New Results

Metabolite T2 relaxation times decrease across the adult lifespan in a large multi-site cohort

Kathleen E. Hupfeld, Saipavitra Murali-Manohar, Helge J. Zöllner, Yulu Song, Christopher W. Davies-Jenkins, Aaron T. Gudmundson, Dunja Simičić, Gizeaddis Simegn, Emily E. Carter, Steve C. N. Hui, Vivek Yedavalli, Georg Oeltzschner, Eric C. Porges, Richard A. E. Edden
doi: https://doi.org/10.1101/2024.06.19.599719

Abstract

Purpose Relaxation correction is crucial for accurately estimating metabolite concentrations measured using in vivo magnetic resonance spectroscopy (MRS). However, the majority of MRS quantification routines assume that relaxation values remain constant across the lifespan, despite prior evidence of T2 changes with aging for multiple of the major metabolites. Here, we comprehensively investigate correlations between T2 and age in a large, multi-site cohort.

Methods We recruited approximately 10 male and 10 female participants from each decade of life: 18-29, 30-39, 40-49, 50-59, and 60+ years old (n=101 total). We collected PRESS data at 8 TEs (30, 50, 74, 101, 135, 179, 241, and 350 ms) from voxels placed in white-matter-rich centrum semiovale (CSO) and gray-matter-rich posterior cingulate cortex (PCC). We quantified metabolite amplitudes using Osprey and fit exponential decay curves to estimate T2.

Results Older age was correlated with shorter T2 for tNAA, tCr3.0, tCr3.9, tCho, Glx, and tissue water in CSO and PCC; rs = −0.21 to −0.65, all p<0.05, FDR-corrected for multiple comparisons. These associations remained statistically significant when controlling for cortical atrophy. T2 values did not differ across the adult lifespan for mI. By region, T2 values were longer in the CSO for tNAA, tCr3.0, tCr3.9, Glx, and tissue water and longer in the PCC for tCho and mI.

Conclusion These findings underscore the importance of considering metabolite T2 changes with aging in MRS quantification. We suggest that future 3T work utilize the equations presented here to estimate age-specific T2 values instead of relying on uniform default values.

1. Introduction

Understanding brain changes across the healthy adult lifespan is critical for preserving brain health in the quickly aging global population and uncovering possible mechanisms of age-related neurological disease. Proton magnetic resonance spectroscopy (1H MRS) is the only methodology that allows non-invasive measurements of endogenous brain metabolite concentrations. However, MRS data are often acquired at echo times (TEs) that are non-negligible compared to metabolite transverse relaxation rates (T2). This results in T2-weighting of the signal, such that metabolite amplitude changes associated with normal aging might be caused by changes in relaxation but misinterpreted as changes in metabolite concentration.

This confound exists even for short-TE MRS, but is particularly a concern for J-difference-edited MRS1, which relies on TEs of >65 ms (constrained by the duration of frequency-selective editing pulses and the J-evolution of target metabolites). Thus, recent consensus1–3 suggests that it is critical to address the confound of T2 relaxation (including for reference signals), particularly in studies of aging and neurodegeneration. However, despite this, most MRS quantification procedures (likely incorrectly) use static reference values, which assume that metabolite T2 relaxation remains constant across the adult lifespan.

Prior work has reported varied relationships between age and T2, primarily for the singlet resonances total N-acetyl aspartate (tNAA), creatine (tCr), and choline (tCho). A majority of prior work at 3 and 4 T has reported shorter metabolite T2s with older age both for metabolites4–7 and tissue water4,7. A few studies8,9 at 1.5 T have reported the opposite effect of longer metabolite T2s with older age; however, one of these studies9 was complicated by overlap among water and metabolite signals, and the other8 examined only the frontal lobe and included only males in the sample. One study10, also at 1.5 T, found longer NAA T2 in the centrum semiovale of older adults; however, this study utilized a linewidth-based approach which has not been validated or used again since publication in 2005. With the exception of work by Brooks and colleagues8, each of these prior studies involved comparison of discrete age groups (young versus older adults) rather than continuous sampling across the adult lifespan, and each used small sample sizes (all n < 20 per age group, with the exception of work by Deelchand and colleagues4 which included 32 young and 26 older adults). Therefore, in the present study, we leveraged a large, multi-site cohort in order to more comprehensively investigate whether metabolite and tissue water T2 values differ across the normal adult lifespan, and to provide statistical models for calculating age-specific T2 values for future integration into MRS quantification procedures.

2. Methods

2.1 Participants

101 healthy adults provided written informed consent to participate at one of two sites: the Johns Hopkins University School of Medicine (n = 51) and the University of Florida (n = 50). The sample included approximately 10 females and 10 males from each of the following decades: 18–29, 30–39, 40–49, 50–59, and 60+ years (Table 1). The Johns Hopkins University and University of Florida Institutional Review Boards approved all study procedures.

View this table:
Table 1. Participant Demographics

Participants first completed the Montreal Cognitive Assessment (MoCA)11, followed by a 1-hour MRI protocol. Of note, four individuals scored below the cut-off score of 23 out of 30 (i.e., indicative of possible mild cognitive impairment12). However, MoCA score was not an a priori exclusion criterion for this study. Moreover, each of these individuals scored 22 (just below the cut-off), and 3 of these 4 reported that English was not their primary language which can negatively impact MoCA performance13 (and it was not feasible to conduct the MoCA in a language other than English). Therefore, we presumed that cognitive impairment was not likely and opted to retain these individuals in the cohort and statistical analyses.

2.2 MRS Acquisition

All scans were performed using a 32-channel head coil on either the Johns Hopkins University 3 T Philips dStream Ingenia Elition MRI scanner or the University of Florida 3 T Philips MR7700 MRI scanner. For voxel positioning, we first collected a T1-weighted structural MRI scan using the following parameters: MPRAGE, TR/TE 2000 ms/2 ms, flip angle 8°, slice thickness 1.0 mm, 150 slices, voxel size 1 mm3 isotropic, total time 2 min 46 sec. Next, we acquired TE series data from two 30 x 26 x 26 mm3 voxels: the white matter (WM) rich centrum semiovale (CSO) and the gray matter (GM) rich posterior cingulate cortex (PCC; Figure 1).

Figure 1.
Figure 1. Voxel Placement.

TE series data were acquired from the centrum semiovale (CSO) and posterior cingulate cortex (PCC). Each participant’s native space binary voxel mask for their CSO and PCC voxels was normalized to standard (MNI) space and overlaid onto the spm152 template. Warmer colors indicate areas of greater overlap between participants (color bar = number of subjects overlapped).

Scan parameters for the TE series included: PRESS localization, TR 2000 ms, 8 logarithmically-spaced TEs 30, 50, 74, 101, 135, 179, 241, and 350 ms, 24 transients per TE sampled at 2000 Hz with 1024 points, and CHESS water suppression (115 Hz bandwidth). Within each series, the TE steps were neither interleaved nor randomized. We also collected a separate series of unsuppressed water reference data at each of the 8 TEs with the same parameters, but with 2 transients per TE and no water suppression. Of note, these voxel sizes and locations were selected to match those collected in our recent cohort of short-TE PRESS metabolite data in 102 individuals ranging from their 20s to their 60s14.

2.2 MRS Data Processing

MRS data were analyzed within MATLAB R2021b using the open-source analysis toolbox Osprey (v2.5.0; https://github.com/schorschinho/osprey/)15. All analysis procedures followed consensus-recommended guidelines1,3. Briefly, analysis steps included: loading the vendor-native raw data (which had already been coil-combined, eddy-current-corrected, and averaged on the scanner at the time of data collection), removing the residual water signal using a Hankel singular value decomposition (HSVD) filter16, and modeling the metabolite peaks at each TE separately as described previously15,17 using TE-specific custom basis sets. The basis sets were simulated by the MRSCloud tool18 (https://braingps.mricloud.org/mrs-cloud).

MRSCloud using a localized 2D density-matrix simulation of a 101 x 101 spatial grid (voxel size 30 x 30 x 30 mm3; field of view 45 x 45 x 45 mm3) and vendor-specific refocusing pulse shape, duration, and sequence timings based on the MATLAB simulation toolbox FID-A19. The basis sets consisted of 18 basis functions: ascorbate (Asc), aspartate (Asp), creatine (Cr), negative creatine methylene (-CrCH2), gamma-aminobutyric acid (GABA), glycerophosphocholine (GPC), glutathione (GSH), glutamine (Gln), glutamate (Glu), lactate (Lac), myo-inositol (mI), N-acetyl aspartate (NAA), N-acetyl aspartyl glutamate (NAAG), phosphocholine (PCh), phosphocreatine (PCr), phosphoethanolamine (PE), scyllo-inositol (sI), and taurine (Tau), as well as 5 macromolecule signals (MM09, MM12, MM14, MM17, MM20) and 3 lipid signals (Lip09, Lip13, Lip20) included as parameterized Gaussian functions17.

We extracted amplitudes for 6 metabolites of interest: tNAA, tCho, tCr3.0 (Cr + PCr), tCr3.9 (Cr + PCr - (-CrCH2)), mI, and Glx (Glu + Gln). We multiplied each metabolite amplitude by Osprey’s internal MRSCont.fit.scale factor for each TE and participant to make the metabolite amplitudes directly comparable across TEs. This scaling factor is applied to the data to ensure an optimal dynamic range between the data and basis set during modeling. It is defined as the ratio of the maximum of the real part of the data and the basis set in the model range. Next, we used lsqcurvefit in MATLAB to fit monoexponential T2 decay functions to the TE series metabolite amplitudes (Equation 1) in order to obtain the T2 decay constant and an R2 value of model fit for each participant for each metabolite. Embedded Image

In Equation 1, yi represents the metabolite amplitude, Ai is a scaling constant, and T2i is the relaxation time to be calculated for the ith subject. Lastly, we created a binary mask of the two MRS voxels in subject space, co-registered these masks to each participant’s T1-weighted structural scan, and segmented the structural scans using SPM1220, in order to calculate the volume fractions of white matter (fWM), gray matter (fGM), and cerebrospinal fluid (fCSF) in each participant’s voxels in subject space (for use in statistical models to control for cortical atrophy and for estimation of tissue water T2).

We then repeated a similar procedure for tissue water. To estimate the water amplitudes, the unsuppressed water data at each TE was modeled using a linear combination model with a simulated water signal18. We used MATLAB’s lsqcurvefit to fit a biexponential decay function (Equation 2) to obtain tissue water T2 and R2 values for each participant. Embedded Image

In Equation 2, yi represents the water amplitude, Ai is a scaling constant, T2wi,tissue is the tissue water relaxation time, and T2wi,CSF is the CSF water relaxation time to be calculated for the ith subject. fCSF is the fraction of CSF within the voxel for the ith subject; 0.4211 weights the first term by the approximate molal concentration of water for non-CSF tissue (40/(40+55)), and 0.5789 weights the second term by the approximate molal concentration of water for CSF (55/(40+55)). The Ai and T2wi,tissue terms were unconstrained, and the T2wi,CSF term was constrained to the range of 50–3000 ms21–23. A data acquisition error occurred at the University of Florida site for the water data; therefore, we included only the Johns Hopkins University participants (n = 51) in statistical analyses of the water T2 data.

2.3 Statistical Analyses

We conducted all statistical analyses using R 4.3.224 within RStudio25. First, we calculated descriptive statistics (mean, standard deviation) by age group for the T2 values for each of the 6 metabolites of interest and tissue water. Next, we examined the correlation between T2 and age for each metabolite and voxel separately. As multiple variables did not satisfy the Pearson correlation normality assumption (Shapiro test p < 0.05), we instead report nonparametric Spearman correlations. To account for multiple comparisons, we applied the Benjamini-Hochberg false discovery rate (FDR) correction to the p-values for each voxel26.

Secondly, we ran a series of linear models, setting each metabolite T2 as the outcome variable and age as the predictor: T2 = β0 + β1*(Age-30). We centered age around 30 years, so that the intercept (β0) from this model would represent the predicted metabolite T2 value at 30 years old, and the slope (β1) would represent the change in T2 for each year of age. The aim of this model was to provide an equation to calculate predicted T2 value for a given metabolite given the age of a participant. As a follow-up analysis, we reran each of these linear models controlling for the potential effects of cortical atrophy with aging: T2 = β0 + β1*(Age-30) + β2*Tissue. As in our recent work examining metabolite T1 changes with aging27, we calculated cortical atrophy as the relative tissue fraction within the voxel, fGM / (fWM + fGM). The purpose of this follow-up model was to ensure that cortical atrophy effects were not a major contributing factor to the observed T2 relationships with age.

In addition, we conducted a series of paired t-tests (followed by FDR correction of the p-values26) to examine differences in metabolite T2 values between the CSO and PCC voxels. We also computed one linear mixed effects model per metabolite in which we set T2 (across both the CSO and PCC voxels) as the outcome variable, age, voxel, and the interaction of age with voxel as the predictors, and a random intercept (ui) for each subject: T2 = β0 + β1*(Age-30) + β2*Voxel + β3*(Age-30)*Voxel + ui. The primary aim of this model was to test for any Age*Voxel interaction effects (i.e., whether the age slope differed by brain region in any cases). The linear mixed effects model and random subject intercepts were necessary because this modeling approach structured the data as ‘repeated measures’ in which each participant had two measurements (CSO T2 and PCC T2).

3. Results

3.1 Data Quality

Creatine (Cr) linewidths were well within the range of consensus-recommended standards (i.e., < 13 Hz for 3 T3) for all spectra except one individual‘s CSO voxel (33-year-old male, Cr linewidth = 14.1 Hz). In addition, for one participant (19-year-old male), the PCC voxel was mistakenly positioned at the wrong location. Thus, these datasets (1 CSO and 1 PCC) were excluded before any statistical analyses. Additional consensus-recommended data quality metrics are presented in Appendix A. Example single-subject spectra at each TE and decay functions for each metabolite are presented in Figure 2. The mean R2 value across the whole cohort for the goodness of fit of the T2 decay model was ≥0.80 for each of the 6 metabolites of interest and tissue water. Table 2 presents descriptive statistics by age group for the T2 values.

Figure 2.
Figure 2. Single-Subject Example Spectra and CSO T2 Decay.

A PRESS spectra for each of the 8 TEs for the CSO (left) and PCC (right) voxels for a representative subject (48-year-old female). This representative subject was determined by averaging the R2 values of model fit across the 6 metabolites and tissue water for each person in the Johns Hopkins cohort (as water data were unavailable for University of Florida subjects) and then taking the group median of this average R2 value. B Example T2 decay plots for the 6 metabolites and tissue water from the CSO voxel for the same representative subject. Blue points represent the metabolite amplitude at each TE, and the dark blue line represents the calculated T2 decay curve.

View this table:
Table 2. T2 Descriptive Statistics by Age Group

3.2 T2 Relationships with Age

Older age was significantly correlated with shorter T2 values for tNAA, tCr3.0, tCr3.9, tCho, Glx, and tissue water in both the CSO and PCC; Spearman r = −0.21 to −0.65, p < 0.05, FDR-corrected for multiple comparisons (Figure 3; Table 3). Age was most strongly correlated with tNAA T2. Age did not correlate with mI T2 for either voxel.

Figure 3.
Figure 3. Metabolite T2 Correlations with Age.

T2 correlations with age are shown for the CSO (above, blue) and PCC (below, orange). Each point represents one participant. The bottom right corner of each plot indicates the Spearman correlation coefficient (rs) and statistical significance of the FDR-corrected p-value for the correlation, *p<0.05, **p<0.01, ***p<0.001. The gray line and shading represent the linear model and 95% confidence interval for the model using age to predict T2 (produced using the geom_smooth function in R).

View this table:
Table 3. T2 Correlations with Age

Next, we fit linear models to predict T2 using age for each metabolite and voxel combination: T2 = β0 + β1*(Age-30); Table 4. As we centered age around 30 years old, the intercept (β0) represents the predicted T2 value for each metabolite at 30 years of age (as opposed to age 0 which would represent an unhelpful extrapolation). The slope (β1) represents the change in the predicted value of T2 for each year of life. For example, for tNAA in the CSO, the predicted T2 value for an individual age 30 years would be: T2 = 288.27 + (30-30)*-0.81 = 288.27 ms, while the predicted T2 value for an individual age 50 years would be T2 = 288.27 + (50-30)*-0.81 = 272.07 ms. (Note that these predicted T2 values also correspond to the gray linear model lines plotted in Figure 3). The slope and intercept values listed in Table 4 can thus be utilized to calculate a predicted T2 for any age in a WM- or GM-rich voxel. Table 4 only includes slopes for the metabolites which were significantly correlated with age.

View this table:
Table 4. Model Coefficients for Estimating T2

Older age was significantly correlated with greater cortical atrophy (calculated as fGM / (fWM + fGM)) in the PCC (rs = −0.25; p = 0.010) but not the CSO (rs = −0.04; p = 0.680). As a follow-up to the linear models presented in Table 4, we reran each model controlling for cortical atrophy with aging: T2 = β0 + β1*(Age-30) + β2*Tissue (see Supplementary Table B1). Including this metric of cortical atrophy in the model did not change the statistical significance of any T2 relationships with age, with the exception of tCr3.9 in the PCC (for which the age-T2 relationship became non-significant, p = 0.094). Independent of the associations between age and T2, greater cortical atrophy was significantly associated with longer metabolite T2 values for tNAA, tCr3.0, and tCr3.9 (CSO only), as well as mI and tissue water (CSO and PCC).

3.2 T2 Differences by Voxel

Paired t-tests revealed differences in T2 values by voxel for all metabolites (as shown in Figure 4). T2 values were higher in the CSO than in the PCC for tNAA, tCr3.0, tCr3.9, Glx, and tissue water, and lower in the CSO for tCho and mI.

Figure 4.
Figure 4. Metabolite T2 Differences by Voxel.

T2 differences by voxel are shown for the CSO (blue) and PCC (orange). Each point represents one participant. The asterisks indicate the statistical significance of the FDR-corrected p-value for the paired t-test, *p<0.05, **p<0.01, ***p<0.001.

As a follow-up analysis, we computed one linear mixed effects model per metabolite (across both the CSO and PCC) to test whether the age slope differed by brain region for any metabolites or tissue water. The Age*Voxel interaction was significant only for tissue water (p=0.007), indicating that, in all but one case, the relationship of age with T2 did not differ based on brain region (Supplementary Table B2). However, for tissue water, the relationship of age with T2 was stronger for the PCC than the CSO.

4. Discussion

Here we present the largest analysis to date examining metabolite and tissue water T2 changes across the healthy adult lifespan. Among 101 adults ages 18-75 and across two sites, metabolite and tissue water T2 values in both the CSO and PCC were generally significantly shortened with age, even when controlling for age-related cortical atrophy. Moreover, T2 values were longer in the CSO, with the exception of tCho and mI which exhibited longer T2 in the PCC. Taken together, these results align with the majority of prior work which also reported T2 declines with normal aging (but in much smaller cohorts). Moreover, the finding of T2 differences based on age and brain region highlights the importance of measuring subject-level T2 during data acquisition or employing estimation methods (such as the statistical models provided here) for calculating age- and region-appropriate T2 values.

Older age was correlated with shorter metabolite T2 values for tNAA, tCr3.0, tCr3.9, tCho, Glx, and tissue water in both the CSO and PCC. This aligns with most prior research in smaller samples which similarly found shorter tNAA, tCr, tCho, and tissue water T2s with older age4–7. As also seen in these prior studies, we identified the strongest age association for tNAA. Most prior reports did not examine Glx. Deelchand and colleagues (2020) reported reduced mI T2 in CSO and PCC in older age, whereas we did not find an association between mI T2 and age. However, it should be noted that Deelchand and colleagues4 compared two age groups (rather than treating age as a continuous variable), and their older cohort (ages 70-83 years) extended beyond our upper age range.

The specific mechanisms underlying these observed T2 changes with aging remain unclear. With the exception of tCr3.9 in the PCC, each of the identified T2 associations with age remained statistically significant when controlling for cortical atrophy, suggesting that age-related atrophy is not a major factor in these findings. Instead, as metabolites are largely intracellular (glial or neuronal)28, their T2 relaxation times are likely influenced by changes in the cellular microenvironment4, i.e. cellular morphology, metabolism, or myelination4,7. It is well established that neurons undergo morphological changes during aging—such as reduction in soma size and loss or regression of dendrites and dendritic spines29— alongside a parallel metabolic shift in astrocytes associated with increased neuroinflammatory response and changes in oxidative metabolism30,31. Furthermore, degeneration of myelin sheath32 and loss of axonal fiber32 with advancing age, accompanied by debris (e.g., protein aggregates) and degraded myelin accumulation33,34 reported in white matter and further supported by in vivo diffusion tensor imaging35,36.

The observed T2 changes could also be influenced by the gradual deposition of iron, particularly Fe3+, in the brain with aging. Although iron is present in the brain in multiple forms, the intracellular non-heme iron (i.e., ferritin) in tissue is thought to cause dephasing of the proton spins and thus a faster T2 decay37. Several studies observed a strong linear correlation between iron concentrations and transverse relaxation (R2)38 values both in vivo and in post-mortem healthy and Alzheimer’s disease brain tissue37, suggesting that faster T2 relaxation is related to age-related iron deposition39. Whilst the precise contribution of each of these mechanisms is unclear, the observed age relationships suggest that T2 measurements are sensitive to various parallel changes in the cellular environment7.

In this dataset, T2 relaxation times were predominantly longer for tNAA, tCr3.0, tCr3.9, Glx, and tissue water in the WM-rich CSO, whereas tCho and mI exhibited longer T2 in the GM-rich PCC. There is relatively limited literature considering GM/WM differences in metabolite T2, and this prior work differs in voxel location, cohort, acquisition and quantification methodology, and statistical approach. Of the seven references we identified6,40–45, five reported longer T2 for NAA in WM as we did6,40–42,45, while one revealed no significant tissue effect43 and one showed the reverse effect44. For tCr3.0, three references found our result of longer T2 in WM6,41,42, three showed no difference40,43,45, and the same study showed the reverse effect44. Only a few studies have measured T2s of tCr3.9 or Glx. For tCr3.9, one paper showed longer T2s in WM45 (as we found) and one no difference42; for Glx, one study that separated Glx as Glu and Gln with J-PRESS found no differences by tissue type in the major component, Glu42. For tCho, four studies found no difference40,41,43,45, two found longer T2 in GM42,44 (as we did), and one longer T2 in WM6. T2 of mI remains less investigated, but the two studies that measured mI T2 also found longer T2 in GM42,44. For tissue water, GM is generally found to have longer T2 values than WM in multi-echo MRI experiments46–48, although the extent to which CSF confounds this result depends on resolution. Regional T2 differences may relate to greater micro- and macro-structural organization in myelinated WM compared to GM49; however, further work is needed to fully understand the mechanisms that govern metabolite T2 relaxation.

We recently performed a meta-regression analysis of 75 manuscripts50 containing 629 unique values to derive a general predictive T2 model, with linear factors for: metabolite, field strength, species, tissue, pulse sequence, and Carr-Purcell Meiboom-Gill filter. The average bias between the (30-year-old intercept) values reported in the present study and the model predicted values was +9 ms (i.e., on average the model predicts shorter T2s than measured here). The average absolute difference was 23 ms which is smaller than the average absolute difference between the predicted model and the T2 training dataset (42 ms). On this basis, we assert that our results are consistent with the diverse T2 literature.

2D modeling of interrelated MRS data has recently gained interest in the MRS community51–53. Most notably, it was found that 2D modeling of synthetic multi-TE MRS data with overlapping peaks led to improved precision due to improved model parsimony achieved through reparametrization51. Applying 2D modeling to our in vivo datasets may improve the T2 estimation of the metabolites reported here and could potentially allow for the T2 estimation of additional low-SNR metabolites. However, it will also require careful reparameterization of the T2 relaxation constants, lineshape estimates, and baseline terms, which will be part of future studies.

There are several limitations to this work. First, we acknowledge that our T2 measurement here is a complex mix of pure T2 and some inhomogeneous broadening factors that are not fully refocused by the two PRESS spin echoes. The goal of the present work was to improve the accuracy of T2 relaxation correction in quantification procedures by understanding age effects on our measure of T2 (rather than to accurately measure pure T2). Second, future work could expand upon the age range to include those younger than 18 and older than 70 years, as well as targeting both normal and pathological aging (e.g., Alzheimer’s and other neurodegenerative diseases). Given the potential of T2 to reflect both micro- and macrostructural organization, the measure may show utility as an early indicator of these changes, as suggested by Kirov and Tal54. Though this was a large cohort with systematic recruitment across the adult lifespan, we only enrolled a few individuals older than age 70 years (the timeframe at which aging effects drastically accelerate). Lastly, we were limited to collecting only two voxels (WM-rich CSO and GM-rich PCC); however, prior evidence suggests that neurochemical changes with aging are highly region-dependent55, and therefore future work might consider probing T2 changes in other brain regions, or across the entire brain.

5. Conclusions

Consistent with prior literature, in a large multi-site cohort sampled systematically across the adult lifespan, we identified a clear age-related decrease in T2 for multiple metabolites and tissue water, as well as differences in T2 between the WM-rich CSO and GM-rich PCC. Together, these findings highlight potential changes in the brain’s cellular microenvironment with normal aging and underscore the critical importance of considering metabolite T2 differences across the adult lifespan in MRS quantification procedures. We suggest that future MRS work leverage the models presented here to estimate age- and region-specific T2 values instead of relying on uniform default values.

Competing Interests

All authors declare that they have no competing interests.

Author Contributions

KH processed all data, conducted all statistical analyses, prepared all figures and supplemental material, and prepared the manuscript. SM contributed to protocol development, manuscript writing and led all revisions of the manuscript. HZ contributed to MRS data processing, and developed Osprey code for the analysis. YS and EC made significant contributions to data collection. CDJ generated the spectra figure and contributed to interpretation of results. AG, DS, and GS contributed to interpretation of results and drafted parts of the Discussion. VY reviewed all structural scans to assess data quality and check for incidental findings. SH set up the scan protocol and oversaw data quality control. GO, EP, and RAE designed the project and led interpretation of the results. All authors participated in revision of the manuscript.

Funding

This work was supported by grants from the National Institute on Aging (K00 AG068440 to KH, R00 AG062230 to GO, and K99 AG080084 to HZ) and grants from the National Institute of Biomedical Imaging and Bioengineering (R21 EB033516 to GO, R01 EB023963 to RE, R01 EB016089 to RE, and P41 EB031771).

Acknowledgements

The authors also wish to thank all of the participants who volunteered their time, as well as support staff at both MRI centers, without whom this project would not have been possible.

References

  1. 1.↵
    Choi IY, Andronesi OC, Barker P, et al. Spectral editing in 1H magnetic resonance spectroscopy: Experts’ consensus recommendations. NMR Biomed. 2021;34(5):e4411.
    OpenUrlCrossRefPubMed
  2. 2.
    Near J, Harris AD, Juchem C, et al. Preprocessing, analysis and quantification in single-voxel magnetic resonance spectroscopy: experts’ consensus recommendations. NMR Biomed. 2021;34(5). doi:10.1002/nbm.4257
    OpenUrlCrossRefPubMed
  3. 3.↵
    Wilson M, Andronesi O, Barker PB, et al. Methodological consensus on clinical proton MRS of the brain: Review and recommendations. Magn Reson Med. 2019;82(2):527–550.
    OpenUrlCrossRef
  4. 4.↵
    Deelchand DK, McCarten JR, Hemmy LS, Auerbach EJ, Eberly LE, Marjańska M. Changes in the intracellular microenvironment in the aging human brain. Neurobiol Aging. 2020;95:168–175.
    OpenUrlCrossRef
  5. 5.
    Jiru F, Skoch A, Wagnerova D, et al. The age dependence of T 2 relaxation times of N-acetyl aspartate, creatine and choline in the human brain at 3 and 4T. NMR Biomed. 2016;29(3):284–292. doi:10.1002/nbm.3456
    OpenUrlCrossRef
  6. 6.
    Kirov II, Fleysher L, Fleysher R, Patil V, Liu S, Gonen O. Age dependence of regional proton metabolites T 2 relaxation times in the human brain at 3 T. Magn Reson Med. 2008;60(4):790–795. doi:10.1002/mrm.21715
    OpenUrlCrossRefPubMed
  7. 7.↵
    Marjańska M, Emir UE, Deelchand DK, Terpstra M. Faster metabolite 1H transverse relaxation in the elder human brain. PloS One. 2013;8(10):e77572.
    OpenUrlCrossRef
  8. 8.↵
    Brooks JCW, Roberts N, Kemp GJ, Gosney MA, Lye M, Whitehouse GH. A Proton Magnetic Resonance Spectroscopy Study of Age-related Changes in Frontal Lobe Metabolite Concentrations. Cereb Cortex. 2001;11(7):598–605. doi:10.1093/cercor/11.7.598
    OpenUrlCrossRefPubMedWeb of Science
  9. 9.↵
    Christiansen P, Toft P, Larsson HBW, Stubgaard M, Henriksen O. The concentration of N-acetyl aspartate, creatine + phosphocreatine, and choline in different parts of the brain in adulthood and senium. Magn Reson Imaging. 1993;11(6):799–806. doi:10.1016/0730-725X(93)90197-L
    OpenUrlCrossRefPubMedWeb of Science
  10. 10.↵
    Kreis R, Slotboom J, Hofmann L, Boesch C. Integrated data acquisition and processing to determine metabolite contents, relaxation times, and macromolecule baseline in single examinations of individual subjects. Magn Reson Med. 2005;54(4):761–768. doi:10.1002/mrm.20673
    OpenUrlCrossRefPubMed
  11. 11.↵
    Nasreddine ZS, Phillips NA, Bédirian V, et al. The Montreal Cognitive Assessment, MoCA: A Brief Screening Tool For Mild Cognitive Impairment. J Am Geriatr Soc. 2005;53(4):695–699. doi:10.1111/j.1532-5415.2005.53221.x
    OpenUrlCrossRefPubMedWeb of Science
  12. 12.↵
    Carson N, Leach L, Murphy KJ. A re-examination of Montreal Cognitive Assessment (MoCA) cutoff scores. Int J Geriatr Psychiatry. 2018;33(2):379–388. doi:10.1002/gps.4756
    OpenUrlCrossRefPubMed
  13. 13.↵
    Ng KP, Chiew HJ, Lim L, Rosa-Neto P, Kandiah N, Gauthier S. The influence of language and culture on cognitive assessment tools in the diagnosis of early cognitive impairment and dementia. Expert Rev Neurother. 2018;18(11):859–869. doi:10.1080/14737175.2018.1532792
    OpenUrlCrossRef
  14. 14.↵
    Gong T, Hui SCN, Zöllner HJ, et al. Neurometabolic timecourse of healthy aging. NeuroImage. 2022;264:119740. doi:10.1016/j.neuroimage.2022.119740
    OpenUrlCrossRef
  15. 15.↵
    Oeltzschner G, Zöllner HJ, Hui SC, et al. Osprey: Open-source processing, reconstruction & estimation of magnetic resonance spectroscopy data. J Neurosci Methods. 2020;343:108827.
  16. 16.↵
    Barkhuijsen H, De Beer R, Van Ormondt D. Improved algorithm for noniterative time-domain model fitting to exponentially damped magnetic resonance signals. J Magn Reson 1969. 1987;73(3):553–557.
    OpenUrlCrossRefWeb of Science
  17. 17.↵
    Zöllner HJ, Považan M, Hui SC, Tapper S, Edden RA, Oeltzschner G. Comparison of different linear-combination modeling algorithms for short-TE proton spectra. NMR Biomed. 2021;34(4):e4482.
    OpenUrlCrossRef
  18. 18.↵
    Hui SCN, Saleh MG, Zöllner HJ, et al. MRSCLOUD: A cloud-based MRS tool for basis set simulation. Magn Reson Med. 2022;88(5):1994–2004. doi:10.1002/mrm.29370
    OpenUrlCrossRef
  19. 19.↵
    Simpson R, Devenyi GA, Jezzard P, Hennessy TJ, Near J. Advanced processing and simulation of MRS data using the FID appliance (FID-A)—An open source, MATLAB - based toolkit. Magn Reson Med. 2017;77(1):23–33. doi:10.1002/mrm.26091
    OpenUrlCrossRefPubMed
  20. 20.↵
    Ashburner J, Barnes G, Chen CC, et al. SPM12 manual. Wellcome Trust Cent Neuroimaging Lond UK. 2014;2464:4.
    OpenUrl
  21. 21.↵
    Piechnik SK, Evans J, Bary LH, Wise RG, Jezzard P. Functional changes in CSF volume estimated using measurement of water T 2 relaxation. Magn Reson Med. 2009;61(3):579–586. doi:10.1002/mrm.21897
    OpenUrlCrossRefPubMedWeb of Science
  22. 22.
    Daoust A, Dodd S, Nair G, et al. Transverse relaxation of cerebrospinal fluid depends on glucose concentration. Magn Reson Imaging. 2017;44:72–81. doi:10.1016/j.mri.2017.08.001
    OpenUrlCrossRef
  23. 23.↵
    Spijkerman JM, Petersen ET, Hendrikse J, Luijten P, Zwanenburg JJM. T 2 mapping of cerebrospinal fluid: 3 T versus 7 T. Magn Reson Mater Phys Biol Med. 2018;31(3):415–424. doi:10.1007/s10334-017-0659-3
    OpenUrlCrossRef
  24. 24.↵
    R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing; 2021. https://www.R-project.org/
  25. 25.↵
    RStudio Team. RStudio: Integrated Development Environment for R. RStudio, PBC; 2021. http://www.rstudio.com/
  26. 26.↵
    Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Ser B Methodol. 1995;57(1):289–300. doi:10.1111/j.2517-6161.1995.tb02031.x
    OpenUrlCrossRefPubMed
  27. 27.↵
    Murali-Manohar S, Gudmundson AT, Hupfeld KE, et al. Metabolite T 1 Relaxation Times Differ across the Adult Lifespan. Neuroscience; 2023. doi:10.1101/2023.01.06.522927
    OpenUrlAbstract/FREE Full Text
  28. 28.↵
    Ronen I, Valette J. Diffusion-Weighted Magnetic Resonance Spectroscopy. In: Harris RK, Wasylishen RL, eds. eMagRes. John Wiley & Sons, Ltd; 2015:733–750. doi:10.1002/9780470034590.emrstm1471
    OpenUrlCrossRef
  29. 29.↵
    Dickstein DL, Kabaso D, Rocher AB, Luebke JI, Wearne SL, Hof PR. Changes in the structural complexity of the aged brain. Aging Cell. 2007;6(3):275–284. doi:10.1111/j.1474-9726.2007.00289.x
    OpenUrlCrossRefPubMedWeb of Science
  30. 30.↵
    Verkerke M, Hol EM, Middeldorp J. Physiological and Pathological Ageing of Astrocytes in the Human Brain. Neurochem Res. 2021;46(10):2662–2675. doi:10.1007/s11064-021-03256-7
    OpenUrlCrossRef
  31. 31.↵
    Jiang T, Cadenas E. Astrocytic metabolic and inflammatory changes as a function of age. Aging Cell. 2014;13(6):1059–1067. doi:10.1111/acel.12268
    OpenUrlCrossRefPubMed
  32. 32.↵
    Peters A. The effects of normal aging on myelin and nerve fibers: A review. J Neurocytol. 2002;31(8/9):581–593. doi:10.1023/A:1025731309829
    OpenUrlCrossRefPubMedWeb of Science
  33. 33.↵
    Safaiyan S, Kannaiyan N, Snaidero N, et al. Age-related myelin degradation burdens the clearance function of microglia during aging. Nat Neurosci. 2016;19(8):995–998. doi:10.1038/nn.4325
    OpenUrlCrossRefPubMed
  34. 34.↵
    Knopman DS, Parisi JE, Salviati A, et al. Neuropathology of Cognitively Normal Elderly. J Neuropathol Exp Neurol. 2003;62(11):1087–1095. doi:10.1093/jnen/62.11.1087
    OpenUrlCrossRefPubMed
  35. 35.↵
    Chad JA, Pasternak O, Salat DH, Chen JJ. Re-examining age-related differences in white matter microstructure with free-water corrected diffusion tensor imaging. Neurobiol Aging. 2018;71:161–170. doi:10.1016/j.neurobiolaging.2018.07.018
    OpenUrlCrossRef
  36. 36.↵
    Hupfeld KE, Geraghty JM, McGregor HR, Hass CJ, Pasternak O, Seidler RD. Differential Relationships Between Brain Structure and Dual Task Walking in Young and Older Adults. Front Aging Neurosci. 2022;14:809281. doi:10.3389/fnagi.2022.809281
    OpenUrlCrossRef
  37. 37.↵
    House MJ, St. Pierre TG, Kowdley KV, et al. Correlation of proton transverse relaxation rates (R 2) with iron concentrations in postmortem brain tissue from alzheimer’s disease patients. Magn Reson Med. 2007;57(1):172–180. doi:10.1002/mrm.21118
    OpenUrlCrossRefPubMedWeb of Science
  38. 38.↵
    Mitsumori F, Watanabe H, Takaya N. Estimation of brain iron concentration in vivo using a linear relationship between regional iron and apparent transverse relaxation rate of the tissue water at 4.7T. Magn Reson Med. 2009;62(5):1326–1330. doi:10.1002/mrm.22097
    OpenUrlCrossRefPubMed
  39. 39.↵
    Schenker C, Meier D, Wichmann W, Boesiger P, Valavanis A. Age distribution and iron dependency of the T2 relaxation time in the globus pallidus and putamen. Neuroradiology. 1993;35(2):119–124. doi:10.1007/BF00593967
    OpenUrlCrossRefPubMedWeb of Science
  40. 40.↵
    Tsai S, Posse S, Lin Y, et al. Fast mapping of the T 2 relaxation time of cerebral metabolites using proton echo-planar spectroscopic imaging (PEPSI). Magn Reson Med. 2007;57(5):859–865. doi:10.1002/mrm.21225
    OpenUrlCrossRefPubMed
  41. 41.↵
    Träber F, Block W, Lamerichs R, Gieseke J, Schild HH. 1 H metabolite relaxation times at 3.0 tesla: Measurements of T1 and T2 values in normal brain and determination of regional differences in transverse relaxation. J Magn Reson Imaging. 2004;19(5):537–545. doi:10.1002/jmri.20053
    OpenUrlCrossRefPubMedWeb of Science
  42. 42.↵
    Wyss PO, Bianchini C, Scheidegger M, et al. In vivo estimation of transverse relaxation time constant (T 2) of 17 human brain metabolites at 3T: T 2 of 17 Human Brain Metabolites at 3T. Magn Reson Med. 2018;80(2):452–461. doi:10.1002/mrm.27067
    OpenUrlCrossRef
  43. 43.↵
    Hetherington HP, Mason GF, Pan JW, et al. Evaluation of cerebral gray and white matter metabolite differences by spectroscopic imaging at 4.1T. Magn Reson Med. 1994;32(5):565–571. doi:10.1002/mrm.1910320504
    OpenUrlCrossRefPubMedWeb of Science
  44. 44.↵
    An L, Li S, Shen J. Simultaneous determination of metabolite concentrations, T 1 and T 2 relaxation times. Magn Reson Med. 2017;78(6):2072–2081. doi:10.1002/mrm.26612
    OpenUrlCrossRef
  45. 45.↵
    Mlynárik V, Gruber S, Moser E. Proton T 1 and T 2 relaxation times of human brain metabolites at 3 Tesla: METABOLITE T 1 AND T 2 IN HUMAN BRAIN AT 3 T. NMR Biomed. 2001;14(5):325–331. doi:10.1002/nbm.713
    OpenUrlCrossRefPubMedWeb of Science
  46. 46.↵
    Deoni SCL, Peters TM, Rutt BK. High-resolution T 1 and T 2 mapping of the brain in a clinically acceptable time with DESPOT1 and DESPOT2. Magn Reson Med. 2005;53(1):237–241. doi:10.1002/mrm.20314
    OpenUrlCrossRefPubMedWeb of Science
  47. 47.
    Schmitt P, Griswold MA, Jakob PM, et al. Inversion recovery TrueFISP: Quantification of T 1, T 2, and spin density. Magn Reson Med. 2004;51(4):661–667. doi:10.1002/mrm.20058
    OpenUrlCrossRefPubMedWeb of Science
  48. 48.↵
    Fujita S, Hagiwara A, Hori M, et al. Three-dimensional high-resolution simultaneous quantitative mapping of the whole brain with 3D-QALAS: An accuracy and repeatability study. Magn Reson Imaging. 2019;63:235–243. doi:10.1016/j.mri.2019.08.031
    OpenUrlCrossRefPubMed
  49. 49.↵
    Knight MJ, McCann B, Tsivos D, Couthard E, Kauppinen RA. Quantitative T1 and T2 MRI signal characteristics in the human brain: different patterns of MR contrasts in normal ageing. Magn Reson Mater Phys Biol Med. 2016;29(6):833–842. doi:10.1007/s10334-016-0573-0
    OpenUrlCrossRef
  50. 50.↵
    Gudmundson AT, Koo A, Virovka A, et al. Meta-analysis and open-source database for in vivo brain Magnetic Resonance spectroscopy in health and disease. Anal Biochem. 2023;676:115227. doi:10.1016/j.ab.2023.115227
    OpenUrlCrossRef
  51. 51.↵
    Tal A. The future is 2D: SPECTRAL-TEMPORAL fitting of dynamic MRS data provides exponential gains in precision over conventional approaches. Magn Reson Med. 2023;89(2):499–507. doi:10.1002/mrm.29456
    OpenUrlCrossRef
  52. 52.
    Zöllner HJ, Davies-Jenkins C, Simicic D, Tal A, Sulam J, Oeltzschner G. Simultaneous multi-transient linear-combination modeling of MRS data improves uncertainty estimation. Published online November 4, 2023. doi:10.1101/2023.11.01.565164
    OpenUrlAbstract/FREE Full Text
  53. 53.↵
    Clarke WT, Ligneul C, Cottaar M, Ip IB, Jbabdi S. Universal dynamic fitting of magnetic resonance spectroscopy. Magn Reson Med. 2024;91(6):2229–2246. doi:10.1002/mrm.30001
    OpenUrlCrossRef
  54. 54.↵
    Kirov II, Tal A. Potential clinical impact of multiparametric quantitative MR spectroscopy in neurological disorders: A review and analysis. Magn Reson Med. 2020;83(1):22–44. doi:10.1002/mrm.27912
    OpenUrlCrossRef
  55. 55.↵
    Eylers VV, Maudsley AA, Bronzlik P, Dellani PR, Lanfermann H, Ding XQ. Detection of Normal Aging Effects on Human Brain Metabolite Concentrations and Microstructure with Whole-Brain MR Spectroscopic Imaging and Quantitative MR Imaging. Am J Neuroradiol. 2016;37(3):447–454. doi:10.3174/ajnr.A4557
    OpenUrlAbstract/FREE Full Text
Back to top
Posted June 24, 2024.
Download PDF

Supplementary Material

Email
Metabolite T2 relaxation times decrease across the adult lifespan in a large multi-site cohort
Kathleen E. Hupfeld, Saipavitra Murali-Manohar, Helge J. Zöllner, Yulu Song, Christopher W. Davies-Jenkins, Aaron T. Gudmundson, Dunja Simičić, Gizeaddis Simegn, Emily E. Carter, Steve C. N. Hui, Vivek Yedavalli, Georg Oeltzschner, Eric C. Porges, Richard A. E. Edden
bioRxiv 2024.06.19.599719; doi: https://doi.org/10.1101/2024.06.19.599719
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area