The interrelationships between neuronal viability, synaptic integrity, microglial responses, and amyloid-beta formation in an in vitro neurotrauma model

The interrelationships between neuronal viability, synaptic integrity, and microglial responses remain in infancy. In dealing with the question, we induced a stretch injury to evaluate the mechanical effects of trauma on rat primary cortical neurons and BV2 microglial cells in a transwell culture system. The viability of primary neurons and BV2 cells was determined by MTT. Synaptic integrity was evaluated by determining the expression of beta-secretase 1 (BACE1), amyloid-beta (Aβ), microtubule-associated protein 2 (MAP2), and synaptophysin (vehicle protein). Both CD16/32-positive (CD16/32+) and CD206-positive (CD206+) microglia cells were detected by immunofluorescence staining. The phagocytic ability of the BV2 cells was determined using pHrodo E. coli BioParticles conjugates and flow cytometry. We found that stretch injury BV2 cells caused reduced viability and synaptic abnormalities characterized by Aβ accumulation and reductions of BACE1, MAP2, and synaptophysin in primary neurons. Intact BV2 cells exhibited normal phagocytic ability and were predominantly CD206+ microglia cells, whereas the injured BV2 cells exhibited reduced phagocytic ability and were predominantly CD16/32+ microglial cells. Like a stretch injury, the injured BV2 cells can cause both reduced viability and synaptic abnormalities in primary neurons; intact BV2 cells, when cocultured with primary neurons, can protect against the stretch-injured-induced reduced viability and synaptic abnormalities in primary neurons. We conclude that CD206+ and CD16/32+ BV-2 cells can produce neuroprotective and cytotoxic effects on primary cortical neurons.


The stretch injury-induced reduction in neuronal viability can be prevented by intact BV2 cells, whereas neuronal viability can be reduced by injured BV2 microglial cells. First, a custom-built
Cell Injury Controller II system was used to apply stretch injury to rat primary cortical neurons in a silastic culture plate (Flex plates) well. After 2nd stretch, we placed injured neurons in the lower well and unchallenged BV2 microglial cells in the insert of a Transwell coculture system (Fig. 1A) for 48 h. Compared to the neuron group, the injured neuron group had significantly lower cell viability (Fig. 1C). Compared to the injured neuron group, the injured neuron + BV2 cell group had significantly higher cell viability (Fig. 1C). There was an insignificant difference in BV2 microglial survival rate between the BV2 group and the neuron + BV2 group or between the BV2 group and the injured BV2 group (Fig. 1D). In a separate experiment, we placed injured BV2 cells and unchallenged cortical neurons in the lower well and insert of a Transwell coculture system, respectively (Fig. 1B). Compared to the BV2 group, the neuron + BV2 group, the injured-BV2 group, or the neuron + injured BV2 group had an insignificant difference (Fig. 1E) in microglia survival rate. However, compared to the neuron group, the neuron + injured BV2 cells, but not the neuron + BV2 group, had a significantly lower neuron survival rate (Fig. 1F).

Stretch injury-induced axonal injury can be attenuated by intact BV2 cells, whereas synaptic integrity can be disrupted by injured BV2 cells.
Microtubule-associated protein 2 (MAP2) is a neuron-specific protein that promotes the assembly and stability of the microtubule network. Synaptophysin (SYP) is a synaptic vesicle protein that regulates vesicle endocytosis in neurons. Next, we asked whether the fluorescence intensity of both MAP2 and SYP in primary cultured rat cortical neurons can be affected by stretch injury. Indeed, we observed that compared to the untreated neuron group, the injured-neuron group had a significantly lower % of both MAP2 and SYP ( Fig. 2A,B). However, compared to the injured-neuron group, the injured neuron + BV2 group had a significantly higher % of both MAP2 and SYP mean fluorescence intensity. In addition, compared to the neuron group, the neuron + injured-BV2 group had a significantly lower mean fluorescence intensity of both MAP2 and SYP (Fig. 2C,D). Fig. 3, injured neurons had significantly lower protein levels of APP than unchallenged neurons in the neuron + BV2 cell group (Fig. 3A, for original images of the blots please see Supplementary Fig. S1) Fig. 3 also shows that in the coculture system, the neuronal levels of APP were significantly decreased by injured BV2 cells (Fig. 3D, for original images of the blots please see Supplementary  Fig. S4). In contrast, the neuronal levels of both BACE1 and Aβ (Fig. 3E,F, for original images of the blots please see Supplementary Figs. S5 and S6) were significantly increased by coculture with injured BV2 cells. These observations suggest that in living brain tissues, some injured microglia (polarized towards the CD16/32 + phenotype) increase the neuronal levels of both Aβ and BACE1 but decrease the neuronal levels of APP in normal neurons, whereas uninjured microglia (polarized towards the CD206 + phenotype) normalize the neuronal levels of these proteins in injured neurons.

Stretch injury-induced neuronal Aβ accumulation can be decreased by intact BV2 cells, whereas the levels of Aβ in intact neurons can be increased by injured BV2 cells. As shown in
The phagocytic capacity of BV2 microglial cells is not affected by injured neurons but is significantly reduced by stretch injury. The efficiency of the BV2 microglial phagocytosis is determined by the pHrodo BioParticles Conjugates and quantified using a flow cytometer (Fig. 4A,B). Compared to the BV2 group, the neuron + BV2 group or the injured-neuron + BV2 group had an insignificant difference in BV2 phagocytic capacity (Fig. 4A,C). However, in all assays, the percentage of bead uptake by BV2 cells treated with Cytochalasin D (Cyto.D, used as a negative control to inhibit phagocytosis) was significantly lower than that by the control Figure 1. Neuronal death can be caused by stretch injury or coculture with injured BV2 microglial cells. Additionally, the stretch injury-induced reduction in neuronal viability was attenuated by coculture with BV2 cells. The rat primary cortical neurons were placed on the lower well and received 1st stretch, then cocultured with BV2 for 24 h. After 2nd stretch, we placed injured rat cortical neurons in the lower well and unchallenged BV2 cells in the insert of a  Fig. 4C). In a separate experiment, we found that the injured BV2 cell and neuron + injured BV2 cell groups exhibited significantly lower phagocytotic capacity than that in the BV2 cell and neuron + BV2 cell groups (Fig. 4B,D, 28.7% vs. 51% bead uptake).
Stretch-injured BV2 cells are predominantly CD16/32 + cells, and unchallenged or intact BV2 microglial cells are predominantly CD206 + cells. Our present study aimed to explore the effects of injured neurons and stretch-injured BV2 cells on the number of CD16/32 + and CD206 + microglia in a coculture system. We found that the percentage of cells in the injured neuron + BV2 cell group that were CD206 + microglia was significantly higher than that in the neuron + BV2 cell group (45.5% vs. 20.4%, Fig. 5A). However, there was no significant difference in the number of CD16/32 + BV2 microglia between the neuron + BV2 cell group and the injured neuron + BV2 cell group had (0.6% vs. 2.3%, Fig. 5A). Figure 5B shows that the percentage of CD16/32 + microglia in the injured BV2 cell group and neuron + injured BV2 cell group was significantly higher than that in the intact BV2 cell group (56.4% vs. 13.7%, 53.7% vs. 13.7%). The change in the percentage of CD16/32 + microglia among the different groups was negligible (Fig. 5B). As shown in Fig. 5, CD206 + microglia predominated during coculture with injured-neuron. However, under stretch injury, CD16/32 + microglial predominated (Fig. 5).

Disscussion
At first, our results showed that moderate stretch injury significantly altered the viability of neurons but not BV2 microglial cells. Although stretch injury did not affect the viability of the BV2 microglial cells, the injured BV2 cells might cause neuronal injury by recruiting more CD16/32 + cells. Next, we asked whether BV2 culture could present synaptic abnormalities following stretch injury. We found that synaptic abnormalities can be caused by stretch injury or culture with injured-BV2 cells. Additionally, stretch injury-induced synaptic abnormalities (e.g., altered microtubule and vesicle proteins) can be reversed www.nature.com/scientificreports/ by unchallenged BV2 cells. Again, it was found that injured-BV2 cells may cause synaptic abnormalities by recruiting more CD16/32 + cells, whereas unchallenged BV2 cells may preserve axonal integrity by recruiting more CD206 + cells. According to a review article, Aβ is an excellent indicator of microglial response after traumatic brain injury (TBI) 9 . BACE1 (or β-secretase) and the gamma-secretase complex are associated with Aβ genesis after TBI. Our present results confirmed that injured neurons had significantly higher protein levels of both Aβ and BACE1 but significantly lower protein levels of APP than did the unchallenged neurons and cells in the neuron + BV2 cell group. Although intact BV2 cells did not affect the levels of Aβ, BACE1 or APP in the neurons, they significantly reversed the stretch injury-induced altered protein levels of Aβ, BACE1, and APP in the neurons. On the other hand, the neuronal levels of both Aβ and BACE1 were significantly increased by cocultures with injured BV2 cells, whereas the neuronal levels of APP were significantly decreased by injured-BV2 cells. These observations suggest that in living brain tissues, some injured microglia (CD16/32 + phenotype) might increase the levels of both Aβ and BACE1 but decrease the levels of APP in normal neurons, whereas uninjured microglia (CD206 + phenotype) might normalize the neuronal levels of these proteins in injured neurons.
Microglia, which resident macrophages of the central nervous system (CNS), mediate primary immune reactions 10 . Microglia phagocytose synaptic debris such as accumulated Aβ. Our present study aimed to elucidate the relationship between microglia phagocytosis and injured neurons in a culture system. Flow cytometric analysis of the phagocytic capacity of BV2 microglial cells using PHrodo E. coli as target particles was performed 11 . We observed that the BV2, BV2 + neuron, and BV2 + injured neuron groups exhibited a similar phagocytic capacity. On the other hand, we found that the injured BV2 cells and neuron + injured-BV2 group exhibited significantly lower phagocytotic capacity than the BV2 and neuron + BV2 groups. Our data indicate that coculturing intact or were enhanced by coculture with injured-BV2. Representative immunoblots are shown. β-actin was used as a loading control. Measurements were made in triplicate, and each bar represents the mean ± SD. One-way ANOVA (for APP and BACE1 expression values) and two-way ANOVA (for Aβ expression values) with Tukey's test were used for multiple comparisons. www.nature.com/scientificreports/ injured neurons with BV2 cellls does not affect the phagocytotic capacity of BV2 cells. However, stretch injury to BV2 cells results in disruption of phagocytosis in injured BV2 cells. Our data further showed that intact B2 cells with normal phagocytic capacity reduced the stretch injury-induced Aβ accumulation and reduced viability in primary cortical neurons. On the other hand, injured BV2 cells with reduced phagocytotic capacity caused neuronal injury and synaptic abnormalities in the cocultured primary neurons. We used pHrodo Green Escherichia coli (E. coli) BioParticles Conjugates in the present study as a marker for phagocytic ability 12,13 . BV2 microglia were cultured, injured, and co-incubated with pHrodo-labelled E. coli over 60 min and their phagocytic and their phagocytic abilities were quantified by flow cytometry. Cytochalasin D (10 μM), an inhibitor for phagocytosis, was used as the negative control. This method assesses engulfment events but does not serve to understand the phagocytosis processing of biologically relevant cargo.
The "find-me" stage of phagocytosis, where target cells release chemotactic signals intracellularly and promote a migratory response from the phagocyte, does not occur when using BioParticles in our transwell cocultured model, nor is the "eat-me" stage which is exposed on the surface of the target cell to directly induce phagocytosis by proximal phagocytes, as BioParticles are not degraded. Therefore the last stage of phagocytosis is not assessed either. By using the E. coli BioParticles assay, we found that the stretch injury-induced neuronal viability and synaptic integrity reductions were attenuated by intact BV2 cells with normal phagocytic. In contrast, both the neuronal viability and synaptic integrity were reduced by coculture with injured-BV2 cells (which possess reduced phagocytic ability). Although microglia have been shown to engulf and clear damaged cellular debris after insult, deficits in microglia function may contribute to synaptic abnormalities in some neurodevelopment diseases. In our present study, microglia with normal phagocytosis may maintain neuronal viability by downregulating both Aβ accumulation and synaptic abnormalities caused by mechanical injury. In contrast, injured microglia with reduced phagocytic ability may reduce neuronal viability by upregulating both Aβ accumulation and synaptic abnormalities following mechanical injury. Therefore, we should investigate the "find-me" and "eat-me" signals in microglia and neurons under stretch stress in the future.
Our results are consistent with previous findings in models of inflammatory neurodegeneration 11,14 . Thus, inhibition of microglial phagocytosis seems sufficient to prevent neuronal death following stretch injury or   15,16 . Microglia can engulf and clear damaged cellular debris after brain insult, wheases deficits in microglia function may contribute to synaptic abnormalities seen in some neurodevelopmental disorders 5 . Although BV2 cells have been used frequently and widely in microglia-relevant studies, recent doubts have been raised about their suitability 17,18 . For example, a more recent study showed that BV2 cells only partially model primary microglia 19 . Thus, our present data using BV2 in microglia-related studies should be carefully considered. www.nature.com/scientificreports/ The Aβ plaques found in TBI patients develop rapidly and can appear within a few hours after injury 20,21 . Additionally, Aβ accumulation after TBI is associated with an increase in BACE1 expression 22 . Microglia may clear Aβ plaques via phagocytosis during TBI 3 . Our present data show that CD206 + BV2 microglia possess a normal phagocytic capacity, allowing them to preserve viability and synaptic integrity in injured neurons via decreasing Aβ formation and increasing phagocytic clearance of Aβ plaques. In contrast, injured CD16/32 + phenotype with reduced phagocytic capacity can cause cell death and synaptic abnormalities in neurons via increasing Aβ accumulation. Our present hypothesis is consistent with several previous investigations. For example, minocycline, an inhibitor of microglial response, can treat TBI by precluding the formation of Aβ through the restoration of the nonamyloidogenic APP processing pathway involving α-secretase 23 . Other anti-inflammatory compounds also exert neuroprotective effects against TBI by enhancing the α-secretase pathway [24][25][26] . CD206 + microglia might contribute to the reduced propagation of Aβ into unaffected neurons or brain tissue 27 via phagocytic removal of Aβ debris.
As depicted in Fig. 6, our results demonstrated that moderate stretch injury causes Aβ accumulation, synaptic abnormalities, and reduced viability in rat primary cortical neurons (Fig. 6A). When cocultured with unchallenged CD206 + BV2 cells with normal phagocytic capacity, stretch injury-induced neuronal Aβ accumulation, axonal injury, and neuronal death are significantly reduced. Although moderate stretch injury does not significantly affect BV2 cell viability, it significantly reduces the phagocytic capacity of BV2 cells. In addition, stretch injury shifts BV2 cells from the CD206 + phenotype cells to the CD16/32 + phenotype. Accordingly, CD16/32 + BV2 cells with reduced phagocytic capacity cause Aβ accumulation, axonal injury, and reduced viability in cortical neurons (Fig. 6B). Microglia establish an intimate contact with Aβ plaques in the brain and become reactive 28,29 . Microglia appears activated in the vicinity of Aβ plagues. Microglia contribute to the propagation of Aβ into affected brain tissue 27 . In the present results, the CD206 + BV2 cells and CD16/32 + cells might delay and accelerate the propagation of Aβ into the unaffected primary cortical neurons, respectively, and result in neuroprotection and cytotoxic effects (Fig. 6AII,BII). To understand microglia functions and the underlying signaling machinery, many attempts were made to employ functional in vitro studies of microglia. The range of available cell culture models is broad, and they come with different advantages and disadvantages for functional assays 30 . They discuss the potentials and shortcomings of transformed cell lines (e.g., BV2 cells) and coculture models for functional studies in vitro. Although our present data indicate that BV2 microglia may improve neuronal survival in stretch injured primary cortical neurons by suppressing Aβ accumulation via maintaining their

Conclusion
Using a microglia-neuron cocultured system, we found that stretch injury-induced amyloid-beta (Aβ) accumulation, axonal injury, and reduced viability in cultured primary rat cortical neurons in vitro. Intact and injured microglia were predominatly CD206 + microglia and CD16/32 + microglia, respectively. Stretch injury or injured BV2 cells caused Aβ accumulation, synaptic abnormalities, and reduced viability in cortical neurons; however, intact BV2 cells protected against stretch-induced injury. Our data suggest that recruitments of more CD206 + or CD16/32 + cells can produce neuroprotective and cytotoxic effects, respectively, on primary cortical neurons.

Coculture of primary cortical neurons and BV2 microglial cells.
We placed unchallenged rat cortical neurons (3 × 10 6 ) in the lower well (neuron group), neurons in the lower well and BV2 microglial cells (2 × 10 6 ) on the insert (BV2 cell + neuron group), injured neurons in the lower well (injured neuron group), or injured neurons in the lower well and BV2 microglial cells in the insert (BV2 cell + injured neuron group) of a Transwell coculture system (Fig. 1A). Forty-eight hours after coculturing, neurons were collected to assess cell viability, the number of synaptophysin-expression neurons, and protein levels. In addition, BV2 microglial cells were collected to evaluate phagocytosis and phenotype.
Statistics. The data are presented as the mean ± SD. Statistical analyses were performed using GraphPad Prism (version 7.04 for Windows; GraphPad Software, San Diego, CA, USA). Data matrices were first tested for normality and homoscedasticity with Shapiro's Wilk and Levene's test. The following parametric tests were applied according to the data characteristics and when the required data assumptions were fulfilled. Cell viability, APP and BACE1 expression, and phagocytosis were analyzed by the one-way ANOVA followed by Tukey's post hoc test. Synaptic plasticity, Aβ expression, and microglia phenotypes were analyzed by the two-way ANOVA followed by Tukey's post hoc test. The significant level was set at P < 0.05.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.