicke’s encephalopathy or Korsakoff’s syndrome. Nonetheless, concomitant complications exacerbate pathology (Baker, Harding, Halliday et al. 1999; Phillips, Harper & Kril, 1987; Torvik & Torp, 1986; Victor, Adams & Collins, 1989). Neuroradiologically, the mainstay of studies of uncomplicated alcoholism concurs with neuropathological findings, especially of white matter involvement (Monnig, Tonigan, Yeo et al. 2013). Quantitative MRI studies report tissue shrinkage typically in the anterior superior vermis, involving lobules I-V with additional sites of volume deficit in the white matter of the cerebellar hemispheres (Sawyer, Oscar-Berman, Mosher Ruiz et al. 2016; Sullivan, Deshmukh, Desmond et al. 2000; for review, Zahr, Pfefferbaum & Sullivan, 2017), also exacerbated by common alcoholism-related concomitant complications (Ritz, Segobin, Lannuzel et al. 2016; Segobin, Ritz, Lannuzel et al. 2015; Sullivan, Deshmukh, Desmond et al. 2000).

Normal aging is another source of cerebellar tissue volume decline (e.g., Persson, Ghisletta, Dahle et al. 2014; Raz, Ghisletta, Rodrigue et al. 2010; Ziegler, Dahnke, Jancke et al. 2012). Longitudinal studies confirm cross-sectional reports of regional volume loss that must be taken into account in assessing the effects of pathology in aging adults. When aging is compounded with alcohol use disorder (AUD), controlled studies have identified age-alcoholism interactions, where age-related cortical gray matter and white matter volume declines accelerate in alcoholic participants compared with unaffected controls (Pfefferbaum, Zahr, Sassoon et al. 2018; Sullivan, Zahr, Sassoon et al. 2018). The age-alcoholism interaction may also be relevant to motor performance, which declines decidedly with aging (e.g.,Holtzer, Epstein, Mahoney et al. 2014; Koppelmans, Hirsiger, Merillat et al. 2015), may accelerate in older alcoholics (Sullivan, Deshmukh, Desmond et al. 2000), and is likely to have a cerebellar substrate (Fitzpatrick, Jackson & Crowe, 2012; Sullivan, Rose & Pfefferbaum, 2006).

MRI findings in abstinent alcoholics describe shrinkage of folia in the anterior superior vermis, notably lobules I-V (e.g., Sawyer, Oscar-Berman, Mosher Ruiz et al. 2016; Sullivan, Deshmukh, Desmond et al. 2000). By contrast, cerebellar volume shrinkage in normal aging occurs more inferiorly and posteriorly, in vermian lobules VI-X (Raz, Dupuis, Briggs et al. 1998), Crus I-II and vermian lobules Vi and VIIa (Yu, Korgaonkar & Grieve, 2017), and lobules V-VI (Ziegler, Dahnke, Jancke et al. 2012). Thus, alcoholics who continue to drink or initiate dependent drinking later in life (cf., Breslow, Castle, Chen et al. 2017; Sullivan, Zahr, Sassoon et al. 2018) are at heightened risk for developing widespread cerebellar volume deficits with ramifications for extensive, cerebellar-based performance deficits.

Quantification of cerebellar gray matter based on T1-weighted MRI remains challenging because of the structural complexity of the folia’s cortical foldings, resulting in partial voluming and poor boundary definition between neighboring regions (e.g., Cardenas, Studholme, Gazdzinski et al. 2007). By contrast, the boundary of the corpus medullare, the major white matter system of the cerebellum, is defined by clear intensity differences from gray matter and affected minimally by partial voluming, thereby enabling reliable measurement of this region with its relatively homogenous signal intensity. Measurement of this structure, which has spatially segregated fiber systems connecting selective cerebellar and cerebral structures and regions (Kelly & Strick, 2003; Schmahmann & Pandya, 1997), provides a foundation for posing specific hypotheses about the in vivo status of cerebellar white matter integrity in alcoholism, its interaction with age, and its relation with function.

Accordingly, using a matched subset of T1-weighted MRIs (N=263) from reports focused on the cerebral structures of alcoholic and control groups (Pfefferbaum, Zahr, Sassoon et al. 2018; Sullivan, Zahr, Sassoon et al. 2018), the cerebellar corpus medullare was subjected to two quantification approaches. The first identified the white matter boundaries of the cerebellum, measuring the full volume of the corpus medullare. Unlike prior volumetric approaches applied to the cerebellum (Cardenas, Durazzo, Gazdzinski et al. 2011; Sawyer, Oscar-Berman, Mosher Ruiz et al. 2016; Ziegler, Dahnke, Jancke et al. 2012), the second approach used deformation-based surface morphometry (Chung, Worsley, Robbins et al. 2003) to quantify structural differences along the white matter - gray matter interface. Along that interface, differences in regional surface areas between AUD and control groups were related to the adjacent lobes of the gray matter. Using these two approaches, this study tested four hypotheses: 1) relative to the control group, the AUD group would exhibit a volume deficit of the corpus medullare; 2) the volume of the AUD group would show a steeper age-related decline than observed in controls; 3) surface area analysis applied to the corpus medullare would reveal regional shrinkage in the AUD, notably, in white matter adjacent to lobules I-V and IX-X, and greater age declines in AUD than controls in selective regions; and 4) poorer performance on tests of gait and balance, manual dexterity, and quality of life would correlate with greater regional surface area deficits in the AUD group.

MATERIALS AND METHODS

Participants

The participants were drawn from our ongoing longitudinal studies of brain MRI (control and alcoholic participants (Sullivan, Zahr, Sassoon et al. 2018). Clinical psychologists or research nurses administered the Structured Clinical Interview for the Diagnostic Statistical Manual - IV (DSM-IV) (American Psychiatric Association, 2013) to all study participants. Only participants meeting DSM-IV criteria for alcohol dependence were included in the patient group. Prospective control participants did not meet DSM-IV criteria for any Axis I disorder. Quantity of lifetime alcohol consumption and date of last drink were obtained from all participants by interview (Pfefferbaum, Lim, Zipursky et al. 1992b; Skinner, 1982; Skinner & Sheu, 1982). The study was conducted from April 11, 2003 to March 3, 2017. All study participants provided written informed consent; the institutional review boards of Stanford University School of Medicine and SRI International approved this study. Participants received modest financial compensation.

Of the 222 AUD participants, 135 had usable MRI data for cerebellar analysis defined by visual inspection (also see the following section). Subsets of those AUD participants also had undergone motor testing (Table 2). A group of 128 control participants with usable MRIs was selected from the overall 199 control participants to match AUD cohort with respect to sex (χ² <.001, p=0.99) and age (two-sample t-test=−0.64, p =0.53). Participants also recorded the Quality of Social Functioning (QSF) composite score, which comprised Quality of Life SF-21 total raw score (Bozzette, Hays, Berry et al. 1995), Global Assessment of Functioning score (current) (Katz, 1983), and Activities of Daily Living (combined Performance and Instrumental scores) (Moos, McCoy & Moos, 2000). Raw scores from these tests were age-corrected based on 66 male and 85 female, healthy controls, aged 20 to 67 at their first examination and expressed as standardized Z-scores. Participants also recorded Body Mass Index (BMI), calculated as one’s weight divided by height squared, as a measure of body fat and nutrition status.

Table 2.

Subsets of the AUD participants with motor test scores

		FFM	AFT	GRIP	Stand Heel-to-Toe	Walk Heel-to-Toe	Stand on One Foot
No.		74	59	31	84	83	82
Sex, No. (%)	Men	45(60.8%)	32(54.2%)	18(58.0%)	52(61.9%)	52(62.7%)	51(62.2%)
	Women	29(39.2%)	27(45.8%)	13(42.0%)	32(38.1%)	31(37.3%)	31(37.8%)
Age, y, mean(SD)		47.9(11.5)	50.1(11.0)	47.6(9.3)	48.2(11.1)	48.2(11.2)	48.2(11.2)
Score, mean(SD)		303.8(53.8)	119.7(20.5)	63.2(20.7)	45.5(44.1)	3.3(3.0)	23.7(21.0)

Compared with the control group, the AUD group had significantly fewer years of education (t= −8,58, p < .0001), lower QSF scores (t=−11.6, p < .0001), and more participants with history of nicotine dependence (χ² = 75.9, p < .0001). However, there was no significant difference in BMI between the two groups.

MRI Acquisition and Processing

An MRI of each participant was acquired on a 3T GE whole-body MR scanner (General Electric Healthcare, Waukesha, WI) using an 8-channel phased-array head coil acquiring T1-weighted Inversion-Recovery Prepared SPGR images (TR=6.55/5.92 ms, TE=1.56/1.93 ms, TI=300/300 ms, matrix = 256×256, thick=1.25 mm, skip=0 mm, 124 slices). Routine phantom data were used to evaluate spatial fidelity; drift was corrected by adjusting scanner calibration parameters when necessary to maintain spatial stability within manufacturer guidelines. Each scan passed visual quality control (Q.Z.), which included checking that signal drop-off artifacts did not significantly affect the segmentation of the cerebellum.

To generate the segmentations of the cerebellum, noise removal (Smith & Nichols, 2009) and image inhomogeneity correction via N4ITK (Coupe, Yger, Prima et al. 2008) were first applied to each MRI. Then the image was skull stripped by a mask, which was generated by performing majority voting (Pfefferbaum, Zahr, Sassoon et al. 2018) at each voxel based on the masks estimated by FSL BET (Smith, 2002), AFNI 3dSkullStrip (Cox, 1996), and FreeSurfer mri_gcut (Sadananthan, Zheng, Chee et al. 2010). Freesurfer v5.3.0 was applied to the skull stripped MRI to segment the cerebellum. This label map was further refined using the cerebellar lobule segmentation tool (Yang, Ye, Bogovic et al. 2016), which parcellated the cerebellum into the corpus medullare and 21 cerebellar lobules according to an atlas (Bogovic, Jedynak, Rigg et al. 2013). The segmentation accuracy associated with cerebellar gray matter was greatly affected by low intensity-contrast of the images (Smith & Nichols, 2009). Therefore, the present study focused on analyzing the structural features of the white matter, i.e., the corpus medullare.

In addition to the volume score, the shape of the corpus medullare of each participant was analyzed by comparing it to the surface of a template via deformation-based surface morphometry (Chung, Worsley, Robbins et al. 2003). First, a template of the corpus medullare was created by aligning the label maps of all participants via group-wise deformable registration (Joshi, Davis, Jomier et al. 2004) and then performing a multi-label fusion on the registered label maps via ANTS 2.1.0 (Avants, Tustison, Song et al. 2011). A surface mesh of the corpus medullare was generated by applying the marching cube algorithm (Lorensen & Cline, 1987) to the template and smoothing the outcome via the Taubin method (Taubin, 1995). This triangle mesh contained 2159 vertices, each of which was labeled according to its adjacent lobule (Fig 1). For each participant, the surface of the template was registered to their surface of the corpus medullare via deformable surface registration (Zhao, Price, Pizer et al. 2015, 2016). From the resulting deformation map, the Tangential Jacobian Determinant (Zhao, Price, Pizer et al. 2015, 2016) was computed for each vertex of the mesh quantifying the difference in area between the surface of the participant and template at that location.

Fig. 1.

Superior, inferior, posterior and lateral view of the template surface of the corpus medullare. The surface was parcellated into 21 regions by labeling each surface vertex according to its adjacent gray-matter lobule.

Motor Testing

Fine Finger Movement Test (FFM) required subjects to turn a knurled rod with their forefinger and thumb, unimanually and then bimanually (Cooper, Sagar, Jordan et al. 1991; Corkin, Growdon, Sullivan et al. 1986). Three, 30-second trials for each condition were administered. Correlational analyses used the mean of the four conditions.

Ataxia testing was based on the Walk-a-Line Battery of gait and balance (Fregly, Graybiel & Smith, 1972), which consisted of three tasks, each performed first with eyes open and then eyes closed, and always with arms folded across the chest. Each condition was tested twice, unless a perfect score was achieved on the initial trial, in which case the subject received full credit on that condition. The conditions were as follows: Stand Heel-to-Toe (AXCRMB) for 60 sec. (maximum score=120 sec.); Walk Heel-to-Toe (AXCSTP) for 10 steps (maximum score=20 steps); and Stand on One Foot (AXCLG), first the right foot and then the left, each for 30-sec. trials (maximum score=60 sec. for each foot).

Alternated finger tapping test (AFT) required the subject to tap a telegraph key as quickly as possible with the index finger of the right hand, left hand, and then alternating with each hand. The number of key presses recorded by a counter during each of the three 15-second trials was recorded and averaged across trials within conditions (Reitan & Davison, 1974).

Unimanual grip strength (GRIP) was measured with a hand dynamometer. The score was the mean of three trials per hand.

Analysis of Corpus Medullare Volume

Group Comparison of Volume Scores (Test 1).

Age, sex, and supratentorial volume (svol) were first regressed out from the volume scores of the corpus medullar based on a general linear model (GLM; glmfit in Matlab 2016b). The GLM was fitted to the control group and then applied to the volume scores of all participants. A two-sample t-test then examined the residuals of that regression for the group difference in the volume of the corpus medullare between AUD and control participants (significant if one-tailed p < 0.05).

Correlation between Volume and Demographics (Test 2).

The effects of nicotine dependence, BMI, QSF, years of education, Life-time-alcohol-consumption and Days-since-last drink on the residual volume scores were tested in the AUD group. Specifically, a two-sample t-test (significant if one-tailed p < 0.05) was applied to the residual scores of AUD participants without history of nicotine dependence versus AUD participants with a history of nicotine dependence. Significance with respect to the remaining demographic and alcohol use factors was determined by applying GLM analysis (correlation significant if one-tailed p < 0.05) to the residual scores of all AUD participants.

Correlation between Volume and Motor Test Scores (Test 3).

The correlation between the volume residuals and each of the 6 motor test scores (Table 2) was examined in the AUD group (correlation significant if one-tailed p < 0.05).

Age-associated Decline in Volume and Age-Alcoholism Interaction (Test 4).

To examine the relation between age and volume, sex and svol were first removed from the volume scores by repeating the GLM analysis in Test 1. Then, the following linear model was fitted with respect to the residual volume scores (VOL_RES)and age

$${\text{VOL}}_{\text{RES}}=\left({\text{b}}_{\text{ctrl}}+{\text{a}}_{\text{ctrl}}\ast \text{age}\right)\ast {\text{δ}}_{\text{ctrl}}+\left({\text{b}}_{\text{AUD}}+{\text{a}}_{\text{AUD}}\ast \text{age}\right)\ast \left(1-{\text{δ}}_{\text{ctrl}}\right).$$

δ_ctrl was a binary variable indicating whether a participant belonged to the control group.b_X and a_X were the intercepts and slopes of the linear aging function associated with group X. Based on this GLM model, age-associated decline in volume was separately examined for the control and AUD group by applying t-tests to a_ctrl and a_AUB. Age-alcoholism interaction was examined by applying a t-test to the slope difference between the two groups, i.e. a_ctrl – a_AUB, with the null hypothesis being a_ctrl – a_AUB = 0 (one-tailed p < .05). This GLM formulation allowed for examining slopes in individual groups and the slope difference across groups within a single model.

Surface-based Analysis

Further examined was the surface-area around each vertex on the surface of the corpus medullare. Specifically, Test 1 and Test 4 were performed on each vertex of the surface based on the Tangential Jacobian determinants of all participants at that vertex. Test 2 and Test 3 were performed only on the vertices that showed significant group difference in surface-area according to Test 1. These vertex-wise correlation/GLM tests were computed by FSL PALM Alpha109 (Permutation Analysis of Linear Models) with 5000 permutations (Winkler, Ridgway, Webster et al. 2014) and corrected for multiple comparisons using the Threshold-free Cluster Enhancement (TFCE) method (Smith & Nichols, 2009) of the PALM software package. Test results were significant if one-tailed t-test p<0.05 after TFCE correction.

RESULTS

Volume-based Analysis

The average volume score of the AUD group was 0.43ml smaller than that of the control group (Test 1, t = −2.25, p = 0.013, Fig. 2a), but the volume score did not significantly correlate with lifetime alcohol consumption, days since last drink, nicotine dependence, BMI, QSF or years of education.

Fig. 2.

(a) Group comparison on the volume of the corpus medullare between the AUD and control group. (b) Age-associated decline of volume in the control (blue) and AUD (red) group. (c-f) Correlation plots between the volume score of the corpus medullare and each of the 4 motor test scores (FFM, GRIP, AXCRMB and AXCSTP) for AUD subjects.

For the motor tests (Test 3), FFM (t = 1.97, p = 0.027), GRIP (t = 2.41, p = 0.011), AXCRMB (t = 2.05, p = 0.021), and AXCSTP (t = 1.92, p = 0.029) correlated with volume scores (Test 3, Fig. 2c-f), but ASCLG and AFT did not. The correlation rate was r = .007ml per unit score for FFM, r = .033ml for GRIP, r = .009ml for AXCRMB and r = .012ml for AXCSTP.

The volume score showed an age-associated decline in the AUD group (Test 4, a_AUD = −0.0385mL/ year, t = −3.17,p = 0.008) that exceeded the age decline of the control group (Test 4, a_AUD – a_ctrl = −0.0373ml/year, t = −2.27, p = 0.012).

All resulting p-values remained significant when replacing GLM analysis with robust linear regression analysis (robustfit in Matlab 2016b). Linear regression analysis was robust to outliers in the volume measurements and the ceiling effect of the ataxia test scores.

Surface-based Analysis

The total surface area of the template of the corpus medullare was 41.5 cm². Surface-area deficits in the AUD group compared with the control group were detected in 16 lobule regions (Test 1, Fig. 3a; Fig. 4a; Table 3) covering a total surface area of 14.5 cm². The average difference in the surface area of the detected clusters between the two groups was 0.32cm². Lobules with > 50% surface region showing significant group difference, where the AUD group was smaller than the control group, were lobules I-V left, lobule IX left and right, and vermian lobule X.

Fig. 3.

p-value maps associated with vertex-wise tests of (a) Group comparison on the Tangential Jacobian Determinant between the AUD and control group; (b) Age-associated decline of Tangential Jacobian Determinant in the AUD group; (c) Age-alcoholism interaction on the Tangential Jacobian Determinant. For all plots, vertex-wise p-values (after multiple comparison correction) were color-coded on the template surface.

Fig. 4.

(a) Shrinkage of surface area in the AUD group compared to the control group; (b) Age-related decline rate of surface area in the AUD group; (c) Accelerated decline rate of surface area in the AUD group compared to the control group.

Table 3.

Area of the surface that showed significant test results (p<0.05) in each lobule and the minimum vertex-wise p-value (after TFCE multiple comparison correction) in each lobule. Lobules with no significant surface area (one tailed p<0.05) are marked in gray

Region Names	Surface Area (cm2)	Group Comparison Area (cm2), %, min-p	Age decline (AUD) Area (cm2), %, min-p	Age-alcoholism interaction Area (cm2), %, min-p
lobules I~V left	4.96	3.09, 62.3%, 0.010	4.50, 90.5%, 0.003	2.15, 43,2%, 0.028
lobules I~V right	5.32	2.59, 48.8%, 0.012	4.59, 86.2%, 0.002	3.55, 66.7%, 0.021
lobule VI vermis	0.58	0.28, 48,6%, 0.015	0.53, 91.0%, 0.003	0.18, 31.5%, 0.035
lobule VI left	3.93	0.65, 16.5%, 0.019	3.73, 94.8%, 0.004	2.67, 68.0%, 0.029
lobule VI right	3.81	0.63, 16.6%, 0.016	3.52, 92.4%, 0.002	3.40, 89.3%, 0.016
lobule VII vermis	0.26	0.02, 9.4%, 0.023	0.26, 100%, 0.003	0.10, 37.2%, 0.031
lobule VII crus 1 left	1.36	0 (0%), 0.226	0.97, 71.8%, 0.004	0.21, 15.1%, 0.033
lobule VII crus 2 left	0.58	0 (0%), 0.057	0.54, 93.2%, 0.004	0.06, 9.6%, 0.035
lobule VIIB left	0.87	0 (0%), 0.057	0.71, 81.6%, 0.003	0.43, 49.6%, 0.027
lobule VII crus 1 right	1.52	0 (0%), 0.071	1.09, 71.8%, 0.004	0.46, 30.5%, 0.022
lobule VII crus 2 right	0.54	0 (0%), 0.084	0.35, 64.3%, 0.006	0.11, 21.3%, 0.034
lobule VIIB right	1.10	0.06, 5.8%, 0.019	0.71, 64.0%, 0.003	0.72, 65.7%, 0.030
lobule VIII vermis	1.30	0.35, 26.9%, 0.019	1.30, 100%, 0.003	0.05, 3.7%, 0.031
lobule VIII left	3.48	0.17, 5.0%, 0.030	3.29, 94.6%, 0.003	2.30, 66.0%, 0.027
lobule VIII right	3.29	0.87, 26.6%, 0.015	3.26, 99.0%, 0.003	3.00, 90.0%, 0.030
lobule IX vermis	0.42	0.17, 40.3%, 0.016	0.22, 52.7%, 0.005	0, 0%, 0.113
lobule IX left	2.54	2.24, 88.0%, 0.010	2.53, 99.4%, 0.003	0.28, 10.9%, 0.028
Lobule IX right	2.44	2.37, 97.0%, 0.010	2.32, 95.1%, 0.002	1.56, 63.9%, 0.024
lobule X vermis	0.66	0.56,85.4%, 0.011	0.36, 55.6%, 0.003	0.02, 2.6%, 0.046
Lobule X left	1.20	0.12, 10.0%, 0.019	1.00, 83.2%, 0.003	0.74, 61.5%, 0.028
Lobule X right	1.35	0.28, 21.7%, 0.020	0.98, 72.3%, 0.004	0.74, 54.7%, 0.024
All	41.54	14.48, 34.9%, 0.009	36.76, 88.5%, 0.002	22.67, 54.6%, 0.016

When performing Tests 2 and 3 to vertices with significant group differences in surface area, no significant correlation was found between regional surface areas and any of the three consumption measures, BMI or the QSF score in the AUD group. However, surface area correlated with 4 out of 6 motor test scores (AXCRMB, AXCSTP, FFM and GRIP). Specifically, correlations with FFM were found in 7 lobule regions with a total surface area of 3.5 cm², GRIP in 12 lobule regions with 9.2 cm², AXCRMB in 13 regions with 7.6 cm², and AXCSTP in 4 lobule regions with 0.34 cm² (Fig. 5, Table 4). The correlation rate was rather uniform inside the detected clusters with r = .002 cm² per unit score for FFM, r = .02 cm² for GRIP, r = .004 cm² for AXCRMB and r = .004 cm² for AXCSTP.

Fig. 5.

p-value maps associated with vertex-wise correlations between the Tangential Jacobian Determinant and each of the 4 motor test scores (FFM, GRIP, AXCRMB, and AXCSTP).

Table 4.

Area of the surface that showed significant correlation with motor test scores in each lobule and the minimum vertex-wise p-value (after TFCE multiple comparison correction) in each lobule. Lobules with no significant area (one tailed p<0.05) are marked in gray

Region Names	Surface Area (cm2)	FFM Area, %, min-p	GRIP Area, %, min-p	Stand Heel-to-Toe Area %, min-p	Walk Heel-to-Toe Area, %, min-p
lobules I~V left	4.96	0, 0%, 0.061	1.03, 20.8%, 0.030	0.22, 4.5%, 0.044	0.08, 1.5%, 0.047
lobules I~V right	5.32	1.55, 29.2%, 0.030	2.48, 46.6%, 0.013	2.48, 46.6%, 0.005	0.07, 1.3%, 0.049
lobule VI vermis	0.58	0.19, 32.2%, 0.046	0.20, 34.8%, 0.023	0.20, 33.9%, 0.005	0 (0%), 0.139
lobule VI left	3.93	0, 0%, 0.056	0.65, 16.5%, 0.028	0 (0%), 0.053	0 (0%), 0.071
lobule VI right	3.81	0.45, 11.8%, 0.045	0.63, 16.6%, 0.017	0.63, 16.7%, 0.005	0 (0%), 0.108
lobule VII vermis	0.26	0, 0%, 0.130	0, 0%, 0.069	0.02, 9.4%, 0.017	0 (0%), 0.176
lobule VII crus 1 left	1.36	0, 0%, NA	0, 0%, NA	0, 0%, NA	0 (0%), NA
lobule VII crus 2 left	0.58	0, 0%, NA	0, 0%, NA	0, 0%, NA	0 (0%), NA
lobule VIIB left	0.87	0, 0%, NA	0, 0%, NA	0, 0%, NA	0 (0%), NA
lobule VII crus 1 right	1.52	0, 0%, NA	0, 0%, NA	0, 0%, NA	0 (0%), NA
lobule VII crus 2 right	0.54	0, 0%, NA	0, 0%, NA	0, 0%, NA	0 (0%), NA
lobule VIIB right	1.10	0, 0%, 0.100	0.04, 3.5%, 0.034	0.06, 6.8%, 0.017	0 (0%), 0.167
lobule VIII vermis	1.30	0, 0%, 0.055	0.35, 26.9%, 0.023	0.33, 25.2%, 0.017	0 (0%), 0.100
lobule VIII left	3.48	0, 0%, 0.106	0.08, 2.2%, 0.025	0 (0%), 0.088	0 (0%), 0.109
lobule VIII right	3.29	0.03, 1.0%, 0.043	0.70, 21.0%, 0.023	0.35, 10.6%, 0.018	0 (0%), 0.078
lobule IX vermis	0.42	0, 0%, 0.080	0 (0%), 0.202	0.13, 30.2%, 0.018	0 (0%), 0.153
lobule IX left	2.54	0, 0%, 0.105	1.24, 48.7%, 0.021	0.78, 30.7%, 0.043	0.17, 6.7%, 0.045
Lobule IX right	2.44	0.99, 40.4%, 0.030	1.50, 61.5%, 0.021	2.0, 80.9%, 0.009	0 (0%), 0.051
lobule X vermis	0.66	0.03, 5.1%, 0.042	0 (0%), 0.054	0.2, 31.0%, 0.019	0 (0%), 0.203
Lobule X left	1.20	0, 0%, 0.175	0 (0%), 0.078	0 (0%), 0.054	0.02, 2.0%, 0.049
Lobule X right	1.35	0.21, 16.2%, 0.030	0.27, 19.7%, 0.021	0.2, 15.2%, 0.009	0 (1.5%), 0.053
All	41.54	3.46, 8.33%, 0.030	9.17, 22.1%, 0.013	7.59, 17.8%, 0.005	0.34, 0.8%, 0.045

Age-associated declines in surface area were detected in the AUD group, where all lobules showed aging effects with at least 50% surface region showing significant surface-area decline (Fig. 3b; Fig. 4b; Table 3). The detected clusters had a total surface area of 36.8 cm² with a decline rate of a_AUD = −0.06 cm²/year. Furthermore, age-AUD interactions occurred in 20 lobules with the single exception of vermian lobule IX. Primarily affected lobules (with > 50% surface region showing significant age-alcoholism interaction) were lobules I-V right, lobule 6 left and right, lobule VIIB right, lobule VIII left and right, lobule IX left, and lobule X left and right (Fig. 3c; Fig. 4c; Table 3). The detected clusters had a total surface area of 22.67 cm², and the decline was faster by a_AUD – a_ctrl = –0.05 cm²/year in the AUD group.

DISCUSSION

The mainstay of the study hypotheses was supported. Specifically, adults with AUD exhibited volume deficits and accelerated aging of the total corpus medullare volume compared with age-matched healthy controls. Surface-area deformations at the white matter boundaries most notably affected were those neighboring and adjacent to lobules I-V on the left, lobule IX on the left and right, and vermian lobule X. These surface areas and total volume of the corpus medullare in the AUD group significantly correlated with motor test performance; however, relations between total or regional white matter with the measure of quality of life functioning were not forthcoming. Accelerated aging of regional surface areas in the AUD participants affected all but one of the 21 surface regions; the exception was lobule IX of the vermis.

AUD-related Volumetric and Regional Surface-area Deficits

The total volume of the corpus medullare, adjusted for age, sex, and supratentorial volume (svol), was smaller by 3.2% in the AUD group than the control group with no residual difference between men and women. This alcoholic-control difference was about half that reported by Sawyer et al. (Sawyer, Oscar-Berman, Mosher Ruiz et al. 2016), who found a 6.1% volume deficit in the total cerebellar white matter, with this volume deficit attributable to alcoholic men. This difference in magnitude of effect could be related to sampling differences and possibly to approaches taken in adjusting for sex differences in intracranial volume (ICV): the Sawyer et al. study used sex as a covariate, whereas we used svol as a regressor, which removes variance related to svol regardless of sex (cf., Mathalon, Sullivan, Rawles et al. 1993; Pfefferbaum, Lim, Zipursky et al. 1992a). Neuropathological studies emphasize white matter as the principal target of chronic alcoholism and note loss of myelin as a significant contributor to volume deficits (de la Monte & Kril, 2014; Sutherland, Sheedy & Kril, 2014) (but see Andersen, Gundersen & Pakkenberg, 2003; Tang, Pakkenberg & Nyengaard, 2004). Critically, Kril and colleagues (Sutherland, Sheedy & Kril, 2014) emphasize that macrostructural damage to the cerebellum is restricted to white matter volume shrinkage in alcoholic cases without signs of Wernicke’s encephalopathy and thiamine deficiency, although the extent of damage in alcoholism is compounded with thiamine deficiency (Baker, Harding, Halliday et al. 1999; Langlais, Zhang & Savage, 1996; human in vivo: Le Berre, Pitel, Chanraud et al. 2014; animal: Pfefferbaum, Adalsteinsson, Bell et al. 2007; postmortem: Phillips, Harper & Kril, 1987; Sullivan, Deshmukh, Desmond et al. 2000; Torvik & Torp, 1986; Victor, Adams & Collins, 1989). Nonetheless, histological analysis indicated Purkinje cell differences in alcoholic cases compared with controls as selective to lobules I, IX, and X of alcoholic cases compared with the rest of the cerebellum (Phillips, Harper & Kril, 1987) and could contribute to white matter shrinkage through atrophy of cell processes.

Merging surface area morphology with a cerebellar atlas provided a finely-grained representation of AUD-related differences from controls. On a lobular basis, significant numbers of vertices ( >50%) were affected in selective regions, including lobules I-V, lobule IX, and vermian lobule X. On a surface area basis, not all regions of those lobules were affected. Additional regions within other lobules with fewer than 50% affected vertices also exhibited deformation deficits; the exceptions were Crus I and II. These relatively preserved regions show functional connectivity with prefrontal cortex in normal controls (O’Reilly, Beckmann, Tomassini et al. 2010) and play a role in compensation for impairment of frontally-based cognitive functions (Chanraud, Pitel, Müller-Oehring et al. 2013; Chanraud & Sullivan, 2014), including verbal (Desmond, Chen, De Rosa et al. 2003) and spatial working memory (Chanraud, Pitel, Pfefferbaum et al. 2010, 2011). In functional MRI studies, alcoholics who recruited lobules VI and VIIA performed at control levels, even though controls recruited only frontal regions while performing the task (Desmond, Chen, De Rosa et al. 2003).

AUD-related Accelerated Aging

Both the control and the AUD groups showed age-related differences in total white matter volumes and in regional vertex measures of the corpus medullare. Over and above white matter shrinkage in normal aging (e.g.,Koppelmans, Hirsiger, Merillat et al. 2015; Persson, Ghisletta, Dahle et al. 2014; Raz, Ghisletta, Rodrigue et al. 2010; Ziegler, Dahnke, Jancke et al. 2012), the AUD group’s difference from the controls was greater in older ages in nearly all 21 lobules. The greatest age-AUD effects occurred in the anterior superior and inferior posterior lobules, regions commonly reported to be affected in AUD in vivo (Fitzpatrick, Jackson & Crowe, 2012; Sullivan, Deshmukh, Desmond et al. 2000; Sullivan, Rose & Pfefferbaum, 2006) and postmortem (Baker, Harding, Halliday et al. 1999; de la Monte & Kril, 2014; Phillips, Harper & Kril, 1987; Torvik & Torp, 1986; Victor, Adams & Collins, 1989).

These cerebellar regional age-alcoholism interactions are also consistent with accelerated aging observed longitudinally in selective frontal cortical volumes of a larger group of controls and AUD participants from which the samples studied herein was drawn (Pfefferbaum, Zahr, Sassoon et al. 2018; Sullivan, Zahr, Sassoon et al. 2018). The frontal regions showing accelerated aging in AUD were the precentral and superior frontal gyri. Relevantly, the precentral cortex has strong functional connection with anterior cerebellar lobules and weaker connection with posterior lobules, and the superior frontal cortex has functional connection with the cerebellar posterior lobules (Buckner, Krienen, Castellanos et al. 2011) including Crus I and II (O’Reilly, Beckmann, Tomassini et al. 2010). The possibility of coordinated aging of cerebral and cerebellar nodes suggests that aging can affect spatially disparate network nodes selectively, with the potential of undermining associated functions with the declining structural integrity.

Motor Deficits and Regional Cerebellar Deformation Differences

Poorer performance on tests of gait and balance, manual dexterity, and grip strength correlated with greater regional surface area deficits in the AUD group. Regarding upper limb motor performance, correlations between finger speed and hand grip strength and regional white matter deficits involved lobules I-V, VI, VIII, IX, and X. These relations overlapped with prior studies using finger tapping, which was affected by lesions of lobules IV-V-VI and VIb-VIIIA (Stoodley, MacMore, Makris et al. 2016) and functional imaging results reporting that activation patterns for sensorimotor tasks using the index finger were localized to medullare lobule V with a secondary focus in lobules VIIIA/B (Stoodley, MacMore, Makris et al. 2016).

The balance measures requiring standing in the Romberg position (AXCRMB) and walking in a tandem gait (AXCTP) correlated with volumes of lobules I-V and VI, which are adjacent to the anterior superior vermis and classically associated with gait and balance (Timmann, Brandauer, Hermsdorfer et al. 2008). Importantly, corpus medullare lobules IX and X were also correlates of balance performance, appropriately, given their role in vestibular function (for review, Stoodley & Schmahmann, 2009). It is notable that none of the motor scores correlated with any of the Crus I or II regions, which have been associated with cognitive and affective processes more than sensorimotor functions (e.g., Buckner, Krienen, Castellanos et al. 2011; Guell, Schmahmann, Gabrieli et al. 2018; O’Reilly, Beckmann, Tomassini et al. 2010; Stoodley & Schmahmann, 2018). These observations are consistent with neuropathological reports of alcoholics with clinical signs of cerebellar damage including gait ataxia who showed Purkinje cell loss in the vermis as associated with WE (Baker, Harding, Halliday et al. 1999). In that study, the vermian white matter of ataxic alcoholics was 42% less than that in control cases (Baker, Harding, Halliday et al. 1999), a percentage also reported in a controlled study but without antemortem behavioral measures (Torvik & Torp, 1986).

Limitations

Despite the novelty of our analysis approach, which used surface points to render finely-grained foci of group differences at the white matter/gray matter interfaces of the cerebellum, the analysis was limited to white matter. Although we sought relations between regional volumes and demographic and consumption variables, we found few correlations other than those with age. Longitudinal studies reported that white matter of the cerebellum, measured with voxel-based deformation-based morphometry, was among the regions showing greatest recovery in abstinent alcoholics (Cardenas, Studholme, Gazdzinski et al. 2007) with longer sobriety associated with greater volume recovery (Segobin, Chetelat, Le Berre et al. 2014) (mean duration of abstinence in the AUD group was 248 days in this study). Unlike previous reports, however, we did not find that cigarette smoking (Durazzo, Mon, Gazdzinski et al. 2017; Gazdzinski, Durazzo, Mon et al. 2010), sex differences (Sawyer, Oscar-Berman, Mosher Ruiz et al. 2016), or amount of alcohol drunk in a lifetime (Le Berre, Pitel, Chanraud et al. 2014; Mon, Durazzo, Gazdzinski et al. 2013; Sullivan, Rose & Pfefferbaum, 2006) accounted for or contributed to group differences. In addition, extent of damage and potential of recovery (Monnig, Tonigan, Yeo et al. 2013; van Eijk, Demirakca, Frischknecht et al. 2013) may have a genetic component (Alexander-Kaufman, Harper, Wilce et al. 2007; Mon, Durazzo, Gazdzinski et al. 2013), which was not tested herein. Further to concepts of inherited traits, we did not have information about family history risk for AUD. To this point, Hill et al. (Hill, Lichenstein, Wang et al. 2016) observed that offspring from multiplex families with a high risk for AUD had larger corpus medullare volumes than those from low-risk families and that poorer performance on a working memory task correlated with larger white matter volumes. Given that direction of difference and relations, our findings showing the opposite relations with smaller regional volumes associated with AUD and poorer motor performance suggest that the AUD difference from controls and relations with performance may actually be related to alcohol consumption. The above discrepant findings may also relate to different abstinence durations across studies.

Conclusion

Surface-based morphometry afforded a viable approach for refined identification of structural loci of differences and potential damage from alcoholism. Merging an atlas onto the Jacobian vertex-wise surface maps revealed patterns of group differences and relations with motor performance that was not restricted to group-determined atlas borders. Rather, an atlas was superimposed onto the surface maps to approximate which lobules were affected. This approach allows for individual differences in determining borders, as has been shown in cortical mapping of the monkey brain (Rajkowska & Goldman-Rakic, 1995) and cerebellar mapping of the human brain with functional connectivity mapping (Buckner, Krienen, Castellanos et al. 2011; Dobromyslin, Salat, Fortier et al. 2012). Overlapping regional effects across individuals, nonetheless, were adequate for finding relations between local white matter deformation and performance on tests of upper motor and lower gait and balance activities, suggesting that surface values of the corpus medullare reflect structurally and functionally segregated networks.

Table 1.

Demographic Data for the Control and AUD Groups

Demographic Variables	Control	Alcohol
No.	128	135
Sex, No. (%)
Men	71 (55.5%)	75 (55.6%)
Women	57 (44.5%)	60 (44.4%)
Self-defined race/ethnicity, No. (%)
Asian	17 (13.3%)	3 (2.2%)
African American	23 (18.0%)	49 (36.3%)
White	84 (65.6%)	81 (60%)
Other (Unknown)	4 (3.1%)	2 (1.5%)
Age, y, mean (SD)
Men	48.5 (13.2)	46.8 (11.8)
Women	48.5 (10.9)	48.5 (9.3)
Handedness, No. (%)
Right-handed	90 (70.3%)	101 (74.8%)
Left-handed	6 (4.7%)	10 (7.4%)
Ambidextrous	6 (4.7%)	3 (2.2%)
Unknown	26 (20.3%)	21 (15.6%)
Years of Education, mean (SD)	16.0 (2.3)	13.3 (2.3)
Quality of life, mean (SD)	0.13 (0.77)	−1.9 (1.63)
Body Mass Index, mean (SD)	25.8 (4.4)	26.8 (5.4)
Nicotine dependence, No. (%)
Never	72 (56.3%)	35 (25.9%)
Past	1 (0.8%)	18 (13.3%)
Current	5 (3.9%)	67 (49.6%)
Unknown	50 (39.0%)	15 (10.6%)
Lifetime alcohol consumption mean (SD), kg	32 (44)	1176 (968)
Days since last drink mean (SD)	NA	248 (630)