Triiodothyronine modulates neuronal plasticity mechanisms to enhance functional outcome after stroke

The development of new therapeutic approaches for stroke patients requires a detailed understanding of the mechanisms that enhance recovery of lost neurological functions. The efficacy to enhance homeostatic mechanisms during the first weeks after stroke will influence functional outcome. Thyroid hormones (TH) are essential regulators of neuronal plasticity, however, their role in recovery related mechanisms of neuronal plasticity after stroke remains unknown. This study addresses important findings of 3,5,3′-triiodo-L-thyronine (T3) in the regulation of homeostatic mechanisms that adjust excitability – inhibition ratio in the post-ischemic brain. This is valid during the first 2 weeks after experimental stroke induced by photothrombosis (PT) and in cultured neurons subjected to an in vitro model of acute cerebral ischemia. In the human post-stroke brain, we assessed the expression pattern of TH receptors (TR) protein levels, important for mediating T3 actions. Our results show that T3 modulates several plasticity mechanisms that may operate on different temporal and spatial scales as compensatory mechanisms to assure appropriate synaptic neurotransmission. We have shown in vivo that long-term administration of T3 after PT significantly (1) enhances lost sensorimotor function; (2) increases levels of synaptotagmin 1&2 and levels of the post-synaptic GluR2 subunit in AMPA receptors in the peri-infarct area; (3) increases dendritic spine density in the peri-infarct and contralateral region and (4) decreases tonic GABAergic signaling in the peri-infarct area by a reduced number of parvalbumin+ / c-fos+ neurons and glutamic acid decarboxylase 65/67 levels. In addition, we have shown that T3 modulates in vitro neuron membrane properties with the balance of inward glutamate ligand-gated channels currents and decreases synaptotagmin levels in conditions of deprived oxygen and glucose. Interestingly, we found increased levels of TRβ1 in the infarct core of post-mortem human stroke patients, which mediate T3 actions. Summarizing, our data identify T3 as a potential key therapeutic agent to enhance recovery of lost neurological functions after ischemic stroke.


Other Supplementary material for this manuscript includes:
Supplementary Videos 1-4

Thyroid hormones effects after experimental stroke (Study I)
For this study, 117 C57BL/6 male mice (20 to 26 g, aged nine to ten weeks, purchased from Charles River) were used. Out of 117 animals, 12 were excluded due to problems during surgery and mortality before entering the treatment phase and 105 animals were randomly assigned into the treatment groups ( Fig. 1). Treatment was initiated on day two after photothrombosis (PT) and every other day until the endpoint of the study. Vehicle (Vh, NaCl 0.9%), T3 (5 or 50 µg/kg) or T4 (5 or 50 µg/kg) were administered by intraperitoneal injection in a total of six administrations. On days two, seven and 14 after stroke onset or sham surgery, animals were evaluated for motor function.

Photothrombosis.
Focal ischemic stroke was induced by PT, as described previously [11,16]. Briefly, animals were anesthetized with isoflurane in N2O / O2 (0.7: 0.3, 5% induction and 1.5 -2 % maintenance) and placed into a stereotactic frame. After local anesthesia the open field test, respectively [8,15]. These assessments were performed in a blinded fashion to the investigator that performed the surgeries and treatments.
Rotating pole test. The RPT was used to assess postural and locomotor asymmetry that results from an unilateral brain lesion [4]. In brief, mice traversed a rotating wooden pole (length 1500 mm, diameter 40 mm, and elevation 700 mm) at zero, three, and ten rotations per minute (rpm), to the right and left sides. Every animal was trained during three days before surgery and tested the day before PT. After stroke or sham surgery, animals were evaluated on day two for randomization into treatment groups. Each trial was video recorded, and videos were used to assess sensorimotor dysfunction by using a zero to six scoring system ( Table 1). Animals that did not perform the behaviour test before the surgery (total score RPT < 20 points) or did not have motor deficits two days after PT (total score RPT > 15 points), were excluded from behaviour analysis. Behavioural analysis was performed in a blinded fashion to the investigator. In total, 42 animals were excluded from behaviour analysis and the following included: PT/Vh, n = 11; PT/T3 5 μg/kg, n = 10; PT/T3 50 μg/kg, n = 11; PT/T4 5 μg/kg, n = 10; PT/T4 50 μg/kg, n = 9.
Open field. The open field test was performed 14 days after stroke to assess both spontaneous post-ischemic locomotor activity and post-ischemic exploration behaviour [14]. Briefly, mice where placed into a square arena (44.5 cm × 44.5 cm) surrounded by 44.5 cm high sidewalls. The mouse was always placed in the center of the box and locomotion was recorded and the total distance traveled measured for five minutes.
Thyroid hormone levels determination. Animals were anesthetized with pentobarbital and plasma was collected 14 days after experimental stroke. Blood was collected from the heart into heparinized syringes and maintained at 2-8 ºC while handling. Plasma was collected after centrifugation 2000 x g for 10 minutes at 4 ºC and further stored at -80 ºC for further analysis. ELISA kit assay was used to determine TH levels in the plasma of mice.
Plasma levels of T3 and T4 were determined by a commercial ELISA kits (ThermoFisher Scientific cat #EIAT3C and cat #EIAT4C, respectively) according to manufacturers instructions. In brief, plasma samples were incubated with specific primary antibodies in donkey anti sheep or goat anti mouse coated 96-well plates. After washing in respective buffer, plates were incubated with 3,3′,5,5′-tetramethylbenzidine substrate and absorbance was measured at 450 nm.

Immunohistochemistry and
Immunofluorescence. Tissue collection for immunostainings was performed as described before [3,9]. Fourteen days after PT animals were deeply anesthetized with pentobarbital and perfused fixed with paraformaldehyde  Infarct size measurement. Coronal brain sections from the start until the end of the infarct and spaced one millimeter were collected and stained for NeuN (rabbit NeuN, Millipore, 1:5000). Brain slices were mounted in Pertex and digitalized (CanoScan 8800F, Canon, Tokyo, Japan). The non-injured portion of the ipsilateral and contralateral hemisphere were encircled and the indirect infarct volume was calculated by integration of areas from serial sections of each brain as described previously [12], using Fiji software [10]. Rabbit c-fos (Santa Cruz, 1:500) positive immunoreactivity (c-fos + ) was accessed using the avidin-biotin-HRP system, as described before.
The following animals were included in this analysis: PT/Vh, n = 7; PT/T3 50 μg/kg, n = 6; and PT/T4 50 μg/kg, n = 4. Bright field images were acquired with 4x magnification objective and Fiji software was used to draw regions of interest, using an optical grid to define the distances and draw the regions. PV + cells and PV + /c-fos + in the peri-infarct somatosensory cortex (area of 0.8 mm 2 ) and homotypic area in the contralateral hemisphere were counted.
The infarct core was identified by the lack of NeuN immunoreactivity in subsequent sections.

Immunoblotting.
Brains from mice were collected as previously described [9]. Fourteen days after PT animals were deeply anesthetized with pentobarbital and brains were immediately frozen (−40°C) in isopentane (Sigma-Aldrich, Taufkirchen, Germany) and further cooled down to −70°C on dry ice for immunoblotting. Fresh frozen brains were placed into a brain matrix and cut (+2.2 mm to -2.2 mm relatively to bregma). For each four millimeters thick section, the tissue correspondent to the infarct core and peri-infarct was collected. The procedure was performed in a refrigerated chamber at -20°C.
Tissue from human brains were dissected out by a pathologist following autopsy. Brain tissues were immediately frozen and stored at -80 ºC, temporarily moved to a refrigerated chamber at -20 ºC to excise small specimens and stored at -80 ºC until protein extraction.

Dynamics of dendritic spines after administration with T3 (Study II)
To study the effects of T3 on dendritic spine dynamics in mouse neocortical neurons after experimental stroke, eight Thy1-yellow fluorescent protein (YFP) transgenic mice (25 to 40 g, aged one year, own breeding), that express YFP in neuronal population were used. Mice were randomly assigned in the following treatment groups: PT/Vh, n = 4; PT/T3 50 μg/kg, n = 4 (Fig. 1). Treatment was administered as described above for Study I. Fourteen days after the surgery, mice were sacrificed, perfusion fixed with PFA 4% and brains were collected for further infarct volume assessment and dendritic spine analysis.

Photothrombosis.
To induce PT in animals for dendritic spine analysis (Study II) the surgical procedure as described for Study I, and the left hemisphere was illuminated with a cold light source (Schott KL 1500 LCD, intensity: 3050 K/4D) through a round aperture measuring 1.5 mm in diameter (equal to an illumination area of 1.767 mm 2 ) for 20 minutes.
This approach induced smaller infarct sizes so that dendritic spines could be analyzed in different regions in the peri-infarct area. The same procedure was performed in Sham operated animals, with saline injection instead of photosensitizing dye.

Detection and classification of dendritic spines from fluorescence Laser Scanning
Microscopy. Brain coronal sections (thickness 30 μm) were rinsed in PBS, mounted on super charged slides, and cover-slipped with PVA-DABCO (Sigma-Aldrich). Images were acquired using an AxioObserver LSM 710 confocal (Carl Zeiss) using a Plan-Apochromat Dendritic spine density and shape classification was accurately quantified and characterized using a three dimensional (3D) computational approach as previously described [5]. After median filter application to reduce noise, radius of 1.0 pixel, we performed proper image 3D deconvolution using interactive deconvolve 3D plugin from Fiji software, after theoretical point spread function (PSF) generation using diffraction PSF 3D plugin. Deconvolution restores image contrast that is lost during image recording due to the optical smearing introduced by the PSF of the microscope, and it is an important systematic error correction for dendritic spine analysis by improvement of signal to noise ratio [5].
For each region, three to five dendritic branches were randomly selected. Dendrites were manually selected, and spines were automatically detected using NeuronStudio software.
Minimum and maximum height for spines were set to 0.2 µm and 2.0 µm, respectively, and voxel dimensions were adjusted for our images (0.098 µm, 0.098 µm and 0.2 µm for xyz, respectively). For spine shape classification, we used the Rayburst algorithm provided by NeuronStudio software [5,6], which allowed to use all the information from a LSM image stack, and also provided procedures as declumping of merged spines and spine stem reattachment, making spine detection more accurate than manual or 2D method, where spines are masked along z axis. Dendritic spines were classified according to head to neck ratio and head diameter as stubby, mushroom or thin [1,5], using default parameters from NeuronStudio. Some detected dendritic spines from neighbor dendritic branches were manually deleted and not included in statistical analysis. Dendritic spine density was calculated with the ratio number of spines / dendrite length.

In vitro modulation of T3 in glutamatergic neurons (Study III)
An in vitro model of cerebral ischemia and electrophysiology studies were performed to study immediate effects of T3 in homeostatic plastic mechanisms, namely modulation of synaptic proteins crucial for neurotransmission and ionotropic glutamate receptors alphaamino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate (NMDA) evoked currents.
Cell cultures. Cultured glutamatergic cortical neurons were used after 7 -8 days in vitro (DIV). Primary cortical neuronal cultures were prepared as described before [7].  Immunobloting. Protein extraction was performed as previously described [2,9]. Western blots were performed as described above for brain extracts, and mouse synaptotagmin (BD Transduction Laboratories, 1:2000) was incubated overnight at 4 ºC. After blocking, the membranes were incubated with a mouse secondary HRP-conjugated antibody (Sigma-Aldrich, Stockholm, Sweden, 1:10000) for one hour at RT.

Electrophysiological recording of membrane currents.
To study ligand-gated channels AMPA and NMDA, we adopted the voltage-ramp method [17]. Electrodes were pulled on a vertical puller (PC-10, Narishige), from borosilicate glass capillaries (Harvard Apparatus). The initial patch microelectrode had a resistance of 5.6 -7.4 MΩ, when filled with internal solution. The electrode was sealed against cells at least 1.0 GΩ, and membrane was ruptured by suction pulses, which allows the recording of the intracellular membrane potential. Only cells with Ra (Access resistance) values < 10 MΩ were included.
Solutions were delivered diluted in the bath solution through a custom-made perfusion system, where capillary tubes with 250 µm inner diameter merged into a common outlet.
Drugs were applied close to cell at approximately 50 to 100 µm, at a rate of 20 µl/min.
Individual currents were recorded in the presence of T3 1 µM (n = 4) or Vh (n = 3), that were incubated 48 hours before the experiments. A sequence of voltage ramps at a rate of 0.23 mV/millisecond were applied at a holding potential of -80 mV. To obtain the agonist induced current-voltage (I-V) relation, ramps I-V curves were constructed applying a 500 milliseconds voltage ramp ranging from -110 mV to +20 mV elicited every 8 seconds.
Voltage ramps were applied in the absence and in the presence of AMPA and NMDA agonist glutamate at 50 µM and co-agonist of NMDA channels glycine at 3 µM, to enable subtraction of leak currents. CNQX and MK-801 were used both at 10 µM as antagonists of AMPA and NMDA channels, respectively. For stabilization of background currents, a minimum of 80 seconds was recorded before agonists and CNQX / MK-801 application.
Cell currents were recorded sequentially in the presence of specific K + -channel blockers tetraethylammonium sodium salt (5 mM) and 4-Aminopyridine (1 mM), that were applied in the perfusion system together with the other drugs. Voltage-gated K + channels needed to be blocked, since those channels were contributing to the conductance as well to the reversal potential obtained.