Mean amplitude of low frequency fluctuations measured by fMRI at 11.7 T in the aging brain of mouse lemur primate

Non-human primates are a critical species for the identification of key biological mechanisms in normal and pathological aging. One of these primates, the mouse lemur, has been widely studied as a model of cerebral aging or Alzheimer's disease. The amplitude of low-frequency fluctuations of blood oxygenation level-dependent (BOLD) can be measured with functional MRI. Within specific frequency bands (e.g. the 0.01–0.1 Hz), these amplitudes were proposed to indirectly reflect neuronal activity as well as glucose metabolism. Here, we first created whole brain maps of the mean amplitude of low frequency fluctuations (mALFF) in young mouse lemurs (mean ± SD: 2.1 ± 0.8 years). Then, we extracted mALFF in old lemurs (mean ± SD: 8.8 ± 1.1 years) to identify age-related changes. A high level of mALFF was detected in the temporal cortex (Brodmann area 20), somatosensory areas (Brodmann area 5), insula (Brodmann areas 13–6) and the parietal cortex (Brodmann area 7) of healthy young mouse lemurs. Aging was associated with alterations of mALFF in somatosensory areas (Brodmann area 5) and the parietal cortex (Brodmann area 7).

Human life expectancy has dramatically increased during the last century. This comes with an increased risk for cerebral alterations leading to neurodegenerative diseases or mild cognitive/motor impairments that impair daily living. Sensorimotor function impairment is one of the stereotypical characteristics of mammalian cerebral aging. It includes slower performance of visual and motor tasks 1 such as walk speed 2 , movement coordination or dexterity 3 . In addition, higher functions such as thinking, memorizing, and speech capacity decline with aging 4 . These functional deficits occur alongside considerable brain alterations that include fiber demyelination 5 and cortical thinning 6 . In non-human primates such as chimpanzees, white matter deterioration was observed, confirming prior results obtained in humans 7 . In addition, several investigations in mammalian models have shown an impact of aging on synapses and dendrites 8,9 . Monitoring functional impairments associated with cerebral ageing with translational imaging methods is critical to understand mechanisms leading to brain alterations and develop new treatments.
Measures of cerebral glucose metabolism using the radiolabeled glucose analog 18 F-fluorodeoxyglucose (FDG) detected by Positron Emission Tomography (PET) imaging is a largely used marker to investigate brain function in elderlies 10,11 and in the mouse model of Alzheimer's disease 12 . The use of PET is however restricted as this requires using radioactive compounds and is limited for animal studies by its low resolution when compared to functional magnetic resonance imaging (fMRI). In consequence, magnetic resonance imaging (MRI) is an interesting alternative to PET.
Low-frequency oscillations (LFO) of blood-oxygen level dependent (BOLD) signal reflects the total power of BOLD signal within the frequency range between 0.01 and 0.1 Hz. The amplitude of low-frequency fluctuations (ALFF) is expected to reflect neuronal activity 13 and was associated with markers of glucose metabolism 14 . Thus, it has been proposed as an MRI-based method to evaluate brain function 13,15 and could be a promising radioactive-free alternative to FDG-PET. Studies in humans have shown that ALFF is negatively correlated with age in several brain regions such as the supplementary motor area, pre-supplementary motor area (Brodmann area (BA) 6), anterior cingulate cortex (BAs 24, 32, 33), bilateral dorsal lateral prefrontal cortex (BAs 9, 46), right  13,15 . However, healthy aging effects on ALFF indexes remain to our knowledge, unexplored in non-human mammalians. The mouse lemur (Microcebus murinus) is a primate attracting increased attention in neuroscience research and particularly for aging studies. This small animal (typical length 12 cm, 60-120 g weight) has a decade-long lifespan 16 and is a model for studying cerebral ageing 17 and Alzheimer's disease 18 . This animal was used to establish the impact of prediabetes in the brain 19 as well as to evaluate interventions modulating cerebral ageing process 16 . The aim of the current study was thus to characterize ALFF in this primate in normal and ageing conditions. We described regional differences of ALFF signal and showed aged-related changes in specific brain regions.

Materials and methods
Animals and breeding. The guidelines of the European Communities Council directive (2010/63/EU) were followed when conducting this study. Our protocol was authorized by the local ethics committees CEtEA-CEA DSV IdF (authorizations 201506051736524 VI (APAFIS#778)). We originally included 33 mouse lemurs (21 males and 12 females) in our study (Table 1). They were bred in our laboratory (Molecular Imaging Research Center, CEA, Fontenay-aux-Roses) after being born at the CNRS/Brunoy, MNHN's France, laboratory breeding colony (UMR 7179 CNRS/MNHN). Four animals that displayed MR images with artefact or brain lesions were removed from the study.
The "young lemur cohort" consisted of 14 animals with an age range of 1.3 to 3.8 years old (mean ± SD: 2.1 ± 0.8 years).
The "old lemur cohort" consisted of 15 animals with an age range of 8.0 to 10.8 years old (mean ± SD: 8.8 ± 1.1 years). www.nature.com/scientificreports/ The animals were housed in cages with one or two lemurs, enrichment for jumping and hiding, temperatures between 24 and 26 degrees Celsius, a relative humidity of 55 percent, and seasonal illumination (summer: 14 h of light, 10 h of darkness; winter: 10 h of light, 14 h of darkness). Food consisted of fresh apples and a handmade blend of bananas, cereals, eggs, and milk. Water supply for animals was freely accessible. None of the animals had ever taken part in invasive research or pharmaceutical trials before.

Animal preparation and MRI acquisition.
To ensure animal stability during the experiment, all animals were scanned once while under isoflurane anesthesia at 1.25-1.5% in air, with respiratory rate monitoring. A 32 °C air heating system was used to maintain body temperature, causing mouse lemurs to go into a state of natural torpor 20 . The body temperature was maintained using a heating pad and measured using a rectal thermometer. The benefit of this is that it prevents reawakening while maintaining a low anesthetic level. Animals were scanned on an 11.7 Tesla Bruker BioSpec MRI machine (Bruker, Ettlingen, Germany) running ParaVision 6.0.1 with a volume coil for radiofrequency transmission and a quadrature surface coil for reception (Bruker, Ettlingen, Germany). We acquired anatomical images with a T2-weighted multi-slice multi-echo (MSME) sequence: TR = 5000 ms, TE = 17.5 ms, FOV = 32 × 32 mm, 75 slices of 0.2 mm thickness, 6 echoes, 5 ms IET, resolution = 200 µm isotropic, acquisition duration 10 min. We acquired resting state time series with a gradient-echo echo planar imaging (EPI) sequence: TR = 1000 ms, TE = 10.0 ms, flip angle = 90°, repetitions = 450, FOV = 30 × 20 mm, 23 slices of 0.9 mm thickness and 0.1 mm gap, resolution = 312.5 × 208.33 × 1000 µm, acquisition duration 7m30s. During the acquisition, the animals were head-fixed using ear bars to minimize head motion. The total duration of anesthesia was approximately one hour.
MRI pre-processing. Data from scanners was exported as DICOM files and then changed to NIfTI-1 format. Then, using the Python program sammba-mri (SmAll MaMmals Brain MRI), spatial preprocessing was carried out 21 , http:// sammba-mri. github. io) and we used nipype for pipelining 22 , leverages AFNI 23 for most steps and RATS 24 for "skullstripping". A study template was made using the mutual registration of anatomical MR images. Images were then registered to a high-resolution anatomical mouse lemur template, built for our previously published functional atlas 25 . Motion, B0 distortion, and slice timing (interleaved) were removed from resting state MR images (per-slice registration to respective anatomical images). Using sequential applications of the transformations from individual anatomical images to the study template and then the transformations from study template to the mouse lemur atlas, all the MR images were placed into the same space. Functional images were further pre-treated using AFNI afni_proc.py 23 . fMRI images were smoothed (0.9 mm), bandpass filtered, detrend corrected (0.01-0.1 Hz) as well as slice timing and motion corrected. TRs with excessive motion of 0.07 mm or where too many voxels were flagged as outliers by 3dToutcount (AFNI), were censored. To ensure steady-state magnetization, the first five volumes were not included in the study. Note that standardization of fMRI pre-processing remains ongoing in human fMRI as in 2012, Carp et al. found 207 different analysis pipelines in 241 studies, suggesting that almost every publication uses a unique analysis pipeline 26 . However, significant efforts have been made through standardized protocol as proposed by fMRIPrep 27 . In rodents, a significant portion (26%, without taking into account pre-processing parameters) of rodent experiments used a specially designed pre-processing pipeline 28 . This suggests that standardization in non-human fMRI is still a far-off goal likely due to the adaptation of specific pre-processing parameters to each species' physiology. mALFF calculation and extraction. LFO measures were performed using the fast Fourier transform index: amplitude of low-frequency fluctuation (ALFF) 29 . As ALFF is sensitive to the scale of raw signal and the unit of BOLD signal is arbitrary, the original ALFF value is not adapted for comparisons between animals. In addition, ALFF can be susceptible to signal fluctuations caused by physiological noise unrelated to brain activity 13 . A standardization procedure has been proposed by dividing the signal of each voxel by the global mean ALFF in each animal 30 . The newly calculated index is called mean ALFF (mALFF). mALFF indexes were calculated for each voxel of the pre-processed EPI images in the low-frequencies range 0.01 to 0.1 Hz using the function "3dLombScargle" and "3dAmpToRSFC" from AFNI 23 . The mALFF signal of each voxel was extracted within the different regions based on the anatomical atlas 31 using NiftiLabelsMasker from Nilearn 32 .
Statistical analysis. Voxel wise analysis was performed using 3dttest++ from AFNI 23 and a clustering approach (-Clusterize) for multiple comparisons. This method is routinely used as statistical correction technique for multiple comparisons in fMRI research 33 . This method is based on the fact that fMRI voxels are not entirely independent and propose clusters of voxels to be evaluated for significance rather than each one separately. The outcome of this analysis (3dttest++) associates a cluster size to an uncorrected significant threshold value in a table form. In our study, we took the highest p-value (0.05) that was associated with a cluster size of 856 voxels. Analysis space was reduced by thresholding the average mALFF map to the 20% highest voxels. Then, using "map_threshold" from nilearn 32 , we extracted from the z-map, any cluster of statistical voxels superior to 856 associated to p < 0.05.

Results
mALFF in young mouse lemurs. mALFF maps were recorded from 14 young mouse lemurs. Individual mALFF maps of each animal were averaged to produce 3D maps of the group (Fig. 1). Automatic extractions of the mALFF signal were then performed in various cerebral regions by using a reference anatomical atlas ( 31 ; Fig. 1A). The highest cortical signal was observed in the temporal cortex (Brodmann area (BA) 20 i.e. secondary visual area, expected to be involved in visual processing and recognition memory), parietal cortex (parietal regions as BA 5 i.e. covering the superior parietal lobule (and a portion of the postcentral gyrus) involved in www.nature.com/scientificreports/ the representation of spatial information of limb movement 34 . High signal was also detected in the primary somatosensory cortex (BA 1-3) and the primary motor cortex (BA 4). More integrative areas such as the insula (BA [13][14][15][16] and in the cingulate cortex (BA 23 and 24) also displayed high signal. In subcortical areas, the basal forebrain exhibited the most elevated signal when compared to the rest of the brain. Conversely, cortical regions such as BA 30 (agranular retrolimbic area involved in vision) and the adjacent BA 27 (area presubicularis involved in vision) or BA 25 (antero-ventral part of the cingulate cortex) displayed the lowest levels of mALFF. Subcortical regions such as substantia nigra, subthalamic nucleus or pituitary gland also displayed the lowest levels of mALFF.
Age-related changes of mALFF. mALFF measures were performed in 15 mouse lemurs (old cohort) and compared to the previously studied young lemurs. Average mALFF maps are displayed in Fig. 2A for young mouse lemurs and in Fig. 2B for old mouse lemurs. Comparison of average mALFF maps between young and old lemurs indicates lower mALFF in the parietal cortex of aged animals and a trend towards higher signal in the www.nature.com/scientificreports/ hypothalamic regions. Voxelwise analysis between groups revealed a significant loss of BOLD signal amplitude in parietal regions involving (BA 5 and BA 7) of old animals (Fig. 2C, D). We used the same analysis to evaluate differences between males and females. No significant cluster was detected. In the absence of a significant effect of gender on mALFF index, the variable "gender" was not considered as a cofactor.

Discussion
Cerebral distribution of mALFF index in mouse lemurs. This study evaluated mALFF index in mouse lemur primates at a high field (11.7 T). The evaluation of mALFF in a young cohort provided 3D maps of the normal distribution of the mALFF indexes. Highest levels of mALFF were detected in cortical structures involved in high-order visual processing (BA 20 in temporal region), somatosensory (BA 5 in parietal region), and in integrative regions involved in visuo-motor coordination (BA 7 in parietal region). High signal was also detected in the primary somatosensory cortex (BA 1-3) and the primary motor cortex (BA 4). More integrative areas such as the insula (BA 13-16) and the cingulate cortex (BA 23 and 24) also displayed a high signal.
Age-related changes of mALFF index. As a second part of the study, we assessed age-related changes of mALFF. In humans, healthy ageing mainly affects ALFF in brain regions involved in motor function such as the supplementary motor area or the pre-supplementary motor area. More integrative regions such as the anterior cingulate cortex, the dorso-lateral prefrontal cortex, the posterior cingulate cortex, and the inferior parietal lobule are also impaired 35 . In mouse lemurs, the only regions significantly affected by healthy ageing were those involved in visuo-somatosensory-motor function (parietal regions as BA 5 i.e. covering the superior parietal lobule (and a portion of the postcentral gyrus) involved in the representation of spatial information of limb movement 34 and BA 7 i.e. an integrative area involved in visuo-motor coordination). These two regions are involved in somatosensory processing, in movement such as grasping 36 or object location 37 . These alterations could participate in the described visuo-somatosensory-motor alterations reported in aged lemurs 38 . Atrophy detected in mouse lemurs impacts the whole brain in the latest stages of ageing but the most prominent atrophied areas include the insular (BAs 13-16), frontal (BA 6), parietal (BAs 5, 7), occipital (BAs 17, 18), inferior temporal (BAs 21, 28) and cingulate cortices (BAs 23,24,25) 17,31 . In consequence, an ageing effect can be detected with mALFF and anatomical measures in both BAs 5 and 7 but BAs such as 17, 18, 13-16 are impacted www.nature.com/scientificreports/ only by atrophy and not by mALFF changes. A better understanding of the significance of mALFF is now needed to further interpret mALFF alterations during aging.
In previous studies, ALFF or fractional ALFF were used to evaluate the effect of simian immunodeficiency virus 39 , spinal injury 40 or anesthesia 41 in macaques. In mice, Huntington-related pathological effects were detected using ALFF 42 . In rats, various effects of stress could also be evaluated using ALFF 43,44 . By showing age-related changes, the current study increases the range of ALFF changes detected in mammals in pathological situations.
Our results cannot be easily be interpreted within the context of the current literature. Indeed, to our knowledge no study evaluated the impact of aging on mALFF in other non-human primate species. In addition, our results do partially fit with the results obtained in humans 35 showing mostly an effect of aging on fractional ALFF in prefrontal regions and in the posterior cingulate cortex (areas particularly involved in the default mode network).
Limitations. Anesthesia is known to interact with brain function. For example, Isoflurane changes functional connections between brain regions in marmosets 45 . In addition, Wu et al. showed that across functionally related but different S1 subregions, isoflurane elicits comparable dose-dependent suppressive effects on the power of rsfMRI signals and local fine-scale functional connectivity 41 . Thus, one of the possible limitations of the study is anesthesia, as it could impact the mALFF signal and change the distribution of mALFF in the brain of healthy young mouse lemurs. Future studies will thus have to be conducted to investigate the impact of anesthesia on mALFF.
Comparisons between young and old animals revealed local age-related differences of mALFF. As both groups were anesthetized, this difference is expected to reflect the ageing effect. We can however not exclude that the suppressive effect of anesthesia may have decreased our ability to detect mALFF changes in some regions affected by ageing.

Conclusion
In conclusion, this study provides evidences suggesting that mALFF can be measured in the whole brain of mouse lemurs and can detect aged-related changes. It highlights mALFF as a tool for the exploration of the cerebral function in mammals as well as an interesting candidate for the longitudinal follow up of age-related cerebral dysfunction in animal models.

Data availability
Raw MRI data in mouse lemurs are available upon request following a formal data sharing agreement required by authors' institution. The template and atlas used in this study maps are available for download in NIfTI-1 format at https:// www. nitrc. org/ proje cts/ mouse lemur atlas.

Code availability
All the software used to perform the analysis are publicly available. In particular, sammba-mri, the code developed by our team for spatial pre-processing is available at http:// sammba-mri. github. io. All the third party codes used in the article are cited.