Microglia Mapping Reveals Cell Interactions Post Stroke

The Microfetti mouse is an established inducible system for multicolor fate mapping of microglia9, created by crossing the Cx3cr1CreER (refs. 17,18) and the R26RConfetti (ref. 19) mouse lines. Upon tamoxifen injection and based on the location of the Cre recombinase activity, microglial cells randomly express one of four different fluorescent proteins (Supplementary Fig. 1a): membrane-tagged cyan fluorescent protein (mCFP), nuclear green fluorescent protein (nGFP), cytoplasmic yellow fluorescent protein (YFP), and cytoplasmic red fluorescent protein (RFP). The labeled cells maintain the expression of these reporters under a CAGG promoter and pass them onto their progeny upon cell division. CD11b+ CD115+ Ly6Clo blood monocytes could also express the Confetti reporters. However, these cells have a half-life of 2-11 days17,18 and were not detectable in blood 5 weeks after tamoxifen injection (Supplementary Fig. 1b). Taken together, the Microfetti mouse allows the tracing of microglia over time and space. It also enables the distinction between microglia and invading blood monocytes.

The Microfetti (Cx3cr1 creER/+ R26R Confetti/+) mouse as a tool for studying microglial proliferation in ischemic stroke

We took advantage of this multicolor fate mapping system to study the dynamics of microglia proliferation in the 30-min proximal intraluminal MCAo/reperfusion model. We have chosen this model because of its high reproducibility, robust induction of microglia responses, and functional recovery, consequently, allowing the study of long-term sequelae of cerebral ischemia. This MCAo model mimics human large vessel occlusion with thrombectomy. The lesions are characterized by selective neuronal cell death limited to the ipsilateral striatum, resulting in a defined impairment of sensorimotor function20. After an intraperitoneal injection of tamoxifen and a waiting period of 6 weeks, mice were subjected to focal brain ischemia. Three days after stroke, MR-imaging was performed, which confirmed that all ischemic lesions included the dorsolateral striatum and showed homogeneous infarct sizes in the volume analysis (median = 10.26 mm3 [IQR: 7.50–13.92]) (Supplementary Fig. 1c). The immunolabeled brain sections were analyzed at six different time points: 2 days, 1 week, 2 weeks, 4 weeks, 8 weeks, and 12 weeks after stroke (Fig. 1a). The ischemic region in the dorsolateral striatum and the corresponding region in the contralateral brain hemisphere were imaged using confocal laser scanning microscopy (Fig. 1b). As we have shown in our previous work, microglial cells in the non-ischemic hemisphere showed predominantly a classical ramified morphology21 (Fig. 1c). In addition, the number of the Confetti+ cells in the non-ischemic contralateral striatum was stable over time with an average of 1653 ± 120 cells per mm3, and a random distribution of the Confetti labeling among Iba-1+ cells (Fig. 1c, e). In contrast, Confetti+ microglia in the ischemic tissue showed substantial dynamic changes over time as regards the number of cells as well as their morphology and spatial distribution (Fig. 1d, e). At 2 days after stroke, Confetti+ microglia exhibited shorter ramifications and larger somata without a significant increase in their number in comparison to the contralateral side (3623 ± 639 vs. 2036 ± 427 per mm3). After 1 week, the number of Confetti+ microglia on the ischemic side increased significantly in comparison to 2 days (11,368 ± 1966 vs. 3623 ± 639 per mm3). Of note, many clusters of cells with a similar color appeared, indicating the formation of clones. These clones were prominent after 2 and 4 weeks without the domination of a single (or very few) clone(s) (Fig. 1d). From 2 weeks on, the number of Confetti+ microglia in the ischemic hemisphere decreased gradually. After 12 weeks, Confetti+ microglia returned to a level that was no longer statistically significant compared to the contralateral side (3425 ± 1503 vs. 1356 ± 386 per mm3). Additionally, far fewer clones were apparent after 8 and 12 weeks.

Microglia undergo polyclonal proliferation after stroke

To investigate whether the observed accumulation of Confetti+ microglia in ischemic tissue was indeed due to clonal proliferation dynamics, we applied a Monte Carlo simulation strategy as previously established9,22. After confocal microscopic image acquisition, 3D images were rendered to extract the location and the Confetti label of microglia in the ischemic region and the contralateral control region (Fig. 2a). Concentric rings, with radii ranging from 20 to 300 µm, were placed around each Confetti+ cell in a stack (Fig. 2b). The density of cells with the same color to the cell in question was calculated for each ring. Then, 1000 Monte Carlo simulations were applied for each stack, where the color labels of cells were randomized. Likewise, the density of cells with the same color in the concentric rings for the simulated data was calculated. The colored regions in Fig. 2b, c represent the areas between the 2nd and 98th percentiles of the Monte Carlo-simulated densities. If the recorded data is higher than the 98th percentile of the simulated data, it means that cells of the same color are closer to each other than they would be if labels were randomly distributed. In other words, it indicates a clonal relation of neighboring microglia. Our results suggest that clones start to appear as early as 2 days after ischemia (Fig. 2c). Clonality peaked after 2 weeks, (i.e., the highest difference between simulated and recorded data; Fig. 2c). Importantly, our data showed an enlargement in many distinct microglial clones indicating polyclonal proliferation dynamics (Fig. 1d and Fig. 2a). At subsequent time points, the recorded data gradually approached the simulated data to reach random levels after 12 weeks. This indicates a gradual re-establishment of the random distribution of labeled cells in the tissue and a disintegration of the clones. In contrast, the contralateral non-ischemic hemisphere showed a steady random distribution of labeled cells and no signs of clonal proliferation dynamics of microglia.

The dynamics of polyclonal proliferation of microglia after 30-min MCAo

To quantify the spatiotemporal dynamics of the polyclonal proliferation, we followed a machine-learning approach. Specifically, we performed a cluster analysis using the density-based spatial clustering of applications with noise (DBSCAN) algorithm. Previous studies estimated the average nearest neighbor distance of microglia and identified 50 µm as a threshold for relatedness or clonality9,11,12. Hence, by specifying a radius of a neighborhood (ε = 50 µm) and a minimum number of cells in one clone (two cells), the DBSCAN algorithm clusters the cells based on their spatial density (Fig. 2d). Cells that express the same Confetti reporter and are closer to each other than 50 µm are considered to belong to the same clone. In contrast, cells with no neighbor closer than 50 µm are considered singlets, i.e., do not belong to any clone, and were excluded from further clonal analysis. This allowed us to estimate the number of microglial clones and their size (number of cells per clone) over time (Fig. 2e, f). The number of clones increased abruptly during the first week after ischemia (605 ± 132 vs. 2272 ± 343 per mm³ at 2 days and 1 week, respectively). After the first week, the number of clones stayed stable until 2 weeks. They decreased gradually afterward and were not statistically significant in comparison to the contralateral hemisphere after 12 weeks. When analyzing the average clone size dynamics, we found that it was about four cells per clone (3.8 ± 0.29) after 1 week, peaked after 4 weeks, reaching about six cells (6.03 ± 0.53), and decreased gradually afterward. In contrast, the contralateral side showed stable dynamics with mostly singlets and occasional small clones that are rarely larger than two cells. The distribution of clone sizes per animal in the stroke region and its contralateral counterpart are presented in Supplementary Fig. 2 and Supplementary Fig. 3. Additional proliferation dynamics were investigated using the Ki-67 marker (Supplementary Fig. 4). The analysis revealed that the probability of microglia proliferation is highest 2 days after brain ischemia, with the majority of microglial proliferation occurring within the first 2 weeks after stroke (Supplementary Fig. 4b). The analysis also revealed an inversely proportional relationship between proliferation index and clone size (Supplementary Fig. 4c). To summarize, our data demonstrate the onset, peak, and resolution dynamics of the local polyclonal proliferation of microglia after ischemic stroke.

Microglia show distinct membrane properties over time after stroke

Microglia express a defined set of ion channels that are correlated to their functional state23. In the normal brain, microglial cells have a characteristic membrane current pattern, and this pattern changes in a time-dependent fashion after a pathological impact24. Therefore, to investigate the functional states of single microglial cells after stroke, we performed whole-cell patch-clamp recordings of microglia in acute Microfetti brain slices. We analyzed the membrane properties at 2 days, 1 week, and 8 weeks after stroke from microglial cells in the ischemic tissue and the corresponding contralateral region (Fig. 3a). The cells from the contralateral hemisphere from all time points were pooled in one control group. Membrane currents were recorded in the voltage clamp mode and voltage steps were applied for 50 ms ranging from −170 mV to +60 mV with 10 mV increment from a holding potential of either −70 mV (gray lines) or −20 mV (black lines) (Fig. 3b). In the contralateral hemisphere, microglia displayed a typical pattern of only small currents as previously described for ramified microglia in acute slices24. In contrast, microglial cells from the ischemic tissue after 2 days exhibited large inward currents at hyperpolarizing voltage steps and pronounced voltage-dependent outward currents that were activated after depolarization from a holding potential of −70 mV. Such currents were similar to those described in LPS-activated microglia in vitro25 or in the facial nucleus in situ 24 h after facial nerve axotomy24. We have previously described these currents in invading monocytes 1 week after MCAo21. In line with our previous work, we observed inward currents after 1 week, while outward currents decreased21. Importantly, our patch-clamp analysis after 8 weeks detected pronounced inward currents; however, with a large variance between recorded cells. In accordance with the current-voltage relationships, microglia displayed a significantly decreased membrane resistance after 2 days that slowly but not completely recovered until 8 weeks after stroke (Fig. 3c). Microglial membrane capacitances were significantly increased only at 2 days after MCAo, indicating changes towards a rounder morphology and/or enlarged size (Fig. 3c). In contrast, the reversal potential did not change in the first week after MCAo but was significantly hyperpolarized after 8 weeks (median −66.56 mV). The analysis of the specific outward conductance revealed an increase for all time points after MCAo compared to control conditions. However, the most pronounced increase was found after 2 days. The specific inward conductance showed a progressive increase with time after MCAo, with significant changes after 1 and 8 weeks and a high variance between the cells at the latter time point. In conclusion, these results elucidate the electrophysiological profiles of microglia from the acute to the chronic phase after stroke and thereby reveal higher heterogeneity in membrane properties after 8 weeks.

The properties of inward and outward currents of microglia

To investigate which types of ion channels are responsible for the observed electrophysiological profiles of microglia after stroke, we characterized the kinetics of the underlying currents. Notably, the outward current displayed the characteristics of voltage-dependent, delayed outward rectifying K+ channels (Supplementary Fig. 5a). The time course of activation (tau) was voltage-dependent, with more rapid activation at more positive potentials. Currents did not inactivate within the 50-ms voltage step. The inward current activated with hyperpolarizing voltage steps at potentials negative to −100 mV showed the typical kinetics of inward rectifying K+ channels (Supplementary Fig. 5b). The inactivation time constant decreased with more negative potentials. In addition, the analysis of the specific inward conductivity of cells with inward but no outward currents further differentiated between cells with moderate and cells with strong inward currents (threshold 0.5 nS/pF) (Fig. 3d).

Thus, by analyzing each cell for the presence of these currents, we found that microglial cells in our data could be classified into four groups: (a) cells with only small currents, (b) cells with moderate inward currents only, (c) cells with strong inward currents only, and (d) cells with inward and outward currents (Fig. 3e). Under control conditions in the contralateral hemisphere, 69% of cells displayed only small currents, while 31% showed moderate inward rectifying currents. At 2 days after stroke, 95% of microglia displayed both inward and outward currents. After 1 week, only 47% of cells displayed pronounced outward currents, while all cells retained their inward currents. Of note, 6% of cells showed strong inward current after 1 week. After 8 weeks, we observed a higher diversity with all four groups being represented. Interestingly, 12% of cells at 8 weeks exhibited only small currents, similar to control cells, while the group with strong inward currents increased to 24%. In addition, 28% of cells showed outward currents after 8 weeks that differed from the above-described voltage-dependent delayed outward-rectifying currents. Those currents were activated by depolarization without a consistent delay, showed a linear current-voltage curve, and were not inactivated at a holding potential of −20 mV. This indicates another ion channel at play 8 weeks after stroke (Supplementary Fig. 5c).

Microglia within one clone share similar membrane properties, while neighboring microglia from different clones show higher heterogeneity 8 weeks after stroke

As described above, microglial cells within the ischemic tissue at 8 weeks after stroke showed high variability in their membrane properties and the types of membrane currents (Fig. 3e and Supplementary Fig. 5c). To investigate whether clonal identity contributes to this heterogeneity in microglial membrane properties, we compared recordings from cells within and between clones and between neighboring cells that were <50 µm apart but had different Confetti markers (Fig. 4a). The distance parameter of 50 µm ensures that patched neighboring cells from different clones share a similar microenvironment, as is the case for patched cells within a clone. Cells from the same clone exhibited very similar membrane properties and currents (Fig. 4b). In fact, the high similarity in electrical properties of cells within one clone was also evident at 1 week after stroke (Supplementary Fig. 6a). In contrast, neighboring cells from different clones are less similar in their electrophysiological profiles (Fig. 4c and Supplementary Fig. 6b).

Revealing the morphological diversity of microglia after stroke using deep learning

Next, we investigated whether microglia exhibit morphological changes that align with the observed electrophysiological dynamics. High-magnification images were acquired of RFP+ microglia in the ischemic striatum and the contralateral hemisphere (Supplementary Fig. 7a). To obtain an accurate image segmentation for such diverse morphologies, we developed an automated pipeline for morphological analysis, which employs a convolutional neural network U-Net26 (Supplementary Fig. 7b). After training the U-Net model, 668 cells were analyzed, and multiple morphological features were extracted for each cell. Some features distinctly differentiated between ipsi- and contralateral microglia and between cells from the ischemic tissue at different time points after stroke (Fig. 5b). Specifically, microglia in the stroke region showed an abrupt decrease in surface area-to-volume ratio at 2 days and a gradual recovery afterward. These changes are congruent with the observed increase in cell capacitance of microglia at 2 days, which also correlated with the presence of the delayed outward rectifying current described in Supplementary Fig. 5a. Based on the morphological feature presented in Fig. 5b, a uniform manifold approximation and projection for dimension reduction (UMAP) analysis was performed. The analysis illustrates an evident distinction between highly ramified cells, predominantly made of cells from the contralateral hemisphere, and cells in the stroke region (Fig. 5c and Supplementary Fig. 7c). It also shows a clear polarity of microglial morphologies in the stroke region where enlarged (primed) cells, mostly present at 2 days and 1 week after stroke, are localized at the upper right pole of the graph. In contrast, small roundish (ameboid) cells are more present at 8 weeks and occupy the lower right pole of the graph. Interestingly, highly ramified cells were absent in the ischemic tissue at 2 days and 1 week after stroke but reappeared at 8 weeks. Such morphological recovery after 8 weeks supports the observation of electrophysiological recovery of some microglial cells, as shown in Fig. 3e.

Microglial inter-clonal cell-cell interactions after stroke

In the morphological analysis, we observed microglial cells in the ischemic tissue in close contact with each other (Fig. 5a and Supplementary Fig. 8a). Therefore, we sought to exploit the multicolor labeling of microglia in the Microfetti mouse to visualize the microscopic interactions between microglial cells in the ischemic tissue. Indeed, when we investigated the ischemic region, we observed cells from different clones in direct interaction with each other. For example, what appears as a single object in Iba-1 immunofluorescence in Fig. 6a is composed of two microglial cells from two different clones. High-resolution confocal images and their 3D renderings showed how the two cells wrapped around and interacted with each other (Fig. 6a and Supplementary Movie 1). In fact, we could differentiate between four different types of cell-cell interactions of microglia in the ischemic tissue (Supplementary Fig. 8b): process-soma, process-process, soma-soma (flat), and entangled soma-soma interactions. The presence of these interaction types was also confirmed with stimulated emission depletion (STED) super-resolution microscopy (Fig. 6b). To explore the temporal dynamics of these interactions, we also employed live-cell imaging on acute brain slices. Indeed, we observed process-to-soma interactions, where a microglial cell extends a process to touch the soma of an adjacent microglial cell (Fig. 6c and Supplementary Movie 2). We also observed process-to-process interactions, where a microglial cell connects with one or more neighboring microglial cells through their processes. Supplementary Fig. 9a and Supplementary Movie 3 show a YFP+ process dynamically sliding over another process from an RFP+ cell. Furthermore, microglial cells showed soma-to-soma interactions. In some cases, the two somata were located next to each other in a parallel or flat manner (Supplementary Fig. 9b and Supplementary Movie 4). In other cases, the two or more microglial cells appeared entangled together in a nodular formation (Supplementary Fig. 9c and Supplementary Movie 5). It is important to emphasize here that those inter-clonal interactions are not due to cell proliferation because, in a proliferation event, both microglial cells share the same labeling color. Since microglia in homeostatic conditions show minimal territorial overlap27, a similar analysis in the contralateral hemisphere is of little relevance.

The present study provides several findings to advance the current understanding of microglial responses to ischemic stroke. First, we show that microglia undergo polyclonal proliferation after ischemic stroke and characterize the dynamics of this process. Second, we demonstrate the heterogeneity in the electrophysiological and morphological profiles of microglia over time after stroke and describe a partial functional and morphological recovery after 8 weeks. Third, our data shows a functional correlation of microglial cells within a given clone and a larger difference between neighboring cells from different clones in the resolution phase, which highlights the impact of clonal identity on microglial function post-stroke. Fourth, we demonstrate inter-clonal cell-cell interactions of microglia after stroke.

Microgliosis after ischemic stroke has been investigated using unicolor reporter mouse models such as Cx3cr1GFP or immunohistochemical labeling with pan-macrophage markers such as Iba-1, F4/80, or isolectin B43,6,7,8. While these methodologies can be employed to quantify the number of microglial cells, they are inadequate for elucidating the underlying clonal dynamics of this process. In this study, multicolor fate mapping was employed to demonstrate that microglia transition from a stochastic mode of proliferation under homeostatic conditions to polyclonal proliferation upon stroke induction. Our electrophysiological experiments suggest that clonality might have an impact on the functional heterogeneity of microglia after brain ischemia. Indeed, even within a distance of 50 µm, microglial neighbors from different clones exhibited more distinct electrophysiological profiles than neighboring microglia sharing the same Confetti color or clone. Single-cell RNA sequencing studies on experimental stroke models have yielded evidence of the heterogeneity of microglia after stroke at a transcriptomic level15,16. The present study provides supportive evidence for such heterogeneity on a functional level as well.