Abstract
Introduction We investigated the interplay between infections and APOE4 on brain glucose hypometabolism, an early preclinical feature of Alzheimer’s Disease (AD) pathology.
Methods Multivariate linear regression analysis was performed on 1,509 participants of the Alzheimer’s Disease Neuroimaging Initiative (ADNI). The outcomes were the rank-normalized hypometabolic convergence index (HCI) and statistical regions of interest (SROI) for AD and mild cognitive impairment (MCI). Further, the HCI and its change in the presence and absence of APOE4 were evaluated.
Results Infections were associated with greater hypometabolism [0.15, 95% CI: 0.03, 0.27, p=0.01], with a more pronounced effect among APOE4 carriers, indicating an interaction effect. A higher HCI (0.44, p=0.01) was observed in APOE4 carriers with multiple infections, compared to (0.11, p=0.08) for those with a single infection, revealing a dose-response relationship. The corresponding estimates for the association of infections with SROI AD and SROI MCI were -0.01 (p=0.02) and -0.01 (p=0.04) respectively.
Conclusion Our findings suggest that infections and APOE4 jointly contribute to brain glucose hypometabolism and AD pathology, supporting a “multi-hit” mechanism in AD development.
1 Introduction
Alzheimer’s disease (AD) is a slowly developing neurodegenerative disorder that is clinically manifested as dementia.1 The current figure for the AD burden in older adults in the United States is 6.7 million, and it is poised to rise to 13.8 million by 2060.2 The preclinical stage of AD can last many years without obvious signs of dementia.3 It is crucial to better understand this preclinical stage in order to develop successful AD prevention.4 Common preclinical features of AD include toxic protein depositions, neuronal apoptosis and reduction in hippocampal volume (brain shrinkage), and brain glucose hypometabolism.5,6 The brain glucose hypometabolism is observed long before the occurrence of overt symptoms in AD and is partly due to mitochondrial dysfunction.7 Measuring glucose utilization in the brain using positron emission tomography (PET) and 18F-fluorodeoxyglucose (FDG) allows for convenient examination of hypometabolic patterns in the brain.8 Brain scans based on FDG PET can effectively detect around 90% of AD-specific metabolic patterns, such as those in the parieto-temporal, frontal, and posterior cingulate regions.9
A large genetic component drives AD (60-80%), and the entire spectrum of the disease can develop over 15-25 years.10 Genetic variations in the APOE gene could single-handedly account for a large part of the risk related to AD in old age.11 On the other hand, addressing modifiable risk factors could reduce or delay up to 40% of dementia risk.12 Therefore, by focusing on the modifiable risk factors, a substantial part of the AD burden could be alleviated at the population level.13 Prevention of certain infections can reduce the risk of chronic diseases, including neurological deficits.14–16 Accumulating evidence suggests that infections could be a significant risk factor for AD that may also facilitate the development of AD pathology at the preclinical stage, though the exact mechanism is unclear and might involve a direct detrimental impact of infection-related factors as well as indirect effects of compromised immunity.17–20
The connection between infections and AD and related pathology may also be influenced by genetic factors.18,21,22 There are also indications that infections can contribute to brain hypometabolism, one of the earliest features of AD pathology; however, research on this topic is scarce.23 Here we explore how infectious diseases may influence brain glucose metabolism in presence and absence of APOE4, the strongest genetic risk factor for AD, in participants of the Alzheimer’s Disease Neuroimaging Initiative (ADNI).
2 Data and Methods
2.1 Study Population
ADNI is a multi-center observational study that began in 2004 under the supervision of Michael W. Weiner. The study recruited individuals within the 55-90 years age range, and enrollment in this cohort occurs in different phases, with previous participants continuing to be in the study and new participants being recruited. To compare and gain knowledge about dementia, this database maintains and updates demographic, phenotypic, biomarker, and genetic data gathered from participants with normal cognition, AD, and other forms of cognitive impairment. The availability of such a wide variety of biomarkers provides sufficient information to learn about the evolution and pathology driving AD.24
More details regarding the study design and objectives can be accessed here (https://adni.loni.usc.edu/study-design/). Broadly, ADNI seeks to integrate information from biomarkers, cognitive measures, and brain scans to improve AD diagnosis and treatment.25 Brain scans were primarily collected to learn about the structural and metabolic functions of the brain, serving as a standard for differentiating the pathological changes seen in AD from those in normal aging.26
2.2 Predictors: Infections and APOE4
Prior infections were determined by combining the information from medical history, baseline symptoms, initial health assessment, and adverse effects datasets. The details of the selected subset of infections included in the final dataset are illustrated in Supplementary Figure 1. Medical history information was collected during screening visit using a questionnaire. Non-harmonious disease names were uniformly labeled for analytical purposes. Duplicated participant information having the same infection and diagnosis date, as well as any infections lacking a diagnosis date, were subsequently excluded. Covariates such as age, sex, education, race, marriage status, and APOE4 information were retrieved from the ADNIMERGE file.
The APOE4 carrier status was identified from DNA extracted by Cogenics from a 3 mL aliquot of EDTA blood extracted from participants during their screening visit.27 Anti-diabetic medications were extracted (list provided in the Supplementary File 2) using the Anatomical Therapeutic Chemical (ATC) classification system coding (https://www.who.int/tools/atc-ddd-toolkit/atc-classification).
Information regarding smoking and alcohol usage was obtained from the medical history file. Finally, we retained infections that only preceded the HCI measurements.
2.3 Outcomes: Brain glucose hypometabolism, AD and MCI
Multiple PET scanners were used to capture brain images based on a standard protocol.28 Measures were taken to correct for the related discrepancies.29 The details regarding the PET scan and related protocols can be viewed elsewhere (https://adni.loni.usc.edu/methods/pet-analysis-method/pet-analysis/). The generated raw PET data are centrally stored at the Laboratory of Neuroimaging (LONI) at the USC Mark and Mary Stevens Neuroimaging and Informatics Institute of the University of California.30
We retrieved the processed study outcomes from the BAIPETNMRCFDG dataset (https://adni.bitbucket.io/reference/baipetnmrc.html). The main outcome of interest was the hypometabolic convergence index (HCI), developed to reflect the AD-specific hypometabolism across regions of the brain by computing voxel-wise z-scores from FDG-PET brain images. Higher HCI values correspond to lower levels of metabolism in the brain.31
Additionally, we examined the associations for infections with statistical regions of interest (SROI) corresponding to AD and Mild Cognitive Impairment (MCI). SROI associations might provide additional insights into the cerebral metabolic rate for glucose (CMRgl) decline in these brain regions, helping to understand the disease-specific pathology they represent.32 The Statistical Parametric Mapping (SPM) software was used to generate the HCI and SROI scores.32,33 The work of Landau et al. provides further details on the generation and development of regions of interest in the ADNI cohort.34 A decline in FDG-PET Region of Interest (ROI) values suggests pathological brain damage and may contribute to the progression of dementia.33
2.4 Statistical Analysis
R version 4.3.2 was used for the data linking and statistical analysis.35 We analyzed the dataset with full covariate and outcome information, without performing any imputations. The ggplot2 package was used to create variable distribution plots.36 The leptokurtic HCI readings were normalized during the RNomni package.37 Multivariate linear regression models were conducted for all specified outcomes separately. Age, education, and allele dosages of APOE ε4 were analyzed as continuous variables. Infections, AD, and diabetes medications were coded as a binary variable (yes or no). Marriage, smoking, and alcohol use were coded as Ever or Never. We explored models with a full set and a reduced set of covariates. The parsimonious model (best explanatory model) was determined using the Akaike Information Criterion (AIC) in the MuMin package.38 A two-sided p-value less than 0.05 was considered to support our hypothesis.
A Random Forest-based model was used to rank the significant variables according to their contributions to the best model.39 The effect modification for infections with HCI by APOE4 carrier status was assessed by visualizing with the rockchalk package.40 Marginal mean estimates were calculated to show the interaction effects for the infections across categories of APOE4 and sex.
2.5 Ethics Approval
The Institutional Review Board of Duke University Health System issued approval for this study (Protocol IDs Pro00109279 and Pro00105389). All participants provided written informed consent. ADNI studies follow Good Clinical Practices guidelines, the Declaration of Helsinki, and United States regulations (U.S. 21 CFR Part 50 and Part 56).
3 Results
3.1 Participant Characteristics
The final sample included information on 1,509 participants after data linking (Supplementary Figure 2). As shown in Table 1, the average age among participants was 73.3 years, with an average education duration of 16.0 years (IQR 14.0-18.0). Over 96% of respondents reported being ever married, and 55.8% were males. There was a relatively lower representation of non-white individuals, totaling 116 (7.6%) in the sample. Percentage of individuals with a history of smoking and alcohol use was 27.1% and 3.3%, respectively. Of these, 215 individuals accounting for 14.2% of the total sample size, reported having infections. The median interval between biomarker assessment and infections was 8.4 (IQR: 3.5 - 28.3). Median HCI was 12.59, and the IQR was 8.4 - 19.3.
Figure 1 shows the distributions of the original HCI and rank-normalized HCI and also a scatterplot of their relationship. For AD and MCI participants, the mean SROI values were 1.15 and 1.03, respectively. About 3% of participants with diabetes were on medication, and 18.3% had an AD diagnosis. Peptic ulcer disease (PUD) (n=152), urinary tract infection (UTI) (n=146), and pneumonia (n=102) were the most frequent among the selected infections. Supplementary Figure 3 shows the difference in the distribution of HCI values for individuals with infections, AD, and APOE4. The median HCI value among individuals with infections was 13.64, while it was lower (12.48) for those without infections. It was also seen that the HCI had a modest positive correlation with APOE4 (Supplementary Figure 4).
3.2 Association of Infections and Other Predictors with the HCI
Supplementary Table 1 shows the regression estimates for all the predictors in the multivariate linear regression full model for HCI outcome. Marriage status, education, smoking, alcohol, and diabetes medication use were not significant predictors of HCI. Table 2 presents the results of the reduced model, which best describes the model variance. AD status predicted the strongest reduction in brain metabolism [β = 1.04, 95% CI 0.92-1.15, p<0.001], followed by age [β = 0.01, 95% CI 0.01-0.02, p<0.001] and APOE4 carrier status [β = 0.32, 0.25-0.38, p<0.001]. Higher variable relevance is indicated by higher values of %INCMSE and INCNodepurity (Supplementary Table 2). The regression coefficient for infections was 0.15 [95% CI 0.02-0.27, p=0.01]. Males and white people were at higher risk of having elevated HCI values. Smoking history was the only non-significant predictor retained in the reduced model. The adjusted R-squared from the reduced model was 26.9%. Males had higher median HCI values.
In the sex-stratified analysis evaluating the effects of infections versus non-infections, males generally demonstrated relatively higher HCI values (Supplementary Table 3). The difference in normalized marginal means between all groups was statistically significant (p<0.001). The combined effects of infections and APOE4 carrier status on HCI levels are shown in Figure 2. This was significantly greater than the effects of either variable alone. Supplementary Table 4 clarifies these results. Specifically, for individuals without infections and APOE4 carrier status, the estimated marginal mean was 0.03 (p=0.53). However, this increased significantly to 0.18 (p<0.001) for APOE4 non-carriers in the presence of infections. Notably, among APOE4 carriers, the estimated marginal mean was substantially higher at 0.62, and this value rose to 0.77 with infections (p<0.001), confirming an interaction between the two factors.
This interaction was further demonstrated in the additional analysis (Supplementary Table 5 and Figure 3), indicating that carriers who experienced multiple infections exhibited greater hypometabolism. Among individuals with more than one prior infection (n=23), the estimate was significantly higher at 0.44 (p=0.01) compared to those with a single infection, which was 0.11 (p=0.08), revealing a dose-response relationship.
3.3 Association of Infections and Other Predictors with the SROI AD
Supplementary Table 6 provides regression estimates for all the factors investigated for SROI AD. Marriage, race, smoking, and alcohol history were not significant predictors of AD-specific hypometabolism. In the reduced model shown in Table 3, AD was associated with increased region-specific hypometabolism (regression coefficient: -0.08, p<0.001). The use of diabetes medications was associated with decreased brain metabolism (-0.03, p=0.02). Similar to previous regression, an increase in APOE4 alleles was a strong risk factor for hypometabolism (-0.02, p<0.001). Male gender showed greater hypometabolism (-0.01, p<0.01). Although education was linked to a better metabolic pattern, this relationship was not profound. Age-specific decreases were not as notable as those observed in HCI (-0.003, p<0.001). While statistically significant, the effect estimate for previous infections was lower for AD (-0.01, p=0.02). These variables collectively predicted 26.8% of the variance in SROI AD.
3.4 Association of Infections and Other Predictors with the SROI MCI
The results of the SROI MCI regression (full model) is presented in the Supplementary Table 7. Generally, the estimates were closer to the SROI AD than HCI. Among the variables that best explained the model (Table 4), AD, APOE4, and diabetes medications had the largest effect estimates. Sex and use of diabetic medications had a marginally greater impact on the MCI region than on the AD region. However, race and education were not identified as significant predictors. Infections were associated with a -0.01 reduction in regional metabolism (p=0.04). The percentage of variation explained by the model for the SROI MCI was also the highest (28.8%) of the three investigated outcomes.
4 Discussion
Results of our study suggest that infections and APOE4 can jointly significantly affect brain glucose metabolism, specifically promote hypometabolism, as measured by the increased values of HCI. A history of infections in this ADNI sample corresponds to a greater hypometabolism, specifically a 0.15 unit increase in rank normalized HCI. However, this estimate rises to 0.44 in the presence of multiple infections. Model inclusion of established confounders such as age, sex, race, and education did not diminish these findings. We also adjusted for AD status, which was more prevalent in the group with no infections and associated with reduced brain metabolism. Adjusting for AD status and APOE4 was necessary to reveal the genuine association of previous infections. Additionally, previous infections were significantly associated with regional brain metabolism specific to AD and MCI in our data.
Our findings agree with previous research suggesting that infections may negatively impact brain metabolism .16,41–43 Infectious diseases, including those addressed in this study, have been previously linked to AD in other data.22,44–46 Our recent paper that used Health and Retirement Study (HRS) data reported associations between AD and various infectious (viral, bacterial, fungal), suggesting that compromised immunity may play a role in AD etiology.20 The connection between infections and brain hypometabolism may also involve pathological immune responses. Some research provides indirect support to this idea by linking brain hypometabolism to microglia activation.47–49
Determining the onset time of infection is a major challenge in AD research. Furthermore, the causal inferences are obscured by the fact that individuals with AD often grapple with a variety of infections due to declining immunity, leading to elevated antimicrobial markers.16 Pathogens have a high affinity to the central nervous system and brain tissues and could affect cognition.50 Given that brain hypometabolism is an early sign of AD, our findings suggest that infections could potentially trigger this process.51 However, the progression of hypometabolism may also depend upon the combination of other risk factors.52 Infections can affect the brain through multiple pathways, both directly and indirectly, particularly when the blood-brain barrier is breached.53,54 Infections propagated through the respiratory route can also reach the brain relatively easily.50,55 However, upon reaching the brain, different infectious agents employ their preferred mechanisms, such as latent activation and the initiation of inflammation, as seen in the case of the Herpes virus.56 Pathogen invasion into the brain leads to chronic inflammation, which can compromise the blood-brain barrier.42 There are distinct differences in inflammatory pathways noted across specific pathogens.19,57 Age-related changes could exacerbate these pathological processes even further.58 Strom and colleagues have also demonstrated that brain hypometabolism correlates with tau pathology and neurodegeneration in crucial dementia-specific regions within the ADNI dataset.59 These mechanisms could potentially explain a significant portion of the biological processes leading up to hypometabolism.
Studies on the relationship between APOE4 and brain metabolism have produced contrasting findings.59–63 In their recently published work, Fortea and colleagues found that simply being homozygous for APOE4 is sufficient, in most cases, to guarantee an AD diagnosis.64 In our analysis, the increase in APOE4 allele was associated with all three outcomes and showed compounding effects with infections and their burden. Even in patients with a single APOE4 variant, which is usually not considered a significant increase in risk compared to homozygous carriers, the presence of infections increases the risk of hypometabolism to nearly the same level as in homozygous APOE4 carriers. One possibility is that the observed effect is due to accelerated neuroinflammation arising from the presence of both risk factors.65 Risk factors for AD tend to cluster in individuals with APOE risk alleles, including a reduction in brain metabolism.66 Amyloid-beta and Tau deposition are higher in APOE4 carriers.67 APOE4 can also accelerate brain degeneration through non-overlapping pathways independent of amyloid deposition and Tau pathology.68–70 APOE4 alleles both promote and resist infections, depending on the type of infection.71 Researchers suggest that APOE4 polymorphisms result in increased lipid production72 and blood-brain barrier loss73, which could facilitate a conducive environment for pathogens.74 Supporting evidence from the Northern Manhattan Study showed that the effect-modifying relationship between APOE4 and infectious burden was correlated with decreased cognition.75 The influence of APOE4 on AD remains incompletely understood, although it is known to engage in intricate interactions with other risk factors for AD, such as age.76,77 However, in stark contrast to these findings, a study reported that the effects of APOE4 on cognition are AD-specific. It singles out the cause of cognitive decline as the interaction between APOE4 and amyloid beta in the hippocampus.78
Our study revealed that diabetes medication was the third-biggest risk factor for AD and MCI-specific brain metabolism, but not for the HCI measure. Previous studies indeed demonstrated that diabetes increases the risk for MCI and AD.79,80 Individuals with diabetes and AD often share common biological pathways.81 Most prominent among these are low-grade chronic inflammation and insulin resistance.82
Sex differences in mechanisms related to AD warrant in-depth study. Usually, females are susceptible to AD and early brain hypometabolism compared to males.83,84 On exposure to prior infections, women are also, particularly at higher risk for reduced hippocampal volume.19 Males overall had a higher HCI value than females in our sample. It is important to note that males with infections had a slightly higher mean age. But this alone cannot explain the gender difference. Importantly, there was no difference in the increase in marginal means due to infections for both sexes. An earlier study reported that brain hypometabolism increased in men after 70 years of age, while this was not seen in females in a normal brain.85 However, the applicability of this finding in the AD context needs confirmation. Some participant characteristics in ADNI may differ from the general population due to voluntary recruitment. Variations in the distribution of AD risk factors among genders might also contribute to this finding.86,87
Given that AD is not curable, prevention stands as the most viable option at present. Vaccinations may potentially alleviate AD risk. Influenza vaccines, in particular, are among the candidates demonstrating this preventive potential.4,79 However, personal genetics could play a role in determining the efficacy and effectiveness of vaccinations. Recent research has revealed that individuals carrying a polymorphism in the NECTIN2 gene exhibit a decreased susceptibility to AD when compared to non-carriers, when receiving vaccinations for pneumonia and flu.22
The availability of medical history information and longitudinally standardized FDG PET measurements were important strengths of our study. We were also able to demonstrate the temporality of association, which was rarely described in earlier human studies.88 There were a couple of study limitations. Of these, the most important is that the medical history is questionnaire-based, suggesting that recall bias may exist and lead to an incorrect exposure classification. Currently, the representation of high-risk groups, such as Afro-American and Hispanic individuals, is limited in the ADNI database, which has constrained the evaluation of the effect modification role of race in relation to infections and brain hypometabolism.89 In this work, we did not specifically explore the heterogeneity in infections and the brain metabolism relationship. However, previous AD studies indicate that there could be subgroups that may be differentially vulnerable.90–92 There may also be a cohort effect, wherein the frequency of infections observed within this group may not accurately fit the current disease landscape. We recommend validating the findings in large cohorts with robust information on prior infections.
5 Conclusion
This study found that infections and APOE4 jointly promote brain glucose hypometabolism in older ADNI participants. In individuals with history of infections who were also carriers of one APOE4 allele, the degree of brain glucose hypometabolism was nearly that seen in APOE4 homozygotes without prior infections. We conclude that prior infections may contribute to AD pathology in synergy with APOE4, thus playing a part in the “multi-hit” mechanism of AD development.
Data Availability
The data used in this manuscript were obtained from the publicly available ADNI database (adni.loni.usc.edu). The ADNI database contains anonymized patient information, making it a secure data repository.
Funding
This research was supported by the National Institutes of Health’s National Institute on Aging (NIA/NIH) grants R01AG076019 and R01AG062623. This content is solely the responsibility of the authors and does not necessarily represent the official views of the NIA/NIH.
Consent for publication
Not applicable
Funding
This research was supported by the National Institute on Aging of the National Institutes of Health under Award Numbers R01AG076019, R01AG070487. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Contributions
S.U., K.G.A., A.I.Y., and A.L.R. were involved in the hypothesis generation, study design, and study supervision, and they critically revised the manuscript. A.L.R. and O.B. were responsible for data cleaning. A.L.R. contributed to the analysis, interpretation of the data, and manuscript writing. All authors read and approved the submitted version of the manuscript.
Competing Interests
The authors declare no competing interests.
Additional Information
Supplementary Material 1
Supplementary Material 2
Acknowledgements
Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.;Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.;Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.
Abbreviations
- AD
- Alzheimer’s disease
- AIC
- Akaike Information Criterion
- APOE
- Apolipoprotein E
- Aβ
- Amyloid βeta
- FDG
- 18F-fluorodeoxyglucose
- GxE
- Gene-environment interaction
- GWAS
- Genome-wide association studies
- HSV
- Herpes Simplex Virus
- %INCMSE
- Percent Increase in Mean Squared error
- IQR
- Interquartile Range
- PET
- Positron Emission Tomography
- pTau
- Phosphorylated Tau
- SD
- Standard Deviation
- SNP
- Single Nucleotide Polymorphism
- SROI
- Statistical Region of Interest
- NECTIN2
- Nectin Cell Adhesion Molecule 2 (gene)
- PUD
- Peptic Ulcer Disease
- UTI
- Urinary Tract Infection
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.↵
- 47.↵
- 48.
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.
- 61.
- 62.
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.
- 92.↵
