Spatio-temporal metabolic rewiring in the brain of TgF344-AD rat model of Alzheimer’s disease

In this work, we aimed to characterise the spatio-temporal metabolic profile changes during the onset of AD-like pathology in the TgF344-AD rat brain. Longitudinal MRS studies were performed in the same cohort of WT and TgF344-AD animals. Four time points, from 9 to 18 months, were evaluated to capture the onset of metabolic changes: three points before the manifestation of gross pathological changes, as described in this model²⁷; and another one later on. To account for specific regional differences in the progression of the disease, MRS data was acquired from four voxel positions, whose accurate location was evaluated based on automatic segmentation of structural MRI.

Based on a strict quality control and processing pipeline, the longitudinal metabolic profiles demonstrated progressive changes in TgF344-AD rats compared to WT, consistent with the major findings reported in clinical AD, such as decreased N-acetylaspartate and glutamate. Specifically, decreased NAA and/or tNAA in the cortex, hippocampus and thalamus, and decreased Glu in the thalamus and striatum were detected in TgF344-AD compared to WT animals, further confirming the suitability of this model for translational studies on AD progression. Additional analysis revealed strong intra-regional couplings of cortical NAA, tNAA and tCr with myo-inositol (either Ins or Ins+Gly) in WT, but with Tau in transgenics rats; whereas in the hippocampus, Tau showed strong correlations with Ins+Gly and tCho only in the WT group. Moreover, divergent inter-regional metabolic couplings in TgF344-AD and WT animals were observed. Specific crosstalks between cortex, hippocampus and striatum were found for Tau in the WT group and for myo-inositol in the TgF344-AD rats. This analysis sheds light into the complex spatio-temporal metabolic rewiring in AD, which would be linked to the pathological cascade associated to the disease^2,48. The different associations of Ins and Tau with other metabolites in WT and TgF344-AD rats, suggest a relevant role for these two metabolites during the course of the pathology. While Ins has been related to neuroinflammation^4,49 modulatory role in microglial activation has been associated to Tau⁵⁰. This fact would be in line with the AD trajectory described in⁴⁸, where it is suggested that Aβ deposition associated to ageing is followed by a protective microglial activation. Therefore, inappropiate microglial reponses leading to inflammatory response could be related with impaired coupling between Tau and Ins observed in the transgenic animals. The lack of effectivity in this early response would be followed by an increase in Aβ deposition, tau protein pathology and neurodegeneration, all of them associated with the decreased levels of NAA⁴ observed in our study, more evident at 15 and 18 months of age. Interestingly, NAA was coupled with Tau in transgenic animals but with Ins in wild-type animals, supporting this dual trajectory depending on the initial microglial response.

Importantly, the age-dependent decrease of cortical and thalamic creatine and its strong correlation with metabolites such as NAA or myo-inositol, reveals its inadequacy as an internal reference for relative quantification methods routinely used in the clinic (e.g. Ins/Cr), highlighting the need for careful interpretation of such results in the context of AD. To mention, the age-dependent evolution of creatine in hippocampus shown a similar pattern than in cortex and thalamus, although not significant differences were identified. Indeed, similar results were reported in a study with a bigger sample size including both male and female TgF344-AD, where significantly decreased levels of hippocampal tCr were observed in the transgenic group³³.

Objective quality control was used to assure the reliability of the absolute metabolite concentrations estimated from each MRS dataset, including spectral selection based on SNR and FWHM, and fitting performance based on Cramér-Rao Lower Bounds assessment. Moreover, the study groups were assessed for potential outliers (based on interquartile range), controlled physiological condition throughout the acquisition (breathing and rectal temperature), and similar percentage of the region of interest covered by the MRS voxel. The slightly lower rectal and brain-estimated temperatures noticed in the cortex and thalamus of the transgenic group are consistent with previous findings in AD patients suggesting limited thermoregulation functions which have been associated with reduced number of suprachiasmatic nucleus neurotensin neurons^51,52. Importantly, these differences, although significant, were not associated with the metabolic changes detected. It deserves to be mentioned that all the acquisitions were performed during the light period to avoid effects associated to the circadian cycle. It has been shown that diurnal variability in MRS measures is low and makes this technique reliable during this period⁵³. Together with this, it has been shown that the use of isoflurane anaesthesia disrupts the differences associated to circadian phases in rats⁵⁴ and therefore the results would not be significantly affected by the day-time of acquisition.

⁠Temporal metabolic profile changes in the TgF344-AD rat brain

Metabolic profile analysis showed different longitudinal evolutions of NAA and/or tNAA in the cortex, hippocampus and thalamus of WT and TgF344-AD rats. Specifically, these metabolites decreased during ageing only in transgenic animals. NAA is a marker of neuronal density and its decrease has been associated with neuronal loss or dysfunction⁴. Consistently, decreased NAA is one of the most reproducible findings in the cortex and hippocampus of AD patients⁴, as well as in rat^28,32,33 and mouse models¹⁵. Progressive neuronal damage has been described in the TgF344-AD model²⁷, aligned with decreased NAA also reported in the hippocampus and thalamus^32,33. Since NAA/Cr has also been negatively correlated with Aβ burden^5,20, decreased NAA at the latest time points in TgF344-AD animals could be also associated with the presence of Aβ plaques, as has been described at 16 months of age by Cohen et al.²⁷ and verified by histopathological analysis in the study cohort at 18 months of age (Supplementary Fig. 3).

A decrease of Glu was observed in the thalamus of TgF344-AD rats during ageing, not observed in WT, while, in the striatum, TgF344-AD showed stable lower levels of Glu than WT rats over the study. Glu is one of the main neurotransmitters in the brain, essential for dendrite and synapse formation. Lower Glu concentration has been associated with neuronal loss and correlated with poor cognitive outcome⁵⁵. Therefore, the observed changes are consistent with the neuronal loss described in the transgenic model²⁷. Indeed, we observed significantly lower Glu levels at 15 months of age in TgF344-AD with respect to WT animals, when clear cognitive impairment has been described²⁷. Glu neurotransmission is crucial for learning and memory function⁵⁶, and therefore, the early lower levels of striatal Glu observed in transgenic rats could be related to the early learning deficits previously described in these rats , such as impairments in long-term spatial memory described at 4 months of age³³, slower learning in a delay-non-match-to-sample test⁴⁴ and impaired reversal learning on the Morris Water Maze⁵⁷ and Barnes Maze²⁷. In mouse models of AD, decreased Glu or Glu/Cr was found in the cortex and hippocampus^13,18,58,59, correlating with Aβ plaque deposition²⁰. In human cohorts, decreased cortical levels of Glx or Glu have been reported in AD, MCI and at-risk populations^4,55, as well as non-significant decreases of Glx/Cr in the striatum and thalamus of AD patients⁶⁰.

The time-course profiles of cortical Cr and thalamic tCr also showed an age-dependent decrease in TgF344-AD animals, not observed in the WT group. In the thalamus, significant lower tCr concentration was observed in transgenic with respect to WT animals at 12, 15 and 18 months of age. These changes are coherent with the decrease in tCr observed in hippocampus recently reported in the same model³³ and might be associated with cell energetics disturbances occurring in neurodegenerative pathologies⁶¹, consistent with a neuroprotective effect of Cr against Aβ toxicity reported in the McGill-R-Thy1-APP rat model of AD⁶². Previous studies in AD or MCI patients reporting lower tCr levels in areas such as the hippocampus^63,64 support our findings, probably related with a similar metabolism of tCr in human and other mammals as described in⁶⁵, where it is suggested the relationship between PCr and glutamate neurotoxicity associated to AD as well as the role of oxidative stress (related to creatine metabolism) in neurodegenerative diseases. Indeed, oxidative stress has been described in the TgF344-AD animals, where decreased antioxidant capacity was reported in these animals⁶⁶. In the same line, ex-vivo analysis revealed lower tCr levels in the hippocampus and frontal cortex in a transgenic mouse model of AD⁶⁷. However, increased tCr levels were reported in the cortex of McGill-R-Thy1 rats²⁸ and in the hippocampus and thalamic regions of TASTPM mice²². These two models showed early amyloid pathology but none of them replicate other aspects of the disease such as neurofibrillary tangles or widespread cell death^26,68, that have been described in the TgF344-AD rats²⁷. Therefore, discrepancies may be related to pathological processes associated to neurodegeneration or tau protein pathology²⁶.

Importantly, these results warn that metabolite to tCr ratios should be analysed with caution, since tCr is not a constant reference in AD and therefore it can lead to conflicting conclusions.

Regarding myo-inositol, a metabolite commonly related to human AD⁴, the MRS literature typically quantifies it as mixed Ins+Gly levels, since the metabolic profiles of both metabolites significantly overlap. Thus, Ins and Gly were both included in our basis set for their individual quantification. Although Gly was not further evaluated after quality control assessment, it improved the estimations of Ins+Gly compared to Ins alone, as expected. Accordingly, we detected significant differences between the age-dependent trajectories of Ins+Gly levels in the thalamus and striatum of TgF344-AD and WT animals, characterized by their stronger increase with ageing in WT rats. Although myo-inositol has been considered a glial marker, with increased levels suggesting gliosis, its role in neuroinflammation or glial activation remains to be discerned^16,49. In this line, while the TgF344-AD model presents early gliosis, already detected at 6 months of age²⁷ and neuroinflammation^32,69, we and others have not detected increased Ins in this model compared to WT³², and an increase in hippocampal Ins was only observed at advanced stages in³³. Notwithstanding, striatal and thalamic median Ins+Gly levels tended to be higher in transgenic than in control animals at early stages, but lower at later ages (Fig. 3). These observations strengthen the importance of accounting for the timing of the disease when comparing literature findings, which might explain some of the discrepancies reported in human cohorts and animal models. For instance, even though Ins increases have been commonly reported in AD patients⁴, significant decreases of Ins/Cr have been also reported in the striatum, hippocampus and other regions in AD or dementia with Lewy’s body, compared to healthy controls⁶⁰. In the APP/PS1 mouse, both decreased Ins/Cr¹⁵ and increased Ins/Cr were reported in the hippocampus^13,58, while no changes were found in Ins absolute concentration in the cortex¹⁸. In the McGill-R-Thy1 rat model of AD, Ins decreased at 3 months of age in the hippocampus, while increased at 9 months of age in the hippocampus and 12 months of age in the frontal cortex²⁸.

With regards to Tau, our results showed significant age-dependent decreases in the cortex of TgF344-AD rats, not observed in the WT group. Indeed, significant lower Tau concentrations were detected in the transgenic group with respect to WT at 15 and 18 months. These results are in line with previous findings as decreased Tau in the hippocampus of this model³³, and decreased Tau/Cr ratio in AD patients⁷⁰ or aged animal models⁷¹. However, a recent study with the TgF344-AD model depicted increased cortical Tau levels at 18 months of age³². This discrepancy may be related to differences in the MRS acquisition protocol (such as its ~ 70% longer echo time, leading to mixed J-coupling and T2 saturation effects, reflected in the detection and quantification of Tau), or different voxel location in the cortex, which in our study was more frontal. In this line, it has been shown that cortical amyloid deposition starts in frontal and temporal areas, affecting medial and posterior cingulate cortex only later on³⁴. In fact, our results suggest a trend towards increased Tau levels in the cortex of transgenic animals before 12 months, but decreased levels at later time points, compared to controls. Thus, the differences could also be associated with regional specific timings of the pathological changes. Tau plays modulatory and regulatory roles in different physiological processes. It has been reported as a neuroprotector in neurodegenerative diseases such as AD and used to ameliorate several neurological disorders⁵⁰. Indeed, Tau supplementation improved cognition in the APP/PS1 mouse model of AD⁷² and enhanced adult neurogenesis under both in vitro and in vivo conditions⁷³.

Finally, cortical tCho revealed different age-dependent evolution in TgF344-AD rats compared to WT, showing (non-significant) increased level at earlier stages which decreased later on. The genotype had a significant effect in the hippocampus with increased tCho values in theTgF344-AD rats, although both groups followed a similar trajectory, in line with the results showed in this region in³³. Accordingly, tCho age-effect differences between TgF344-AD and control rats in the hypothalamus and hippocampus were also reported in³². Choline compounds are found in myelin sheets and cell membranes and variations in tCho have been related to white matter integrity and membrane turnover, although opposite findings have been reported in the cortex of AD subjects⁴. Indeed, decreased tCho/Cr was detected in the prefrontal cortex and thalamus⁶⁰, while increased levels were reported in the posterior cingulate cortex⁷⁴. Moreover, increased tCho has been reported with ageing in both humans and animal models^49,75,76, associated with demyelination, inflammation, and functional changes including cognitive decline⁷⁷. As for myo-inositol, higher cortical tCho in young transgenic animals might suggest early damage, being demyelination and inflammation processes less evident at more advanced ages. Choline is liberated during cell membrane turnover, and therefore increases are related to neuronal degeneration, and has been observed in brain tumours, demyelinating disease or after traumatic brain injury^33,78. On the other hand, decreased tCho at later time points could also be associated with neuronal damage and subsequent cognitive impairment, as described in aged TgF344AD rats²⁷; also aligned with decreased hippocampal tCho observed in a drug-induced memory deficit model⁷⁹. This strengthens again the importance of accounting for the timing of the disease in metabolic profile investigations.

Intra-regional correlations of brain metabolic profiles

We further investigated the associations between different metabolites in each brain region, from a systems biology perspective. Thus, we found different metabolic correlation patterns in transgenic and wild-type groups.

As expected, there were significant correlations between NAA and tNAA, Glu and Glx and Ins and Ins+Gly, regardless of the brain region and group; except for Ins and Ins+Gly in the thalamus. The latter could be related to the significant age effect observed in this region for Ins+Gly, but undetected in Ins. Although Gly could not be evaluated independently in our study, its neuroprotective properties in neurodegeneration and memory impairment⁸⁰ merit further investigation with more sensitive technique in the context of AD.

The strong correlations between creatine with other metabolites altered in TgF344-AD rats represent one of the main findings of this study. Although creatine has been commonly used as a reference in MRS analysis of AD, recent studies suggest the involvement of Cr in neuropathological mechanisms underlying dementia and neurodegeneration^33,63,64,65, which would encourage to a careful interpretation of results based on ratios to Cr. Our results strengthen this rationale, since tCr is shown to be correlated with metabolites associated to pathological processes described in AD, such as neurodegeneration (NAA), neuroinflammation (Ins) or glutamate toxicity (Glu, Glx)⁶⁵. In the cortex, tCr was highly correlated with NAA and tNAA in both TgF344-AD and WT groups, whereas correlations with Glu and Glx were specific to TgF344-AD, and with Ins+Gly and Ins were only found in WT animals. In the hippocampus, tCr correlated strongly with NAA in both groups; but with Ins, Ins+Gly and Glx only in TgF344-AD rats, and with Glu only in WT animals. Interestingly, in the thalamus and hippocampus, tCr only correlated with tNAA in the transgenic group. Altogether, these results suggest an important role for Cr in AD-like processes, since it was highly correlated with metabolites showing pathology specific time-course changes. This supports the idea that the commonly considered metabolic ratios to tCr, such as decreased NAA/tCr and increased Ins/tCr^4,67, might not be appropriate in the case of AD-like pathologies, given the potential confounding effect of altered tCr levels.

Moreover, the correlations between taurine and myo-inositol with other metabolites also differed between TgF344-AD and WT animals. Several studies have suggested an important role of either Tau or Ins in AD pathology^{4,15,18,28,33,58,60,70}, highlighting the need to better understand their changes and apparent discrepancies reported in the literature. Beyond concentration differences in specific regions, intra-regional metabolite correlation analysis provided new insights into this issue. Thus, cortical levels of Tau in TgF344-AD rats were highly correlated with NAA, tNAA and tCr, suggesting a link between Tau decreases and neuronal dysfunction; while in WT animals, cortical NAA, tNAA and Cr correlated with Ins+Gly and Ins. Due to the minimal sample size criteria defined, Tau concentration was not evaluated in the thalamus and, therefore, no correlations could be investigated in this case. Such metabolic correlation patterns could reflect compensatory effects between neuroinflammation (Ins)^4,49 and neurotrophic (Tau)^50,72,73 processes associated with normal ageing, potentially mitigated in transgenic animals. In TgF344-AD, impaired neurotrophic processes could be related to neuronal dysfunction, manifested by decreased NAA. This pathological mechanism would be in line with the AD trajectory proposed in⁴⁸.

Further correlations were found in the hippocampus of WT animals, between myo-inositol and NAA and tNAA and between Tau and both Ins+Gly and tCho. While the metabolic interplay between Ins, tCho and Tau remains to be unravelled, these results suggest a role for Tau in the normal ageing process, similar to Ins and tCho in inflammation and degeneration during normal ageing, as described in humans and animal models^62,63,64. Tau has a modulatory and regulatory function in physiological processes including modulation of neuroinflammation. Its neuroprotective and anti-neuroinflammatory role has been reported in neurodegenerative diseases⁵⁰. The coupling between Tau and the two metabolites related with inflammation observed in wild type animals was lost in the transgenic animals (correlation coefficients were 0.24 between Tau and Ins+Gly and 0.09 between Tau and tCho), which might reflect a dysfunction associated to the lack of the modulatory effect of Tau to cope with neuroinflammation.

Highly significant (p < 0.001) correlations between Glu (or Glx) and NAA, tNAA and/or tCr were detected in the cortex and hippocampus, as well as between Glx and NAA in the striatum only WT rats (while correlation between Glx and NAA was 0.62 in WT animals, it decreases to 0.45 in transgenic rats). In this region, TgF344-AD rats showed lower levels of Glu compared to WT, which could partially be related with this decrease in the correlation values (indeed, the correlation between Glu and NAA decreases from 0.6 in WT to 0.4 in transgenic rats; while an increase is observed in the correlation coefficient between Gln and NAA: 0.11 in WT and 0.29 in transgenic). All this could potential reflecting Gln-Glu cycle changes, associated with the neuronal loss described in this model^27,32,65.

Inter-regional correlations of brain metabolic profiles

Finally, we investigated the correlation between metabolite concentration across different brain regions. In WT animals, Tau levels were strongly correlated between the cortex, striatum and hippocampus, and with Ins+Gly in the hippocampus. In turn, TgF344-AD rats exhibited the same inter-regional metabolic crosstalk between cortex, striatum and hippocampus for Ins, instead of Tau. This could be related to the specific correlations of cortical NAA (and tNAA) with Tau found in transgenic animals, but with Ins (or Ins+Gly) in the cortex and hippocampus of WT animals. Such metabolic couplings could reflect a whole brain imbalance in Tau levels associated with region-specific onsets of neuronal dysfunction^27,34,57. Altogether, this would support the hypothesis of a balance between neuroinflammation (Ins) and neurotrophic and modulatory (Tau) processes during normal ageing impaired in TgF344-AD rats, what could be related with the hypothesis presented in⁴⁸, suggesting that differences in microglial response to the deposition of β-amyloid are related with AD progression.

Interestingly, while no significant group or interaction effects were found for GSH or Gln, highly significant correlation of striatal GSH with cortical NAA in the WT group and with Gln in transgenic animals were observed. GSH is the most prevalent anti-oxidant in the brain^49,81 and decreased levels have been reported in MCI and AD patients⁸¹ and in the APPTg2576 model of AD¹⁸. Decreased antioxidant capacity which has been previously reported in the TgF344-AD rats⁶⁶ could be related with the genotype specific correlation between GSH and other metabolites. Assuming that the GSH cycle compensates for decreased excitatory neurotransmission during Glu-Gln shuttle inhibition⁸², the correlation between Gln and GSH in TgF4344-AD rats could be related to alterations in the normal Glu-Gln cycle, consistent with the decreased Glu levels observed in this group.

Limitations

We acknowledge some limitations in our study. Firstly, the relatively small sample size. While only 18 animals were included, each one was evaluated at 4 time points, and consequently the fitting of the LME model included up to 72 samples. In addition, the rigorous criteria for LME analysis ensured enough significance and effect sizes, although some changes at specific ages might remain undetected. Since our study focused on the analysis of the longitudinal differences of brain metabolic profiles between control and transgenic animals, rather than differences at a specific age, this allowed us to increase the statistical power analysis, as well as being more sensitive to the timing of the pathologic changes. Due to the exploratory character of the study, we have not applied multiple comparison correction. Nevertheless, as shown in Supplementary tables 4 and 5, the effect size of the observed differences was large in most of the evaluated metabolites.

Some brain or cognitive alterations have been described after 18 months of age in the TgF344-AD rats. While the age-dependent pathological processes involved increased presence of Aβ plaques and stronger cognitive impairment shown at 26 months of age with respect to 18 month-old animals²⁷, other alterations such as deficits in novel object recognition²⁷and visual alterations have been reported only at 24 and 19 months of age respectively⁸³.Therefore, another potential limitation could be the chosen endpoint of the study.

The cohort evaluated in this work belongs to a wider study, where the animals underwent cognitive evaluation every 3 months by a delayed non match to sample (DNMS) task as previosuly described¹², and further research should be done to evaluate if the repetition of this task could have an impact in brain metabolism. In addition, only male animals were included in our experiments, and consequently we did not account for sex-related differences in the pathology. We acknowledge this is a limitation of our study, and female animals have been included in our ongoing investigations. To mention, a recent study including both female and male TgF344-AD reported limited effect of sex on MRS-based results in hippocampus, while a sex effect was identified in volumetric measures³³.

Finally, as a general limitation of animal models of AD, they might not mimic all the changes occurring in patients. Although our findings are mostly consistent with metabolic changes previously described in AD patients, further research should ascertain the reproducibility of the novel results reported here for the first time, such as the intra- and inter-regional correlations.