Summary
Rehabilitation is the most effective treatment for promoting the recovery of motor deficits after stroke. One of the most challenging experimental goals is to unambiguously link brain rewiring to motor recovery prompted by rehabilitative therapy. Here, we investigated which aspects of cortical remodeling are induced by rehabilitation by combining optical imaging and manipulation tools in a mouse model of stroke. We revealed that the stabilization of peri-infarct synaptic contacts fostered by rehabilitation goes along with increased vascular density induced by angiogenesis. Furthermore, we showed the progressive formation of a new motor representation in the peri-infarct area where temporal and spatial features of cortical activation recovered towards pre-stroke condition. In the same animals we observed the reinforcement of inter-hemispheric connectivity after rehabilitation. The present work provides the first evidences that rehabilitation promotes the combined recovery of structural and functional features distinctive of healthy neuronal networks.
Introduction
Every year, several million stroke victims worldwide are impaired by long-term disability, imparting a large societal and personal burden in terms of lost productivity, lost independence and social withdrawal (Mozaffarian et al., 2015). When stroke affects the motor-associated cortices, the recovery of motor function is often partial, owing to the limited degree of spontaneous brain repair (Carmichael et al., 2017). Strategies that promote brain plasticity, like pharmacological treatments and physical training, can enhance neural rewiring and dramatically improve functional motor recovery. Pharmacological treatment both in the peri-infarct and contralesional areas has been shown to aid the restoration of function (for instance, see (Clarkson et al., 2010)). Another highly effective rehabilitative approach after stroke is physical training (Bütefisch, 2006; Jones and Adkins, 2015). The effectiveness of rehabilitative therapies can be maximized through a combination of treatments (Adkins-Muir and Jones, 2003; Dancause and Nudo, 2011; Hesse et al., 2007; Plautz et al., 2003). A few animal studies provided valuable insights into the anatomical adaptations in rodents and primates induced by coupled therapies (Fang et al., 2010; Lee et al., 2004; Plautz et al., 2003; Wahl et al., 2014). In a recent study on mice, Spalletti et al. took advantage of a rehabilitation paradigm that combines motor training and pharmacological inhibition of the contralesional primary motor cortex (M1) with Botulinum Neurotoxin E (BoNT/E) (Spalletti et al., 2017). BoNT/E injection reduced the excessive transcallosal inhibition exerted from the healthy to the stroke side (Spalletti et al., 2017). The combination of motor training and BoNT/E silencing was superior to either treatment alone in promoting recovery of motor skills in stroke mice. Importantly, this combined therapy led to motor improvements that generalized to multiple motor tasks (Spalletti et al., 2017). The induction of a generalized functional gain, i.e. the recovery of motor functions beyond the ones that are trained, is crucial when evaluating the efficacy of rehabilitative therapies.
In the last decade, fluorescence imaging provided valuable insights into spontaneous cortical plasticity after stroke. Structural plasticity after stroke has been evaluated by two-photon fluorescence (TPF) microscopy (Brown et al., 2007b; Johnston et al., 2013; Mostany et al., 2010; Sakadzic et al., 2015) showing that the extent of synaptic recovery varied along the distance from the infarct core (Sigler and Murphy, 2010) and that redistribution of blood vessel and dendrite orientation was confined to the peri-infarct area (Brown et al., 2007a). In parallel, Murphy and colleagues performed a number of acute optical imaging studies showing the persistence of cortical functional remapping over months after stroke (Brown et al., 2009; Harrison et al., 2013). The availability of a new generation of experimental tools, like optogenetics (Ayling et al., 2009; Lim et al., 2014) and genetically encoded functional indicators (Chen et al., 2013) provided new means of exploring neuronal rewiring.
To our knowledge, no investigation has yet described how rehabilitation after stroke affects cortical structural and functional plasticity in vivo and on the long-term scale. Many unanswered questions need to be addressed, like how rehabilitation molds neuronal structural plasticity and cortical functional maps. Moreover, it is still to be defined if neuronal and vascular plasticity act in concert to aid functional recovery.
In the present study, we dissected how rehabilitative treatment activates concurrent modalities of cortical plasticity in mice by using a combination of cutting-edge optical techniques. Our experiments revealed that a combination of physical training with pharmacological manipulation of the contralesional activity helps the preservation of dendritic architecture along with stabilization of spines in the peri-infarct region. We found in the same double-treated subjects the increase in the density of blood vessels primarily focused in the peri-infarct area. By longitudinally monitoring the calcium functional maps, we found that combined rehabilitation prompted the recovery of essential features of pre-stroke activation profiles, both in spatial and temporal terms. In the same animals presenting large-scale remapping of the injured hemisphere, we observed a significant enhancement of transcallosal functional connectivity after one month of rehabilitative therapy. Overall, our study provides an unprecedented view of complementary aspects of cortical anatomical and functional adaptation induced by rehabilitation after stroke.
Results
Experimental design
We used a photothrombotic model of focal stroke applied to the right M1 of GCaMP6f or GFPM mice (STROKE, TOXIN, ROBOT and REHAB groups). To restore the interhemispheric balance perturbed by stroke, we performed a focused inactivation of the forelimb motor cortex in the hemisphere contralateral to the ischemic lesion by means of intracortical injections of the synaptic blocker Botulinum Neurotoxin E (BoNT/E, TOXIN group). BoNT/E is a bacterial enzyme that enters synaptic terminals and reversibly blocks neurotransmission by cleaving SNAP-25, a main component of the SNARE complex (Caleo et al., 2007). We next tested the impact of physical rehabilitation by daily training the affected forelimb of stroke mice with controlled, repeatable and targeted exercises, guided by a robotic platform, namely the M-Platform (ROBOT group, (Spalletti et al., 2014)). Rehabilitation-associated training on the M-Platform consisted of repeated cycles of passively actuated contralesional forelimb extension followed by its active retraction triggered by an acoustic cue (Figure 1A). Finally, we tested the synergistic effect of combined therapy. The combined rehabilitation paradigm of the REHAB group consists in a combination of motor training on the M-Platform and pharmacological inactivation of the contralesional hemisphere via BoNT/E (Figure 1B).
ROBOT and REHAB mice were trained for 4 weeks starting 5 days after injury (Figure 1C), in line with the overall consensus that the initiation of rehabilitative training 5 or more days after stroke is mostly beneficial and has no adverse effects (Krakauer et al., 2012). The motor task was rapidly learned and easily performed. Indeed, the amplitude and slope of the force peaks exerted during the voluntary forelimb-pulling task were not significantly different across groups, neither within a week nor across weeks of training (Figure 1D).
Combined rehabilitation preserves pyramidal neurons architecture and promotes the stabilization of synaptic contacts in the peri-infarct area
In the present study we assessed how different rehabilitative approaches prompt large- and small-scale cortical plasticity. In particular, we wondered if the improved performance in generalized motor tasks induced exclusively by the combined treatment (Spalletti et al., 2017) resulted in distinct features of cortical reorganization. We investigated synaptic dynamics, vascular network remodeling and remapping of motor representation.
First, we explored structural rewiring of dendrites and spines. Heightened spontaneous structural remodeling was observed in mice at both presynaptic (axonal fibers and terminals) and postsynaptic (dendrites and dendritic spines) sites in the peri-infarct cortex in the weeks following stroke (Brown et al., 2010; Carmichael et al., 2001; Dancause et al., 2005; Hsu and Jones, 2006; Ueno et al., 2012). In this set of in vivo experiments, we performed TPF microscopy of pyramidal apical dendrites and axons (0-100 μm deep from the pial surface) with increasing distances (up to 4 mm) from stroke core in the cortex of GFPM mice (Figure S1A). Orientation of dendrites and axons was evaluated on the 4th week after performing the lesion (STROKE, TOXIN, ROBOT and REHAB mice).
A strong orientation gradient of dendrites induced by the collapse of dead tissue in the ischemic core was evident in the STROKE group (Figure 2A, right panels and Figure 2B). Reorientation was not detectable at distal regions from the ischemic core (>1 mm; Figure S1B and S1C). A strong alignment of neurites in the peri-infarct region was observed in stroke mice treated with BoNT/E (TOXIN group, Figure 2B). Four weeks of robotic training (ROBOT group) were not sufficient to prevent the formation of this structural abnormality. In contrast, reorientation was not visible in REHAB animals, nor in proximal (Figure 2B) or distal (Figure S1C) regions. Indeed, the randomness in dendrite orientation in REHAB mice resembled the one observed in healthy (CTRL) subjects (Figure 2B).
Alterations induced by ischemic damage on neural circuitry are known to affect dendrites as well as synaptic contacts, producing a large increase of spine turnover in peri-infarct area of spontaneously recovering mice (Brown et al., 2007a; Mostany et al., 2010). We next asked how different rehabilitative approaches modified synaptic turnover. We thus monitored the appearance and disappearance of apical dendritic spines (Figure 2C) by performing a frame-by-frame comparison of spines in the mosaic we acquired along the rostro-caudal axis with TPF microscopy. We first evaluated the peri-infarct plasticity of pyramidal neurons in spontaneously recovering mice (STROKE) and verified that this focal lesion triggers an increased instability of synaptic contacts. Surprisingly, the single treatments by inhibition of the contralesional hemisphere or physical therapy alone were not able to reinstate physiological levels of spine plasticity (Figure 2D).
On the contrary, spines in the peri-infarct region of REHAB animals exhibited increased synaptic stability (Figure 2D, left panel) and a lower turnover (Figure 2D, middle panel) than STROKE mice, that were comparable to healthy CTRL values. The combined rehabilitative treatment also resulted in higher synaptic densities thereby recapturing an important feature of pre-stroke conditions (Figure 2D, right panel). The stabilization of synaptic contacts induced by rehabilitation was stronger at the proximal level and weakened with increasing distance from ischemic core (Figure S1D, upper and middle panels). Interestingly, the change in density of spines extended beyond peri-infarct region (Figure S1D, lower panel).
In brief, longitudinal imaging of cortical neurons revealed that the combined rehabilitative therapy was necessary to preserve the organization of dendritic arbors in peri-infarct cortex and to restore dendritic spine plasticity.
Combined rehabilitative treatment stimulates angiogenesis in the peri-infarct area
Since stroke profoundly altered the spatial distribution of superficial blood vessels (Figure 2A, left panels), we wondered if and how rehabilitation altered vascular milieu in the ipsilesional cortex. To test this, we longitudinally observed mouse cortices of all groups under both cranial window and thinned skull preparations for one month after stroke. We found that the injured area progressively shrunk owing to the collapse of dead tissue in STROKE and REHAB mice (Figure 2A, 3A and S2). In addition, a very bright area appeared in the peri-infarct region of both groups of GCaMP6 mice, possibly representing an excitotoxic response associated with calcium dysregulation (Ankarcrona et al., 1995) elicited by the photothrombotic stroke (see the left panel of Figure 3A; 13 out of 14 STROKE and REHAB mice). The enhanced brightness gradually diminished and disappeared after the acute period (6-19 days after injury in STROKE and REHAB mice; see examples in Figure 3A, right panel and Figure S2), and was accompanied by a large shrinkage of the necrotic tissue. The consequent displacement of the peri-infarct area was associated with a substantial remodeling of blood vessels (white arrowheads in Figure 3A), in agreement with previous studies (Brown et al., 2007a).
We quantified the structural reshaping of cortical vasculature with high resolution by performing 3D reconstructions with TPF microscopy. Specifically, after the last motor training session, we stained the brain vasculature of GFPM mice with a vessel-filling fluorescent dye, albumin-fluorescein isothiocyanate (FITC, Figure 3B). By imaging Thiodiethanol (TDE) cleared cortical slices (see Methods), we obtained high-resolution 3D maps of blood vessels of the injured, right hemisphere (Figure 3C, Video S1). The reorganization of cortical tissue (Figure 2A) that led to a considerable alignment of dendrites towards the stroke core produced an analogous re-orientation of blood vessels in peri-infarct area of STROKE mice (Figure 3D). Interestingly, neither the single treatment (TOXIN and ROBOT mice) nor combined rehabilitation (REHAB) could rescue random vascular orientation of pre-stroke conditions in the peri-infarct (proximal, <500 μm) region (Figure 3D). A less pronounced orientation was visible in all groups at distal regions (1000 to 1500 μm from core, Figure S3A).
We further guessed if rehabilitation triggered angiogenesis in the ipsilesional cortex. We found that the single treatments (TOXIN and ROBOT groups) did not have any significant impact on blood vessel numbers (Figure 3E, left panel). Conversely, the density of blood vessels near the infarct site was significantly higher in REHAB mice compared to both STROKE and CTRL groups.
We then evaluated the vascular density over the entire cortical depth to define layer-specific contributions to the overall increase in blood vessel density. Blood vessel distribution was quite uniform in CTRL mice. Increases in density were localized mainly to the middle layers of the cortex (300-500 μm deep) in REHAB group (Figure 3E, middle and right panels). We found that the augmented vascular density was mainly observed in regions near the stroke core (Figure 3E, right panel), whereas distal regions did not show any significant increase (Figure S3B).
We then asked if the increased density was due to an increase in average diameter of blood vessels. Our results show that there is no significant difference in distribution of lumen sizes of cortical blood vessels for the five experimental groups (Figure 3F), ruling out possible influences of vasodilation or vasoconstriction in the evaluation of density. Alternatively, the increased density could be due to ramification of existing vessels, leading to an increase in the complexity of the vascular maps. Nevertheless, the evaluation of the number of branching points showed no significant differences between the spontaneously recovering mice (STROKE) and all the treatments (TOXIN, ROBOT, REHAB) nor in proximal (Figure 3G) or distal (Figure S3C) regions, thus ruling out this hypothesis.
Finally, we asked if rehabilitation fostered an increase in endothelial proliferation that eventually led to augmented vascular density. We thus compared a new group of REHAB mice after 10 sessions of daily robotic training with spontaneously recovering mice (STROKE group) 16 days after stroke. To evaluate endothelial proliferation, we performed ex vivo double labeling on brain slices with CD31 and Ki67 (see Supplementary Material section). Figure 3H shows that the number of double stained cells is higher in the peri-infarct region of the cortex of REHAB compared to STROKE mice. The augmented endothelial cell proliferation in combination with the increased density of blood vessels (higher than healthy CTRL mice) and the non-significant change in lumen diameter in REHAB mice strongly suggest that the combined treatment promotes angiogenesis in the peri-infarct area.
Taken together, these results on cortical vasculature indicate that the combination of pharmacological and physical therapy triggers substantial blood vessel growth via the promotion of endothelial cell proliferation in the peri-infarct region.
In conclusion, our results suggest that generalized functional recovery is associated with a pro-angiogenic effect on the vasculature of the peri-infarct area induced by the combined rehabilitative treatment.
Combined rehabilitation treatment restores cortical activity patterns disrupted by stroke
We next assessed if combined rehabilitation modulated cortical motor representation activated by voluntary movement of the injured forelimb. We implemented an integrated system for simultaneous imaging of the calcium indicator GCaMP6f over the injured hemisphere and recording of forces applied by the contralesional forelimb during the training sessions on the M-Platform (Figure 4A-C). Calcium imaging was used as a measure of cortical activity in the brains of GCaMP6f mice. We focused on the analysis of calcium response activated in the same time-window of the voluntary retraction movement (see example in Figure 1C, right panel). The imaging area overlapped with the region visualized in GFPM mice (Figure 4B). Wide-field calcium imaging showed that a small area located in the motor-sensory region reproducibly lit up in CTRL mice during the forelimb retraction movement on the M-platform (examples of cortical activation are reported in Figure 4D and Figure S4A). On the contrary, a large area covering most of the cortical surface of the injured hemisphere was activated while performing the task in non-treated (STROKE) mice one month after stroke. Remarkably, calcium activation in REHAB mice was similar to healthy controls (CTRL) in terms of extension, location, timing and amplitude (Figure 4D-H, Figure S4A-D).
We analyzed the extension and location of the motor representation by overlapping the movement-triggered activation maps obtained on every day of the training week (ROIg, see Figure S4B-C and Supplemental Procedures). Stroke expanded the motor representation from M1 toward more caudal regions not specifically associated with motor control (STROKE group in Figure S4C, 2nd panel). Interestingly, the extension of motor representation in REHAB mice was consistently reduced (Figure S4C 3rd panel and Figure S4D). In terms of location, the motor representation of REHAB mice was centered on the motor-associated region in the peri-infarct area (Figure S4C 3rd panel and Figure S4D). The localization and spread of motor maps changed progressively along the weeks of training (Figure S4F). In most cases (5 out of 6 REHAB mice) we found a higher correlation (i.e. augmented functional connectivity) in the activity of spared motor-associated areas (Figure S4E) at the end of the training period (REHAB/4W) compared to the beginning (REHAB/1W). We hypothesized that focalized versus spread motor representations could be associated with different patterns of propagation of calcium activity. To quantify the concurrent recruitment of motor-associated and other functional areas, we analyzed the temporal profile of movement-triggered calcium transients over the injured hemisphere. We assumed that an extended activation (as in STROKE mice) implied a synchronicity in the activation of the rostral peri-infarct and the caudal areas. Indeed, we found that the delay between the maximum calcium peak of the rostral and caudal regions was slightly negative in STROKE mice, indicating that the activation of the caudal region somewhat preceded the activation of the rostral (peri-infarct) one (Figure 4E). Conversely, the delay was positive in CTRL and REHAB mice. Longitudinal imaging showed that the temporal pattern of rostro-caudal calcium activation in REHAB mice was gradually recovered towards pre-stroke conditions during 4 weeks of training (Figure S4G, left panel).
To sum up, the combined rehabilitative treatment focalized the motor representation to the peri-infarct region. The motor representation of REHAB mice closely resembled the one observed in CTRL animals both spatially and temporally.
We further analyzed calcium transients in peri-infarct area during contralateral forelimb retraction in CTRL, STROKE, and REHAB mice. At a glance, the average traces in Figure 4F show that stroke (in STROKE mice) and the combined treatment (REHAB mice) modified several features of the calcium transients. While a significant reduction in the amplitude of calcium transients during forelimb retraction was evident one month after stroke in the STROKE group, REHAB mice partially recovered to pre-stroke conditions (Figure 4G). In detail, the combined treatment led to a progressive increase in the amplitude of the calcium response over the training period (Figure S4G, middle panel). Moreover, REHAB animals showed progressively steeper slope of calcium activation along the weeks of training (Figure S4G, right panel). As compared to STROKE animals, a faster rise of calcium transient in the peri-infarct region was observed in REHAB mice at the end of the training period (Figure 4H).
In brief, our results showed that combined rehabilitative treatment promoted the formation of a new motor representation in the peri-infarct area where temporal and spatial features of cortical activation recovered towards pre-stroke condition.
Combined rehabilitation improves inter-hemispheric functional connectivity
We hypothesized that the combined treatment alters the functional connectivity of the new motor representation with the contralesional motor cortex. Several studies in mice and humans have shown that interhemispheric M1 connectivity is reduced after stroke (Bauer et al., 2014; van Meer et al., 2010a; van Meer et al., 2010b). Therefore, it is hypothesized that increased interhemispheric connectivity positively correlates with the recovery of motor performances in the subacute stage after stroke in humans (Carter et al., 2010). Based on this hypothesis, we tested whether transcallosal projections were altered by combined rehabilitation by using an all-optical approach that combined optogenetic activation of the intact M1 with calcium imaging on the injured hemisphere. In these experiments, the intact (left) M1 of GCaMP6f mice was injected with AAV9-CaMKII-ChR2-mCherry to induce the expression of ChR2 in pyramidal neurons (Figure 5A).
One month after stroke, we optogenetically stimulated the intact M1. To avoid ChR2 activation while exciting GCaMP6f fluorescence, we partially occluded the 505 nm LED in our custom-made wide-field microscope (Figure 5B-C; see also (Conti et al., 2019)). Optogenetic stimulation was achieved with a second excitation path in which a 470 nm laser was focused on the AAV-transfected region via acousto-optic deflectors (AODs) (Crocini et al., 2016).
Laser stimulation of the contralesional M1 reproducibly triggered the activation of calcium transients in the injured hemisphere (Figure 5D, Video S2), mainly spreading from the homotopic M1 of CTRL mice, or in the peri-infarct area of STROKE and REHAB animals (Figure S5). Cortical activity then propagated to functionally connected regions that were located either anterior or posterior to the primary source of activation in all 3 groups of animals (Video S2). By quantifying the delay between the start of the optogenetic stimulation on the left hemisphere and the peak of calcium activity in the right hemisphere, we found no significant difference between the STROKE and CTRL mice (Figure 5E). In contrast, this delay was significantly reduced in REHAB mice as compared to both CTRL and STROKE mice (Figure 5E). In addition, the success rate of contralateral activation in response to optogenetic stimulation was higher in REHAB mice than in CTRL and STROKE animals (Figure 5F). Therefore, this series of all-optical experiments suggests that combined rehabilitation strengthens the functional connectivity between the spared motor cortex and the perilesional cortex one month after stroke.
Discussion
In the present study we showed that rehabilitative treatment combining physical training and BoNT/E inhibition of contralesional hemisphere stabilized perilesional synaptic contacts, restored features of cortical activation typical of pre-stroke conditions and promoted the formation of an enriched vascular environment to feed the neuronal population that has been recruited to perform motor control. These results offer novel insights on cortical plasticity induced by rehabilitation in vivo. To the best of our knowledge, this study provides first-time evidence on the correlation between rehabilitation-induced neuronal and vascular reshaping in the mouse cortex after stroke.
The combined rehabilitation paradigm used here was recently characterized in terms of behavioral assessment of motor recovery (Spalletti et al., 2017). The authors previously developed the M-Platform, a mechatronic device for mouse forelimb training (Spalletti et al., 2014) that mimics a robot for upper limb rehabilitation in humans, the “Arm-Guide” (Reinkensmeyer et al., 2001). In Spalletti et al., 2017, robotic rehabilitation paradigm per se was not able to generalize recovery to untrained motor tasks (i.e. Schallert Cylinder test). They also showed that BoNT/E silencing, active in the motor cortex for about 10 days after injection, produced a small and transient improvement in general forelimb motor tests (Gridwalk and Schallert Cylinder test). Nevertheless, the guide of an appropriate motor rehabilitation regime was necessary to achieve a complete functional recovery. Here we show that incomplete motor recovery produced by the single treatments is associated with non-physiological synaptic turnover and reduced spine density. The synergy between BoNT/E treatment and robotic training on other hand was able to restore the stability and density of spines of pre-stroke conditions. Finally, our results show that the generalized recovery of motor function exclusively induced by the combined therapy was associated with increased vascular proliferation in the peri-infarct area.
A commonly agreed hypothesis is that the genesis of an enriched vascular milieu around the infarct area might promote synaptogenesis and dendritic remodeling, in addition to axonal sprouting (Brown et al., 2010; Carmichael, 2006; Kelley and Steward, 1997). In parallel, a recent review by Wahl and colleagues suggested that rehabilitative training might shape the spared and the new circuits by stabilizing the active contacts (Wahl et al., 2014). In support of these hypotheses, we show that augmented vascular density induced by the combined rehabilitative treatment is associated with increased density of synaptic contacts and stabilization of synaptic turnover. Although there are no other in vivo studies examining rehabilitation-induced spine plasticity after stroke to make a direct comparison with, our results are in agreement with recent postmortem histological studies where rehabilitation was shown to determine significant increases in spine density of distal apical dendrites in corticospinal neurons (Wang et al., 2016).
In this study, we have shown for the first time that rehabilitation promotes the recovery of structural features of healthy neuronal networks. Only the combined therapy could rescue the random orientation of pyramidal neurons while supporting the stabilization of synaptic contacts. Interestingly, in the same REHAB mice where stroke profoundly altered the spatial distribution of blood vessels around the core (Figure 3D), the randomness of dendrite orientation resembled healthy (CTRL) mice (Figure 2B). Further analysis on the acute phase after stroke will allow understanding if random neuronal orientation is maintained or later recovered by rehabilitation.
We hypothesized that the neural plasticity we observed in rehabilitated mice could be supported by revascularization. This assumption is based on the relationship existing between the proangiogenic state and neurological improvement in patients with stroke (Arkuszewski et al., 2009; Hermann and Chopp, 2012; Xiong et al., 2010). The reason behind it could be that the region of active neovascularization acts as a niche for enhanced neuronal remodeling and tissue repair (Prakash and Carmichael, 2015; Shen et al., 2008). Here, we found that vascular density in the peri-infarct area is comparable to healthy animals in stroke non-treated mice (STROKE) and single treatment groups (TOXIN, ROBOT), while exceed physiological values in mice under double treatment (REHAB). Our results suggest that the transient proangiogenic state induced in response to an ischemic insult is enhanced by combined rehabilitation. This hypothesis is supported by our findings on increased endothelial cell proliferation in REHAB mice. In conjunction with the more efficient neuronal activation and stabilized synaptic turnover we measured in the same areas, our study thus supports the hypothesis that enduring recovery from stroke might result from the association of angiogenesis with neuronal plasticity (Ergul et al., 2012).
We then investigated the changes induced by the combined rehabilitation paradigm on cortical activity. It has been reported that motor-targeted focal stroke induces abnormally scattered cortical maps that persist for months after stroke (see (Harrison et al., 2013)). In our study, the diffuse structure of motor representation was observed in non-treated animals (STROKE group). The combined rehabilitation paradigm progressively re-established a cluster of neuronal activity in peri-infarct areas where location, timing and amplitude parameters highly resembled those of healthy control animals. Several studies recently addressed the contamination of the blood volume changes on fluorescence signal of calcium indicators like GCaMP (see, for instance, (Ma et al., 2016a; Ma et al., 2016b; Makino et al., 2017; Murphy et al., 2018; Wright et al., 2017)). The recent paper by Ma et al. (Ma et al., 2016b) showed that the hemodynamic contribution is slower than the GCaMP6f signal, with an average temporal delay of 0.86 ± 0.05 s (representing the phase shift between neural activity and total hemoglobin concentration). More recently, by simultaneously recording (wide-field) calcium dynamics and hemodynamics, Wright et al. (Wright et al., 2017) showed that sensory evoked responses corresponding to oxygenated and deoxygenated hemoglobin have (i) a very slow onset after stimulus presentation, (ii) delayed peaks and (iii) low peak magnitudes (see also (Murphy et al., 2016)) compared to GCaMP6 fluorescence. These studies suggest that the hemodynamic fluctuations, much slower than calcium-associated neuronal activity, affect the late phase of calcium transients more than the rise. Last, by performing wide-field imaging on awake Thy1-GFP mice, Makino an colleagues (Makino et al., 2017) confirmed that the majority of movement-evoked fluorescence changes recorded in Thy1-GCaMP6s mice are due to calcium rather than hemodynamic signals or other artifacts. In our study, we focused on the analysis of the rise phase of the calcium transient (i.e. slope and amplitude), where the hemodynamic contamination should be very low. Nevertheless, we cannot exclude that hemodynamic fluctuations could give a small contribution to the fluorescence signal we recorded on GCaMP6f mice.
By performing optogenetic stimulation on the same group of stroked-afflicted, rehabilitated mice, we demonstrated that the progressive spatial and temporal refocusing of motor control in the ipsilesional cortex is associated with increased interhemispheric connectivity. The functional coupling between homotopic motor cortices was restored 4 weeks after injury in STROKE mice, suggesting that some form of spontaneous recovery compensated for the transcallosal projections lost after stroke. The interhemispheric connectivity is further enhanced after the rehabilitative therapy, suggesting a possible correlation between the recovery of activation patterns in the perilesional area and the reinforcement of connectivity with pyramidal neurons projecting from healthy M1. While there are no studies examining, on the same animals, cortical functionality and transcallosal connectivity to draw direct comparisons with, our results are consistent with the hypothesis that rehabilitative training of the paretic upper limb can, on one side, increase the functional activation of motor regions of the ipsilesional cortex (Hodics et al., 2006; Hubbard et al., 2015) and, on the other, decrease interhemispheric inhibition from the contralesional motor cortex (Harris-Love et al., 2011). We speculate that refocusing of the motor representation could be supported by dendritic rewiring in the peri-lesional area (as shown by our in vivo imaging result on structural plasticity), as well as by the activity of contralesional pyramidal cells projecting to the peri-infarct area. Further analysis on targeted neuronal populations is needed to clarify how rehabilitation-induced changes in transcallosal connectivity are specifically linked to alterations in the excitatory/inhibitory balance in the ipsilesional spared tissue.
The mouse model of stroke used in this study provides several advantages over more traditional models. Recently, Lim and colleagues used voltage-sensitive dyes combined with optogenetics to demonstrate a spontaneous partial recovery of cortical functional connectivity 8 weeks after stroke (Lim et al., 2014). The combination of transgenic GCaMP6 mice with AAV-induced expression of ChR2 in the homotopic M1, has the advantage of enabling on one side, the reliable control and on the other, the stable monitoring of neuronal activity over weeks and months. We anticipate that the all-optical approach we used will be further extended to the longitudinal exploration of the modified transcallosal axonal projections from the injured to the heathy hemisphere in our rehabilitation paradigm.
Our multi-scale investigation brought to light complementary aspects of the structural and functional plasticity induced by rehabilitation that may lead to the development of more efficient therapies and improve post-stroke recovery in patient populations.
Experimental procedures
Animal and surgery
Animal protocols were approved by the Italian Minister of Health. GFP transgenic mice were implanted with a cranial window in the right hemisphere from the motor cortex up to the visual area. A thinned skull preparation was used on GCaMP6f mice. Each group contained comparable numbers of male and female mice, and the age of mice was consistent between the groups (4-12 months). The surgery was followed by the first imaging session under the two-photon microscope.
Photothrombotic Stroke
Mice were injected with a Rose Bengal solution (0.2 ml, 10 mg/ml solution in Phosphate Buffer Saline (PBS)). Five minutes after intraperitoneal injection a white light from an LED lamp was focused with a 20X objective and used to illuminate the M1 for 15 min to induce unilateral stroke in the right hemisphere. During the same procedure, on REHAB mice we injected 500 nl of BoNT/E (80 nM) divided in 2 separate injections of 250 nl at (i) +0.5 anteroposterior, +1.75 mediolateral and (ii) +0.4 anteroposterior, +1.75 mediolateral at 700 μm cortical depth. For optogenetics experiments, we delivered 1 μl of AAV9-CaMKII-ChR2-mCherry (2.48*1013 GC/mL) 700-900 μm deep inside the cortex (Conti et al., 2019).
Robotic Platform
The animals were trained by means of the M-Platform (Spalletti et al., 2014), which is a robotic system that allows mice to perform a retraction movement of their left forelimb. The motor rehabilitation consists in a pulling task: first the animal forelimb is passively extended by the linear actuator of the platform and then the animal has to pull back the forelimb up to the resting position. The motor training is composed by 15 movements and after each movement the animal receives a liquid reward.
Labelling of brain vasculature
After perfusion with PFA 4% followed by a perfusion with 10 ml of fluorescent gel, the vasculature was stained using the protocol described by Tsai et al.(Tsai et al., 2009), except that we replaced fluorescein (FITC)-conjugated albumin with 0.05% (w/v) tetramethylrhodamine (TRITC)-conjugated albumin in order to avoid spectral overlap between GFP and FITC (Di Giovanna et al., 2018).
Statistical Analysis
Data were analyzed using Origin Pro. The specific tests used are stated alongside all probability values reported.
Author Contributions
A.L.A.M. and M.C. conceived the study. A.L.A.M., E.C., A.P.D., C.S and C.A. performed experiments. S.L., A.L.A.M., E.C., A.S. and A.P.D. processed data. A.P., S.L. and L.S. developed the integrated microscope and M-platform system. F.S.P., S.M. and M.C. obtained funding support. A.L.A.M. and E.C. wrote the paper. All authors approved the paper.
Declaration of Interests
The authors declare no competing interests.
Acknowledgement
We thank Alessio Masi and Marie Caroline Muellenbroich for very useful discussion about the manuscript; Giuseppe De Vito for assistance on statistics analysis. We thank Dr. Zanier from the Mario Negri Institute (Milan) for providing reagents for immunohistochemical analysis. We thank the mechanics and electronics workshops at LENS. This project has received funding from the H2020 EXCELLENT SCIENCE -European Research Council (ERC) under grant agreement ID n. 692943 BrainBIT”. In addition, it was supported by the European Union’s Horizon 2020 research and innovation program under grant agreements No. 720270 (Human Brain Project) and 654148 (Laserlab-Europe). Part of this work was performed within the framework of the Proof of Concept Studies for the ESFRI research infrastructure project Euro-BioImaging at the PCS facility LENS.
Footnotes
↵8 Lead Contact