High iron in mouse olfactory ensheathing cells at the glia limitans in the olfactory bulb underlies MRI T₂* hypointensity

The glia limitans covers the entire surface of the brain and spinal cord on the external face, toward the pia mater (glia limitans superficialis), and internally around brain blood vessels (glia limitans perivascularis) (Mason et al., 2021). The olfactory nervous system glia limitans is unique as it consists of olfactory ensheathing cells (OECs) and astrocytes (Beiersdorfer et al., 2020; Nazareth et al., 2019). In contrast, the glia limitans in the rest of the nervous system consists only of astrocytes. Rather than being myelinated, OECs at the olfactory nerve layer (ONL) wrap around bundles of olfactory nerve axons to support neural transmission. Olfactory sensory neurons (OSNs) at the nasal epithelium have a special ability to continuously regenerate from progenitor cells throughout life (Beecher et al., 2018; Saglam et al., 2021) and OECs are important to support this regeneration. OECs can participate in innate immune responses and release several signaling molecules, such as TNF and IL-1β, to recruit macrophages which may be important for function as part of the olfactory glia limitans or for degrading dying axons that are replaced throughout life (Jiang et al., 2022). Thus, OECs play a variety of roles in maintaining a permissive environment for regeneration and modulating inflammation and neurodegeneration, making them a unique type of glia cell (Denaro et al., 2022). In previous work studying microbleeds and immune cell trafficking into the mouse olfactory bulb during viral infection (Liu et al., 2022), we noticed a large decrease in signal in T₂*-weighted MRI in the ONL of control mice.

T₂*-weighted MRI (gradient-recalled echo or susceptibility-weighted imaging) is useful for the in-vivo characterization of brain iron in development, neuroinflammatory diseases, brain bleeding, and aging (Bulk et al., 2020; Duyn & Schenck, 2017; Griffin et al., 2019; Liu et al., 2022; Tisdall et al., 2022; Ward et al., 2014). Iron plays an important role in many metabolic processes, including oxygen transport, DNA synthesis, mitochondrial respiration, myelin synthesis, and neurotransmitter synthesis and metabolism (Ward et al., 2014). Iron levels are low in the brain at birth (Drayer et al., 1986), then increase rapidly between youth and middle age. Then, iron increases slowly with advancing age in healthy individuals (Bartzokis et al., 2007; Dusek et al., 2022). Iron varies around the human brain with highest levels found in the substantia nigra (Lee & Lee, 2019). Oligodendrocytes are a major location of non-heme iron in the brain likely due to the large metabolic demands for myelination (Cheli et al., 2020; Tisdall et al., 2022). Iron-laden activated macrophages and microglia have been reported in the human brain with amyotrophic lateral sclerosis (ALS) and multiple sclerosis (MS) (Kwan et al., 2012; Nair et al., 2020). Microglia ferroptosis is reported to contribute to neurodegeneration (Ryan et al., 2023). Indeed, increased brain iron is associated with many neurological disorders. In Parkinson’s Disease (PD) patients, several studies have reported decreased T₂* intensity in the substantia nigra (Cho et al., 2011; Schwarz et al., 2018). In Alzheimer’s Disease (AD), a T₂* MRI study of post-mortem human brains and a mouse model detected increased iron concentrations which accompany amyloid β (Aβ) aggregation in the hippocampus (Jack et al., 2005; Nabuurs et al., 2011). The redox chemistry of iron has been shown to play a role in free radical formation in tissues and thus may play a role in neuroinflammation, such as in substantia nigra neurons during neuron loss (Faucheux et al., 2003), the production of Aβ (Sayre et al., 2000), and the activation of macrophages and microglia (Kwan et al., 2012; Ward et al., 2014). Understanding the links between iron accumulation and neuroinflammation would be helpful for the management of these disorders.

Here, T₂*-weighted MRI detected a large decrease in signal at the inner nerve fiber layer (iNFL) of the ONL in normal mouse from 3 to 12 weeks consistent with iron accumulation in this region. Iron histochemistry, immunohistology, and electron microscopy (EM) demonstrated that there was iron accumulation in OECs that underlies the T₂*-weighted MRI signal decrease in this region. Based on MRI T₂* relaxivity, the iron concentration in the iNFL was estimated to be 2–3 mM. Based upon electron microscopy (EM) of individual OECs, the ferritin concentration in the cytosol of OECs was estimated to be 2–12 µM and iron concentration was estimated to be 1.6–12 mM. In both cases, these represent high levels of iron and ferritin. Interestingly in rats, MRI did not detect evidence of iron accumulation in the ONL of the olfactory bulb until about 11 months of age. In mice and older rats, the high iron levels were associated with microglia activation and accumulation of lipofuscin within microglia. Lipofuscin is a known marker of brain inflammation usually associated with degeneration (Moreno-Garcia et al., 2018). It is not clear whether the iron accumulation detected is required for function in mouse OECs or is a result of OEC function in mouse, in particular phagocytosis of dying axons as they are replaced by olfactory nerve regeneration. The fact that iron accumulates to high levels in this region at a young age should enable studying the connections between iron and inflammation.

All mice and rats in this study were handled in accordance with the Institute of Laboratory Research guidelines and were approved by the Animal Care and Use Committee of the National Institute of Neurological Disorders and Stroke. Adult male (59) and female (25) C57BL/6J (B6) mice were purchased from The Jackson Laboratory. Adult male (20) and female (12) Sprague-Dawley rats were purchased from Harlan Laboratories. A flow chart with the animal numbers used in each experiment carried out in this study is shown in the Supplementary Chart 1.

MRI experiments were carried out on an 11.7T animal MRI [Magnex magnet, Resonance Research Incorporated 12 cm ID gradients and Bruker electronics (Billerica, MA)]. A four element Bruker receive only CryoProbe was used for detection. T₂*-weighted 3D gradient-recalled echo sequences were used for acquisitions.

For whole-mouse brain in-vivo imaging, the following parameters were used: isotropic resolution = 75 µm, TE/TR = 10/30 ms, Flip Angle (FA) = 10°, Field of View (FOV) = 19.2 x 14.4 x 9.6 mm, NA = 2, and scan time = 24 min.

For whole-brain ex-vivo MRI, mice or rats were perfused transcardially with 5% formalin. The heads were post-fixed with 10% formalin overnight. 24 h before ex-vivo MRI acquisition, heads were transferred to PBS buffer. During acquisition, mice heads were submerged in fomblin or PBS in a 15 mL falcon tube. Rat heads were wrapped closely with plastic wrapping to avoid water evaporation, because of the size limitation of CryoProbe. For all ages of mice and rats, the parameters for ex-vivo MRI were: isotropic resolution = 33 µm, TE/TR = 20/40 ms, FA = 15°, FOV = 25.6 x 15 x 12.8 mm, NA = 12, and scan time = 15 h 8 min. For multi-gradient echo (MGE) imaging, the parameters were: TE = 5 ms, TR = 60 ms, Echo Spacing = 9 ms, Echo Images = 4, FA = 40°, FOV = 25.6 x 15 x 12.8 mm, NA = 10, and scan time = 17 h 4 min.

The Medical Image Processing, Analysis, and Visualization (MIPAV) program (http://mipav.cit.nih.gov) was used to quantify the changes of hypointensity at the olfactory bulb (Saar et al., 2015). A line of voxels of interest (VOI) was drawn from external plexiform layer (EPL) through the glomerular layer (GL) and ONL of the bulb (Supplementary Fig. S2). The % decreased intensity was calculated relative to the EPL intensity using the formula: % Signal Decrease = (Intensity_iNFL – Intensity_EPL) / Intensity_EPL.

T₂* maps were made from MGE images (parameters as shown above) using the Bruker ParaVision 360. The iron concentration (mM) in the iNFL ([Fe]_iNFL) was calculated using the formula: (1/T₂*_iNFL – 1/T₂*_Cortex) / r₂* + [Fe]_Cortex. The iron concentration in the cortex of mouse varied from 0.1 to 0.2 mM (Hare et al., 2012).

We estimated the value of transverse relaxivity, r₂*, of iron at 11.7T by using two methods. The r₂* value was calculated to be 33 mM^-1s^-1 at 11.7T from published measurement of transverse relaxation rate, R₂*, of human tissue with respect to iron concentrations in different brain regions as a function of field strength (Yao et al., 2009). A similar strategy was used for the second method, where r₂* was determined to be 24 mM^-1s^-1 based on T₂* maps generated from the mouse brain (Supplementary Fig. S3) and using previously reported iron concentrations in different mouse brain regions (Hare et al., 2012).

Animals were perfused transcardially with 5% formalin. Heads or brains were isolated and post-fixed with 10% formalin at 4°C for 24 h. In order to isolate intact ONL, decalcification in EDTA (0.5 M, pH 8.0) at 4°C for 5–7 days was performed (Moseman et al., 2020). EDTA solution was changed every 2 days. Brains or decalcified heads were transferred into 15% sucrose for 1 day, followed by 30% sucrose for 1 day. Brains were cut coronally and embedded in Tissue-Tek O.C.T. Compound (Sakura). Cryosections were cut coronally (30 µm) by a cryostat (Leica CM 1850). Brain sections were stained using a standard procedure for free-floating immunohistochemistry. Briefly, brain sections were rinsed with TBS 3 x 5 min, followed by blocking in 5% normal serum in TBS with 0.5% Triton X-100 for 2 h at 4°C. Then, the sections were stained in the following antibody solutions in TBS with 3% normal serum and 0.3% Triton X-100 overnight at 4°C. Rabbit anti-ferritin (anti-ferritin heavy chain, 1:100 dilution, Abcam) was used to stain ferritin. Goat anti-Sox10 (10 µg/mL, R&D Systems) was used to stain OECs. To study microglia activation, rabbit anti-IBA-1 antibody (0.4 µg/mL, FUJIFILM Wako) and rat anti-CD68 antibody (5 µg/mL, clone FA-11, Bio-Rad) were used. When lipofuscin quenching was needed, TrueBlack® Plus Lipofuscin Autofluorescence Quencher (Biotium) was used after IHC staining according to the supplier’s protocol. Astrocytes were stained with rabbit anti-GFAP (6 µg/mL, polyclonal, Agilent Dako). Axons in the ONL were stained with goat anti-olfactory marker protein (OMP, 1:200 dilution, FUJIFILM Wako). Tissue sections were put on the Superfrost microscope slides (Fisherbrand). After being dried overnight at room temperature, slides with brain sections were coverslipped using ProLong Gold antifade reagent with or without DAPI (Thermo Fisher Scientific). Images were captured using a Nikon Eclipse Ti microscope or a Leica Stellaris 8 Confocal microscope.

Iron staining of brain tissue was performed using an Iron Stain Kit (Abcam), according to the supplier’s protocol with some modification. Cryosections were cut coronally (20 µm), and tissue sections were put on the Superfrost microscope slides directly. After being dried on a slide heater for 2 h and cooled down for 30 min, slides were rinsed in distilled water x2. Slides were incubated in the working Iron Staining Solution for 5 min followed by rinsing in distilled water thoroughly. Slides were stained in Nuclear Fast Red Solution for 5 min followed by rinsing in distilled water x4. Slides were then dehydrated in 95% alcohol followed by absolute alcohol. Slides with brain sections were coverslipped using ProLong Gold antifade reagent without DAPI. Staining results were imaged immediately by using a Nikon Eclipse Ti microscope.

Twelve-week-old mice (N = 4) were examined. Three sections of olfactory bulb tissue from each mouse were used for quantification. Images were acquired under 10x views (0.2 mm x 0.2 mm) in the ONL. The number of each cell type was divided by the total number of DAPI + cells. The average value of the 3 tissue sections was used to present the value of one mouse.

Lipofuscin were identified by its broad emission spectrum autofluorescence that covers wavelengths between 480 and 660 nm (Warburton et al., 2007). Images obtained under 10x view (0.2 mm x 0.2 mm) were used for quantification. The number of lipofuscin particles was counted manually.

Twelve-week-old mice were perfusion-fixed with 4% paraformaldehyde (PFA) in PBS at room temperature. The fixed brains with intact olfactory bulbs were dissected and stored in the same fixative until the olfactory bulbs were vibratomed coronally into 100 µm slices. The total fixation time was kept between 30 and 40 min, from the start of fixation to the finish of slicing. The olfactory bulb slices were kept in PBS between 1 and 10 days and then processed for pre-embedding immunogold labeling (Tao-Cheng et al., 2021). Briefly, slices were blocked and permeabilized in 5% normal goat serum with 0.1% saponin for 30–60 min, incubated with goat-anti-Sox10 primary antibodies or mouse monoclonal antibody against synaptic vesicle glycoprotein 2 (SV2, clone 10H3, a gift from Dr Erik S. Schweitzer) as control in 5% NGS with 0.05% saponin in PBS for 1 h, washed, and incubated with donkey-anti-goat secondary or goat-anti-mouse Nanogold antibody (Nanoprobes) at 1:200 in 5% NGS with 0.05% saponin in PBS for 1 h, washed, and fixed with 2% glutaraldehyde in PBS. Samples were then stored at 4˚C for at least overnight. Samples were washed in deionized water and silver enhanced (HQ silver enhancement kit, Nanoprobes), then treated with 0.2% osmium tetroxide in 0.1 M phosphate buffer at pH 7.4 on ice for 30 min, washed and treated with 0.25% uranyl acetate in acetate buffer at pH 5.0 at 4°C for 1 h, washed and dehydrated in a grade series of ethanol, and embedded in epoxy resins. Thin sections were cut at ~70 nm and counterstained with lead citrate. Images were examined on a JEOL 1200 EX transmission electron microscope at 60 KV, and photographed with a bottom-mounted digital CCD camera (AMT XR-100, Danvers, MA, USA). OEC, axon bundles, endothelial cells, pericytes, and astrocytes were identified based on their structural characteristic from previous EM studies. The size and appearance of detected iron particles in the present study were consistent with a previous report (Zhang et al., 2005), where iron-containing ferritin molecules were identified by electron energy loss spectroscopy.

To estimate the ferritin concentration in OECs, the number of ferritin-iron particles in the OECs cytosol (N = 20 OECs) were counted manually from marked areas of the EM images. The volume of the cytosol was calculated using the formula: length of the marked area x width of the marked area x 70 nm (thickness of thin sections).

Statistical analysis was carried out with the one-way analysis of variance (ANOVA) with Tukey’s multiple-comparisons test. All statistical analyses were performed using GraphPad Prism 10.2.3. The statistical reports were shown in the Supplementary Material.

In normal 12-week-old mice (both male and female), T₂*-weighted MRI detected a large decrease in signal from the edge of the GL into the ONL. Representative in-vivo and ex-vivo images of male mice are shown in Figure 1A and B. Representative ex-vivo images of male and female mice are shown in Supplementary Figure S1. There was a 48% decrease in signal from the EPL to the ONL. No significant differences between sexes were observed (p > 0.05) (Supplementary Fig. S1). MGE MRI was used to identify the GL due to the shorter T₁ of this layer at a short TE (TE = 5 ms, Fig. 1C). The largest signal decrease was identified at the iNFL of the ONL, with some hypointensity at the GL and outer nerve fiber layer (oNFL) (Fig. 1C).

The MRI signal decrease in the olfactory bulb was studied as a function of age in mice and rats (Fig. 2). The decreases of intensity from the EPL to the ONL were quantified in Figure 2B and D. Line VOI intensity graphs were shown in Supplementary Figure S2. As quantified in Figure 2B, the decreases of intensity at the iNFL at 3 weeks, 6 weeks, 12 weeks, and 2 years were 16.5 ± 4.9%, 30 ± 5.5%, 48.2 ± 11.5%, and 73.1 ± 8.4%, respectively. In 2-year-old mice, the hypointense region continued to grow in the iNFL and many other locations of the bulb, such as the granule cell layer (GCL) and the mitral cell layer (MCL). The iNFL still showed the largest decrease in signal.

This rapid decrease of signal in T₂*-weighted MRI in the ONL of young mice was not observed in young rats of comparable age (Fig. 2C, D). A 17.3 ± 4.6% decrease of intensity was detected in 12-week-old rats, compared with 48.2 ± 11.5% in 12-week-old mice. In rats at 11 months of age, there was a 54.6 ± 8.4% decrease in MRI signal intensity, similar to signal changes found in 3-month-old mice. These changes were similar to those detected in 12-week-old mice.

To determine if iron levels were increased in region of signal loss, iron Perl’s staining and ferritin immunohistology was performed. In 3-week-old mice, when no hypointensity was detected in the ONL, no high iron levels and very little ferritin was detected (Fig. 3A–C). In 12-week-old mice, at the area where large signal loss was detected in the ONL, especially in the iNFL, high iron and ferritin accumulation was detected (Fig. 3G, H). At other areas of the ONL and GL where no hypointensity was detected, no iron and very little ferritin was detected (Fig. 3E, F). Based on T₂* maps (Supplementary Fig. S3), the iron concentration in the iNFL was calculated to be in the range 2–3 mM depending on whether a transverse relaxivity, r₂*, for iron in vivo of 24 or 33 mM^-1s^-1 (see Section 2). Compared to cortical iron of 0.1–0.2 mM in mouse (Hare et al., 2012), this indicates a very high iron in the iONL.

To determine the cellular origin of the high intrinsic iron, olfactory bulb tissues of 12-week-old mice were stained for OECs (Sox10), astrocytes (GFAP), microglia (IBA1), oligodendrocytes (Oligo2), and axons (olfactory marker protein, OMP). We found that OECs were the major cell type (84.2 ± 6.2% of DAPI+ cells) in the ONL (Fig. 4A). There was a small number of astrocytes (6.5 ± 0.8%) and microglia (4.5 ± 0.5%). Oligodendrocytes were barely detected, which is consistent with previous reports that OECs, not oligodendrocytes, wrap the bundles of axons at the ONL (Boyd et al., 2005; Reshamwala et al., 2019). OECs were also the major cell type in the ONL of 3-week-old mice (Fig. 4B; Supplementary Fig. S4B) and 12-week-old rats (Supplementary Fig. S4C). A ferritin heavy chain antibody that works efficiently in both mice and rats (Supplementary Fig. S4A) was used in this study. Co-staining of OECs (Sox10) and ferritin showed that there was a high level of ferritin in the OECs in mice of 12 weeks (Fig. 4C), but not 3 weeks (Fig. 4B; Supplementary Fig. S4B). There were not high levels of ferritin in 12-week-old rats (Supplementary Fig. S4C). Ferritin is the most important iron storage protein, and high levels of ferritin are usually associated with high levels of iron. This high level of ferritin was not in the astrocytes (GFAP+ cells) (Fig. 4D). The high ferritin did not appear to be located in axons (Fig. 4E). Due to the complex shape of OECs and their ability to wrap olfactory neuron axons, higher-resolution EM studies were performed to further identify the origins of the high iron in 12-week-old mice.

OEC cell bodies were identified by Sox10-immunogold labeling in the nuclei in EM (Fig. 5A1, B2, D1; Supplementary Fig. S5A2, B2). The structural characteristics of the nuclear chromatin configuration resembled those from previous EM reports (Au et al., 2002; Doucette, 1989; Valverde et al., 1992). Consistent with IHC staining results (Fig. 4A1), almost every nucleus encountered in the ONL was Sox-10 positive (26 out of 29). Figure 5A–C and Supplementary Figure S5 show several representative views of an OEC cell body adjacent to a blood vessel and OEC processes wrapping around axon bundles. High concentrations of dark particles with diameters of ~6 nm (yellow circles) were present in the cytoplasm of Sox-10 positive OECs (Fig. 5A, B). Although no available ferritin antibody worked for immuno-EM in our hands to further verify the identity of these particles, the size and shape of these dark particles are identical with those of iron-containing ferritin molecules identified by electron energy loss spectroscopy (Zhang et al., 2005). Thus, these particles here are interpreted as iron-ferritin particles.

Gap junctions between OECs (Fig. 5D1,2) were also an identifying feature of OECs (Rela et al., 2010), and the presence of iron-ferritin particles in gap junction-containing cellular processes added more evidence that OECs are the cell type containing high iron in ONL. OECs form sheet-like processes (highlighted in yellow in Figure 5A, C and Supplementary Fig. S5) to wrap around axon bundles. Iron-ferritin particles (yellow circles in Figure 5A–D at high magnification) were prominently present in the cytosol of OECs, but not in the axons (Fig. 5A3, C). Interestingly, iron-ferritin particles were also observed in pericytes (Fig. 5A2 and Supplementary Fig. S5A5, labeled in blue) next to endothelial cells.

Based on EM of individual OECs (N = 12), ferritin concentrations in the cytosol of OECs were estimated to be 2–12 µM. Assuming 800–1000 irons per ferritin (Strbak et al., 2021), this estimates iron concentration between 1.6 and 12 mM.

The MRI data, iron staining, and ferritin staining indicated relatively high levels of iron in the iNFL, and some iron in the GL and oNFL. Iron is associated with inflammation in number of neurodegenerative disorders and therefore it was determined if there was evidence of neuroinflammation. Near the high iron areas in the iNFL, large irregular lipofuscin granules, a neuroinflammation marker, were observed in the GL and ONL, near the GL, in mice as early as 6 weeks (Fig. 6A). Lipofuscin was barely detected in 3-week-old mice or in 6-month-old rats (Fig. 6B). At these early timepoints, the large MRI signal loss was not yet detected. The number of lipofuscin granules increased from 11 ± 3 (6-week-old mice) to 43 ± 7 (12-week-old mice) per view (0.2 mm x 0.2 mm) as did the size of lipofuscin granules. The T₂* hypointensity became more prominent over this same period (Fig. 1).

In rats, the correlation between MRI hypointensity and increasing lipofuscin was also observed (Figs. 2C and 6B). T₂* hypointensity was barely detected in 12-week-old rats where no lipofuscin was observed. A little more signal loss was detected in 6-month-old rats, where a few lipofuscin particles, 7 ± 2 per view could be observed. When a larger signal loss and increased ferritin and iron was detected in 11-month-old rats, significantly more lipofuscin, 46 ± 7 per view, was observed. Interestingly, the degree of hypointensity and the number of lipofuscin particles were similar in the 3-month-old mice and the 11-month-old rats.

In addition to lipofuscin, CD68 + activated microglia could be detected at the GL and ONL (Fig. 6C). TrueBlack was applied to stain lipofuscin and block their autofluorescence (Supplementary Fig. S6). Lipid spots from TrueBlack staining were observed in the enlarged lysosome of microglia, as indicated by CD68 (Fig. 6C). EM also identified lipofuscin in the lysosome of microglia at the GL and ONL (Fig. 6D). In other areas of GL and ONL where large MRI signal loss was not detected, lipofuscin and microglia activation were also not observed (Fig. 6E2).

Figure 7 summarizes the locations of T₂* hypointensity (A), OECs stained by Sox10 (B), and lipofuscin (C) at the GL and ONL (iNFL and oNFL) in 12-week-old mice. MRI signal loss and OECs were mostly located at the iNFL. Lipofuscin was mostly located at the GL.

T₂*-weighted MRI has been used to guide histopathology studies to study brain iron accumulation (Bulk et al., 2020; Griffin et al., 2019; Liu et al., 2022; Tisdall et al., 2022). It is well established that increases in iron cause decreases in signal in T₂*-weighted MRI. In the present study, MRI detected a large decrease of signal at the iNFL of the ONL in young, normal mice from 3 to 12 weeks (Fig. 2A). This T₂* hypointensity guided us to identify high iron in OECs. It is likely that this high iron is the cause of the MRI changes detected. The other major contributor to T₂* signal loss in brain MRI is myelin (Nair et al., 2020), but the ONL is devoid of myelin, likely indicating that the change in T₂*-weighted signal intensity was due to the high iron. Interestingly, rats developed the MRI signature of high iron in the olfactory bulb at a much slower rate than mice (Fig. 2C).

What is the origin of the high intrinsic iron in OECs? Based on EM of ferritin in the cytosol of OECs and assuming that each ferritin has 800–1000 irons (Strbak et al., 2021), iron concentrations in the cytosol of OECs ranged from 1.6 to 12 mM. Quantitating T₂* changes in the iNFL compared to cortex of the brain was used to estimate an iron concentration to be 2–3 mM. In both cases, this is very high compared with typical iron concentration in the mouse brain cortex of 0.1–0.2 mM. The highest brain area containing iron in humans is the substantia nigra (nigrosome 1) which has about 2.7 mM iron in healthy controls and 4.3 mM in Parkinson’s Disease (Friedrich et al., 2021). Thus, the iron in the OECs in the iNFL is within range of the highest iron concentrations found in the human brain. It has been reported that OECs are the main phagocytic cells that remove axon debris during development of the olfactory system (Nazareth et al., 2015). Macrophages contributed only a minor role to clearing the axon debris. Likely phagocytosis of axonal debris can lead to increases of iron. This is likely why microglia iron has been associated with neurodegeneration (Kwan et al., 2012; Porras & Rouault, 2022). The rapid accumulation of high iron in the OECs might indicate that there is a high turnover rate of OSNs as OECs engulf a large amount of axon debris in the iNFL of ONL in mice between 3 and 12 weeks. This would predict that rat olfactory neurons turn over slower than mouse. There are few estimates of olfactory neuron turnover in rodents or higher species, and these are usually measured in response to injury (Brann & Firestein, 2014).

Increases in brain iron are a hallmark of neuroinflammatory and neurodegenerative diseases (Urrutia et al., 2021). Neuroinflammation and brain iron accumulation can potentially enhance each other through multiple mechanisms. For example, in PD, a number of possible mechanisms for increased iron accumulation in the substantia nigra have been proposed, including: leaky blood–brain barrier (Faucheux et al., 1999), increased expression of proteins critical for iron uptake, for example, transferrin receptor 1 and divalent metal transporter 1, in dopamine neurons (Ma et al., 2021; Salazar et al., 2008), decreased expression of the iron exporter (ferroportin-1) (Ma et al., 2021), or increased actoferrin receptors in neurons and small vessels (Faucheux et al., 1995). Studies have shown an increase in redox-active iron associated accumulation of neuromelanin-iron complex and aggregation of α-synuclein in substantia nigra neurons of PD patients (Faucheux et al., 2003; Wise et al., 2022). Ferroptosis, which is an iron-dependent form of cell death driven by iron-dependent phospholipid peroxidation, has been implicated in neurodegeneration (Ryan et al., 2023). Identifying iron accumulation and neuroinflammation in the olfactory bulb has been investigated as early biomarkers for neurodegeneration, such as AD and PD, aiding in early diagnosis and intervention. Strategies to modulate iron levels and reduce neuroinflammation could be potential therapeutic approaches to mitigate neurodegeneration in AD/PD and other related disorders.

Interestingly, large irregular lipofuscin granules and activated microglia were found near the high iron region in young mice where there is not believed to be pathological neurodegeneration. However, there is normal olfactory neuron turnover (Figs. 6 and 7). Lipofuscin has been shown to form through reactive oxygen species (Barbouti et al., 2021). A necessary precondition for reactive free radical generation and lipofuscin formation is the intracellular availability of ferrous iron (Fe2+) (“labile iron”), catalyzing the conversion of weak oxidants, such as peroxides and lipids, to extremely reactive ones like hydroxyl (HO•) or alcoxyl (RO•) radicals. Iron is also predominant among the metals in lipofuscin (Gray & Woulfe, 2005). In young mice, most lipofuscin granules were found at the GL (Figs. 6A and 7C), where some MRI signal loss was detected (Figs. 1B and 7A). This is an area where the density of OECs was low (Figs. 4A and 7B). In comparison, much fewer numbers of lipofuscin granules were found in the iNFL, where prominent MRI signal loss was found and the density of OECs was high. This might indicate that OECs may be able to safely store high iron in ferritin which can prevent the formation of reactive oxygen species. Lipofuscin was not found in the olfactory system of young rats (Fig. 6B), which is consistent with previous reports (Kosaka et al., 2009), but was found in older rats when MRI signal loss occurred. Microglia activation and formation of lipofuscin could be due to the turnover of the axonal terminals of olfactory neurons in the glomerular layer due to microglia being the main phagocytic cell in this part of the olfactory bulb.

OECs are part of the glia limitans and act as a physical barrier against unwanted cells, viruses, or molecules attempting to enter the brain via the olfactory nerve system (Beiersdorfer et al., 2020). In the olfactory system, OSNs are directly connected from the nasal epithelium through the cribriform plate to the olfactory bulb and brain. The access through the cribriform plate to the brain makes the olfactory system a source of bacteria, virus, or fungal infection for the brain. Many studies suggest that OECs are the main phagocytic cells and exert crucial roles in protecting the olfactory nervous system against invasion by pathogenic organisms (Harris et al., 2009; Nazareth et al., 2015). Thus, some of the iron accumulation could be due to phagocytosis of pathogenic organisms in the mouse. Why this would be the case for younger mice but not rats is not clear. The early accumulation of iron in mouse OECs raises questions about its role in OEC function or its relation to phagocytic activity. OECs can modulate microglia-astrocyte responses by secreting anti-inflammatory cytokines and are also capable of down-regulating pro-inflammatory factors (Denaro et al., 2022). Due to their supportive role in supporting regeneration of OSNs and their immunomodulatory functions, OECs are a candidate for cell therapy for CNS injuries (Oieni et al., 2022). Superparamagnetic iron oxide labeled OECs have been traced in vivo by MRI to study their functions in cell therapy (Dunning et al., 2004). The fact that OECs can accumulate high iron may make it possible to use MRI to track OECs in cell transplantation without adding any contrast agent.

In conclusion, the ONL of the mouse develops a prominent MRI T₂* signal loss from 3 to 12 weeks of age. There is a high level of iron in the OECs at the iNFL of the ONL of mice at a time when MRI hypointensities become prominent and the iron accumulation is the likely cause of the MRI changes detected. Interestingly, lipofuscin accumulation and microglia activation occur near the iron accumulation. It is not clear whether the iron accumulation plays a role in the lipofuscin and microglia activation or whether both are independently occurring due to ongoing degeneration and regeneration of OSNs. Interestingly, rats accumulate the MRI signature of high iron about three times slower than mice. It should be possible to survey other species quickly using MRI to determine if they also accumulate iron and at what age. The differences may have to do with a higher rate of OSN regeneration or a higher metabolic rate in the mouse than the rat. The connection between iron and inflammation in young, healthy mice warrants further investigation. The iron accumulation in OECs associated with activated microglia in young animals should enable the links between iron accumulation and inflammation to be studied without the confounds of studying older animals. Understanding the role of iron in the context of neuroinflammation and neurodegeneration provides valuable insights into the underlying mechanisms of neurodegeneration and informs the development of targeted therapies.

• Data are available on https://figshare.com. The links of data of each Figure are:

• Data of Figure 1 https://figshare.com/s/b95a99573c5467969252.

• Data of Figure 2 https://figshare.com/s/8bfd3efe031be8e07268.

• Data of Figure 4 https://figshare.com/s/97c115e95fc2b861bc35.

• Data of Figure 5 https://figshare.com/s/22bc4d976b51f8b7ea60.

• Data of Figure 6 https://figshare.com/s/8a4801bde726630dda1e.

L.L. performed the experiments and wrote the manuscript with input from other authors. J-.H.T-.C. performed EM experiments. H.R. contributed imaging analysis and calculated iron concentration. S.D. performed MRI experiments. N.B. performed perfusion experiments. A.P.K. designed experiments, wrote the manuscript, and supervised the project.

This research was supported by the intramural program at the National Institute of Neurological Disorders and Stroke (NINDS), National Institutes of Health (NIH).

All authors declare that no competing interests exist.

The authors would like to thank Dr. Peijun Zhang of the University of Oxford for suggestions and discussions on visualizing iron-bound ferritin particles in EM experiments. The authors also thank Mrs. Sandra Moreira for her assistance in the EM work.

Supplementary material for this article is available with the online version here: https://doi.org/10.1162/imag_a_00299