Clusters of secretagogin-expressing neurons in the aged human olfactory tract lack terminal differentiation

Abstract

Expanding the repertoire of molecularly diverse neurons in the human nervous system is paramount to characterizing the neuronal networks that underpin sensory processing. Defining neuronal identities is particularly timely in the human olfactory system, whose structural differences from nonprimate macrosmatic species have recently gained momentum. Here, we identify clusters of bipolar neurons in a previously unknown outer “shell” domain of the human olfactory tract, which express secretagogin, a cytosolic Ca²⁺ binding protein. These “shell” neurons are wired into the olfactory circuitry because they can receive mixed synaptic inputs. Unexpectedly, secretagogin is often coexpressed with polysialylated–neural cell adhesion molecule, β-III-tubulin, and calretinin, suggesting that these neurons represent a cell pool that might have escaped terminal differentiation into the olfactory circuitry. We hypothesized that secretagogin-containing “shell” cells may be eliminated from the olfactory axis under neurodegenerative conditions. Indeed, the density, but not the morphological or neurochemical integrity, of secretagogin-positive neurons selectively decreases in the olfactory tract in Alzheimer's disease. In conclusion, secretagogin identifies a previously undescribed cell pool whose cytoarchitectonic arrangements and synaptic connectivity are poised to modulate olfactory processing in humans.

Olfaction is an ancient sense of vertebrates pivotal for the individual's environmental adaptation, competitiveness, and survival. The olfactory system is the phylogenetically oldest part of the forebrain, with its proportion to the total brain volume progressively decreasing from rodents to humans (1–3). Rodents maintain and continuously refine their olfactory system by adding new neurons throughout life to olfactory glomeruli via the rostral migratory stream (RMS) (4, 5). The microsmatic character of humans is frequently associated with the reduced size of the rhinencephalon. The lack of rodent-like lifelong plasticity in the human olfactory circuitry may be related, at least in part, to the neonatal disappearance of continued chain migration of immature neurons in the human homolog of the RMS (6–10).

Mitral and tufted cells, output neurons of the main olfactory bulb (1), send olfactory information to interrelated allo- or neocortical areas via their axons coursing in the olfactory tract. The olfactory tract is particularly well formed and elongated in humans (and primates; ref. 11) and is believed to be neuron-free except from dispersed cells displaced from retrobulbar and prepiriform regions (12).

Ca²⁺ binding and sensor proteins sculpt fundamental nervous system functions and are used as selective markers for neuronal subpopulations (13). Secretagogin is a recently cloned Ca²⁺ binding protein (14) whose Ca²⁺ sensor functions are becoming increasingly appreciated (15). During the analysis of the primate olfactory system, we found secretagogin-positive (secretagogin⁺) neurons in the RMS and olfactory bulb (16). Therefore, we hypothesized that secretagogin may reveal previously undescribed cellular identities and cytoarchitectural arrangements in the human olfactory axis. Here, we report the existence of a neuron-rich cellular niche in the outer “shell” domain of the human olfactory tract, containing bipolar secretagogin⁺ cells neurochemically resembling deep-layer and periglomerular olfactory interneurons (17) and synaptically integrated into the olfactory circuitry. Clustered “shell” cells retain molecular markers of immature neuronal identity in a descending gradient toward the olfactory bulb.

Olfactory dysfunction is a prevalent premonitory sign of Alzheimer's disease (AD) due to damage to olfactory centers or regions interrelated with the olfactory pathway (18). We show the selective decline of secretagogin⁺ shell cells in the olfactory tract in AD. In contrast, secretagogin⁺ periglomerular neurons survive and remain devoid of amyloid β (Aβ) or tau pathology. Overall, we define a unique neuronal subtype in the human olfactory system whose loss can be implicated in impaired olfactory information processing during aging and under neurodegenerative conditions.

Results and Discussion

Secretagogin Identifies Bipolar Cells in the Human Olfactory Tract.

We have recently shown secretagogin⁺ periglomerular interneurons in the olfactory bulb of the gray mouse lemur (Microcebus murinus, Primates) (16). However, whether secretagogin is expressed in the human olfactory system remains unknown. Here, we address secretagogin's distribution and the identity of cells in the human olfactory tract and bulb (Table S1), rostral from the olfactory trigone and excluding presumed proliferative zones (Fig. 1A), by using affinity-purified polyclonal antibodies whose specificity has been established (16) (Table S2).

Fig. 1.

Secretagogin in the human olfactory tract and bulb. (A) Schematic outline of the human olfactory system. Vertical red line denotes the approximate tissue cutoff at the level of the olfactory trigone. Red circles indicate the positions of identified secretagogin⁺ cells. (B and C) Immunostained perikarya and processes load the olfactory trigone (B), tract (C), and bulb. (B) Secretagogin⁺ cells along the opposite margins of the olfactory stria. (C) Bipolar neurons align the human olfactory tract. Processes (arrowheads in C₁–C₄) occasionally ramify (C₁). (D) Independent of age or Alzheimer's severity, secretagogin⁺ neurons invariably accumulate in the outer shell domain (100–300 μm from the surface) of the olfactory tract. Arrows indicate direction toward the olfactory bulb. [Scale bars: 250 μm (B and C), 20 μm (C₁), and 10 μm (C₂–C₄).]

Secretagogin immunoreactivity was localized to oval-shaped cells with a somatic diameter of 8–15 μm and distributed along the olfactory trigone (Fig. 1B), as well as the entire length of the olfactory tract (Figs. 1 C–C₄ and 2 A and B). Clusters of four to five secretagogin⁺ cells were often observed. This cytoarchitectural arrangement of secretagogin⁺ cells did not resemble the dense continuum of chain-migrating neuroblasts in the rodent or primate RMS (4, 19). Secretagogin⁺ cells exhibited bipolar morphology (Fig. 1C) with a process emanating at the cell surface facing the olfactory bulb. Although we cannot exclude the occasional pruning of peripheral dendrites during histochemistry, this forward-facing process typically ramified into two or three branches of equivalent length (Fig. 1C₁). Nonbipolar phenotypes were only observed in the proximal part of the olfactory tract (Fig. 2B). Secretagogin filled the entire cytoplasm, including processes of >100 μm in length. Because transversal sections of the olfactory striae showed uninterrupted secretagogin immunoreactivity (Fig. 1B), we propose that, besides aligning longitudinally along the opposite margins of the olfactory tract (Fig. 1 B and C), secretagogin⁺ cells populate an onion skin-like shell proximal to the superficial (outer) surface of the olfactory tract (Fig. 2A). Secretagogin⁺ cells invariably populated the outer region of the olfactory tract, irrespective of the lifespan (ranging from 60 to 95 y) of the cases studied (Fig. 1D). Although a bipolar neuronal phenotype in the olfactory system could allude to migratory behaviors, an equally appealing hypothesis is that intercalated mitral cell axons force the morphological adaptation of secretagogin⁺ cells. The latter notion is supported by the association of Ca²⁺-binding protein-containing bipolar neurons to axonal tracts in the corpus callosum (20) and anterior medullary velum (21).

Fig. 2.

Secretagogin⁺ neurons of the olfactory tract synaptically integrate into the olfactory circuitry. (A) Secretagogin⁺ bipolar shell cells populate the margins of the human olfactory tract. (B) Secretagogin⁺ cells are predominantly bipolar, except a subset of multipolar calretinin (CR)⁻/PSA-NCAM⁻ shell cells at locations proximal to the trigone. Somatodendritic afferents are in green. (C) The percentage of secretagogin⁺/CR⁺ and secretagogin⁺/PSA-NCAM⁺ neurons declines toward the distal olfactory tract. **P < 0.01. (D and D₁) Dendritic (arrowheads), but not somatic MAP-2, immunoreactivity was seen in secretagogin⁺ neurons. (E–H) Synaptobrevin⁺ (E–E₂), synaptophysin⁺ (F and F₁), or GAD65/67⁺ (G and G₁) profiles (arrowheads) on secretagogin⁺ somata and dendrites. (H–K) Similar to the human olfactory tract, SNAP25⁺ (H), synaptobrevin-2⁺ (I), VGLUT1⁺ (J and J₁), or GAD65/67⁺ (K–K₂) boutons contact secretagogin⁺ somata or processes in the mouse lemur olfactory tract. Note that postsynaptic GABA_ARα1 subunits (31) appose GAD65/67⁺ afferents (K₂). (L) Markers of cellular identity and afferent synapses on secretagogin⁺ neurons (n.d., nondetectable due to e.g., epitope mismatch). [Scale bars: 10 μm (B), 3 μm (D, E₂, F, G, and G₁), 2 μm (D₁, F₁, H, I, I₁, J, K, K₁, and K₂), and 1 μm (E and E₁).]

Secretagogin⁺ Bipolar Cells Receive Synaptic Input.

Secretagogin⁺ cells must receive synaptic inputs if they are integrated in the olfactory circuitry. Typically, relay cells receive dense afferentation on their somatodendritic compartments (22). We found secretagogin⁺ cells expressing MAP-2, a somatodendritic marker of neurons (Fig. 2 D and D₁), and being apposed by presynapses (Fig. 2B) immunoreactive for synaptobrevin-2 (Fig. 2 E–E₂), synaptophysin (Fig. 2 F and F₁), or glutamic acid decarboxylase [65/67-kDa isoforms (GAD65/67); Fig. 2 G and G₁].

Next, we tested whether secretagogin⁺ neurons were restricted to the human olfactory tract or might have also been present in an analogous structure of primates by studying the gray mouse lemur (Fig. S1 A and B). We identified secretagogin⁺ bipolar neurons along the shaft of the lemur's olfactory tract (Fig. S1B), in which the neurochemical attributes and synaptic innervation patterns (Fig. 2 H and I and Fig. S1 D–F) were reminiscent of those of human secretagogin⁺ neurons. Because shell cells received both excitatory [vesicular glutamate transporter 1⁺ (VGLUT1⁺); Fig. 2 J and J₁) and inhibitory (GAD65/67⁺; Fig. 2 K–K₂) afferents and expressed corresponding postsynaptic receptor subunits (GABA_ARα1; Fig. 2K₂), we suggest that these cells can be synaptically wired into the olfactory circuitry.

Secretagogin⁺ Cells Express Neuronal Plasticity Markers.

The scarce innervation of secretagogin⁺ cells together with their lack of mature neuronal markers [NeuN (23) or system A amino acid transporter 2 (SAT2), preferentially labeling excitatory neurons (24); Fig. 2L and Fig. S2, but see Fig. 5A] prompted us to test whether secretagogin⁺ cells represent a cell cohort that had retained immature characteristics (4, 7, 25).

Fig. 5.

Secretagogin is retained in the olfactory bulb in AD. (A) Secretagogin⁺ cells, likely interneurons (arrowheads), intermingled with SAT2⁺ mitral cells in the olfactory bulb. (B) Secretagogin and CR often coexist in periglomerular cells (arrowheads). (C–C₄) Quantitative assessment of the total (C₁) and secretagogin⁺ cell number (C₂) in olfactory glomeruli (C, Braak VI) in AD (n.s., nonsignificant). Note that the percentage number of secretagogin⁺ cells per glomerulus fails to correlate with either age (C₃) or Alzheimer's stage (C₄). (D) Secretagogin⁺ periglomerular neurons do not contain hyperphosphorylated tau (open arrowheads). (E and E₁) Complementary distribution of β-amyloid (arrowheads) and secretagogin. (F) The probability of secretagogin's colocalization with other Ca²⁺ binding proteins. (G–G₂) Unaltered distribution of secretagogin⁺ neurons in the olfactory bulb despite robust amyloid plaque load (G₂) in APdE9 mice. GL, glomerular layer; EPL, external plexiform layer; ML, mitral layer; IPL, inner plexiform layer; GRL, granular layer. [Scale bars: 300 μm (E₁), 140 μm (G₁), 50 μm (A–C), 30 μm (D), and 25 μm (G₂).]

Doublecortin immunoreactivity could not be detected unambiguously in the adult human olfactory system. Where noticeable, as in primates (16), secretagogin⁺ neuroblasts lacked doublecortin expression. We found unexpectedly high levels of polysialylated–neural cell adhesion molecule (PSA-NCAM), a widely accepted neuronal and structural plasticity marker (8, 25), in the olfactory tract (Fig. 3A) and bulb (Fig. S3 A and A₁) of aged humans. We demonstrated that PSA-NCAM⁺ shell cells frequently coexpressed secretagogin (Fig. 3 A–A₄). Secretagogin coexisted with calretinin (CR), another neuron-specific Ca²⁺ binding protein (13, 26), in shell cells, particularly in superficial cell assemblies near the outer margin of the olfactory tract (Fig. 3 B and C). We observed a gradual decrease in the density of PSA-NCAM⁺/secretagogin⁺ or CR⁺/secretagogin⁺ cells, uniformly bipolar in morphology (Fig. 2B), in the olfactory tract, which had tailed off toward the olfactory bulb (Fig. 2C). Irrespective of the age of the study subjects, secretagogin⁺ neurons in the olfactory tract exhibited low to moderate levels of β-III-tubulin (TUJ1) immunoreactivity (Fig. 3 D1 and D₂). The lack of glial fibrillary acidic protein (GFAP) expression (Fig. 3 E–E₂) in secretagogin⁺ cells in the human olfactory system reinforces our hypothesis that this cell population is of neuronal origin. The presence of PSA-NCAM⁺/secretagogin⁺ neurons in the adult primate olfactory tract (Fig. S1C) supports the notion that PSA-NCAM expression is unlikely to represent a cellular response to injury or neurodegeneration. Overall, PSA-NCAM, CR, and TUJ1 expression by secretagogin⁺ cells suggests that these may retain some characteristics of immature neurons but are synaptically connected in the olfactory system.

Fig. 3.

Secretagogin identifies bipolar cells with a makeup of immature neurons. (A) PSA-NCAM enwraps secretagogin⁺ bipolar cells (note that secretagogin is cytosolic). Open rectangles pinpoint the positions of A₁–A₃Insets and a terminal specialization, likely growth cone (A₄, open arrowheads). (B) CR⁺ and secretagogin⁺ processes run in parallel. (C) Secretagogin colocalizes with CR (arrowheads). (D₁ and D₂) β-III-tubulin coexists in secretagogin⁺ shell cells (arrowheads). Arrows in B–D and D₁ point toward the olfactory bulb. (E–E₂) Secretagogin⁺ neurons lack GFAP (arrowhead points to glial cell) immunoreactivity. [Scale bars: 20 μm (B and C), 10 μm (A and D₂), 7 μm (E₁), 5 μm (A₂), and 2 μm (A₄).]

Secretagogin Labels Periglomerular Cells in the Human Olfactory Bulb.

Recently, we identified secretagogin as a marker of periglomerular and deep-layer olfactory interneurons in the mouse and monkey (16). Therefore, we hypothesized that secretagogin may be expressed along the entire human olfactory axis. Indeed, a dense plexus of secretagogin⁺ neurons was seen throughout the human olfactory bulb (Fig. S3A). Secretagogin⁺ interneurons were often encountered in plexiform, granular, and, to a lesser extent, mitral cell layers (Fig. 5 A–C and Fig. S3B).

Loss of Secretagogin Expression in the Olfactory Tract in AD.

The temporally precise and coordinated activity of periglomerular interneurons participates in the continuous refinement of olfactory inputs (5, 19). One of the early clinical signs of AD is perturbed olfactory processing (27). Therefore, we asked whether secretagogin⁺ olfactory neurons might be affected or lost under conditions of AD pathology. We used a patient cohort of 16 AD cases and 4 age-matched controls (Table S1). The density of AT8⁺ hyperphosphorylated tau-bearing olfactory neurons exhibited a significant positive correlation with advancing Braak stages (ρ = 0.876, P < 0.01; Table S1).

Secretagogin⁺ cells contained hyperphosphorylated tau in the olfactory tract (Fig. 4 A–A₁″). Their number significantly declined with age (Fig. 4B). The loss of shell cells became accentuated upon binning our patient material as per AD staging (P < 0.05 vs. control; Fig. 4B₁). In contrast, the pattern of secretagogin immunoreactivity in the perikarya or dendrites in the periglomerular layer of the olfactory bulb (Fig. 5 B and C) did not show significant changes in AD cases relative to controls (Fig. 5C). Both the total number of cells (Fig. 5C₁) and the percentage of secretagogin⁺ neurons in individual olfactory glomeruli (Fig. 5C₂) remained unperturbed with no correlation to age (Fig. 5C₃) or to one another (Fig. 5C₄). These data suggest that secretagogin⁺ neurons in the olfactory tract focally and selectively succumb to neurofibrillary pathology in AD.

Fig. 4.

Secretagogin⁺ neurons in the olfactory tract are lost in AD. (A–A₁″) Secretagogin⁺ neurons (open arrowheads) can exhibit neurofibrillary tangle pathology (AT8⁺) (filled arrowheads) in the human olfactory tract. (B and B₁) Age (R² = 0.62, P < 0.01) and Alzheimer's stage-related decline of secretagogin⁺ shell cells in the olfactory tract. *P < 0.05. [Scale bars: 20 μm (A) and 10 μm (A₁″).]

Next, we tested whether secretagogin⁺ neurons were affected by AD-related cytoskeletal modifications. Although dendritic fields within olfactory glomeruli exhibited AT8 immunoreactivity (likely corresponding to the primary dendritic tuft of mitral cells and/or primary olfactory afferents), secretagogin⁺ periglomerular interneurons did not accumulate hyperphosphorylated tau (Fig. 5D). The periglomerular layer was devoid of extracellular Aβ plaques in all cases (27) (Fig. 5 E and E₁), excluding the study of a direct relationship between Aβ toxicity and secretagogin expression in the human olfactory system.

The olfactory bulb harbors neurochemically distinct subsets of interneurons (1, 17). The expression of CR, parvalbumin, and calbindin-D28k by olfactory interneurons is used to subclassify these cells (17). Rodent and primate studies revealed secretagogin's propensity to coexist with CR, but less so with parvalbumin or calbindin-D28k (16). Here, we assessed whether secretagogin⁺ interneurons retained their ability to coexpress CR in AD. By determining the probability of colocalization of these Ca²⁺ binding proteins on sufficiently large and randomized cell populations in the glomerular, external plexiform, and granular cell layers (Fig. 5F and Fig. S3B), we found a transient (albeit nonsignificant) increase in secretagogin/CR colocalization in moderate AD. The density of secretagogin⁺/CR⁺ interneurons remained unchanged in severe AD relative to controls (Fig. 5F and Fig. S3C). Parvalbumin⁺ interneurons were devoid of secretagogin expression in either AD or control cases. We conclude that the identity of secretagogin⁺ cells and their laminar distribution pattern in the olfactory bulb remain unaffected in AD. In view of the lack of Aβ deposits in the human olfactory glomerular layer, the analysis of a transgenic mouse model presenting Aβ-induced olfactory impairments (28) is warranted.

Secretagogin Expression in Mice with AD-Like Pathology.

APdE9 mice are characterized by robust and progressive amyloid plaque deposition in the olfactory bulb (Fig. S4 A, B, and B₁). Because mice do not have an equivalent to the primate olfactory tract, we focused on Aβ-laden forebrain territories containing undifferentiated secretagogin⁺ neuroblasts and/or terminally differentiated neurons to test whether Aβ can induce the loss of secretagogin⁺ olfactory neurons.

We show that transgene-driven Aβ accumulation in APdE9 mice failed to impact the laminar distribution or cellular integrity of secretagogin⁺ neurons in the mouse olfactory bulb (Fig. 5 G–G₂) or indusium griseum (Fig. S4 D₁ and D₁'). High-resolution confocal microscopy revealed chain-migrating secretagogin⁺ neuroblasts in the RMS, populating the olfactory bulb in both APdE9 mice and wild-type controls (Fig. S4 C₁ and D₂). The density of secretagogin⁺ neuroblasts entering the olfactory bulb did not differ between APdE9 mice and wild-type littermates [8.4 ± 1.3 (APdE9) vs. 6.6 ± 0.6 (wild-type) cells per 10⁴ μm²; P > 0.1]. Despite the severe Aβ burden, neither the morphological phenotype nor the packing density of secretagogin⁺ periglomerular neurons was qualitatively different in either the accessory (Fig. S4 C₃ and D₃) or the main olfactory bulb (Fig. S4 C₂ and D₄). Our experimental data suggest that secretagogin expression is unaffected by Aβ, because secretagogin⁺ neuroblasts and neurons remain unaltered both in their contingents and phenotypes, recapitulating histopathological findings in AD. These data raise the possibility that a phylogenetically segregated subset of secretagogin⁺ neurons is selectively lost during AD, implicating a spatially-confined cellular locus for olfactory impairment in aged humans.

Conclusions

Understanding the neuronal substrates of olfaction, particularly the neuronal subtypes conferring critical network modalities, received limited attention in humans because of the lack of neurochemical cell identity markers allowing the precise mapping of spatial and temporal modifications to the olfactory circuitry. The continued identification of molecularly and morphologically diverse neuronal subtypes in the human olfactory circuitry, like in rodents (29), may be rewarding because it can uncover striking phylogenetic differences (30). Here, we show the existence of spatially segregated neurons expressing secretagogin, as well as a battery of markers nonconventional for differentiated neurons that had integrated into a synaptic circuitry. These shell cells are unique to primates and humans, emphasizing interspecies differences in the cytoarchitectural organization of the olfactory circuitry. Our hypothesis is that secretagogin⁺ shell cells did not reach their final destination in the olfactory bulb during development, yet synaptically integrated to refine some aspects of olfactory information processing. Their atypical localization in the olfactory tract may render secretagogin⁺ neurons sensitive to noxious insults during AD, giving rise to a spatially restricted primary locus triggering olfactory impairment as AD progresses. Although secretagogin's cellular function(s) remain unknown, an appealing conjecture is that this Ca²⁺ binding protein can contribute to shaping synaptic responsiveness as Ca²⁺ sensor through intermolecular interactions (15).

Materials and Methods

Tissues and Histochemistry.

Olfactory bulb and adjoining tract tissues from AD patients and age-matched controls (without clinical signs of neuropsychiatric disease; n = 20 cases in total, both sexes) were assigned to this study (Table S1). Experimental protocols on human, primate, and mouse specimens were approved by local authorities (SI Materials and Methods).

Immunohistochemistry.

Rabbit anti-secretagogin antibodies were generated as described (16). Multiple immunofluorescence histochemistry with mixtures of primary antibodies (Table S2) and Sudan Black B counterstaining to quench tissue autofluorescence were according to published protocols (16) as described in SI Materials and Methods.

Imaging.

Images of sections processed for chromogenic detection of secretagogin or Aβ were captured by using a Nikon Eclipse 90i microscope equipped with a motorized research microscope system and digital camera head (DS-Qi1Mc; Nikon) and analyzed by using Nikon NIS Elements imaging software. Procedural details of the morphometric analysis are referred to in SI Materials and Methods. Single x–y plane images were captured by laser-scanning microscopy (710LSM; Zeiss).