Association of spermidine plasma levels with brain aging in a population-based study

Abstract

INTRODUCTION:

Supplementation with spermidine may support healthy aging, but elevated spermidine tissue levels were shown to be an indicator of Alzheimer’s disease (AD).

METHODS:

Data from 659 participants (age range: 21–81 years) of the population-based Study of Health in Pomerania TREND were included. We investigated the association between spermidine plasma levels and markers of brain aging (hippocampal volume, AD score, global cortical thickness (CT), and white matter hyperintensities (WMH)).

RESULTS:

Higher spermidine levels were significantly associated with lower hippocampal volume (ß=−0.076; 95% CI: −0.13 to −0.02; q=.026), higher AD score (ß=0.118; 95% CI: 0.05 to 0.19; q=.006), lower global CT (ß=−0.104; 95% CI: −0.17 to −0.04; q=.014), but not WMH volume. Sensitivity analysis revealed no substantial changes after excluding participants with cancer, depression, or hemolysis.

DISCUSSION:

Elevated spermidine plasma levels are associated with advanced brain aging and might serve as potential early biomarker for AD and vascular brain pathology.

1. INTRODUCTION

Polyamines are naturally occurring cell metabolites of living organisms that have gained much attention in recent years due to their function in cellular longevity and their crucial role in protein synthesis, nucleic acid synthesis, protection of cells and genes from oxidative damage, regulation of DNA methylation, and general cell development.^1,2 They can be synthesized endogenously or ingested with food.³ The main representatives of polyamines are spermidine and spermine.

Interestingly, polyamines have been linked to positive as well as negative health effects. On the one hand, increased dietary polyamine intake has been reported to improve cognitive and general health and to slow the aging process, and has been promoted for the prevention of dementia.⁴ Especially dietary spermidine was shown to be associated with benefits on different systems (i.e., cardio-vascular system, brain structure and function, and immune system) in animal models.⁴

Consistent with previous research in animal models, there is also evidence that dietary spermidine intake might improve memory performance in older individuals with subjective cognitive decline (SCD) and patients with mild cognitive impairment (MCI).^5–7 In addition, a positive relationship has been shown between dietary spermidine intake and hippocampal volume, as well as cortical thickness (CT) in older adults.⁸ Finally, there is evidence that increased spermidine intake might be associated with a lower mortality rate in the general population.⁹ However, supplementation studies were small in sample size and therefore results may not be generalizable. In fact, a recently published larger longitudinal clinical trial failed to show a positive relationship between spermidine supplementation and memory performance in their primary analysis.¹⁰ Moreover, it is still controversial if and to what degree spermidine is able to cross the blood-brain barrier.^11,12

On the other hand, higher tissue polyamine levels have been reported in various diseases like dementia due to Alzheimer’s Disease (AD),^13–15 Parkinson’s disease,¹⁶ cancer,¹⁷ and mental disorders like depression.¹⁸ For instance, several postmortem studies could show that polyamine tissue levels (i.e., putrescine, spermine, and spermidine) were elevated in frontal and occipital lobes,¹⁹ or in temporal cortex, white matter and thalamus in patients with AD.^14,20 Moreover, Paik and colleagues demonstrated that urinary polyamine levels in AD patients were significantly elevated compared to healthy individuals.²¹ Although the exact mechanisms between polyamine metabolism and the development of AD are not yet clear, it is assumed that an increase of ornithine decarboxylase as a key enzyme of polyamine synthesis, as well as a maladaptive stress response of the polyamine metabolism due to peptide-associated free radical damage, may possibly underlie its pathogenic role.¹⁵ Polyamines are assumed to be toxic to brain cell cultures, and polyamine injections have been shown in animal models to damage the blood brain barrier.²² Finally, using pathway enrichment analysis, Graham and colleagues demonstrated that patients with MCI who converted to dementia had a higher spermidine and spermine production at baseline compared with MCI patients who remained stable over time.¹³ Additionally, based on blood-based high-resolution mass spectrometry metabolomics, they were able to accurately predict those MCI patients converting to dementia two years prior to dementia diagnosis.

Although previous studies investigated the role of polyamine levels in brain tissue, urine, and blood in patients with AD and in animal models of AD, little is known about the role of polyamine blood levels during aging in the general population. In addition, there is conflicting evidence on the effect of spermidine supplementation versus spermidine tissue levels in brain health. Finally, the presumed role of spermidine as a predictive biomarker for conversion from MCI to dementia due to AD raises the possibility that spermidine blood levels may also be a promising biomarker to predict onset of neurodegenerative disease in healthy cohorts. Therefore, the aim of the present work was to investigate the association between spermidine plasma levels and brain health in a population-based cross-sectional study by using neuroimaging markers for AD, to gain more insights into the role of spermidine across the lifespan.

We hypothesized that higher spermidine plasma levels would be associated with less advantageous grey and white matter patterns related to advanced brain aging. In particular, we hypothesized that higher spermidine blood levels would be associated with lower hippocampal volume and lower global CT, higher AD score (i.e., greater AD-related brain atrophy), and greater white matter hyperintensities (WMH) volume.

2. METHODS

2.1. Study sample

The Study of Health in Pomerania (SHIP) was designed to determine the prevalence of common risk factors and diseases in a population in northeastern Germany.²³ For this purpose, a sample was randomly drawn from local registers. Between 1997 and 2001, 4,308 individuals participated in the study. In parallel to the original SHIP-START study, a new independent sample was drawn to conduct research of similar scope (SHIP-TREND). The data used in our analyses were derived from this SHIP-TREND cohort, which was initiated 10 years after SHIP-START in the same region. A total of 4,420 individuals participated in the SHIP-TREND study. All participants gave written informed consent. The study was approved by the ethics committee of the University Medicine Greifswald (no. of ethics vote: BB39/08) and complies with the Declaration of Helsinki. Data used in our analyses originates from baseline examinations, which took place between 2008 and 2012 (SHIP-TREND-0). A detailed description of the assessment of all included variables can be found in the Supplementary Material.

2.2. Statistical analysis

Statistical analysis was performed using R version 3.6.1.²⁴ For multivariable regression analyses, we used the ‘lm’ function for fitting ordinary least squares regression models.²⁴ MRI markers of brain aging (i.e., global hippocampal volume, AD score, global CT, and WMH) were used as outcome variables. Three nested models were implemented: The first basic model examined the relationship between the exposure variable spermidine and the respective outcome variable, adjusted for total intracranial volume (ICV). A second model was additionally adjusted for age, sex, red blood count (RBC), and the interaction of age and sex (spermidine confounder model), which were previously identified as possible cofounder variables of spermidine. For instance, previous studies showed that spermidine levels decrease with increasing age,²⁵ and 90% of free circulating spermidine is associated with red blood cells.²⁶ Moreover, age and sex are known to be associated with brain aging.^27,28 A final model was additionally adjusted for the exposure variables C-reactive protein, waist-to-hip-ratio, years of education, dietary habits, alcohol consumption, smoking status, and ApoE4 (dementia risk factor model) to account for well-known dementia risk factors.²⁹ Total ICV was added to each model since it highly correlates with brain structure sizes.³⁰ Additionally, in the spermidine confounder and dementia risk factor model age and the interaction of age and sex were modelled not only linearly but also quadratically to account for non-linear age effects on the respective outcome variable. Finally, to reach normal distribution a log transformation was applied to spermidine, C-reactive protein, and WMH volume. We used False Discovery Rate (FDR) to control for multiple testing.³¹ Results were considered statistically significant at a q-value of < 0.05. Participants with missing data on one of the variables of interest were excluded from analysis.

As sensitivity analyses, we reran the dementia risk factor model for each brain aging marker after excluding participants with cancer diagnosis as well as moderate and severe depression (N = 56, see Suplementary Materials for details on the assessment) since it is well known that individuals with cancer and depression show elevated spermidine blood level values compared to healthy individuals.^17,18

Additionally, to rule out that hemolysis had affected our results, we re-ran the dementia risk factor model for each brain aging marker after excluding participants with elevated lactat dehydrogenase or potassium values (N = 27, see Supplementary Material for details on the assessment), since both blood markers are known to indicate hemolysis and 90% of free circulating spermidine is known to be associated with red blood cells.²⁶

3. RESULTS

3.1. Descriptive statistics

Descriptive characteristics of the study sample are shown in Table 1. Our final sample comprised 659 participants (mean age 50.1 years ± 13.5 years, 58% women). Participants with structural abnormalities of the brain (e.g., tumors, cysts, ventricular dilatations; N = 44) and low segmentation quality of WMH (e.g., over- or under-segmented WMH; N = 115) were excluded. Additionally, participants with spermidine plasma values exceeding ± 3 SD of the mean plasma concentration were also excluded from the study sample (N = 2). See Fig. 1 for a flowchart showing the selection of the study sample. Most participants (75.3%) indicated to be physically active (i.e., including exercise during summer and winter time). The PCS-12 and MCS-12 scores revealed averages of 51.11 and 55.45 points, respectively, with 100 points being the maximum score corresponding to excellent physical and mental health.³² A moderate and recommended dietary habit was present in 19.1% and 15.6% of the participants, respectively. Additionally, 2.4% of participants showed moderate or severe symptoms of depression. A total of 44% and 35.5% were never smokers and former smokers, respectively. Furthermore, 23.1% of participants were ApoE ε4 allele carriers (i.e., either ε3/ε4 or ε4/ε4) and 5.5% had received a cancer diagnosis at least once in their lifetime.

Table 1:

Descriptive characteristics of study sample

Characteristics	Median (95% CI) or N (%)
Sociodemographic parameters
Age, [yr]	51.00 (40.50, 60.00)
Female, N (%)	380 (58)
Education, [yr]	13 (11, 15)
Blood parameters
Red blood count [Tpt/l]	4.60 (4.30, 4.90)
White blood count [Gpt/l]	5.46 (4.74, 6.42)
CRP [mg/l]	1.13 (0.61, 2.47)
Spermidine [μM]	0.12 (0.10, 0.14)
Brain imaging measures
Alzheimer’s Disease Score	−4.48 (−5.39, −3.80)
White Matter [ml]	601.71 (549.86, 655.31)
Grey Matter [ml]	574.60 (530.28, 618.69)
Global Hippocampal volume [mm3]	6918.34 (6514.07, 7353.55)
Global Cortical Thickness [mm]	4.70 (4.54, 4.84)
White Matter Hyperintensities volume [mm3]	171 (69, 421)
Intracranial Volume [L]	1.57 (1.46, 1.67)
Lifestyle
Dietary Habit, N (%)
poor	430 (65.3)
moderate	126 (19.1)
recommended	103 (15.6)
Alcohol consumption [ethanol in g/d]a	3.92 (1.26, 9.78)
Smoking status, N (%)
non-smoker	290 (44)
previous smoker	234 (35.5)
smoker	135 (20.5)
Physical activity, N (%)
not active	163 (24.7)
active	496 (75.3)
Waist-to-Hip-Ratio	0.86 (0.81, 0.93)
BMI [kg/m2]	26.79 (24.13, 29.62)
Mental health
Depressionb, N (%)	16 (2.44)
moderate	11 (1.68)
severe	5 (0.76)
Mental health scorec	55.45 (50.83, 57.29)
Physical health
Physical health scorec	51.11 (45.92, 54.83)
Cancer diagnosis, N (%)	36 (5.5)
Genetics
ApoE4, N (%)	152 (23.1)

Values are presented as median and interquartile range or N and %.

^aAlcohol consumption was calculated as alcohol consumption during the last 30 days.

^bN = 654.

^cN = 648.

Abbreviations: APOE4, Apolipoprotein ε4 allele carrier; CRP, C-reactive protein; BMI, body mass index.

Figure 1.

Flowchart showing the selection of the study sample. The final sample comprised 659 adults from the SHIP-TREND-0 study. MRI, magnetic resonance imaging

3.2. Association between Spermidine and Markers of Brain Aging

Table 2 shows the association between spermidine and the four different markers of brain aging considered in this study. Results for all regression models including all covariates can be found in the Supplementary Material (Table S1–S4).

Table 2:

Association between Spermidine and Markers of Brain Aging

	Spermidine
	ß (95% CI)	q-value
Global Hippocampal volume
Basic model	−0.063 (−0.12, −0.01)	.047
Spermidine confounder model	−0.076 (−0.13, −0.02)	.013
Dementia risk factor model	−0.076 (−0.13, −0.02)	.026
Alzheimer’s Disease Score
Basic model	0.103 (0.03, 0.18)	.026
Spermidine confounder model	0.117 (0.04, 0.19)	.005
Dementia risk factor model	0.118 (0.05, 0.19)	.006
Global Cortical Thickness
Basic model	−0.125 (−0.20, −0.05)	.004
Spermidine confounder model	−0.105 (−0.17, −0.04)	.004
Dementia risk factor model	−0.104 (−0.17, −0.04)	.014
White Matter Hyperintensities volume
Basic model	0.073 (0.00, 0.15)	.093
Spermidine confounder model	0.082 (0.02, 0.15)	.045
Dementia risk factor model	0.081 (0.03, 0.16)	.103

ß: standardized regression coefficient; 95% CI with lower and upper bound; q: q-value. A log transformation was applied to spermidine, C-reactive protein, and white matter hyperintensities volume. Basic model: adjusted for ICV. Spermidine confounder model: adjusted for basic model variables and age, age², sex, RBC, the interaction age × sex and age² × sex. Dementia risk factor model: adjusted for spermidine confounder model variables and C-reactive protein, waist-to-hip-ratio, years of education, dietary habits, alcohol consumption, smoking status and ApoE4. All p-values are FDR-corrected for multiple testing (e.g., q-value). Abbreviations: CI, confidence interval; FDR, False Discovery Rate; ICV, total intracranial volume; RBC, red blood count; APOE4, Apolipoprotein ε4 allele carrier. Detailed results for all regression models including all variables can be found in the Supplementary Materials (Tables S1–S4).

3.2.1. Association between spermidine and global hippocampal volume

To evaluate the relationship between spermidine blood values and global hippocampal volume in our sample we ran a first regression model (basic model), which revealed an inverse association between both variables (ß = −0.063, q =.047). We ran a second partially adjusted multivariable regression model (spermidine confounder model). This regression model explained 54% of the variance [R² =.54, adj. R² = .53, F(8, 650) = 93.74] and revealed an inverse association between spermidine and hippocampal volume (ß = −0.076, q = .013). The final fully adjusted regression model (dementia risk factor model) explained 54% of the variance [R² =.54, adj. R² = .53, F(17, 641) = 44.99] and revealed again an inverse association between spermidine and hippocampal volume (ß = −0.076, q = .026). Thus, all three regression models consistently indicate that higher spermidine blood levels are associated with lower hippocampal volume. Detailed results are shown in Table S1.

3.2.2. Association between spermidine and AD score

The basic model revealed a positive association between spermidine and AD score (ß = 0.103, q = .026). To further investigate this relationship, we ran the spermidine confounder model. This regression model explained 21% of the variance [R² =.21, adj. R² = .20, F(8, 650) = 22.05] and showed a positive association between spermidine and AD score (ß = 0.117, q = .005). The fully adjusted dementia risk factor model explained 25% of the variance [R² =.25, adj. R² = .23, F(17, 641) = 12.27] and again revealed that spermidine was positively associated with AD score (ß = 0.118, q = .006). All three regression models consistently indicate that higher spermidine blood levels are associated with higher AD scores. Detailed results are shown in Table S2.

3.2.3. Association between spermidine and global cortical thickness

The basic model revealed an inverse association between spermidine and global CT (ß = −0.125, q = .004). The spermidine confounder model explained 38% of the variance [R² =.38, adj. R² = .37, F(8, 650) = 49.45]. Spermidine blood levels were inversely associated with global CT (ß = −0.105, q = .004). Finally, the dementia risk factor model explained 40% of the variance [R² =.40, adj. R² = .38, F(17, 641) = 24.69] and confirmed again the already found inverse association between spermidine and global CT (ß = −0.104, q = .014). All three regression models consistently indicate that higher spermidine blood levels are associated with lower CT. Detailed results are shown in Table S3.

3.2.4. Association between spermidine and WMH volume

Finally, the basic model revealed no significant association between spermidine and WMH volume (ß = 0.073, q = .093). The spermidine confounder model explained 34% of the variance [R² =.34, adj. R² = .33, F(8, 650) = 41.53]. Spermidine plasma levels were positively associated with WMH volume (ß = 0.082, q = .045). The fully adjusted dementia risk factor model explained 35% of the variance [R² =.35, adj. R² = .33, F(17, 641) = 19.89]. After FDR-correction for multiple testing, spermidine was not significantly associated with WMH volume (ß = 0.081, q = .103, p = .017). Taken together, the spermidine confounder model indicates that higher spermidine blood levels are associated with greater WMH volume. Although the association between spermidine and WMH volume in the dementia risk factor model was not significant after FDR-correction, the standardized regression coefficient for this model supports the relation between spermidine and WMH volume found in the spermidine confounder model. Detailed results are given in Table S4.

3.3. Sensitivity analyses

Since elevated spermidine levels are also common in individuals with cancer and depression in comparison to healthy individuals, we performed a sensitivity analysis excluding participants with cancer diagnosis and moderate and severe depression. This analysis did not substantially change results related to the association between spermidine and brain imaging markers. Detailed results including all variables are given in Table S5.

Furthermore, as 90% of free circulating spermidine is associated with red blood cells²⁶, the presence of hemolysis could have affected spermidine plasma levels. Therefore, in a second sensibility analysis we excluded participants with indication of possible hemolysis. Again, this analysis did not substantially change results related to the association between spermidine and brain imaging markers. Detailed results including all variables are given in Table S6.

4. DISCUSSION

The aim of the present study was to investigate the association of spermidine blood levels and brain aging across the lifespan in a community-based cross-sectional sample. In particular, we were interested in the association of spermidine with four different imaging-based biomarkers of brain aging and AD, i.e., global hippocampal volume,³³ global CT,³⁴ WMH volume,³⁵ and AD score.³⁶

We found that increased spermidine levels were associated with advanced brain aging independent of dementia risk factors (i.e., age, C-reactive protein, waist-to-hip-ratio, dietary habits, alcohol consumption, smoking, and ApoE4 allele carrier status) and potential confounders of spermidine levels in peripheral blood (i.e., age, sex, and RBC). More specifically, elevated spermidine plasma levels were associated with lower hippocampal volume, higher AD score and lower CT volume in the dementia risk factor model. Additionally, spermidine plasma levels were associated with greater WMH volume in the spermidine confounder model. Although the association between spermidine and WMH volume failed to reach significance after FDR-correction for multiple testing in the dementia risk factor model, the standardized β-estimate revealed the same inverse tendency.

Our results are substantiated by previous cross-sectional evidence on the role of spermidine tissue and blood levels in patients with MCI and with dementia due to AD. Evidence on the association of elevated polyamine, in particular spermidine, tissue levels and AD is mainly based on post-mortem studies.^14,19,37 It has been shown that spermidine tissue levels were elevated in the temporal cortex, white matter and thalamus in patients with AD compared to spermidine tissue levels of participants without neurodegenerative disease.^14,20 This increase of spermidine in brain tissue from deceased patients with dementia due to AD indicates that an alteration of the polyamine metabolism might be involved in AD pathology.

However, evidence on the role of spermidine blood levels in AD is limited and partially contradicting.^13,38,39 In line with our findings, the study of Graham and colleagues showed that polyamine metabolism was altered in MCI patients who later transitioned to dementia leading to elevated spermidine (and spermine) production.¹³ Similarly, Sternberg and colleagues showed that polyamine plasma levels were increased in patients with MCI, as compared to healthy controls.³⁹

In contrast, a 4-year longitudinal study including patients with MCI, dementia due to AD and healthy controls observed that spermidine plasma levels were lower in patients with MCI and dementia compared to healthy controls, whereas spermine plasma levels were increased in patients with MCI.³⁸ Moreover, our results seem to contradict previous findings on the effects of higher dietary spermidine intake and of spermidine supplementation on brain health, including brain structure and memory performance.^5–8

However, one study showed that higher dietary spermidine intake was associated with smaller CT and smaller left hippocampal volume,⁴⁰ and a larger supplementation study in humans could not demonstrate an improvement in memory performance.¹⁰ In fact, there is still controversy as to whether spermidine can cross the blood-brain barrier in non-diseased states. Some reports have pointed out that spermidine only crosses the blood-brain barrier in animal models suffering from fatal stress,⁴¹ but there is also some evidence that spermidine can cross the blood brain barrier in healthy aged mice.¹² Moreover, supplementation studies have not consistently shown an increase in spermidine blood levels following dietary spermidine supplementation.^6,7,42–44 One possible explanation could be that polyamine metabolism is highly regulated and can thus preserve individual polyamine concentrations in blood over a wide range of dietary intake.⁴ Therefore, short-term increases of polyamine intake will not change polyamine blood levels;⁴⁵ rather, an increase in dietary polyamine intake over at least 26 weeks might be necessary to detect significant changes in blood levels, as shown in mice.⁴²

Mechanisms underlying the detrimental association of spermidine blood levels with brain health still remain to be fully elucidated. Interestingly, disruption of blood-brain barrier has been reported in animal models after polyamine injections.²² Accordingly, it may be possible that elevated polyamine blood levels could disrupt the blood-brain barrier under certain circumstances. This in turn may facilitate the penetration of free radicals, viruses, and bacteria into the central nervous system. A theory also suggested by Zheng and colleagues.⁴⁶ In addition, mechanistic explanations for the beneficial effects of polyamines on brain and overall health have been questioned, such as their putative ability to directly induce autophagy by inhibiting lysine acetyltransferase EP300.⁴⁷ Other studies have shown that in the presence of ruminant serum, exogenously added polyamines do not play a direct role in autophagy induction. Rather, autophagy may have been induced artificially by toxic metabolites such as aldehydes, peroxides and ammonia, byproducts resulting from the oxidation of spermidine and spermine by bovine serum amine oxidase.⁴⁸ To ensure that our results were not driven by other possible pathological processes that may lead to elevated spermidine plasma levels, we performed two sensitivity analysis excluding participants with cancer diagnosis and moderate and severe depression or hemolysis, showing almost the same pattern as in our main analyses. Therefore, our findings indicate the robustness of the association of spermidine for brain aging beyond pathologies in which polyamines play an already well-described decisive role.

Taken together, our findings consistently show an association between higher blood-derived spermidine levels and lower brain health, as indicated by several imaging biomarkers for brain aging and AD. To the best of our knowledge, this is the first study that investigated the role of spermidine blood level values across the lifespan in a community-based cross-sectional sample. While previous studies have focused on the role of elevated spermidine levels in populations with preexisting neurodegenerative diseases like MCI or dementia due to AD and Parkinson’s disease, our study is the first to show that these inverse associations already exist in individuals without previously diagnosed neurodegenerative disease or cognitive impairment. Moreover, we show for the first time that elevated spermidine blood levels are associated with white matter pathology, besides the already known association with grey matter atrophy. Our findings support the hypothesis that spermidine could not only be an indicator of incipient pathological states but may also serve as a potential biomarker for preclinical AD, and a predictor for further clinical deterioration.¹³ Given the currently ongoing intensive search for blood-based biomarkers for AD that entail lower costs and less invasive procedures compared to biomarkers found in cerebrospinal fluid,⁴⁹ and still unsatisfactory sensitivity and specificity of amyloid beta, phosphorylated tau, neurofilament light chain, or glial fibrillary acidic protein in peripheral blood,⁵⁰ spermidine might be an interesting addition or possibly even alternative to these parameters.

4.1. Limitations

Several limitations should be considered when interpreting our results. First, this study is cross-sectional; therefore, it is not possible to draw causal conclusions. Nevertheless, our study, which includes a large data set drawn from the general population, provides important insights on how elevated spermidine plasma levels are associated with markers of brain aging. Moreover, we currently have no information if participants included in this study developed neurodegenerative diseases at a later time point, an issue to be addressed in future studies.

4.2. Conclusion

Our results show for the first time that higher spermidine levels are associated with more advanced cortical and subcortical brain aging in the general population. Thus, physiological spermidine blood levels do not reflect the beneficial health effects observed with supplementation in animal models. Rather, elevated spermidine blood levels are associated with more advanced brain aging and might even play a role as a potential biomarker for AD and vascular brain pathology. Future longitudinal research must show to which extent elevated spermidine blood levels are specific and prognostic for neurodegenerative diseases.