Dissociation between iron accumulation and ferritin upregulation in the aged substantia nigra: attenuation by dietary restriction

Study 1: Ex vivo characterization of the normally aging mouse brain

Study 2: Dietary restriction (DR)-induced attenuation of age-related brain changes in the SN

DISCUSSION

This is the first study to show DR-induced amelioration of iron dyshomeostasis in the SN, a region known to be particularly vulnerable to disease, intimating a potential pivotal role for SN in development of the aging phenotype. Our results highlight iron chelation therapies as potential putative “anti-brain aging” interventions, which may also attenuate the risk of developing neurodegenerative diseases, particularly Parkinson’s disease.

In the STR, we observed a continual increase in R2* with aging, consistent with human studies [22,38]. We show increased R2* was correlated with increased iron, at least up to 19 months of age. Maximum iron content was observed at 19 months in mice, consistent with similar observations in man [6] at corresponding ages [39]. Whilst human studies showed significant correlations of R2* to histochemical-detected iron [6], we show R2* correlations with direct semi-quantitative iron measurements, as well as within the same spatial region of the same sample.

In the GP, rapid increases in R2* [22] and iron [6] during early life have been shown in humans, followed by a plateau at 30 - 40 years of age. No initial rapid increase in R2* or iron in the GP was observed in our study. However, Haacke et al., (2010) [38] did report similar findings to those reported here in humans of a comparable age range to mice in our study, i.e., with the youngest being young adults as opposed to children as in the Aquino et al., (2009) study [22]. Whilst full correspondence to human investigations cannot necessarily be expected, these results intimate a need for developmental aging to be distinguished from senescence, at least when considering changes in iron and relaxation rates.

Our finding of increased R2* in the STR and GP between 19 - 27 months without concomitant increased accumulation of iron is noteworthy, although in the latter, there was a trend towards increased iron albeit significance was not reached. The increased R2* in the STR, alongside unchanged levels of iron, suggest iron is partitioning into different forms with aging. Iron is often associated with ferritin but may also bind to hemosiderin and magnetite. Hemosiderin-iron is an aggregate that induces greater spin dephasing [40] and so has a greater effect on T2* (and thus R2*) than ferritin-iron, which acts mostly through T2 relaxation [40]. Iron associated with biogenic magnetite also has a higher magnetic moment than that stored in ferritin and better at enhancing T2 relaxation [41]. A shift in the proportion of iron to these more magnetically active forms in the STR could thus account for R2* increases, without concomitant increases in iron. One study has reported increased hemosiderin in the aged (>65 years of age) putamen, a sub-region of the STR [42].

Aquino et al., (2009) showed age-related increases in SN R2* over the lifespan (1 – 80 years) [22], similar to that reported here. However, one post-mortem study showed the majority of iron accumulation occurred very early in life [43] and another, a more linear increase [44]. The apparently contradictory findings between these two human studies most likely arise from differences in sample age ranges, again highlighting the need to distinguish developmental from biological aging. The trajectory of iron accumulation in our study is consistent with that of Zecca et al., (2004) [44], who used a similar sample age range.

Both R1 and R2 values in the basal ganglia exhibited an overall increase with age. Whilst both measures significantly correlated to iron as in human studies [45], we showed a reduced correlation between R1 and iron as previously reported [46,47]. Interestingly, for R1 and R2 particular, nonlinear regression analysis showed these relaxation times tended to plateau at higher iron concentrations, suggesting there is a saturation point for the influence of iron on these relaxation rates: this saturation effect was not readily observed for R2*. It was noted R2 remained a reliable predictor for changes in brain iron during senescence, with a greater influence of other tissue factors on R2 occurring up to 13 months of age (see below).

R2 values were similar between 6 - 13 months in the GP and SN, with a significant decrease observed in the STR, despite iron accumulation, which would increase R2. Astrogliosis has been shown to increase T2 (decrease R2) [48] potentially by enlarging the extracellular space and increasing interstitial fluid content [49–51]. Indeed, astrocytes are more reactive with aging [52] and the demonstrable increased astrogliosis shown in the STR here likely accounts for decreases in R2 observed between 6 and 13 months, without a measurable decrease in STR volume. The decreased R2 in the STR may also result from neuronal loss, although unlikely, given the lack of atrophy over 6 – 19 months. Alterations in cellular content and tissue structure clearly play a prevalent role on age-related change in R2, at least up to 13 months. This may potentially mask the influence of iron accumulation on R2 to this point and may also explain why studies with narrower age ranges have reported no age-related alterations in T2 [53,54]. Of note, after 13 months, R2 changes in all the basal ganglia regions mirrors R2*, demonstrating R2 is as suitable a proxy for iron content as R2* in advanced aging.

Overall, the smallest effect of age was seen on R1 relaxation rate, which we have shown to have the least, albeit significant, correlation to iron. The overall pattern of R1 change across the mouse lifespan reflects the increase in iron, but without the specificity of R2* or even R2.

DR has previously been demonstrated as a robust “anti-brain aging” intervention [27]. We have highlighted the SN as a potentially crucial region of deleterious iron accumulation in aging, especially due to its high neuronal vulnerability [11]. In this context, our most pertinent finding is that DR ameliorates in vivo age-related R2 increases in the SN alongside inhibited iron-induced up-regulation of ferritin. Other studies have also shown DR-amelioration of increases in T2 with aging [16], but not in the serial manner here or with corroborating ferritin IHC measurements. The demonstrable strength of ex vivo R2 to predict elevations in iron (particularly during senescence) in the current investigation lends credence that the amelioration of age-related increases in in vivo R2 following 12 months pertain to DR-attenuation of iron accumulation.

Under normal homeostatic control, increased cellular iron induces expression of ferritin to increase iron-storage capacity [55]. Indeed, increased iron in the GP throughout aging, alongside increased ferritin-positive cells shown here, has been documented previously [56]. In the STR, higher ferritin-immunopositive cell numbers preceded iron accumulation over 2 – 6 months, but did not continue to increase over 6 – 19 months, despite increased iron, suggesting that ferritin is only partially loaded with iron in the STR of younger brains. Ferritin can store up to 4500 atoms of iron and in peripheral tissues at least, ferritin remains only partially saturated with iron [57]. Increased ferritin-immunopositive cell numbers at 27 months compared to 6 suggest that existing ferritin levels are insufficient to accommodate the increasing iron levels. Studies in rodents and man have also shown increased ferritin in the STR with aging [55,56], with one rat study showing a similar pattern of iron and ferritin change in the STR as we have reported here [55].

Unexpectedly, the increased iron levels in the SN over 6 and 19 months were accompanied by a decrease, rather than an increase, in ferritin-immunopositive cells, suggesting iron dysregulation. The impaired iron handling/storage with aging in the SN may predispose this region to oxidative damage and subsequent neurodegeneration, and contribute to the select neuronal vulnerability of this region [11]. Rather than being sequestered in ferritin, the excess iron may be sequestered in neuromelanin (NM), given the prevalence of NM-containing dopaminergic neurons in the SN [58]. Indeed, NM is known to associate with iron and shown to increase with aging [43,44]. Furthermore, Zecca et al., (2004) reported impaired ferritin expression in aged SN neurons in man alongside significantly large amounts of NM-bound iron. NM may protect neurons against toxic quinone and semi-quinone, by-products of dopamine metabolism [59], with the SN having the highest dopaminergic neuronal content. However, at high iron concentrations, NM becomes saturated, leading to a potentially pro-oxidant environment [60,61] and indeed high levels of NM, have been associated with SN neurodegeneration in middle-aged mice [62]. Whether excess iron is indeed sequestered by NM in aging and if so, whether the iron level reported here is sufficient to saturate NM in the SN, warrant future investigation.

Iron mishandling in the SN may have prevalent effects throughout the brain as SN dopaminergic projections modulate striatal neurons [63], affecting sensorimotor learning and memory [64]. Further, decreases in nigral neurons with aging [65] may precipitate impaired dopaminergic innervation and consequent behavioral abnormalities [66]. Indeed, studies have shown D₁ and D₂ dopamine receptor losses in the STR with aging [67], as well as a reduced ligand binding [68]. Others have shown that the caudal and lateral STR, receiving afferent input from the A9 SN neurons, exhibit decreases in dopamine content [69]. In light of this, iron dyshomeostasis in the SN demonstrated here, may play a prevalent and crucial role in stimulating oxidative damage of the melanized dopaminergic neurons, contributing to their vulnerability and explaining the extrapyramidal symptoms of the aging brain.

In PD, death of dopaminergic neurons in the SN pars compacta decreases dopamine levels in the prefrontal cortex and STR [70], leading to cognitive and motor dysfunction [71]. A feature of PD is the greater iron accumulation in the SN than in aging [5,72] and similar to the current investigation, iron-induced ferritin expression is also inhibited [73]. Additionally, increased redox activity in NM aggregates of melanized neurons has been observed in PD, with NM-positive staining shown to positively correlate with ferric iron content [61]. The potential role of NM in PD pathophysiology poses an interesting question regarding its role in iron sequestration with respect to normal aging, given the similarity in iron alterations of our study to that reported previously in PD [73].

We have shown age-related increases in GFAP-positive astrocytes in the basal ganglia regions, consistent with previous studies [74,75]. The expanded endfeet of astrocytes encompass the brain capillary endothelial cells (BCECs) [76] and facilitate the transport of various metabolites and metals, including iron, into the neuropil [76]. Astrocytes not only take up iron as transferrin-bound iron, they can also import iron via the divalent metal transporter 1 (DMT1), which are expressed at astrocytic end feet [77]. We show that the time-course of the increases in GFAP-immunopositive cells during aging was similar to that for iron accumulation in the STR and the SN, and may arise from the known uptake and storage of iron in astrocytes [75]. Indeed, one third to three quarters of the iron in the brain is stored in glial cells [78]. The increases in iron in the STR, possibly from increased astrocytic uptake and storage, is associated with a corresponding increase in ferritin-immunopositive cells, suggesting the increased iron is stored with ferritin. However, in the SN, the increases in both astrocytes and iron levels are not concomitant with an increase in ferritin-positive cells. As mentioned before, the excess iron does not appear to be bound by ferritin and may become dependent on NM for non-toxic storage.

In the GP, however, astrocyte levels were comparable except between 2 and 27 months, despite increased iron during this time. Iron can still be accumulating in astrocytes in the GP, without a corresponding increase in astrocytes numbers, as the astrocytes numbers in the GP at 2 months are relatively high compared to those in the STR, and appear to be sufficient to accommodate the increase in iron.

A pro-inflammatory state has been demonstrated in aging, with increases in glial cell number [79] and an activated, dystrophic microglial phenotype previously reported. Stereotactic injections of lipopolysaccharide in the STR have been shown to increase microglial activation, ferritin expression and total nigral iron content in aged rats [80], with activated microglia shown to increase iron uptake and decrease export [81]. Thus, the microgliosis observed with aging in the three basal ganglia regions may, at least in part, explain the iron accumulation in these areas. Further, microgliosis in the STR was less pronounced and correlated with lower levels of iron accumulation. Increased microglial iron uptake may induce pro-inflammatory cytokine release and neurodegeneration [82,83]. Activated microglia produce superoxide, which is capable of stimulating iron release from ferritin [84], leading to increased labile iron, enhancing pro-oxidative capacity that is central to neuroinflammatory and neurodegenerative processes in aging. Indeed, Hayashi et al., (2008), highlighted microglia as a major source of oxidation products in the aged mouse brain [85].

The decreased ferritin storage potential of iron in the SN reported in the current investigation alongside astrogliosis, combined with the tendency for aged microglia to become dystrophic and render iron more redox-active [84,86], cements the SN as a particularly vulnerable region in aging. The interplay between iron accumulation and glial cell populations requires furthe exploration to elucidate the full extent to which this region contributes to the aged phenotype.

Whether microgliosis precedes and induces the increases in iron uptake and/or whether iron accumulation occurs initially and contributes to microglial activation/dystrophy remains to be seen. In regard to the latter, the role of microglia in long term storage of iron has previously been noted, with storage of iron potentially increasing microglial dystrophy [86]. Reduced iron accumulation, as suggested by T2, has been associated with improved motor performance [16]. Iron may have a critical role in the age-related increase and activation of microglia in aging. In light of this, the demonstrable reduction of microglia in the SN by DR, combined with apparent amelioration of iron dyshomeostasis, highlights that restoration of iron homeostasis could be a key facet of the beneficial effects of DR [35]. Determining if a causal relationship exists between the apparent restoration of iron homeostasis and reduced microglia, or if the reverse is true, could help prove the efficacy of iron chelation therapies or anti-inflammatory treatments, respectively. However, combining the two therapeutic strategies may prove to be synergistic. Indeed, interactions of inflammatory signaling molecules and iron in PD have been discussed in Pretorius et al., (2014) [87] and suggested to enhance eryptosis in the periphery. Whether the effects of DR in the SN shown here can incite beneficial changes in other regions known to be affected during aging, such as the STR that receives significant dopaminergic innervation from the SN, remains to be seen.

In addition, the lack of an astrocytic reaction, concomitant with iron dyshomestasis and microgliosis certainly supports the idea that microglia rather than astrocytes, plays a more dominant role in the handling of iron in the SN. The lack of effect of DR on astrocytes in the SN is interesting given previous reports of DR reducing age-related increases GFAP, albeit not specifically in the SN [88].

The integration of in vivo and in vitro iron measures alongside ferritin and glial immunohistochemistry in the aged mouse brain provides formative insight into the regulation of brain iron in aging, but also the importance of iron dyshomeostasis and neuroinflammation on brain aging and as possible targets for novel interventions. Further, we reveal the particular vulnerability of the SN to iron dyshomeostasis in aging, potentially mediating/resulting from age-related inflammation and contributing to the susceptibility of this brain area to neurodegenerative diseases, especially PD. In this context, DR, and in the future DR analogues, may serve as putative interventions against such deleterious aging processes.

Ethical statement

All experimental procedures performed on mice were in compliance with the local ethical review panel of King’s College London and the U.K. Home Office Animals Scientific Procedures Act 1986.

Animals and Treatment

Study 1:Ex vivo characterization of the aging mouse brain

Male C57BL/6J mouse brains were obtained from Shared Aging Research Models, (ShARM, Sheffield, UK). Mice were culled at 2 (n = 11), 6 (n = 8), 13 (n = 11), 19 (n = 11) and 27 months (n = 8) of age by rising CO₂ inhalation (different numbers of mice were included at each age, depending on tissue availability.). Ages of mice chosen for study were equivalent to different stages of adulthood from just post-adolescence, adult, midlife, elderly and the very elderly. The heads were removed and fixed in 4% paraformaldehyde (1 week) and stored in phosphate-buffered saline (PBS, 4°C) with 0.05% sodium azide prior to MRI, SRXRF and immunohistochemistry (see below).

Study 2: DR-induced attenuation of age-related brain changes

Animals were maintained under standard laboratory conditions (room temperature automatically maintained at 21°C ± 1°C; 12h light:dark cycle). Mouse cages were provided with wood shavings and shredded paper bedding for enrichment. Body weights and food intake data were recorded weekly.

Male C57BL/6J mice (n = 24, aged 7 months old; Harlan, UK), were singly housed and acclimatized for 7 days before allocation into two groups of 12. Following in vivo baseline MRI (see below), one group was fed ad libitum (AL), while the other, underwent an intermittent fasting regimen involving fasting on alternate days from 17:30 – 11:00 h. Mice were scanned again in vivo at 9, 12, 15 and 19 months of age (see section on in vivo MRI below). After MRI at 19 months, mice underwent transcardial perfusion with heparinized (50 IU/ml) saline to flush out blood from the vasculature before perfusion-fixation with 4% paraformaldehyde under terminal anaesthesia. Heads were decapitated and fixed for a further week in 4% paraformaldehyde before storage in 0.05% azide-supplemented PBS at 4^oC until ex vivo MRI (see below).

MR acquisition

Ex vivo MRI: Brains within the skull were prepared for MRI (see below) by removal of, the skin, muscle, lower jaw, tongue, nasal bones and zygomatic arches, and embedded in perfluoropolyether (Galden™, Performance Fluids, UK, Solvoy Solexis) to minimize susceptibility and prevent tissue dehydration during scanning. Care was taken to avoid air bubbles. MRI was performed using a quadrature volume radiofrequency coil (33 mm internal diameter; Rapid Biomedical, Rimpar, Germany) on a 7T horizontal bore MRI scanner (Agilent Technologies Inc, Walnut Creek, CA, USA).

T1 relaxometry was achieved with a 2D spin-echo (SE) sequence with repetition times (TR) of 670, 800, 1000, 1200, 2000 and 4000 ms; echo time (TE), 11 ms and 2 averages. A 2D multi-echo, SE sequence was run with TE, 11.6, 34.8, 58.0, 81.2, 104.4, 127.6, 150.8 and 174.0 ms; TR, 7600 ms and 4 averages, to measure T2. To account for stimulated echoes only data from the odd TEs were included in the T2 fitting (see below). T2* was assessed with a 2D multi-echo gradient echo sequence with five echoes and TE, 2.5, 8.5, 14.5, 20.5, 26.5 ms; TR, 2000 ms and 10 averages, flip angle 90°. Contiguous coronal slices (40, 0.5 mm thick) were recorded with field of view (FOV), 30 × 30 mm and matrix size, 256 × 256 for all relaxometry.

In vivo MRI: T2 relaxometry was performed as for brain samples at baseline (7 months), 9, 12, 15 and 19 months but with TE (10.6, 21.3, 31.9, 42.6, 53.2, 63.9, 74.5, 85.2, 95.8, 106.5, 117.1, 127.8, 138.4, 149.1, 159.7 and 170.4 ms) and TR, 5217 ms; 2 averages and 30 contiguous coronal slices at 0.5 mm thick; FOV, 20 × 20 mm and matrix size = 128 × 128.

MRI image analysis

T1, T2 and T2* maps were generated by pixel-by-pixel non-linear fitting to equations: y = A (1-exp^(-TR/T1)), y = B(exp^(-TE/T2)) and y = C(exp^(-TE/T2*)), respectively (A, B and C are constants that represents signal components that were unchanged in each experiment) in JIM 5.0 (Xinapse Systems, Alwincle, UK). R1, R2 and R2* maps (R2 only for study 2) were then calculated by performing the reciprocal function for each maps. Structural images were produced by summing the shortest 8 echoes of the T2 relaxometry data using proprietary scanner software and these images used for manual tracing of the striatum for volumetric assessment.

Regions of interest (ROIs) were manually drawn in the STR, GP and SN on the high quality structural images (guidelines listed in Supplemental Table 2, according to a standard mouse brain atlas [89]) using ImageJ (National Institute of Health, Maryland, USA), and then placed onto each of the directly registered relaxation rate maps. Note the structural images were produced from the T2 relaxometry data (see above). As all relaxometry data was collected in the same scan session without moving the brains, accurate registration between the structural image and all relaxometry maps was possible. Relaxation rates for each anatomical ROI were obtained and averaged for each age before statistical testing in Prism v5 (GraphPad, SD, USA). For volume analysis, striatal ROIs were delineated as previously described [90].

Cryoprotection and tissue sectioning

Following MRI acquisition, brains from 5 mice per group from both studies (not including 13 month old mice from study 1 as these brains were obtained sometime after synchrotron access) were cryoprotected in 30% sucrose (PBS, 0.05% sodium azide). Subsequently, cryosectioning was performed coronally to produce 40 μm thick frozen sections that were mounted onto 4 μm thick Ultralene film (Spex Sample-Prep, NJ, USA) secured to a customized holder for SRXRF. Cryosections sections of 20 μm thick were also obtained and mounted onto Superfrost plus microscope slides for IHC. Not all the brains at the various ages that had previously undergone MRI were analyzed by SRXRF as synchrotron access is very limited. Likewise, only sufficient numbers of brains at the various ages needed for statistical testing were analyzed by IHC.

SRXRF elemental mapping and analysis

SRXRF of whole/right hemisphere of brain tissue sections was performed on the I18 beamline at the Diamond Light Source synchrotron radiation facility (Didcot, UK). The beam energy was tuned to 11 keV and focused to 100 × 100 μm (determining resolution) and the brains scanned in a raster manner. The brain samples were mounted at a 45° angle with respect to the incoming X-ray beam and the detector to minimize scatter contribution. The raw data consist of full energy dispersive spectra for each sample point exposed to the beam. The spectra were subsequently fitted and the net peak areas of the characteristic K_α and K_β peaks of iron were evaluated using PyMca [91]. Quantification was performed by measuring a reference metal film (AXO, Dresden, GmbH), which allows estimation of the photon flux on the samples. SRXRF elemental maps of pixel-by-pixel iron concentrations (mg/kg wet weight) were manually aligned to the corresponding relaxation rate maps using ImageJ. Thus, ROIs, similar to that for the relaxation maps, could be placed at the equivalent spatial locations on the elemental iron maps to quantify regional iron.

Immunohistochemistry

Standard DAB IHC was performed. Endogenous peroxidase activity was blocked by incubation with 1% H₂O₂ in Tris- buffered saline, pH 7.4 containing 0.2% Triton X-100 (TBS+; 30 mins, RT). After 2 X 5 mins washes in TBS+, non-specific binding was blocked using 10% skimmed milk power (SMP) in TBS+ (2 h, RT). Sections were then incubated with either rabbit anti-ferritin antibody (F6136; Sigma-Aldrich, Poole, Dorset, UK), rabbit polyclonal anti-Iba-1 antibody (019-19741; Wako Pure Chemical Industries, Richmond, VA), rabbit polyclonal anti-GFAP antibody (DAKO, High Wycombe, UK) in 5% SMP in TBS+ (overnight, 4^°C). After washes in TBS+, slides were incubated with secondary biotinylated anti-rabbit IgG (1:200, DAKO, High Wycombe, UK) in 5% SMP in TBS+ (2 h, RT). Avidin-biotin binding was then performed using the ELITE ABC kit (Vector, Peterborough, UK) for ferritin and Iba1 immunohistochemical protocols, respectively, prior to the chromogenic reaction using the Impact DAB peroxidase kit (Vector, Peterborough, UK). Slides were subsequently dehydrated in ascending grades of ethanol, cleared in xylene and mounted using DPX (Sigma-Aldrich, St Louis, MO, USA).

Histological image analysis

Sections were scanned on a LEICA SCN400F scanner (University College London, London) to produce 20× magnification digital images from 4 sections per mouse in the STR (Bregma +1.10 to 0.50 mm), GP (Bregma -0.22 to -0.70mm) and SN (Bregma -3.08 to -3.64 mm) from the right hemisphere for each section. Depending on the region, 2 – 4 optical fields were taken, each comprising a 628.0 × 278.5 µm² area. Ferritin, GFAP or iba1 immunopositive cells in each optical field were manually counted using the cell counter plugin of ImageJ and expressed as number of immunopositive cells per unit area.

Data analysis

Spearman’s rank-order correlation and nonlinear regression analysis was performed to assess the strength of the association between R1, R2 and R2* with SRXRF measured elemental iron in all basal ganglia regions. Statistical testing of correlations was performed using cocor [92].

One way analysis of variance (ANOVA) with post-hoc Tukey’s correction (Prism v5, GraphPad, San Diego, CA, USA) was used to test for differences in individual relaxation rates, iron, ferritin, Iba1 and GFAP-immunopositive cells and normalized STR volume with aging (study 1). A two-way ANOVA with repeated measures and post-hoc Sidak’s multiple comparisons test was used for testing differences in in vivo R2 with aging and feeding regimen (study 2). Two-way ANOVA was also used to test for differences in ferritin, Iba1 and GFAP-immunopositive cell numbers between AL and DR mice in study 2.

All values quoted are mean ± standard error of mean (SEM) and significance was set at P < 0.05.