APOL1 renal-risk variants associate with reduced cerebral white matter lesion volume and increased gray matter volume

Abstract

To assess apolipoprotein L1 gene (APOL1) renal-risk-variant effects on the brain, magnetic resonance imaging (MRI)-based cerebral volumes and cognitive function were assessed in 517 African American-Diabetes Heart Study (AA-DHS) MIND and 2,568 hypertensive African American Systolic Blood Pressure Intervention Trial (SPRINT) participants without diabetes. Within these cohorts, 483 and 197 had cerebral MRI, respectively. AA-DHS participants were 60.9% female, mean age of 58.6 years, diabetes duration 13.1 years, estimated glomerular filtration rate of 88.2 ml/min/1.73m², and a median spot urine albumin to creatinine ratio of 10.0 mg/g. In additive genetic models adjusting for age, sex, ancestry, scanner, intracranial volume, body mass index, hemoglobin A1c, statins, nephropathy, smoking, hypertension, and cardiovascular disease, APOL1 renal-risk-variants were positively associated with gray matter volume (β 3.4×10⁻³) and negatively associated with white matter lesion volume (β −0.303) (an indicator of cerebral small vessel disease) and cerebrospinal fluid volume (β −30707) (all significant), but not with white matter volume or cognitive function. Significant associations corresponding to adjusted effect sizes (β/standard error) were observed with gray matter volume (0.16) and white matter lesion volume (−0.208), but not with cerebrospinal fluid volume (−0.251). Meta-analysis results with SPRINT MIND participants who had cerebral MRI were confirmatory. Thus, APOL1 renal-risk-variants are associated with larger gray matter volume and lower white matter lesion volume suggesting lower intracranial small vessel disease.

Introduction

The apolipoprotein L1 gene (APOL1) G1 and G2 renal-risk-variants (RRVs) are strongly associated with a spectrum of non-diabetic kidney diseases in African Americans (AAs).^1-3 Whether APOL1 RRVs also associate with altered risk for cardiovascular disease (CVD) remains controversial.⁴ APOL1 G1 and G2 variants are associated with lower levels of calcified atherosclerotic plaque (CP) in coronary and carotid arteries.^5;6 CP is a measure of subclinical CVD and an important predictor of CVD events and mortality.⁷ The African American-Diabetes Heart Study (AA-DHS) and the Jackson Heart Study (JHS) both detected lower levels of CP in AAs with increasing numbers of APOL1 RRVs.^5;6 However, observations linking APOL1 RRVs with outcomes in these cohorts yield opposing results. AA-DHS identified improved survival with increasing numbers of APOL1 RRVs.⁶ In contrast, JHS (validated in the Women’s Health Initiative [WHI]) observed increased risk for myocardial infarction and CVD events in those with two APOL1 RRVs.^4;5 The paradoxically higher CVD risk with APOL1 in JHS and WHI could relate to confounding from APOL1 association with nephropathy. APOL1 was not associated with chronic kidney disease (CKD) in the type 2 diabetes (T2D)-affected AA-DHS cohort, thereby minimizing the potential for nephropathy-related CVD confounding effects.⁶ Nonetheless, the impact of APOL1 RRVs on the extra-renal vasculature requires additional study.

Cerebral structure can be accurately assessed using magnetic resonance imaging (MRI).⁸ Therefore, the AA-DHS MIND and the Systolic Blood Pressure Intervention Trial (SPRINT) MIND studies were initiated to improve understanding of environmental and inherited risk factors for subclinical cerebrovascular disease and cognitive performance.^9-11 AA-DHS further aimed to assess relationships between conventional and non-conventional CVD risk factors with cerebral volumes and cognitive performance in the understudied AA population.¹² Based on potential systemic vascular effects of APOL1, the G1 and G2 RRVs were tested for association with cerebral structure and cognitive function in AA-DHS MIND participants with T2D and hypertensive AA SPRINT MIND participants without diabetes.

Results

Demographic, clinical and imaging data in 517 AA-DHS MIND participants is displayed in Table 1. Of the total, 483 completed laboratory analysis, MRI scans, and cognitive testing (34 completed cognitive testing and laboratory analysis, but not the MRI due to claustrophobic symptoms). As in general populations, 13.9% of all AA-DHS MIND participants had two APOL1 RRVs; this group also had significantly higher overall African ancestry proportion, more kidney disease, and greater percentages of high school and college graduates. Although TICV was similar across the three genotype groups, significantly lower unadjusted WMLV was seen in those with two APOL1 RRVs. The APOL1 renal-risk genotype association with nephropathy in AA-DHS MIND differs from the lack of nephropathy association in a prior AA-DHS report, because 220 new participants were recruited for MRI scans.⁶ AA SPRINT (and SPRINT MIND MRI) participant demographic, clinical and MRI imaging data are presented in Table 2; 14% of the 2,568 AA SPRINT participants had two APOL1 RRVs. As in AA-DHS MIND, these individuals had higher overall African ancestry proportion; they also had a lower estimated glomerular filtration rate (eGFR) and higher urine albumin:creatinine ratio (UACR).

Table 1.

Baseline demographic, laboratory, neuropsychological testing, and MRI data in the AA-DHS MIND

Variables	APOL1 0 risk alleles		APOL1 1 risk allele		APOL1 2 risk alleles		P-value1
	N	Mean (SD) median	N	Mean (SD) median	N	Mean (SD) median
Age, years	212	58.4 (10.2) 57.1	234	58.8 (9.1) 58.0	70	58.6 (9.6) 58.9	0.43
Female, %	212	[57.5%]	235	[64.3%]	70	[60.0%]	0.32
African ancestry proportion, %	212	0.78 (0.14) 0.8	235	0.78 (0.15) 0.81	70	0.83 (0.11) 0.84	0.02
Diabetes duration, years	211	13.4 (8.2) 11.7	233	13.0 (7.3) 11.8	70	12.3 (7.1) 11.2	0.41
Body mass index, kg/m2	212	35.2 (9.0) 33.6	235	35.5 (8.5) 34.0	70	34.6 (7.0) 33.9	0.74
History of cardiovascular disease, %	212	[22.2%]	234	[22.6%]	70	[24.3%]	0.75
Hypertension, %	207	[84.1%]	231	[88.7%]	69	[82.6%]	0.61
Current/former smoker, %	212	[55.7%]	232	[53.9%]	70	[55.7%]	0.85
Systolic blood pressure, mmHg	212	132.9 (17.8) 132.8	235	130.1 (17.3) 129.0	70	131.4 (20.5) 130.5	0.1
Diastolic blood pressure, mmHg	212	77.8 (11.5) 78.5	235	76.1 (10.9) 76.0	70	77.0 (12.1) 76.0	0.18
Statin medications, %	212	[42.0%]	234	[37.2%]	70	[35.7%]	0.24
Urine albumin:creatinine ratio, mg/g	211	157.1 (539.3) 10.2	231	148.5 (549.1) 9.1	70	301.7 (944.3) 17.3	0.33
Urine ACR >300 mg/g, %	211	[10.0%]	231	[9.1%]	70	[14.3%]	0.57
Serum creatinine, mg/dl	211	0.98 (0.3) 0.91	233	0.97 (0.3) 0.89	70	1.0 (0.23) 1.0	0.18
Estimated GFR, ml/min/1.73m2	211	89.2 (23.4) 90.0	233	89.0 (23.9) 91.0	70	82.4 (21.3) 82.0	0.09
Kidney disease2, %	211	[64.0%]	233	[59.2%]	70	[77.1%]	0.02
Fasting glucose, mg/dl	211	154.8 (64.3) 141.0	233	142.4 (60.4) 127.0	70	161.4 (70.3) 146.5	0.39
Hemoglobin A1c, %	210	8.1 (2.1) 7.4	233	7.9 (1.9) 7.4	70	8.4 (2.2) 7.8	0.83
High School or less, %	212	[45.3%]	235	[38.7%]	70	[24.3%]	1.2×10−3
Technical/AS degree, %	212	[42.0%]	235	[40.9%]	70	[52.9%]	1.2×10−3
College graduate, %	212	[12.7%]	235	[20.4%]	70	[22.9%]	1.2×10−3
Modified Mini Mental State (0-100)	212	85.6 (9.0) 87.0	235	85.5 (10.1) 87.0	70	86.2 (8.3) 87.5	0.8
Montreal Cognitive Assessment (0-30)	212	19.3 (3.9) 19.0	235	19.3 (4.0) 19.0	70	19.3 (4.1) 19.5	0.91
Digit Symbol Coding (0-135)	207	49.0 (17.1) 50.0	231	50.0 (16.0) 49.0	70	52.2 (15.7) 49.5	0.32
MRI Variables
Total Intracranial Volume, cm3	196	1305 (141) 1302	220	1292 (132) 1271	67	1292 (130) 1280	0.28
White Matter Volume, cm3	196	494.7 (68.4) 491.2	220	493.0 (62.6) 486.1	67	488.0 (62.2) 476.8	0.47
Gray Matter Volume, cm3	196	564.0 (58.7) 564.0	220	563.1 (58.9) 557.7	67	569.3 (60.9) 563.9	0.98
Cerebrospinal Fluid Volume, cm3	196	253.8 (53.9) 240.8	220	242.8 (46.3) 236.4	67	240.3 (35.0) 241.2	0.11
White Matter Lesion Volume, cm3	194	8.6 (16.3) 2.8	217	6.3 (11.1) 1.1	67	4.7 (9.7) 1.2	6.6×10−3

¹Based on the Cochran-Armitage trend test;

²kidney disease defined as eGFR <60 ml/min/1.73m² and/or UACR >30 mg/g;

ACR – albumin:creatinine ratio; GFR – glomerular filtration rate.

Table 2.

Baseline demographic, laboratory, neuropsychological testing, and MRI data in African American SPRINT participants

Variables	APOL1 0 risk alleles	APOL1 1 risk alleles	APOL1 2 risk alleles	P-value
	(N=1,065)	(N=1,143)	(N=360)
Age, years	64.4 ± 9.1	64.5 ± 9.0	63.5 ± 9.0	0.1482
Female, %	482 (45.3)	515 (45.1)	169 (46.9)	0.8149
African ancestry proportion, %	0.79 (0.71-0.86)	0.81 (0.73-0.87)	0.84 (0.77-0.88)	<0.0001
Body mass index, kg/m2	31 ± 6.5	30.7 ± 6.2	31.5 ± 6.7	0.0612
History of cardiovascular disease	181 (17.0)	178 (15.6)	60 (16.7)	0.6462
Current/Former smoker	823 (77.3)	881(77.1)	274(76.1)	0.9004
Systolic blood pressure, mmHg	139.8 ± 16.5	139.7 ± 16.3	139.4 ± 15.8	0.9011
Diastolic blood pressure, mmHg	80.9 ± 12.3	81.2 ± 12.5	82.3 ± 12.6	0.1815
Statin medications	366 (34.7)	378 (33.2)	123 (34.3)	0.769
Urine albumin:creatinine ratio, mg/g	8.7 (4.9-21.7)	8.8 (5.1-21.2)	12.2 (6.5-32.8)	<0.0001
Urine albumin:creatinine ratio >300 mg/g, %	37 (3.6)	29 (2.7)	24 (6.9)
Serum creatinine, mg/dL	1.06 (0.89-1.25)	1.05 (0.88-1.27)	1.09 (0.89-1.38)	0.0268
Estimated glomerular filtration rate, ml/min/1.73m2	76.6 ± 22.7	76.9 ± 22.5	73.3 ± 24.6	0.0283
Kidney Disease1	366 (35.5)	377 (34.6)	145 (41.4)	0.0616
Fasting glucose, mg/dL	96 (89-104)	96 (88-104)	95 (88-102)	0.3234
Education				0.3658
< High School (HS) education	157 (14.7)	163 (14.3)	66 (18.3)
HS graduate or GED	222 (20.8)	241 (21.1)	79 (21.9)
Some training beyond HS, but no college degree	449 (42.2)	460 (40.2)	149 (41.4)
College graduate	102 (9.6)	115 (10.1)	30 (8.3)
Some training beyond college degree	135 (12.7)	164 (14.3)	36 (10.0)
Montreal Cognitive Assessment (0-30)	22 (19-25)	22 (19-25)	22 (19-25)	0.9132
Digit Symbol Coding (0-135)	46 (37-56)	46 (36-57)	45 (36-56.8)	0.9476
MRI Variables	(N=83)	(N=94)	(N=20)
Total Intracranial Volume, cm3	1356.3 (1230.4-1442.1)	1349.0 (1278.6-1444.5)	1350.3 (1279.4-1399.7)	0.742
White Matter Volume, cm3	518.3 (467.8-554.5)	518.1 (482.2-551.5)	501.9 (482.2-536.2)	0.793
Gray Matter Volume, cm3	616.9 (556.1-662.0)	628.9 (590.8-668.3)	623.3 (601.5-642.2)	0.359
Cerebrospinal Fluid Volume, cm3	1.3 (1.1-1.7)	1.23 (1.1-1.6)	1.2 (1.1-1.5)	0.690
White Matter Lesion Volume, cm3	1.0 (0.2-2.3)	0.9 (0.3-2.1)	0.6 (0.2-1.5)	0.679

Numbers are n (%), mean ± standard deviation, or median (interquartile range);

¹kidney disease reflects eGFR <60 ml/min/1.73 m² and/or UACR >30 mg/g.

Table 3 displays adjusted AA-DHS MIND association results between APOL1 RRVs with MRI volumes and cognitive testing results in additive genetic models (Supplementary Table S1 displays recessive genetic models). As seen with CP associations in AA-DHS, APOL1 RRV associations with gray matter volume (GMV), cerebrospinal fluid volume (CSFV), and white matter lesion volume (WMLV) were stronger in additive models. In fully-adjusted models, APOL1 RRVs were significantly and positively associated with GMV (parameter estimate [β] 3.50×10⁻³ standard error [SE] 1.48×10⁻³; p=0.018) and negatively associated with WMLV (β −0.303, SE 0.099; p=2.33×10⁻³) and CSFV (β −30707, SE 8264; p=2.28×10⁻⁴), but not with white matter volume (WMV), Montreal Cognitive Assessment (MoCA), or modified mini-mental status examination (3MS) cognitive test performance. These associations were observed with Box-Cox transformed outcomes and correspond to adjusted effect sizes (β/SE) of 0.16 GMV, −1.208 WMLV, and −0.251 CSFV (p-values 0.028, 0.01 and 0.93, respectively). The significant association observed with the transformed CSFV outcome and APOL1 was no longer significant after applying the inverse of the transformation.

Table 3.

Relationships between cerebral volumes and cognitive testing with numbers of APOL1 renal-risk variants in AA-DHS MIND

MRI cerebral volume
Outcome	Adjustment	Transformed scale			Original scale
		Coefficient	SE	p-value	Scaled Coefficient	SE	p-value
White Matter Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	1.009	2.283	0.659	0.029	0.069	0.67
	Above + BMI, HbA1c, statins, smoking, CVD, KD and hypertension	1.059	2.314	0.648	0.031	0.07	0.661
Gray Matter Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	3.4×10−3	1.47×10−3	0.022	0.153	0.071	0.032
	Above + BMI, HbA1c, statins, smoking, CVD, KD and hypertension	3.5×10−3	1.48×10−3	0.018	0.16	0.073	0.028
CSF Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	−29472.6	8241.959	3.85 ×10−4	−0.238	2.696	0.93
	Above + BMI, HbA1c, statins, smoking, CVD, KD and hypertension	−30706.7	8263.5	2.28 ×10−4	−0.251	2.859	0.93
WM Lesion Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	−0.303	0.099	2.33 ×10−3	−0.204	0.079	9.8×10−3
	Above + BMI, HbA1c, statins, smoking, CVD, KD and hypertension	−0.303	0.099	2.33 ×10−3	−0.208	0.081	0.010
Cognitive function
Outcome	Adjustment	Coefficient			SE	p-value
MoCA	Age, sex, African ancestry proportion, and education	−0.029			0.022	0.173
	Above + BMI, HbA1c, statins, smoking, CVD, KD, and hypertension	−0.032			0.022	0.135
3MS	Age, sex, African ancestry proportion, and education	−0.047			0.037	0.198
	Above + BMI, HbA1c, statins, smoking, CVD, KD, and hypertension	−0.055			0.037	0.133
DSC	Age, sex, African ancestry proportion, and education	0.022			0.0182	0.225
	Above + BMI, HbA1c, statins, smoking, CVD, KD, and hypertension	0.023			0.0184	0.211

WM – white matter; CSF – cerebrospinal fluid; TICV – total intracranial volume; MoCA – Montreal Cognitive Assessment; 3MS – Modified Mini Mental State; DSC – Digit Symbol Coding; CVD – cardiovascular disease; KD – kidney disease (defined as eGFR <60 ml/min/1.73m² and/or UACR >30 mg/g); BMI – body mass index; HbA1c – hemoglobin A1c.

Cognitive testing was performed in 2,568 AA SPRINT participants who also had APOL1 and ancestry informative marker genotypes; MRIs were performed in 197. Among the 197 AA SPRINT participants with an MRI, 10.2% (N=20) had two APOL1 RRVs. APOL1 RRV associations with cerebral volumes in the smaller number of non-diabetic SPRINT MIND participants were non-significant but trended in the same directions as in AA-DHS MIND (Table 4). In the fully-adjusted additive genetic models, APOL1 RRVs had a non-significant positive relationship with GMV (p=0.16) and negative relationships with both WMLV (p=0.54) and CSFV (p=0.29). Supplementary Table S2 displays results with the recessive genetic model. As in AA-DHS MIND, significant associations were not observed between APOL1 RRVs and cognitive performance in SPRINT.

TABLE 4.

Relationships between cerebral volumes and neuropsychological testing with numbers of APOL1 renal-risk variants in African American SPRINT participants

MRI cerebral volume
Outcome	Adjustments	Scaled Coefficient	SE	p-value
White Matter Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	−0.0973	0.1147	0.397
	Above + BMI, statins, CVD, smoking, and KD	−0.0906	0.1174	0.440
Gray Matter Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	0.2058	0.1158	0.075
	Above + BMI, statins, CVD, smoking, and KD	0.1632	0.1173	0.164
CSF Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	−0.1132	0.1114	0.310
	Above + BMI, statins, CVD, smoking, and KD	−0.1211	0.1145	0.290
WM Lesion Volume	Age, sex, TICV, African ancestry proportion, and MRI scanner	−0.059	0.1114	0.597
	Above + BMI, statins, CVD, smoking, and KD	−0.0699	0.1146	0.542
Cognitive function
Outcome	Adjustments	Coefficient	SE	p-value
MoCA	Age, sex, education, and African ancestry proportion	0.0078	0.0126	0.536
	Above + BMI, statins, history of CVD, smoking status, and KD	0.005	0.0128	0.695
DSC	Age, sex, education, and African ancestry proportion	0.0046	0.0081	0.566
	Above + BMI, statins, history of CVD, smoking status, and KD	0.0046	0.0083	0.581

WM – white matter; CSF – cerebrospinal fluid; TICV – total intracranial volume; MoCA – Montreal Cognitive Assessment; DSC – Digit Symbol Coding; CVD – cardiovascular; KD – kidney disease (defined as eGFR <60 ml/min/1.73m² and/or UACR >30 mg/g); BMI – body mass index.

A meta-analysis was performed in the full sample of AA-DHS MIND and SPRINT participants (Table 5). Significant associations were again observed between APOL1 renal-risk variants with lower WMLV and higher GMV in this well-powered analysis.

TABLE 5.

Relationships between cerebral volumes and neuropsychological testing with numbers of APOL1 renal-risk variants in a meta-analysis including all African American AA-DHS MIND and SPRINT participants

Outcome	Covariates	AA-DHS-MIND			SPRINT			Meta-analysis
		β	SE	P-value	β	SE	P-value	β	SE	P-value	Q	P-value*	I2
MRI cerebral volume
WMV	Age, sex, TICV, African ancestry proportion, and MRI scanner	0.029	0.069	0.67	−0.1	0.11	0.4	−0.005	0.06	0.94	0.89	0.35	-
GMV	Age, sex, TICV, African ancestry proportion, and MRI scanner	0.153	0.071	0.032	0.21	0.12	0.08	0.17	0.06	0.006	0.15	0.70	-
CSFV	Age, sex, TICV, African ancestry proportion, and MRI scanner	−0.238	2.696	0.93	−0.11	0.11	0.31	−0.11	0.11	0.31	0.002	0.96	-
WMLV	Age, sex, TICV, African ancestry proportion, and MRI scanner	−0.204	0.079	9.8×10−3	−0.06	0.11	0.6	−0.16	0.06	0.02	1.13	0.29	0.11
Cognitive function
MoCA	Age, sex, African ancestry proportion, and education	−0.029	0.022	0.173	0.01	0.01	0.54	−0.001	0.01	0.91	2.11	0.15	0.53
DSC	Age, sex, African ancestry proportion, and education	0.022	0.018	0.225	0.002	0.01	0.57	0.007	0.007	0.31	0.76	0.38	-

WMV – white matter volume; GMV – gray matter volume; CSFV – cerebrospinal fluid volume; WMLV – white matter lesion volume; TICV – total intracranial volume; MoCA – Montreal Cognitive Assessment; DSC – Digit Symbol Coding; β – parameter estimate; SE – standard error

Q – Cochran’s Q;

^*P-value associated with Cochran’s Q;

I² - heterogeneity.

Finally, associations between hypertension (defined as systolic blood pressure >140 mmHg, physician report, or taking anti-hypertensive medications), eGFR, and UACR with WMLV and GMV were assessed in AA-DHS MIND. Supplementary Table S3 reveals that hypertension and albuminuria were positively associated with WMLV; eGFR was negatively associated with WMLV. In contrast, hypertension did not significantly impact GMV; eGFR and UACR were positively and negatively associated, respectively, with GMV. This is consistent with findings in a prior AA-DHS MIND report containing approximately half of the cohort.¹³ While not specific to African Americans, a recently published study of brain perfusion in SPRINT MRI participants also indicated a positive association between albuminuria and WMLV, with the highest WMLV seen in individuals with UACR >30 mg/g and eGFR <60 ml/min/1.73m².¹⁴ We note that further adjustment for hypertension, eGFR and UACR did not alter the association between APOL1 RRVs on GM and WMLV (adjusted p-values=0.03 and 4.8×10⁻⁴, respectively). In the model with further adjustment for hypertension, eGFR and UACR, the APOL1 RRVs explained 2.0% and 0.5%, of the total variation observed in WMLV and GMV, respectively.

Discussion

APOL1 G1 and G2 renal risk variants have pronounced effects on risk for non-diabetic glomerulosclerosis; odds ratios for genetic association with common forms of nephropathy range from 7.3 to 29.^1;15 It appears likely that renal gene expression, not circulating APOL1 proteins, contribute to nephropathy susceptibility based on studies in kidney transplantation and HIV-associated nephropathy.^16-21 These nephropathy risk variants also have extra-renal effects on large blood vessels where they are associated with protection from CP, a marker of subclinical CVD.^5;6 The present analyses in AA-DHS MIND (483 participants) extend these observations to the brain. As for systemic subclinical CVD, APOL1 renal risk variants were associated with lower WMLV (less severe cerebral small vessel disease) and larger gray matter volumes. In contrast to nephropathy susceptibility, APOL1 RRVs appear to have generally protective effects on the brain and cerebral vasculature. A smaller sample of 197 AA SPRINT participants was also analyzed, this sample alone was likely underpowered to show independent effects. The association between GM and APOL1 and between WMLV and APOL1 remained significant in a meta-analysis including all participants, a result likely driven by the strength of association observed in AA-DHS MIND. Results in both studies appear homogeneous, with Cochran Q=0.15, p-value=0.70 and Q=1.13, p-value=0.29, for the associations between GM and APOL1 and between WMLV and APOL1, respectively. The proportion of variance associated with heterogeneity (I²) was 0 and 0.11, respectively. In addition, few if any other studies contain AAs genotyped for APOL1 with cerebral MRI and cognitive performance data. Independent of APOL1, the literature supports inverse relationships between GMV and WLMLV in normal aging, brain metabolism, Alzheimer’s Disease, and diabetes.^22-25 Presumably, loss of cortical cells (lower GMV) relates in part to cerebrovascular disease (higher WMLV).

This AA-DHS MIND APOL1 analysis was adjusted for multiple covariates potentially associated with WMLV and other cerebral volumes. Adjustments included brain size (total intracranial volume, TICV), MRI scanner, African ancestry proportion, age, sex, blood sugar control, smoking, statins, hypertension, nephropathy, and prior CVD. APOL1 associations were independent from these established cerebrovascular disease risk factors. In addition to measuring automated WMLV readings, subjective WMLV scoring was also performed by two radiologists at the Wake Forest School of Medicine (WFSM) who were blinded to participant characteristics; the average of the two readings was considered. Subjective WMLV readings ranged from 0 (absent) to 10 (severe). These WMLV scores were also significantly lower in those with increasing numbers of APOL1 RRVs; mean (SD) median subjective WMLV scores were 2.4 (3.9) 2.5 in those with 0 RRVs; 1.7 (6.1) 2.0 with 1 RRV; and 2.2 (1.5) 2.0 with 2 RRVs, (p=0.03 additive model; p=0.10 recessive model). Thus, automated and subjectively scored WMLV results yielded consistent and significant associations. Results in SPRINT showed similar trends in non-diabetic individuals at risk for cerebrovascular disease.

APOL1 RRVs appear to impact vascular health in the white matter of the brain and in the systemic vasculature. However, whether circulating APOL1 variant proteins are protective to blood vessels or altered APOL1 gene expression in vascular cells and/or circulating blood cells (monocytes) contributes to this protection remains unknown. Inconsistent results for APOL1 genetic association with CVD have been reported.⁴ APOL1 was associated with higher risk for adjudicated CVD events in the JHS and WHI, but with improved survival (likely fewer CVD events) in AA-DHS.^5;6 SPRINT failed to detect an association between APOL1 and prevalent CVD in high risk for CVD (non-diabetic) AA participants.²⁶ One possible contributor to these discordant results is confounding by nephropathy. APOL1 was strongly associated with kidney disease in JHS and WHI; nephropathy predisposes to CVD.^27;28 In contrast, the initial AA-DHS sample lacked APOL1 association with kidney disease because it was a sample universally affected with diabetes.⁶ We note that the AA-DHS MIND sample recruited 220 participants who were not in the parent AA-DHS; AA-DHS MIND detected an APOL1 association with nephropathy (Table 1). APOL1 associations with nephropathy at baseline in SPRINT were weak based on the low frequency of two RRV carriers (14% in SPRINT vs. 13% in the general AA population) and low level proteinuria in SPRINT participants based on enrollment criteria.²⁶ Hence, confounding of CVD by nephropathy risk was far less likely in AA-DHS and SPRINT; this may have allowed for demonstration of protective or neutral effects of APOL1 on subclinical and clinical CVD. Results in these MIND studies in AAs support protective effects of APOL1 RRVs on the brain, including larger volumes of gray matter and smaller volumes of white matter lesions and CSF (suggesting less cerebral atrophy). Cognitive performance on the MoCA and 3MS did not differ in AA-DHS participants (or the meta-analysis) based upon APOL1.²⁹

WMLV in these studies corresponds to cerebral white matter lesions (WML) or leukoaraiosis on brain MRI. Cerebral hypoperfusion with resultant ischemia secondary to white matter small vessel disease is believed to be the major contributing factor in their formation.³⁰ WML are especially common in the elderly and in those with hypertension and diabetes. Pathologically, lesions demonstrate rarefaction of the white matter (axonal and myelin density decrease) and perivascular gliosis; they do not show frank infarction.³¹ There is a significant reduction in CD31 staining, a marker of endothelial cell integrity, in individuals with WML.³² Greater WMLV is further associated with decreased cognitive³³ and physical function, including gait impairment.³⁴

This is the first assessment of APOL1 effects on the brain in AAs; however, limitations exist. Despite detailed phenotypes in participants with recent African ancestry who are at high risk for cerebrovascular disease, sample sizes were modest. The lack of statistical significance within SPRINT MIND alone may relate to small numbers of AA participants; however, consistent directions of associations relative to the larger AA-DHS MIND were reassuring. Findings support APOL1 RRV associations with larger GMV and lower WMLV in AAs with and without diabetes, given significant associations in the larger AA-DHS sample with diabetes and the meta-analysis. APOL1 was associated with baseline nephropathy in both the AA-DHS MIND and SPRINT MIND; however, associations with GMV and WMLV remained significant after adjusting for kidney disease. Due to the relatively small sample size and heterogeneous underlying diseases (participants with and without diabetes), additional studies are warranted. It is critical that appropriate adjustments be made for the effects of APOL1 on kidney disease, because nephropathy may contribute to white matter disease and reductions in GMV. To address this issue, AA-DHS MIND data were re-analyzed excluding 100 participants with UACR >30 mg/g and/or eGFR <60 ml/min/1.73m²; significant associations between APOL1 RRVs with WMLV and GMV remained (data not shown). Although imaging protocols and analysis methods differed between AA-DHS MIND and SPRINT MIND, the similar direction of findings for the SPRINT MIND and the meta-analysis support the main results in this study. Although cognitive testing was performed, we lacked data on gait speed and gait disturbances in AA-DHS MIND. There is also the potential that overmatching bias, survival bias, and Berkson’s fallacy could have impacted findings.^35-37

In conclusion, APOL1 renal-risk variants are associated with lower WMLV and larger GMV in AAs. The mechanism(s) whereby APOL1 mediates protection from vascular disease in the cerebral and systemic circulation is unknown. It is opposite in direction to APOL1 association with risk for kidney disease. Additional studies on APOL1-mediated protection from cerebrovascular disease and large vessel calcified atherosclerotic plaque should be undertaken.

Methods

Participants

AA-DHS MIND

AA-DHS MIND is a single center study performed at the WFSM.^10;12;13 As reported, AA-DHS MIND recruited unrelated participants with clinically diagnosed T2D based upon age at onset >30 years in the absence of diabetic ketoacidosis, with either active diabetes treatment (insulin and/or oral hypoglycemic agents), fasting blood sugar ≥126 mg/dL, non-fasting blood sugar ≥200 mg/dL, or hemoglobin A1c (HbA1c) ≥6.5%. Hypertension was considered present if diagnosed by a physician, anti-hypertensive medications were prescribed, or clinic blood pressures were >140/90 mmHg. The study was approved by the WFSM Institutional Review Board and all participants provided written informed consent. Individuals with a known serum creatinine concentration >2 mg/dl were not recruited.

Examinations were performed in the WFSM Clinical Research Unit. Subjects had fasting blood work for measurement of chemistries, HbA1c, lipid profiles, high sensitivity C-reactive protein, and a spot urine albumin and creatinine concentration for UACR (LabCorp, Burlington, NC). Estimated GFR was computed using the Chronic Kidney Disease-Epidemiology Collaboration (CKD-EPI) equation.³⁸ After a morning snack, cognitive testing and cerebral MRI were performed.^13;39

African American SPRINT participants

SPRINT is a two-armed, multicenter, randomized, open label, clinical trial including 9,361 participants designed to test whether a strategy to treat systolic blood pressure (SBP) to <120 mmHg will reduce CVD outcomes in non-diabetic hypertensive participants compared to treating to the currently recommended SBP target <140 mmHg.⁹ The trial was stopped early, due to the significant benefit of the lower BP arm.¹¹ Participants are men and women aged ≥50 years with a baseline SBP between 130 and 180 mm Hg taking a maximum of four antihypertensive medications and with at least one additional CVD risk factor. Clinical CVD criteria included a history of myocardial infarction, acute coronary syndrome, coronary revascularization, carotid endarterectomy/stenting, and/or peripheral artery disease with revascularization. Patients with Framingham Risk Score >15, age ≥75 years, and/or CKD who met the SBP eligibility criteria were eligible for enrollment. Individuals with diabetes, proteinuria ≥1 g/day, history of stroke, eGFR <20 ml/min/1.73m², symptomatic heart failure within the past 6 months, or left ventricular ejection fraction <35% were excluded.

SPRINT Memory and Cognition in Decreased Hypertension (MIND) and SPRINT MIND MRI are nested sub-studies designed to test whether the lower SBP goal influences rates of incident dementia and mild cognitive impairment, global and domain-specific cognitive function, and cerebral small vessel ischemic disease. SPRINT MIND MRI includes 664 SPRINT participants enrolled at University of Alabama at Birmingham, Boston University, Case Western Reserve University, Miami University, University of Pennsylvania, Vanderbilt University, and Wake Forest School of Medicine. All AA SPRINT (N=2,568) and AA SPRINT MIND MRI (N=197) participants with APOL1 genotyping were included in the analyses.

Cerebral Magnetic Resonance Imaging (MRI)

AA-DHS MIND

The initial 73 scans were performed on a 1.5 Tesla (1.5T) GE EXCITE HD scanner with twin-speed gradients using a neurovascular head coil (GE Healthcare, Milwaukee, WI). Due to a change in scanners at the WFSM Center for Biomolecular Imaging, the subsequent 410 scans were performed on a 3.0 Tesla (3.0T) Siemens Skyra MRI scanner using a high resolution 20 channel head/neck coil (Siemens Healthcare, Erlangen, Germany). The imaging protocol included high resolution 3D volumetric T1-weighted images and Fluid-attenuated inversion recovery (FLAIR) images as previously described.¹⁰

Structural T1 images were segmented into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF), normalized to Montreal Neurologic Imaging (MNI) space, and modulated with the Jacobian determinants (non-linear components only) of the warping procedure to generate volumetric tissue maps using the Dartel high-dimensional warping and the SPM8⁴⁰ new segment procedure as implemented in the VBM8 toolbox (http://dbm.neuro.uni-jena.de/vbm.html). Total GM, WM and CSF volumes, and TICV (GMV + WMV + CSFV) were determined from the VBM8 automated segmentation procedure which outputs a text file with values for native space total GM, WM, and CSF volumes. All volumes were reported in cc by summing the voxel values, multiplying by the voxel volume (in mm³) and dividing by 1000.

WML segmentation was performed using the lesion segmentation toolbox (LST) toolbox⁴¹ for SPM8 at a threshold (k) of 0.25. The LST was validated in AA-DHS MIND against expert manual segmentation, as well as to identify the optimum threshold in this population.⁴² Normalization to MNI space was accomplished by co-registration with the structural T1 and applying the normalization parameters computed in the VBM8 segmentation procedure. WML volume, reported in cc, was determined by summing the binary lesion maps and multiplying by the voxel volume.

To account for between-scanner variation, 15 AA-DHS MIND participants underwent 1.5T and 3.0T MRI scans. Imaging data underwent identical processing for both datasets in each participant. Adjustments in all brain volumes except WMLV were made based on any systematic differences in volumetric measures between scanners.¹³

SPRINT MIND

The standardized MRI protocol was conducted on 3.0T scanners and included sagittal 3D FLAIR, T2-, and T1-weighted sequences with whole brain coverage. The image-analysis methodology uses an automated computer program that classifies all supratentorial brain tissue into either normal or abnormal GM or WM, and assigns the tissue type to each of 154 anatomic regions of interest (ROI) of the cerebrum. Tissue within each anatomic ROI was further classified as either normal or abnormal. T1-weighted volumetric MRI scans were first pre-processed according to a standardized protocol for removal of extra-cerebral tissue, using an in-house method.⁴³ The brain was segmented into GM, WM, and CSF using MICO, a method proposed for joint bias field estimation and tissue segmentation.⁴⁴ The lesion segmentation is done using WMLS.⁴⁵ Briefly, multiparametric MRI sequences (i.e., T1, T2 and FLAIR) are first co-registered and intensity normalized. A support vector machine classifier is trained on local features extracted from normal and abnormal tissues on these images. Voxels on a new image are classified as either lesion or normal according to the training model. An ROI segmentation method,⁴⁶ based on multi-atlas registration using DRAMMS⁴⁴ and label fusion, was used to derive regional volumetric measurements. The method partitioned the T1 image into 154 anatomical regions, which are organized within a hierarchical structure to allow derivation of volumetric measurements in various resolution levels. These ROI are organized in an anatomically hierarchal system and then collapsed into four anatomic ROI for this analysis: total brain, total GM, total WM, and Deep Gray/White Matter (DGW; operationally defined as caudate nucleus, putamen, globus pallidus, internal capsule, and thalamus).

Cognitive Testing

Interviewers in AA-DHS MIND and SPRINT MIND were trained, certified, and assessed for quality control by a single investigator in AA-DHS MIND (KMS) and a team at the SPRINT MIND Coordinating Center (including KMS). General cognitive function was assessed with the MoCA in both studies. The MoCA is a screening instrument for mild cognitive dysfunction comprised of items to evaluate memory, visuospatial abilities, executive function, attention, concentration, working memory, and language, as well as orientation to time and place.⁴⁷ AA-DHS MIND further administered the 3MS.⁴⁸ Both studies administered the Wechler Adult Intelligence Scale Digit Symbol Coding (DSC) task measuring processing speed and working memory.

Statistical Analyses

For AA-DHS MIND, calibration equations between the 1.5T and 3.0T MRI measures were estimated using robust linear regression.¹³ These equations were used to determine the corresponding 3.0T MRI value from an MRI scan performed on a 1.5T scanner. R-squared values (R²) obtained from the calibration equations also provide an estimate of the reliability of 1.5T measurements as predictors of the 3.T MRI values. Since the MRI measurements serve as the dependent variables in the model, additional variation in the calibrated 1.5T measurements do not bias the parameter estimates; they only reduce the power of the association test.

For both studies, generalized linear models were used to test for associations between the number of APOL1 RRVs (additive and recessive models were tested) with MRI and neuropsychological test outcomes. Variables derived from the brain MRI were TICV, WMLV, GMV, WMV, and CSFV. The Box-Cox method was applied to identify the appropriate transformation best approximating the distributional assumptions of conditional normality and homogeneity of variance of the residuals.⁴⁹ Because different Box-Cox transformations were applied within AA-DHS and SPRINT, the resulting model coefficients are on different transformed scales, and so are difficult to directly compare. However, to aid in interpretation, scaled coefficients $(\widehat{\beta }∕\widehat{\sigma })$, where $\widehat{\sigma }$ denotes the estimated residual standard deviation from the Box-Cox transformed regression model are presented.^50;51 For both studies, we first fit a minimally adjusted model including age, sex, TICV, African ancestry proportion, and MRI scanner as covariates. In AA-DHS MIND, a second model was then fit including body mass index (BMI), HbA1c, use of statins, hypertension, prior CVD, nephropathy defined as eGFR<60 ml/min/1.73m² and/or UACR >30 mg/g, and smoking (current/former) as additional covariates. In SPRINT, the fully-adjusted model included BMI, use of statins, prior CVD, nephropathy, and smoking as additional covariates; HbA1c was not measured in non-diabetic SPRINT participants, all of whom were hypertensive at initial randomization. Diagnostic tests performed on the model residuals showed that the assumptions underlying the GLM were met following the application of Box-Cox transformations.

To account for the discrete nature of the neuropsychological test scores and to allow for over-dispersion, negative binomial regression models were used for the MoCA, 3MS (AA-DHS MIND only), and Digit Symbol Coding tasks. The logarithm function was used to link the mean of the outcome with the predictors included in the model. Age was modeled as a cubic polynomial to allow for a non-linear association with each of the neuropsychological test scores. Diagnostic tests based on the deviance residuals were used to assess whether model assumptions were appropriate. The models used to test for the association between APOL1 RRVs and neuropsychological test scores were adjusted for age, sex, African ancestry proportion, MRI scanner and TICV. An inverse variance weighted meta-analytic approach was used to estimate the overall effect of APOL1 on the MRI and cognitive function variables. Finally, effects of hypertension (yes/no), eGFR, and UACR on cerebral volumes and cognitive testing were performed using the same methods described above.