This study performed a behavioral experiment and comprehensive morphological assessment and quantitative synaptic ultrastructural analysis to determine pre- and postsynaptic protein expression in euthyroid HT mice. Morris water maze experiment is a classic method to evaluate the ability of spatial learning and memory by forcing experimental animals to look for the underwater hiding platform [25]. In the present study, mice in the HT group took a significantly longer time to find the hidden platform, had a reduced time spent in the target platform, and a decreased number of crossing the platform compared to mice in the Con group. This confirmed that HT itself can induce spatial learning and memory impairment regardless of thyroid functions in mice, which is in accordance with our previous study. Concurrently, our results provide initial evidence that euthyroid HT induces synaptic loss and alters the synaptic ultrastructure of mice. Interestingly, we observed microglia activation and enhanced phagocytosis of synaptic materials, in addition to an abnormal apposition of microglia to neurons in euthyroid HT mice, suggesting a mechanism of microglia-mediated synaptic loss, which may underlie the deleterious effects of HT on spatial learning and memory.
Synapses act as interfaces in which networks of neurons interact during physiological brain function [54]. We found that euthyroid HT produced a reduction in synaptic density, and reduced the number of vesicles, active zone length, PSD thickness, and synaptic curvature in the frontal lobe at the ultrastructural level. The changes indicated that euthyroid HT induced synaptic loss and impaired synaptic structures in mice. Active regions and PSD are key elements of complex synaptic signals that mediate synaptic plasticity [55]. Curvature can be observed at the interface of pre- and post-synaptic areas, designed to enhance surface area, ensuring direct delivery of the delivered transmitters as opposed to diffusion across the peripheral space. This enhances the efficacy of neurotransmission [56]. Consistent with these data, the quantification of SYN and PSD95 revealed synaptic loss in the frontal lobes of HT mice. Studies have confirmed that SYN, PSD95 and other synaptic proteins regulated the release of neurotransmitters and dictated the plasticity of synapses [20]. Altering these parameters contributes to altered plasticity at the structural and functional level, impairing the efficiency of synaptic transmission between neurons in HT, leading to neuropsychological HT alterations.
How HT leads to a loss of synapses in the euthyroid remains unclear. Thyroid dysfunction can induce neuronal deficits, including synaptic alterations [57, 58]. Our previous studies have highlighted that experimental hypothyroidism induces synaptic deficit in the hippocampus [59] and prefrontal lobe [34]. Interestingly, local deficiency (e.g., in the hippocampus) of thyroid hormone was observed prior to the decrement in plasma in an animal model of subclinical hypothyroidism [50]. In this study, T3 and T4 levels were detected in the frontal lobe but no differences were found between groups, indicating that the synapse loss in the euthyroid HT mice had nothing to do with differences in local thyroid hormones. On the other hand, thus far, synapse loss has been proposed to occur via neuron-autonomous mechanisms (for example, apoptosis) [60, 61] or synaptic stripping mediated by the glia [37, 62]. Evidence for either scenario or a combination of both exists. As an exemplar, Albert et al. showed a loss of dentate afferent synapses and from ultrastructural data highlighting autonomous and glia-mediated degradation of synapses in chronic multiple sclerosis (MS) [63]. In this study, synapse loss was unlikely to be due to local neuronal cell loss in HT mice, because neuron numbers were within the normal range for this brain area. This was consistent with our previous findings showing a lack of neuronal apoptosis or ultrastructural damage in the frontal lobe of euthyroid HT mice [13].
Synaptic loss has also been reported in other animal models of autoimmune illness, such as systemic lupus erythematosus (SLE) [64] and MS [63]. The close-association of synapses and microglia may contribute to synapse loss [64, 65] a factor that has been poorly investigated in HT mice. During activation, microglia undergo clear morphological changes. Our findings indicated that the number of round/amoeboid Iba1-labeled cells, a reactive microglia phenotype [44] significantly increased in the frontal lobe of HT mice. Microglial activation enhances phagocytosis. An intriguing hypothesis is therefore that HT enhances microglial phagocytosis of synaptic elements, contributing to synapse loss in euthyroid HT. Consistent with this hypothesis, increased PSD95 engulfment occurred in HT mice. We interpreted the presence of PSD95-labeled puncta in microglia as a consequence of active phagocytosis of synapses. Indeed, EM observations showed that membrane-bound compartments containing synaptic vesicles were visible within the microglia of HT mice. Similar EM observations have been reported in lupus-prone mice, in which microglial cells are reactive and engulf synaptic material, suggesting microglia-dependent synaptic loss [64].
The activation of microglia is accompanied by chemotactic responses and migration towards areas of neuronal damage [66, 67]. We thus explored the physical associations between microglia and neurons in the frontal lobe. Microglial attachment to neuronal perikarya increased in HT mice. In addition, at the ultrastructural level, microglial cells directly contacted neuronal cell bodies lacking synapses a phenomenon rarely observed in control mice. An abnormal association of microglia and neurons has been reported in MS [68] and Alzheimer’s disease [69]. A consequence of close apposition between microglial cells and neurons is the interruption of synaptic contacts [25, 26, 41], termed synaptic stripping [67], also reported in animal models of cortical inflammation [26]. Morphological studies [21, 70] have revealed direct contact between microglia and synaptic elements supporting the possibility that microglia strip synaptic elements. Consistent with these data, we provide initial morphological evidence that reactive microglia migrate to and strip presynaptic terminals from neuronal soma in the context of euthyroid HT.
Synaptic elimination is mediated by microglial phagocytosis as a consequence of direct synaptic removal [71]. However, microglial cells secrete synaptotoxic factors, including TNF-α and IL-1β [72]. In vitro studies revealed that conditioned medium from reactive microglia led to a loss of synaptophysin in primary neuronal cultures [73]. In addition, some soluble neurotrophic factors [23], which were secreted by microglia, could be crucial for synapse maintenance, therefore, dysfunction in homeostatic microglial activity may result in a lack of such support, thereby contributing to synaptic loss. Our previous work demonstrated that euthyroid HT increased the expression of TNF-α and IL-1β in the frontal lobes of mice [13]. Thus, synaptic loss following euthyroid HT may be, at least partially, attributed to activated microglia. This suggests that the inhibition of microglial activation offers neuroprotection during euthyroid HT.
Our study also has some limitations. We assessed the role of microglia in synaptic remodeling in the frontal lobe. However, the synaptic loss that accompanies synaptic ultrastructure damage in this brain area may not be only attributable to reactive microglia. Astrocytes are the most numerous cell in the central nervous system that provide trophic support for neurons, promote formation and function of synapses, prune synapses by phagocytosis, and fulfil a range of other homeostatic maintenance functions[74–77]. Previous studies found that neuroinflammation and ischaemia induced two different types of reactive astrocytes that they termed A1 and A2, respectively. A1 astrocytes highly upregulate many classical complement cascade genes previously shown to be destructive to synapses, so they postulated that A1 astrocytes might be harmful. By contrast, A2 astrocytes upregulated many neurotrophic factors, and they therefore postulated that A2 astrocytes were protective. Recent data revealed that A1 neurotoxic astrocytes were induced by activated microglia [78]. In fact, our previous study revealed increased activation of astrocytes using the same HT model, in which we identified a greater number of activated cells and a greater area of GFAP expression in HT mice than in controls [13]. Astrocytes lead to synaptic loss both directly (through recognition of a phagocytosis signal from synaptic components) and/or indirectly (by inducing the deposition of certain complement proteins at the synapses, which are subsequently eliminated by microglia) [79]. Further studies are now required to determine whether astrocytes contribute to synaptic loss in euthyroid HT. But this is not the topic of our present study.