The Alzheimer– E. coli Axis: What Can We Learn from an Electronic Health Record Platform

Abstract

Background:

Alzheimer’s disease (AD) is a neurodegenerative disease with unclear etiology. Recent studies have demonstrated a potential role for gut microbiome. There is, however, a significant dearth in epidemiological correlation between gut bacteria and AD.

Objective:

To investigate the association between Escherichia coli (E. coli) infection and AD.

Methods:

Counts of patients with ICD 10 diagnoses of AD, E. coli, urinary tract infection, and comorbidities were retrieved from the electronic health records at the University of Florida Health Center.

Results:

The relative risk for AD with a previous event of E. coli was 5.17 (95%CI 4.0786 to 6.5446, p < 0.0001). In the unadjusted association, patients with E. coli infection had odds ratio (OR) of 20.83 to have AD (95%CI, 17.7–24.34; p < 0.0001); after adjusting for gender (OR = 12.71; 95%CI, 10.82–14.83; p < 0.0001), race (OR = 13.97; 95%CI, 11.84–16.36; p < 0.0001), age group (OR = 11.51; 95%CI, 9.73–13.54; p < 0.0001), diabetes (OR = 9.23; 95%CI, 7.79–10.87; p < 0.0001), stroke (OR = 5.31; 95%CI, 4.47–6.28; p < 0.0001), and hypertension (OR = 4.55; 95%CI, 3.86–5.32; p < 0.0001).

Conclusion:

These results should be taken cautiously. This retrospective cross-sectional study cannot infer causality and had used aggregate data that did not allow simultaneous adjustments of covariates. Future studies are warranted to investigate the link between gut bacteria and AD.

Keywords

Alzheimer’s disease association infection

INTRODUCTION

Alzheimer’s disease (AD) is a neurodegenerative disease characterized by memory and language disorders, and the accumulation of amyloid-β and tau protein in the brain [1, 2]. AD is associated with various diseases such as type 2 diabetes, periodontal disease, and obesity [3]. Multiple investigators have suggested infectious agents to be associated with AD [4]. The human gut and brain form a network called “brain–gut–microbiota axis,” and it is suggested that the gut microbiota is involved in brain diseases. The gut bacteria can impact the brain and the affected brain by AD can decrease gut mobility and cause increase in gut bacterial infections [5]. Herpes simplex, herpes zoster, spirochetes. and gram-negative bacteria have been proposed to be initiating factors [6]. Recently, few studies have proposed that gut bacteria may be associated with AD by the leaky gut theory, that suggests that gut bacteria or their byproducts may leak in to body fluids and impact the brain [7].

These bacterial pathogens may directly cross a compromised blood-brain barrier, reach the central nervous system, and cause neurological damage by triggering neuroinflammation. Pathogens may also cross a weakened intestinal barrier, reach vascular circulation, and then cross the blood-brain barrier or cause low grade chronic inflammation and subsequent neuroinflammation [8].

Escherichia coli (E. coli) bacterial amyloid functions as a trigger to initiate misfolded alpha-synuclein [9]. Rats exposed to curli-producing bacteria displayed increased neuronal α-synuclein deposition in both gut and brain and enhanced microgliosis and astrogliosis [10]. LPS and E. coli K99 were detected immunocytochemically in brain parenchyma and vessels in all AD and control brains, DNA sequencing confirmed E. coli DNA in human control and AD brains [11]. The MMWR publication reported that there is an increase in both Shiga toxin-producing E. coli and AD as the cause of death in United States [12]. Although the literature on the association between AD and E. coli and gut microbiota is rife, it includes mostly basic research about the plausibility of the brain gut association and critical reviews on the subject. We could not identify clinical or epidemiological studies that reported on the strength of the association.

In the present study, we aimed to investigate whether the Alzheimer-E. coli axis can be demonstrated by increased association between AD and E. coli by getting data on patients from electronic health records (EHR) platform.

MATERIAL AND METHODS

The aggregated data extracted from Informatics for Integrating Biology and the Bedside (i2b2), http://www.i2b2.org) is an NIH-supported National Center for Biomedical Computing that developed a scalable informatics framework designed for translational research with collaboration of the University of Florida (UF). The i2b2 reports based on EHR were used in logistics regressions. We have used the international coding of diseases (ICD) 10 platform for our queries. The dependent variable was AD (ICD 10 G30), while the independent variables included gender, race, age, smoking history, hypertension (ICD 10, I10-I11), E. coli infection (ICD 10, B96.2), obesity, (ICD10, E65–E68), cardiovascular disease (ICD 10, I00–I99), diabetes mellitus (ICD 10, EO8-E11), stroke (ICD10, I63), and hypercholesterolemia (ICD10, ED78.0). The platform enables to retrieve counts for each variable and up to 3 combinations of variables. The UF Integrated Data Repository i2b2 platform provided the demographic data. The total study population (n) included patients attending the different UF Health centers across the state of Florida in the period of January 2015–June 2020 corresponding to the total ICD 10 diagnoses. The study was exempted by the UF Health Center Institutional Review Board as it did not include personal health information. Logistic regression was conducted using the aggregated counts as weight in the model and calculate the odds ratio (OR). Limited by I2b2 data, we can only adjust for one covariate at a time. Obesity, diabetes, circulatory disease, stroke, hypertension smoking status, and demographics (gender, age groups, and race) are all known risk factors for AD were adjusted as covariates one at a time. SAS 9.4, USA procedure logistic was used to perform the analysis.

Fitting of logistic regression on aggregate/count data is no different from fitting logistic regression on standard data, given that the aggregated data can always be expanded into the standard form where each row represents an individual record. Therefore, it can be viewed as a standard logistic regression. Therefore, the covariates were accounted for by being included in the logistic models one at a time [13].

The relative risk for AD when there was a previous infection of E. coli was calculated using the MedCalc statistical software (MedCalc Software Ltd. Relative risk calculator. https://www.medcalc.org/calc/relative_risk.php (Version 20.009; accessed July 27, 2021).

RESULTS

The demographic data on the studied population is presented in Table 1. All the cases that were diagnosed with E. coli infection and AD had also urinary tract infection and the infection is typically confirmed by bacterial culture. The majority of these patients were older than 85; however, 79 patients were at the age of 55–84 (Table 1). Because the ICD 10 database includes patients diagnosed only from 2015 to present, we believe that the number of younger ages group will be diagnosed with AD in the near future.

Table 1

Demographic information of patients

		Total number of patients (n = 1,458,444)	Number of patients with AD	Number of patients with no AD	Number of patients with E. coli			Number of patients with AD, E. coli, and UTI	Number of patients first E. coli and secondly AD
					AD positive	AD negative	Total
Sex (missing = 1,523)	Female	782,173	2,629	779,544	135	3,177	3,312	130	55
	Male	676,271	1,394	674,877	38	1,484	1,522	25	13
Race (missing = 7)	African American	147,262	350	146,912	35	956	991	30	13
	White	701,039	2,123	698,916	128	3,317	3,445	125	55
	Other	570,135	171	569,964	10	270	280	0	0
Age group (missing = 77)	Ages 0–54 y	822,094	14	822,080	0	1,773	1,773	0	0
	Ages 55–64 y	217,739	132	217,607	8	772	780	8	4
	Ages 65–74 y	218,538	401	218,137	10	945	955	10	3
	Ages 75–84 y	135,685	980	134,705	54	732	786	50	22
	Ages > 85 y	65,482	1,136	64,346	101	443	544	76	39

AD, Alzheimer’s disease; UTI, urinary tract infection

As shown in Table 2, patients with E. coli infection were 20.83 times more likely to have AD than patients without E. coli infection (95%CI, 17.7–24.34; p < 0.0001).

Table 2

Odds ratio for Alzheimer’s disease in patients with E. coli infection after adjustments for race, age, gender, and comorbidities

	Odds Ratio	95%Wald Confidence Limits
E. coli versus Non E. coli^*	20.83	17.7	24.34	< 0.0001
E. coli versus Non E. coli¹	12.71	10.82	14.83	< 0.0001
Male versus Female	0.63	0.59	0.67	< 0.0001
E. coli versus Non E. coli²	13.97	11.84	16.36	< 0.0001
African American versus White	0.77	0.68	0.86	< 0.0001
Other versus White	0.1	0.09	0.12	< 0.0001
E. coli versus Non E. coli³	11.51	9.73	13.54	< 0.0001
Age 55–64 versus < 54	35.11	21	63.75	< 0.0001
Age 65–74 versus < 54	105.63	64.55	188.93	< 0.0001
Age 75–84 versus < 54	413.39	254.21	736.16	< 0.0001
Age > 85 versus < 54	987.08	607.31	1757.15	< 0.0001
E. coli versus Non E. coli⁴	15.64	13.28	18.3	< 0.0001
Smoking versus No Smoking	2.84	2.62	3.07	< 0.0001
E. coli versus Non E. coli⁵	15.85	13.37	18.67	< 0.0001
Obese versus No Obese	2.04	1.82	2.28	< 0.0001
E. coli versus Non E. coli⁶	4.86	4.13	5.69	< 0.0001
Cardiovas versus No Cardiovas	20.26	18.42	22.32	< 0.0001
E. coli versus Non E. coli⁷	9.23	7.79	10.87	< 0.0001
DM versus No DM	6.43	5.86	7.03	< 0.0001
E. coli versus Non E. coli⁸	5.31	4.47	6.28	< 0.0001
Stroke versus No Stroke	18.19	16.67	19.82	< 0.0001
E. coli versus Non E. coli⁹	4.55	3.86	5.32	< 0.0001
Hypertension versus No Hypertension	17.54	16.16	19.05	< 0.0001
E. coli versus Non E. coli¹⁰	16.42	13.91	19.26	< 0.0001
Hypercholes versus No Hypercholes	6.37	5.58	7.23	< 0.0001

Model marked with ^* is the raw model that evaluates the association between Alzheimer’s disease (AD) status with E. coli infection without adjusting for any covariates. Model ^{1,2,3,4,5,6,7,8,9,10}adjust for gender, race, age, smoking, obesity, cardiovascular disease (Cardiovas), diabetes (DM), stroke, hypertension, and hypercholesterolemia (Hypercholes), respectively and one at a time. In this project, the primary outcome is to define the association between AD status with E. coli infection. After fitting logistic regression models, we found that there is a significant association between E. coli infection and AD. The odds of having AD for people with E. coli infection was 20.83 (95%CI: 17.7, 24.34) times of the odds for people without E. coli infection. When gender, race, age, and other risk factor diseases are included for adjustment into the model, the odds of having AD with E. coli infection are still significant. The secondary outcome is to define the risk factor diseases that associate with AD status. By comparing the odds ratio (OR) from the models, we figured Cardiovas (OR 20.23), stroke (OR 18.19), hypertension (OR 17.54), DM (OR 6.43), hypercholes (OR 6.37), smoking (OR 2.84), and obesity (OR 2.04) are significant risk factors associated with AD.

As shown in Table 2, the association decreased dramatically but remained significant after adjusting for gender (OR = 12.71; 95%CI, 10.82–14.83; p < 0.0001), race (OR = 13.97; 95%CI, 11.84–16.36; p < 0.0001), age group (OR = 11.51; 95%CI, 9.73–13.54; p < 0.0001), and smoking history (OR = 15.64; 95%CI, 13.28–18.3; p < 0.0001).

As shown in Table 2, the association decreased dramatically but remained significant after adjusting for obesity (OR = 15.85; 95%CI, 13.37–18.67; p < 0.0001), cardiovascular disease (OR = 4.86; 95%CI, 4.13–5.69; p < 0.0001), diabetes mellitus (OR =9.23; 95%CI, 7.79–10.87; p < 0.0001), stroke (OR = 5.31; 95%CI, 4.47–6.28; p < 0.0001), hypertension (OR = 4.55; 95%CI, 3.86–5.32; p < 0.0001), and hypercholesterolemia (OR = 16.42; 95%CI, 13.91–19.26; p < 0.0001).

All conditions were significantly associated with AD (p < 0.0001). Black or African American race was on average 0.77 times less likely to have AD than white race (p < 0.0001). Male patients were 0.63 times less likely to have AD than female patients (p < 0.0001). Also, older patients were more likely to have AD (p < 0.0001).

Of the 173 patients with both E. coli and AD, we have identified 68 patients that had first a diagnosis of E. coli infection and later with AD. The relative risk for these patients to develop AD was calculated to be 5.17 (95%CI 4.07 to 6.54, p < 0.0001).

DISCUSSION

In the present cross-sectional retrospective study of EHR aggregate platform, we have found that patients with E. coli infection were 20.83 times more likely to have AD than patients without E. coli infection (95%CI, 17.7–24.34; p < 0.0001). The association was robust even after adjustments to age, race, and gender and significant known risk factors for AD such as cardiovascular diseases, hypertension, diabetes, obesity, and history of smoking.

The study only assessed population seeking care at the UF’s health centers. The results may be influenced by social disparities since socio-economic factors may affect the decision of certain population to seek medical care in this particular health system [14]. In spite of the ubiquity of information on the brain-gut axis, we could not identify published studies who reported of clinical or epidemiological information on the strength of association between gut infections in the form of gastroenteritis and AD. Although the theoretical basis for this theory exist and bacteria have been shown to form in vitro and animal studies protein structures that are similar to those in the AD brains [15], no studies on increased odds ratio for AD in in patients with E. coli infections were published.

The finding that E. coli produce extracellular amyloids known as curli fibers, composed of the major structural sub-unit CsgA, are a common secretory component that facilitate surface adhesion, biofilm development, and protection against host defenses is a pivotal pillar of the brain-gut axis theory [16].

Early life E. coli infection was shown to cause an inflammatory response in a mouse brain starting shortly after infection and showed increased susceptibility to synapse damage and cognitive impairment induced by low doses of induced by amyloid-β oligomers (AβOs), neurotoxins found in AD brains and suggest that neonatal infections can modulate microglial response to AβOs in adult mice, thus contributing to amyloid-β-induced synapse damage and cognitive impairment [17].

The i2b2 platform does not allow us to identify the origin site of the E. coli infection associated with the ICD-10 code B89.2. We assume that in the majority of the cases the origin is the GI tract, and the clinical scenario might have gastroenteritis or urinary tract infection. However, we were focused on the fact that these were all cases associated with the E. coli family and not necessary a specific strain.

This type of a retrospective cross-sectional study suffers from a few weaknesses. First, there was no access to individual data of patients, secondly the aggregate data did not allow adjustments for all potential confounding factors/covariates simultaneously, and finally it cannot imply any direction of the associations inferred. On the positive side, this type of aggregate data allows to analyze data from as many as 2,840 AD patients that would be very difficult to achieve with a conventional statistical analysis.

In conclusion, further studies are warranted to elucidate the role of gram-negative bacteria in AD.

DISCLOSURE STATEMENT

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/21-5004r2).

References

Alkasir

, Li

, Jin

, Zhu

(2017) Human gut microbiota: The links with dementia development. Protein Cell 8, 90–102.

Bloom

(2014) Amyloid-β and tau: The trigger and bullet in Alzheimer disease pathogenesis. JAMA Neurol 71, 505–508.

Long

, Holtzman

(2019) Alzheimer disease. An update on pathobiology and treatment strategies. Cell 179, 312–339.

Mawanda

, Wallace

(2013) Can infections cause Alzheimer’s disease? Epidemiol Rev 35, 161–180.

Liang

, Wu

, Jin

(2018) Gut-brain psychology: Rethinking psychology from the microbiota-gut-brain axis. Front Integr Neurosci 12, 33.

Harris

, Harris

(2015) Herpes simplex virus type 1 and other pathogens are key causative factors in sporadic Alzheimer’s disease. J Alzheimers Dis 48, 319–353.

Kowalski

, Mulak

(2019) Brain-gut-microbiota axis in Alzheimer’s disease. J Neurogastroenterol Motil 25, 48–60.

Panza

, Lozupone

, Solfrizzi

, Watling

, Imbimbo

(2019) Time to test antibacterial therapy in Alzheimer’s disease. Brain 142, 2905–2929.

Zhan

, Stamova

, Jin

, DeCarli

, Phinney

, Sharp

(2016) Gram-negative bacterial molecules associate with Alzheimer disease pathology. Neurology 87, 2324–2332.

10.

Bartels

, Choi

, Selkoe

(2011) α-Synuclein occurs physiologically as a helically folded tetramer that resists aggregation. Nature 477, 107–110.

11.

Chen

, Stribinskis

, Rane

, Demuth

, Gozal

, Roberts

, Jagadapillai

, Liu

, Choe

, Shivakumar

, Son

, Jin

, Kerber

, Adame

, Masliah

, Friedland

(2016) Exposure to the functional bacterial amyloid protein curli enhances alpha-synuclein aggregation in aged Fischer 344 rats and Caenorhabditis elegans. Sci Rep 6, 34477.

12.

Johnson

, Hayes

, Brown

, Hoo

, Ethier

; Centers for Disease Control and Prevention (CDC) (2014) CDC National Health Report: Leading causes of morbidity and mortality and associated behavioral risk and protective factors–United States, 2005-2013. MMWR Suppl 63, 3–27.

13.

Chow

, Shao

, Wang

(2008) Sample size calculations in clinical research, second edition. Chapman & Hall/CRC, Boca Raton, Florida.

14.

Coughlan

, Coutrot

, Khondoker

, Minihane

, Spiers

, Hornberger

(2019) Toward personalized cognitive diagnostics of at-genetic-risk Alzheimer’s disease. Proc Natl Acad Sci U S A 116, 9285–9292.

15.

Chapman

, Robinson

, Pinkner

, Roth

, Heuser

, Hammar

, Normark

, Hultgren

(2002) Role of Escherichia coli curli operons in directing amyloid fiber formation. Science 295, 851–855.

16.

Schwartz

, Boles

(2013) Microbial amyloids–functions and interactions within the host. Curr Opin Microbiol 16, 93–99.

17.

Frost

, Barros-Aragão

, da Silva

, Venancio

, Matias

, Lyra Silva

E NM

, Kincheski

, Pimentel-Coelho

, De Felice

, Gomes

FCA

, Ferreira

, Figueiredo

, Clarke

(2019) Neonatal infection leads to increased susceptibility to Aβ oligomer-induced brain inflammation, synapse loss and cognitive impairment in mice. Cell Death Dis 10, 323.