Alzheimer’s disease (AD) is the most common form of neurodegenerative dementia and a significant source of caregiver morbidity and healthcare burden. The neuropathological hallmarks of AD consist of extracellular deposits of β-amyloid-containing neuritic plaques, as well as intracellular neurofibrillary tangles derived from aggregated, abnormally hyperphosphorylated tau proteins. Besides the aforementioned plaques and tangles, AD is also characterized by synaptic loss (Selkoe, 2002), dysregulated neuroinflammation (Wyss-Coray and Rogers, 2012) and neuronal degeneration (Scheltens et al., 2016). Additionally, cerebrovascular disease (CeVD) has been suggested to play a pathogenic role in dementias including AD (Kling et al., 2013). Several neuroimaging markers of CeVD, such as cortical infarct, lacunes and especially white matter hyperintensities (WMH) are associated with AD and may contribute to disease deterioration (Jellinger, 2002; Moghekar et al., 2012; Toledo et al., 2013; Attems and Jellinger, 2014). Given the prolonged prodromal stages where potential disease-modifying therapies may more likely be efficacious (Jack et al., 2013), the availability of reliable, easily accessible diagnostic and prognostic biomarkers is essential for optimal clinical management, and will also help advance and assess new therapeutic strategies for AD. While cerebrospinal fluid (CSF) and positron emission tomography (PET)-based biomarkers are available for AD pathology (β-amyloid and tau; Villemagne et al., 2017; Zetterberg, 2017) which may be associated with small vessel CeVD (Kester et al., 2014), these are relatively costly and invasive, and research efforts to uncover blood-based biomarkers are ongoing (Henriksen et al., 2014; Zetterberg, 2017).

Hepatocyte growth factor (HGF, also known as scatter factor), originally isolated from liver and known to play a role in liver regeneration (Nakamura et al., 1984), has recently been shown to also regulate various brain functions, including axon outgrowth, neuronal survival, synaptic function and plasticity (Ebens et al., 1996; Nakamura and Mizuno, 2010; Wright and Harding, 2015). While previous studies reported increased HGF levels in the brain and CSF of patients with AD (Fenton et al., 1998; Tsuboi et al., 2003), it is unclear whether peripheral HGF may be related to cognitive impairment and dementia. Furthermore, HGF can regulate vascular functions including angiogenesis (Xin et al., 2001; Morishita et al., 2002), and has also been shown to reduce ischemia-associated infarct volume (Tsuzuki et al., 2001; Shimamura et al., 2004), perhaps by preventing delayed neuronal death (Miyazawa et al., 1998). However, no study has yet examined the association of serum HGF in patients with prodromal dementia (cognitive impairment no dementia, CIND), AD, or concomitant CeVD. Hence, in the present study we examined the association of serum HGF levels with MRI markers of CeVD and cognitive impairment/AD in a memory clinic population.

A total of 383 subjects (95 NCI, 164 CIND, 124 AD) were recruited between August 2010 and May 2014, of whom 66 (16 NCI, 23 CIND and 27 AD) had insufficient blood samples and seven (3 CIND, 4 AD) had no neuroimaging (none, incomplete, or ungradable MRI), thus leaving 310 (79 NCI, 138 CIND and 93 AD) with available data for analyses. Table 1 shows the demographic and disease factors of the study cohort. Compared with NCI, CIND and AD were older, and had higher prevalence of diabetes mellitus, hypertension and cardiovascular disease. Serum HGF concentrations in the cohort ranged from 58.4 pg/mL to 3020.0 pg/mL. As shown in Table 1, levels of HGF were highest in AD and lowest in NCI, with intermediate levels in CIND. The differences reached statistical significance between NCI vs. AD (P = 0.007, Kruskal-Wallis ANOVA with post hoc Dunn’s tests). Table 2 shows the multivariate analyses of associations between serum HGF levels and cognitive impairment. Higher levels of HGF were significantly associated with both CIND and AD after correction for age, gender, APOE ε4 carrier, hypertension, diabetes, and cardiovascular diseases. Finally, Table 3 demonstrates the associations between HGF and MRI markers of CeVD. Higher levels of HGF were significantly associated with higher counts of lacunes, cortical microinfarcts and microbleeds, as well as more severe WMH. In contrast, there was no association between HGF and cortical infarct counts.

In this study, we reported higher levels of HGF in both CIND and AD, independent of demographic covariates and cardiac risk factors. These findings are in line with previous reports of higher HGF levels in the brain and CSF of patients with AD (Fenton et al., 1998; Tsuboi et al., 2003), although it is at present unclear whether peripheral HGF changes reflect HGF variations in the central nervous system (CNS). Additionally, the increased HGF were associated with CeVD in CIND and AD, but interestingly, these associations were restricted to markers of small vessel (lacunes, cortical microinfarcts, microbleeds, WMH), but not large vessel (cortical infarct), disease. Previous studies have shown that small vessel disease substantially contributes to long term morbidity in AD and constitutes an important therapeutic target (Arvanitakis et al., 2016; van Veluw et al., 2017). As they affect cerebral arterioles, venules as well as capillaries, detection and quantification of small vessel disease burden remain challenging on neuroimaging (Wardlaw et al., 2013), and the potential availability of complementary blood-based markers for small vessel CeVD may thus be a useful tool for long term management. Interestingly, the persistence of these associations even after controlling for cardiovascular risk factors suggest distinct interactions between HGF and the cerebrovasculature. Furthermore, the specificity of associations between HGF and small vessel CeVD suggests an etiological link between HGF and small, but not large, vessel disease, although the precise underlying mechanisms are at present unclear. Besides its beneficial effect on axon outgrowth, neuronal survival, synaptic function and plasticity (Ebens et al., 1996; Nakamura and Mizuno, 2010; Wright and Harding, 2015), HGF has been reported to attenuate cerebral ischemia-induced blood-brain barrier (BBB) disruption and loss of tight junction protein expression, both of which are features of small blood vessels of the brain (Date et al., 2006). Furthermore Knowland et al. (2014) showed that ischemia-induced tight junction breakdown associated with BBB disruption, but only at the level of small venules. The increased levels of HGF seen in SVD-associated cognitive impairment and AD cases in our study may thus indicate adaptive plasticity responses to small-vessel mediated neuronal damage or degeneration. However, follow-up studies are needed to validate HGF correlations with small vessel CeVD in AD, as well as their associations with other biomarkers, for e.g., those of BBB disruption and ischemic damage.

The strengths of this study include: (1) a relatively large cohort; (2) inclusion of covariates such as age, gender, APOE ε4 carrier, hypertension, diabetes and cardiovascular diseases in analyses to reduce confounding; and (3) use of neuroimaging platforms to detect and quantify CeVD. However, this study has several limitations as well. First, whilst our finding of HGF association with CIND suggests that HGF changes may be an early feature of AD, it is not possible to establish a temporal association between HGF changes and the development of cognitive impairment and AD because of the cross-sectional design of the study. Follow-up longitudinal studies can help address this limitation as well as elucidate the prognostic utility of serum HGF in predicting the course of CeVD and cognitive decline. Second, while we specifically focused on HGF’s association with CeVD in CIND and AD, it is important to follow-up with similar association studies on vascular or post-stroke dementias in order to further evaluate the nature of HGF changes in response to CeVD, and hence its diagnostic utility. Third, the NCI control group was relatively younger and has less burden of CeVD compared to case groups which could have resulted in selection bias and residual confounding. Lastly, whilst the dementia cases have been assessed using consensus criteria for clinical AD diagnosis (NINCDS-ADRDA), there is a lack of neuropathological confirmation at present.

In conclusion, the present study suggests that serum HGF has potential clinical utility as a peripheral biomarker of small vessel CeVD in cognitive impairment and AD. However, follow-up longitudinal studies are needed for further validation.

YZ, CPC and MKPL: conception and design; YZ and SH: development of methodology; YZ, SH, MKI, NV and CPC: acquisition of data; YZ, YLC and SH: analysis and interpretation of data; YZ, YLC, SH and MKPL: writing of manuscript; CPC and MKPL: study supervision. All authors have read, revised and approved the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.