All animals in this study were housed and treated in strict accordance with the recommended NIH guidelines for animal care and safety and were approved by the Ethics Committee for Animal Research at the Third Military Medical University (Army Medical University).

Lipopolysaccharide (LPS; Solarbio, L8880), IL-4 (Sino Biological, 11846-HNAE), TGF-β1 (Sino Biological, 10804-HNAC) and BrdU (Sigma-B5002) were dissolved in sterile water with 1,000× stock concentration. Y27632 (Peprotech, 1293823) and verteporfin (MedChem Express, CL318952) were dissolved in DMSO with 1,000× stock concentration. For regent treatments, corresponding vehicle was used as the control.

The following primary antibodies were used in immunocytochemical (ICC) staining and Western blotting (WB). S100 (rabbit, Sigma, SAB5500172; ICC, 1:800), GFAP (mouse, Abcam, ab10062; ICC, 1:600), p-MLC (rabbit, Abcam, ab157747; WB, 1:2,000), Ki67 (mouse, Abcam, ab8191; ICC, 1:600), BrdU (mouse, Abcam, ab8152; ICC, 1:500), Tuj1 (mouse, Abcam, ab78078; ICC, 1:600), Tau (mouse, Abcam, ab80579; ICC, 1:400), p75NTR (mouse, Abcam, ab3125; ICC, 1:400), YAP1 (rabbit, Abcam, ab39361; ICC, 1:600; WB: 1:2,000), YAP1 (phospho S127; rabbit, Abcam, ab76252; WB, 1:2,000), CTGF (mouse, Santa Cruz, sc-365970; WB, 1:3,000), integrin beta 5 (rabbit, Abcam, ab15459; WB, 1:1,000), FAK (rabbit, Abcam, ab40794; WB, 1:3,000), FAK (phospho Y397; rabbit, Abcam, ab81298; WB, 1:2,000), L1-CAM (mouse, Abcam, ab24345; WB, 1:1500), GAPDH (rabbit, Abcam, ab181602; WB, 1:3,000) and β-actin (mouse, Boster, BM0627; WB, 1:4,000). For ICC, Fluor-488- or 568-conjugated anti-mouse or rabbit immunoglobulin-G (IgG; Invitrogen, Carlsbad, CA, USA; A-11008, A32723, A-11004 and A-11011) diluted 1:600 in PBS were used. For WB, horseradish peroxidase (HRP)-conjugated secondary antibodies (Abcam, ab6789 and ab205718; 1:3,000) were used. Primary L1-CAM antibody (mouse, R&D Systems, MAB5674) was used for blocking L1-CAM functionally.

Primary OECs were prepared from the OB of adult Long Evans rats and purified by differential cell adhesiveness as described previously (Liu Y. et al., 2010; Li et al., 2017). Briefly, after meninges were removed from the OB, the olfactory nerve layer (ONL) was peeled away from the glomerular and deeper layers of the OB. Tissues were chopped and digested in 0.25% trypsin (Sigma) for 15 min. After ending digestion by DMEM/F-12 (Hyclone) containing 10% fetal bovine serum (FBS, Gibco), these chopped tissues were further mechanically dissociated into single cell suspension. The cell suspension was plated on uncoated 6 cm culture dishes (NEST) twice, each for 24 h. These non-adhesive cells were collected, spun down, resuspended in culture medium and seeded onto poly-D-lysine (PDL, 0.1 mg/mL, Sigma) pre-coated 6-well plates.

The morphological classification of OECs has been described previously (Huang et al., 2008). Briefly, process-bearing OECs were defined as those with two processes and a long fusiform bipolar morphology. Flattened OECs were defined as those with orientated processes and flat sheet-like morphology. Flattened OECs include two subtypes according to the location of the nucleus. Flattened type 1 OECs exhibit a fan-like shape with a nucleus lying at the edge of cytoplasm. Flattened type 2 OECs exhibit a round shape with a nucleus lying at the center of cytoplasm.

Purified OECs were seeded on glass slides and pre-treated with 10 μg/mL LPS, 10 ng/mL IL-4, 10 ng/mL TGF-β1 or 10μM Y27632 in DMEM/F12 containing 3% FBS for 24 h. Cortical neurons were cultured from embryonic day 18 (E18) Long Evans rats as previously described (Dietrich et al., 2003; Su et al., 2013). Briefly, brains were washed using ice-cold Hanks’ balanced salt solution (HBSS, Sigma). Cortical tissues were demembrated, chopped into small pieces, and incubated with 15 U/mL papain (Sigma) for 10 min. Then, cells were gently triturated using fire-polished Pasteur pipettes. Neurons were resuspended in DMEM/F12 containing 15% FBS after being centrifuged for 5 min at 300 g. After large fragments were allowed to settle for 10 min, neurons in supernatant were seeded onto OEC cultures. After 4 h, the culture medium was changed to reduce cell fragments. The co-cultures of cortical neurons and OECs were further cultured in DMEM/F12 containing 3% FBS for 3 days and subsequently examined.

Purified OECs were seeded onto PDL-coated 6-well plates at a low density (5,000 cells per well). The back of the 6-well plate was lined to make sure of the locations of OECs that were observed during the time-lapse imaging. The isolated cells with typical OEC morphology were selected for time-lapse imaging. The purified OECs were treated with 10 μg/mL LPS, 10 ng/mL IL-4 and 10 ng/mL TGF-β1 in DMEM/F12 containing 3% FBS. The morphological changes of OECs were recorded with a CCD camera attached to the inverted phase microscope (LEICA DM IRE2). Two shape factors were used to describe the morphological properties of OECs with ImageJ software. The average process length was the average of leading and trailing processes of OECs. The circularity was a function of the perimeter P and the area A:

The circularity of a circle is 1 and much less than 1 for a process-bearing morphology.

OECs on glass slides were fixed with 4% paraformaldehyde (PFA) and permeabilized with 0.3% TritonX-100. For BrdU staining, OECs were labeled with BrdU (10 μM) for 24 h, fixed in cold methanol and DNA was denatured with 2 mol/L HCl. After being blocked with 3% bovine serum albumin (BSA), OECs were incubated with primary antibodies at 4°C for 16 h. After washes with PBS, OECs were incubated with Fluor-488- or 568-conjugated anti-mouse or rabbit IgG (Invitrogen) diluted 1:600 in PBS at 37°C for 1 h. Cell nuclei were counterstained with DAPI (Beyotime). For F-actin staining, OECs were incubated with fluorescein-conjugated phalloidin (10 ng/mL; Sigma, P5282) for 30 min. Pictures were taken with a fluorescence microscope (BX51; Olympus) or confocal microscope (FV1000; Olympus), and analyzed with ImageJ software.

An in vitro scratch assay was used to measure OECs’ migrating ability as well as neurite outgrowth and extension (Liang et al., 2007; Vukovic et al., 2009). For the migrating ability assay, purified OECs were seeded in 12-well plates and cultured for 24 h. Scratches across the center of the wells were generated with sterile 1 mL pipette tips using a ruler as a guide. Detached cells were removed by several washes. OECs were treated with 10 μg/mL LPS, 10 ng/mL IL-4, or 10 ng/mL TGF-β1. The images from the same field of view were captured at 24 and 48 h after the initial scratch injury. The images of the initial scratches were also captured as the initial control, which was represented as dotted lines. The number of OECs that migrated into scratches per field (50×) was measured and averaged at each time point with ImageJ software. To examine neurite outgrowth and extension, neurons cultured on astrocytes were scratched and detached cells were removed by several washes. After OECs were seeded with or without Y27632, extended neurites across scratches were visualized by Tau staining.

The matrigel drop migration assay was performed according to an agarose drop assay that has been described previously (Ogier et al., 2006; Gueye et al., 2011). Briefly, purified OECs were resuspended in a matrigel basement membrane matrix (BD Bioscience). 0.5 μL matrigel drops containing OECs were seeded onto 6-well plates and placed at 37°C to allow the matrigel to solidify. DMEM/F12 was added to cover the solidified matrigel drop. Then, OECs were treated with 10 μg/mL LPS, 10 ng/mL IL-4, or 10 ng/mL TGF-β1 in DMEM/F12 containing 3% FBS. The migration of OECs was carefully observed and photographed through a LEICA DM IRE2 inverted phase microscope. The cell migration distance index and migration area index were analyzed using ImageJ software. The matrigel drop area and the area of migrating cells were measured and the migration area index was represented as the ratio of the migrating cells area to the matrigel drop area. The longest migration distances out of the matrigel drops in each quadrant were measured, and the migration distance index was represented as the ratio of average cell migration distance to the radius of the matrigel drop.

The trans-well migration assay was used to analyze the migratory response of purified OECs toward neurons (Su et al., 2009). The primary cortical neurons were seeded and cultured in 12-well plates. Purified OECs were counted and seeded into the upper chamber of trans-well inserts with a polyethylene terephtalate filter membrane with 8 μm-diameter pores. After pre-treatment with LPS, IL-4 and TGF-β1 in DMEM/F12 containing 3% FBS for 12 h, the trans-well inserts with OECs were transferred to the 12-well plates culturing primary cortical neurons. After co-culture for 48 h, the non-migratory OECs on the upper membrane surface were removed with cotton swabs, while the migrated OECs on the underside surface of the membrane were fixed with 4% PFA and stained with DAPI. For quantitative assessment, the number of nuclei stained migrating OECs was counted at five random fields per filter membrane in three independent experiments.

To investigate neurite outgrowth and extension, the number of neurites, branches and the length of neurites were measured (Witheford et al., 2013). Briefly, four randomly-chosen fields in each glass slide were photographed through a fluorescent microscope. All Tuj-1 immunofluorescent cells were selected to measure the number of neurites, number of branches and length of neurites. At least 20 neurons in each glass slide were assessed for each group. The numbers of primary, secondary and tertiary neurites were assessed. The primary neurite was defined as the neurite that extended from the body of neurons. The secondary neurite was defined as the neurite that extended from the primary neurite, and tertiary neurite was defined as the neurite that extended from the secondary neurite. The number of neurites was the number of primary neurites. The number of branches was the total number of secondary and tertiary neurites. Neurite length was defined as the length between the body and the farthest tip of the neurite.

Protein samples of OECs were reconstituted in a sample buffer and denatured by boiling at 100°C for 5 min. Protein lysates were loaded on SDS-PAGE for electrophoresis and electrophoretically transferred onto a methanol-activated polyvinylidene difluoride (PVDF) membrane (Millipore). Non-specific protein was blocked by incubation with 5% BSA diluted in Tris-buffered saline (TBS) containing 0.05% Tween-20 (TBST) for 2 h. The membrane was probed by primary antibodies at 4°C for 16 h and subsequently recognized by incubation with HRP-conjugated secondary antibodies for 2 h at room temperature. The immuno-reactive bands were presented by reacting with chemiluminescence reagents (ECL, Beyotime).

YAP siRNAs and scrambled siRNA controls were obtained from OriGene Technologies (China). OECs were seeded in 6-well plates, incubated overnight to a density of 70% and transfected using Lipofectamine 2000 (Thermo Fisher Scientific Waltham, MA, USA) for 72 h according to the manufacturer’s instructions. Afterwards, the medium was replaced with a fresh medium for subsequent protein analysis.

All data were represented as mean ± SD from at least three independent experiments. Statistical analysis was performed using Student’s t-test, one-way analysis of variance (ANOVA) or two-way ANOVA with Dunnett’s post test and χ² test. Statistical significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

According to the criteria of process-bearing and flattened OECs (Huang et al., 2008), we first analyzed the effects of pro-inflammatory (LPS) and anti-inflammatory (IL-4 and TGF-β1) conditions on the cell morphology of OEC using low-density OECs cultured on PDL-coated plates by time-lapse images. OECs under control conditions (DMEM/F12 containing 3% FBS) maintained a flattened appearance with short processes. When under pro-inflammatory conditions, OECs changed into a process-bearing shape with extremely long processes (Figure 1A). However, when under anti-inflammatory conditions, OECs changed to a flattened shape and showed a flat sheet-like shape within 8 h (Figure 1A). The morphological changes of OECs cultured at high-density were similar to low-density cultures (Figure 1B). The average process length of OECs increased and the circularity decreased under pro-inflammatory conditions, indicating an increased degree of cell elongation of OECs. In contrast, the average process length of OECs decreased under anti-inflammatory conditions (Figures 1C,D). Furthermore, pro-inflammatory conditions resulted in an increased proportion of process-bearing OECs, whereas anti-inflammatory conditions caused an increased proportion of flattened OECs (Figure 1E).

As the ECM modulated cell morphology (Tisay and Key, 1999; Vukovic et al., 2009), we next analyzed morphological changes of OECs on different substrates under pro- and anti-inflammatory conditions. Similar results were observed when OECs were grown on gelatin or laminin-coated surfaces (Supplementary Figure S1). Taken together, it strongly suggested that pro- and anti-inflammatory culture conditions were sufficient to induce morphological changes of OECs.

Reversible morphological changes have been reported between process-bearing and flattened OECs independent of any environmental stimuli (spontaneously) or induced by applying endothelin-1 or increasing the intracellular level of cAMP (Huang et al., 2008; van den Pol and Santarelli, 2003; Vincent et al., 2003). In the present study, OECs switched to a process-bearing shape within 24 h by adding LPS. When LPS was replaced with TGF-β1, process-bearing OECs returned to a flattened shape within 12 h (Figures 2A,C,D). Conversely, when TGF-β1 was replaced with LPS, flattened OECs partially returned to a process-bearing shape within 36 h (Figures 2B–D). LPS treatment increased the proportion of process-bearing OECs, and the proportion of process-bearing OECs significantly decreased by replacing LPS with TGF-β1 within 12 h. Interestingly, the proportion of process-bearing OECs could be partially reversed by replacing TGF-β1 with LPS within 36 h (Figure 2E). These results suggested that pro- and anti-inflammatory conditions induced partially reversible morphological changes of OECs. In addition, OECs maintained their process-bearing or flattened shape for at least 9 days under constant pro- or anti-inflammatory conditions (Figure 2F).

The Rho GTPases (including RhoA, Rac1 and Cdc42) play important roles in regulating cell morphology and the RhoA/ROCK/Myosin/F-actin pathway mediates the morphological plasticity of cultured OECs (Etienne-Manneville and Hall, 2002; Huang et al., 2011b). Thus, we evaluated whether morphological changes of OECs under pro- and anti-inflammatory conditions were attributed to different expression patterns of stress fibers by F-actin staining. OECs under pro-inflammatory conditions (LPS) formed few stress fibers. In contrast, OECs under anti-inflammatory conditions (IL-4 and TGF-β1) formed bundles of actin filaments at the edge and thick parallel bundles of actin filaments that spanned the entire cell body (Figure 3A). These results indicated that anti-inflammatory conditions induced morphological changes of OECs towards a flattened shape through cytoskeletal re-arrangements.

Next, we investigated whether the ROCK pathway mediated morphological changes of OECs. WB showed that LPS treatment decreased protein levels of phosphorylated MLC (p-MLC), one of the main substrates of ROCK. In contrast, TGF-β1 and IL-4 treatments increased p-MLC protein levels (Figure 3B), indicating that the ROCK pathway was involved in the morphological changes of OECs under pro- and anti-inflammatory conditions.

In addition, Y27632, a ROCK-specific inhibitor, promoted OECs to switch towards a process-bearing shape and markedly inhibited the formation of stress fibers and membrane ruffles induced by TGF-β1 (Figures 3C–F). Furthermore, Y27632 significantly decreased the p-MLC protein level and blocked morphological changes of OECs towards the flattened shape induced by TGF-β1 (Figures 3C–G). Taken together, these results suggested that the ROCK/F-actin pathway mediated TGF-β1-induced cytoskeletal re-arrangements and morphological changes of OECs.

The OECs’ morphology is associated with a specific functional phenotype (Vincent et al., 2005; Ekberg et al., 2012; Wang and Huang, 2012). Here, we explored the relationship between the morphological changes and functions of OECs under pro- and anti-inflammatory conditions. The proliferation of OECs was tested by Ki67 and BrdU staining. No significant influences on proliferation were observed when OECs were treated with LPS, IL-4 or TGF-β1 in comparison with the control (Supplementary Figures S2A–D). The average cell number of OECs showed no significant difference either (Supplementary Figure S2E). This suggested that pro- and anti-inflammatory conditions had no significant influences on the proliferation of OECs. However, the CCK-8 assay, detecting cellular metabolic activities of mitochondrial enzymes, showed significant difference between pro- and anti-inflammatory conditions, as pro-inflammatory conditions enhanced the cellular metabolic activity of OECs (Supplementary Figure S2F).

As morphological phenotypes of OECs displayed differential migratory properties and process-bearing OECs were more motile than flattened OECs (Huang et al., 2008; Wang and Huang, 2012), we then examined whether OECs under pro- and anti-inflammatory conditions also displayed differential migratory activities. In an in vitro scratch assay, process-bearing OECs under pro-inflammatory conditions covered more regions of scratch gaps and more OECs migrated into scratch gaps (Figures 4A,B). In contrast, flattened OECs under anti-inflammatory conditions showed less OECs migrating into scratch gaps, indicating that process-bearing OECs under pro-inflammatory conditions are more motile than flattened OECs under anti-inflammatory conditions (Figures 4A,B). In the matrigel drop assay (Figure 4C), spontaneous and un-directional movements of OECs under pro- and anti-inflammatory conditions were examined. Process-bearing OECs under pro-inflammatory conditions covered more area and migrated longer distances than OECs under control and anti-inflammatory conditions (Figures 4D–F).

Previous studies showed that migration of OECs was directly related to the rate of olfactory axon growth (Chehrehasa et al., 2010; Windus et al., 2011), so we examined the directional movement of OECs towards neurons using a trans-well chamber migration assay. OECs that migrated through the insert filter membrane increased by 20% under pro-inflammatory conditions compared to those under control and anti-inflammatory conditions (Figures 4G,H). Thus, it showed that process-bearing OECs under pro-inflammatory conditions displayed stronger migratory activities than flattened OECs under anti-inflammatory conditions.

To further explore the relationship between the morphology and function of OECs, cortical neurons were co-cultured with OECs for 3 days, and neurite outgrowth and extension were examined using Tuj1 staining (Figure 5A). Almost all neurites extended on OEC surfaces rather than on OEC-free regions (Figures 5B,C). In addition, neurites outgrew and extended along processes of OECs (Figure 5D). These results indicated that process-bearing OECs aided and promoted neurite outgrowth and extension, and that the cell contact-mediated effect of OECs was involved.

Next, we evaluated whether morphological changes of OECs under pro- and anti-inflammatory conditions also affected neurite outgrowth and extension. Cortical neurons grew longer and branched more neurites on OECs pre-treated with pro-inflammatory conditions (LPS) for 24 h (Figures 5E,F). Consistent with the above results, neurons outgrew and extended along the processes of OECs (Figure 5F). In contrast, neurons on OECs pre-treated with anti-inflammatory conditions (TGF-β1) showed shorter unbranched neurites (Figure 5G). Additionally, OECs under anti-inflammatory conditions (TGF-β1) grew in clusters and had short or no processes (Figure 5G). Furthermore, neurons on OECs pre-treated with LPS grew with more neurites and branches than those grown on control OECs and OECs pre-treated with TGF-β1 (Figures 5H,I). The total lengths of neurites grown on OECs pre-treated with LPS were longer than those grown on control OECs and OECs pre-treated with TGF-β1 (Figure 5J). These results suggested that OECs pre-treated with pro- and anti-inflammatory conditions differentially affected neurite outgrowth, branch and extension, and morphological shifts of OECs were involved in this neurite outgrowth-promoting effect.

As the ROCK/F-actin pathway mediated the morphological changes of OECs, we used ROCK inhibitor Y27632 to induce OECs towards a process-bearing shape, and examined neurite outgrowth and extension. OECs were pre-treated with Y27632 for 12 h to induce OECs towards a process-bearing shape, and then cortical neurons were co-cultured with OECs (OECs + Y27632 group). Alternatively, cortical neurons were co-cultured with OECs and the OEC/neuron co-cultures were pre-treated with Y27632 for 12 h (OECs/Neuron + Y27632 group). OEC/neuron co-cultures without Y27632 treatment (OECs group) were defined as the control. Compared with the OECs group, cortical neurons grew with more neurites and branches and extended longer neurites in the OECs + Y27632 group (Figures 6A–D). In addition, adenylate cyclase agonist forskolin, which also induced OECs towards a process-bearing shape, also promoted neurite extension along processes of OECs (data not shown). Strikingly, cortical neurons in the OECs/Neuron + Y27632 group grew with more neurites and branches than in OECs + Y27632 group (Figures 6A–C). Furthermore, a combination of OECs and Y27632 promoted neurites to extend across injured regions using an in vitro scratch assay (Figure 6E). All these results suggested that ROCK inhibitor Y27632 induced process-bearing OECs and enhanced neurite outgrowth and extension.

As neurites grew out and extended along the processes of OECs, we wondered whether direct cellular contact with OECs and ECM or adhesion molecules of OECs played key roles in the enhanced neurite outgrowth. To test this hypothesis, we fixed OECs with methanol to preserve their structure (ECM and adhesion molecules) and to eliminate the effect of the factors secreted by OECs. We found that neurites still grew out and extended along the processes of methanol-fixed OECs, indicating that direct cellular contact but not the factors secreted by OECs played a role in the outgrowth of neurites along the processes of OECs (Figure 6F). When the ECM and adhesion molecules of OECs were extracted and disrupted with urea, neurites grew out and extended randomly rather than along the processes of OECs. Furthermore, consistent with a previous study (Witheford et al., 2013), a block of neural adhesion molecule L1-CAM with antibody reduced the extension of neurites along the processes of OECs (Figure 6F). Together, these results suggested that process-bearing OECs induced by ROCK inhibitor Y27632 enhanced neurite outgrowth and extension via cellular contact and L1-CAM-dependent mechanisms.

The activity of transcriptional co-activator YAP is correlated to the stability of the actin cytoskeleton and cell tension, which were controlled by Rho GTPase pathways and myosin II (Wada et al., 2011; Nardone et al., 2017). Thus, we wondered whether YAP played a role in morphological shift and the enhanced neurite outgrowth-promoting property of OECs induced by ROCK inhibition. YAP was mostly localized in the nuclei of flattened OECs and in the nuclei and cytoplasm of process-bearing OECs (Figure 7A). Consistent with this, TGF-β1 treatment shifted OECs towards a flattened shape with a larger cellular area, formed more stress fibers and resulted in nuclear YAP accumulation. In contrast, ROCK inhibitor Y27632 shifted OECs towards a process-bearing shape with a smaller cellular area, inhibited the formation of stress fibers and excluded nuclear YAP to cytoplasm (Figures 7A–D). Consistent with YAP distribution, TGF-β1 and IL-4 treatments increased protein levels of YAP and decreased the phosphorylation of YAP at Ser127. Accordingly, the expression of CTGF, a well-characterized target gene of the YAP-TEAD complex, was dramatically up-regulated, indicating that TGF-β1 and IL-4 activated YAP (Figure 7E). In contrast, Y27632 and LPS treatments increased the phosphorylation of YAP at Ser127 and down-regulated the expression of CTGF, indicating that Y27632 and LPS inhibited YAP (Figure 7E). Furthermore, the increased protein levels of integrin and FAK autophosphorylation at the Tyr397 site were found in TGF-β1 and IL-4-treated OECs, indicating activated integrin-focal adhesion (FA) signaling and resultant cytoskeleton stability (Figure 7E). In contrast, a marked reduction in the protein levels of integrin and FAK autophosphorylation were found in Y27632 and LPS-treated OECs, indicating inactivated integrin-FA signaling and disrupted cytoskeleton (Figure 7E). Thus, these results indicated that stress fibers consisting of F-actin mediated the morphological shift of OECs and regulated YAP distribution.

As shown above, L1-CAM was involved in the enhanced neurite outgrowth-promoting property of OECs. Interestingly, Y27632 treatment inhibited YAP activity and increased the protein levels of L1-CAM of OECs (Figure 7E). Similarly, verteporfin, a YAP-TEAD inhibitor, also increased the L1-CAM protein level of OECs (Figure 7F). To further validate the role of YAP in the regulation of L1-CAM, we used siRNA to downregulate YAP protein levels endogenously. YAP-siRNA transfection decreased the YAP protein level of OECs and upregulated the L1-CAM protein level of OECs (Figure 7G). Collectively, these results suggested that YAP signaling downstream of the ROCK pathway mediated morphological shift and enhanced the neurite outgrowth-promoting properties of OECs through upregulating L1-CAM.