Causal Effects of Plasma Haptoglobin Levels on Alzheimer’s Disease: A Two-Sample Mendelian Randomization Study

Abstract

Background:

A connection between plasma levels of haptoglobin (Hp) and Alzheimer’s disease (AD) has been shown in several observational studies. It is debatable, nonetheless, how the two are related causally.

Objective:

To establish the causal relationship between Hp and AD using a two-sample Mendelian randomization (MR) study.

Methods:

From the extensive genome-wide association studies and FinnGen dataset, summaries and statistics pertaining to AD were gathered. We investigated the possibility of a causal link between Hp and AD using a two-sample MR study. Inverse variance weighting was used as the primary analytical technique, and it was supported by the joint application of complementary analyses and fixed effects meta-analysis to combine results from various sources.

Results:

Genetically determined Hp was causally associated with AD [odds ratio (OR), 1.05; 95% confidence interval (CI), 1.02 to 1.09; p = 8.96×10^–4]; Inverse variance-weighted estimates coming from different data sources were combined in a meta-analysis with consistent findings (OR, 1.03; 95% CI, 1.01 to 1.05; p = 2.00×10^–3). The outcomes of the inverse MR analysis showed that AD had no appreciable causal impact on Hp.

Conclusion:

The present MR analysis shows that higher plasma Hp leads to an increased risk of AD. Strategies for plasma Hp testing may open up new doors for the early diagnosis and prevention of AD.

Keywords

Alzheimer’s disease causality haptoglobins Mendelian randomization analysis

INTRODUCTION

Alzheimer’s disease (AD) is a neurodegenerative disease that leads to progressive cognitive decline and progressive behavioral disorders. Its main pathological features are senile plaques caused by extracellular amyloid-β (Aβ) deposition and neurofibrillary tangles caused by excessive phosphorylation of tau protein which is the most abundant microtubule-associated protein.

Although the etiology of AD is yet unknown, the theory that oxidative stress and inflammatory mechanisms can trigger AD progression has been widely accepted [1, 2]. Haptoglobin (Hp) is one of the acute phase response proteins and is widely present in the circulatory system with a variety of functional properties, including antioxidant and anti-inflammatory activities [3, 4]. In recent years, several observational studies have shown elevated levels of Hp in the serum and cerebrospinal fluid of AD patients compared to healthy controls [5 –7]. In a case-control study of sporadic AD conducted by Philbert et al., it was observed that Hp levels were higher than normal in most regions of the brain among patients with sporadic AD [8].

Even though there is a wealth of literature supporting the substantial link between Hp and AD, their perspectives largely revolve around conventional observational epidemiology. Because it is difficult to collect data, there are few implementation options, and observational studies are typically not randomized, the results are not yet conclusive proof of a cause-and-effect relationship between Hp and AD. Mendelian randomization (MR) is a cutting-edge and potent research technique that eliminates reverse causation [9], avoids prior external confounding, and is a more economical bioinformatics approach [10]. It has lately become a dependable new methodological tool for epidemiological investigations, based on the theory of employing genetic variables as metrics to validate causal links between exposure factors and outcome indicators [11]. Here, we performed a two-sample MR study to carry out the initial data analysis. We initially performed data analysis using a two-sample MR study, which makes use of genetic datasets from two separate populations, to establish the causal relationship between Hp and AD. Additionally, we repeated the main analysis using datasets from different sources. This study has a theoretical basis and clinical translational value, and the findings may provide a reference to guide the development of preclinical AD risk prediction and treatment tools.

MATERIALS AND METHODS

Study design

We defined single nucleotide polymorphisms (SNPs) as instrumental variables (IVs) [12] and selected circulating Hp levels for MR analysis to investigate the causal relationship of Hp with AD. In our study, three basic assumptions must be met for any MR analysis to be reliable (Fig. 1) [13]. Every piece of information used in this study came from publicly accessible databases. All initial research received informed permission from all study participants as well as ethical approval from the relevant medical ethics boards. Additionally, this study precisely followed the STROBE-MR statement’s guideline for reporting observational studies in epidemiology using MR (Supplementary Table 1) [14].

Fig. 1

Diagrammatic representation of the Mendelian randomization investigation. SNPs, single-nucleotide polymorphisms; AD, Alzheimer’s disease; IVs, instrumental variables; Hp, haptoglobin; MR, Mendelian randomization.

Data sources and IVs selections

Based on 997 people of European origin from the Cooperative Health Research in the Region of Augsburg (KORA) public database, which includes age, sex, and body mass index as variables, we retrieved the SNPs with Hp-related phenotypes for this MR investigation (Table 1) [15]. Some researchers have suggested selecting more than 10 independent SNPs as IVs for MR analysis to maintain sufficient statistical power [16]. Therefore, the criteria of p < 5×10^–6, linkage disequilibrium r² < 0.1, and clump = 250 kb have been set to obtain an adequate number of SNPs. 23 SNPs were strongly connected to Hp out of the 95 SNPs that were substantially linked to Hp at the Bonferroni significance level. We further used PhenoScanner V2 [17] to test whether the above SNP loci were associated (p < 5×10^–8) with risk factors for AD. SNPs related to LDL [18] (rs2000999, rs217181, rs8044476, rs9302635), diabetes [19], inflammation [20], height [21] (rs4788808), hypertension [22], and body mass index [23] will be specifically excluded. Then, we computed R² for each SNP using the previously mentioned approach to account for the amount of variation [24]. For each SNP, we assessed the power of the remaining SNPs using the F statistics [F = R²(N-2)/(1-R²)], where N represents the sample size of the GWAS conducted for the SNP-Hp association [25]. The result files contained information on every single SNP. In this study, no SNPs associated with the outcome were found (p < 5×10^–8) [26], and proxy SNPs were not used. In this, two loci having palindromic sequences were disregarded (rs17286411, rs2266943). Finally, all selected SNPs had minor allele frequencies greater than 0.01 (Supplementary Table 2).

Table 1

Detailed information of studies and datasets used for analyses

Data source		Phenotype	Sample size	Cases	Case Mean AAO	Controls	Control Mean AAE	Population
KORA study		Hp	997	/	/	/	/	European
IGAP	ADGC	AD	28,990	14,428	71.1	14,562	76.2	European
	CHARGE	AD	15,611	2,137	82.6	13,474	76.7	European
	EADI	AD	8,871	2,240	75.4	6,631	78.9	European
	GERAD	AD	10,454	3,177	73.0	7,277	51.0	European
FinnGen		AD	211,678	2,191	/	209,487	/	European

KORA, Cooperative Health Research in the Region of Augsburg; Hp, haptoglobin; IGAP, International Genomics of Alzheimer’s Project; AD, Alzheimer’s disease; ADGC, Alzheimer Disease Genetics Consortium; CHARGE, Cohorts for Heart and Aging Research in Genomic Epidemiology Consortium; EADI, The European Alzheimer’s Disease Initiative; GERAD, Genetic and Environmental Risk in AD/Defining Genetic; AAO, Age at onset; AAE, Age at an examination or last follow-up.

We drew on pooled clinically diagnosed late-onset AD genome-wide association studies (GWAS) data from a GWAS meta-analysis of the International Genomics of Alzheimer’s Project (IGAP) stage 1 discovery study, which reported more than 10 million SNPs in 21,982 AD patients with clinical or autopsy-documented and 41,944 controls (Table 1) [27]. The GWAS dataset can be found in the GWAS Catalog (https://www.ebi.ac.uk/gwas/). Participants were sourced from multiple populations, as detailed in Table 1. Using the FinnGen consortium’s results (https://www.finngen.fi/), which yielded full genomic variant data for 500,000 Finnish residents, the study was then replicated. The FinnGen database can be accessed and downloaded from the IEU OpenGWAS project (https://gwas.mrcieu.ac.uk/) using the GWAS ID “finn-b-F5_ALZHDEMENT”. The diagnostic criteria for AD were defined using the “F00*” codes in the International Classification of Disease-10 (ICD-10), and the “20910” codes in ICD-8. The dataset included 2,191 cases and 209,487 controls from European populations.

Mendelian randomization analysis

The TwoSampleMR [28], MendelianRandomization [29], and MR-PRESSO [30] algorithm packages were used to run all analyses in R version 4.1.1. The inverse of the result variance was employed for each SNP as the weight in the study’s primary statistical model, inverse variance weighting (IVW), which was used to examine the causality between exposure (Hp) and outcome (AD). This traditional approach would be regarded as the most statistically reliable if the MR assumptions were satisfied [11]. The IVW estimations may be skewed by up to 90% of MR analyses, according to simulation studies, which have demonstrated that pleiotropy can affect MR analysis. To account for various multi-effects patterns and to evaluate the reliability of the findings, this study used a variety of MR analyses, including MR-Egger regression [31], weighted median [32], simple mode [33], and simple median [34]. MR-Egger analysis was used to examine its intercept for potential polyvalence levels [35]. If the requirement that null SNPs do not account for more than 50% of the weight of the MR effect estimates is fulfilled, the weighted median technique yields consistent effect estimates [32].

The MR-PRESSO overall test [30] and the PhenoScanner V2 [17] (as described above) were used to further examine horizontal pleiotropy as well as the MR-Egger analysis. Primarily, any pleiotropic effects were statistically tested using the MR-Egger intercept test. The MR-IVW approach’s heterogeneity was then evaluated using Cochran’s Q test and quantified using the I² statistic [36]. The IVW fixed effects model was largely applied where there was no potential for horizontal pleiotropy heterogeneity [11]. In the presence of heterogeneity, a random effects model was chosen [37]. Identification of SNPs indicating inverse association was investigated using preliminary MR-Steiger filtering [38] (Supplementary Table 3). Eventually, after several tests, we set a two-sided p-value of less than 0.025 (= 0.05/2 outcomes) as the significance threshold, and calculated statistical power using the web tool (https://cnsgenomics.com/shiny/mRnd/) [39].

RESULTS

The exposure and outcome samples did not overlap because they were obtained from two different consortia (Table 1). We identified 16 distinct SNPs by screening the aforementioned IVs. These 16 SNPs are estimated to account for 71.38% of the diversity in Hp phenotypes (Table 2). All instrumental factors’ F-statistics were more than 10 (ranging from 20 to 161), which showed that the SNPs discovered had high power and that the biases caused by weak instrumental variables had no effect on the subsequent MR analysis (Table 2).

Table 2

Characteristics of instrumental variables for Hp and their association with AD

SNP	Hp			AD-IGAP			AD-FinnGen			R²(%)	F
	Beta	SE	p	Beta	SE	p	Beta	SE	p
rs10976376	–0.28	0.06	3.50E-06	–0.03	0.02	0.14	0.00	0.04	0.96	2.24	22.83
rs11646091	–0.75	0.13	1.20E-08	–0.03	0.04	0.42	0.05	0.06	0.40	4.70	49.09
rs11648622	–0.61	0.05	8.50E-36	–0.04	0.02	0.01	–0.05	0.04	0.20	13.94	161.10
rs12149769	0.72	0.07	1.70E-24	0.06	0.02	0.01	0.02	0.07	0.74	10.81	120.65
rs12935091	0.54	0.11	2.00E-06	0.03	0.04	0.45	–0.04	0.08	0.57	2.28	23.24
rs17604185	–1.07	0.14	2.70E-15	0.03	0.05	0.49	–0.22	0.19	0.24	3.58	36.97
rs212165	0.47	0.08	1.40E-09	–0.01	0.02	0.76	0.06	0.07	0.37	4.17	43.25
rs2550043	–0.38	0.07	1.10E-07	–0.05	0.02	0.04	–0.04	0.06	0.52	2.47	25.19
rs28402078	–0.89	0.19	1.90E-06	–0.10	0.06	0.08	–0.01	0.06	0.81	4.43	46.13
rs35063109	–0.53	0.11	1.90E-06	–0.05	0.04	0.26	–0.07	0.09	0.45	1.99	20.22
rs3803704	–0.26	0.05	9.00E-07	–0.03	0.02	0.06	0.03	0.04	0.41	2.52	25.77
rs4238965	–0.37	0.06	2.10E-10	0.00	0.02	0.98	0.00	0.04	0.92	4.23	43.94
rs7187638	–0.59	0.08	1.70E-12	–0.07	0.03	0.02	–0.17	0.07	0.02	4.86	50.82
rs8046337	0.28	0.05	2.10E-09	0.03	0.02	0.08	–0.06	0.04	0.08	3.49	35.96
rs9556683	0.26	0.05	1.20E-06	–0.02	0.02	0.26	0.03	0.04	0.41	2.57	26.26
rs9568631	0.41	0.08	5.30E-07	–0.04	0.03	0.10	0.10	0.05	0.03	3.09	31.73

Hp, haptoglobin; SNP, single-nucleotide polymorphism; Chr, chromosome; EA, effect allele; OA, other allele; EAF, effect allele frequency; MAF, minor allele frequency; R², percentage of the variation of asthma explained by the SNP; Beta, estimate of the effect of the association; SE, standard error; IGAP, International Genomics of Alzheimer’s Project; AD, Alzheimer’s disease; F, F statistic.

According to the IVW method’s findings, the IGAP database’s elevated plasma Hp levels predicted by genetic variations were linked to a higher risk of AD (Fig. 2). Odds statistical analysis based on the FinnGen dataset returned a similar direction with wider CIs (OR = 1.01; 95% CI = 0.96 to 1.06; p = 0.76; Figs. 2 3). The primary results were made more reliable by the fact that the results of the meta-analysis that combined the IGAP and FinnGen datasets also passed several Bonferroni adjustments (OR = 1.03; 95% CI = 1.01 to 1.05; p = 2.00×10^–3; Fig. 3). The general direction of the six analytical techniques remained consistent for the causal link of Hp on AD, despite some of the supplemental analytical methods’ findings falling short of the p-value of 0.025 and having wider CIs (Table 3).

Fig. 2

Scatter plot of the MR estimates for the association of serum Hp levels with risk of AD based on IGAP (A) and FinnGen (B). The lines of different colors in the figure represent the results of different MR analysis methods. The slope represents the MR estimate of Hp on AD risk. Each point represents an SNP, and the horizontal and vertical lines intersecting at each point display the 95% confidence interval for each SNP. IGAP, International Genomics of Alzheimer’s Project; AD, Alzheimer’s disease; Hp, haptoglobin; MR-PRESSO, MR Pleiotropy Residual Sum and Outlier.

Fig. 3

Association of genetically predicted serum Hp levels with risk of AD. IGAP, International Genomics of Alzheimer’s Project; AD, Alzheimer’s disease; Hp, haptoglobin; OR, odds ratio; CI, confidence interval.

Table 3

Association of genetically predicted circulating Hp levels with AD risk in complementary analyses

Data source	Outcome	SNPs, n	Method	OR	95% CI	p
IGAP	AD	16	IVW (fixed effects)	1.05	1.03, 1.08	2.27×10^–05
		16	Weighted median	1.08	1.04, 1.12	1.56×10^–04
		16	MR Egger	1.06	0.98, 1.15	1.78×10^–01
		16	Simple mode	1.11	1.04, 1,18	8.16×10^–03
		16	Simple median	1.08	1.04, 1.12	3.13×10^–05
		16	MR-PRESSO^*	1.05	1.02, 1.09	4.65×10^–03
FinnGen	AD	16	IVW (random effects)	1.01	0.95, 1.17	0.79
		16	Weighted median	1.04	0.97, 1.13	0.27
		16	MR Egger	0.99	0.84, 1.16	0.89
		16	Simple mode	1.06	0.95, 1.19	0.29
		16	Simple median	1.02	0.95, 1.10	0.61
		16	MR-PRESSO^*	1.01	0.95, 1.07	0.80

IGAP, International Genomics of Alzheimer’s Project; AD, Alzheimer’s disease; Hp, haptoglobin; SNPs, single-nucleotide polymorphisms; IVW, inverse-variance weighted; MR-Egger, Mendelian randomization-Egger; MR-PRESSO, MR-pleiotropy residual sum and outlier; CI, confidence interval; The odds ratio (OR) represents the increase in genetic prediction for each 1-standard deviation (SD) increase in exposure. ^*No outliers detected.

Table 4

Evaluation of heterogeneity and directional pleiotropy using different methods

Data source	Outcome	Heterogeneity		Pleiotropy
		IVW Cochran’s Q	MR-Egger	MR-Egger	MR-PRESSO
		p	Cochran’s Q	intercept	global test
			p	p	p
IGAP	AD	0.058	0.041	0.867	0.098
FinnGen	AD	0.145	0.111	0.793	0.166

IGAP, International Genomics of Alzheimer’s Project; AD, Alzheimer’s disease; IVW, inverse-variance weighted; MR-Egger, Mendelian randomization-Egger; MR-PRESSO, MR-pleiotropy residual sum and outlier.

A series of sensitivity analyses supported that the causal effect of Hp and AD was robust. Regarding IVW and MR-Egger regression statistics revealed no significant heterogeneity among SNPs (p > 0.05; Table 4). However, Cochran’s Q test statistics indicated a slight heterogeneity between the genetic predictive SNPs based on the Hp levels and AD from the IGAP dataset (p < 0.05, Table 4). In contrast, no heterogeneity was observed in the Finngen dataset (p > 0.05; Table 4). For the objects exhibiting heterogeneity, we selected a random-effects model to analyze the causal relationship. The MR-Egger intercept and 0 did not differ statistically significantly (p > 0.05; Table 4), and the results of the MR-PRESSO global test were also not statistically significant (p > 0.05; Table 4). Thus, it is reasonable to believe that the observed SNPs do not exhibit horizontal pleiotropy. In addition, no outliers were visible in the MR-PRESSO test data. The MR-Steiger filtering test results are presented in Supplementary Table 3, and no SNPs with inverse correlation were discovered. This study demonstrated sufficient power for the IGAP Consortium AD, while for the FinnGen Consortium AD, the estimated power of this study was 80% when the expected OR was 1.079 (or 0.929) (Supplementary Table 4).

DISCUSSION

AD is the most frequent cause of dementia in older people and is extremely dangerous to health [40]. There are not any sensitive and specific early diagnostic methods or criteria for AD, which is a crafty and irreversible disease. When the disease has often advanced to the middle and late stages, patients rely on clinical symptoms and imaging signs to make a diagnosis. The greatest strategy to lower AD morbidity and mortality, given its epidemiological features, is by early avoidance of its risk factors and early identification based on useful biomarkers.

Hp is now often found in human bodily fluids, where it binds to free hemoglobin in the blood with a high affinity and is strongly linked to, among other things, inflammatory and oxidative stress reactions in the body. For the first time, a two-sample MR method was utilized to predict plasma Hp levels genetically. Our results demonstrate that a higher plasma Hp level is linked to a higher risk of AD. Our funding indicates that for each 1-standard deviation increase in plasma Hp, there is a 3% increased risk of AD (OR, 1.03; 95% CI, 1.01 to 1.05; p = 2.00×10^–3). In several complementary analyses, the positive outcomes were stronger and generally consistent. No horizontal pleiotropy was found. To address the observed heterogeneity in Cochran’s Q statistics during our bidirectional MR analysis, we employed the IVW random-effects method as our primary MR approach. This approach helps mitigate biases in the results caused by heterogeneity.

In various neurobehavioral tests, it was discovered that Hp-deficient mice had fewer neurological impairments [41]. Another case-control investigation revealed that in individuals with sporadic AD, Hp was found to be considerably higher in the majority of brain areas after death [8]. By combining multidimensional liquid chromatography and gel electrophoresis, a newly developed study method discovered higher Hp in AD patients as compared to controls [42]. Wang et al. obtained similar results by capillary electrophoresis immunoassay and laser-induced fluorescence detection techniques [43]. Zhu et al. showed an Area Under the Curve of 0.921 (p < 0.001, 95% CI = 0.873–0.969) for Hp in the blood of AD patients based on a subject operating characteristic curve analysis, and these findings are consistent with the results obtained by our MR [7]. However, negative results were obtained in some previous observational studies [44] with results that could be attributed to residual confounders or reverse causality [45].

It is worth noting that the neuroinflammatory response is the common thread of pathogenesis of various neurological diseases [46]. Hp, as a widespread inflammatory response protein, has been reported by many pieces of evidence. Along with AD, it is also elevated in those who have Parkinson’s disease, Huntington’s disease, schizophrenia, and other neurological diseases [47 –49]. Thus, it may not be a specific marker for AD but may be a useful indicator of the inflammatory status of the disease and has potential early predictive value for AD in the clinical setting. Interestingly, Hp is a pleiotropic protein, which has also been reported to be amyloid-associated [50]. Hp is associated with Aβ plaque deposition in AD patients and Hp may affect cellular uptake and clearance of Aβ by preventing Aβ binding and/or by reducing its aggregation [51]. Spagnuolo et al. provide evidence that Hp is involved in promoting the binding of Apolipoprotein E and Aβ in cells, thereby limiting Aβ neurotoxicity and promoting its clearance [52]. Although our findings provide strong evidence for the existence of a causal relationship between elevated circulating levels of Hp and increased risk of AD, exactly how Hp affects the development of AD remains to be determined. Therefore, further clinical and fundamental research is required to interpret these results in the future.

This study using MR structures provides some benefits. First, the rule of segregation, which is the foundation of MR analysis, may successfully obviate possible confounders and reverse causation. Second, a larger sample of GWAS was employed in this work, and several sample sizes were merged for the meta-analysis, enabling a better overall analysis of AD occurrences. To make sure that its theories are solid, this MR study also passed acceptable and stringent tests of horizontal pleiotropy and heterogeneity. The current study does, however, have several drawbacks. First, although eliminating the prejudice associated with ethnic bias, our study was only conducted on people of European ancestry, which raises questions about its application to populations of other races. Second, previous studies have indicated that Hp increases with age in both humans and experimental animals [53, 54]. Age is closely associated with the onset and progression of AD, and it may serve as an unavoidable confounding factor. Both the KORA and IGAP datasets have accounted for important individual characteristics, including age, in their analyses, thus reducing the influence of confounding factors. However, the two-sample MR design used in this study does not allow for further elucidation of this relationship. Future investigations could explore this further using multivariable MR or other approaches. Third, considering the limited number of genetic variants included in the study, the variance explained by these variants in Hp is constrained, and there may be an underlying genetic variation that remains unaccounted for. This is because Hp, as a plasma protein, does not exist in isolation but rather functions within various protein networks [55]. However, current research capabilities do not provide a definitive explanation, and therefore, larger sample sizes or more precise detection techniques are required to robustly confirm the missing variations in this study. Some of the primary analysis’s statistical power fell below the 80% cutoff (Supplementary Table 4). This could be also attributed to the insufficient sample sizes or relatively small proportion of cases in the FinnGen dataset. As a result, caution should be used while interpreting the data. Finally, despite the benefit of less intrusive and accessible human blood samples in clinical settings, the blood-brain barrier may prevent some proteins in plasma from accurately and quickly detecting changes in the brain. We will investigate the function of proteins from the intracellular environment or Hp of cerebrospinal fluid in more detail in the future.

CONCLUSION

Taken together, the genetic evidence from the results of this MR Study reveals a potential causal relationship between higher plasma Hp and increased risk of AD. Hp may be a novel high-sensitivity plasma biomarker for the diagnosis of AD, providing new insights into the early prevention of AD.

Footnotes

ACKNOWLEDGMENTS

The authors wish to thank the KORA study, the IGAP consortium, and the FinnGen consortium for providing summary-level data. We would like to express our gratitude to Mr. Chen (12218519@zju.edu.cn) for his MR training course. His excellent sharing of the MR process makes it easier for us to accomplish this work.

FUNDING

This study was funded by grants from the Basic Research Project of Shanghai Sixth People’s Hospital (General Cultivation Project, ynms202207).

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

All the datasets used in the present study are openly available. The data generated or analyzed during this study have been included in this published article and its supplementary information files.

The supplementary material is available in the electronic version of this article: .

References

Heppner

, Ransohoff

, Becher

(2015) Immune attack: The role of inflammation in Alzheimer disease. Nat Rev Neurosci 16, 358–372.

Pohanka

(2018) Oxidative stress in Alzheimer disease as a target for therapy. Bratisl Lek Listy 119, 535–543.

Sadrzadeh

, Bozorgmehr

(2004) Haptoglobin phenotypes in health and disorders. Am J Clin Pathol 121 Suppl, S97–S104.

Bowman

, Kurosky

(1982) Haptoglobin: The evolutionary product of duplication, unequal crossing over, and point mutation. Adv Hum Genet 12, 189-261, 453–184.

Ayton

, Janelidze

, Roberts

, Palmqvist

, Kalinowski

, Diouf

, Belaidi

, Stomrud

, Bush

, Hansson

(2021) Acute phase markers in CSF reveal inflammatory changes in Alzheimer’s disease that intersect with pathology, APOE ɛ4, sex and age. Prog Neurobiol 198, 101904.

Yerbury

, Wilson

(2010) Extracellular chaperones modulate the effects of Alzheimer’s patient cerebrospinal fluid on Abeta(1-42) toxicity and uptake. Cell Stress Chaperones 15, 115–121.

Zhu

, Jiang

, Chen

, Zhou

, Cheng

(2018) Serum haptoglobin in Chinese patients with Alzheimer’s disease and mild cognitive impairment: A case-control study. Brain Res Bull 137, 301–305.

Philbert

, Xu

, Unwin

, Dowsey

, Cooper

GJS

(2021) Widespread severe cerebral elevations of haptoglobin and haemopexin in sporadic Alzheimer’s disease: Evidence for a pervasive microvasculopathy. Biochem Biophys Res Commun 555, 89–94.

Emdin

, Khera

, Kathiresan

(2017) Mendelian randomization. JAMA 318, 1925–1926.

10.

Lawlor

, Harbord

, Sterne

, Timpson

, Davey Smith

(2008) Mendelian randomization: Using genes as instruments for making causal inferences in epidemiology. Stat Med 27, 1133–1163.

11.

Burgess

, Butterworth

, Thompson

(2013) Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 37, 658–665.

12.

Davey Smith

, Hemani

(2014) Mendelian randomization: Genetic anchors for causal inference in epidemiological studies. Hum Mol Genet 23, R89–98.

13.

Burgess

, Scott

, Timpson

, Davey Smith

, Thompson

(2015) Using published data in Mendelian randomization: A blueprint for efficient identification of causal risk factors. Eur J Epidemiol 30, 543–552.

14.

Skrivankova

, Richmond

, Woolf

BAR

, Yarmolinsky

, Davies

, Swanson

, VanderWeele

, Higgins

JPT

, Timpson

, Dimou

, Langenberg

, Golub

, Loder

, Gallo

, Tybjaerg-Hansen

, Davey Smith

, Egger

, Richards

(2021) Strengthening the reporting of observational studies in epidemiology using Mendelian randomization: The STROBE-MR statement. JAMA 326, 1614–1621.

15.

Suhre

, Arnold

, Bhagwat

, Cotton

, Engelke

, Raffler

, Sarwath

, Thareja

, Wahl

, DeLisle

, Gold

, Pezer

, Lauc

, El-Din Selim

, Mook-Kanamori

, Al-Dous

, Mohamoud

, Malek

, Strauch

, Grallert

, Peters

, Kastenmüller

, Gieger

, Graumann

(2017) Connecting genetic risk to disease end points through the human blood plasma proteome. Nat Commun 8, 14357.

16.

Zhu

, Zheng

, Zhang

, Wu

, Trzaskowski

, Maier

, Robinson

, McGrath

, Visscher

, Wray

, Yang

(2018) Causal associations between risk factors and common diseases inferred from GWAS summary data. Nat Commun 9, 224.

17.

Kamat

, Blackshaw

, Young

, Surendran

, Burgess

, Danesh

, Butterworth

, Staley

(2019) PhenoScanner V2: An expanded tool for searching human genotype-phenotype associations. Bioinformatics 35, 4851–4853.

18.

Benn

, Nordestgaard

, Frikke-Schmidt

, Tybjærg-Hansen

(2017) Low LDL cholesterol, PCSK9 and HMGCR genetic variation, and risk of Alzheimer’s disease and Parkinson’s disease: Mendelian randomisation study. BMJ 357, j1648.

19.

Meng

, Wang

, Ji

, Shen

(2022) Causal association evaluation of diabetes with Alzheimer’s disease and genetic analysis of antidiabetic drugs against Alzheimer’s disease. Cell Biosci 12, 28.

20.

Twarowski

, Herbet

(2023) Inflammatory processes in Alzheimer’s disease-pathomechanism, diagnosis and treatment: A review. Int J Mol Sci 24, 6518.

21.

Larsson

, Traylor

, Burgess

, Markus

(2017) Genetically-predicted adult height and Alzheimer’s disease. J Alzheimers Dis 60, 691–698.

22.

DeCarli

(2021) The link between blood pressure and Alzheimer’s disease. Lancet Neurol 20, 878–879.

23.

Buchman

, Wilson

, Bienias

, Shah

, Evans

, Bennett

(2005) Change in body mass index and risk of incident Alzheimer disease. Neurology 65, 892–897.

24.

Shim

, Chasman

, Smith

, Mora

, Ridker

, Nickerson

, Krauss

, Stephens

(2015) A multivariate genome-wide association analysis of 10 LDL subfractions, and their response to statin treatment, in 1868 Caucasians. PLoS One 10, e0120758.

25.

Burgess

, Thompson

(2011) Avoiding bias from weak instruments in Mendelian randomization studies. Int J Epidemiol 40, 755–764.

26.

Yang

, Yu

, Xu

, Xie

, Zheng

, Feng

, Wang

, Xu

(2023) No evidence of a genetic causal relationship between ankylosing spondylitis and iron homeostasis: A two-sample Mendelian randomization study. Front Nutr 10, 1047640.

27.

Kunkle

, Grenier-Boley

, Sims

, Bis

, Damotte

, Naj

, Boland

, Vronskaya

, van der Lee

, Amlie-Wolf

, et al. (2019) Genetic meta-analysis of diagnosed Alzheimer’s disease identifies new risk loci and implicates Aβ, tau, immunity and lipid processing. Nat Genet 51, 414–430.

28.

Hemani

, Zheng

, Elsworth

, Wade

, Haberland

, Baird

, Laurin

, Burgess

, Bowden

, Langdon

, Tan

, Yarmolinsky

, Shihab

, Timpson

, Evans

, Relton

, Martin

, Davey Smith

, Gaunt

, Haycock

(2018) The MR-Base platform supports systematic causal inference across the human phenome. Elife 7, e34408.

29.

Yavorska

, Burgess

(2017) MendelianRandomization: An R package for performing Mendelian randomization analyses using summarized data. Int J Epidemiol 46, 1734–1739.

30.

Verbanck

, Chen

, Neale

, Do

(2018) Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50, 693–698.

31.

Burgess

, Thompson

(2017) Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol 32, 377–389.

32.

Bowden

, Davey Smith

, Haycock

, Burgess

(2016) Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 40, 304–314.

33.

, Niu

, Guo

, Liu

, Zheng

, Liu

, Yang

, Wang

, Li

, Hou

(2022) A mild causal relationship between tea consumption and obesity in general population: A two-sample Mendelian randomization study. Front Genet 13, 795049.

34.

Yuan

, Yang

(2018) Genetically determined serum uric acid and Alzheimer’s disease risk. J Alzheimers Dis 65, 1259–1265.

35.

Bowden

, Davey Smith

, Burgess

(2015) Mendelian randomization with invalid instruments: Effect estimation and bias detection through Egger regression. Int J Epidemiol 44, 512–525.

36.

Greco

, Minelli

, Sheehan

, Thompson

(2015) Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome. Stat Med 34, 2926–2940.

37.

Bowden

, Del Greco

, Minelli

, Davey Smith

, Sheehan

, Thompson

(2017) A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization. Stat Med 36, 1783–1802.

38.

Hemani

, Tilling

, Davey Smith

(2017) Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet 13, e1007081.

39.

Brion

, Shakhbazov

, Visscher

(2013) Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 42, 1497–1501.

40.

Hugo

, Ganguli

(2014) Dementia and cognitive impairment: Epidemiology, diagnosis, and treatment. Clin Geriatr Med 30, 421–442.

41.

Leclerc

, Li

, Jean

, Lampert

, Amador

, Diller

, Tolosano

, Doré

(2019) Temporal and age-dependent effects of haptoglobin deletion on intracerebral hemorrhage-induced brain damage and neurobehavioral outcomes. Exp Neurol 317, 22–33.

42.

Zhang

, Barker

, Pinchev

, Marshall

, Rasamoelisolo

, Smith

, Kupchak

, Kireeva

, Ingratta

, Jackowski

(2004) Mining biomarkers in human sera using proteomic tools. Proteomics 4, 244–256.

43.

Wang

, Yang

, Lu

, Chen

(2015) Utilization of magnetic nanobeads for analyzing haptoglobin in human plasma as a marker of Alzheimer’s disease by capillary electrophoretic immunoassay with laser-induced fluorescence detection. Anal Chim Acta 865, 76–82.

44.

Giometto

, Argentiero

, Sanson

, Ongaro

, Tavolato

(1988) Acute-phase proteins in Alzheimer’s disease. Eur Neurol 28, 30–33.

45.

Smith

, Ebrahim

(2003) ‘Mendelian randomization’: Can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 32, 1–22.

46.

Voet

, Srinivasan

, Lamkanfi

, van Loo

(2019) Inflammasomes in neuroinflammatory and neurodegenerative diseases. EMBO Mol Med 11, e10248.

47.

Argüelles

, Venero

, García-Rodriguez

, Tomas-Camardiel

, Ayala

, Cano

, Machado

(2010) Use of haptoglobin and transthyretin as potential biomarkers for the preclinical diagnosis of Parkinson’s disease. Neurochem Int 57, 227–234.

48.

Huang

, Wu

, Tseng

, Chen

, Hsieh

, Chen

(2011) Increased prothrombin, apolipoprotein A-IV, and haptoglobin in the cerebrospinal fluid of patients with Huntington’s disease. PLoS One 6, e15809.

49.

Wan

, La

, Zhu

, Yang

, Jiang

, Chen

, Feng

, Li

, Sang

, Hao

, Zhang

, He

(2007) Abnormal changes of plasma acute phase proteins in schizophrenia and the relation between schizophrenia and haptoglobin (Hp) gene. Amino Acids 32, 101–108.

50.

Yerbury

, Rybchyn

, Easterbrook-Smith

, Henriques

, Wilson

(2005) The acute phase protein haptoglobin is a mammalian extracellular chaperone with an action similar to clusterin. Biochemistry 44, 10914–10925.

51.

Maresca

, Spagnuolo

, Cigliano

(2015) Haptoglobin modulates beta-amyloid uptake by U-87 MG astrocyte cell line. J Mol Neurosci 56, 35–47.

52.

Spagnuolo

, Maresca

, La Marca

, Carrizzo

, Veronesi

, Cupidi

, Piccoli

, Maletta

, Bruni

, Abrescia

, Cigliano

(2014) Haptoglobin interacts with apolipoprotein E and beta-amyloid and influences their crosstalk. ACS Chem Neurosci 5, 837–847.

53.

Pan

, Wang

, Zheng

, Zhang

, Lin

, Shao

, Zhuge

, Jin

(2017) Aging systemic milieu impairs outcome after ischemic stroke in rats. Aging Dis 8, 519–530.

54.

Ding

, Kopchick

(2011) Plasma biomarkers of mouse aging. Age (Dordr) 33, 291–307.

55.

Jaeger

, Lucin

, Britschgi

, Vardarajan

, Huang

, Kirby

, Abbey

, Boeve

, Boxer

, Farrer

, Finch

, Graff-Radford

, Head

, Hofree

, Huang

, Johns

, Karydas

, Knopman

, Loboda

, Masliah

, Narasimhan

, Petersen

, Podtelezhnikov

, Pradhan

, Rademakers

, Sun

, Younkin

, Miller

, Ideker

, Wyss-Coray

(2016) Network-driven plasma proteomics expose molecular changes in the Alzheimer’s brain. Mol Neurodegener 11, 31.