Oral Cnm-positive Streptococcus Mutans Expressing Collagen Binding Activity is a Risk Factor for Cerebral Microbleeds and Cognitive Impairment

Cerebral microbleeds (CMBs) are an important risk factor for stroke and dementia. We have shown that the collagen binding surface Cnm protein expressed on cnm-positive Streptococcus mutans is involved in the development of CMBs. However, whether the collagen binding activity of cnm-positive S. mutans is related to the nature of the CMBs or to cognitive impairment is unclear. Two-hundred seventy nine community residents (70.0 years) were examined for the presence or absence of cnm-positive S. mutans in the saliva by PCR and collagen binding activity, CMBs, and cognitive function were evaluated. Cnm-positive S. mutans was detected more often among subjects with CMBs (p < 0.01) than those without. The risk of CMBs was significantly higher (odds ratio = 14.3) in the group with S. mutans expressing collagen binding activity, as compared to the group without that finding. Deep CMBs were more frequent (67%) and cognitive function was lower among subjects with cnm-positive S. mutans expressing collagen binding activity. This work supports the role of oral health in stroke and dementia and proposes a molecular mechanism for the interaction.


Investigation items concerning lifestyle and body measurement (clinical check-up).
Lifestyle habits were evaluated using a self-administered questionnaire checked by trained members of the study staff. This questionnaire consisted of items concerning medical past history, drug compliance, alcohol intake, smoking status and educational background. Hypertension was defined as having a history of hypertension and/or a systolic pressure ≥ 140 mmHg and/or a diastolic pressure ≥ 90 mmHg. Hyperlipidemia was defined as having a history of hyperlipidemia and/or showing a triglyceride ≥ 150 mg/dl, an HDL < 40 mg/dl and/or an LDL ≥ 140 mg/dl. Diabetes mellitus was defined as showing an HbA1c ≥ 6.5% and/or use of anti-diabetic medication. It was determined which subjects had a history of stroke/cardiovascular disease or had received treatments for those conditions. Those who were currently drinking alcohol every day were designated as the alcohol drinking group. Those who were smoking every day were designated as the smoking group. The subjects were also divided into 2 groups according to the number of years of schooling (≤ 12 years and > 12 years). Four subjects did not answer the question concerning education. Systemic atherosclerosis was evaluated by measuring brachial-to-ankle pulse wave velocity (baPWV) (form PWV/ABI; Omron Healthcare Co., Ltd., Kyoto, Japan) 16,17 . Dental evaluations. We collected information concerning the number of remaining teeth, presence/absence of dental caries, and community periodontal index (CPI) 18,19 by oral examination, which was performed by a single dentist (FM). Dental caries (+ ) was defined as having at least 1 tooth with dental caries. Concerning the CPI, the oral cavity of each participant was divided into 6 areas, and the pocket depth was measured in the tooth specified in each area by probing using a WHO periodontal probe. CPI were graded according to following categories; (0) no sign of disease, (1) gingival bleeding after gentle probing, (2) supra-or subgingival calculus, (3) pathologic pockets 4 or 5 mm deep, and (4) pathologic pockets ≥ 6 mm, and the maximum value of the 6 areas was regarded as the CPI of the subject 19 . Based on the results, the subjects were divided into those with and without pathologic pockets (CPI ≥ code 3).
Laboratory evaluations. Peripheral blood samples were obtained from all participants. We measured high-sensitivity C-reactive protein (hs-CRP) and apolipoprotein E (APOE) genotype in addition to basic hematological and biochemical assays. Serum hs-CRP concentrations were measured using an established commercially available manufacturer's protocol by latex coagulation nephelometry as part of the health screening check-up (Kyoto Biken Laboratories, Inc., Kyoto, JAPAN). APOE genotyping was performed by polymerase chain reaction (PCR) (Funakoshi Co., Ltd., Tokyo, Japan). Of the 279 subjects, 272 (97%) consented to APOE genotyping. APOE genotype was classified into 2 groups, according to the presence or absence of the APOE ε 4 allele.

Isolation of S. mutans strains and molecular biological detection of cnm encoding Cnm protein.
We sampled oral bacteria from all subjects from saliva and plaque around the teeth. Each oral salivary specimen was streaked on Mitis-Salivarius agar containing bacitracin (MSB), a selective medium for mutans streptococcal species, and bacterial colonies were picked up. Bacterial DNA was extracted from the colonies in accordance with our previously reported methods and stored as DNA extracts 13,20 . The DNA extracts were subjected to PCR for confirming S. mutans using S. mutans-specifics primers 21 . Using cnm-specific primers, the presence or absence of the cnm gene in the S. mutans isolates was determined. Details of these assay methods are described in supplementary materials.
Scientific RepoRts | 6:38561 | DOI: 10.1038/srep38561 Measurement of the Cnm protein-binding activity of cnm-positive S. mutans. We performed a collagen-binding assay to determine the collagen-binding activity of the cnm-positive S. mutans isolated from the subjects using methods previously reported 13,20 . For analysis of Cnm protein-binding activity, standard strains (SA83 and SA137) with binding activity of 100% was prepared, and the binding activity of the sample was expressed as the rate relative to the activity of the standard sample. The Cnm protein-binding activity was judged to be positive when this rate was 10% or higher. Details of these assay methods are described in supplementary materials.
CMBs were evaluated by susceptibility-weighted imaging (SWI) (3d-Tlffe 4 mmTHK/Gapless) of brain MRI. By SWI, CMBs can be detected because hemosiderin present in the foci of CMBs is visualized as punctate and round hypointense areas 24,25 . A representative MRI of microbleeds based on the SWI image is shown in Fig. 1. As shown in the figure, microbleeds were assessed based on susceptibility-weighted imaging (SWI) of brain MRI. We classified CMBs into deep, lobar, and mixed types 26 . This protocol is the same as that employed in our previous clinical MRI studies 13,27 . The absence of a large cerebral hemorrhage, pathological brain atrophy, or other findings suggesting an infectious disease or neurodegenerative condition was established in all subjects. We carefully performed all the MR evaluations blinded to information about the Cnm-related data. The examiners (a radiologist and neurologist) discussed the findings to obtain a consensus in case of discordant MR evaluations. Because CMBs may be confused with other signals such as blood vessels or calcifications, we carefully evaluated whether or not the findings represent CMBs using axial and sagittal scanning. Along with the total-grading score on the Fazekas classification and CMBs, the intraclass correlation coefficient (ICC) was calculated as an index of inter-rater reliability; a high ICC of 0.92 and 0.93 were obtained for both DWMLs and PVH. Thus, the data obtained had a favorable inter-rater reliability, which suggested that this grading was sufficiently reliable to be used as an accurate grading tool.

Cognitive function evaluations.
We used the Mini-Mental State Examination (MMSE) 28 and letter fluency tasks 29,30 . Specifically, the subjects are instructed to orally recall as many words that start with a specified letter during a 1-minute period, and the words are counted. In this study, we counted Japanese words that start with "Ta" and "Ka" 31 . These tests of cognitive function were performed on the same day as brain MRI, and the results were recorded by trained neurologists and neuropsychologists.
Analytical methods. The results of each factor examined were compared between sexes and according to the presence or absence of CMBs, cnm-positive S. mutans, and collagen binding activity. These group-wise comparisons were performed using the t-test and χ 2 -test. We also evaluated the risk of CMBs by multivariate analysis using sex, age, lifestyle, and the results of body measurement, oral examination, blood tests, and brain MRI as adjustment factors. These analyses were performed at the 0.05 level of significance using SPSS 19.0 J for Windows (SPSS Japan Inc., Tokyo, Japan).

Results
Characteristic of the subjects. Table 1 shows the characteristics of the subjects. The mean age was 70.0 ± 6.1 years, BMI was 22.7 ± 3.0 kg/m 2 , and MMSE score was 28.3 ± 2.1. Fifty-eight subjects (21%) were APOEε 4 allele carriers. Concerning the dental check-up findings, the mean number of remaining teeth was 22.7 ± 7.5, 86 subjects (31%) had dental caries, and the CPI was ≥ code 3 in 77 subjects (28%). Laboratory evaluations showed that S. mutans, cnm-positive S. mutans, and collagen binding activity were positive in 94%, 33%, and 25%, respectively, of the subjects. On brain MRI, CMBs were detected in 73 subjects (26%).
The age, BMI, frequency of subjects with a history of treatment for diabetes, frequency of those with drinking history, frequency of those with smoking history, number of years of schooling, frequency of those with dental caries, and MMSE score were significantly higher in males, and the frequency of those with S. mutans was significantly higher in females.  "Ka" (score) ± SD 9.9 ± 3.4 9.5 ± 3.3 9.8 ± 3.4 Background of CMBs (+) and cognitive function. Table 2 compares the CMBs (+ ) (n = 73) and CMBs (− ) (n = 206) groups. The CMBs (+ ) group showed a significantly higher BMI (p < 0.05) and significantly higher frequencies of those with a CPI ≥ code 3 (p < 0.05), cnm-positive S. mutans-isolated subjects (p < 0.01), DWMH ≥ grade 2 (p < 0.05) and collagen binding activity (+ ) subjects (p < 0.01). However, no significant difference was observed between the two groups in regard to hypertension, hyperlipidemia, diabetes, oral antiplatelet therapy and PVH grade. We did not see any relationship between CMBs and hs-CRP. The type of CMBs were deep in 44 (60%), lobar in 20 (27%) and mixed in 9 (12%). We also evaluated the laterality of CMBs but found no right-left differences. On cognitive function evaluations, the CMBs (+ ) group showed significantly lower MMSE score and lower "Ta" score of the letter fluency tasks (p < 0.05 for both). Table 3 shows the results of logistic regression analysis concerning the occurrence of CMBs. Crude analysis demonstrated a CPI ≥ code 3 (OR = 1.83, 95% CI: 1.03-3.25, p < 0.05) and collagen binding activity (+ ) (OR = 9.11, 95% CI: 4.94-16.8, p < 0.01) as significant risk factors for the occurrence of CMBs. In Model 2, in  "Ta" (score) ± SD 8.5 ± 3.5 7.6 ± 2.9* "Ka" (score) ± SD 10.0 ± 3.5 9.2 ± 3.0 Background and cognitive function of the cnm-positive S. mutans group and the collagen binding activity ≥10% group. Comparing the cnm-positive (n = 91) with the cnm-negative (n = 188) groups, no significant difference was observed in the clinical characteristics examined (Table 4). Cnm protein was collagen binding activity ≥ 10% in 78%. The distribution of CMBs was deep in 70%, lobar in 17% and mixed in 13% in the cnm-positive group with significant differences between the two groups. Deep CMBs were observed more frequently in the cnm-positive group, but lobar CMBs were observed more frequently in the cnm-negative group.
Cognitive function evaluations showed no significant difference in the MMSE score, but the "Ka" score of the letter fluency tasks was significantly decreased in the cnm-positive group (p < 0.05 for "Ka"). Comparing the collagen binding activity (+ ) (n = 71) with (− ) (n = 208) groups, no marked difference was observed in clinical characteristics (Table 4). But the frequency of deep CMBs was significantly higher in the collagen binding activity ≥ 10% group (p < 0.01). Also, the percentage of subjects with dental caries was significantly higher (p < 0.01) in the collagen binding activity ≥ 10% group. Concerning cognitive functions, the "Ta" and "Ka" scores of the letter fluency tasks were significantly lower in the collagen binding activity ≥ 10% group (p < 0.05 for "Ta", p < 0.01 for "Ka").

Discussion
We have previously reported that the collagen binding activity of Cnm protein produced by cnm-positive S. mutans is involved in the development of CMBs 13 . Our data showed that chronic presence of cnm-positive S. mutans in mice hematogenously induces cerebral hemorrhage through disruption of the blood-brain barrier 10 . In cerebral hemorrhage, endothelial cells of cerebral vessels are damaged, and collagen is exposed on the surface of the damaged blood vessel. Hemostasis caused by this exposed collagen is important as mechanism against cerebral hemorrhage, but, if cnm-positive S. mutans enters the blood from the mouth, its collagen binding activity may induce cerebral hemorrhage by adherence to and damage to the vascular endothelium 10 . However, whether these relationships between chronic cnm-positive S. mutans infection and cerebrovascular damage demonstrated in animals also applies to humans is unclear.
CMBs can be detected as punctate hypointensities by susceptibility-weighted imaging (SWI), a technique of brain MRI that is becoming increasingly prevalent, and pathological studies have confirmed that these hypointensities are hemorrhagic hemosiderin deposits 32,33 . CMBs have been observed in 4-6% of healthy individuals and 57-64% of patients with intracerebral hemorrhage 6,7 . CMBs were positive in 15% of the general population in the Rotterdam Scan Study, and their prevalence increased with age 34   CMBs positivity was relatively high at 26% in our study perhaps because of the high average age of the subjects (70 ± 6.1 years) and the high risk of hypertension, smoking and alcohol use. Deep CMBs were observed in 60% of our subjects. Bokura et al. reported the frequencies of deep, lobar, and mixed CMBs at 53%, 13%, and 34%, respectively, in an epidemiological survey of local residents 7 , and we obtained similar results. However, we noted no relationship between the prevalence of CMBs and hypertension or hyperlipidemia, which have been reported as risk factors of CMBs 6,7,35,36 . The mean age of our subjects was 70, which increases the prevalence of hypertension and CMBs. We noted marked differences in the distribution of CMBs on brain MRI between cnm-positive and -negative groups, following the similar results in the collagen activity groups. It has been previously reported that the hs-CRP level, a biomarker of chronic inflammation, was high in the group with CMBs, suggesting their relationship with increased vulnerability of the vascular wall due to inflammation 37 . The percentage of patients with high hs-CRP was higher in the CMBs (+ ) group, but the difference was not significant. Nevertheless, the frequent occurrence of deep CMBs in patients with cnm-positive S. mutans strongly suggests an involvement of bacteria    in secondary inflammation or injury of perforating arteries 38 . In subjects positive for S. mutans from the Japanese hospital-based cohort study 11,14 , the increased rate of collagen binding activity was similarly correlated with the number of deep CMBs, which is consistent with our results. A limitation of our work is that we did not measure inflammatory components other than hs-CRP, such as cytokines. Also, as a previous report suggested the relationship between the arterial stiffness and CMBs 39 , we performed evaluation of PWV, reflecting arteriosclerosis of peripheral vessels, but noted no significant difference. Regarding the cognitive functional evaluations, the group of CMBs (+ ) showed low scores on the MMSE and letter fluency task starting with "Ta", whereas it did not show statistical lower scores in letter fluency task starting with "Ka", suggesting that it was not statistical significant because of lack of sufficient numbers. Another possibility is that the letter fluency task may be influenced by the affected cerebral location by CMBs in accordance with the functional representation of language.
Secondly, the infection rate of cnm-positive S. mutans was high in the CMBs (+ ) group. Also, collagen binding activity ≥ 10% showed the highest odds ratio for the occurrence of CMBs after adjustment for other factors. Thus, collagen binding activity of cnm-positive S. mutans was an independent risk factor of the occurrence of CMBs and in the CMBs (+ ) group, the percentage of patients with CPI ≥ code 3 was high. Comorbidities, such as gingivitis and periodontitis, increase vascular vulnerability, and entry of cnm-positive S. mutans into the blood, whoch may result in damage to cerebral vessels 40 . Also, the percentage of dental caries (+ ) patients was significantly higher in the collagen binding activity ≥ 10% group. Nomura et al. suggested a strong relationship between the occurrence of dental caries and the cnm gene 20 . Individuals with dental caries, often neglect appropriate brushing or dental treatment permitting proliferation of cnm-positive S. mutans, which binds to collagen, a component of dentin, increasing the percentage of collagen binding activity (+ ) in subjects in the dental caries (+ ) group 11 .
Thirdly, cognitive scores were decreased in the CMBs (+ ) group. Moreover, the scores of letter fluency tasks were significantly lower in the collagen binding activity ≥ 10% group. Yakushiji et al. reported that that CMBs cause cognitive impairment by damaging the basal ganglia-and the frontal-subcortical neuronal circuits 9 . Charidimou et al. observed that CMBs in the basal ganglia are often associated with microangiopathy, and are related to cognitive impairment in older people 41,42 . We suggest that collagen binding (+ ) S. mutans activates an indirect inflammation-mediated mechanism, which exacerbates microangiopathy in the region supplied by the perforator branches in the basal ganglia, thus increasing deep CMBs, and cognitive impairment. Moreover, because we confirmed that the MMSE scores were lower in females compared to males, and males had a smaller frequency of S. mutans positivity, we performed logistic analysis including sex as an adjusting factor. We also added education history as an adjusting factor. However, sex differences were not statistically significant.
We evaluated the relationships of PVH/DWMH, which reflect chronic ischemic changes of the brain, with cnm-positive S. mutans and the collagen binding activity. No significant relationship was noted in this study. In addition, APOE4 is widely known to be an allele that enhances cerebral vascular vulnerability and cognitive impairment 43,44 , and CMBs are observed more frequently in APOE4 carriers 45,46 . However, we found that the presence APOE4 was not significantly related to the presence of CMBs.
This study has the following limitations. Various confounding factors other than the items evaluated in this study may be involved in the associations involving cognitive impairment, CMBs and cnm-positive S. mutans. Because of the cross-sectional nature of this study, it was difficult to evaluate the relationships of brain MR or cognitive function with oral findings including cnm-positive S. mutans exposure. Longitudinal studies of the relationships of Cnm-positive S. mutans and CMBs will be of interest. Also we have not evaluated other factors related to oral hygiene besides the Cnm protein which may be conferring the occurrence of CMBs with cognitive decline. The data were obtained from a single center enrolling consecutive subjects responding to a random mailing in order to minimize the effect of individual selection bias. Although we used a random mailing to recruit subjects, some degree of selection bias was unavoidable. In addition, the number of subjects of this study was small and evaluation of a larger population is necessary. We understand that MMSE does not provide a sensitive examination of cognitive deficits. Further examination of executive functions will be advisable such as TMT (Trail Making Test) and the FAB (Frontal Assessment Battery). Furthermore, a comprehensive inflammatory profile was not obtained.

Conclusions
The collagen binding activity of cnm-positive S. mutans was closely related to the occurrence of deep CMBs and may be a risk factor for cognitive impairment. These results are novel findings that link chronic oral infections with geriatric disorders, such as stroke and cognitive impairment, and suggest cnm-positive S. mutans, in particular, is a novel factor of cognitive impairment associated with CMBs. We now clarify in this population-based survey that the rate of cnm-positive S. mutans is also high in asymptomatic cerebral microbleeds, and is related to a decline of cognitive function. An intervention study focused on oral care and the microbiota in CMBs subjects will be of interest. These data further support the important influence of the oral microbiota on neurological disease and emphasize the importance of collaboration between dental and medical researchers.