A glibenclamide-sensitive TRPM4-mediated component of CA1 excitatory postsynaptic potentials appears in experimental autoimmune encephalomyelitis

The transient receptor potential melastatin 4 (TRPM4) channel contributes to disease severity in the murine experimental autoimmune encephalomyelitis (EAE) model of multiple sclerosis and to neuronal cell death in models of excitotoxicity and traumatic brain injury. As TRPM4 is activated by intracellular calcium and conducts monovalent cations, we hypothesized that TRPM4 may contribute to and boost excitatory synaptic transmission in CA1 pyramidal neurons of the hippocampus. Using single-spine calcium imaging and electrophysiology, we found no effect of the TRPM4 antagonists 9-phenanthrol and glibenclamide on synaptic transmission in hippocampal slices from healthy mice. In contrast, glibenclamide but not 9-phenanthrol reduced excitatory synaptic potentials in slices from EAE mice, an effect that was absent in slices from EAE mice lacking TRPM4. We conclude that TRPM4 plays little role in basal hippocampal synaptic transmission, but a glibenclamide-sensitive TRPM4-mediated contribution to excitatory postsynaptic responses is upregulated at the acute phase of EAE.

The transient receptor potential melastatin 4 (TRPM4) channel contributes to disease severity in the murine experimental autoimmune encephalomyelitis (EAE) model of multiple sclerosis and to neuronal cell death in models of excitotoxicity and traumatic brain injury. As TRPM4 is activated by intracellular calcium and conducts monovalent cations, we hypothesized that TRPM4 may contribute to and boost excitatory synaptic transmission in CA1 pyramidal neurons of the hippocampus. Using single-spine calcium imaging and electrophysiology, we found no effect of the TRPM4 antagonists 9-phenanthrol and glibenclamide on synaptic transmission in hippocampal slices from healthy mice. In contrast, glibenclamide but not 9-phenanthrol reduced excitatory synaptic potentials in slices from EAE mice, an effect that was absent in slices from EAE mice lacking TRPM4. We conclude that TRPM4 plays little role in basal hippocampal synaptic transmission, but a glibenclamide-sensitive TRPM4mediated contribution to excitatory postsynaptic responses is upregulated at the acute phase of EAE.
In this study we tested the hypothesis that transient receptor potential melastatin 4 (TRPM4) channels are activated during, and contribute to, excitatory synaptic responses in CA1 hippocampal neurons. As we previously demonstrated that TRPM4-lacking mice have improved outcomes in the EAE model of multiple sclerosis and that Trpm4 -/neurons are protected from excitotoxicity, we also examined the role of TRPM4 in synaptic transmission at the acute phase of EAE 1 .
TRPM4 are non-selective monovalent cation channels activated by intracellular calcium and modulated by ATP, calmodulin, protein kinase-c (PKC), phosphatidylinositol 4,5-bisphosphate, and H 2 O 2 2-7 . It has been suggested that the TRPM4 channel mediates the Ca 2+ -activated non-selective cation current (I CAN ) in a number of cell types 8 . TRPM4 channels are ubiquitously expressed in both excitable and non-excitable cells where they contribute to heart sinus rhythm 9,10 , boost smooth muscle cell depolarization 11 , and regulate calcium by decreasing the driving force in T cells and dendritic cells 12 . In the brain, TRPM4 antagonists reduce bursting in neurons of the substantia nigra, layer V of the medial entorhinal cortex and the thalamic reticular nucleus [13][14][15] . Similarly, in the pre-Bötzinger complex, TRPM4 contributes to bursting and inspiratory drive but not breathing rhythm generation [16][17][18] . In the hippocampus, knock-out of TRPM4 impairs theta-burst induced long-term potentiation (LTP) of synaptic transmission 19 and in cerebellar Purkinje cells, both TRPM4 and TRPM5 contribute to, but are not absolutely required for the depolarization-induced slow current 20 . Recently, CA1 pyramidal neurons have been reported to express TRPM4 along the apical dendrites, accompanied by a TRPM4-dependent inward current 21 . TRPM4 may therefore be synaptically activated in CA1 neurons by calcium entering downstream of NMDA and metabotropic glutamate receptors. The additional depolarization due to TRPM4 could then further increase synaptically induced calcium entry. Such a positive-feedback process could play a physiological role, but if left unchecked may contribute to excitotoxicity. Indeed, TRPM4 expression is altered in spinal neurons, glial cells, and the neuro-vasculature in response to injury [22][23][24] . In animal models of spinal cord injury and brain hemorrhage [25][26][27][28] , pharmacological blockade or knockout of TRPM4 improves clinical outcomes and also protects neurons against glutamate-induced excitotoxicity 1,29 . A recent study found that NMDA receptors and TRPM4 directly couple and disruption of this interaction is protective against excitotoxicity and in a mouse model of stroke 28 . Additionally, TRPM4 reportedly co-assembles with the sulfonylurea receptor 1 (SUR1) 30 . SUR1 is a member of the ATP-binding cassette transporter family and also co-assembles with the Kir6.2 inwardly-rectifying ATP-sensitive K + channel 31 . In spinal cord injury, SUR1 but not Kir6.2 is upregulated 32 . SUR1 and TRPM4 upregulation have been reported in brain injury and in murine EAE 27,33,34 . The anti-diabetic drug, glibenclamide, antagonizes SUR1/Kir6.2, SUR1/ TRPM4 and TRPM4 channels 35 and the neuroprotective effects of glibenclamide treatment are similar to TRPM4 or SUR1 knock-out in EAE and brain injury models 1,33,36,37 . In the hippocampus, the contribution of TRPM4 to basal synaptic transmission at the CA3-CA1 Schaffer collateral synapse remains poorly understood. Given that NMDAR-dependent calcium influx could activate TRPM4 and drive further depolarization and calcium entry, we measured CA3-CA1 excitatory postsynaptic calcium transients (EPSCaTs) in wild-type mouse organotypic hippocampal slice cultures and excitatory postsynaptic potentials (EPSPs) in acute hippocampal slices from wild-type and Trpm4 -/healthy and EAE mice. We found no evidence for the involvement of TRPM4 in basal synaptic transmission. In EAE, we observed a reduction in EPSPs in the presence of glibenclamide in wild-type but not Trpm4 -/or healthy mice suggesting a glibenclamide-sensitive TRPM4-mediated contribution to EPSPs during CNS inflammation.

Results
TRPM4 antagonism does not reduce calcium influx at individual CA1 pyramidal neuron spines or their parent dendrites. To study the role of TRPM4 and calcium influx at the spine and dendrite, mouse organotypic hippocampal slice cultures were biolistically transfected with tdimer2 as a morphology marker and GCaMP6f. to monitor intracellular calcium of CA1 pyramidal neurons (Fig. 1a). Excitatory postsynaptic calcium transients (EPSCaTs) at single spines and dendrites in response to electrical stimulation of the Schaffer collaterals were monitored using a two-photon microscope (Fig. 1b). With the expectation that TRPM4 activation requires and contributes to large calcium transients, we aimed to induce not only isolated transients at single spines but responses at the dendrite as well (Fig. 1b-d). Washing in the TRPM4 antagonist, 9-phenanthrol, did not significantly reduce spine EPSCaT peak amplitude ( Fig. 1e; F (2,26) = 0.6477, p = 0.53, RM-ANOVA) or area under the curve (AUC) ( Fig. 1e; F (2,26) = 0.5971, p = 0.56, RM-ANOVA). Nor was there any reduction in dendritic CaT peak amplitude ( Fig. 1f; F (2,8)  TRPM4 antagonists do not reduce CA1 excitatory synaptic potentials in healthy wild-type or Trpm4-deficient mice. We next prepared acute hippocampal slices from adult wild-type and Trpm4 -/mice and made whole-cell patch-clamp recordings from CA1 pyramidal neurons. As action potentials, inhibitory potentials, and calcium-activated small conductance potassium (SK) channels might mask a TRPM4-mediated component of EPSPs, QX-314 was included in the intracellular solution to block voltage-gated Na + channels, and bicuculline in the extracellular solution blocked GABA A and SK channels 38 . D-serine was included in the extracellular solution to ensure that the glycine site of NMDA receptors was occupied. Large EPSPs (~ 40 mV deflections), expected to activate calcium channels, were recorded from CA1 pyramidal neurons in response to Schaffer collateral stimulation in stratum radiatum. There was no effect of 9-phenanthrol on EPSP amplitude or AUC in hippocampal slices from either wild-type ( Fig. 2a The TRPM4 antagonist glibenclamide also had no effect on the EPSP peak amplitude or AUC in either wild-type (Fig. 2e,f, Supplemental Fig. 2; Wilcoxon paired test amplitude p = 0.59, AUC p = 0.63) or Trpm4 -/mice (Fig. 2g,h; Wilcoxon paired test amplitude p = 0.31; AUC p = 0.09). There were also no differences due to genotype in the cell parameters and baseline EPSP amplitude or AUC (Table 1). These results suggest that a TRPM4-mediated conductance does not significantly contribute to EPSPs. Glibenclamide but not 9-phenanthrol reduces EPSP peak amplitude in wild-type EAE mice. Several studies have suggested that TRPM4 function plays an important role in diseased states, including models of multiple sclerosis 1,29 , ischemia 25,39 , and brain injury 40 . Therefore, we induced EAE in Trpm4 +/+ and Trpm4 -/littermates to test whether TRPM4-dependent currents only significantly contribute to synaptic transmission in an inflamed state. Clinical scores and body weight were measured daily (Fig. 3a,b). Acute slices were prepared at the peak of the acute phase, i.e. a clinical score of at least 3 or a decreasing score, blind to genotype (Fig. 3a). There were no differences due to genotype in the clinical score or body weight at the peak of acute EAE, as previously reported (Fig. 3a EPSPs were recorded from EAE mice following the same procedure as for the healthy mice. Table 1 summarizes the baseline cell parameters, EPSP amplitude and AUC of CA1 neurons from healthy and EAE, Trpm4 +/+ and Trpm4 -/mice. There were no differences due to genotype or health in nearly all parameters (two-way ANOVA). Only cell resistance showed a significant difference due to genotype (Table 1; F (1,42) = 6.38, p = 0.015) and between healthy and EAE mice (Table 1; F (1,42) = 11.54, p = 0.0015) and a post-hoc Sidak's test for multiple comparisons revealed a significant difference between healthy Trpm4 +/+ and EAE Trpm4 +/+ (Table 1; p = 0.046). Interestingly, in the EAE slices there was a significant negative correlation of cell resistance with clinical score in Trpm4 +/+ neurons that was absent in Trpm4 -/neurons ( Fig. 3d; Trpm4 +/+ : www.nature.com/scientificreports/ As in healthy animals, 9-phenanthrol had no apparent effect on EPSP amplitude or AUC in slices made from EAE Trpm4 +/+ mice ( Fig. 4a,b, Supplemental Fig. 2; amplitude p = 0.63; AUC p = 0.25). Given there was no drug effect in Trpm4 +/+ neurons, we had no need to test for off-target effects and elected not to treat slices from Trpm4 -/animals with 9-phenanthrol. In contrast to 9-phenanthrol, glibenclamide significantly though modestly reduced the EPSP amplitude ( TRPM4 mRNA levels do not change in EAE. As the recordings from EAE mice were suggestive of a change in TRPM4 function, localization or regulation during EAE, we were specifically interested in whether neuronal TRPM4 expression might be altered in EAE. We isolated cortical neuron nuclei from healthy and EAE mice and performed a quantitative polymerase chain reaction analysis to determine mRNA expression levels of TRPM4. Interestingly, TRPM4 mRNA levels did not change under EAE (Fig. 5, t-test, p = 0.19) . Given no change in mRNA levels between healthy and EAE mice, we sought to determine whether we could detect differences in TRPM4 localization or protein levels. We first performed immunohistochemistry in hippocampal CA1 from healthy Trpm4 +/+ and Trpm4 -/mice.We used 3 commercially available antibodies and one antibody generously provided by Simard 24,35 . The Alomone antibody was raised in rabbits against a short peptide of amino acids (aa) 5-17 of human TRPM4 (intracellular N-terminus), the Abcam antibody was raised in rabbits  www.nature.com/scientificreports/ against a synthetic peptide from the C-terminal region of mouse TRPM4, the Santa Cruz antibody was raised in goat against a peptide mapping within an internal region of human TRPM4, and the antibody from Dr. J. M. Simard's lab was raised in rabbits against aa 1-612 of mouse TRPM4 (intracellular N-terminus). Disappointingly, none of the four antibodies showed differential staining patterns in hippocampal slices of Trpm4 +/+ and Trpm4 -/mice although it appeared there was clear staining around the soma and dendrites (Supplemental Fig. 3). Thus, we could not determine the expression pattern of TRPM4 protein in CNS tissue and did not attempt to stain tissue from EAE mice.

Discussion
In this study, we found that TRPM4 does not participate in basal synaptic transmission at the CA3-CA1 Schaffer collateral synapse in organotypic hippocampal slice cultures or in acute hippocampal slices made from healthy mice. We did however, detect a small TRPM4-dependent effect at the acute phase of EAE. We undertook this study expecting to find that TRPM4 indirectly boosted synaptic EPSCaTs and directly increased EPSPs in the hippocampus, particularly when synaptic stimulation was strong enough to cause dendritic calcium rises typical of NMDA spikes or large EPSPs 41 . We blocked postsynaptic action potentials, inhibitory GABA A receptors and SK channels, which are also activated by calcium and hyperpolarize rather than depolarize spines and dendrites, and still did not see evidence of a TRPM4 conductance in slices from healthy mice. Surprisingly, the antagonist glibenclamide, but not 9-phenanthrol, decreased EPSP amplitude in EAE Trpm4 +/+ mice. Importantly, glibenclamide did not affect EPSPs in EAE Trpm4 -/mice. Glibenclamide is considered less specific to TRPM4 than 9-phenanthrol 42 and glibenclamide is widely known for its ability to antagonize SUR1-Kir6.2 (i.e. K ATP ) channels 43 . It has been shown that SUR1-TRPM4 channels also co-assemble and that glibenclamide has a 100-fold higher efficacy for SUR1-TRPM4 channels than for TRPM4 homomeric channels 35 . Interestingly, in hypoglycemic conditions glibenclamide decreases EPSCs in cortical neurogliaform cells but not pyramidal neurons via K ATP channels 44 . While cortical and hippocampal pyramidal neurons should not be considered identical, they are similar morphologically and electrophysiologically and TRPM4 seems not to participate in excitatory synaptic transmission in both. We cannot rule out that K ATP (SUR1-Kir6) channels Table 1. Only cell resistance significantly varies between genotype or health status at baseline. All measurements are taken from the baseline prior to drug wash-in. Data are mean ± SEM. The degrees of freedom are (1,42) for all parameters except clinical score and body weight. Upper pair of F and p values are for genotype (Trpm4 +/+ vs Trpm4 -/-) comparisons and the lower pair are for treatment (EAE vs. healthy) comparisons. For clinical score and body weight, unpaired t-tests were used. n-size is indicated within each cell.

Cell parameters at baseline
Healthy Trpm4 +/+ Healthy www.nature.com/scientificreports/ are also down-regulated in the Trpm4 -/mice although we think it more likely that glibenclamide in EAE reduces EPSPs via SUR1-TRPM4 or TRPM4 channels in the slices from EAE mice. Although TRPM4 mRNA is present in neurons 1,28 we were unable to detect TRPM4 protein in brain slices by immunostaining. The fluorescent signals seen with several antibodies was unspecific as it was the same in slices from Trpm4 -/and Trpm4 +/+ mice. It may be that protein expression of TRPM4 is simply too low in the hippocampus to be detected by immunohistochemistry. Alternatively, antigenic sites may be masked in tissue, or the antibodies recognize some endogenous protein that is not TRPM4 and is therefore present in both the wild-type and TRPM4-lacking mice. A recent study demonstrated somato-dendritic immunostaining against TRPM4 in CA1 neurons of the hippocampus, however, there was no published validation of these antibodies in Trpm4 -/tissue 21 .
Our findings that TRPM4 plays little to no role in hippocampal synaptic transmission, support a recent study, describing a direct coupling of NMDA receptors with TRPM4 28 . When this coupling, which is suggested to underlie the role of TRPM4 in excitotoxicity, is disrupted there is no change in calcium influx or in basal synaptic transmission. Although TRPM4 antagonism is a different action than disruption of NMDA/TRPM4 interaction, in neither ours nor their study was there an apparent involvement of TRPM4 in normal synaptic transmission. The NMDA/TRPM4 coupling is hypothesized to occur specifically at extrasynaptic sites 28  www.nature.com/scientificreports/ to only subtle inflammatory changes in the hippocampus during EAE. Interestingly, we did not observe a change in the mRNA levels of TRPM4 in cortical pyramidal neurons between healthy and EAE animals. While this does not preclude an increase in TRPM4 protein in hippocampal pyramidal neurons, it suggests that a re-organization rather than upregulation may occur as we observed in spinal cord 1 .
In conclusion, we observed that TRPM4 does not significantly contribute to basal synaptic transmission at the Schaffer collateral synapse of healthy mice, but TRPM4 may contribute to EPSPs in EAE. In multiple sclerosis and EAE there are clear hippocampus-specific cognitive impairments, synaptic dysfunction and atrophy both in patients and in animal models 46,47 . TRPM4 might prove an interesting therapeutic target if it is found that cognitive function after i.e. EAE is improved by TRPM4 inhibitors as in chronic cerebral hypoperfusion 48 . The relatively small contribution to synaptic transmission may cause fewer side effects and prove advantageous over direct targeting of, for instance, NMDA receptors.

Materials and methods
Animals. All transgenic mice (Trpm4(tm1.1Mfre) 49 and wild type C57BL/6 were housed and bred at the University Medical Center Hamburg-Eppendorf (UKE) animal facility. Both female and male mice were used as indicated. All mice were housed with a 12-h light/dark cycle and had water and food ad libitum. All procedures performed in mice were in compliance with German law and according to the guidelines of Directive 2010/63/ EU. Protocols were approved by the local ethics committee (Behörde für Justiz und Verbraucherschutz (BJV), Lebensmittelsicherheit und Veterinärwesen). The ARRIVE guidelines for reporting the experiments were followed. No a priori determination of sample size was performed.
Mouse organotypic hippocampal slice cultures. Hippocampal slice cultures were prepared from P4-P7 male wild type C57BL/6 mice following a published protocol 50 . Pups were anesthetized with 80% CO 2 /20% O 2 and decapitated. Hippocampi were dissected in cold slice culture dissection medium containing (in mM): 248 sucrose, 26 NaHCO 3 , 10 glucose, 4 KCl, 5 MgCl 2 , 1 CaCl 2 , 2 kynurenic acid and 0.001% phenol red. The solution was saturated with 95% O 2 , 5% CO 2 , pH 7.4, 310-320 mOsm kg −1 . Tissue was cut into 400 µM thick sections on a tissue chopper and cultured on porous membranes (Millipore PICMORG50) at 37 °C in 5% CO 2 . No antibiotics were added to the slice culture medium which was partially exchanged (60-70%) twice per week and contained Plasmid construction and expression. DNA encoding GCaMP6f and tdimer2 were each subcloned into a neuron-specific expression vector (pCI) under the control of the human synapsin-1 promoter using the Hin-dIII and NotI restriction sites for GCaMP6f and the Acc65I and EcoRI restriction sites for tdimer2. To express GCaMP6f and tdimer2 in CA1 pyramidal neurons, gold particles (1.6 μm, 2.75 μg DNA per mg gold) were coated with an 8:3 ratio of expression vectors encoding GCaMP6f and tdimer2 (respectively). At DIV 7-9, slice cultures were ballistically transfected using a Helios gene gun (Bio-Rad). Experiments were conducted between DIV 14-21.
Two-photon microscopy. The two-photon imaging setup was custom built based on an Olympus BX51WI microscope. An Olympus LUMPlan W-IR2 60 × 0.9 NA objective was used and image acquisition was controlled by the open-source software package ScanImage 3.7 51 , which we modified to allow user-defined arbitrary line scans. A pulsed Ti:Sapphire laser (Chameleon, Coherent) was used to excite GCaMP6f and tdimer2 at 980 nm. Emitted photons were collected through the objective and oil-immersion condenser (1.4 NA, Olympus) with two pairs of photomultiplier tubes (H7422P-40, Hamamatsu). 560 DXCR dichroic mirrors and 525/50 and 607/70 emission filters (Chroma) were used to separate green and red fluorescence. Excitation light was blocked by short-pass filters (ET700SP-2P, Chroma).
Calcium imaging. An organotypic hippocampal slice culture was transferred to the chamber at the twophoton microscope and bathed in oxygenated ACSF containing (in mM): 135 NaCl, 2.5 KCl, 4 CaCl 2 , 4 MgCl 2 , 10 Na-HEPES, 12.5 D-glucose, 1.25 NaH 2 PO 4 . To identify spines and dendrites responding to electrical stimulation of Schaffer collateral axons (two 0.2 ms paired pulses with an interpulse interval of 40 ms), frame scans (at 15.625 Hz) of oblique dendrites were acquired. After finding a responding spine, user-defined arbitrary line scans (500 Hz) that crossed the responding spines and dendrite were used to quantify the calcium response to stimulation, improving time resolution and signal to noise. Responses to stimulation were acquired at 0.0167 Hz. A stable baseline was acquired for at least twenty minutes. After baseline, 9-phenanthrol (30 μM) was washed into the chamber and responses were acquired for another twenty minutes. Finally, fresh ACSF containing no drugs was washed into the chamber and responses were acquired for a final twenty minutes. For analysis, regions of interest (ROI) were drawn over spines and dendrites in ImageJ and the relative percent change in GCaMP6f. green fluorescence was calculated as 100 * (F-F 0 ) / F 0 where F is the fluorescence intensity and F 0 is the average fluorescence intensity prior to stimulation. Peak amplitude and area under the curve (AUC) were extracted from within the response time-window after smoothing in Matlab with a moving average span of 5. Baseline, drug and washout responses over each twenty minute period were then averaged (Fig. 1). www.nature.com/scientificreports/ Immunohistochemistry. Trpm4 +/+ and Trpm4 -/mice were anaesthetized with an intraperitoneal injection of 10 mg ml −1 esketamine hydrochloride (Pfizer), 1.6 mg ml −1 xylazine hydrochloride (Bayer) in water (100 µl per 10 g of body weight). Mice were then perfused with 0. Slices were then kept in the same solution at room temperature until used. After at least 1 h at room-temperature, a slice was placed in the recording chamber and continuously perfused with oxygenated ACSF supplemented with D-Serine (30 μM) to occupy the second agonist binding site of NMDA receptors, bicuculline-methochloride (10 μM) to block GABA A receptors and SK-channels 38 and 0.1% DMSO (see below). Recordings were performed using either a Multiclamp 700B or an Axopatch 200B amplifier (Molecular Devices). Recordings were controlled and digitized using National Instruments A/D boards and Ephus software in the Matlab environment 52 . Current-clamp recordings from CA1 pyramidal cells were performed at 23-25 °C (Figs. 2, 3) or at 30-32 °C (Fig. 4). Patch pipettes with a tip resistance of 3-5 MΩ were filled with (in mM): 135 K-gluconate, 10 HEPES, 4 MgCl 2 , 4 Na 2 -ATP, 0.4 Na-GTP, 10 Na 2 -phosphocreatine, 3 L-ascorbic acid and 3 QX-314 chloride (to block voltage-gated sodium channels). For recordings done in healthy Trpm4 +/+ and Trpm4 -/mice, the experimenter was not blind to genotype. Recordings made from EAE mice were performed blind to the genotype. To evoke excitatory postsynaptic potentials (EPSPs), a monopolar electrode was placed in the stratum radiatum and 0.2 ms pulses were delivered using an ISO-Flex stimulator (A.M.P.I.). The stimulation intensity was set to produce an EPSP of approximately 40 mV. After a stable baseline was achieved (at least five minutes), a TRPM4 antagonist, either 9-phenanthrol (30 μM) or glibenclamide (20 μM), was washed into the bath. Both 9-phenanthrol and glibenclamide were dissolved in DMSO and resulted in a final 0.1% DMSO total concentration in the ACSF. An equivalent concentration of DMSO was included in the drug-free ACSF. Trials were acquired at 0.08-0.1 Hz.

Analysis & statistics.
Calcium imaging experiments were analyzed using ImageJ and repeated-measures ANOVA was used for statistical analysis. Electrophysiology analysis was completed in Matlab. Individual trials were then averaged into one minute bins. For statistics, averages were made from the last five minutes of baseline and within 15-22 min after washing in the drug. Analysis of electrophysiological recordings from EAE ani- www.nature.com/scientificreports/ mals was conducted blind. All statistics were performed using Graphpad Prism (version 8). For paired analyses, the Wilcoxon matched pairs signed rank test was used. For analysis of the role of genotype and clinical score, a two-way ANOVA was used for each parameter. p less than 0.05 was considered significant. 5000 bootstrap samples were taken; the confidence interval is bias-corrected and accelerated. To calculate estimation statistics, multiple paired differences were calculated using the estimation statistics app 53 . The P-value(s) reported are the likelihood(s) of observing the effect size(s), if the null hypothesis of zero difference is true. For each permutation P-value, 5000 reshuffles of the control and test labels were performed.