Non-Alcoholic Fatty Liver Disease,Liver Fibrosis,and Regional Amyloid-β and Tau Pathology in Middle-Aged Adults: The Framingham Study

INTRODUCTION

Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver condition, affecting approximately 25% of the population in the developed world [1]. Metabolic dysregulation, including type 2 diabetes, insulin resistance, and hyperlipidemia are closely associated with NAFLD in a bidirectional manner [2]. Thus, NAFLD prevalence may reach  68% in individuals with type 2 diabetes [3] and up to 80% in those with morbid obesity [4]. Evidence suggests that NAFLD may be directly related to multiple extra hepatic health conditions [5], which include measures of vascular dysfunction [6–8], and cardiovascular morbidity and mortality [9, 10]. Furthermore, an increasing amount of evidence suggests that NAFLD is independently associated with measures of brain health. Indeed, NAFLD has been previously related to smaller brain volume in the Framingham Study’s Offspring cohort [11] as well as in other populations [12]. In addition, individuals with NAFLD demonstrated poorer cognitive function [13] and reduced brain activity [14]. In contrast, other studies have failed to identify a link of NAFLD with cognitive function [15] and incident dementia [16].

The natural history of NAFLD is diverse and can include various stages of liver fibrosis [17]. Histologically, NAFLD encompasses a broad range of pathologies ranging from simple steatosis, with no or minimal inflammation, to nonalcoholic steatohepatitis that is characterized by necroinflammation and an increased progression of fibrosis [18]. Similarly to NAFLD, liver fibrosis is often clinically silent, and is present in up to 9% of individuals without known liver disease [19]. It is increasingly recognized that liver fibrosis, rather than the existence of hepatic steatosis per se, is a strong prognostic factor for long-term complications including liver-related outcomes, cardiovascular mortality [20] and stroke [21]. Recent evidence also stresses the possible important role of liver fibrosis in cognitive function and dementia risk [22, 23].

Despite the growing support for the implications of NAFLD and liver fibrosis to brain health, the connection with Alzheimer’s disease (AD) is unknown. In vivo quantification of amyloid-β (Aβ) and tau deposition are valid biomarkers for prodromal AD and can emerge decades prior to AD clinical diagnosis [24]. Thus, in the current study, we utilized data from PET imaging to examine the association of NAFLD and liver fibrosis with regional Aβ and tau deposition. We hypothesized that NAFLD and liver fibrosis will be associated with increased Aβ and tau PET retention in brain regions in which AD pathology initially emerges, and that these associations will be independent of cardio-metabolic measures.

METHODS

Study sample

The study sample is based on participants from the Offspring [25] and third generation [26] of the Framingham Heart Study (FHS). Figure 1 presents a flow chart of the study sample. We included a total of 5,841 participants, of them 2,430 Offspring and 3,411 third generation, who attended exams 9 (2011–2014) and 2 (2008–2011), respectively. Of these, 4,991 participated in the multi-detector CT 2 sub-study for evaluation of ectopic fat, including liver fat, between September 2008 and December 2011 or had information on all components of the liver fibrosis score (FIB-4). We excluded 803 participants with excessive alcohol consumption defined as self-report of >14 alcoholic drinks/week for men and >7 alcoholic drinks/week for women because we were interested in non-alcoholic fatty liver disease. Thus, 4,188 were eligible for inclusion, among them PET imaging was obtained for a representative sample (with respect to vascular risk) of 230 eligible participants. Of them, 169 (73%) had information on NAFLD, 177 (77%) had information on all components of FIB-4, and 116 (50%) had information on both NAFLD and all components of FIB-4. Eligibility for the PET imaging sub-study included absence of significant neurological conditions including clinical stroke, dementia, and multiple sclerosis. As previously described, FHS participants undergo routine cognitive screening and comprehensive monitoring for continual surveillance of dementia [27], which was exclusionary for participation in the PET imaging sub-study and therefore also in the current study. Data were obtained under a protocol approved by the institutional review board of the Boston University Medical Center, and written informed consent was obtained from all participants.

OPEN IN VIEWER

Fig. 1

Flow chart of the study sample.

RESULTS

The characteristics of the total study sample (N = 230) and of those who were not included in our analyses are presented in Supplementary Table 1, and the characteristics of subgroups with NAFLD, fibrosis and both conditions are presented in Supplementary Table 2. The age of the participants in the study sample who had information on NAFLD or fibrosis (n = 230) was 50±10 years, and 112 (49%) were women. The prevalence of NAFLD was 34% (57 out of 169). Intermediate or high risk for advanced fibrosis (FIB-4 > 1.3) was present in 19% of the total sample with FIB-4 information (34 of 177) and in 17% in those with prevalent NAFLD (7 out of 41).

Table 1 presents the characteristics of participants with and without NAFLD. Those with NAFLD were more likely to be men and to have diabetes, insulin resistance, and hypertension. In addition, compared to individuals without NAFLD, those with NAFLD had higher BMI and visceral adipose tissue volume, as well as increased systolic blood pressure and higher levels of total and HDL cholesterol, serum C-reactive protein and ALT and AST. There were no significant differences in liver fibrosis index between those with and without NAFLD (Table 1). Additionally, individuals with intermediate or high risk for advanced fibrosis (FIB-4 > 1.3) were older, had lower C-reactive protein levels, and as expected, had higher ALT and AST, lower platelets levels and higher fibrosis score compared to those with lower risk for advanced fibrosis (Supplementary Table 3).

DISCUSSION

Our study explored the association of NAFLD and non-invasive liver fibrosis index with Aβ and tau deposition in the brain. We observed no associations between NAFLD and brain Aβ and tau. However, in the total sample, high risk for advanced liver fibrosis was significantly related to tau deposition in the rhinal brain region. Among individuals with NAFLD, advanced liver fibrosis was related to tau pathology in the inferior temporal, parahippocampal, entorhinal and rhinal regions as well as to Aβ deposition overall and in the inferior temporal and parahippocampal brain regions.

In line with our findings, a recent review of the available evidence for cognitive dysfunction in NAFLD concluded that despite insufficient evidence on the link between the whole NAFLD spectrum and cognitive dysfunction, simple steatosis may not be an independent risk factor for cognitive dysfunction and that a more severe NAFLD, with involvement of fibrosis and hepatitis may be required to affect cognition [13]. This includes a previous study from our group, which found no significant association between NAFLD and cognitive function, yet risk for advanced liver fibrosis assessed using the NAFLD fibrosis score was associated with poorer executive function and abstract reasoning among individuals with NAFLD [23]. Findings are also consistent with regard to dementia risk. Indeed, a cohort study with histological data demonstrated no association between NAFLD and incident dementia, yet, histological indicators of fibrosis improved dementia risk prediction beyond that of conventional dementia risk factors [16]. Accordingly, a community-based cohort study conducted in Germany found no significant link between NAFLD and all-cause or vascular dementia [49], while in the Italian longitudinal study, NAFLD fibrosis score among the general population (i.e., not restricted to NAFLD patients) was related to increased dementia risk in those who were physically frail. Although the latter studies did not assess NAFLD and fibrosis together, they imply that liver fibrosis rather than NAFLD per se is a risk factor for dementia. Lastly, few studies exist showing that the presence of white matter hyperintensities is associated with the fibrosis severity among patients with NAFLD [50, 51], which again, consistent with our report, highlights the importance of liver fibrosis to brain health.

The pathophysiology of both NAFLD and AD are complex and multi-factorial, and the mechanistic links between them are speculative. Moreover, it should be noted that the assessment of liver fibrosis in our study is based on a non-invasive score that may indicate liver function in general rather than liver fibrosis specifically. Because the liver is responsible for activation, clearance, and processing of multiple molecules, there may be various mechanisms in which liver dysfunction may affect the brain. For example, liver damage may lead to change in cholesterol catabolism through its conversion to primary bile acids, which in turn may be linked with increased AD risk, both in animals [52] and humans [53].

One major hypothesis for the link between NAFLD and dementia to date was that dementia risk in NAFLD is driven primarily by vascular forms. This hypothesis was based on studies showing a direct association of NAFLD, and particularly NAFLD with advanced fibrosis, with subclinical vascular damage including atherosclerosis, endothelial dysfunction, arterial stiffness, vascular calcification [54–56], and cerebral small vessel disease [50, 51]. However, recent explorations of AD temporal dynamics suggest that accumulation of Aβ and tau precedes cerebral small vessel disease [57, 58]. Thus, the strong link between fibrosis and AD pathology in our study may highlight the additional contribution of other pathophysiological mechanisms. One possible pathway for the link of liver fibrosis with Aβ and tau deposition in our study may be through systemic inflammation, which characterizes advanced liver fibrosis in particular [59]. Indeed, a network clustering and pathways enrichment pointed to the Interleukin signaling pathway as a key pathway underlying both NAFLD and AD [60]. Furthermore, according to preclinical studies, systemic inflammation may increase dementia risk not only through elevation in atherosclerosis and cerebral small vessel disease [20, 61], but also by triggering neuroinflammation that promotes cerebral Aβ accumulation directly [62, 63]. Further support for this hypothesis arrives from the Atherosclerosis Risk in Communities (ARIC) - PET Study, which demonstrated a link between systemic inflammation and cerebral Aβ deposition, albeit only in specific race and sex groups [64]. Moreover, a recent study among the Alzheimer’s Disease Neuroimaging initiative (ADNI) participants with mild cognitive impairment demonstrated an association between elevated inflammatory markers and tau, but not Aβ pathology [65]. These results are in accordance with the link of liver fibrosis with tau but not Aβ pathology in our general sample, and may highlight the importance of inflammation as an underlying mechanism. An additional explanation for our findings may be that hepatic dysfunction and insulin resistance lead to insufficient Aβ clearance in the blood, possibly through reduced expression of Low-density lipoprotein receptor–related protein 1 (LRP1) [66], which in turn contributes to brain Aβ accumulation [67]. Lastly, toxic metabolites produced in the injured liver may cross the blood brain barrier and lead to Aβ and tau pathology. For example, ceramides, bioactive sphingolipids, are activated in the course of NAFLD [68] and liver fibrosis [69], and were recently linked with incident AD and Aβ load on PET in the Framingham Study [70].

We show that in the general sample (i.e., not restricting to those with NAFLD), higher risk of advanced fibrosis is coupled with tau but not with Aβ deposition. Although previous studies demonstrated that amyloidosis is required for the subsequent elevation of tauopathy [71, 72], recent PET assessments confirm autopsy studies [73] by highlighting the contribution of the initial medial temporal tau deposition to AD natural history [40]. Specifically, the first signal of tau PET is thought to appear independently of Aβ burden in the rhinal cortex, which in our study showed strong association with FIB-4 index. Subsequent to the tau deposition in the rhinal cortex, tau is thought to spread to temporal neocortex but this stage is dependent on the primary tau accumulation and on global Aβ load [40]. Thus, our findings may advance the literature by showing that liver fibrosis is linked with early markers of AD, upstream to the development of vascular pathology and Aβ aggregation. Furthermore, tau deposition in these brain regions are highly correlated with early AD clinical phenotypes [72, 74–76].

Focusing on a subsample with confirmed NAFLD yielded a small sample size and therefore limited power to adjust for the whole set of confounders. However, after adjustment for age, sex, time between exposure and PET and camera, we found additional associations of advanced fibrosis with overall Aβ load as well as in the inferior temporal and parahippocampal regions. Due to between-studies heterogeneity in Aβ detection techniques (e.g., PET versus autopsy) as well as in population characteristics and study settings, the spatiotemporal ordering of Aβ deposition is not fully understood [47]. Yet, recent PET studies suggest that Aβ accumulates initially in the medial frontal and cingulate regions [47] which, in our sensitivity analysis, were not related to liver fibrosis. In contrast, overall Aβ load and its deposition in brain regions that were linked with liver fibrosis in our study (e.g., middle temporal) are considered to accumulate in relatively later AD stages [77, 78]. However, these accumulations of Aβ may be needed for the observed spread of tau beyond the middle temporal lobe to neocortical regions [40, 72].

The inconsistency in results between the general sample and the NAFLD subsample (i.e., associations of FIB-4 with tau in the first and with tau and Aβ in the latter) may imply that Aβ deposition is influenced by a synergistic effect of liver steatosis and fibrosis. Alternatively, the different findings may reflect the differential capacity of the FIB-4 index to non-invasively identify advanced fibrosis in these two samples. Of note, FIB-4 has been extensively studied in ethnically diverse NAFLD populations where it showed good ability to discriminate advanced versus non-advanced fibrosis [79, 80]. On the other hand, it may have limited ability to estimate risk of advanced fibrosis in the general population [81]. In addition, the FIB-4 cutoff for advanced fibrosis has been determined among persons with NAFLD [45], which is in line with our findings that dichotomous FIB-4 using this cutoff is related to AD markers only in the NAFLD subsample. Thus, despite the small sample of participants who had both PET and NAFLD information, the significant link between FIB-4 and both Aβ and tau may more specifically indicate the consequences of advanced liver fibrosis.

The strengths of our study include the predominantly middle-aged sample with information on Aβ and tau PET imaging, the CT-based ascertainment of NAFLD and the well-characterized cohort of individuals with a wide variety of metabolic and lifestyle covariates, including visceral adipose tissue. We also acknowledge several limitations of our study: first, this is a cross-sectional design that does not allow inference on temporal relationship between liver conditions and AD pathology. Second, liver fibrosis was assessed using a non-invasive index with limited validity compared to liver biopsy as the gold standard, and we lacked imaging information such as magnetic resonance elastography or ultrasound elastography (i.e., FibroScan) [82]. However, FIB-4 has the advantage of being a simple tool composed of routinely collected biochemical variables. Third, we lacked information on fibrosis etiology other than NAFLD such as hepatotoxic medications, viral hepatitis, autoimmune liver disease, or alcohol abuse (due to underreporting of alcohol consumption). A fourth limitation is the restricted external validity of our results as the sample was predominantly of European ancestry, from one geographic area and of a relatively high socioeconomic status. Lastly, the statistical power was limited due to small number of persons in some subgroup analyses, and residual confounding could exist due to incomplete adjustment.

In conclusion, our study highlights the importance of liver fibrosis severity to extrahepatic health conditions by demonstrating a link between liver fibrosis severity and early stages of AD. If validated in other studies, these findings suggest that FIB-4 may improve risk stratification models in AD and may help identifying individuals at risk for tau and Aβ accumulation, who may benefit from preventive strategies. The public health implications of these findings may be particularly significant because liver fibrosis is prevalent and yet often underdiagnosed. In addition, liver fibrosis can be managed through lifestyle modifications as well as through existing therapeutics and numerous drugs that are currently under development [18]. Future investigations are necessary to explore whether therapeutic strategies that target liver fibrosis can lead to decreased AD burden and to clarify the underlying mechanisms linking liver fibrosis to AD.