For this study we used eight outbred Wistar Han rats purchased from Charles River (France). We divided the animals into four age-groups, 3-months, 6-months, 12-months, and 18-months, each containing two rats. All procedures were approved by the Norwegian Animal Research Authority, and follow the European Convention for the Protection of Vertebrate Animals used for Experimental and Other Scientific Purposes. We kept the animals on a 12-h light/dark cycle under standard laboratory conditions (19–22 °C, 50–60% humidity), with access to food and water ad libitum. The human case was from archival material obtained from King’s College London Biobank.

Processing and immunohistochemistry. We fully anaesthetised the animals by placing them in a chamber containing 5% isoflurane gas (Baxter AS, Oslo, Norway), followed by an intraperitoneal injection of pentobarbital (0.1 ml per 100 g; SANIVO PHARMA AS, Oslo, Norway). Subsequently we carried out transcardial perfusions with Ringer’s solution at room temperature (in mM: NaCl, 145; KCl, 3.35; NaHCO3, 2.38; pH ~ 6.9) immediately followed by circulation of 4% freshly depolymerised paraformaldehyde (PFA; Merck Life Science AS/Sigma Aldrich Norway AS, Oslo, Norway) in a 125 mM phosphate buffer (PB) solution (VWR International, Pennsylvania, USA), pH 7.4. The brains were extracted and post-fixed overnight in PFA at 4 °C, and then placed in a freeze protective solution (dimethyl sulfoxide in 125 mM PB with 20% glycerol, Merck Life Science AS/Sigma Aldrich Norway AS, Oslo, Norway) until sectioning. We sectioned the brains coronally at 40 μm into six series, using a freezing sledge microtome.

Immunohistochemistry was done on free-floating sections. To improve access to the antigen, the sections were subjected to gentle heat induced antigen retrieval by placement in PB at 60 °C for two hours. For chromogenic immunoenzyme-labelling: To avoid unspecific labelling the sections were first incubated for 2 × 10 min in a solution of PB with 3% hydrogen peroxide (Merck Life Science AS/Sigma Aldrich Norway AS, Oslo, Norway), and next incubated for 1 h in a solution of PB with 5% normal goat serum (Abcam Cat# ab7481, RRID: AB_2716553). Subsequently, free-floating sections were incubated with the primary antibody, rabbit anti-BNIP3 (BioNordika, Oslo, Norway, Cat# CST-3769S, RRID:AB_2259284) in a 1:500 dilution in PB containing 0.4% Saponin (VWR International AS, Bergen, Norway) and 5% normal goat serum, for 48 h. We also tested a primary incubation time of 20 h (i.e. overnight), which gave slightly less overall signal intensity. We opted for 48 h to make sure we did not miss any potentially weakly positive neurons or neuropil. Sections were then rinsed for 3 × 5 min in Tris-buffered saline (50 mM Tris, 150 mM NaCl, pH 8.0) containing 0.4% Saponin (TBS-Sap), and then incubated with biotinylated secondary antibody goat anti-rabbit (1:400; Sigma-Aldrich Cat#B8895, RRID: AB_258649) in TBS-Sap, for 90 min at room temperature. Subsequently, the sections were rinsed in TBS-Sap, and then incubated in Avidin–Biotin complex (ABC, Vector Laboratories Cat# PK-4000, RRID: AB_2336818) for 90 min at room temperature. The sections were then rinsed again, and then incubated in a solution containing 0.67% 3,3ʹ-Diaminobenzidine (Sigma/Merck, Cat# D5905) and 0.024% hydrogen peroxide for 4 min. We safeguarded against potentially subtle differences that can arise from small variations in the time of incubation or the solution with the chromogen by incubating sections from different age groups in parallel, using the same solution of Diaminobenzidine. For immunofluorescence labelling: All tissue was first subjected to Heat Induced Antigen Retrieval, by immersion in PB at 60 °C for 2 h, and then blocking against unspecific labelling was done by placing the tissue in 5% goat serum in PB for 1 h. For reelin and Bnip3, we incubated the sections overnight with primary antibodies (mouse anti-reelin G10 clone (1:1000), Millipore, Cat# MAB5364, RRID:AB_2179313 and rabbit anti-BNIP3 (1:500), BioNordika, Oslo, Norway, Cat# CST-3769S, RRID:AB_2259284) in PB solution containing 0.4% Saponin (VWR, Cat# 27,534.187) and with 5% goat serum at 4 degrees Celsius. The sections were subsequently incubated for two hours with secondary antibodies, including Alexa 546 Goat anti-Mouse (Thermo Fisher, Cat#A11003, RRID: AB_10562732) and Alexa 488 Goat anti-Rabbit (Thermo Fisher Scientific Cat# A-11008, RRID:AB_143165), both at 1:500 in a solution containing 0.4% Saponin and 5% goat serum at room temperature. For Bnip3 and tyrosine hydroxylase (TH) double-immunolabeling, we used the same approach as above with rabbit anti-BNIP3 (1:500) and mouse anti-tyrosine hydroxylase (1:1000, Millipore Cat# MAB318, RRID:AB_2201528). For the triple-labelling experiments (Supplementary Fig. 2) we first incubated sections with the primary antibodies against reelin and Bnip3 overnight, before incubating them for two hours with secondary antibodies as detailed above. The sections were subsequently incubated overnight with primary antibody rabbit anti-calbindin (1:5000, Swant Cat# CB38, RRID:AB_10000340), followed by secondary antibody Alexa 647 Donkey anti-Rabbit (1:500, Thermo Fisher, Cat#A-31573, RRID: AB_2536183) for two hours. The clear segregation of the Bnip3 labelling vs the calbindin labelling means that we can safely assume that the Alexa 488 Goat anti-Rabbit secondary antibody had fully saturated the epitopes of the rabbit anti-Bnip3 primary, such that this primary was not bound by the subsequent application of the Alexa 647 Donkey anti-Rabbit secondary antibody.

Processed tissue was mounted on Superfrost™ glass slides (Thermo Fisher Scientific) from a solution of 50 mM tris(hydroxymethyl)aminomethane (Millipore, Cat# 1.08382.1000) with hydrochloric acid, at pH 7.6, and then left to dry overnight before being coverslipped using a mixture of Xylene (VWR International AS, Bergen, Norway) and entellan (Merck KGaA, Cat# 1.07960.0500).

Imaging and quantification. All tissue sections were scanned on an automated Zeiss Axio Scan.Z1 system with a 20 × objective (Plan-Apochromat 20x/0.8 M27), using identical settings. Analyses and quantifications were done in Zen software (2.6, Blue Ed.). For quantifications, we used a series of sections covering the rostrocaudal extent of EC from each of three animals, and double-immunolabeled against reelin and Bnip3. While visualizing only the channel used to scan the reelin signal (546-channel), we selected cell profiles using the Zen Circle Tool, taking care to obtain good coverage also of the rostrocaudal axis, but avoiding the part at the ventral extreme (i.e. the part furthest away from the rhinal fissure) since the signal here is very weak relative to background. After this selection, we switched to the channel for the Bnip3 signal (488-channel) and identified each circle with a clearly positive signal. By this we obtained the ratio between Bnip3-positive neurons and reelin-positive neurons, i.e. \(\frac\).

Controlling for background labeling by the secondary antibodies. To control for possible background labeling by the secondary antibodies we immunolabeled sections containing EC from different age groups and omitted the primary antibody in one section per run. We found no signs of background labeling in our setup (Fig. 1).

Fig. 1

No evident background labeling in our setup. a Immunolabeling with the Bnip3 antibody reveals a high level of signal in LEC layer II of Wistar rats (example from 18-month-old animal). b Running the same experiment without the Bnip3 antibody while keeping all else equal shows a complete absence of signal. The same is the case across the age groups. Scale bars indicated on each figure. Dashed line marks border between LEC and the piriform cortex (Pir)

Orientation of sections. To aid visualisation of the position of the sections shown in the Results, we provide a cartoon (Fig. 2).

Fig. 2

Guide to orientation of sections. a Cartoon of rat brain seen from a laterocaudal perspective, with the situation of the entorhinal cortex (EC) indicated, and an illustration of a virtual resection and compression of EC into a flatmap. b Matching flatmap of EC layer II with reelin-expressing fan cells illustrated for the lateral entorhinal cortex (LEC) and reelin-expressing stellate cells illustrated for the medial entorhinal cortex (MEC). The decrease in color intensity for the cells indicate the decrease in reelin-expression as one moves successively away from the rhinal fissure (RF). Blue dashed lines (i–iii) indicate the locations of the coronal sections displayed in Fig. 3a–c, respectively

Bnip3 expression in the entorhinal cortex

We found that the expression of Bnip3 is near undetectable in the cell somata of most parts of the forebrain. This is in line with previous reports (He et al. 2019; Lu et al. 2014; Zhang et al. 2011; Althaus et al. 2006) and the finding that pathological conditions such as hypoxia or ischemia are required to trigger increased expression (Lu et al. 2014; Zhang et al. 2011; Althaus et al. 2006). However, in stark contrast to this general finding, a major subset of neurons in the part of EC LII that is located close to the rhinal fissure (i.e., dorsally) expressed high levels of Bnip3 in all age groups (Fig. 3 and Supplementary Fig. 1). Since the only difference between young and old rats appears to be a modest overall increase in staining intensity with age, the description to follow holds true for all age groups. The expression of Bnip3 in both LEC and MEC manifests a striking gradient (Fig. 3b and c). While the expression is high in LII-neurons located close to the rhinal fissure, the expression level gradually decreases at positions successively further away from the rhinal fissure, to the point where the level in the most ventrally located LII-neurons is almost below the detection limit.

Fig. 3

A major subset of ECLII neurons expresses Bnip3. a Example of the dorsolateral EC (DLE, i.e. the anterior tip of LEC), in which LII-neurons are particularly strongly stained. b The expression of Bnip3 gradually decrease at positions successively further away from the rhinal fissure (RF), irrespectively of whether one is located in LEC or MEC (see Fig. 1 as a guide to how the sections relate to each other across EC and between LEC vs MEC). c Example at the caudal extreme (CE) of EC where only the part of MEC located close to the rhinal fissure is present in coronal sections. (d–f) shows the corresponding labeling for 18 month old animals. Per Perirhinal cortex, PaS Parasubiculum, Por Postsubiculum. Scale bars are indicated for each figure, note that for the insets in (b) and (e) the scale bar in the leftmost inset applies to all

Bnip3-expression is restricted to the reelin-positive ECLII-population

Assuming that the level of Bnip3 expression serves as a proxy of the neuronal mitochondrial turnover rate, and that the turnover rate is positively correlated with the neuronal metabolic rate, it follows from the data (Fig. 3) that the overall neuronal metabolic rate is substantially higher in EC LII than in other layers of the EC, as well as in most other parts of the rat brain (see below). And the data clearly suggest that energy expenditure is highest in neurons located close to the rhinal fissure.

Since the strength of the Bnip3 signal clearly appeared to align with the previously reported reelin expression gradient in ECLII (Kobro-Flatmoen et al. 2016), we double-immunolabeled against reelin and Bnip3. The results (Fig. 4) clearly suggest that high Bnip3 expression is restricted to the Re + ECLII neurons.

Fig. 4

Double-immunolabeling against reelin and Bnip3. a Top panel shows labeling for reelin, bottom panel shows labeling for Bnip3. b Higher powered insets (numbers correspond to those in (a)) reveal that Bnip3 follows the same gradient as reelin, thus being strong towards the rhinal fissure (RF) and increasingly less so as one moves successively further away from the rhinal fissure. Further, it is apparent from these images that Bnip3 is confined to the population of Re + ECLII neurons. Scalebar for a applies to both images; scalebar for b1 applies to all six images

We therefore took one series of sections from each of three animals and quantified the degree of co-localization between Bnip3 and reelin throughout ECLII. Applying stringent criteria for what qualifies as a positive Bnip3 signal, we found that Bnip3 expression was restricted to Re + ECLII neurons, and that approximately 90% of Re + ECLII neurons were positive for Bnip3 (\(\frac\); animal 1\(: \frac=0.88\); animal 2:\(\frac=0.91\); animal 3:\(\frac=0.90\)). This is probably a conservative estimate, and we expect that more extensive analyses are likely to find that the degree of overlap is even higher. Given that the Bnip3 expression level serves as a proxy for neuronal metabolic rate, it appears that the data support our prediction (see Introduction) that the observed reelin-gradient in ECLII reflects the overall energy expenditure of the neurons along this gradient. To visually demonstrate that Bnip3 is specifically co-localized with Re + ECLII neurons and hence does not co-localize with Cb + ECLII-neurons, we performed triple-immunolabeling experiments against Bnip3, reelin and calbindin. These experiments further demonstrate the selective co-localization of Bnip3 with Re + ECLII neurons (Supplementary Fig. 2).

Indication that Bnip3 may be expressed also in human EC LII neurons

In a pilot test, we also found Bnip3-expression in human ECLII neurons (Fig. 5). But since we do not currently have access to complete human EC samples, we were prevented from assessing whether all our results obtained in Wistar rats extend to humans. However, as in rats, reelin is expressed in human principal ECLII neurons following the same topographical gradient (Kobro-Flatmoen and Witter 2019). Therefore, we consider it likely that a close overlap between reelin expression and Bnip3 expression will also be the case in humans.

Fig. 5

Bnip3-expression in human ECLII neurons. a Immunolabeling against Bnip3 in EC of a human subject appears to reveal a signal that is b most notable in LII (the insets in (b) match those in (a)). Arrows in (b) indicate examples of what are likely positive neurons. Scale bars are indicated in figures, (b 1’) also applies to (b 2’)

Other neurons expressing elevated levels of Bnip3 in rats

We then used Bnip3 as a probe to identify neurons or neuronal clusters that putatively have higher than average mitochondrial activity (Fig. 6). LII-neurons of the piriform cortex and occasional neurons with high levels in the preoptic nucleus display moderate and high levels of Bnip3 expression, respectively (Fig. 6a). In the hippocampal formation, the expression of Bnip3 is generally low or absent, except for field CA3, where a substantial number of neurons in the pyramidal layer express moderate levels. Notable levels of Bnip3-expression is also found in the outer 2/3ds of the molecular layer of the dentate gyrus (DG) and in stratum lacunosum moleculare of CA3/2, which contains the Re + ECLII neuron-terminals; we also observe that there is a tendency for more intense labeling in the free (lower) blade of DG (Fig. 6b). Furthermore, in parts of the visual cortex (Fig. 6c), the retrosplenial, somatosensory and motor cortices (Fig. 6d), and the frontal pole (Fig. 6e) we find subsets of neurons where Bnip3 expression is clearly above background, sometimes reaching moderate levels.

Fig. 6

Moderate levels of Bnip3-expression is present in certain structures outside of the entorhinal cortex. This includes a neurons in the piriform cortex and the preoptic nucleus, b CA3 (note the neuropil labeling in the outer 2/3ds of the molecular layer of DG, such labeling is also present in CA3 and CA2, albeit to a lesser degree as is clear from the insets), c visual cortex, d retrosplenial-, motor-, and somatosensory cortex, and e frontal cortex. Scale bars are indicated in each figure, note that for the related insets, the leftmost scale bar applies to all. a HDB Nucleus of the horizontal limb of the diagonal band, b gl granular layer of dentate gyrus, iml inner molecular layer of dentate gyrus, mml middle molecular layer of dentate gyrus, oml outer molecular layer of dentate gyrus, hf hippocampal fissure, rad stratum radiatum, pyr stratum pyramidale, or stratum oriens

Bnip3-expression in neurons of the dorsomedial part of the locus coeruleus (LC) is also high (Fig. 7a, b). We confirmed that these Bnip3 high-expressing neurons are indeed part of LC by double-labeling against tyrosine hydroxylase (Fig. 8). The moderate to high Bnip3 expression found in non-Re + ECLII neurons suggests that these neurons also have higher than average metabolic rate, but it is beyond the scope of this paper to elaborate on the reasons for this.

Fig. 7

High levels of Bnip3-expression in neurons of the locus coeruleus (LC) shown in a a three month old rat and b an 18 month old rat. Equally high levels are found in a subset of neurons belonging to the mesencephalic trigeminal nucleus (ME). Scale bars are indicated in each figure

Fig. 8

Bnip3-expression in neurons of LC co-localize with tyrosine hydroxylase. Scale bar in leftmost image applies to all