Neuroinflammation and amyloid deposition in the progression of mixed Alzheimer and vascular dementia

Highlights

• •Neuroinflammation and amyloid beta (Aβ) deposition represent two distinct pathophysiological pathways in a cohort of elderly participants with mixed Alzheimer’s disease (AD) and vascular contributions to cognitive impairment and dementia (VCID) pathologies.
• •Neuroinflammation, but not Aβ deposition, contributes to white matter hyperintensities (WMH) baseline volume and progression.
• •Both neuroinflammation and Aβ deposition independently contribute to the progression of cognitive impairment.

Abstract

Background

Alzheimer’s disease (AD) and vascular contributions to cognitive impairment and dementia (VCID) pathologies coexist in patients with cognitive impairment. Abnormal amyloid beta (Aβ) deposition is the hallmark pathologic biomarker for AD. Neuroinflammation may be a pathophysiological mechanism in both AD and VCID. In this study, we aimed to understand the role of neuroinflammation and Aβ deposition in white matter hyperintensities (WMH) progression and cognitive decline over a decade in patients with mixed AD and VCID pathologies.

Methods

Twenty-four elderly participants (median [interquartile range] age 78 [64.8, 83] years old, 14 female) were recruited from the Knight Alzheimer Disease Research Center. ¹¹C-PK11195 standard uptake value ratio (SUVR) and ¹¹C-PiB mean cortical binding potential (MCBP) were used to evaluate neuroinflammation and Aβ deposition in-vivo, respectively. Fluid-attenuated inversion recovery MR images were acquired to obtain baseline WMH volume and its progression over 11.5 years. Composite cognitive scores (global, processing speed and memory) were computed at baseline and follow-up over 7.5 years. Multiple linear regression models evaluated the association between PET biomarkers (¹¹C-PK11195 SUVR and ¹¹C-PiB MCBP) and baseline WMH volume and cognitive function. Moreover, linear mixed-effects models evaluated whether PET biomarkers predicted greater WMH progression or cognitive decline over a decade.

Results

Fifteen participants (62.5%) had mixed AD (positive PiB) and VCID (at least one vascular risk factor) pathologies. Elevated ¹¹C-PK11195 SUVR, but not ¹¹C-PiB MCBP, was associated with greater baseline WMH volume and predicted greater WMH progression. Elevated ¹¹C-PiB MCBP was associated with baseline memory and global cognition. Elevated ¹¹C-PK11195 SUVR and elevated ¹¹C-PiB MCBP independently predicted greater global cognition and processing speed declines. No association was found between ¹¹C-PK11195 SUVR and ¹¹C-PiB MCBP.

Conclusions

Neuroinflammation and Aβ deposition may represent two distinct pathophysiological pathways, and both independently contributed to the progression of cognitive impairment in mixed AD and VCID pathologies. Neuroinflammation, but not Aβ deposition, contributed to WMH volume and progression.

1. Introduction

Vascular contributions to cognitive impairment and dementia (VCID) represents a spectrum of underlying pathological processes leading to cerebrovascular brain injury and dementia. Cerebral small vessel disease (CSVD) is the most common contributor to VCID (Gorelick et al., 2011). Vascular risk factors of CSVD include hypertension, hyperlipidemia, diabetes, and smoking (Dickie et al., 2016). On the other hand, Alzheimer’s disease (AD) is a progressive neurodegenerative disease caused by the accumulation of amyloid plaques and neurofibrillary tangles. Though the primary putative pathogeneses of these diseases are distinct, AD and VCID pathologies commonly coexist in the community-dwelling elderly (Schneider et al., 2007). Mixed pathology increases the odds of dementia by almost three times. White matter hyperintensities (WMH) of presumed vascular and ischemic origin (Kang et al., 2022) are often found in patients with AD and VCID (Alber et al., 2019, Lee et al., 2016). The pathogenic mechanisms underlying disease risk and progression are not completely understood in mixed AD and VCID.

Amyloid beta peptide (Aβ) aggregates defines a key pathological feature of AD and can be measured using ¹¹C-Pittsburgh compound B (¹¹C-PiB) PET imaging (Cohen and Klunk, 2014). Emerging literature has suggested the role of neuroinflammation in AD and VCID (Werry et al., 2019, Wu et al., 2013). Systemic serum and cerebrospinal fluid markers of inflammation, including C-reactive protein (CRP) and matrix metalloproteinases, are elevated in patients with CSVD (Mitaki et al., 2016, Rosenberg, 2017). Elevated CRP during midlife may promote late-life white matter structural injury and cognitive decline over 20 years (Walker et al., 2019, Walker et al., 2018). PET imaging using ¹¹C-PK11195, a radioligand that binds to the 18 kDa translocator protein (TSPO) located on the membrane of microglia and astrocytes, has also been used to investigate neuroinflammation in patients with AD (Wang et al., 2022) and CSVD (Walsh et al., 2021).

It has been reported that pathologic cerebral Aβ deposition is associated with activated macrophages, microglia and astrocytes as well as the release of pro-inflammatory cytokines in AD (Dorey et al., 2014) and CSVD (Rosenberg, 2017). There is a dearth of literature investigating both neuroinflammation and Aβ aggregation from the same patients in AD and VCID imaging and clinical manifestations. It is not readily discernible whether neuroinflammation and Aβ deposition are distinct or entangled pathophysiological mechanisms in patients with mixed AD and VCID pathologies. In this study, we hypothesized that neuroinflammation, manifesting as elevated ¹¹C-PK11195 standard uptake value ratio (SUVR), is associated with larger WMH volume and predicts greater WMH progression in elderly participants with a range of vascular risk factors. We also hypothesized that neuroinflammation, as an independent pathophysiological process, contributes to cognitive decline in conjunction with elevated Aβ deposition measured by ¹¹C-PiB mean cortical binding potential (MCBP).

2. Methods

2.1. Participant and data acquisition

This study is a retrospective analysis using data from twenty-four elderly participants who underwent ¹¹C-PK11195 PET, ¹¹C-PiB PET, and MRI scans at the Washington University Knight Alzheimer Disease Research Center (ADRC). This study was approved by the Institutional Review Board of Washington University School of Medicine with written informed consent.

¹¹C-PK11195 PET and ¹¹C-PiB PET images were acquired on a Siemens 962 HR + PET scanner (Siemens AG, Knoxville, TN, USA) from all participants between 2004 and 2007. Sixty-minute dynamic PET scans were acquired immediately after the tracer injection (injection dose: 11.0 [8.4, 12.1] mCi for ¹¹C-PK11195 and 10.8 [7.6, 13.0] mCi for ¹¹C-PiB).

Serial MR images were collected between 2004 and 2022. All baseline MRI were acquired on a Siemens 1.5T Vision scanner (Siemens Healthineers, Erlangen, Germany) from 2004 to 2006, and all follow-up MRI were acquired on a Siemens 3T Tim Trio, a 3T Biograph mMR, or a 3T Prisma scanner from 2009 to 2022. T1 magnetization-prepared rapid gradient-echo (MPRAGE) images were acquired from all participants (baseline 1.5T scan parameters: echo time (TE)/repetition time (TR)/inversion time (TI) = 4/9.7/20 ms; follow-up 3T scan parameters: TE/TR/TI = 3.16/2400/1000 ms). Nineteen participants underwent baseline fluid-attenuated inversion recovery (FLAIR) scans, and nine had follow-up FLAIR scans (baseline 1.5T scan parameters: TE/TR/TI = 120/10000/2600 ms; follow-up 3T scan parameters: TE/TR/TI = 76/9000/2500 ms).

The Mini-Mental State Examination (MMSE) and Clinical Dementia Rating (CDR®) scores were collected. Serial cognitive tests were performed using standard Knight ADRC cognitive battery (Johnson et al., 2008) at an approximately 12-month interval between 1987 and 2018. Clinical evaluations were performed by following the ADRC clinical core codebook (https://knightadrc.wustl.edu/wp-content/uploads/2021/07/Clinical-Core-Codebook-UDS2.pdf). The presence of vascular risk factors, including hypertension, hypercholesterolemia, diabetes, stroke or transient ischemic attack, and smoking, were determined by self-reported medical history or by medical chart review. Similar to a previous work (Gottesman et al., 2017), an overall vascular risk score, with 1 point for each risk factor, was calculated. Of note, self-reported stroke may underestimate its true incidence (Day et al., 2020).

The longitudinal study timeline is provided in Fig. 1. Individual participant’s ¹¹C-PK11195 PET scan time was used as the timepoint zero. ¹¹C-PiB PET images were acquired at −15.5 [−20, −12.5] months (negative numbers indicate that data was acquired earlier than the ¹¹C-PK11195 PET scan). Baseline MR images were acquired at −9 [−17.5, −6] months. Follow-up MR images were acquired from 9 participants 3 [3, 5] times up to 121 [100, 138] months. The baseline MMSE and CDR scores were collected at 0 [−5, 3] months. Twenty-one participants had follow-up MMSE up to 87 [57, 108] months and follow-up CDR up to 87 [62, 108] months, both at an approximately 12-month interval. The first cognitive data were acquired at −64.5 [−139.2, −31.8] months. The cognitive data closest to the ¹¹C-PK11195 PET scan were acquired at 0 [−2, 4.2] months and used as baseline cognitive data in this study. Beyond the baseline cognitive assessments, twenty-one participants had 7 [5, 9] follow-up cognitive assessments up to 88 [61, 135] months. The baseline clinical evaluation was performed at 2 [−1.2, 5.5] months. Twenty-two participants had follow-up clinical evaluations up to 76.5 [54.5, 88.2] months at an approximately 12-month interval.

Fig. 1.

Longitudinal study timeline. Timepoint zero was defined by individual participant’s ¹¹C-PK11195 PET scan.

2.2. MRI data processing

Baseline and follow-up T1 MPRAGE segmented and parcellated using Freesurfer 5.3 (Martinos Center for Biomedical Imaging, Charlestown, MA, USA), with careful quality control and manual editing with consistent criteria. WMH lesions were delineated manually by a board-certified neurologist (P. Kang) on FLAIR images using Medical Image Processing, Analysis, and Visualization software (https://mipav.cit.nih.gov). FLAIR images and lesion masks were co-registered to T1 MPRAGE images using FMRIB’s Linear Image Registration Tool (FLIRT) in the FSL toolbox (Jenkinson et al., 2002). To account for brain volume variation across participants and time, normalized WMH volume was computed as a ratio of WMH volume to brain tissue volume. The brain tissue volume was defined as the summation of cerebral GM and WM volumes. Linear regression was performed between longitudinal WMH volume and age to compute the average annual WMH progression (Table 1).

Table 1.

Participant characteristics.

	Variables	N = 24a
Demographics	Age at the time of 11C-PK11195 Scan	78 [64.8, 83]
	Female	14 (58.3%)
	African American	3 (12.5%)
	Education	16 [12 18]
Vascular risk factors	Hypertension	14 (58.3%)
	Hyperlipidemia	12 (50.0%)
	Diabetes Mellitus	3 (12.5%)
	Stroke or transient ischemic attack	2 (8.3%)
	Smoking	1 (4.2%)
	Participants with at least one vascular risk factor	20 (83.3%)
Imaging characteristics	PiB Positive (11C-PiB MCBP > 0.18)	17 (70.8%)
	Baseline WMH volume (ml)	4.06 [2.14, 19.21]
	Baseline normalized WMH volume	0.52% [0.22%, 2.30%]
	Participants with baseline WMH volume >1 ml	17/19 (89.5%)
	WMH volume progression rate (ml per year)b	0.23 [0.12, 0.42]
	Normalized WMH volume progression rate (per year)b	0.03% [0.02%, 0.04%]
	Lacunar infarct	2/21 (9.5%)
	Lobar microbleeds	1/12 (8.3%)
	Deep microbleeds	1/12 (8.3%)
	Cortical superficial siderosis	1/12 (8.3%)
	Enlarged perivascular spaces	9/15 (60%)
Baseline cognitive and functional status	MMSE	28.5 [26.8 30]
	Participants with MMSE ≥ 24	21 (87.5%)
	CDR	0 [0, 0.5]
	Participants with CDR = 0	17 (70.8%)

^aValues are shown as Median [Interquartile range] or numbers (%).

^bN = 9.

Baseline FLAIR images and WMH lesion masks were co-registered to the International Consortium for Brain Mapping atlas using Advanced Normalization Tools (Fonov et al., 2009). A WMH lesion density map was then calculated from all 19 participants. A group cumulative WMH region of interest (ROI) was defined in white matter with a lesion density greater than zero (Fig. 2A, color overlay). A group common normal appearing white matter (NAWM) ROI was defined in white matter with a lesion density of zero (Fig. 2B, blue overlay). Other widely accepted CSVD MRI features, including lacunar infarct, lobar and deep microbleeds, cortical superficial siderosis, and enlarged perivascular spaces were reviewed and summarized in Table 1 (Wardlaw et al., 2013).

Fig. 2.

Baseline WMH lesion density map (A) and group common NAWM ROI (B, marked in blue). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

2.3. PET data processing

¹¹C-PK11195 and ¹¹C-PiB PET data were reconstructed dynamically using filtered back projection algorithm with 24 × 5-second frames, 9 × 20-second frames, 10 × 1-minute frames, and 9 × 5-minute frames. After inter-frame motion correction, ¹¹C-PK11195 standardized uptake value ratio (SUVR) maps were computed from the 10–60 min frames using the median cerebellar cortex as the reference region. After co-registering to the baseline T1 MPRAGE images using FSL’s FLIRT, average ¹¹C-PK11195 SUVR was calculated in the whole brain (WB), cerebral gray matter (GM), WMH and NAWM. A recent study suggested that neuroinflammation in precuneus, posterior cingulate, lateral temporal and inferior parietal cortices had high accuracy in differentiating participants with normal cognition, mild cognitive decline, and AD (Pascoal et al., 2021). Therefore, we also evaluated ¹¹C-PK11195 SUVR in these four cortical regions defined by Freesurfer. ¹¹C-PiB data was processed using a PET Unified Pipeline (https://github.com/ysu001/PUP). The binding potential was determined by Logan plot (Logan et al., 1990) using the frames from 0 to 60 min with the cerebellar cortex as the reference region. Regional spread function-based partial volume correction was performed (Rousset et al., 2008). As a index of Aβ burden, MCBP was calculated by averaging binding potential in the prefrontal cortex, gyrus rectus, lateral temporal, and precuneus regions defined by Freesurfer (Mintun et al., 2006).

2.4. Cognitive data processing

Similar to previous works (Cutter et al., 1999, Wang et al., 2018), the raw scores of each cognitive test were first normalized to z-scores using the group mean and standard deviation of the first cognitive test scores across all participants. Three standardized composite z-scores, zGlobal, zSpeed, and zMemory, were calculated by averaging the z-scores of individual tests. zGlobal was calculated if ≥8 of 11 following tests were available: Wechsler Adult Intelligence Scale (WAIS) Block Design, WAIS Information, Category Fluency (Animals), Wechsler Memory Scale (WMS) Digits Forward, WMS Digits Backward, Boston Naming Test, Trailmaking A, Trailmaking B, WAIS-R Digit Symbol, Associate Memory (WMS Associates Recall (Easy) divided by 2+ WMS Associates Recall (Hard)), and the total free recall score from the Free and Cued Selective Reminding Test (FCSRT). zSpeed was calculated if ≥2 of 3 following tests were available: Trailmaking A, Trailmaking B, and WAIS-R Digit Symbol. zMemory was calculated as an average z-score of two memory tests, Associate Memory and the total free recall score from the FCSRT. Scores for Trailmaking A and B were inverted such that higher scores indicated better performance. Higher composite z-scores indicate better performance. Twenty-two, twenty-two, and twenty-one participants had baseline zGlobal, zSpeed and zMemory scores, respectively. Eighteen participants had 5 [4, 7] follow-up zGlobal scores. Eighteen participants had 5 [4, 7] follow-up zSpeed. Twenty-one participants had 7 [4, 9] follow-up zMemory scores. Linear regression was performed between longitudinal cognitive score and age to compute the average annual cognitive change.

2.5. Statistical analysis

Statistical analyses were performed using R 3.6.1 (Foundation for Statistical Computing, Vienna, Austria). Two-tailed significance was set to 0.05 a priori. All models were evaluated for multicollinearity using variance inflation factors (VIF).

To evaluate the association between PET imaging biomarkers of neuroinflammation (regional ¹¹C-PK11195 SUVR) and Aβ deposition (¹¹C-PiB MCBP) and baseline normalized WMH volume and cognitive function, multiple linear regression models controlling for the time difference between PET and baseline MR scans or baseline cognitive test were utilized. In addition to PET imaging biomarkers, we also evaluated the possible entry of age, sex, education, and vascular risk score into these models. For the continuous variables such as age, vascular risk score, and education, Pearson correlation between an independent variable and the dependent variable (baseline WMH volume or cognitive function) was used to determine its entry into the multi-variable models. The entry of the categorical variable sex was determined by whether the dependent variable was different between males and females using a t-test. A univariate P < 0.2 was required for entry into the multiple linear regression model. The multiple linear regression models used in this study are summarized in Table 2 and Table 4. We also evaluated whether regional ¹¹C-PK11195 SUVR was associated with ¹¹C-PiB MCBP using multiple linear regression controlling for age and time difference between the two acquisitions.

Table 2.

Multiple linear regression models between baseline WMH volume and PET biomarkers.

Model: Baseline WMH volume ∼ Regional 11C-PK11195 SUVR + Age + Vascular Risk Score
PK Regiona	PK SUVRb
Whole Brain ROI	0.281 [0.008, 0.555] *
NAWM ROI	0.197 [−0.060, 0.454]
WMH ROI	NAc
Cerebral Gray Matter ROI	0.240 [0.018, 0.462] *
Precuneus ROI	0.117 [−0.066, 0.300]
Posterior Cingulate ROI	0.086 [−0.040, 0.212]
Lateral Temporal Cortex ROI	NAc
Inferior Parietal Cortex ROI	NAc

*P < 0.05.

^aEach row presents results from a separate model evaluating the effect of ¹¹C-PK11195 SUVR in a specific region.

^bValues are shown as beta coefficient [95% confidence internal].

^cNot included in the multiple linear regression model because P > 0.2 in the univariate test.

Table 4.

Multiple linear regression models between baseline cognitive function and PET biomarkers.

Model: Baseline zGlobal ∼ Regional 11C-PK11195 SUVR + 11C-PiB MCBP + Age + Education + Vascular Risk Score
PK Regiona	PK SUVRb	PiB MCBPb
Whole Brain ROI	−0.287 [−10.388, 9.813]	−0.568 [−1.032, −0.103]*
NAWM ROI	1.267 [−7.696, 10.231]	−0.589 [−1.050, −0.128]*
WMH ROI
Cerebral Gray Matter ROI
Precuneus ROI
Posterior Cingulate ROI	NAc	−0.569 [−0.993, −0.144]*
Lateral Temporal Cortex ROI
Inferior Parietal Cortex ROI
Model: Baseline zSpeed ∼ Regional 11C-PK11195 SUVR + Age + Education + Vascular Risk Score
PK Regiona	PK SUVRb
Whole Brain ROI	−5.873 [−21.245, 9.499]
NAWM ROI	−3.533 [−17.378, 10.312]
WMH ROI	NAc
Cerebral Gray Matter ROI	−5.351 [−18.828, 8.126]
Precuneus ROI
Posterior Cingulate ROI	NAc
Lateral Temporal Cortex ROI
Inferior Parietal Cortex ROI
Model: Baseline zMemory ∼ Regional 11C-PK11195 SUVR + PiB MCBP + Age
PK Regiona	PK SUVRb	PiB MCBPb
Whole Brain ROI	−5.075 [−16.009, 5.859]	−0.798 [−1.327, −0.268]**
NAWM ROI	−3.550 [−13.990, 6.889]	−0.804 [−1.343, −0.266]**
WMH ROI
Cerebral Gray Matter ROI
Precuneus ROI
Posterior Cingulate ROI	NAc	−0.818 [−1.360, −0.276]**
Lateral Temporal Cortex ROI
Inferior Parietal Cortex ROI

*P < 0.05.

**P < 0.01.

^aEach row presents results from a separate model evaluating the effect of ¹¹C-PK11195 SUVR in a specific.

^bValues are shown as beta coefficient [95% confidence internal].

^cNot included in the multiple linear regression model because P > 0.2 in the univariate test.

Linear mixed-effects models with an intercept only random effect for participant, controlling for sex, education, and vascular risk score were used to evaluate whether PET imaging biomarkers predicted greater WMH progression and cognitive decline. Normalized WMH volumes or cognitive scores acquired at baseline and all follow-up visits were included in these longitudinal analyses. The mixed-effects models are summarized in Table 3 and Table 5. Models were assessed for normality quantitatively with the Shapiro-Wilk and Anderson-Darling tests and graphically using the histogram and QQ-plot of model residuals. The interaction between regional ¹¹C-PK11195 SUVR and age and the interaction between ¹¹C-PiB MCBP and age were investigated as the primary independent variables.

Table 3.

Longitudinal prediction of WMH progression using mixed-effects models with a random effect for participant.

Model: WMH volume ∼ Regional 11C-PK11195 SUVR × Age + 11C-PiB MCBP × Age + Sex + Vascular Risk Score
PK Regiona	PK SUVR × Ageb	PiB MCBP × Ageb
Whole Brain ROI	0.030 [0.004, 0.054] *	−0.001 [−0.002, 0.001]
NAWM ROI	0.014 [−0.004, 0.033]	−0.000 [−0.002, 0.001]
WMH ROI	−0.004 [−0.018, 0.010]	−0.001 [−0.003, 0.001]
Cerebral Gray Matter ROI	0.027 [0.010, 0.044] **	−0.000 [−0.002, 0.001]
Precuneus ROI	0.013 [0.002, 0.024] *	−0.001 [−0.003, 0.000]
Posterior Cingulate ROI	0.009 [−0.000, 0.017]	−0.001 [−0.002, 0.001]
Lateral Temporal Cortex ROI	0.012 [−0.002, 0.027]	−0.000 [−0.002, 0.002]
Inferior Parietal Cortex ROI	0.006 [−0.004, 0.016]	−0.001 [−0.003, 0.001]

*P < 0.05.

**P < 0.01.

^aEach row presents results from a separate model evaluating the effect of ¹¹C-PK11195 SUVR in a specific region.

^bValues are shown as beta coefficient [95% confidence internal].

Table 5.

Longitudinal prediction of cognitive decline using mixed-effects models with a random effect for participant.

Model: zGlobal ∼ Regional 11C-PK11195 SUVR × Age + 11C-PiB MCBP × Age + Sex + Education + Vascular Risk Score
PK Regiona	PK SUVR × Ageb	PiB MCBP × Ageb
Whole Brain ROI	−0.840 [−2.302, 0.622]	−0.155 [−0.222, −0.088] ***
NAWM ROI	−0.491 [−1.780, 0.798]	−0.153 [−0.230, −0.077] ***
WMH ROI	−0.228 [−1.002, 0.546]	−0.163 [−0.230, −0.097] ***
Cerebral Gray Matter ROI	−0.461 [−1.453, 0.530]	−0.172 [−0.236, −0.108] ***
Precuneus ROI	−0.629 [−1.249, −0.008] *	−0.164 [−0.227, −0.100] ***
Posterior Cingulate ROI	0.301 [−0.238, 0.840]	−0.171 [−0.234, −0.107] ***
Lateral Temporal Cortex ROI	−0.959 [−1.866, −0.052] *	−0.184 [−0.249, −0.119] ***
Inferior Parietal Cortex ROI	−0.958 [−1.498, −0.419] ***	−0.161 [−0.222, −0.101] ***
Model: zSpeed ∼ Regional 11C-PK11195 SUVR × Age + 11C-PiB MCBP × Age + Sex + Education + Vascular Risk Score
PK Regiona	PK SUVR × Ageb	PiB MCBP × Ageb
Whole Brain ROI	−2.512 [−4.621, −0.404] *	−0.253 [−0.361, −0.146] ***
NAWM ROI	−2.100 [−3.627, −0.572] **	−0.239 [−0.352, −0.126] ***
WMH ROI	−1.062 [−2.224, 0.100]	−0.282 [−0.388, −0.177] ***
Cerebral Gray Matter ROI	−1.106 [−2.738, 0.526]	−0.291 [−0.396, −0.186] ***
Precuneus ROI	−0.809 [−1.788, 0.169]	−0.275 [−0.382, −0.168] ***
Posterior Cingulate ROI	−0.025 [−0.847, 0.798]	−0.308 [−0.414, −0.202] ***
Lateral Temporal Cortex ROI	−1.770 [−3.122, −0.418] *	−0.316 [−0.420, −0.212] ***
Inferior Parietal Cortex ROI	−1.200 [−2.062, −0.337] **	−0.254 [−0.356, −0.153] ***
Model: zMemory ∼ Regional 11C-PK11195 SUVR × Age + 11C-PiB MCBP × Age + Sex + Education + Vascular Risk Score
PK Regiona	PK SUVR × Ageb	PiB MCBP × Ageb
Whole Brain ROI	−0.753 [−2.284, 0.777]	−0.010 [−0.077, 0.058]
NAWM ROI	−0.391 [−1.632, 0.850]	−0.002 [−0.077, 0.073]
WMH ROI	−0.222 [−1.029, 0.584]	−0.001 [−0.069, 0.067]
Cerebral Gray Matter ROI	−0.524 [−1.673, 0.626]	−0.035 [−0.112, 0.042]
Precuneus ROI	−0.543 [−1.263, 0.177]	−0.049 [−0.127, 0.029]
Posterior Cingulate ROI	0.194 [−0.372, 0.760]	−0.034 [−0.111, 0.042]
Lateral Temporal Cortex ROI	−0.685 [−1.657, 0.286]	−0.038 [−0.115, 0.040]
Inferior Parietal Cortex ROI	−1.029 [−1.653, −0.405] **	−0.054 [−0.126, 0.018]

*P < 0.05.

**P < 0.01.

***P < 0.001.

^aEach row presents results from a separate model evaluating the effect of ¹¹C-PK11195 SUVR in a specific region.

^bValues are shown as beta coefficient [95% confidence interval].

3. Results

3.1. Participant characteristics

Participant characteristics are summarized in Table 1. Twenty participants (83.3%) had at least one vascular risk factor. Seventeen participants (70.8%) were PiB positive (¹¹C-PiB MCBP > 0.18) (Mintun et al., 2006). Fifteen participants (62.5%) had both a positive PiB scan and at least one vascular risk factor. The baseline WMH volume was 4.06 [2.14, 19.21] ml, corresponding to normalized WMH volume of 0.52% [0.22%, 2.30%]. Seventeen of nineteen participants (89.5%) had >1 ml WMH volume. Enlarged perivascular spaces were found in 60% of participants, while lacunar infarct, microbleeds, and cortical superficial siderosis were less frequent at baseline. Among participants with follow-up clinical evaluations (N = 22), one, three, and two participants developed hypertension, hyperlipidemia, Diabetes Mellitus, respectively. One participant became a smoker. Twenty-one (87.5%) and seventeen (70.8%) participants had normal baseline MMSE (>=24) and CDR (=0), respectively (Lezak et al., 2004). Among participants with follow-up CDR (N = 21), eight participants had CDR progression from 0 to 0.5 (N = 1), from 0 to 1 (N = 3), from 0 to 2 (N = 3), and from 0.5 to 2 (N = 1). Thirteen participants had no change (CDR = 0: N = 10, CDR = 0.5: N = 3). Among participants with follow-up MMSE (N = 21), four participants progressed from normal MMSE (>=24) to abnormal MMSE (<24). As shown in Table 1, this study cohort has a wide presence of mixed AD and VCID pathologies. The baseline WMH lesion density map and group common NAWM ROI are shown in Fig. 2. WMH lesion density was the highest in the periventricular and deep white matter regions, consistent with previous findings (Kang et al., 2022).

3.2. Association between PET biomarkers and baseline and longitudinal WMH volume

Fig. 3 shows baseline FLAIR images and PET ¹¹C-PK11195 SUVR maps from two representative participants with WMH volumes of 1.89 ml (Fig. 3A) and 45.44 ml (Fig. 3B). The WB ¹¹C-PK11195 SUVR were 0.94 (Fig. 3C) and 1.07 (Fig. 3D), respectively. The multiple linear regression between PET imaging markers and WMH volume is summarized in Table 2. ¹¹C-PiB MCBP did not meet the criteria for entering the multiple linear regression model. ¹¹C-PK11195 SUVR in WB (β = 0.281, P = 0.044) and GM (β = 0.240, P = 0.036) were associated with WMH volume after controlling for age and vascular risk score.

Fig. 3.

Baseline MR FLAIR images (A, B) and PET ¹¹C-PK11195 SUVR maps (C, D) from two representative participants. WMH volumes were 1.89 ml and 45.44 ml, and the whole brain ¹¹C-PK11195 SUVR were 0.94 and 1.07 in Participant 1 and Participant 2, respectively. ¹¹C-PK11195 SUVR was higher in participants with larger WMH volume.

The longitudinal progression of WMH is demonstrated in Fig. 4A. WHM volume increased by 0.23 [0.12, 0.42] ml per year, corresponding to normalized WMH volume of 0.03% [0.02%, 0.04%] per year over 11.5 years (Fig. 4A). The association between PET biomarkers and longitudinal WMH progression is summarized in Table 3. Elevated WB ¹¹C-PK11195 SUVR together with age (β = 0.030, P = 0.027), elevated GM ¹¹C-PK11195 SUVR together with age (β = 0.027, P = 0.004) and elevated precuneus ¹¹C-PK11195 SUVR together with age (β = 0.013, P = 0.028) respectively predicted greater WMH progression, while elevated ¹¹C-PiB MCBP together with age did not.

Fig. 4.

Longitudinal WMH progression (A) and cognitive decline (B-D). Different colors indicate data from different participants. Solitary gray circles indicate participants who had only one baseline measurement. Normalized WMH volume increased by 0.03% [0.02%, 0.04%] per year over 11.5 years (A). Global, processing speed and memory function decreased by 0.10 [0.01 0.32], 0.15 [0.02, 0.59], and 0.07 [0.02, 0.30] per year over 7.5 years, respectively (B-D).

VIFs were smaller than 2.1 for all independent variables in all baseline and longitudinal models, demonstrating no multicollinearity in the analyses.

3.3. Association between PET biomarkers and baseline and longitudinal cognition

The multiple linear regression between PET biomarkers and baseline cognitive function is summarized in Table 4. ¹¹C-PiB MCBP but not ¹¹C-PK11195 SUVR was associated with baseline zGlobal (P < 0.05) and zMemory (P < 0.01) in multiple linear regression models. Neither regional ¹¹C-PK11195 SUVR nor ¹¹C-PiB MCBP was associated with baseline zSpeed.

The longitudinal decline of cognitive function is demonstrated in Fig. 4 (B-D). Global, processing speed and memory function decreased by 0.10 [0.01 0.32], 0.15 [0.02, 0.59], and 0.07 [0.02, 0.30] per year over 7.5 years, respectively (Fig. 4(B-D)). The associations between PET biomarkers and longitudinal cognitive decline are summarized in Table 5. Elevated precuneus ¹¹C-PK11195 SUVR together with age (β = -0.629, P = 0.050), elevated lateral temporal cortex ¹¹C-PK11195 SUVR together with age (β = −0.959, P = 0.041), and elevated inferior parietal cortex ¹¹C-PK11195 SUVR together with age (β = −0.958, P < 0.001), respectively, predicted greater zGlobal decline independent of¹¹C-PiB MCBP together with age (P < 0.001). Elevated WB ¹¹C-PK11195 SUVR together with age (β = −2.512, P = 0.021), elevated NAWM ¹¹C-PK11195 SUVR together with age (β = −2.100, P = 0.008), elevated lateral temporal cortex ¹¹C-PK11195 SUVR together with age (β = −1.770, P = 0.011), and elevated inferior parietal cortex ¹¹C-PK11195 SUVR together with age (β = −1.200, P = 0.007), respectively predicted greater zSpeed decline independent of ¹¹C-PiB MCBP together with age (P < 0.001). Elevated ¹¹C-PK11195 SUVR in inferior parietal cortex together with age (β = −1.029, P = 0.002) predicted greater zMemory decline independent from ¹¹C-PiB MCBP together with age (P = 0.145).

VIF was <2.7 for independent variables in all baseline and longitudinal models, demonstrating no multicollinearity in the analyses.

3.4. Association between ¹¹C-PK11195 SUVR and ¹¹C-PiB MCBP

Fig. 5 shows scatter plots of ¹¹C-PK11195 SUVR from various ROIs and ¹¹C-PiB BP in the MC ROI (MCBP). Regional ¹¹C-PK11195 SUVR was not associated with ¹¹C-PiB MCBP. Notable, ¹¹C-PK11195 SUVR was not correlated with ¹¹C-PiB in the same MC region (Fig. 5D).

Fig. 5.

Scatter plots between regional ¹¹C-PK11195 SUVR and ¹¹C-PiB MCBP. ¹¹C-PK11195 SUVR was not associated with ¹¹C-PiB MCBP in all regions.

4. Discussion

Neuroinflammation is emerging as an important pathomechanism in AD and VCID, and may have direct implications in the development of potential therapeutics. Chronic inflammation may result in a pathologic cascade, ultimately resulting in inappropriate inflammatory infiltration and tissue destruction. Post-mortem studies show inflammatory changes such as microgliosis and astrogliosis in patients with vascular cognitive impairment in the late stages of CSVD (Gouw et al., 2011, van Veluw et al., 2022). Low et al. found neuroinflammation, measured by regional ¹¹C-PK11195 binding was associated with various CSVD markers in a cohort of participants consisting of healthy controls, mild AD patients, and amyloid-positive MCI patients (Low et al., 2021). Malpetti et al. demonstrated that elevated anterior temporal ¹¹C-PK11195 binding predicts cognitive decline in patients with symptomatic Alzheimer’s disease pathology (Malpetti et al., 2020). While several large-cohort studies have shown an association between systemic inflammatory biomarkers and white matter injury or cognitive decline over 20 to 30 years (Walker et al., 2019, Walker et al., 2018), other studies did not find this association (Metti et al., 2014, Weuve et al., 2006). These mixed findings are likely owing to differences in age, vascular risk factors, disease severity, inflammatory markers investigated, and possible heterogeneity in disease mechanisms (Ridker, 2004). Our study used ¹¹C-PK11195 PET to measure regional neuroinflammation directly, which is more specific than the systemic inflammatory markers. We found elevated ¹¹C-PK11195 SUVR across the whole brain in participants with higher WMH volume, suggesting widespread neuroinflammation.

AD and VCID pathologies commonly coexist in elderly patients with cognitive impairments (Schneider et al., 2007). Post-mortem studies show CSVD pathology in autopsy-confirmed AD (Bangen et al., 2015). It is exceedingly difficult to clinically differentiate between AD and VCID pathomechanisms (Cipollini et al., 2019, Rosenberg et al., 2019). It is crucial that we understand the extent to which AD and VCID pathomechanisms contribute toward the neuroimaging manifestations and cognitive impairment in a personalized manner as targeted therapeutics are developed (Cipollini et al., 2019, Rosenberg et al., 2019). We investigated the roles of neuroinflammation and Aβ deposition in the progression of WMH and cognitive decline in participants with a range of vascular risk factors and Aβ deposition. Neuroinflammation, but not pathologic Aβ deposition, was associated with baseline WMH volume and predicted its progression over 11.5 years. Our results suggest that neuroinflammation is closely linked with imaging abnormalities in patients with mixed AD and VCID pathologies. Moreover, Aβ deposition, but not neuroinflammation, was associated with baseline global cognitive and memory dysfunctions. Both neuroinflammation and Aβ deposition independently predicted cognitive decline over 7.5 years after controlling for age, education, and vascular risk factors. Notably, there was no association between ¹¹C-PK11195 SUVR and ¹¹C-PiB MCBP. Taken together, our findings suggest that neuroinflammation and pathologic Aβ deposition represent distinct pathophysiological mechanisms, and both contribute to disease progression in AD and VCID mixed etiology cognitive impairment. We did not find an association between neuroinflammation and Aβ deposition, suggesting that these two processes act independently from each other in disease progression. Immunohistochemical studies in post-mortem brains from AD patients show that although clustered microglial cells can be observed in and around Aβ plaques, the presence of activated microglia is not consistently associated with Aβ deposition (Haga et al., 1989, Streit, 2005). On the other hand, in-vivo PET studies show positive (Dani et al., 2018), negative (Yokokura et al., 2011) or no (Kreisl et al., 2016) association between neuroinflammation and Aβ deposition in AD or mild cognitive impairment. Of note, these studies focused on AD pathology and did not investigate potential vascular-related risk factors and pathologies. Our findings support the independent roles of neuroinflammation and pathological Aβ deposition in mixed AD and VCID.

Our study has several limitations. Firstly, the sample size of this study was relatively small (N = 24). Moreover, only nine participants had longitudinal FLAIR scans, and only 21 participants had longitudinal cognitive data. Secondly, ¹¹C-PK11195, which is a first-generation PET tracer to measure neuroinflammation, has a low signal-to-noise ratio and high non-specific binding (Chauveau et al., 2008). Notably, we did not find an association between NAWM or WMH ¹¹C-PK11195 SUVR and WMH lesion progression, which may be partially due to the low signal-to-noise ratio in white matter. The second- and third-generation TSPO tracers have improved specific binding but suffer large inter-individual variability in binding affinity due to TSPO genetic polymorphisms (Owen et al., 2011). More recently, PET tracers binding to other targets involved in neuroinflammatory response beyond TSPO have been developed. For example, ¹¹C-CS1P1 is a tracer targeting sphingosine-1-phosphate receptor subtype 1, a cell membrane G-protein coupled receptor that plays an essential signaling role in the inflammatory response (Jiang et al., 2021). Thirdly, the baseline MRI was acquired using a 1.5T scanner while the follow-up MRI was acquired using a 3T scanner. The use of 1.5T scanner for baseline and 3T scanner at follow-up scans was not ideal but inevitable for a longitudinal study over more than a decade since MRI scanners might be upgraded over the course of the study. The FLAIR imaging protocols are the same for all the 3T scans. The WMH lesions were manually delineated by a board-certified neurologist. It is possible that the WMH obtained at baseline and follow-up may be affected by the use of two different MR fields. We expect small variations for all the 3T images despite three different 3T MR scanner models being used. Finally, the usage of different Freesurfer versions and pipelines may lead to different segmentation results, especially in small cortical ROIs. Due to low signal-to-noise ratio in the ¹¹C-PK11195 PET, results in small ROIs need to be interpreted with caution. Future prospective studies with a larger sample size and using improved tracers are needed to further elucidate neuroinflammation’s role in mixed etiology dementia and aim for the development of relevant therapeutics.

5. Conclusions

In conclusion, we demonstrate that neuroinflammation and Aβ deposition represent two distinct pathophysiological pathways in a cohort of elderly participants with mixed AD and VCID pathologies. Neuroinflammation plays a role in the development of WMH early in the course of the disease. Both neuroinflammation and Aβ deposition independently contribute to the progression of cognitive impairments.

Funding

This work was supported by grants from the National Institutes of Health (NIH): P01AG026276 (J.C.M.), P01AG003991 (J.C.M.), P30AG066444 (J.C.M.), 1R01AG054567 (T.L.B.), R01HL129241 (A.L.F.), 1R01NS082561 (H.A.), 1P30NS098577 (Imaging Core, H.A.), RF1NS116565 (H.A. and A.L.F.), R21NS127425 (H.A.), R01NS085419 (J.M.L.), U24NS107230 (J.M.L.), KL2TR002346 (P.K.), R03AG072375 (Q.W.) and R01AG074909 (Q.W.). Additional support was generously provided by the Charles and Joanne Knight Alzheimer’s Research Initiative and by the Fred Simmons and Olga Mohan Fund and the Paula and Rodger Riney Funding.

CRediT authorship contribution statement

Chunwei Ying: Data curation, Investigation, Methodology, Software, Visualization, Writing – original draft. Peter Kang: Investigation, Methodology, Writing – original draft. Michael M. Binkley: Investigation, Visualization, Writing – review & editing. Andria L. Ford: Investigation, Writing – review & editing. Yasheng Chen: Investigation, Writing – review & editing. Jason Hassenstab: Investigation, Writing – review & editing. Qing Wang: Writing – review & editing. Jeremy Strain: Writing – review & editing. John C. Morris: Writing – review & editing. Jin-Moo Lee: Writing – review & editing. Tammie L.S. Benzinger: Conceptualization, Writing – review & editing. Hongyu An: Conceptualization, Funding acquisition, Project administration, Supervision, Writing – original draft.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data will be made available on request.