Diosgenin promotes axonal growth in 5XFAD mouse
We assessed whether diosgenin has axonal growth effect in vivo using cortex-axotomized mice. No neurofilament-H (NF-H)-positive staining was detected in the lesion area at 1 h following axotomy (Supplementary Fig. 1A). To avoid overlaid estimation of neuroprotection and axonal growth, diosgenin administration started from 7 days after axotomy. Since administration of diosgenin for 14–20 days recovered memory deficits in 5XFAD mice [10, 14], diosgenin or a vehicle solution was administered for 15 days. Axonal density in cortical layers I–IV was quantified in the lesion area. Diosgenin administration significantly promoted axonal growth (Supplementary Fig. 1B), with no significant change in the lesion area size (Supplementary Fig. 1C).
Next, we focused on a long axonal tract related to memory formation, from the HPC to the PFC [16, 17] in 5XFAD mice. Since neural circuit from the dorsal HPC, but not the ventral HPC, to the PFC plays a role in short-term working memory , axon-growing neurons were visualized by sequentially injecting two colors of retrograde Dextran tracers into the PFC to determine whether diosgenin administration promotes axonal growth from the dorsal HPC in 5XFAD mouse (Fig. 1A–C). Seven days prior to initiating drug administration, the originally projected axons from the HPC to the PFC were labeled with Dextran Texas Red. After 14-day diosgenin administration, Dextran fluorescein isothiocyanate (FITC), was injected in exactly the same PFC position of the first tracer; thus, naive neurons were labeled with two tracers (merged as yellow), and axon-growing neurons were labeled with Dextran FITC only. Complete match of the injected positions for the two tracers was confirmed for all analyzed mice (Supplementary Fig. 2A). When these tracers were sequentially injected into the PFC, they spread similarly to the HPC (Supplementary Fig. 2B), confirming the accuracy of the injecting techniques and analogous spreading ranges of two tracers. The dentate gyrus was excluded from the analyses to exclude neurogenesis-induced increase of axonal projection . Tracer-positive neuronal cell bodies were determined by neuronal nuclei (NeuN)- and 4’,6-diamidino-2-phenylindole (DAPI)- copositive staining. The number of mono-positive neurons for the second tracer Dextran FITC in the CA1 and CA3 regions were significantly increased by diosgenin administration (Fig. 1D, F), showing that diosgenin promoted axonal growth in 5XFAD mice. Interestingly, axon-degenerating neurons (Texas Red+, FITC–) was significantly increased in 5XFAD compared with wild-type mice; however, axonal degeneration was significantly lower in the CA1 (Fig. 1E) and tended to be lower in the CA3 (Fig. 1G) by diosgenin administration. The number of naive neurons (Texas Red+, FITC+) was tended to be increased by diosgenin administration in 5XFAD mice, indicating the axon-protective effect of diosgenin (Supplementary Fig. 3A, 4A). The originally projected axons (total Texas Red+) in vehicle- and diosgenin-administered 5XFAD mice did not differ, but it was significantly lower than that of wild-type mice, showing that the same degree of axonal retraction in this neural circuit had occurred in 5XFAD mice before drug administrations (Supplementary Figs. 3B and 4B). The number of total FITC+ neurons was significantly decreased in vehicle-treated 5XFAD compared with wild-type mice, whereas it was significantly increased by diosgenin, suggesting that the axonal density in this circuit was recovered to equal level as wild-type in diosgenin-treated 5XFAD mice. The number of total NeuN+ neurons and DAPI+ cells in the CA1 and CA3 did not differ among the groups (Supplementary Figs. 3D, E and 4D, E). High magnification images of the CA1 and CA3 were shown in Supplementary Figs. 3 and 4.
In vivo transcriptome to assess gene expression profile of axon-growing neurons
To investigate the gene expression profile of axon-growing neurons in vivo, laser capture microdissection (LCM) was performed to collect individually naive and axon-growing neurons from the CA1 and CA3 of vehicle- (n = 3) or diosgenin-treated (n = 3) 5XFAD mice (Fig. 2A). Due to scanty neurons for each mouse, neurons from three mice were combined for DNA microarray. The total number of collected neurons was 660 naive neurons and 720 axon-regenerated neurons from brain sections of three mice.
Genes with stable expression levels (log2 > 8) were analyzed using transcriptome analysis console (Thermo Scientific). Hierarchical clustering and scatter plot analysis revealed a striking difference in the expression patterns (fold change > 5 or < −5) between the two neuron types (Fig. 2B, C). In total, 67 genes in the axon-growing neurons were 5-fold up- or downregulated compared with naive neurons (Supplementary Table 1). The database of all the genes is shown in Supplementary Table 2. We focused on the gene with the highest expression level in axon-growing neurons, secreted protein acidic and rich in cysteine (SPARC). The SPARC protein level was confirmed using HPC neuron lysates. Diosgenin treatment for 4 days significantly enhanced the expression of SPARC in cultured HPC neurons (Fig. 2D, Supplementary Fig. 5). In addition, expression level of SPARC in CA1 neurons was significantly lower in 5XFAD than wild-type mice, whereas it was significantly increased by 14-day-diosgenin administration in 5XFAD mice (Fig. 2E, F).
To investigate whether SPARC upregulation is necessary for diosgenin-induced axonal growth, a SPARC knockdown experiment was performed using siRNA transfection. To distinguish the siRNA-transfected neurons, GFP vectors were mixed together with siRNAs, and the expression of SPARC and axonal length were measured only in GFP+ neurons. Transfection of 30 nM siRNA for SPARC (siSPARC) for 3 days significantly reduced SPARC expression without affecting axonal growth in HPC neurons (Supplementary Fig. 6A–C). Three days after siRNA transfection, neurons were treated with vehicle or diosgenin for another 4 days (Fig. 2G). As results, upregulation of SPARC mediated by diosgenin was significantly inhibited in siSPARC-transfected neurons (Fig. 2H). In addition, diosgenin-induced axonal growth was completely diminished by SPARC knockdown (Fig. 2I), indicating that SPARC is an essential molecule for diosgenin-induced axonal growth.
SPARC overexpression recovers memory deficits and promotes axonal growth in 5XFAD mice
To investigate a direct contribution of SPARC upregulation on axonal growth and memory recovery, the SPARC gene was overexpressed neuron-specifically by adeno-associated virus serotype 9 (AAV9) delivery. AAV9 were engineered to express Cerulean-tagged SPARC, or Cerulean control under a Syn1 promotor. At first, SPARC overexpression was induced in vitro to test axonal growth in cultured HPC neurons. Seven days after AAV-SPARC treatment, SPARC expression was significantly increased at doses of more than 5 × 106 GC/µl (Fig. 3A, B). At this time, SPARC overexpression (5 × 106 GC/µl AAV-SPARC treatment) significantly increased the axonal length in neurons (Fig. 3C, D).
Next, we investigated whether SPARC overexpression in HPC neurons contributes to memory recovery in 5XFAD mice. At 14, 21, and 28 days post-AAV-SPARC injection to the HPC, SPARC expression was significantly increased in NeuN-positive neurons compared with that in AAV-control-treated mice (Supplementary Fig. 7A–C). Wild-type or 5XFAD mice were injected with AAVs in the CA1 region, and memory tests were performed from 21 days later. In object recognition memory test (Fig. 3E), all mice showed similar exploratory behaviors toward the two identical objects A1 and A2 during training. In the test session, AAV-SPARC-injected 5XFAD mice showed a significantly higher preferential index to the novel object B1 while AAV-control-injected 5XFAD mice did not. To test spatial memory, an object location test was performed 23 days after AAV injection (Fig. 3F). SPARC-overexpressed 5XFAD mice showed a significant improvement in the object location memory compared with AAV-control-injected 5XFAD mice. The total distance traveled (cm), turn angle (degrees), and immobility time (s) were not significantly different between the groups in a locomotion test on day 24 (Supplementary Fig. 8A–C). No significant changes in body weight were observed in any groups (Supplementary Fig. 8D). Taken together, SPARC overexpression in the HPC neurons recovers memory impairment in 5XFAD mice. Twenty-five days after AAV injection, these mice were served to confirm SPARC expression. Similar with Fig. 2E, expression level of SPARC in HPC neurons was significantly lower in 5XFAD than wild-type mice; however, it was significantly increased by AAV-SPARC injection in 5XFAD mice (Supplementary Fig. 8E, F), showing that SPARC overexpression in the HPC was maintained during a series of behavioral tests. Furthermore, synapse formation from the HPC to the PFC were evaluated in these mice (Fig. 3G). Since axons of HPC neurons would be labeled with AAV-derived Cerulean, colocalization of Cerulean, synaptophysin+ pre-synapse, and PSD95+ post-synapse on NeuN+ neurons (blue dotted line) in the PFC were calculated. As a result, synapse formation was significantly decreased between HPC neuron’s pre-synapse and PFC neuron’s post-synapse in 5XFAD compared with wild-type mice; however, SPARC overexpression significantly increased synapse formation in this circuit (Fig. 3H).
Next, we investigated whether overexpression of SPARC in the HPC neurons promotes axonal growth in 5XFAD brains using similar methods as Fig. 1. Dextran Texas Red was injected into the PFC and AAVs were injected into the CA1 7 days later. Twenty-one days after AAV injections, Dextran FITC was injected into the PFC region (Fig. 3I, J). Axonal projection from the CA1 to the PFC was significantly increased in AAV-SPARC-injected 5XFAD mouse brains (Fig. 3K). The number of axon-degenerating neurons (Texas Red+, FITC–) was significantly rescued by SPARC overexpression (Fig. 3L). The number of naive neurons (Texas Red+, FITC+) was also increased by SPARC overexpression in 5XFAD mice (Fig. 3M). The originally projected axons (total Texas Red+) in AAV-control and AAV-SPARC-injected 5XFAD mice were not different, showing that the same degree of axonal degeneration had occurred in these 5XFAD mice before AAV injections (Fig. 3N). The number of total FITC+ neurons was significantly increased by SPARC overexpression with equal level as wild-type mice (Fig. 3O). The number of total NeuN+ neurons did not differ among the groups (Fig. 3P). A significant increase in axonal growth and decrease in axonal degeneration from the CA3 to the PFC were observed by SPARC overexpression in 5XFAD mice compared with AAV-control-injected mice (Supplementary Fig. 9). These data suggested that SPARC overexpression in the HPC neurons enhanced long-distance axonal projection in 5XFAD mouse.
SPARC overexpression-induced axonal growth contributes to memory recovery in 5XFAD mice
To confirm direct contributions of SPARC-driven axonal growth on memory recovery, DREADDs  was used to silence neural activity of SPARC-overexpressed axons projected from the HPC to the PFC. AAV9 was engineered to express Cerulean-tagged SPARC with inhibitory muscarinic receptor variant hM4Di (AAV-SPARC-hM4Di) under a Syn1 promotor to overexpress both SPARC and hM4Di in neurons. hM4Di-expressing neurons respond to clozapine-N-oxide (CNO) to activate Gi signaling. Therefore, AAVs were injected into the HPC, and CNO was microinjected through a cannula inserted in the PFC.
At 21 and 28 days post-AAV-SPARC-hM4Di injection to the CA1, SPARC expression was significantly increased in NeuN-positive neurons (Supplementary Fig. 10). Additionally, AAV-derived hM4Di (Cerulean) was observed in the PFC at the location where the axons of HPC neurons terminate (Supplementary Fig. 10). A cannula for the CNO infusion was inserted in the PFC (Fig. 3Q). Object recognition memory tests were performed 21 and 23 days after AAV injection immediately after microinjection of saline or 1 mM CNO, which is the reported dose with no abnormal behaviors by itself . Microinjection of saline in the PFC did not interfere with memory retention in wild-type mice and AAV-SPARC-hM4Di-injected 5XFAD mice (Fig. 3R). Microinjection of CNO had no influence on the memory function of hM4Di-absent AAV-Cont-empty-injected wild-type mice. However, CNO injection impaired object recognition memory in AAV-Cont-hM4Di-injected wild-type mice, showing that the HPC-PFC circuit is indispensable to memory formation. Importantly, when CNO was microinjected into the PFC of AAV-SPARC-hM4Di-injected 5XFAD mice, SPARC overexpression-induced memory recovery was completely diminished. These results indicate that SPARC overexpression-induced axonal growth directly contributes to memory recovery in 5XFAD mice.
The results of the locomotion test on day 28 showed that total distance traveled (cm), turn angle (degrees), and immobility time (s) were not significantly different between the groups (Supplementary Fig. 11A–C). The cannula was located in identical positions within the PFC (Supplementary Fig. 11D). SPARC overexpression was maintained in the CA1 neurons of AAV-SPARC-hM4Di-injected 5XFAD mice, and hM4Di was observed in the axons of HPC neurons terminating to the PFC in all groups (Supplementary Fig. 11E, F).
SPARC expression is increased on the axonal area
We investigated the intracellular localizations of SPARC in diosgenin-treated neurons compared with normal control and Aβ25–35-treated neurons. HPC neurons were treated with Aβ25–35 for 3 days, and diosgenin or vehicle solution was applied for 4 days (Fig. 4A). Immunostaining detected intrinsic expression of SPARC in neuronal bodies and axonal shafts in the control neurons, whereas Aβ treatment significantly decreased the SPARC expression, particularly in axons (Fig. 4B). However, diosgenin treatment significantly increased the expression of SPARC on axons compared with Aβ25–35/vehicle groups. Since we have previously identified 1,25D3-MARRS as a direct binding receptor of diosgenin to induce axonal regrowth and memory recovery [9, 10, 14], we investigated whether SPARC upregulation on axons was mediated by 1,25D3-MARRS signaling. Three days after culturing HPC neurons, Aβ25–35 was treated for 3 days followed by 1,25D3-MARRS neutralizing antibody or control IgG and vehicle or diosgenin treatments for 4 days. As results, diosgenin-induced axonal regrowth and SPARC upregulation were diminished by 1,25D3-MARRS neutralizing antibody treatment (Supplementary Fig. 12A, B), suggesting that that diosgenin-induced upregulation of SPARC on axons was mediated by 1,25D3-MARRS signaling.
The expression of SPARC on axons from the HPC to the PFC was investigated by anterograde tracing of biotinylated dextran amines (BDA). Vehicle solution or diosgenin was administered daily for 21 days, and on administration day 14, BDA was injected in the CA1 (Fig. 4C). Injected BDA colocalized with hypophosphorylated NF-H (pNF-H)-positive axons in the PFC (Supplementary Fig. 13). The quantitative values of BDA-positive axons in the PFC were significantly reduced in vehicle-treated 5XFAD mice compared with wild-type mice (Fig. 4D). However, diosgenin administration significantly increased axonal projection to the PFC. In the PFC of wild-type mice, high expression of SPARC on BDA-positive axons was observed (Fig. 4C). However, the number of SPARC-expressing axons in the 5XFAD PFC was significantly decreased (Fig. 4E), and a large part of BDA-positive axons did not express SPARC (Fig. 4C, white arrowheads). When 5XFAD mice were treated with diosgenin, SPARC-expressing BDA-positive axons significantly increased to the equal level as that in wild-type mice. These data indicated that upregulation of SPARC was significant in axonal terminal area by diosgenin treatment.
Elevated axonal SPARC levels induce accurate pathfinding of injured axons
Based on evidence showing high expression of SPARC in regrowing axons, we speculated that accurate pathfinding of injured axons brought by SPARC elevation may require an interacting extracellular counterpart molecule for arrival at the terminating site. SPARC consists of three domains, namely the acidic, follistatin-like, and extracellular domains , and is expressed in the cytoplasm , extracellular matrix [24, 25], and plasma membrane . Regarding SPARC located on axonal membranes, SPARC possibly interact with extracellular molecules via its extracellular domain. One of the main ligands of the extracellular domain of SPARC is the collagens;  collagen types I, III, and IV were reported. The collagen family promotes axonal growth, axonal guidance, and synapse formation . However, the effect of the interaction of collagen with SPARC on axonal formation has never been studied. Since the expression of collagen type I (collagen I) is high in the brain, SPARC-collagen I interaction on axonal regeneration was investigated.
Short-term culture of HPC neurons could rarely induce the expression of collagen I, whereas its expression was increased in long-term-cultured neurons (Supplementary Fig. 14A, B). Therefore, HPC neurons were cultured for 14 days and treated with Aβ25–35 for 3 days followed by 4-day treatment with diosgenin. In control neurons, collagen I was distributed in a fiber-like shape and colocalized with axons (Fig. 4F). However, some collagen I existed like axonal wreckages and did not colocalize with axons in Aβ25–35-treated neurons (blue arrowheads). Since collagen I is produced and secreted from neurons [29, 30], we speculated that this neuron-derived collagen I remained in the extracellular space even after axonal disruption. Quantitative graphs showed that length of pNF-H+ axons that colocalize with collagen I-positive fibers was significantly decreased by Aβ25–35 treatment compared with control neurons; however, colocalization of axons with collagen I-positive fiber significantly increased by diosgenin (Fig. 4G). SPARC signal was not detected when only secondary antibody for SPARC were used for immunostaining (Supplementary Fig. 14C). In addition, a negative control peptide for Aβ25-35, Aβ35-25, did not influence on levels of colocalization of axons and collagen I (Supplementary Fig. 14D) and expression of SPARC on axons (Supplementary Fig. 14E). A significant colocalization of axons with collagen I was also detected in the PFC of wild-type mice (Fig. 4H); however, collagen I-positive but pNF-H-negative dots were increased in 5XFAD mice (white arrows). Furthermore, SPARC on axonal membranes and its colocalization with extracellular collagen I were confirmed by non-permeable immunocytochemistry (Fig. 4I). SPARC signals were detected on axonal membranes (yellow arrowheads) and colocalized with extracellular collagen I in control neurons. In Aβ25-35-treated neurons, the expression of SPARC on axonal membranes was drastically decreased, and extracellular collagen I alone remained (blue arrowheads). However, diosgenin treatment increased the expression of SPARC on axonal membranes, colocalizing with collagen I (yellow arrowheads). These results led us to hypothesize that even if axons were degenerated by Aβ, extracellular collagen I still be located on their original tracks; therefore, direction-specific axonal regeneration may occur when elevated levels of SPARC on axonal membranes interact with extracellular collagen I.
To test this hypothesis, the effect of SPARC-collagen I interaction on axonal growth was investigated. Prior to culturing the neurons, collagen I was overlaid on poly-D-lysine (PDL)-coated dishes, and HPC neurons were then treated with AAV-control or AAV-SPARC 4 h after culturing (Fig. 4J). Axonal length was significantly increased by 7-day-SPARC overexpression in the PDL-coated dish. Interestingly, the SPARC-driven axonal growth was significantly increased further by collagen I coating. These data indicated that the SPARC-collagen I interaction promotes axonal growth.
To investigate whether extracellular collagen I relates to accurate pathfinding of axons in SPARC-overexpressed neurons, collagen I was coated in direction-limited (Fig. 4K). Collagen I was overlaid on the right side of the dish, and HPC neurons were seeded on a collagen I-free space using device chambers. After removal of device chambers, AAVs were treated for 14 days. Axons growing onto the PDL- and collagen I-coated regions were quantified. AAV-control-treated neurons showed similar axonal growth preference to the PDL and collagen I. Meanwhile, axonal growth of SPARC-overexpressed neurons showed a directivity toward collagen I.
Furthermore, we tested whether interaction of SPARC on axonal membranes and extracellular collagen I actually promoted directional regrowth of axons (Fig. 4L). HPC neurons were seeded on the soma space (gray) of a triple chamber neuron device and treated with 5 × 107 GC/µl AAV-control to visualize axons by Cerulean fluorescence in live imaging. Since only axons could extend into microgrooves of the device chambers, neuron-derived extracellular collagen I would remain in the microgrooves where axons originally elongated. Ten days after neuron culture, Cerulean+ axons were observed using a fluorescence microscope to confirm axonal extension into the microgrooves. Then, Aβ25–35 was treated to soma (gray) and axonal space (pink) for 3 days (11–13 DIV [days in vitro]). Live imaging in 13 DIV revealed that exactly same Cerulean+ axons that originally extended into the microgrooves were atrophied by Aβ25–35 treatment. After that, triple chamber neuron devices were removed from the bottom of the dishes, and 5 × 106 GC/µl AAV-control or AAV-SPARC were treated together with SPARC neutralizing antibody (SPARC-Ab) or control IgG (IgG). Since device chambers were removed, axons would freely to regrow everywhere. Axonal densities in the original microgroove space (axons pursued extracellular collagen I) or out of microgroove space (axons did not pursue extracellular collagen I) were quantified in 20 DIV. As results, SPARC overexpression significantly regrew axons to pursue collagen I in control IgG-treated group (Fig. 4M). On the other hand, axonal regrowth into collagen I direction was completely inhibited by SPARC-Ab treatment, suggesting that membranal SPARC was required for axons to regrow into their original tracks. Interestingly, as we expected, density of axons without pursuing collagen I (gray arrowheads) was significantly increased by SPARC-Ab treatment in SPARC-overexpressed neurons (Fig. 4N), which failed to induce direction-specific axonal regrowth. Extracellular collagen I was existed exactly along the axons ran in the microgroove area in all groups even after axonal atrophy (Fig. 4L). All images were captured in exact same region throughout all time points in each group. These data together demonstrated that elevated levels of SPARC on axonal membranes interact with the guidepost molecule, extracellular collagen I, resulting in accurate axonal regrowth.